If a simple systematic way to derive or guess (either mathematically or by a combination of physical arguments and mathematics) that one of the Casimir operator of Poincare group is $W^2\equiv W_\mu W^\mu$ where $$W^\mu=\frac{1}{2}\epsilon^{\mu\nu\sigma\rho}P_\nu J_{\sigma\rho}\tag{1}.$$ In physics textbooks, (1) is given as a definition, and from which one can check that $W^2\equiv W_\mu W^\mu$ is indeed a Casimir. But I find this definition of $W^\mu$ to be quite non-trivial to guess. So I'm not looking for a rigorous derivation and if there are physical arguments to achieve this, it will do for me.


2 Answers 2


The intuitive meaning of Pauli-Lubanski's four-vector is easily explained for massive particles.

For a massive particle there is a reference frame at rest with it. In that reference frame, the four momentum $P_a$ has only its temporal component $a=0$ with value $m$. In that reference frame, looking at the formula you wrote, you see that $W^a$ has only three nonvanishing components $a=1,2,3$ with the evident meaning of angular momentum at rest with the particle (multiplied with the mass) if you take the meaning of $J_{0a}=-J_{a0}$ into account. The value of $W_aW^a$ is independent from the reference frame, so it can be computed at rest with the particle producing $m^2$ times the square of the angular momentum at rest with the particle: the squared spin times $m^2$.

Since the action of the Poincaré group is infinitesimally implemented by the generators of the group, the fact that $W_aW^a$ is an invariant just means that it commutes with all generators of the group, i.e., it is a Casimir operator.


The guess is based on the requirement of the translational invariance. Really, $J_{\mu\nu}$ doesn't commute with translations operator $P_{\mu}$. This means that the candidates $J_{\mu\nu}J^{\mu\nu}$, $\epsilon^{\mu\nu\alpha\beta}J_{\mu\nu}J_{\alpha\beta}$ on the role of the Casimir operator aren't translational-invariant.

In order to construct the translational-invariant Casimir operator, we can define $C = W_{\mu...}W^{\mu...}$, where $W_{\mu...}$ is translational-invariant operator. The only possible candidate (since $[P_{\mu},P_{\nu}] = 0$) is $$ W^{\mu} \sim \epsilon^{\mu\nu\alpha\beta}P_{\nu}J_{\alpha\beta} $$


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.