Show that the imaginary part of the Dirac Lagrangian [density] is a total derivative.

My attempt:

The Dirac Lagrangian density is given by, $$ \mathcal{L} \ = \ -\bar{\psi}(\gamma^{\mu}\partial_{\mu} + m)\psi $$ Now, as far as I am aware, the hermiticity conditions of the gamma matrices depend on which representation one uses - we used the Weyl representation in class but with the $(-, +, +, +)$ metric, so there was a factor of $-i$ in every matrix - essentially meaning that the matrices would be anti-hermitian (and therefore, their complex conjugate should furnish a negative sign). Including this is not necessary for my solution, but I will still do it.

The spinors themselves are complex (as far as I know) and all I can say is that the product $\psi^{\dagger}\psi$ should be real. Thus, $$ \mathcal{L}^{*} \ = \ -\bar{\psi}^{*}(-\gamma^{\mu}\partial_{\mu} + m)\psi^{*}. $$ This implies, $$ \mathcal{L} - \mathcal{L}^{*} = -\bar{\psi}(\gamma^{\mu}\partial_{\mu} + m)\psi - \bar{\psi}^{*}(\gamma^{\mu}\partial_{\mu} - m)\psi^{*}$$ $$ = -\partial_{\mu}(\bar{\psi}\gamma^{\mu}\psi + \bar{\psi}^{*}\gamma^{\mu}\psi^{*}) + (\partial_{\mu}\bar{\psi} \gamma^{\mu} - m\bar{\psi})\psi + (\partial_{\mu}\bar{\psi}^{*} \gamma^{\mu} + m\bar{\psi}^{*})\psi^{*} .$$ Here, the expressions in the brackets in the second and third terms vanish if I impose the equations derived from Dirac Lagrangian itself. This means, $$ \frac{1}{2i}(\mathcal{L} - \mathcal{L}^{*}) = \mathrm{Im}(\mathcal{L}) = \partial_{\mu}f^{\mu} $$ This is a total derivative as I am supposed to prove.

My issues with my solution:

In order to derive this, I impose the dynamical equations derived from the starting Lagrangian - is it at all valid? The Dirac equation says, $$ (\gamma^{\mu}\partial_{\mu} + m)\psi = 0.$$ This clearly says that if I impose this equation on the starting density, I get $\mathcal{L} = 0$ - this does not seem to be different from what I have done in my own solution, except that I imposed the equation later instead of at the beginning. The imaginary part being $0$ is of course a valid solution (since any constant field will have a vanishing divergence) but it is a trivial solution.

At the same time, the only other option I see is to use both the idea that the chosen representations of the gamma matrices are anti-hermitian and the plane wave solutions of the Dirac field to derive something - but there are similar problems there too (since then I would need to look into all the momentum states, $u(p), \ v(p)$) and the entire procedure would be tedious and I am not even sure if I will get an answer. The second issue to that is that then the proof is not be general - I will have assumed a form of both the matrices and the spinors.

Any help will be appreciated. Thanks!


Due to the comments by @MadMax, I feel like I should provide some other information that I skipped (because I felt it was not needed). One of them is the negative sign in front of the Lagrangian that I included in this edit.

In our course, we defined the gamma matrices as, $$ \gamma^{0} = -i\begin{pmatrix} 0 & 1\\1&0\end{pmatrix},\ \gamma^{i} = -i\begin{pmatrix} 0 & \sigma^{i}\\-\sigma^{i}&0\end{pmatrix}$$ The Dirac adjoint was defined as, $$ \bar{\psi} = i\psi^{\dagger}\gamma^{0}$$ And the metric, as I mentioned earlier, is $(-,+,+,+)$. Finally, the Lagrangian is defined as, $$ \mathcal{L} = -\bar{\psi}(\gamma^{\mu}\partial_{\mu} + m)\psi $$

As @MadMax mentions, the mass term would have picked up a factor of $i$, but since our definition of the Dirac adjoint was different, it doesn't.

  • $\begingroup$ For metric $(-, +, +, +)$, instead of $\mathcal{L} \ = \ \bar{\psi}(\gamma^{\mu}\partial_{\mu} + m)\psi$, should your Lagrangian be $\mathcal{L} \ = \ i\bar{\psi}(\gamma^{\mu}\partial_{\mu} + m)\psi$? Assuming that we are talking about Grassmann-odd $\psi$ field, there is a missing i, right? $\endgroup$
    – MadMax
    Sep 25, 2023 at 18:30
  • $\begingroup$ We didn't have an i in class; but we defined the gamma matrices with an extra factor of i. $\endgroup$
    – ShKol
    Sep 25, 2023 at 19:25
  • $\begingroup$ "We didn't have an i in class", then you might consider asking your professor for refund of your tuition :) $\endgroup$
    – MadMax
    Sep 25, 2023 at 19:30
  • $\begingroup$ "but we defined the gamma matrices with an extra factor of i", because of this extra i factor in the gamma matrices, the mass term picks up an extra i since $\bar{\psi} = \psi^{\dagger}\gamma_0$ and there is an extra i coming from $\gamma_0$. As for the kinetic term, the extra i from $\gamma_0$ cancels out the extra i from $\gamma_\mu$, therefore it stays the same as the metric (+,-,-,-) format. $\endgroup$
    – MadMax
    Sep 25, 2023 at 19:36
  • 1
    $\begingroup$ For the Lagrangian, you can take a look at Edward Witten's paper referenced in this post: physics.stackexchange.com/questions/513768/… $\endgroup$
    – MadMax
    Sep 25, 2023 at 21:03

1 Answer 1


Give a representation of the Gamma matrices, there exists some matrix $\beta$ which satisfies $$ (\gamma^\mu)^\dagger = \beta \gamma^\mu \beta^{-1} , \qquad \beta^\dagger = - \beta . $$ The Dirac conjugate is defined by $$ {\bar \psi} \equiv i \psi^\dagger \beta . $$ Now, let's consider the Dirac Lagrangian, $$ {\cal L} = {\bar \psi} ( \gamma^\mu \partial_\mu + m ) \psi $$ Taking a complex conjugating and manipulating further, we find \begin{align} {\cal L}^* &= [ {\bar \psi} ( \gamma^\mu \partial_\mu + m ) \psi ]^* \\ &= [ i (\psi^\dagger \beta ) ( \gamma^\mu \partial_\mu + m ) \psi ]^\dagger \\ &= - i \psi^\dagger ( ( \gamma^\mu )^\dagger \overleftarrow{\partial_\mu} + m ) (\psi^\dagger \beta )^\dagger \\ &= - i \psi^\dagger ( \beta \gamma^\mu \beta^{-1}\overleftarrow{\partial_\mu} + m )\beta^\dagger \psi \\ &= - {\bar \psi} ( \gamma^\mu \overleftarrow{\partial_\mu} + m ) \psi \\ &= {\bar \psi} ( \gamma^\mu \partial_\mu + m ) \psi - \partial_\mu [ {\bar \psi} \gamma^\mu \psi ] \\ &= {\cal L} - \partial_\mu [ {\bar \psi} \gamma^\mu \psi ] . \end{align}

  • $\begingroup$ I understand the steps here - but I don't understand why taking a complex conjugate is akin to taking the Hermitian. Is it because the Lagrangian is a scalar and taking the conjugate transpose just means taking a conjugate? Also, why does there always exist such a matrix like $\beta$? In the usual conventions for the gamma matrices and such, it always seems like the relation holds coincidentally. Thanks! $\endgroup$
    – ShKol
    Sep 26, 2023 at 4:04
  • $\begingroup$ Your calculation is wrong. By $- {\bar \psi} ( \gamma^\mu \overleftarrow{\partial_\mu} + m ) \psi = {\bar \psi} ( \gamma^\mu \partial_\mu + m ) \psi - \partial_\mu [ {\bar \psi} \gamma^\mu \psi ]$, you changed the sign of the mass term. Please double check your math. $\endgroup$
    – MadMax
    Sep 26, 2023 at 15:05

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.