The Pauli-Lubanski pseudo-vector is defined as:

$$W_{\mu}=\frac{1}{2}\epsilon_{\mu \nu \lambda \rho}J^{\nu \lambda}P^{\rho}$$

Where the rotation and translation operators transform as:

\begin{align} U(\Lambda,a)P^{\mu}U^{-1}(\Lambda,a)&=\Lambda^{\mu}_{\nu}P^{\nu}\\ U(\Lambda,a)J^{\mu \nu}U^{-1}(\Lambda,a)&=(\Lambda^{-1})^{\mu}_{\lambda}(\Lambda^{-1})^{\nu}_{\rho}(J^{\lambda \rho}-a^{\lambda}P^{\rho}+a^{\rho}P^{\lambda}) \end{align}

I'm working on how the Pauli-Lubanski transforms under Lorentz, to show that it's in fact Lorentz covariant. I saw a solution in one of the answers of this post Calculating the commutator of Pauli-Lubanski operator and generators of Lorentz group the following:

\begin{align} U(\Lambda,a)W_{\mu}U^{-1}(\Lambda,a)&=\frac{1}{2}\epsilon_{\mu \nu \lambda \rho}(\Lambda^{\nu}_{\alpha}\Lambda^{\rho}_{\beta}J^{\alpha \beta}+a^{\nu}\Lambda^{\rho}_{\alpha}P^{\alpha}-a^{\rho}\Lambda^{\nu}_{\rho}P^{\rho})\Lambda^{\sigma}_{\delta}P^{\delta}\\ &=\frac{1}{2}\Lambda^{\alpha}_{\mu}\epsilon_{\alpha \nu \rho \sigma}J^{\nu \rho}P^{\sigma}\end{align}

However, I'm having trouble handling the $\Lambda's$ in order to get from step one to step two. Could you offer me some guidance?

  • $\begingroup$ Hint: are you comfortable with the transformation properties of $\epsilon$? $\endgroup$ Commented Sep 6, 2018 at 5:36
  • $\begingroup$ I only know that it is an antisymmetric tensor (equal to +1 for even permutations and -1 for odd permutations). I also read it is Lorentz invariant, so applying a Lorentz transformation doesn't affect it. $\endgroup$
    – Charlie
    Commented Sep 6, 2018 at 7:01

1 Answer 1


Since you're interested in the transformation law of $W^{\mu}$ under the Lorentz group, let us set $a^{\mu}=0$ in your equations. Also, notice that you've got the first one wrong; it should read

$$ U(\Lambda)P^{\mu}U(\Lambda)^{-1}=(\Lambda^{-1})^{\mu}_{\nu}P^{\nu} $$

Using the latter together with the transformation law for $J^{\mu\nu}$, we write

$$ U(\Lambda)W_{\mu}U(\Lambda)^{-1}=\frac{1}{2}\epsilon_{\mu\nu\lambda\rho}U(\Lambda)J^{\nu\lambda}P^{\rho}U(\Lambda)^{-1}=\frac{1}{2}\epsilon_{\mu\nu\lambda\rho}U(\Lambda)J^{\nu\lambda}U(\Lambda)^{-1}U(\Lambda)P^{\rho}U(\Lambda)^{-1}=\frac{1}{2}\epsilon_{\mu\nu\lambda\rho}(\Lambda^{-1})^{\nu}_{\tau}(\Lambda^{-1})^{\lambda}_{\sigma}(\Lambda^{-1})^{\rho}_{\alpha}\ J^{\tau\sigma}P^{\alpha} $$

Now, observe that from the definition of the determinant

$$ \epsilon_{\mu\nu\lambda\rho}(\Lambda^{-1})^{\mu}_{\beta}(\Lambda^{-1})^{\nu}_{\tau}(\Lambda^{-1})^{\lambda}_{\sigma}(\Lambda^{-1})^{\rho}_{\alpha}=\det(\Lambda^{-1})\ \epsilon_{\beta\tau\sigma\alpha} $$

follows the equality

$$\epsilon_{\mu\nu\lambda\rho}(\Lambda^{-1})^{\nu}_{\tau}(\Lambda^{-1})^{\lambda}_{\sigma}(\Lambda^{-1})^{\rho}_{\alpha}=\det(\Lambda^{-1})\ \Lambda_{\mu}^{\beta}\ \epsilon_{\beta\tau\sigma\alpha}$$


$$ U(\Lambda)W_{\mu}U(\Lambda)^{-1}=\det(\Lambda^{-1})\Lambda_{\mu}^{\beta}\ \frac{1}{2}\ \epsilon_{\beta\tau\sigma\alpha}\ J^{\tau\sigma}P^{\alpha}=\det(\Lambda^{-1})\Lambda^{\beta}_{\mu}W_{\beta} $$

Rearranging the indices and using the defining property of the $\Lambda$'s, $\Lambda^{T}\eta\Lambda=\eta$, we find

$$ U(\Lambda)W^{\mu}U(\Lambda)^{-1}=\det(\Lambda^{-1})(\Lambda^{-1})_{\beta}^{\mu}W^{\beta} $$

Since $\det(\Lambda^{-1})=\pm 1$ depending on the Lorentz transformation, we find

$$ U(\Lambda)W^{\mu}U(\Lambda)^{-1}=\pm(\Lambda^{-1})_{\beta}^{\mu}W^{\beta} $$

which is the transformation law for a pseudo-vector.

  • $\begingroup$ I see, then I can still then rearrange the components in the desired order (I also see there was a mistake in my lecture notes regarding the transformation of $P^{\mu}$). Also, I didn't think about using the definition of the determinant but I clearly see now that's the transformation property I was missing about the Levi-Civita tensor, and that's the part where one can find the pseudo-vector character of $W^{\mu}$ appears. I actually had to compute the general Lorentz transform $U(\Lambda,a)$ but I can work it from here now that I understand better how this tensor transforms. Thank you. $\endgroup$
    – Charlie
    Commented Sep 6, 2018 at 11:51
  • $\begingroup$ You totally got the point. You're welcome. P.S.: If you think my answer was enough, you may want to accept it. Just saying! Have a good day! $\endgroup$ Commented Sep 7, 2018 at 12:08

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.