Let $n \in \mathbb{N}$, and let $S^{n} \subseteq \mathbb{R}^{n+1}$ be the unit $n$-sphere.

Does there exist a dense subset $D$ of $S^n$ such that $\textbf{x} \in D \implies -\textbf{x} \notin D$?

This is true for $n=1$. Note that $S^1 \cong A := [0,1]/\mathbb{Z}$, where the equivalent question is: does there exists a dense subset $D'$ of $A$ such that $x \in D \implies x \pm \frac{1}{2} \notin D'$?

Consider the function $f : [0,1) \to \{0,1\}$,

$$f(x)= \begin{cases} 0 & x< \frac{1}{2},\text{ any decimal expansion of } x \text{ has finitely many 3s} \\ 1 & x < \frac{1}{2}, \;x \text{ has a decimal expansion with } \infty \text{ many 3s} \\ 1 & x = \frac{1}{2} \\ 1 & x > \frac{1}{2}, \text{ any decimal expansion of } x \text{ has finitely many 3s}\\ 0 & x > \frac{1}{2}, \;x \text{ has a decimal expansion with } \infty \text{ many 3s} \end{cases} $$

Then $D' = f^{-1}(\{0\})$ is a suitable dense set.

However, I am not sure how to approach the cases $n >1$, since $S^n$ is no longer homeomorphic to a very simple object and my topological toolkit is quite unsophisticated. If the question can be answered with elementary methods, I welcome any hints. I am not even sure whether or not I expect such a set $D$ to exist for $n> 1$.

  • 1
    $\begingroup$ It is also equivalent to ask the analogue of the question in $\Bbb R^n$, if you’ve figured out what the antipodes look like after stereographic projection $\endgroup$
    – FShrike
    Aug 7 at 14:42
  • $\begingroup$ I don't have time for a full answer, but you could consider the stereographic projection $\mathbb{S}^n\setminus\{(0,\ldots,0,1)\}\rightarrow\mathbb{R}^n$, then look at where the intersection $H_\pm^{n+1}=\{x\in\mathbb{R}^{n+1}|\operatorname{sign}(x_0)=\pm 1\}$ with $\mathbb{S}^n$ lands, and then consider the primages of their intersection with $\mathbb{Q}^n$ and $\mathbb{R}^n\setminus\mathbb{Q}^n$ respectivly. $\endgroup$ Aug 7 at 14:58
  • $\begingroup$ @SamuelAdrianAntz I almost understand your comment, but it's not quite clear. I am a little suspicious of your last step because note that $D$ quite naturally bijects with its complement, so both are uncountable. I understand the stereographic projection to be a bijection, so the preimage of $\mathbb{Q}^n$ would be countable? Or have I misunderstood? $\endgroup$ Aug 7 at 15:19
  • $\begingroup$ @FShrike That's certainly a possible approach. But even if I could figure out where they antipodes go, I am not certain that this simplifies matters much (or enough for me). After all, $\mathbb{R}$ is still much simpler than $\mathbb{R}^n$, $n>1$. $\endgroup$ Aug 7 at 15:28

1 Answer 1


This answer is an expansion of my comment, but I simplified it a bit without using the stereographic projection.

First of all, the obvious way of taking the intersection of $\mathbb{Q}^{n+1}$ with $\mathbb{S}^n\subset\mathbb{R}^{n+1}$ won't work, as it can even be empty as explained in this answer here. Consider the half-spaces $\mathbb{H}_\pm^{n+1}=\{x\in\mathbb{R}^{n+1}|\operatorname{sign}(x_0)=\pm 1\}$ and the (surjective) projections: $$\operatorname{pr}_\pm\colon \mathbb{H}_\pm^{n+1}\cap\mathbb{S}^n \twoheadrightarrow\mathbb{D}^n, (x_0,x_1,\ldots,x_n)\mapsto(x_1,\ldots,x_n)$$ flattening down both half-spheres.

The subsets $A=\operatorname{pr}_-^{-1}(\mathbb{D}^n\cap\mathbb{Q}^n)$ and $B=\operatorname{pr}_+^{-1}(\mathbb{D}^n\cap(\mathbb{R}^n\setminus\mathbb{Q}^n))$ are both dense (in the closure of their respective half-sphere) as preimages of dense sets under continuous maps. $A\cup B$ is therefore dense in $\mathbb{S}^n$ as the closure operator commutes with finite unions (See here).

For $x\in A$, we have $x_0<0$, therefore $-x\notin A$, as well as $x_1,\ldots,x_n\in\mathbb{Q}$, therefore $-x\notin B$.

For $y\in B$, we have $y_0>0$, therefore $-y\notin B$, as well as $y_1,\ldots,y_n\notin\mathbb{Q}$, therefore $-y\notin A$.

  • 1
    $\begingroup$ So one can use effectively the same idea as for $n=1$. Fair enough. I convinced myself it would be much more difficult and didn't try it. Also, your profile description is fairly entertaining. $\endgroup$ Aug 7 at 21:26
  • 1
    $\begingroup$ As a minor correction, $A$ and $B$ are images, not preimages. $\endgroup$ Aug 7 at 21:33
  • 1
    $\begingroup$ Thanks a lot! :-) The mistake with $A$ and $B$ is also corrected now. $\endgroup$ Aug 7 at 21:40
  • 1
    $\begingroup$ No worries, and thanks for the answer. I'm sorry, but I think I have just observed another typo. I think the projection maps surject onto $D_1(0)$ in $\mathbb{R}^{n-1}$, the interior of the $(n-1)$-sphere. $\endgroup$ Aug 7 at 21:45
  • 1
    $\begingroup$ Oh, yes! Honestly, I'm really not that good in writing down my thoughts in the correct notation. $\endgroup$ Aug 7 at 21:47

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.