It puzzles me that Zee uses throughout the book this definition of covariant derivative: $$D_{\mu} \phi=\partial_{\mu}\phi-ieA_{\mu}\phi$$ with a minus sign, despite of the use of the $(+---)$ convention.

But then I see that Srednicki, at least in the free preprint, uses too the same definition, with the same minus sign. The weird thing is that Srednicki uses $(-+++)$

I looked too into Peskin & Schröder, who stick to $(+---)$ (the same as Zee) and the covariant derivative there is:

$$D_{\mu} \phi=\partial_{\mu}\phi+ieA_{\mu}\phi$$

Now, can any of you tell Pocoyo what is happening here? Why can they consistently use different signs in that definition?

  • $\begingroup$ To complement what is said below, using $(+---)$, you can see that starting from the Lorentz force you are able to deduce that $D_\mu=\partial_\mu+iq A_\mu$ with $q$ the charge of the particle. I think if you are being technical, you should use something consistent with this, but it is not stricly necessary because for most Lagrangians (I don't know if all of them) the important thing is the difference in charge's sign among particles, not every particles' sign. $\endgroup$ Oct 3, 2022 at 14:45
  • $\begingroup$ For the record, the definition $D_\mu$ is independent of the sign of the metric. With (-+++) is the same definition of $D_\mu$. $\endgroup$ Oct 3, 2022 at 15:05
  • $\begingroup$ @BorisValderrama that seems to contradict the answer below by Qmechanic $\endgroup$ Dec 5, 2023 at 8:21
  • $\begingroup$ @infinitezero you are right, I think I am wrong about the independence of the metric. I suppose I didn't check it at the moment $\endgroup$ Jan 3 at 6:32

2 Answers 2


We will work in units with $c=1=\hbar$. The $4$-potential $A^{\mu}$ with upper index is always defined as

$$A^{\mu}~=~(\Phi,{\bf A}). $$

1) Lowering the index of the $4$-potential depends on the sign convention

$$ (+,-,-,-)\qquad \text{resp.} \qquad(-,+,+,+) $$

for the Minkowski metric $\eta_{\mu\nu}$. This Minkowski sign convention is used in

$$\text{Ref. 1 (p. xix) and Ref. 2 (p. xv)} \qquad \text{resp.} \qquad \text{Ref. 3 (eq. (1.9))}.$$

The $4$-potential $A_{\mu}$ with lower index is $$A_{\mu}~=~(\Phi,-{\bf A}) \qquad \text{resp.} \qquad A_{\mu}~=~(-\Phi,{\bf A}).$$

Maxwell's equations with sources are

$$ d_{\mu}F^{\mu\nu}~=~j^{\nu} \qquad \text{resp.} \qquad d_{\mu}F^{\mu\nu}~=~-j^{\nu}. $$

The covariant derivative is

$$D_{\mu} ~=~d_{\mu}+iqA_{\mu}\qquad \text{resp.} \qquad D_{\mu} ~=~d_{\mu}-iqA_{\mu}, $$

where $q=-|e|$ is the charge of the electron.

2) The sign convention for the elementary charge $e$ is

$$e~=~-|e| ~<~0 \qquad \text{resp.} \qquad e~=~|e|~>~0.$$

This charge sign convention is used in

$$\text{Ref. 1 (p. xxi) and Ref. 3 (below eq. (58.1))} \qquad \text{resp.} \qquad \text{Ref. 2.}$$


  1. M.E. Peskin and D.V Schroeder, An Introduction to QFT.

  2. A. Zee, QFT in a nutshell.

  3. M. Srednicki, QFT.

  • $\begingroup$ FYI: M.E. Peskin and D.V Schroeder, An Introduction to QFT, has Minkowski sign convention $(+,-,-,-)$ (p. xix); charge sign convention $e=-|e|$ (p.809); and covariant derivative $D_{\mu}=d_{\mu}+iqA_{\mu}$ (p.303). M.D. Schwartz, QFT & the standard model, has Minkowski sign convention $(+,-,-,-)$ (p.12,817); charge sign convention $e=|e|$ (p.818); and covariant derivative $D_{\mu}=d_{\mu}-iqA_{\mu}$ (p.140,224,818). $\endgroup$
    – Qmechanic
    Feb 24, 2013 at 23:17
  • 3
    $\begingroup$ FYI: W. Siegel, Fields, has Minkowski sign convention $(-,+,+,+)$ (p.55); charge sign convention $e=|e|$ (p.184,204); and covariant derivative $D_{\mu}=d_{\mu}+iqA_{\mu}$ (p.184,204), which is opposite. [Also note that Siegel's definition (p.169ff) of the action $S=\int\! dt ({\rm Pot.terms - Kin.terms})$ is opposite of the standard definition.] $\endgroup$
    – Qmechanic
    Feb 25, 2013 at 21:25
  • 2
    $\begingroup$ FYI: (i) C. Itzykson and J.-B. Zuber, QFT, has Minkowski sign convention $(+,-,-,-)$ (eq.A-1); charge sign convention $e=-|e|$; and covariant derivative $D_{\mu}=d_{\mu}+iqA_{\mu}$ (eq.4-77), like e.g. Ref. 1, and e.g. Bjorken and Drell. (ii) S. Weinberg, The Quantum Theory of Fields, has Minkowski sign convention $(-,+,+,+)$ (p.xxv); has charge sign convention e=|e| (p.xxvi); and covariant derivative $D_{\mu}=d_{\mu}-iqA_{\mu}$ (eq.8.1.21) $\endgroup$
    – Qmechanic
    Feb 26, 2013 at 16:48
  • 1
    $\begingroup$ thanks. If it is of any use, some books in my personal library follow: Schutz 2009(-+++); Chetaev 1989 (---+); Einstein 1921(---+); Wald 1984(-+++); Dirac 1967(+---); Susskind&Lindesay 2005(+---); Choudhuri 2010(-+++); Carroll&Ostlie 2007(+---); Tong 2007 classnotes on QFT (+---); Tong 2009 classnotes on ST (-+++...+), 't Hooft 2009 notes on BHs (-+++); Schneider&Ehlers&Falco 1989 (+---); Zee 2010 (+---). I write the signature below the title, so that I don't need to find it out again and again every time I consult something, I guess I'll have to add the electron charge now in QFT books. $\endgroup$ Feb 26, 2013 at 17:21
  • 5
    $\begingroup$ Why the h*** don't they stick to the original convention in the very first paper for everything? It is really so painful, or do they have some need to appear as original? The Susskind lectures have now the + sign, and the David Tong notes have God knows what convention in the Faraday tensor. Everytime I try to cross some details between books and specially the first time I use a new document or internet page, I have to spend the hell of a time figuring out which is the arbitrary convention of that author. The first ten times it was even funny, now it is a pain in the neck. Damn them all! $\endgroup$ May 15, 2013 at 0:06

A late answer, but important in my opinion.

There is another aspect: the sign in the covariant derivative also depends on the sign convention used in the gauge transformation!

This is something that is overlooked a lot.

If the Dirac field transforms as $$ \psi \rightarrow e^{ig\alpha} \psi, $$ then the covariant derivative is defined as $$ D_\mu = \partial_\mu - ig A_\mu. $$

But if the Dirac field transforms as $$ \psi \rightarrow e^{-ig\alpha} \psi, $$ then the covariant derivative is defined as $$ D_\mu = \partial_\mu + ig A_\mu. $$

It is interesting to see that in both cases, the gauge field transforms as $$ A_\mu \rightarrow A_\mu + \partial_\mu \alpha. $$

Peskin and Schroeder use the first convention, with coupling constant $g=-|e|$ (this is indeed a bit confusing, but makes sense from a physical point of view, as the electromagnetic coupling to an electron should be negative). They start using the more general definitions with a $g$ from the moment they go to non-Abelian theories in chapter 15, and keep using the first convention.

The second convention is used eg. in Collins' new book "Foundations on pQCD". This has somewhat become the de-facto standard for the TMD community (transverse momentum dependent PDFs), so people should realise that they cannot simply combine formulas from this book with eg. Peskin and Schroeders'.

Btw, in the non-Abelian case, the sign change also propagates into the definition of the gauge field tensor (in front of the interaction part)

In these examples I assumed $(+---)$ for all, as is standard for particle physics (while $(-+++)$ is standard for GR and string theory/susy).

  • $\begingroup$ If you want to know in which formulas this sign convention possibly makes a difference, just substitute $g\rightarrow -g$. $\endgroup$ Aug 16, 2014 at 10:49
  • $\begingroup$ Why the downvote? $\endgroup$ Sep 6, 2014 at 12:45
  • 1
    $\begingroup$ I downvoted the answer because it doesn't contain anything that answers the question. It only adds a relationship to one more question, about another sign, and therefore deepens, not clarifies, the confusion. $\endgroup$ Sep 6, 2014 at 12:58
  • $\begingroup$ @freddieknets that seems wrong. Peskin & Schroeder define their gauge covariant derivative as $D_\mu\equiv\partial+ieA_\mu(x)$ on page 78, and define $e=-|e|$ on the same page. They also define $\psi(x)\to e^{i\alpha(x)}\psi(x),\quad A_\mu\to A_\mu-\frac 1 e \partial_\mu \alpha(x)$. $\endgroup$ Jul 15, 2019 at 23:16
  • $\begingroup$ @alexchandel This is still the same. For historical reasons they don’t extract the coupling constant from the gauge transformation potential. Replace $\alpha(x)$ in P&S with $-e \,\alpha(x)$ and you get the same formulation as me. $\endgroup$ Jul 16, 2019 at 22:30

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.