Maybe I still have trouble understanding what physicists mean when something is a vector but here is how I see it. I use the Einstein summation convention throughout.

Given a set of basis vectors $B = \{\vec{x}_i\}$ and a set of coordinates $C = \{c_i\}$, we represent a physical quantity $\vec{v} = c_i \vec{x}_i$. Now I am free to change my basis to some $\vec{x}'_i = R_{ij}\vec{x}_j$. In this case, I must have $c'_i = R^{-1}_{ji}c_j$ so that I still refer to the same physical quantity after the transformation.

With this in mind, I'm trying to understand what it means when people say that one can make a vector out of the Pauli matrices as explained here https://en.wikipedia.org/wiki/Pauli_matrices#Pauli_vector.

What is special about the Pauli matrices such that when I apply the transformation $R$ on basis of Pauli matrices, it "works"? What exactly wouldn't work if I took some other arbitrary set of 2x2 matrices?

  • $\begingroup$ This is e.g. explained in my Phys.SE answer here. $\endgroup$ – Qmechanic Apr 2 '18 at 17:16

Take $$ R=\left( \begin{array}{cc} e^{-\frac{1}{2} i (\alpha +\gamma )} \cos \left(\frac{\beta }{2}\right) & -e^{-\frac{1}{2} i (\alpha -\gamma )} \sin \left(\frac{\beta }{2}\right) \\ e^{\frac{1}{2} i (\alpha -\gamma )} \sin \left(\frac{\beta }{2}\right) & e^{\frac{1}{2} i (\alpha +\gamma )} \cos \left(\frac{\beta }{2}\right) \\ \end{array} \right)=R_z(\alpha)R_y(\beta)R_z(\gamma)\, . $$ Then, for instance, \begin{align} R\cdot\sigma_z\cdot R^{-1} = \sigma_z\cos\beta + \sigma_x\cos\alpha\sin\beta +\sigma_y\sin\alpha\sin\beta \end{align} which is the same as the rotation $R$ applied to $\hat z$. The other Pauli matrices also transform into linear combinations of themselves as the corresponding basis vectors do.

  • $\begingroup$ Sorry, could you explain why the $R_i(\theta)$ are 2x2 matrices since they rotate vectors in three dimensional space? And if they're 3x3, then how do you compute $R^{-1}\sigma_i R$? $\endgroup$ – user1936752 Apr 3 '18 at 6:17
  • $\begingroup$ @user1936752 The dimension of the vector space is $3$ because each Pauli matrix is a basis vector in that space (of $2\times 2$ hermitian matrices). If you want you can map $\sigma_x\mapsto \hat x= \left(\begin{array}{c} 1 \\ 0 \\ 0\end{array}\right)$ etc and then make $\bar R$ into a $3\times 3$ matrix representation of $R$ with $R\cdot \sigma_k \cdot R^{-1}\mapsto \bar R\hat k$. $\endgroup$ – ZeroTheHero Apr 3 '18 at 9:41
  • $\begingroup$ Is your point then that such a map exists for Pauli matrices but not arbitrary 2x2 matrices? If yes, can I trouble you to explicitly show how one gets the mapping between $\sigma_x$ and $\hat{x}$ mapping? $\endgroup$ – user1936752 Apr 3 '18 at 10:04
  • $\begingroup$ @user1936752 any $2\times 2$ hermitian traceless matrix can be expressed as a linear combination of the $3$ Pauli matrices, just like any vector in $3D$ can be expanded as a linear combination of $3$ linearly independent vectors. The Pauli matrices are trace orthogonal $\hbox{Tr}[(\sigma_k)^\dagger \sigma_m]=2\delta_{km}$ so function as basis vectors for this vector space. $\endgroup$ – ZeroTheHero Apr 3 '18 at 12:32

Suppose you start out looking for a vector $\vec{ \sigma}$, that you want to satisfy:

$$ {\vec{\sigma} \times \vec{\sigma}} = 2i \vec{ \sigma} $$

So it's a priori a vector with a non-zero cross product with itself. Expanding the cross product:

$$ \sigma_a\sigma_b - \sigma_b\sigma_a = 2i\epsilon_{abc}\sigma_c $$

You can solve that with 2 indices ($\alpha, \beta \in (+\frac 1 2, -\frac 1 2)$):

$$ [\sigma_a]_{\alpha\beta} = (\alpha + \beta)\delta_{a3} +|\alpha-\beta|\delta_{a1} + i(\beta-\alpha)\delta_{a2}$$

which are the Pauli matrices, forming a vector. The challenge now is convincing yourself that they describe an internal degree of freedom that is spin 1/2.


The formulae you see in textbooks for the spin matrices are just one representation of an algebra satisfying $\sigma_a\sigma_b =\delta_{ab}I_2+i \sum_c \epsilon_{abc}\sigma_c$. This condition is preserved under the usual rotational transformation, which obtains a different representation of this algebra. The same one-among-many disclaimer applies to other matrix-valued constants in physics (although the preserved rule will be different), such as Gell-Mann matrices & gamma matrices.


protected by Qmechanic Apr 2 '18 at 19:57

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

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