I am trying to check that the "1" component of the Pauli-Lubanski vector for a massless particle with $P^{\mu} = (E, 0, 0, E)^{\mu}$ is $E(-J_1 + K_2)$, but I keep getting $E(-J_1 - K_2)$.

Starting from: $W_{\mu} = \frac{1}{2}\epsilon_{\mu \alpha \beta \gamma}P^{\alpha}M^{\beta \gamma}$,

$M_{0 i} = K_i$,

and $\frac{1}{2}\epsilon_{ijk}M_{jk} = J_i$.

I get,

$W_1 = \frac{1}{2}\epsilon_{1 \alpha \beta \gamma}P^{\alpha}M^{\beta \gamma}$

$= \frac{1}{2}\epsilon_{1 0\beta \gamma}EM^{\beta \gamma} + \frac{1}{2}\epsilon_{1 3 \beta \gamma}E M^{\beta \gamma}$

$= \frac{E}{2}(- \epsilon_{0 1 \beta \gamma}M^{\beta \gamma} + 2 \epsilon_{1 3 0 2 }M^{02}) $

$= \frac{E}{2}(- \epsilon_{0 1 i j}M^{ij} + 2\epsilon_{1 3 0 2 }M^{02})$

$= \frac{E}{2}(- \epsilon_{0 1 i j}M_{ij} - 2\epsilon_{1 3 0 2 }M_{02})$

$= E(-J_1 - \epsilon_{1 3 0 2 }M_{02})$

$= E(-J_1 - \epsilon_{1 3 0 2 }K_2$

$= E(-J_1 + \epsilon^{1 3 0 2 }K_2$

$= E(-J_1 - K_2)$.

But the answer is supposed to be $W_1 = E(-J_1 + K_2$.

I would appreciate it if someone could point out what I'm doing wrong please.


1 Answer 1


Check how you define the Levi-Civita symbols, there are a lot of conventions. In your derivation, you use the following relations: $$\varepsilon^{1302}=-1, \qquad \varepsilon_{01ij} = \varepsilon_{1ij}$$ Now, the first condition is equivalent to $\varepsilon^{0123}=+1$, so you are using the convention $\varepsilon_{0123}=-1$. On the other hand, the second condition is only consistent if you define $\varepsilon_{123} =-1$. Even if you could define $\varepsilon_{123}$ this way, I've never seen it, so I will assume it is not your intention.

This means that either $\varepsilon^{1302}=-1$ is wrong or $\varepsilon_{01ij} = \varepsilon_{1ij}$ is wrong. In particular, if you choose the convention $\varepsilon_{0123}=+1$, which is also very common, you would obtain $\varepsilon^{1302}=+1$ solving your sign problem.

EDIT: If you want to define $\varepsilon^{0123}=+1$ and $\varepsilon_{123} =+1$, then you need to modify the second relation to be $\varepsilon_{01ij} = -\varepsilon_{1ij}$, and you would obtain the result $$W_1 = E(J_1 - K_2)$$ which has the correct relative sign. Probably other conventions can account for the last $-$ sign.

  • $\begingroup$ Thank you. The convention I am following is $\epsilon^{0123} = 1$, so thank you for uncovering at least one of my errors. However, if I then correct the error in the $\epsilon_{01ij}$ term, is the final result not unchanged? $\epsilon_{01ij}M^{ij} = -\epsilon^{01ij}M^{ij}= 2 J^1 = -2 J_1$..? And it certainly doesn't affect the $K_2$ term, which is off by a sign. $\endgroup$
    – Gleeson
    Feb 15, 2023 at 16:13
  • $\begingroup$ Then I don't see any other error in your derivation. But as I said, there are a lot of conventions involved here. For example, if I remember correctly, in Ryder's QFT book, the PL pseudovector is defined with a - sign (and I think he also uses the convention $\varepsilon^{0123}=+1$). So check that you are following all the conventions correctly. $\endgroup$
    – Gaussian97
    Feb 15, 2023 at 16:17
  • $\begingroup$ Once one has chosen on the 4d convention for the epsilon (whether epsilon(0123)= 1 when the indices are up, or down), then how does one infer the inherited implications for the 3d symbol when the symbols are up or down? I am comfortable in the regular 3d case where up/down didn't matter, but it seems to happen often that in a 4d calculation we need to "restrict" to 3d and it is not clear to me how this plays out. Is there a straightforward way? For example, if $\epsilon^{0123} = 1$, then is $\epsilon^{0ijk} = \epsilon^{i1jk} = \epsilon^{ij2k} = \epsilon^{ijk3} = \epsilon^{ijk}$? $\endgroup$
    – Gleeson
    Feb 15, 2023 at 17:07
  • $\begingroup$ I think the most used convention for the 3D symbol is precisely to ignore whether the indices are up or down, simply define cyclic as +1. Of course, you can also define $\varepsilon_{123}=+1$ and change the sign every time you raise an index (though I haven't seen this convention very often). At the end, conventions are completely arbitrary. Once you have defined how the 3D and 4D symbols are defined, finding a relation between them should be trivial. If you want to use $\varepsilon_{0123}=-1$ and the usual 3D symbol, then you need the relation $\varepsilon_{0ijk}=-\varepsilon_{ijk}$. $\endgroup$
    – Gaussian97
    Feb 15, 2023 at 17:23

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.