I've learned the Banach-Tarski paradox as following:

  • The points on the sphere (but not the fixed points) are drawn as a square grid, form each point there are three new directions plus the direction back to the point it came form.
  • The grid is cut in four pieces:
    • (A) The x-axis, plus the points where you can come when turning right
    • (B) The points where you can come when turning left, excluding x-axis
    • The points where you can come by going up
    • The points where you can come by going down

Then there is a new sphere created by moving all points on A one place to the left. This creates together with B the new sphere.

Now my question is: Can we create a complete sphere by moving all points on A countably infinitely many points to the left?


The question is a bit light on details, but if I understand it correctly the answer is no.

First let me fill in some of the details of the usual argument. We take a copy of $F_2$, the free group on two generators, contained in $\text{SO}(3)$, the group of isometries of the sphere $S^2$. Excluding the set $D$ of fixed points (meaning points that are fixed by some nonidentity element of $F_2$) the action of $F_2$ on the remaining points is free. There are only countably many fixed points, so there is a trick for dealing with these later; let's ignore them.

In $S^2 \setminus D$, we use $\mathsf{AC}$ to choose, for each orbit, a point in that orbit. This allows us to consider each orbit of $S^2 \setminus D$ as a copy of $F_2$ and to transfer a paradoxical decomposition of $F_2$ to each copy simultaneously, giving a paradoxical decomposition of $S^2 \setminus D$.

The question seems to be considering a single orbit (copy of $F_2$) on its own, so it essentially asking about paradoxical decompositions of $F_2$. (By the way, I think it is a bit misleading to say "square grid": in a square grid if I go right, up, left, and then down, then I am back where I started, whereas in $F_2$ if I multiply $a$, $b$, $a^{-1}$, and $b^{-1}$ in that order, then I do not get the identity element.)

Now for your question:

Can we create a complete sphere by moving all points on A countably infinitely many points to the left?

I don't think it makes sense to move "infinitely many points to the left". In $F_2$ there is no element $b^{-1}b^{-1}b^{-1}\cdots.$ You might think that because we are consider a copy of $F_2$ that is contained in the sphere, there is a metric and so we could try to take a limit: let $x \in S^2 \setminus D$ be the point we have chosen from this orbit and consider the sequence of points $x, b^{-1}x, b^{-1}b^{-1}x,\ldots.$ However, this sequence is not convergent (in fact its set of limit points forms a circle.)

You may be confused by diagrams that show $F_2$ contained in the plane, and in which points approach a limit as they move to the left:

enter image description here

However, the metric that $F_2$ inherits from the plane $\mathbb{R}^2$ in this picture is not the same one that it inherits from $S^2$. In fact, this picture does not represent an isometric action of $F_2$ at all (note that the edges do not all have the same length.) So the usefulness of this picture for understanding the paradoxical decomposition of the sphere is limited.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.