3
$\begingroup$

The Hausdorff paradox is an incredibly counter-intuitive consequence of the axiom of choice; it is also important for demonstrating the non-existence, under AC, of a rotation-invariant measure on the discrete $\sigma$-algebra of the sphere.

But there is a part of the Hausdorff paradox that seems consistent with the existence of a rotation-invariant measure on the discrete $\sigma$-algebra, and yet still seems (to me) very counter-intuitive; so I am curious as to whether it can be proved without AC:

Is it a theorem of "ZF + Dependent Choice" that there exists a non-empty set $A \subset \mathbb{S}^2$ and rotations $R,R' \colon \mathbb{S}^2 \to \mathbb{S}^2$ such that $R(A)$ is disjoint from $A$ and $R'(A)=A \cup R(A)\,$?

Is it, at least, a theorem of "ZF + Dependent Choice" that there exists a non-empty set $A \subset \mathbb{S}^2$ and a rotation $R \colon \mathbb{S}^2 \to \mathbb{S}^2$ such that $A \subsetneq R(A)\,$?

(I fear there's some really obvious construction of such a set $A$ that I've missed - but I'll take the risk and ask the question!)

$\endgroup$
4
  • 2
    $\begingroup$ With only ZF+DC, it is consistent that everything is measurable, so you cannot have identity of that sort that break "conservation of measure". But what you are asking is probably possible as one can take $A$ to have zero measure. for example, take a faithful action of the free group $F_2$ on the sphere, pick a point $p$ which is not a fixed point of any element of $F_2$, then the orbit of $p$ is isomorphic as a $F_2$ set to $F_2$ itself, and hence you are going to be able to find inside the orbit something very close to what you are asking using a paradoxical decomposition of $F_2$ $\endgroup$ Feb 26, 2017 at 0:50
  • 4
    $\begingroup$ Yes to the second question, by A = {(cos(n), sin(n), 0): n in N}, let S = clockwise rotation by 1 radian. $\endgroup$
    – user44143
    Feb 26, 2017 at 3:57
  • $\begingroup$ $\aleph_0+\aleph_0=\aleph_0$ is a weaker (but still incredibly counterintuitive) version of the Hausdorff paradox which can be proved without AC. $\endgroup$
    – bof
    Mar 29, 2017 at 2:28
  • $\begingroup$ I disagree with @bof. $\endgroup$ Mar 29, 2017 at 6:46

1 Answer 1

2
$\begingroup$

Following the suggestion in the first comment below my question (and with the help of the second comment), I can give an example of a scenario that is "even worse" than what I requested, where $A \cup R(A)$ is a proper subset of $R'(A)$.

(Finding an example where we actually have equality is proving remarkably difficult to me, so I still invite answers regarding having equality, i.e. $A \cup R(A)=R'(A)$.)

Regard $\mathbb{S}^2$ as the unit sphere about the origin in $\mathbb{R}^3$.

Fix $\alpha \in \mathbb{R}$ such that $\cos(\alpha)$ is transcendental, and define the matrices \begin{align*} M_1 \ &= \ \left( \begin{array}{c c c} \cos(\alpha) & -\sin(\alpha) & 0 \\ \sin(\alpha) & \cos(\alpha) & 0 \\ 0 & 0 & 1 \end{array} \right) \\ & \\ M_2 \ &= \ \left( \begin{array}{c c c} \cos(\alpha) & 0 & -\sin(\alpha) \\ 0 & 1 & 0 \\ \sin(\alpha) & 0 & \cos(\alpha) \end{array} \right). \end{align*}

According to p227 of here, since $\cos(\alpha)$ is transcendental, $\{M_1,M_2\}$ forms a generating set of a free group contained in $\mathrm{SO}(3)$; and by Euler's rotation theorem, each non-trivial member of the group $G$ generated by $\{M_1,M_2\}$ fixes exactly 2 points. Hence (since $\mathbb{S}^2$ is uncountable) there exists a point $x^\ast \in \mathbb{S}^2$ which is not fixed under any non-trivial member of $G$.

Let $G^+$ be the monoid generated by $\{M_1,M_2\}$, and take \begin{align*} A \ &= \ \{M_1Mx^\ast : M \in G^+\} \\ R \ &= \ M_2M_1^{-1} \\ R' \ &= \ M_1^{-1}. \end{align*} Then $R(A)=\{M_2Mx^\ast : M \in G^+\}$ and $R'(A)=\{Mx^\ast : M \in G^+\}$. So $A$ is disjoint from $R(A)$, and $$ R'(A) \ = \ A \cup R(A) \cup \{x^\ast\}. $$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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