For any electromagnetic field, it is easy to impose the Coulomb gauge condition ${\bf\nabla}\cdot{\bf A}=0$. To start with, if ${\bf \nabla}\cdot{\bf A}_{\rm old}\neq 0$, the trick is to make a gauge transformation $${\bf A}_{\rm old}\to {\bf A}_{\rm new}={\bf A}_{\rm old}+{\bf \nabla}\theta({\bf x},t)\tag{1a}$$ such that the new vector potential ${\bf A}_{\rm new}$ satisfies, ${\bf \nabla}\cdot{\bf A}_{\rm new}=0$. This is achieved by choosing the gauge function $\theta({\bf x},t)$ to be a solution of the Poisson's equation $$\nabla^2\theta({\bf x},t)=-{\bf \nabla}\cdot{\bf A}_{\rm old}\tag{1b}$$ which can almost always be solved.

Now let us consider a slightly more nontrivial job. For a free electromagnetic field, it is always possible to choose a gauge such that both $\phi_{\rm new}=0$ and ${\bf \nabla}\cdot{\bf A}_{\rm new}=0$. I wonder if this can be achieved by a single gauge transformation. To start with, again if $\phi_{\rm old}\neq 0$ and ${\bf \nabla}\cdot{\bf A}_{\rm old}\neq 0$, can we make a single gauge transformation $$\phi_{\rm old}\to\phi_{\rm new}=\phi_{\rm old}-\frac{\partial\theta}{\partial t}, {\bf A}_{\rm old}\to {\bf A}_{\rm new}={\bf A}_{\rm old}+\nabla\theta\tag{2}$$ to achieve $$\phi_{\rm new}=0, ~{\bf \nabla}\cdot{\bf A}_{\rm new}=0\tag{3}$$ in one shot i.e. by choosing a single gauge function $\theta({\bf x},t)$? Unlike the previous case, $\theta$ has to satisfy two equations now: $$\frac{\partial}{\partial t}\theta({\bf x},t)=\phi_{\rm old}, ~~ \nabla^2\theta({\bf x},t)=-{\bf \nabla}\cdot{\bf A}_{\rm old}.\tag{4}$$ What is the gurrantee that both the equations in $(4)$ can be simulaneusly solved? Or should it really be done in two succesive steps?

In a two-step process, if we do a first gauge transformation to make $\phi=0$, what is the guarantee that by doing a second gauge transformation to satisfy $\nabla\cdot{\bf A}=0$, I would not regenerate $\phi$? It isn't very transparent.

  • 4
    $\begingroup$ Gauge transformations form a group, the two gauge transformations performed one after the other can be viewed as a single gauge transformation. $\endgroup$
    – jacob1729
    Apr 15, 2020 at 14:57
  • $\begingroup$ Sure. But I was wondering if we do a first gauge transformation to make $\phi=0$, what is the guarantee that by doing a second gauge transformation to satisfy $\nabla\cdot{\bf A}=0$, I would not regenerate $\phi$. It isn't very transparent. @jacob1729 $\endgroup$
    – SRS
    Apr 15, 2020 at 15:02
  • $\begingroup$ I believe jacob's point (or at least the one I would make) is that the phrase "single gauge transformation" or the idea of "successive steps" is meaningless. You're just asking whether $\phi = 0$ and $\nabla\cdot A = 0$ is an admissible gauge, i.e. whether this state of the field can be achieved with gauge transformations at all. $\endgroup$
    – ACuriousMind
    Apr 15, 2020 at 15:09
  • $\begingroup$ @ACuriousMind That's fair! But for the first case, I have sketched the intermediate step i.e. how to go to the desired gauge. In the second case, how to go to the desired gauge i.e., $\phi=\nabla\cdot{\bf A}=0$ is not very clear to me. $\endgroup$
    – SRS
    Apr 15, 2020 at 15:14
  • $\begingroup$ Possible duplicates: physics.stackexchange.com/q/1250/2451 and links therein. $\endgroup$
    – Qmechanic
    Apr 15, 2020 at 16:31

1 Answer 1


EDIT: This answer has been substantially edited since first posted.

For the two-step process. This gauge transformation is only possible if we are in the absence of charges. Then $\nabla \cdot \mathbf E=0$, which means for a general gauge $$\nabla^2 \phi +\frac{\partial}{\partial t} \nabla \cdot \mathbf A =0$$ Now do a gauge transformation that removes $\phi$. This gauge transformation can be both space and time dependent. Now you are in a gauge where $\phi=0$ and $\nabla \cdot \mathbf A \neq 0$. The above implies

$$\frac{\partial}{\partial t} \nabla \cdot \mathbf A =0$$

Therefore, you can now do a time-independent gauge transformation to set $\nabla \cdot \mathbf A =0$. Since the transformation is time-independent, it won't affect $\phi$.

For the single-step process, you want to solve $$\frac{\partial}{\partial t}\theta({\bf x},t)=\phi_{\rm old}, ~~ \nabla^2\theta({\bf x},t)=-{\bf \nabla}\cdot{\bf A}_{\rm old}$$ Roughly speaking, the first equation is solved by $$\theta({\bf x},t) = \int_{t_0}^t \phi({\bf x},\tau) d\tau+f(\mathbf x)$$ where $f(\mathbf x)$ is arbitrary. Then $$\nabla^2 \theta = \int_{t_0}^t \nabla^2\phi({\bf x},\tau) d\tau+\nabla^2 f(\mathbf x)$$ Now we can use the condition above, valid in absence of charges, to get, $$\nabla^2 \theta = -\int_{t_0}^t \frac{\partial}{\partial \tau} \nabla \cdot \mathbf A({\bf x},\tau) d\tau+\nabla^2 f(\mathbf x)=$$ $$=-\nabla \cdot \mathbf A({\bf x},t)+\nabla \cdot \mathbf A({\bf x},t_0)+\nabla^2 f(\mathbf x)$$ Now you can pick $f$ so that $$\nabla \cdot \mathbf A({\bf x},t_0)+\nabla^2 f(\mathbf x)=0$$ and then the condition $$\nabla^2 \theta =-\nabla \cdot \mathbf A$$ is satisfied.

  • $\begingroup$ How is an equation of the form $\frac{\partial}{\partial t}\theta_1(t)=\phi_{\rm old}(t,{\bf x})$ make sense where RHS is a function of ${\bf x}$ (in general) but not the LHS? Am I missing something? $\endgroup$
    – SRS
    Apr 15, 2020 at 15:11
  • $\begingroup$ Apologies, that was incorrect. I have updated my answer. $\endgroup$
    – John Donne
    Apr 15, 2020 at 17:28

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.