How do you prove that the universal covering of $SO(3, \mathbb{R})$ is $S^3$ ? Or equivalently, that it is diffeomorphic to $P_3\mathbb{R}$ ?

Thank you for your answers.

  • 8
    $\begingroup$ The group $SU(2) \cong S^3$ of unit quaternions acts by conjugation on the imaginary quaternions. Check that this gives a surjective homomorphism $SU(2) \twoheadrightarrow SO(3)$ with kernel $\pm \operatorname{id}$. See here for an explicit formula. See also this nice blog post by Qiaochu Yuan. $\endgroup$
    – t.b.
    Mar 7, 2012 at 16:00

2 Answers 2


Any element of $SO(3)$ is rotation about an axis in $\mathbb{R}^3$ - that is each element can be represented by an axis of rotation and an angle of rotation.

$\mathbb{RP}^3$ is $\mathbb{D}^3$ with antipodal points identified. Given a point in $\mathbb{D}^3$ it is some distance (between -1 and 1) on a vector from the origin. This vector gives you the axis of rotation for a point in $\mathbb{RP}^3$. We still need the angles of rotation, but these will be given by the distance between -1 to 1. Scale those values to be $-\pi$ to $\pi$, Note that then antipodal points of $\mathbb{D}^3$ are mapped to the same point in $\mathbb{RP}^3$

  • $\begingroup$ Thank you for your answer. Your first statement seems false : i think you forgot the elements which are a relexion on an axis and a rotation-reflexion on the orthogonal plane. $\endgroup$ Mar 7, 2012 at 16:16
  • 1
    $\begingroup$ Wouldn't those change orientation? $\endgroup$
    – Aru Ray
    Mar 7, 2012 at 16:18
  • $\begingroup$ I don't think so. Its eigenvalues are $\{-1,-1,1\}$ $\endgroup$ Mar 7, 2012 at 16:40
  • 2
    $\begingroup$ @SelimGhazouani: Note that this is simply a rotation by $180°$ around the third eigenvector, e.g. it can be written $$\begin{pmatrix} \cos\theta & \sin \theta & 0\\ -\sin \theta & \cos \theta & 0 \\ 0 & 0 & 1 \end{pmatrix}$$ for $\theta = \pi$. aru is right. It is a standard theorem in linear algebra that any matrix $A\in SO(3)$ can be brought into the above form with some $\theta$ by a orthogonal change of coordinates. You can also look at it this way: Given $A$, we can find an eigenvector $v$ to the eigenvalue $1$ and the restriction of $A$ to the 2-dim subspace $v^\perp$ must be a rotation. $\endgroup$
    – Sam
    Mar 7, 2012 at 18:03
  • $\begingroup$ How do you find an eingenvector to the eingenvalue 1 ? Can't the case $\{-1, e^{i\theta}, -e^{-i\theta}\} $ occur ? $\endgroup$ Mar 8, 2012 at 7:32

One can also show that the adjoint representation of su(2) is so(3) and find the covering map quite explicitly.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .