4

Lie group $SO(4)$ is doubly covered by $SU(2) \times SU(2)$, I want to know the map from $SU(2) \times SU(2)$ to $SO(4)$.

The map from $SU_{2}$ to $SO(3)$ is $\begin{pmatrix} \alpha & \beta \\-\overline{\beta} & \overline{ \alpha} \end{pmatrix} \longrightarrow \begin{pmatrix} \frac{1}{2}(\alpha^{2}-\beta^{2}+\overline{\alpha}^{2}-\overline{\beta}^{2} & \frac{i}{2}(-\alpha^{2}-\beta^{2}+\overline{\alpha}^{2}+\overline{\beta}^{2} & -\alpha\beta-\overline{\alpha}\overline{\beta}\\\frac{i}{2}(\alpha^{2}-\beta^{2}-\overline{\alpha}^{2}+\overline{\beta}^{2} & \frac{1}{2}(\alpha^{2}+\beta^{2}+\overline{\alpha}^{2}+\overline{\beta}^{2} & -i(\alpha\beta-\overline{\alpha}\overline{\beta})\\ \alpha\overline{\beta}+\overline{\alpha}\beta & i(-\alpha\overline{\beta}+\overline{\alpha}\beta)& \alpha\overline{\alpha}-\beta\overline{\beta} \end{pmatrix}.$ Therefore what is the image of

$(\begin{pmatrix} \alpha & \beta \\-\overline{\beta} & \overline{ \alpha} \end{pmatrix},\begin{pmatrix} \alpha' & \beta' \\-\overline{\beta}' & \overline{ \alpha}' \end{pmatrix})\in SU_{2}\times SU_{2}$

guojm
  • 345
  • 1
    Do you already know a nice map for the double cover $SU(2)\rightarrow SO(3)$? The one I know (using quaternions) can be easily tweaked to give the double cover of $SO(4)$. – Jason DeVito - on hiatus Sep 28 '18 at 14:58
  • Thanks. I know the map $SU(2) \rightarrow SO(3)$, can you explain clearly the map $SU(2) \times SU(2) \rightarrow SO(4)$? – guojm Sep 28 '18 at 15:01
  • 1
    The way I think of the map $SU(2)\rightarrow SO(3)$ is by conjugation using quaternions. Instead of conjugation, use two separate quaternions. That is, thinking of $SU(2)$ as $Sp(1)$, the double cover is $(p,q)\in Sp(1)^2\mapsto (v\mapsto pvq^{-1})$ where we think of $v\in \mathbb{R}^4\cong \mathbb{H}$. – Jason DeVito - on hiatus Sep 28 '18 at 15:32
  • Thanks. I re-edit the question. Could you please tell me what is the image of it? – guojm Sep 30 '18 at 08:44
  • The image must be all of $SO(4)$ for general theory reasons ;-). If I had to do the computation, I would use some sort of Computer Algebra System.... – Jason DeVito - on hiatus Sep 30 '18 at 20:29

2 Answers2

3

In physics, this construction is called the spinor map, or the Weyl representation, or the nonlinear (chiral) σ model.

Consider the unit determinant 2×2 matrix mapped out of the 3-vector $\vec r \equiv (x,y,z)$,
$$ X = \begin{pmatrix} \sqrt{1+\vec r ^2} + z & x - iy \\ x + iy & \sqrt{1+\vec r ^2} - z \end{pmatrix}= \sqrt{1+\vec r ^2}~ 1\!\!1 +x\sigma_x +y\sigma_y+z\sigma_z , $$ and your unimodular ones $$ L=\begin{pmatrix} \alpha & \beta \\-\overline{\beta} & \overline{ \alpha} \end{pmatrix},\qquad R=\begin{pmatrix} \alpha' & \beta' \\-\overline{\beta}' & \overline{ \alpha}' \end{pmatrix}, $$ so $LXR$ is also unimodular. You may painstakingly do the multiplication and compare entries, to specify the 3×3 matrix M sending $\vec r \equiv (x,y,z)$ to the 3-vector specified by $LXR$. Each element of M will be linear in the unprimed (L) and also primed (R) elements.

Normally, in physics, this is done in a smoother, more efficient language involving the Lie algebra, and the simple expression of group elements in terms of Pauli matrices, $\exp(i \vec r \cdot \vec{\sigma}) = 1\!\! 1\cos{r} + i (\vec r \cdot \vec{\sigma}) \sin{r}/r $ , but this procedure should be equivalent.

3

Here's a sketch of the construction; I'll leave the (more or less straightforward) algebraic details to the reader.

Let $\Bbb V$ be the $4$-dimensional (real) vector space of complex $2 \times 2$ matrices $z \in M(2, \Bbb C)$ satisfying $$z^* = J z^{\top}J^{-1}, \qquad J := \pmatrix{0&-1\\1&0} .$$

Now, realize $SU(2)$ as the group of matrices $g \in M(2, \Bbb C)$ that satisfy $g^*g = \Bbb I$, and define an action of $SU(2) \times SU(2)$ on $\Bbb V$ by $$(g, h) \cdot z := g z h^* .$$ This action preserves the positive-definite bilinear form $$\langle z, w \rangle := \operatorname{Re}\operatorname{tr}(z w^*)$$ on $\Bbb V$ and so defines a map $SU(2) \times SU(2) \to SO(\Bbb V) \cong SO(4)$; this is the desired double cover.

Computing in a basis for $\Bbb V$---the basis $\left(\pmatrix{1&0\\0&1}, \pmatrix{i&0\\0&-i}, \pmatrix{0&1\\-1&0}, \pmatrix{0&i\\i&0}\right)$ is both convenient for computation and (up to an overall rescaling) orthonormal with respect to the bilinear form---gives an explicit formula analogous to that for the map $SU(2) \to SO(3)$ in the question statement.

For more details and similar constructions of the other sporadic coverings among Lie groups (both complex and real), see Paul Garrett's useful notes Sporadic isogenies to orthogonal groups [pdf].

Travis Willse
  • 99,363