31
$\begingroup$

I've been writing some linear algebra problems with colleagues, and the following question occurred to us:

Let $S^2$ denote the unit sphere in $\mathbb{R}^3$. Does there exist a partition $S^2=\bigsqcup_\alpha B_\alpha$ such that each $B_\alpha$ is an orthonormal basis for $\mathbb{R}^3$?

The analogous question for the unit circle in the plane is easily affirmative, but that approach doesn't seem to work here.

$\endgroup$
5
  • $\begingroup$ The three-sphere is a Lie group, so it can be can be partitioned into orthonormal bases for $\mathbb{R}^4$. What about $S^5$? (I have no ideas about $S^2$.) $\endgroup$
    – Sam Nead
    Sep 30, 2021 at 16:18
  • $\begingroup$ How does the Lie group structure help? $S^1$ is a Lie group too but the only decomposition into bases that I can think of is still patched together in the sense that the bases do not form a differentiable (or even continuous) curve. $\endgroup$
    – M. Winter
    Sep 30, 2021 at 22:14
  • 1
    $\begingroup$ @M.Winter It seems like the quaternion structure of $S^{3}$ gives a partition into an orthonormal basis. Define an equivalence relation on $S^{3}$ where $x\simeq y$ iff $x=ay$ where $a\in\{1,i,j,k,-1,-i,-j,-k,\}$. Then for each equivalence class $P$ containing $x$, $P$ can be partitioned down further into two orthonormal bases, namely $\{x,ix,jx,kx\},\{-x,-ix,-jx,-kx\}.$ $\endgroup$ Oct 1, 2021 at 0:38
  • 3
    $\begingroup$ I am curious why this question is interesting? what is its application? $\endgroup$
    – C.F.G
    Oct 1, 2021 at 7:53
  • $\begingroup$ The partitions of $S^{2}$ into orthonormal bases are in a canonical one-to one correspondence with the functions $f:S^{2}\rightarrow S^{2}$ that satisfy the following functional equations: $f(z)\times(z\times f(z))=z,(z\times f(z))\times z=f(z),f(f(z))=z\times f(z),f(z\times f(z))=z.$ $\endgroup$ Oct 4, 2021 at 4:52

1 Answer 1

26
$\begingroup$

Using the Axiom of Choice, yes you can.

To get such a partition, start by enumerating all the points of the sphere with order type $\mathfrak{c}$ (the least ordinal number with the same cardinality as the sphere): say $\langle p_\alpha :\, \alpha < \mathfrak{c} \rangle$ is such an enumeration. Now we define our partition elements one at a time, via a transfinite recursion of length $\mathfrak{c}$. At stage $\alpha$ of the recursion, suppose we've already selected, at previous stages of the recursion, some bases $\{ B_\xi :\, \xi < \alpha \}$ that will be in our partition. Now consider the point $p_\alpha$. There are two possibilities: either we already put $p_\alpha$ into one of the $B_\xi$'s for some $\xi < \alpha$, or we didn't. In the first case, we do nothing at stage $\alpha$ of the recursion: formally, we could define $B_\alpha = \emptyset$ in this case. In the second case, we choose an orthonormal basis $B_\alpha$ that contains $p_\alpha$, and that is disjoint from everything we put into our partition at an earlier stage. This is possible there is a $\mathfrak{c}$-sized collection $\mathcal C$ of orthonormal bases, any two of which intersect only in $p_\alpha$ (just rotate a basis a bit around $p_\alpha$); because $\bigcup_{\xi < \alpha}B_\xi$ contains $\leq 3 \cdot |\alpha| < \mathfrak{c}$ points, one of the bases in $\mathcal C$ contains no points from $\bigcup_{\xi < \alpha}B_\xi$. We choose some such basis to be $B_\alpha$. In the end, $\{ B_\alpha :\, \alpha < \mathfrak{c} \} \setminus \{\emptyset\}$ is the desired partition.

$\endgroup$
3
  • 10
    $\begingroup$ For the record, I would still be interested to know if there is a more hands-on way to do this. For example, is there a Borel partition of the sphere into orthonormal bases? (I can't seem to think of one.) $\endgroup$
    – Will Brian
    Sep 30, 2021 at 20:47
  • 1
    $\begingroup$ The argument of Will Brian is analogous to the proof in Komjáth and Totik, , Problems and Theorems in Classical Set Theory, Chapter 13, Problem 13, that $\mathbb{R}^3$ can be partitioned into unit circles. I believe it is still open whether this result can be proved without the Axiom of Choice. $\endgroup$ Oct 3, 2021 at 14:18
  • 1
    $\begingroup$ Re Richard Stanley's comment, see mathoverflow.net/questions/28647/… $\endgroup$
    – aorq
    Oct 4, 2021 at 0:21

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.