In the book Advanced Classical Electromagnetism by Robert Wald, it was shown if $\psi$ satisfy the homogeneous wave equation $$\square\psi=0$$ then we have $$\psi\left(x^{\mu}\right)=-\frac{1}{c} \int_{x^{\prime 0}=0}\left[\psi\left(x^{\prime \mu}\right) \partial_{0}^{\prime} G_{\mathrm{ret}}\left(x^{\mu}, x^{\prime \mu}\right)-G_{\mathrm{ret}}\left(x^{\mu}, x^{\prime \mu}\right) \partial_{0}^{\prime} \psi\left(x^{\prime \mu}\right)\right] d^{3} x^{\prime}\tag{5.133}$$ where the retarded Green function is defined as $$G_{\text {ret }}\left(t, x ; t^{\prime}, x^{\prime}\right)= \begin{cases}0, & \text { for } t<t^{\prime} \\ \frac{\delta\left(t-t^{\prime}-\left|x-x^{\prime}\right| / c\right)}{4 \pi\left|x-x^{\prime}\right|}, & \text { for } t>t^{\prime}\end{cases}\tag{5.50}$$ but in the next line, he suddenly has $$\psi(t, \boldsymbol{x})=\frac{1}{4 \pi} \int_{S^{\prime}}\left[\frac{1}{r^{\prime 2}} \psi\left(\theta^{\prime}, \varphi^{\prime}\right)+\frac{1}{r^{\prime}}\left(\frac{1}{c} \frac{\partial \psi}{\partial t}\left(\theta^{\prime}, \varphi^{\prime}\right)+\hat{\boldsymbol{r}}^{\prime} \cdot \nabla \psi\left(\theta^{\prime}, \varphi^{\prime}\right)\right)\right] r^{\prime 2} \sin \theta^{\prime} d \theta^{\prime} d \varphi^{\prime}\tag{5.134}$$ for $t>0$. I find this step particularly hard to follow, here are some of my questions:

  1. Can we assume that the usual rule defining derivative of delta function (distribution) continue to hold in this case, such that $\delta' f=-\delta f'$?
  2. If so, where does the first and third term of the integrand of (5.134) come from? Since the derivative is taken with respect to the zeroth (i.e. time) component in (5.133), I feel there shouldn't be a gradient term in (5.134). Also, I wonder why the first term scale as $1/r'^2$...

1 Answer 1

  1. You can integrate by part the $\delta$-function as usual, under an integral. Here, the $\delta'$ which appears is integrated by part in a integral over the spatial variable $x$. Another way of saying this is that because the $\delta$ function is a function of $t-t'-|x-x'|/c$, the time derivative is changed into a spatial derivative.

  2. Let us compute.

$(5.133)$ is : $$\psi(t,x) = -\frac{1}{c^2}\int\Big[\psi(x',0) \frac{\delta'(t-|x-x'|/c)}{4\pi |x-x'|} - \frac{\delta(t-|x-x'|/c)}{4\pi|x-x'|}\frac{d\psi}{dt}(x',0)\Big]\text d^3x'$$

Take spherical coordinates around the point $x$, so that $\text d^3 x' = r'^2 \sin(\theta')\text d\theta' \text d\varphi'\text dr'$ and :

$$\psi(t,x) = -\frac 1{c^2}\int \Big[\psi(r',\theta',\varphi')\frac{\delta'(t-r'/c)}{4\pi r'}- \frac{\delta(t-r'/c)}{4\pi r'}\frac{d\psi}{dt}(r',\theta',\varphi')\Big]r'^2 \sin(\theta')\text d\theta' \text d\varphi'\text dr'$$

We can then integrate by part over $r$ in the first term to get :

\begin{align} \psi(t,x) &= \frac{1}{4\pi c^2}\int \Big[ c\frac{d}{dr'}\Big(r'\psi(r',\theta',\varphi')\Big)\delta(t-r'/c)+ \delta(t-r'/c)r'\frac{d\psi}{dt}(r',\theta',\varphi')\Big] \sin(\theta')\text d\theta' \text d\varphi'\text dr' \\ &=\frac{1}{4\pi}\int_{S'} \Big[\psi(\theta',\varphi') + r'\frac{d\psi}{dr'}(\theta',\varphi') + \frac{r'}c \frac{d\psi}{dt}(\theta',\varphi')\Big]\sin(\theta')\text d\theta' \text d\varphi' \end{align}

Now, we use $\frac{d\psi}{dr'} = \mathbf{\hat r'}\cdot\nabla \psi$ and factorize the $r'^2$ into the measure on the right, to get :

$$\psi(t,x)= \frac{1}{4\pi}\int_{S'} \Big[\frac{1}{r'^2}\psi(\theta',\varphi') + \frac{1}{r'}\big[\mathbf{\hat r'}\cdot\nabla \psi(\theta',\varphi') + \frac{1}c \frac{d\psi}{dt}(\theta',\varphi')\big]\Big]r'^2\sin(\theta')\text d\theta' \text d\varphi'$$

  • $\begingroup$ Thank you for your detailed derivation, but I am still a little bit bugged by the "integration by parts" argument. So long as I know, the derivative of the delta function is defined such that $\delta'f=-f'\delta$ holds, so I feel this change of variable is not justified by the integration by parts. I think you are suggesting we think $\partial_{t'}\delta(t-t'-r'/c)=c\partial_{r'}\delta(t-t'-r'/c)$, because this is a $\delta$ function, so we do not need to differentiate between $t'$ and $r'/c$. Is my interpretation correct? $\endgroup$
    – Sofvar
    Commented May 4, 2022 at 1:57
  • $\begingroup$ Yes, this last equation is correct. $\endgroup$ Commented May 4, 2022 at 6:47

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.