What Is the Banach-Tarski Paradox? Can You Double a Sphere?

Take a solid ball. Cut it into a handful of pieces. Move those pieces around, only rotating and sliding them, never stretching. Put them back together.

You now have two solid balls. Each one is exactly the same size as the ball you started with.

That's not a trick with mirrors or a rounding error. It's a real theorem, proven in 1924 by Stefan Banach and Alfred Tarski. You can double a solid ball using nothing but rotations, translations, and five pieces.

Here's the thing: it's not really about spheres. It's about what happens to "size" when you're allowed infinitely many, infinitely scattered points to work with. Let's build up to it.

Wait, doesn't this break math?

Your first reaction is probably: that can't be right. Volume is conserved. Cut a cake into five pieces and reassemble it, and you still have one cake, not two.

You're right, for cake. Cake is made of a finite number of crumbs, and every piece you cut has a well-defined volume that adds up. But the pieces in Banach-Tarski aren't cake crumbs. They're not solid chunks at all.

They're infinitely tangled clouds of points, so scattered and jagged that they don't have a volume to begin with. You can't measure them, so there's nothing for the "volume must be conserved" rule to apply to. That's the loophole. And to see why it exists, we need to spend a minute with infinity.

Warm-up: the hotel that's never full enough to turn a guest away

Imagine a hotel with infinitely many rooms, numbered 1, 2, 3, and so on forever. Every single room is occupied. A guest walks in and asks for a room.

Normally you'd say no vacancy. But here's what the manager does instead: she asks every guest to move from room $n$ to room $2n$.

Guest 1 moves to room 2. Guest 2 moves to room 4. Guest 3 moves to room 6. Every guest still has a room; nobody got kicked out. But now every odd-numbered room is empty, and there are infinitely many of them. The new guest checks into room 1. So could seven more. So could infinitely many more.

Nothing was added, nothing was destroyed. Slide the guests over, and empty rooms just... appear. This is Hilbert's Hotel, and the lesson is the one we'll need for everything that follows: for infinite collections, "full" and "has room for more" aren't opposites. Banach-Tarski is this same trick, done to a sphere instead of a hallway.

The trick that makes it work: a group that duplicates itself

Now for the real engine. Forget spheres for a second and think about words.

Take two letters, $a$ and $b$, plus their "undo" moves, $a^{-1}$ and $b^{-1}$. Build every possible word by stringing these letters together, with one rule: never let a letter sit next to its own undo. So $aab$ is fine, $abb^{-1}$ is not, because $bb^{-1}$ cancels to nothing.

This collection of words is called the free group on two generators, and it has a strange self-similar shape. Every word (except the empty one) starts with exactly one of $a$, $a^{-1}$, $b$, $b^{-1}$. Group all the words by their starting letter, and you get four buckets, plus the empty word, that together contain everything.

Watch what happens when you switch to "Copy 1." Take just the words starting with $a$, plus just the words starting with $a^{-1}$. Now apply $a$ to every word in that second bucket. Every word like $a^{-1}x$ becomes $x$, so that whole bucket transforms into "everything that doesn't start with $a$", meaning identity, and everything starting with $a^{-1}$, $b$, or $b^{-1}$. Combine that with the original $a$-bucket you kept untouched, and:

Two of the four buckets, one of them rotated, rebuild the entire group. And the other two buckets, $S(b)$ and $S(b^{-1})$, do the exact same thing with a rotation by $b$. One set of words just produced two full copies of itself, out of pieces it already had. That's the paradox, in its purest form, no geometry required yet.

But a sphere isn't a group... or is it?

Here's the bridge. Take two rotations of a sphere, one around one axis, one around another, chosen so they never accidentally cancel out (this is provable, and holds for almost any two axes you'd pick). These two rotations, and their inverses, behave exactly like $a$, $a^{-1}$, $b$, $b^{-1}$: combining them the same way always produces a genuinely different rotation. No shortcuts, no coincidences.

Now pick a point on the sphere's surface. Apply every possible combination of those rotations to it, and you get an entire orbit, a countable, infinitely scattered set of points. Sort every point in every orbit by which rotation-word produced it, using the same four buckets from before, and you've just split the sphere's surface into four paradoxical pieces, the same way we split the group of words.

Rotate two of those pieces the way we rotated the word-buckets, and they rebuild the whole sphere. Rotate the other two, and they build a second, independent copy of the whole sphere. Two spheres, out of the material of one, no stretching involved.

Patching the leftover points

There's one loose end. A few points on the sphere sit exactly on a rotation axis, so rotating around that axis doesn't move them anywhere, it fixes them in place. Those fixed points don't fit neatly into any orbit, and there are only countably many of them, but they still need a home.

This is exactly the Hilbert's Hotel trick again, just dressed differently. Instead of doubling room numbers, we rotate a countable, infinitely long orbit of points by a fixed angle.

Rotate the whole orbit forward by that angle, and every point slides into the next point's old spot, except the very first one, whose spot empties out, and a brand new spot appears at the far end. The rotation absorbs exactly one extra point, for free, the same way the hotel absorbed one extra guest. Do this for the countable set of awkward fixed points, and they slot neatly into the construction. No leftovers.

From a hollow sphere to a solid ball

Everything so far duplicates the surface of the sphere, a hollow shell. But we started this whole thing wanting to double a solid ball.

The fix is almost anticlimactic: draw a straight line from the center of the ball to every point on the shell. Duplicating the shell drags an entire solid radius line along with it, and stacking up all those duplicated lines fills in a full, solid ball. The only snag is the center point itself, which every rotation leaves fixed, but that's a single point, and it slots into the construction the same way the leftover fixed points did.

Putting it all together: one ball becomes two

Here's the full picture. Split the solid ball into five pieces: four that rotate and reshuffle like the word-buckets, and one that handles the center point and the countable exceptional set. Move the pieces, purely with rotations and slides, no stretching, no shrinking.

Two of the pieces regroup into a full ball. The other pieces regroup into a second full ball, the exact same radius as the one you began with. The dots in that animation aren't solid chunks, they're a stand-in for infinitely scattered dust. That distinction matters more than it looks like it should, and it's exactly why this doesn't work on your kitchen table.

Why this doesn't work with an actual orange

If you tried this with a real orange, you'd fail immediately, and not because you're bad with a knife. Every piece you could ever actually cut, with a knife, a laser, anything physical, has a well-defined volume. Cut an orange into five pieces and the volumes add up to the orange's original volume, always. There's no way around it.

The five pieces in Banach-Tarski don't have a well-defined volume at all. They're what mathematicians call non-measurable sets, infinitely intricate scatterings of points so pathological that the very concept of "volume" refuses to apply to them. You can't draw them, you can't approximate them, and you can't construct them by any step-by-step recipe.

In fact, you need the Axiom of Choice just to prove they exist, a foundational assumption in set theory that guarantees you can always make infinitely many simultaneous, unspecified selections, one from each of infinitely many buckets, even when there's no rule for how to pick. Banach-Tarski isn't a construction you could ever carry out. It's a proof that such a decomposition exists, using tools that are, quite deliberately, not constructive.

That's the resolution. Volume conservation isn't violated. It just never applied to these pieces in the first place, because "volume" was never defined for them to begin with.

The short version

Banach-Tarski works because infinite, non-measurable sets don't play by the rules that finite, measurable objects do. Hilbert's Hotel shows infinite collections can absorb extras without anyone noticing. The free group on two generators shows that a self-similar infinite structure can be split into pieces that each regrow into a full copy of the whole thing. Map that structure onto a sphere with two independent rotations, patch a countable set of leftover fixed points using the same absorb-for-free rotation trick, and extend from a hollow shell to a solid ball with straight radial lines.

The result: one solid ball, cut into five pieces nobody could ever actually cut, reassembles into two solid balls identical to the original. It's not a loophole in geometry. It's a reminder that "size" is a much more fragile idea than it feels like, the moment infinity gets involved.

All visualizations are interactive React components running entirely in your browser, animated with requestAnimationFrame. No libraries beyond React.