I'm trying to obtain the commutation relation between the Pauli-Lubanski vector operator and the generators of the Lorentz Group:

$$[W^\mu,P_\sigma]=[\frac{1}{2}\epsilon^{\mu\nu\lambda\rho} P_\nu M_\lambda\rho,P_\sigma]\\ \hspace{2.3cm}= \frac{1}{2}\epsilon^{\mu\nu\lambda\rho}[ P_\nu M_{\lambda\rho},P_\sigma]\\ \hspace{3.9cm}=\frac{1}{2}\epsilon^{\mu\nu\lambda\rho}P_\nu[ -\eta_{\lambda\sigma}P_\rho+\eta_{\rho\sigma}P_\lambda]\\ \hspace{5.cm}=\frac{1}{2}\epsilon^{\mu\nu\lambda\rho}\eta_{\rho\sigma}P_\nu P_\lambda-\frac{1}{2}\epsilon^{\mu\nu\lambda\rho}\eta_{\lambda\sigma}P_\nu P_\rho $$

Now, I know a priori that this commutator is zero and I'm trying to change indices accordingly. For instance, in the first term I want to rename the dummy indices $\lambda$ to $\rho$ and vice-versa. This will permute the respective last indices in the Levi-Civita tensor. Since I'm only renaming indices, I assume that I don't have to put a minus sign when doing so: $$=\frac{1}{2}\epsilon^{\mu\nu\rho\lambda}\eta_{\lambda\sigma}P_\nu P_\rho-\frac{1}{2}\epsilon^{\mu\nu\lambda\rho}\eta_{\lambda\sigma}P_\nu P_\rho$$

But now, to get the same form for the Levi-Civita tensor in both terms I permute the last two indices of that tensor in the first term, taking into account that it is an antisymmetric tensor: $$=-\frac{1}{2}\epsilon^{\mu\nu\lambda\rho}\eta_{\lambda\sigma}P_\nu P_\rho-\frac{1}{2}\epsilon^{\mu\nu\lambda\rho}\eta_{\lambda\sigma}P_\nu P_\rho=-\epsilon^{\mu\nu\lambda\rho}\eta_{\lambda\sigma}P_\nu P_\rho$$ , which is not $0$.

Have I gone wrong somewhere?

If I do the same for the second term instead of the first I get the same result but positive. Since the results must be equal, the only possibility for something of the form $+\text{final result}=-\text{final result}$ is for $\text{final result}=0$. Does it make sense?

  • 1
    $\begingroup$ Everything is zero, every term is separately zero, including the thing you said was not zero, because e.g. in $\varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho}$ the $\varepsilon^{\mu \nu \rho \sigma}$ is anti-symmetric in $\nu$ and $\rho$ while $P_{\nu} P_{\rho}$ is symmetric in $\nu$ and $\rho$. $\endgroup$
    – bolbteppa
    Commented Mar 17, 2020 at 20:56
  • 1
    $\begingroup$ Note $\varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} = \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} + \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} = \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} - \frac{1}{2} \varepsilon^{\mu \rho \nu \sigma} P_{\nu} P_{\rho} = \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} - \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\rho} P_{\nu} = \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} - \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} = 0$ $\endgroup$
    – bolbteppa
    Commented Mar 17, 2020 at 20:59
  • 1
    $\begingroup$ Does this answer your question? Commutator of the Pauli-Lubanski with a vector $P^{\mu}$ $\endgroup$ Commented Mar 17, 2020 at 22:22
  • $\begingroup$ @bolbteppa Consider writing this into an answer. It's more easily searchable, permanent and its quality can be assessed by up/downvotes. Even if you think it is trivial, your comment is a complete answer to the question that OP asked. $\endgroup$
    – MannyC
    Commented Mar 17, 2020 at 23:16
  • $\begingroup$ Thank you very much for your answers @bolbteppa! I still have one more question regarding the renaming of dummy indices. If I have $\epsilon^{\mu\nu\rho\sigma}P_\rho P_\sigma$ and rename $\rho$ to $\sigma$ and $\sigma$ to $\rho$, is the result $\epsilon^{\mu\nu\sigma\rho}P_\sigma P_\rho$ or $-\epsilon^{\mu\nu\sigma\rho}P_\sigma P_\rho$? $\endgroup$
    – RicardoP
    Commented Mar 18, 2020 at 12:07

1 Answer 1


Every term of the form $$\varepsilon^{\mu \nu \rho \sigma} P_{\rho} P_{\sigma}$$ in your calculation is zero, including the thing you said was not zero, because e.g. in $\varepsilon^{\mu \nu \rho \sigma} P_{\rho} P_{\sigma}$ the $\varepsilon^{\mu \nu \rho \sigma}$ is anti-symmetric in $\rho$ and $\sigma$, $\varepsilon^{\mu \nu \rho \sigma} = - \varepsilon^{\mu \nu \sigma \rho}$, while $P_{\rho} P_{\sigma}$ is symmetric in ν and ρ, $P_{\rho} P_{\sigma} = P_{\sigma} P_{\rho}$.

The simplest way to show that this is zero is to prove that $A = - A$ so that $2A = 0$. For simplicity we consider the two-dimensional analogue $$\varepsilon^{\mu \nu} P_{\mu} P_{\nu}$$ where $\mu, \nu = 0,1$. I want to show that $\varepsilon^{\mu \nu} P_{\mu} P_{\nu} = - \varepsilon^{\mu \nu} P_{\mu} P_{\nu}$. The calculation is as follows: \begin{align} \varepsilon^{\mu \nu} P_{\mu} P_{\nu} &= + \varepsilon^{\nu \mu} P_{\nu} P_{\mu} \ \ (1) \\ &= - \varepsilon^{\mu \nu} P_{\nu} P_{\mu} \ \ (2) \\ &= - \varepsilon^{\mu \nu} P_{\mu} P_{\nu} \ \ (3) \end{align} In line $(1)$ I used the fact that I can just re-label dummy indices whatever way I want, since they are dummy indices, and here I want to have them written in reverse order so I can later invoke anti-symmetry on $\varepsilon^{\mu \nu}$ and symmetry on $P_{\mu} P_{\nu}$. To see very explicitly why I can re-label dummy indices, just write it out: \begin{align} \varepsilon^{\mu \nu} P_{\mu} P_{\nu} &= \varepsilon^{0 \nu} P_{0} P_{\nu} + \varepsilon^{1 \nu} P_{1} P_{\nu} \\ &= (\varepsilon^{00} P_{0} P_{0} + \varepsilon^{0 1} P_{0} P_{1}) + (\varepsilon^{1 0} P_{1} P_{0} + \varepsilon^{1 1} P_{1} P_{1}) \\ &= (\varepsilon^{0\mu} P_{0} P_{\mu}) + (\varepsilon^{1 \mu} P_{1} P_{\mu}) \\ &= \varepsilon^{\nu \mu} P_{\nu} P_{\mu}. \end{align} Note I have done absolutely nothing but write it out so that there are no dummy indices, then collect the terms up with dummy indices again, but now using a different labelling.

In going from $(1)$ to $(2)$ I used the anti-symmetry of $\varepsilon^{\mu \nu}$ and in going from $(2)$ to $(3)$ I used the symmetry of $P_{\mu} P_{\nu}$. Now I have $A = - A$ so that $2A = 0$.

Another way to prove this result is to write it in the form $A = \frac{1}{2} A + \frac{1}{2} A = \frac{1}{2}A - \frac{1}{2} A = 0$ which is just a longer way of doing the above calculation, and clearly uses the above calculation in going from the $+$ to the $-$, but it is often used (e.g. to derive the angular momentum operators in special relativity/classical mechanics etc...) so it's good to be aware: \begin{align} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} &= \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} + \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} \ \ (1) \\ &= \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} - \frac{1}{2} \varepsilon^{\mu \rho \nu \sigma} P_{\nu} P_{\rho} \ \ (2) \\ &= \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} - \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\rho} P_{\nu} \ \ (3) \\ &= \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} - \frac{1}{2} \varepsilon^{\mu \nu \rho \sigma} P_{\nu} P_{\rho} \ \ (4) \\ &= 0. \end{align} In line $(1)$ I know that $\varepsilon^{\mu \nu \rho \sigma}$ is anti-symmetric while $P_{\nu} P_{\rho}$ is symmetric so that the whole thing is immediately zero, and I want to show this explicitly by turning it into something like $A = \frac{1}{2} A + \frac{1}{2} A = \frac{1}{2} A - \frac{1}{2} A = 0$, so I introduce the $1/2$ just to get two copies of it which I expect will cancel one another. In going $(1)$ to to $(2)$ I just used the anti-symmetry of $\varepsilon^{\mu \nu \rho \sigma}$ to write one of them with a $-$ sign. In going from $(2)$ to $(3)$ I then re-labelled the dummy indices so that I would have $\varepsilon^{\mu \nu \rho \sigma}$ in both terms, In going from $(3)$ to $(4)$ I then used commutativity of $P_{\mu}$ and $P_{\nu}$. Note $(4)$ is now in the form $A = \frac{1}{2}A - \frac{1}{2}A = 0$.


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.