The Most Outrageous Theorem in Mathematics
Take a solid ball. Cut it into five pieces. Now rearrange those five pieces — no stretching, no squishing, just rigid motions, the same moves you'd use to rearrange furniture — and you get two solid balls, each identical to the original. You have doubled something from nothing.
If your first reaction is "that's obviously nonsense," congratulations — you have good instincts. It sounds like nonsense. It sounds like a violation of basic conservation laws, the kind of claim that should get a mathematician's license revoked. And yet it's true. It's been proved. It's been proved so thoroughly that no serious mathematician disputes it, even though many of them find it deeply unsettling.
The Banach-Tarski paradox — named after Stefan Banach and Alfred Tarski, who published it in 19241 — is what happens when the infinite gets weird. Not ordinary infinite, the kind you encounter in calculus class where limits politely converge and everything works out. This is the other kind of infinite, the kind that makes you question whether the foundations of mathematics are as solid as you thought.
Let me be precise about what the theorem actually says, because the precision matters. Given any solid ball in three-dimensional space, it is possible to partition it into a finite number of disjoint subsets, and then reassemble those subsets using only rotations and translations into two balls, each of the same size as the original. The minimum number of pieces required is five.2
The catch — and there is always a catch — is what those "pieces" look like. They are not the kind of pieces you could cut with a knife, or a laser, or any physical instrument whatsoever. They are so infinitely complicated, so unfathomably jagged and scattered, that they don't even have a well-defined volume. They are what mathematicians call non-measurable sets, and they are the heart of everything strange about this theorem.
The Warm-Up: Hilbert's Hotel
Before we tackle spheres, let's build some intuition for how the infinite misbehaves. You've probably heard of Hilbert's Hotel — the thought experiment where a hotel with infinitely many rooms, all full, can still accommodate a new guest by shifting everyone down one room. Guest in room 1 moves to room 2, guest in room 2 moves to room 3, and so on. Room 1 is now free.
This is already strange. A full hotel made room without removing anyone. But here's the really disturbing part: you can accommodate infinitely many new guests. Move everyone from room n to room 2n. Now all the odd-numbered rooms are empty, and you have infinitely many vacancies in your previously full hotel.
Hilbert's Hotel is a one-dimensional party trick. Banach-Tarski is what happens when you try the same stunt in three dimensions, with rotations instead of room-shuffling. The geometry of three-dimensional space has enough room for the infinite to really run wild.
The Banach-Tarski paradox: one ball becomes two, using only rigid motions.
Free Groups and Paradoxical Decompositions
The mathematical engine behind Banach-Tarski is a beautiful piece of algebra called a free group. Here's the idea. Take two rotations of the sphere — call them a and b. Choose them carefully: specifically, make sure they're rotations by irrational multiples of π around different axes. These two rotations generate a group — the set of all possible combinations of a, b, a⁻¹, and b⁻¹.
A "word" in this group is something like aab⁻¹a⁻¹bb — a sequence of rotations applied one after another. The crucial property of a free group is that no nontrivial word equals the identity. You can't combine these rotations in any way and accidentally end up back where you started (with the single exception of explicitly undoing each step).
This free group has a paradoxical property that Felix Hausdorff discovered in 1914, a full decade before Banach and Tarski.3 You can partition it into four pieces and rearrange them (by group multiplication) to get two copies of the entire group. It's Hilbert's Hotel on steroids, but now with the rich structure of rotations in space.
Every point on the sphere (minus a countable set of "problem points") belongs to an orbit under the free group. Using the Axiom of Choice, we pick exactly one representative from each orbit. Then we can sort all the remaining points by which word in the free group carries the representative to that point. This sorting creates the non-measurable pieces — and because of the paradoxical structure of the free group, these pieces can be rearranged into two copies of the (almost) entire sphere.
The "problem points" — a countable set — are handled separately, absorbed into the pieces with a clever rotation trick.
Here's where it gets philosophically interesting. That step — "pick exactly one representative from each orbit" — requires making uncountably many simultaneous choices. There is no rule, no formula, no algorithm that tells you which representative to pick. You just… choose. All of them. At once. This is the Axiom of Choice in action, and it is the single ingredient that makes Banach-Tarski possible.
A Simplified Decomposition
We can't draw non-measurable sets — they're too wild for any picture. But we can build intuition with a simpler analogy. The interactive below shows a circle being decomposed and rearranged. Think of it as a cartoon version of the paradox: the spirit is right even if the mathematical details are simplified.
Should We Believe This?
The Banach-Tarski paradox has been called the most surprising result in mathematics, the most counterintuitive, and — by its detractors — the most suspicious. Because the question it forces you to confront isn't just "is this true?" (it is). The question is: should we accept the axiom that makes it true?
The Axiom of Choice says: given any collection of nonempty sets, you can form a new set by choosing exactly one element from each. That sounds utterly innocuous. If I give you a bunch of boxes, each containing at least one marble, obviously you can pick one marble from each box. What's controversial about that?
Nothing — when the collection is finite, or even countably infinite. The trouble starts with uncountable collections. When you need to make more choices than there are natural numbers, more choices than you could ever specify, even in principle, even with infinite time — that's when things get strange. You're not choosing marbles anymore. You're asserting the existence of a function that nobody could ever write down.4
The Axiom of Choice: from every set, pick one. Simple for finitely many — but for uncountably many?
Here's the thing about the Axiom of Choice: it doesn't just enable Banach-Tarski. It enables huge swaths of mathematics that everyone uses and nobody finds paradoxical. Every vector space has a basis? That's the Axiom of Choice.5 Every ring has a maximal ideal? Axiom of Choice. Tychonoff's theorem that arbitrary products of compact spaces are compact? Axiom of Choice. If you ripped the Axiom of Choice out of modern mathematics, the damage would be catastrophic.
This is the dilemma. You can have a clean, paradox-free mathematical universe — Robert Solovay showed in 1970 that there's a model of set theory (without the Axiom of Choice) where every set of real numbers is measurable, and Banach-Tarski is simply false.6 But you lose a lot of the tools mathematicians depend on. Most mathematicians have made their peace: they accept the Axiom of Choice and regard Banach-Tarski not as a bug, but as a feature — a reminder that infinite sets are genuinely, irreducibly stranger than finite ones.
— Jerry Bona (all three are logically equivalent)
Explore the Axiom of Choice
The interactive below lets you explore what the Axiom of Choice does — and then cast your own vote on whether it should be accepted.
Finite Choice: Choose one element from each of 8 sets. Click a box to select.
No axiom needed — you can just… do it.
Selected: 0 / 8
Infinitely many sets, each with elements. We need a rule — like "always pick the smallest." Countable Choice is less controversial but still not provable from ZF alone.
Uncountably many sets — one for every point on a line. No algorithm can enumerate them all. The Axiom of Choice asserts a choice function exists, even though nobody can construct one. This is what enables Banach-Tarski.
Should mathematics accept the Axiom of Choice?
What Banach-Tarski Doesn't Mean
Let me head off the objection I know you're formulating: "But conservation of mass! But physics!"
You're right. Banach-Tarski doesn't apply to physical matter. Not because the mathematics is wrong, but because the "pieces" in the decomposition are non-measurable sets — objects that have no volume, no mass, no physical meaning whatsoever. They're not lumps of stuff you could pick up. They're infinitely scattered, fractal-dust clouds of points so pathological that Lebesgue measure — the standard mathematical formalization of "volume" — simply throws up its hands and refuses to assign them a size.
This is the deep lesson: measure theory only works on "nice" sets. The Lebesgue-measurable sets form a rich and well-behaved family. It includes every set you can construct by starting with basic geometric shapes and applying countable unions, intersections, and complements. Every set you will ever encounter in physics, engineering, or everyday life is measurable. Non-measurable sets require the Axiom of Choice to even prove they exist — you cannot exhibit a single concrete example.
Why Physicists Can Sleep at Night
The pieces in Banach-Tarski are non-measurable sets — they have no well-defined volume. Physical matter is made of atoms, which are discrete and countable. You cannot physically construct a non-measurable set, because it requires making uncountably many choices simultaneously. The paradox lives entirely in the realm of mathematical abstraction. It tells us something about sets, not about stuff.
Think of it this way. The real number line contains more points than you could ever list. Between any two points there are uncountably many others. Mathematical objects built from this continuum can behave in ways that have no physical analogue, because physical matter is fundamentally granular — atoms, quarks, Planck lengths. The continuum is a model, a beautifully useful model, but Banach-Tarski reminds us that the model contains monsters.
The Hausdorff Precedent
Banach and Tarski didn't come out of nowhere. In 1914, Felix Hausdorff proved a related result: the sphere (just the surface, not the solid ball) can be partitioned into three pieces, plus a countable set, such that the three pieces are all congruent to each other — and also congruent to the complement of one of them.3 This already violated measure-theoretic common sense: if three pieces are each one-third of the sphere, how can any piece also be two-thirds of the sphere?
Hausdorff's paradox was the first crack in the wall. Banach and Tarski drove a truck through it.
A timeline of the paradoxes: from the Axiom of Choice to Banach-Tarski and beyond.
Living with the Infinite
What should we do with the Banach-Tarski paradox? I think the answer is: let it humble you.
Mathematics is not just a collection of useful computation techniques. It is an exploration of logical consequences — a vast game of "if this, then that." And sometimes the consequences are startling. The Axiom of Choice is a single, intuitively plausible statement about sets. From it flows an enormous amount of useful, beautiful, everyday mathematics. And also, from the same axiom, you get the ability to decompose a ball into five pieces and reassemble them into two balls.
The paradox doesn't mean mathematics is broken. It means our intuitions about "volume" and "size" and "conservation" are finite intuitions, forged by evolution in a world of discrete objects at human scales. When we extend our reasoning to genuinely infinite collections — not just "very large" but actually, truly infinite — we should expect surprises. Banach-Tarski is the biggest surprise of all.
Giuseppe Vitali was the first to construct a non-measurable set, in 1905, just a year after Zermelo formalized the Axiom of Choice.7 His construction used the same basic trick as Banach-Tarski: partition the reals into equivalence classes, choose one representative from each class, and watch as the resulting set defies all attempts at measurement. The pattern was clear from the start. Once you accept that you can make infinitely many simultaneous unmotivated choices, you have to accept that some of the objects you create will be monsters.
Most mathematicians are fine with that. They treat the Axiom of Choice the way a carpenter treats a power saw: useful, effective, and yes, you could hurt yourself if you're not careful. The key is knowing when the results it gives you are "real" (applicable to measurable, constructive objects) and when they're artifacts of the infinite (non-measurable exotica that live only in the abstract).
As Stan Wagon writes in his definitive book on the subject: "The Banach-Tarski Paradox is more a comment on the richness of rigid motions in three dimensions and the pervasiveness of non-measurable sets than it is a statement about the physical world."8
One ball becomes two. It's outrageous. It's proved. And it's a perfect example of why mathematics is not just about getting the right answer on your tax return — it's about understanding the deep structure of logical possibility, including the parts that make you slightly uncomfortable.
Embrace the discomfort. That's where the good stuff is.