Let $(\eta^{\mu\nu})=\operatorname{diag}(+1,-1,-1,-1)$. The Dirac matrices $\gamma^\mu$, $\mu=0,1,2,3$ satisfy by definition $$\{\gamma^\mu,\gamma^\nu\}=2\eta^{\mu\nu}\tag{1}\label{1}$$ where $\{A,B\}=AB+BA$ is the anti-commutator.

If we have another collection $\tilde\gamma^\mu$ of matrices of the same size satisfying the same commutation relations \eqref{1} then the Skolem–Noether theorem implies that there exists an invertible matrix $C$ such that $$\tilde\gamma^\mu=C\gamma^\mu C^{-1}\mbox{ for all } \mu.\tag{2}\label{2}$$

I am interested in the following situation. Let us require in addition to \eqref{1} that for Hermitian conjugates one has $$(\gamma^0)^*=\gamma^0, (\gamma^i)^*=-\gamma^i \text{ for }i=1,2,3.\tag{3}\label{3}$$ Is it possible to classify four tuples of matrices $\gamma^\mu$, $\mu=0,\dotsc,3$, satisfying \eqref{1} and \eqref{3}? Is there a generalization of such a classification to higher dimensions?

Remark 1. If one takes in \eqref{2} the matrix $C$ to be unitary then $\tilde\gamma^\mu$ also satisfy \eqref{3}.

Remark 2. My understanding is that the motivation to consider conditions \eqref{2} goes back to the original motivation of Dirac to construct an equation on the wave function $\psi$ of a free electron which would be not only relativistic and implied the Klein–Gordon equation, but also $\lvert\psi\rvert^2$ would be the time component of a conserved 4-current.


1 Answer 1


If you require that the matrix $C$ in \eqref{2} preserves the involutions \eqref{3}, then you get the consequence $\tilde{\gamma}^\mu = C C^* \tilde{\gamma}^\mu (C C^*)^{-1}$, which then implies that $C C^* = u$ is some unit in the center of the Clifford algebra such that $u = u^*$. Equivalently $C^* = u C^{-1}$. Rescaling $C=v D$, where $v$ is another scalar unit (which obviously cancels in the formula $C \gamma C^{-1} = D \gamma D^{-1}$), we get $D^* = \frac{u}{v^* v} D^{-1}$. So we can normalize the involution properties of $C$ by choosing $u/(v^* v)$ to be some canonical value.

The center of a complex Clifford algebra is either 1- or 2-dimensional, in a pattern that depends on the number of generators (which I'm always too lazy to look up). But supposing the center is 1-dimensional, that is consisting of scalar matrices $\mathbb{C} I$, for any $u$ we can choose $v$ such that $u/(v^* v) = \pm 1$. In that case, if you can find one tuple of $\gamma$-matrices satisfying \eqref{3}, the rest are classified by \eqref{2} where it is sufficient to take $C^* = \pm C^{-1}$. The case of the 2-dimensional center can be left as an exercise.

  • $\begingroup$ Thank you. As far as I understand there are no matrices such that $C^*=-C^{-1}$. Indeed $0\leq tr(C^*C)=tr(-I)<0$. $\endgroup$
    – asv
    Sep 23 at 5:45
  • $\begingroup$ @asv Of course, you're right. I missed the positivity condition on $u$, in addition to its self-adjointness. $\endgroup$ Sep 23 at 15:31

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.