My question: why the momentum of the $A_0$-field in non abelian gauge theories is $0$?

In my class on gauge theories, we end up with this Lagrangian:

$$ \mathcal{L}=F_a^{\mu \nu}F^a_{\mu \nu} $$


$$F_a^{\mu \nu}=\partial^{\mu} A_a^{\nu}-\partial^\mu A_a^\mu-g f^{abc} A_b^\mu A_c^\nu$$

At order $0$ in $g$, we find the same term than the abelian case.

At order $1$ in $g$, we find :

$$ 2g f^{abc}(\partial_\mu A_\nu^a-\partial_\nu A_\mu^a)A^\mu_b A^\nu_b$$

Where $f^{abc}$ are totally antisymmetric : $[T_a,T_b]=if_{abc}T_c$ where $T_k$ are the generators of the Lie algebra.

My question is :

In my course, it is written that we have the same problem about canonical quantization here than in abelian theories : $\pi_0^a=0$ (the momentum associated to the fields is $0$).

For the order $0$ I totally agree bc it has the same form as in the Abelian theory.

But for the order $1$ it is not $0$ ?

Indeed :

$$B=f^{abc}(\partial_\mu A_\nu^a-\partial_\nu A_\mu^a)A^\mu_b A^\nu_b$$

$$ \frac{\partial B}{\partial \dot{A}^a_{\beta}}=f^{abc}A^0_b A^{\beta}_c-f^{abc}A_b^{\beta}A^0_c=f^{abc}A^0_b A^{\beta}_c-f^{acb}A_c^{\beta}A^0_b=A^0_b A^{\beta}_c(f^{abc}-f^{acb})\neq 0$$

It is not $0$ because $f^{abc}$ is totally antisymmetric...

Thus, why in non Abelian gauge theories we have the same problem about canonical quantization than in the Abelian theories ?

  • $\begingroup$ What do you mean the momentum is zero for Abelian gauge fields? For electromagnetism in the Weyl gauge, for example, the canonically conjugate field momentum to the vector potential is (up to a sign) the electric field. $\endgroup$ Dec 7, 2017 at 17:32
  • $\begingroup$ @SeanE.Lake I mean that if we compute : $\pi_\mu=\frac{\partial \mathcal{L}}{\partial (\partial_0 A^{\mu})}$ we find $0$ for the Abelian gauge fields. Because we have for the Abelian case : $\mathcal{L}=(\partial_\mu A_\nu-\partial_\nu A_\mu)(\partial^\mu A^\nu - \partial^\nu A^\mu)$ $\endgroup$
    – StarBucK
    Dec 7, 2017 at 17:41
  • 1
    $\begingroup$ That's certainly true for $\mu=0$, but not for $\mu\in\{1,2,3\}$. The Lagrangian is given by $E_i E_i -B_iB_i=(\partial_i A_0 +\partial_0A_i)(\partial_i A_0 +\partial_o A_i) - (\epsilon_{ijk} \partial_j A_k)(\epsilon_{ilm}\partial_lA_m)$. Obviously, there's no time derivative of $A_0$, but there are quadratic time derivatives of $A_i$. $\endgroup$ Dec 7, 2017 at 17:48
  • $\begingroup$ I think the problem here is that what you're calling "the momentum of the field" is only the momentum of the 4-potential's time-like component. The other components have non-trivial momenta. $\endgroup$ Dec 7, 2017 at 17:52
  • $\begingroup$ @SeanE.Lake Huh yeah you are right ! It is only true for the $0$ components then we have the exact same "problem" as in Abelian cause theory. Sorry I was a little confused. Thanks !! $\endgroup$
    – StarBucK
    Dec 7, 2017 at 17:55

1 Answer 1


Not all of the canonical field momenta in the Abelian gauge field are zero. The electromagnetic Lagrangian (the prototypical example of an Abelian field) is \begin{align} \mathcal{L} & = -\frac{1}{4} F^{\mu\nu}F_{\mu\nu} \\ & = \frac{1}{2} \vec{E}\cdot\vec{E} - \frac{1}{2}\vec{B}\cdot\vec{B} \\ & = \frac{1}{2} \left(\partial_i A^0 + \partial_0 A^i\right)\left(\partial_i A^0 + \partial_0 A^i\right) - \frac{1}{2}\left(\epsilon_{ijk}\partial_j A^k\right)\left(\epsilon_{ilm}\partial_l A^m\right) \end{align} In Heaviside-Lorentz units with $c=1$ and metric signature $(+,-,-,-)$. The field momentum canonically conjugate to $A_\mu$ is given by \begin{align} \Pi^\mu & \equiv \eta^{\mu\nu}\frac{\partial \mathcal{L}}{\partial \left(\partial_0 A^\nu\right)} \Rightarrow\\ \Pi^i & = -\partial_i A^0 - \partial_0 A^i. \end{align} Note that $\Pi^0$ is identically $0$ because $\partial_0 A^0$ doesn't even appear in the Lagrangian. For this reason, $A^0$ can be regarded as a non-dynamical, or auxiliary, field. This is a manifestation of the Gauge symmetry - there is an entire field worth of freedom in $A^\mu$ that has no inertia, and can therefore take any value the problem needs it to have at any time. The form that these unconstrained degrees of freedom take is set by the gauge fixing condition used. This can be seen in gauges that are frequently fixed by adding a term to the Lagrangian that can add $\partial_0 A^0$ terms to the Lagrangian (e.g. $R_\xi$ gauges).

Non-Abelian fields act like multiple interacting Abelian fields, so this picture of the Lagrangian not depending on $\partial_0A^0$ extends to those fields, too.

All that said, you could imagine integrating the cross terms from expanding $\vec{E}^2$ by parts to swap the space and time derivatives to get $$\mathcal{L} = \frac{1}{2} \partial_iA^0 \partial_iA^0 + \partial_0A^0 \partial_iA^i + \frac{1}{2} \partial_0A^i \partial_0A^i + \vec{B}\cdot\vec{B} + (\mathrm{surface\ terms}).$$ I haven't actually looked at this last Lagrangian (I literally just derived it), but I don't think it would say anything particularly interesting since it differs from the Lagrangian without $\partial_0A^0$ by a couple of surface terms ($-\partial_0\left[A^0\partial_iA^i\right]+\partial_i\left[A^0\partial_0A^i\right]$), and is thus equivalent. Odd to note, though, that in this Lagrangian $\Pi^0=\nabla\cdot\vec{A}$ is exactly what is set to zero in the Coulomb gauge.


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.