4
$\begingroup$

We identify $S^3$ with the unit quaternions and $S^2$ the unit pure quaternions, and the conjugate of $z=a+bi+cj+dk$ is defined as $z^*=a-bi-cj-dk$. Then we consider the map $$\pi:S^3\ni z\mapsto ziz^*\in S^2.$$ The reason why $\pi$ maps into $S^2$ is given in this answer, but I'm stuck as to how to show it is surjective (in fact, a surjective submersion)?


Thanks to @Xipan Xiao's comment now I understand why $\pi$ is surjective: in fact for each $p\in S^2$ we can pick $\tilde q=\cos(\theta/2)+q\sin(\theta/2)\in S^3,\, q=e_1\times p/\|e_1\times p\|\in S^2$ so that $p=\tilde qi\tilde q^*$.

However, I'm still quite at a loss how to show $\pi$ is a submersion? Can we possibly avoid explicitly expanding the quaternions and using the stereographical projection charts to compute the Jacobian?

$\endgroup$
  • 1
    $\begingroup$ let $a=\cos\frac{\phi}{2}$, and $z=\cos\frac{\phi}{2}+V\sin\frac{\phi}{2}$ then $zxz*$ is the rotation of $\frac{\phi}{2}$ along the axis $V$. So the map is to rotate $(1, 0, 0)$ and it's onto since we can set $V$ to any unit pure quaternion and $\phi$ to any value. $\endgroup$ – Xipan Xiao Apr 23 '17 at 5:48
  • $\begingroup$ @XipanXiao thanks for your comment, which clarifies. $\endgroup$ – Vim Apr 23 '17 at 7:11
3
$\begingroup$

You can compute the derivative directly by viewing the spheres as submanifolds in $\mathbb H$ and the space of purely imaginary quaternions, respectively. Indeed, the map $f(z)=ziz^*$ makes sense as a map from $\mathbb H$ to imaginary quaternions, so you can differentiate it as any map $\mathbb R^4\to\mathbb R^3$, and then think about the restriction to the unit spheres. In partiuclar, you can compute $Df(z)(w)$ as the derivative at $t=0$ of $f(z+tw)$. Oberserving that quaternionic multiplication is $\mathbb R$ bilinear, while conjugation is $\mathbb R$-linear, you get $Df(z)(w)=wiz^*+ziw^*$, and this is twice the imaginary part of $wiz^*$.

So what you actually have to prove is that if you have a purely imaginary quaternion $q$ perpendicular to $ziz^*$ (i.e. tangent to $S^2$ in that point), then you can write it as $wiz^*+ziw^*$, for some $w$ which is perpendicular to $z$ (and hence tangent to $S^3$ in $z$). Now this is easy for $z=1$. Here $q$ must be a linear combination of $j$ and $k$, while $w$ has to be purely imaginary, and you immediately write out an explicit solution. For general $z$, the orthogonal space to $ziz^*$ is spanned by $zjz^*$ and $zkz^*$ and you can similarly find an explicit solution.

Having verified that $\pi$ is a submersion, it follows that it is surjective. Since $d\pi(z)$ is surjective, the implicit function theorem shows that the image of $\pi$ contains an open nighborhood of $z$ in $S^2$ and since this works for each $z$, you see that $\pi(S^3)$ is open in $S^2$. On the other hand, continuity of $\pi$ shows that $\pi(S^3)$ is a compact subset of $S^2$ and thus closed. Since $S^2$ is connected, you get $\pi(S^3)=S^2$.

$\endgroup$

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.