4

How to visualize why the group of rotational symmetries (orientation preserving isometries) of the sphere $S^2$ is $$S^3/(\mathbf{Z}/2) ?$$

I meant that because that the rotational symmetry group is $SO(3)=RP^3=S^3/(\mathbf{Z}/2) $. I know this fact, but how to convince ourselves that by visualization that there is a $\mathbf{Z}/2$ mod out?

It may be useful to use the fibration fact that the $S^3/(\mathbf{Z}/2)$ can be a lens space obtained by $S^1$ fiber over $S^2$? But how is that relevant to the isometry group of the sphere $S^2$?

p.s. I certainly know $(3)=^3=^3/(/2)$. My point here is how to visualize the rotational group has a quotient out $(/2)$ part in $^3/(/2)$ for the rotational group? Which path of the loop of the rotational group element gives that $\pi_1(^3/(/2))=(/2)$?

Dan Asimov
  • 1,157
  • In my (very own) opinion, the isomorphisms $SO(3) \simeq \mathbb{R}P^3 \simeq S^3/\mathbb{Z}_2$ are "exceptional" and does not say much about the symmetries of the 2-sphere. The only geometric information on the symmetries of $S^2$ read in $SO(3)$ as it is: a group of rotations. – Didier Sep 04 '21 at 11:03
  • I certainly know $(3)=3=^3/(/2)$. My point here is how to visualize the rotational group has a quotient out $(/2)$ part in $^3/(/2)$ for the rotational group? Which path of the loop of the rotational group element gives that $\pi_1(^3/(/2))=(/2)$? – Марина Marina S Sep 04 '21 at 13:20
  • yes I know this trick but it will be good to have an intuitive answer for my question. How is that Dirac string related to the rotational group in terms of 3/(/2)? – Марина Marina S Sep 04 '21 at 15:55
  • It's unclear to me what the question really is here. Are you asking for a geometric proof of a homeomorphism between $SO(3)$ and $RP^3$? (See here for a nice geometric proof.) The last paragraph of your question suffers from some basic English language problems (" how to visualize the rotational group has a quotient out", "Which path of the loop of the rotational group element"). Please, rewrite these. Or, write these questions in your native language so someone can translate. – Moishe Kohan Sep 05 '21 at 18:07
  • Actually sorry, I am asking a geometric visualization of a homeomorphism (also diffeomorphism?) between (3) and 3. – Марина Marina S Sep 08 '21 at 17:14

2 Answers2

4

Consider the ball $B$ in $\mathbb{R}^3$ of radius $2\pi$, centered on the origin.

Any point $u\in B$ represents a rotation through an angle $|u|$ radians about the directed axis represented by $u$. Note there is no ambiguity with $0\in B$ as a rotation through $0$ radians is the identity map - e.g. we do not need an axis.

Note also that all rotations through $2\pi$ radians are the same regardless of axis, so we may quotient the boundary of $B$ to a point, to obtain a sphere $S^3$ and still have every element of $S^3$ representing a well defined rotation.

Any rotation in $SO(3)$ will be represented by precisely two points in $S^3$: a non-trivial rotation will be $\theta$ radians clockwise about some directed axis represented by a unit vector $u$, with $\theta\in [0,2\pi]$, so will be represented by $\theta u\in S^3$. However it will also be represented by $-(2\pi-\theta)u\in S^3$.

Thus $SO(3)\cong S^3/(\mathbb{Z}/2)$ in a visual way - a rotation is represented by a point in $S^3$, where the distance from the identity is the angle of rotation, and direction from identity is axis of rotation.

An easy way to see the natural fibration $SO(3)\to S^2$ is to define it as as evaluation on a point of $S^2$. That is fix a point $x\in S^2$ and map a rotation $A\in SO(3)$ to $Ax\in S^2$.

Finally, to visualise the non-trivial element in $\pi_1(SO(3))$ simply pick an axis and rotate about it going form $0$ radians to $2\pi$ radians. This lifts to a path in $S^3$ going from the origin to the point representing rotations by $2\pi$ radians. Thus it represents a non-trivial element of $\pi_1(SO(3))$ (to be trivial it would have to lift to a loop).


Let $f:S^3\to SO(3)\to S^2$ denote the composition of the quotient map $S^3\to SO(3)$ and the fibration described above, sending each $A\in SO(3)$ to $Ax$.

The fiber $F_y$ in $SO(3)$ above a point $y\in S^2$ is the set of rotations which map to $y$. That is $F_y$ is the set of rotations $\{C| Cx=y\}$.

If $Ax=y$ then $F_y=\{AT_\theta |\theta\in[0,2\pi]\}$, where $T_\theta$ denotes a rotation through $\theta$ radians about $x$.

Consider in particular $f^{-1}(x)$. This is all points in $S^3$ which map to a $T_\theta$. That is precisely the line through the origin and $x$. Note this line is actually a circle, as the boundary points of $B$ are identified. In other words $2\pi x=-2\pi x$ in $S^3$.

Now consider the fiber $F_y$ where $y$ is the point in $S^2$ opposite $x$. The fiber above $y$ in $SO(3)$ is all $\pi$ radian rotations about an axis perpendicular to $x$. These are precisely the rotations taking $x\mapsto y$.

Thus $f^{-1}(y)$ is the circle about the origin of radius $\pi$, perpendicular to $x$. Thus we can see that $f^{-1}(x)$ and $f^{-1}(y)$ form a pair of linked circles in $S^3$.

enter image description here

Note that for any points $z\neq w\in S^2$ there are paths in $S^2$ from $x$ to $z$ and from $y$ to $w$ which do not cross. Following these paths gives us isotropies of the fibers from $f^{-1}(x) \to f^{-1}(z)$ and $f^{-1}(y) \to f^{-1}(w)$, which also do not cross. Thus any pair of fibers $f^{-1}(z)$ and $f^{-1}(w)$ will be linked in the same way as in the picture above.

tkf
  • 11,563
3

One way is to think of S3 as the group of unit quaternions.

Then for any q ∊ S3, you can define the map

fq : S3 → S3

via fq(x) = q x q-1. The purely imaginary unit quaternions form a S2 ⊂ S3. Restricting fq to this S2 can be seen to define a rotation of S2.

Then the assignment

q → fq

defines a map S3 → SO(3) that is a surjective Lie group homomorphism whose kernel can be seen to be just {±1}, which is what you wanted to prove.

Dan Asimov
  • 1,157
  • thanks + 1 I am looking into this now. – Марина Marina S Sep 08 '21 at 20:35
  • 2
    @МаринаMarinaS This can also be visualized geometrically. A unit quaternion has the form $$q=\cos\left(\frac\theta2\right)+\sin\left(\frac\theta2\right)u,\qquad \theta\in [0,2\pi]$$ where $u$ is a purely imaginary unit quaternion. Then $f_q$ is just a rotation about $u$ through $\theta$ radians. Thus this is a different perspective on essentially the same construction as mine. – tkf Sep 08 '21 at 22:44
  • @tkf, Actually I wonder whether any of you twos answer can visualize the Hopf link fiber $S^1$ linked in the $S^3$? When we take some of two sets of rotations. – Марина Marina S Sep 09 '21 at 00:33
  • How does the link represent inside the $SO(3)$ or the $B3$ picture of tkf? – Марина Marina S Sep 09 '21 at 00:34
  • also btw- are we talking about homeomorphic or diffeomorphic here? – Марина Marina S Sep 09 '21 at 01:49
  • I do know Hopf fibration from this picture eg https://nilesjohnson.net/hopf.html Is your picture the same as his? – Марина Marina S Sep 09 '21 at 02:37
  • The "purely imaginary" is "b i +c j + d k"? I am confused by your expression – Марина Marina S Sep 10 '21 at 14:44
  • For this map, $q → f_q$, it seems that only $q \in S^3$ → to the restricted $f_q: S^2 \to S^2$ as $SO(3)$ is a surjective Lie group homomorphism. – Марина Marina S Sep 10 '21 at 16:11
  • Hi all, @tkf, but how about the case $q \in S^3$ → to the general $f_q: S^3 \to S^3$? What will be this map $q$ to $f_q: S^3 \to S^3$ not surjective Lie group homomorphism? – Марина Marina S Sep 10 '21 at 16:13
  • 1
    @МаринаMarinaS $f_q(\cos\phi+(\sin\phi) v)=\cos \phi+(\sin\phi) f_q(v)$, as reals are central in the quaternions. In other words $f_q$ restricts to the same rotation on each latitude. – tkf Sep 10 '21 at 16:49
  • 1
    @МаринаMarinaS Note these rotations have two eigenvalues equal to $1$ (corresponding to eigenvectors along the real axis and the axis of rotation). In general rotations of $S^3$ need not satisfy this (e.g. they can eigenvalues $\lambda,\bar{\lambda},\mu,\bar{\mu}$). Thus $f_q$ is certainly not surjective as a map $S^3\to SO(4)$. However the map $S^3\times S^3\to SO(4)$, sending $(q_1,q_2)\mapsto q_1_q_2$ is surjective and has kernel ${(1,1),(-1,-1)}$. – tkf Sep 10 '21 at 17:17