Related to this earlier A common standard model Lagrangian mistake?

Here I am treating Dirac equation of Dirac field as QFT.

You may want to consider the quantized version or the classical version.

Now my question is:

Is $\bar{\psi} \psi$ its own complex (*)? its own transpose ($T$)? its own hermitian conjugate ($\dagger$)?

Naively in the Dirac lagrangian we have its mass term: $m\bar{\psi} \psi =m \psi^\dagger \gamma^0 \psi$ based on the standard notation in Peskin's QFT book (using the Weyl representation).

Since $\bar{\psi} \psi$ is a Lorentz scalar number (that depends on the spacetime point $(t,x)$) and integrated over the 4d spacetime, I believe it is a real number as a Lorentz scalar (yes), so $$ \bar{\psi} \psi \in \mathbb{R} (?). $$ If this is true, we may check that $T$ as a transpose and $*$ as a complex conjugation, these should be true $$ (\bar{\psi} \psi)^T=\bar{\psi} \psi ? \tag{(i)} $$ $$ (\bar{\psi} \psi)^*=\bar{\psi} \psi ? \tag{(ii)} $$ However, a simple calculation shows that: $$ (\bar{\psi} \psi)^T=(\psi^\dagger \gamma^0 \psi)^T={\psi}^T \gamma^0 \psi^*= \sum_{a,b}{\psi}_a \gamma^0_{ab} \psi^*_b= \sum_{a,b} \{{\psi}_a,\psi^*_b\} \gamma^0_{ab} - \sum_{a,b} \psi^*_b \gamma^0_{ab} {\psi}_a =-\psi^\dagger \gamma^0_{} {\psi} = -\bar{\psi} \psi $$ here $a,b$ are 4-component spinor indices. We use the fermi anticommutation relation $\{{\psi}_a,\psi^*_b\}={\psi}_a\psi^*_b +\psi^*_b \gamma^0_{ab} {\psi}_a =\delta^3$ which is up to a delta function for the equal time anticommutation relation. We derive the last equality based on $\gamma^0_{ab}=\gamma^0_{ba}$, so $\sum_{a,b} \psi^*_b \gamma^0_{ba} {\psi}_a = -\bar{\psi} \psi$. This means we derive that: $$ (\bar{\psi} \psi)^T=-\bar{\psi} \psi. \tag{(i')} $$ Similarly, a simple calculation shows that: $$ (\bar{\psi} \psi)^*=(\psi^\dagger \gamma^0 \psi)^*={\psi}^T \gamma^0 \psi^*= \sum_{a,b}{\psi}_a \gamma^0_{ab} \psi^*_b= \sum_{a,b} \{{\psi}_a,\psi^*_b\} \gamma^0_{ab} - \sum_{a,b} \psi^*_b \gamma^0_{ab} {\psi}_a =-\psi^\dagger \gamma^0_{} {\psi} = -\bar{\psi} \psi $$ here $a,b$ are 4-component spinor indices. This means we derive that: $$ (\bar{\psi} \psi)^*=-\bar{\psi} \psi. \tag{(ii')} $$

question 1:

I imagine that the Dirac path integral implies a $e^{ -i m \bar{\psi} \psi}$ phase with $ \bar{\psi} \psi$ shall be a real number. But it turns out to lead to a conflict between the equations (i), (ii) to that of (i'), (ii'). Is there a simple explanation?

question 2:

Is the delta function important here? the equality above is not precise since we use $\{{\psi}_a,\psi^*_b\}={\psi}_a\psi^*_b +\psi^*_b \gamma^0_{ab} {\psi}_a =\delta^3$.

question 3:

However, regardless the conflict of its complex (*) and its transpose ($T$), we always have its hermitian conjugate ($\dagger$) to be itself $$ (\bar{\psi} \psi)^\dagger=+\bar{\psi} \psi. \tag{(iii)} $$ Why is that?

There may be good physics intuition behind.

  • $\begingroup$ Not sure who voted down -- what I asked is reasonable thing about taking "complex (*)? its transpose (𝑇)? its hermitian conjugate (†)?" in which coordinates/spinors/matrix/space? -- please leave an answer if you find what I said is trivial or obvious $\endgroup$ Mar 29, 2022 at 2:31
  • $\begingroup$ $\psi$ is a Grassmannian variable. All your manipulations are valid only for complex-valued spinors. You have to be careful about the extra minus signs you get when you move fermions across each other (eg. In your transpose calculation) $\endgroup$
    – Prahar
    Mar 29, 2022 at 8:47
  • $\begingroup$ To be clear, if $A$ and $B$ are Grassmann-valued, then $(AB)^T = - BA$ and $(AB)^* = - A^*B^*$. Hermitian conjugate is complex conjugation + transpose so $(AB)^\dagger = ((AB)^T)^* = - (B^TA^T)^* = + B^\dagger A^\dagger$ so the formula for Hermitian conjugation is exactly the same as for complex-valued matrices. $\endgroup$
    – Prahar
    Mar 29, 2022 at 8:54
  • $\begingroup$ Thanks Prahar I agree with you -- your answer agrees with mine. I think the other answer by Gold only treats $\psi$ as a classical variable. Does Grassmann-valued count as a quantized variable? Or can Grassmann-valued be a classical variable? $\endgroup$ Mar 29, 2022 at 12:44
  • $\begingroup$ How about the commutator $\{{\psi}_a,\psi^*_b\}={\psi}_a\psi^*_b +\psi^*_b \gamma^0_{ab} {\psi}_a =\delta^3$? Is the $\delta^3$ involved? Should it be equal time on a space $\delta^3$ or should it be a spacetime $\delta^4$? Do we need to worry about the $\delta$ function? could you be so kind to write an answer? (I think Gold answer may be partial only.) $\endgroup$ Mar 29, 2022 at 12:47

2 Answers 2


In classical field theory, $\psi$ is a Grassmann-valued field so you have to be super careful with complex conjugation and transpose. If $A$ and $B$ are Grassmann-valued matrices, then $$ (AB)^T = - B^TA^T , \qquad (AB)^* = - A^* B^* , \qquad (AB)^\dagger = + B^\dagger A^\dagger $$ As a comparison, for complex-valued matrices, the corresponding formulae are $$ (AB)^T = + B^TA^T , \qquad (AB)^* = + A^* B^* , \qquad (AB)^\dagger = + B^\dagger A^\dagger $$ If you keep track of these extra minus signs, you will find that ${\bar \psi} \psi$ is a real-valued number.

  • $\begingroup$ I think 𝜓¯𝜓 is a real valued Grassman number? thanks! $\endgroup$ Mar 29, 2022 at 13:55
  • $\begingroup$ Product of two Grassmann numbers is a complex number so ${\bar \psi} \psi$ is real-valued. It is NOT Grassmann-valued. $\endgroup$
    – Prahar
    Mar 29, 2022 at 13:56
  • $\begingroup$ But I thought that only $\int [D {\psi}][D \bar{\psi}]𝜓¯𝜓$ or $\int [D {\psi}][D \bar{\psi}] \exp(a 𝜓¯𝜓)$ gives a number (NOT Grassmann)? The $𝜓¯𝜓$ is still a real valued Grassman number? $\endgroup$ Mar 29, 2022 at 13:58
  • $\begingroup$ Product of two Grassmann numbers commutes with ALL other numbers (complex or Grassmann). Therefore, it is not a Grassmann number (Grassmann numbers ANTICOMMUTE with other Grassmann numbers). $(\theta \eta) \psi = - \theta \psi \eta = + \psi ( \theta \eta)$. Thus, $\theta\eta$ commutes with $\psi$. This implies that $\theta\eta$ is a complex number. $\endgroup$
    – Prahar
    Mar 29, 2022 at 14:01
  • $\begingroup$ The action has to be a real number, right? How can you have a term of the form ${\bar \psi} \psi$ in the action if it's not real?!?!? $\endgroup$
    – Prahar
    Mar 29, 2022 at 14:02

Well, by definition we have, $$\bar \psi = \psi^\dagger \gamma^0=(\psi^T)^\ast \gamma^0\tag{1}.$$

Now, $\psi$ is a $4\times 1$-matrix, so $\psi^T$ is a $1\times 4$-matrix. Clearly $(\psi^T)^\ast$ will also have the same dimensionality. Since $\gamma^0$ is a $4\times 4$-matrix it follows $(\psi^T)^\ast\gamma^0$ is again a $1\times 4$-matrix. So this establishes that $\bar \psi$ is just a row vector.

In that case $\bar \psi \psi$ is a row vector times a column vector. The result is a $1\times 1$-matrix, or just a number. This by itself already establishes that $(\bar \psi \psi)^T = \bar \psi \psi$.

So: Is $\bar \psi \psi$ its transpose? Yes.

Now even though $\bar \psi \psi $ is just a number it is expressed as a product of matrices and therefore we can use the following identity $$(\bar \psi \psi)^\dagger = \psi^\dagger \bar \psi^\dagger = \psi^\dagger (\psi^\dagger \gamma^0)^\dagger = \psi^\dagger(\gamma^0)^\dagger \psi = \psi^\dagger \gamma^0 \psi = \bar \psi \psi\tag{2}$$

where we have used that $\gamma^0$ is hermitian.

So: Is $\bar \psi \psi$ its hermitian conjugate? Yes.

Finally combining the two we answer the other question: Indeed we have that $$(\bar \psi \psi)^\ast = [(\bar \psi \psi)^T]^\ast = (\bar \psi \psi)^\dagger = \bar \psi \psi\tag{3}.$$

So: Is $\bar \psi \psi$ its complex conjugate? Yes.

  • $\begingroup$ thanks so much for your answer, but there may be different meanings when we take complex conjugation and transpose. $\endgroup$ Mar 28, 2022 at 22:26
  • $\begingroup$ I think you are doing the classical version (but there is an issue that the Dirac $\psi$ is not a normalized wavefunction see Peskin chap 3). But I am doing the quantized version where when I exchange the fermion operator, there are (-1) sign. Maybe we can compare our calculations carefully :) $\endgroup$ Mar 28, 2022 at 22:28
  • $\begingroup$ My calculation follows closely a related one as eq. 3.147 of Peskin QFT chap 3. $\endgroup$ Mar 28, 2022 at 22:29
  • $\begingroup$ Not sure who voted down -- what I asked is reasonable thing about taking "complex (*)? its transpose (𝑇)? its hermitian conjugate (†)?" in which coordinates/spinors/matrix/space? $\endgroup$ Mar 29, 2022 at 2:31
  • $\begingroup$ thanks, I upvote anyway. $\endgroup$ Mar 29, 2022 at 12:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.