The homomorphism between $SO(3)$ and $SU(2)/\mathbb{Z}_2$ suggests that using $SO(3)$ to describe the real world (by real I mean classical) is simply a convention and that the exact same physics can be described using spinors (using a $2$ dimensional complex vector space instead of a $3$ dimensional real vector space).

My confusion is the following. For $G_i \in SU(2)$, we can write

$$ SU(2)/\mathbb{Z}_2 = \Big\{\pm G_1,\;\pm G_2,\;\pm G_3,\dots\Big\}\tag{I.} $$

where $G_i$ are $2\times 2$ matrices and therefore act on a $2$ dimensional vector space say $V$. For $H_i \in SO(3)$, $H_i$ are $3\times 3$ matricies acting on a $3$ dimensional real vector space say $U$. I understand that

$$SU(2)/\mathbb{Z}_2 \cong SO(3) \tag{II.}$$

which shows that the $3$ dimensional representation of $SO(3)$ is $SU(2)/\mathbb{Z}_2$, however, in equation (I.), I have explicitly constructed a $2$ dimensional representation and it is not clear how the vector spaces are related. I should in theory be able to map a vector from $U\rightarrow V$ but I'm not sure how to do this.

When people talk about spinors, are they talking about the vectors acted on by the fundamental representation of $SU(2)$ and not $SU(2)/\mathbb{Z}_2$? If this is the case, does that mean that spinors are only relevant in quantum systems i.e. where the projective representation can be considered due to the inner product of two states being invariant under phase?


Perhaps this is a possible way of solving.

It has been pointed out in the answer section that one can write $$G_i = \exp \imath \alpha_k^{(i)} \sigma_k/2$$ $$H_i = \exp \imath \alpha_k^{(i)} L_k$$ where $L_k$ are the generators forming the Lie algebra of the group $SO(3)$, $\sigma_k$ are the Pauli matricies which are the generators forming the Lie algebra of the group $SU(2)$ and $\alpha^{(i)}_k$ is an element of $\mathbb{R}^3$.

If I have the constants $$\alpha^{(1)}_k = (1,0,0)$$ corresponding to an element $H_1$ of $SO(3)$ (where $H_1$ rotates around the $x$ axis), the vector $$X= \begin{pmatrix} 1 \\ 0 \\0 \end{pmatrix} \in U$$

should be left invariant under the transformation $H_1$ since the vector along the axis of rotation is always left invariant. If I can show that the following

  1. There is exactly one eigenvector of $SU(2)$ for each $\alpha$ therefore the vector $U$ of $SO(3)$ that is invariant under a particular element of $SO(3)$ gets mapped to the invariant vector of $SU(2)$ for the same $\alpha$.

then would that be a correct way of mapping a vector from $U\rightarrow V$?


2 Answers 2

  1. There is not a linear map$^1$ $$\mathbb{R}^3~\cong~ \color{red}{U}\quad\to\quad \color{red}{V}~\cong~\mathbb{C}^2\tag{1}$$ of vector spaces, but instead a surjective 2:1 Lie group homomorphism $$ SU(2)~\cong~SU(\color{red}{V})\quad\to\quad SO(\color{red}{U})~\cong~SO(3).\tag{2} $$ For details, see e.g. my Phys.SE answer here.

  2. However, there is a non-linear map $$\begin{align}\color{red}{V}~\cong~\mathbb{C}^2~\ni~\psi\quad\mapsto\quad\sigma~=~&\psi\psi^{\dagger} - \psi^{\dagger}\psi{\bf 1}_{2\times 2}\cr ~\in~&su(2)~\stackrel{(4)}{\cong}~\mathbb{R}^3~\cong ~\color{red}{U}\end{align}\tag{3}$$ if we use a bijective isometry (4) from 3D space $(\mathbb{R}^{3},||⋅||^2)$ to the space of $2\times 2$ traceless Hermitian matrices $(su(2),\det(⋅))$, $$\begin{align} \mathbb{R}^3 ~\cong ~& su(2)\cr ~:=~&\{\sigma\in {\rm Mat}_{2\times 2}(\mathbb{C}) \mid \sigma^{\dagger}=\sigma \wedge {\rm tr}(\sigma)=0\} \cr ~=~& {\rm span}_{\mathbb{R}} \{\sigma_1,\sigma_2,\sigma_3\},\cr \mathbb{R}^3~\ni~&\vec{r}~=~(x^1,x^2,x^3) \cr &\quad\mapsto \quad\sigma~=~\sum_{j=1}^3 x^j\sigma_j~\in~ su(2), \cr ||\vec{r}||^2 ~=~&\det(\sigma),\end{align}\tag{4}$$ cf. the answer by Thomas Fritsch. For a relation to the Bloch sphere, see also e.g. this related Phys.SE post.

  3. As a consistency check, one may show that if we transform a spinor $\psi\in \mathbb{C}^2$ with an element of $SU(2)$, then the corresponding vector $\vec{r}\in\mathbb{R}^3$ from eq. (3+4) is rotated with the corresponding element of $SO(3)$ from eq. (2).


$^1$ The rotation vector $\vec{\alpha}\in \mathbb{R}^3$ that OP mentions represents an element of the Lie groups (2) rather than the vector spaces (1).

  • $\begingroup$ How do you tell if it's going to be a linear map or not? Even if it's non-linear, does that pose any problems? Also, when considering the Dirac equation, how do we make any predictions if we don't even know what the value of each of the spinor components represent i.e. if we don't know how to map them to $\mathbb{M}^4$? Is this just because they are spinor fields i.e. $\phi(x^\mu)$, which are naturally dependent on $x^\mu$ explicitely? $\endgroup$
    – DIRAC1930
    Commented Aug 3, 2021 at 15:30
  • $\begingroup$ I updated the answer. $\endgroup$
    – Qmechanic
    Commented Aug 4, 2021 at 8:07
  • $\begingroup$ I'm a bit confused about how this mapping is worked out. Could you explain explicitly where it comes from? $\endgroup$
    – DIRAC1930
    Commented Aug 4, 2021 at 14:51
  • $\begingroup$ I updated the answer. $\endgroup$
    – Qmechanic
    Commented Aug 4, 2021 at 15:14
  • $\begingroup$ Thanks. So to relate to the last point of my original question, is it just a convention to use $SO(3)$ on $\mathbb{R}^3$ and not $SU(2)$ on $\mathbb{C}^2$ to describe classical rotations on $\mathbb{E}^3$? Correct me if I'm wrong, but I'm assuming this changes when describing quantum phenomenon on $\mathbb{M}^4$ because $SO(1,3)$ is only homomorphic to $SU(2) \times SU(2)$ up to a phase (projective representation). For example, $(1/2,0)$ does not exist in classical physics because it is an artefact of the projective representation which is only valid quantum mechanically? Is this correct? $\endgroup$
    – DIRAC1930
    Commented Aug 4, 2021 at 15:45

I should in theory be able to map a vector from $U\rightarrow V$ but I'm not sure how to do this.

Let $$|\psi\rangle=\begin{pmatrix} a \\ b \end{pmatrix}$$ be a spinor in $V=\mathbb{C}^2$. Then you can construct a corresponding vector in $U=\mathbb{R}^3$ by using the Pauli matrices (see A. Steane - An introduction to spinors, page 4, equation 12):

$$\begin{align} x&=\langle\psi|\sigma_x|\psi\rangle \\ y&=\langle\psi|\sigma_y|\psi\rangle \\ z&=\langle\psi|\sigma_z|\psi\rangle \end{align}$$

or explicitly spelled out $$ \vec{r}= \begin{pmatrix} x \\ y \\ z \end{pmatrix}= \begin{pmatrix} a^*b+b^*a \\ -ia^*b+ib^*a \\ a^*a-b^*b \end{pmatrix}$$

You see, it is not a linear map (as already pointed out in @Qmechanic's answer), but it is more a square-like map. Especially, all spinors $e^{i\alpha}|\psi\rangle$ (with $\alpha\in\mathbb{R}$) result in the same vector $\vec{r}$.

When the spinor $|\psi\rangle$ is transformed by a matrix from $SU(2)$, then the corresponding vector $\vec{r}$ is transformed by a matrix from $SO(3)$.

The inverse mapping (from vector $\vec{r}\in\mathbb{R}^3$ to spinor $|\psi\rangle\in \mathbb{C}^2$) is more difficult. And as noticed above, it is not unique.
First convert the Cartesian coordinates ($x,y,z$) to spherical coordinates ($r,\theta,\phi$). From these you can then calculate the components of $|\psi\rangle$ (see A. Steane - An introduction to spinors, page 3, equation 1): $$\begin{align} a&=\sqrt{r}\cos(\theta/2)e^{i(-\alpha+\phi)/2} \\ b&=\sqrt{r}\sin(\theta/2)e^{i(-\alpha-\phi)/2} \end{align}$$ where $\alpha\in\mathbb{R}$ is any arbitrary phase.

  • $\begingroup$ Thanks. Could you explain how you found this mapping or the reason behind why it works? Also, does this work in general for a lie group homomorphism? $\endgroup$
    – DIRAC1930
    Commented Aug 4, 2021 at 14:44
  • $\begingroup$ @DIRAC1930 For more background see the linked article in my updated answer. This mapping is special to the relation between $SU(2)$ and $SO(3)$. Other pairs of Lie groups will have other mappings. $\endgroup$ Commented Aug 4, 2021 at 14:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.