Tell me more ×
Physics Stack Exchange is a question and answer site for active researchers, academics and students of physics. It's 100% free, no registration required.

In classical electrodynamics, what is the radiated power from a generalized source (consisting of charge density $\rho$ and current density $\vec{J}$) in vacuum?

Let us define $V_s$ to be the smallest sphere that completely encloses all source points -- that is, it is the smallest sphere outside which $\rho\left(\vec{r},t\right) = \vec{J}\left(\vec{r},t\right) = 0$ for all $t$. Let us say $\mathcal{R}_s$ is the radius of $V_s$.

Now let us define another sphere $V$ concentric with $V_s$. Let us say the radius of this sphere is $\mathcal{R}$, and $\mathcal{R} > \mathcal{R}_s$.

So what is power crossing the surface of $V$ at any instant? In other words, what is the value of the following? $$ P\left(\mathcal{R},t\right) = \oint_{\partial V} \vec{S}\left(\vec{r},t\right) \cdot \space d\vec{s}\left(\vec{r}\right) = \frac{1}{\mu_0} \oint_{\partial V} \left( \vec{E}\left(\vec{r},t\right) \times \vec{B}\left(\vec{r},t\right) \right) \cdot \space d\vec{s}\left(\vec{r}\right) $$

Below, I have provided an answer for the 'far field' approximation ($\mathcal{R} \gg \mathcal{R}_s$). Are there 'sanitized' expressions for cases where this approximation does not hold?

Even if there isn't a 'clean' expression for the general case, what I am looking for is the following. Let us say that there are two instants $t_1$ and $t_2$ such that for all $\vec{r} \in V_s$ we have $\rho\left(\vec{r}, t_1\right) = \rho\left(\vec{r}, t_2\right)$ and $\vec{J}\left(\vec{r}, t_1\right) = \vec{J}\left(\vec{r}, t_2\right)$. If we now define: $$ \mathcal{P}_{\infty}\left(t\right) = \lim_{\mathcal{R}\to\infty} P\left(\mathcal{R},t + \frac {\mathcal{R}} {c}\right) \\ \mathcal{P}_{\delta}\left(\mathcal{R},t\right) = P\left(\mathcal{R},t + \frac {\mathcal{R}} {c}\right) - \mathcal{P}_{\infty}\left(t\right) $$ ... can the following be proven? $$ \int_{t_1}^{t_2} \mathcal{P}_{\delta}\left(\mathcal{R},t\right)\space dt = 0 $$ Also, is it true (and can it be proven) that if $\hat{n}\left(\vec{r}\right)$ is the unit normal at any $\vec{r} \in V$, there will always be one axis along which $\vec{S}\left(\vec{r},t\right) \cdot \hat{n}\left(\vec{r}\right) = 0$?

share|improve this question

2 Answers

up vote 1 down vote accepted

For $\mathcal{R} \gg \mathcal{R}_s$, if we define: $$ \vec{Y}\left(t\right)= \iiint_{V_s} \vec{J}\left(\vec{r},t\right) \space dV\left(\vec{r}\right) ,\qquad \vec{Z}\left(t\right)= \iiint_{V_s} \frac{\partial \vec{J}\left(\vec{r},t\right)} {\partial t} \space dV\left(\vec{r}\right) = \frac {d \vec{Y}\left(t\right)} {dt} $$ and also $$ t_{\mathcal{R}} = t - \frac{\mathcal{R}} {c} $$ Assuming that the origin is at the center of $V$, so that $\hat{n}\left(\vec{r}\right) = \hat{r}$, the results are: $$ \vec{S}\left(\vec{r},t\right)\cdot\hat{n}\left(\vec{r}\right) = - \frac { \left( \vec{Y}\left(t_\mathcal{R}\right) \times \hat{n}\left(\vec{r}\right) \right) \cdot \left(\vec{Z}\left(t_\mathcal{R}\right) \times \hat{n}\left(\vec{r}\right) \right) } {16 \pi^2 \epsilon_0\mathcal{R}^3 c^2} - \frac { \left| \vec{Z}\left(t_\mathcal{R}\right) \times \hat{n}\left(\vec{r}\right) \right|^2 } {16 \pi^2 \epsilon_0\mathcal{R}^2 c^3} \\ P\left(\mathcal{R},t\right) = \frac {1} {6 \pi \epsilon_0 c^2} \left( \frac {\vec{Y}\left(t_{\mathcal{R}}\right)\cdot\vec{Z}\left(t_{\mathcal{R}}\right)} {\mathcal{R}} + \frac {\left|\vec{Z}\left(t_{\mathcal{R}}\right)\right|^2} {c} \right) \\ \mathcal{P}_{\infty}\left(t\right) = \frac {\left|\vec{Z}\left(t\right)\right|^2} {6 \pi \epsilon_0 c^3} ,\qquad \mathcal{P}_{\delta}\left(\mathcal{R},t\right) = \frac {\vec{Y}\left(t\right)\cdot\vec{Z}\left(t\right)} {6 \pi \epsilon_0 \mathcal{R} c^2} $$

Derivation

Given the following definitions: $$ \vec{R} = \vec{r} - \vec{r}_s, \qquad R = \left | \vec{R} \right |, \qquad t_r = t - \frac {R} {c} $$ From Jefimenko's Equations, the value of $\vec{E}$ and $\vec{B}$ for any $\vec{r}$ and any $t$ is as follows: $$ \vec{E}(\vec{r},t) = \frac {1} {4 \pi \epsilon_0} \iiint_{V_s} {\left( \frac {\rho (\vec{r}_s, t_r)} {R^3} \vec{R} + \frac {1} {R^2 c} \frac {\partial \rho (\vec{r}_s, t_r) } {\partial t} \vec{R} - \frac {1} {R c^2} \frac {\partial \vec{J} (\vec{r}_s, t_r) } {\partial t} \right)} \space dV\left(\vec{r}_s\right) $$ $$ \vec{B}(\vec{r},t) = \frac {\mu_0} {4 \pi} \iiint_{V_s} {\left( \frac {\vec{J} (\vec{r}_s, t_r)} {R^3} \times \vec{R} + \frac {1} {R^2 c} \frac {\partial \vec{J} (\vec{r}_s, t_r) } {\partial t} \times \vec{R} \right)} \space dV\left(\vec{r}_s\right) $$ Using this, we can write down the expression for $\vec{S}\left(\vec{r},t\right)$, which has six terms, each of which is a product of two volume integrals. We are, however, interested in $\vec{S}\left(\vec{r},t\right) \cdot \hat{n}\left(\vec{r}\right)$, where, if we designate the center of $V$ (and $V_s$) as $\vec{r}_0$: $$ \hat{n}\left(\vec{r}\right) = \frac {\vec{r}-\vec{r}_0} {\left|\vec{r}-\vec{r}_0\right|} $$ Now if we choose a frame of reference where $\vec{r}_0$ is the origin, we find that $\hat{n}\left(\vec{r}\right) = \hat{r}$. When $\mathcal{R} \gg \mathcal{R}_s$, we can consider $\dfrac{\left|\vec{r}_s\right|}{\left|\vec{r}_s\right|} \approx 0$. Thus, irrespective of $\vec{r}_s$ under this 'far field' approximation, $\vec{R}\approx\vec{r}-\vec{r}_s$, and is effectively independent of $\vec{r}_s$, so it can be taken outside of the volume integrals.

Then, because $\left(\vec{a}\times\left(\vec{b}\times\vec{a}\right)\right)\cdot\vec{a}=0$ four of the six terms in $\vec{S}\left(\vec{r},t\right)\cdot\hat{n}\left(\vec{r}\right)$ turn out to be zero.

Next, we substitute $\vec{Y}$ and $\vec{Z}$ into the expression for $\vec{S}\left(\vec{r},t\right)\cdot\hat{n}\left(\vec{r}\right)$ and get the expression listed above.

To help us with the surface integral, we define a spherical coordinate system $(\mathscr{r},\theta,\phi),\space 0\le \mathscr{r}\lt\infty,\space 0\le\theta\le\pi,\space 0\le\phi\le 2\pi$ with its center at $\vec{r}_0$, and orient it such that: $\vec{Z}\left(t\right)$ is $( |\vec{Z}\left(t\right)|, 0, 0)$ and $\vec{Y}\left(t\right)$ is $( |\vec{Y}\left(t\right)|, \gamma\left(t), 0\right)$. Note that the orientation of this coordinate system changes with time. In this coordinate system, therefore, $\hat{n}\left(\vec{r}\right)$ at any instant $t$ has to be represented in terms of the retarded time as $\left(1,\vartheta\left(\vec{r},t_r\right), \varphi\left(\vec{r},t_r\right)\right)$

Working out $\vec{S}\left(\vec{r},t\right)\cdot\hat{n}\left(\vec{r}\right)$ in this coordinate system, we get: $$ \vec{S}\left(\vec{r},t\right)\cdot\hat{n}\left(\vec{r}\right) = \frac {\left|\vec{Z}\left(t_{\mathcal{R}}\right)\right|^2} {16 \pi^2 \epsilon_0 {\mathcal{R}}^2 c^3} \sin^2\vartheta\left(\vec{r},t_{\mathcal{R}}\right) + \frac {\vec{Y}\left(t_{\mathcal{R}}\right)\cdot\vec{Z}\left(t_{\mathcal{R}}\right)} {16 \pi^2 \epsilon_0 {\mathcal{R}}^3 c^2} \sin^2\vartheta\left(\vec{r},t_{\mathcal{R}}\right) \\ - \frac {\left|\vec{Y}\left(t_{\mathcal{R}}\right)\times\vec{Z}\left(t_{\mathcal{R}}\right)\right|} {32 \pi^2 \epsilon_0 {\mathcal{R}}^3 c^2} \sin2\vartheta\left(\vec{r},t_{\mathcal{R}}\right) \cos\varphi\left(\vec{r},t_{\mathcal{R}}\right) $$

Finally, we compute the surface integral. For a sphere $V$ of radius $\mathcal{R}$ we get: $$ \oint_{\partial V} \sin^2\vartheta\left(\vec{r}\right) \space ds(\vec{r}) = \frac {8 \pi \mathcal{R}^2} {3},\qquad \oint_{\partial V} \sin 2\vartheta\left(\vec{r}\right) \space\cos\varphi\left(\vec{r}\right) \space ds(\vec{r})= 0 $$

Substituting, we get our result.

Observations

For the record,because of the way we've defined $V_s$ and because of conservation of charge, $Q_s$ is the (constant) total charge inside $V_s$, $$ \iiint_{V_s} \rho\left(\vec{r},t\right) \space dV\left(\vec{r}\right) = Q_s ,\qquad \iiint_{V_s} \frac{\partial \rho\left(\vec{r},t\right)} {\partial t} \space dV\left(\vec{r}\right) = \frac {d Q_s} {dt} = 0 $$ ... but we don't need these terms in the derivation.

Also, $\int_{t_1}^{t_2} \mathcal{P}_{\delta}\left(\mathcal{R},t\right) \space dt = 0$ because: $$ \int_{t_1}^{t_2} \vec{Y}(t)\cdot\vec{Z}(t) dt = \int_{t_1}^{t_2} \vec{Y}(t)\cdot\frac{d\vec{Y}}{dt}(t) dt = \frac12 \int_{t_1}^{t_2} \frac{d}{dt}|\vec{Y}(t)|^2 dt = \frac{1}{2} \left[|\vec{Y}(t)|^2\right]_{t_1}^{t_2} $$ and $$\forall\vec{r}\in V: \vec{J}\left(\vec{r},t_1\right) = \vec{J}\left(\vec{r},t_2\right) \implies \vec{Y}\left(t_1\right) = \vec{Y}\left(t_2\right)$$

The value of $\vec{S}\left(\vec{r},t\right)\cdot\hat{n}\left(\vec{r}\right)$ is zero whenever $\vec{Z}\left(t_{\mathcal{R}}\right) \times \hat{n}\left(\vec{r}\right) = 0$, that is, wherever the unit normal is parallel to $\vec{Z}\left(t_{\mathcal{R}}\right)$

share|improve this answer

My previous answer addresses the case where $\mathcal{R} \gg \mathcal{R}_s$, so $\dfrac{\left|\vec{r}_s\right|}{\left|\vec{r}\right|}$ for all $\vec{r}_s\in V_s$ and $\vec{r}\in \partial V$. What if it isn't so?

Unfortunately, things no longer stay simple. If we define the following new integrals: $$ \vec{W}_3\left(\vec{r},t\right)= \iiint_{V_s} \left( \frac {\vec{a}} {\left|\vec{r}-\vec{a}\right|^3} {\rho\left(\vec{a}, t - \dfrac{\left|\vec{r}-\vec{a}\right|} {c} \right)} \right) \space dV\left(\vec{a}\right) ,\\ \vec{X}_2\left(\vec{r},t\right)= \iiint_{V_s} \left( \frac{\vec{a}} {\left|\vec{r}-\vec{a}\right|^2} \frac{\partial} {\partial t} \rho\left(\vec{a}, t - \dfrac{\left|\vec{r}-\vec{a}\right|} {c} \right) \right) \space dV\left(\vec{a}\right) ,\\ \vec{Y}_{3}\left(\vec{r},t\right)= \iiint_{V_s} \left( \vec{J}\left(\vec{a}, t - \dfrac{\left|\vec{r}-\vec{a}\right|} {c} \right) \times \frac {\vec{r} - \vec{a}} {\left|\vec{r}-\vec{a}\right|^3} \right) \space dV\left(\vec{a}\right) ,\\ \vec{Z}_{2}\left(\vec{r},t\right)= \iiint_{V_s} \left( \frac{\partial} {\partial t} \vec{J}\left(\vec{a}, t - \dfrac{\left|\vec{r}-\vec{a}\right|} {c} \right) \times \frac {\vec{r} - \vec{a}} {\left|\vec{r}-\vec{a}\right|^2} \right) \space dV\left(\vec{a}\right) ,\\ \vec{Z}_1\left(\vec{r},t\right)= \iiint_{V_s} \left( \frac {1} {\left|\vec{r}-\vec{a}\right|} \frac{\partial} {\partial t} \vec{J}\left(\vec{a}, t - \dfrac{\left|\vec{r}-\vec{a}\right|} {c} \right) \right) \space dV\left(\vec{a}\right) $$ We can derive that: $$ \vec{S}\left(\vec{r},t\right)\cdot \hat{r} = \frac {1} {16 \pi^2 \epsilon_0 \mathcal{R} } \left( \vec{Y}_3 \left(\vec{r},t\right) + \frac {\vec{Z}_2 \left(\vec{r},t\right)} {c} \right) \cdot \left( \vec{W}_3 \left(\vec{r},t\right) \times \vec{r} + \frac {\vec{X}_2 \left(\vec{r},t\right) \times \vec{r}} {c} - \frac {\vec{Z}_1 \left(\vec{r},t\right) \times \vec{r}} {c^2} \right) $$

The problem, now, is that unlike $\vec{Y}$ and $\vec{Z}$, the functions $\vec{W}_3$, $\vec{X}_2$, $\vec{Y}_3$, $\vec{Z}_2$ & $\vec{Z}_1$ all contain $\vec{r}$ as a parameter, and so cannot be brought out of the surface integral over $\partial V$.

If we could take the approximation $\vec{r} - \vec{a} \approx \vec{r}$, we could say: $$ t_{\mathcal{R}} = t - \frac{\mathcal{R}} {c} ,\\ \vec{W}\left(t\right)= \iiint_{V_s} \rho\left(\vec{r},t\right) \space\vec{r} \space dV\left(\vec{r}\right) ,\qquad \vec{X}\left(t\right)= \iiint_{V_s} \frac{\partial \rho\left(\vec{r},t\right)} {\partial t} \vec{r} \space dV\left(\vec{r}\right) = \frac {d \vec{W}\left(t\right)} {dt} ,\\ \vec{Y}\left(t\right)= \iiint_{V_s} \vec{J}\left(\vec{r},t\right) \space dV\left(\vec{r}\right) ,\qquad \vec{Z}\left(t\right)= \iiint_{V_s} \frac{\partial \vec{J}\left(\vec{r},t\right)} {\partial t} \space dV\left(\vec{r}\right) = \frac {d \vec{Y}\left(t\right)} {dt} $$ ... and from there ... $$ \vec{W}_3\left(\vec{r},t\right)=\frac {\vec{W}\left(t_\mathcal{R}\right)} {\mathcal{R}^3} ,\qquad \vec{X}_2\left(\vec{r},t\right)=\frac {\vec{X}\left(t_\mathcal{R}\right)} {\mathcal{R}^2} ,\qquad \vec{Z}_1\left(\vec{r},t\right)=\frac {\vec{Z}\left(t_\mathcal{R}\right)} {\mathcal{R}} ,\\ \vec{Y}_3\left(\vec{r},t\right)=\frac {\vec{Y}\left(t_\mathcal{R}\right) \times \vec{r}} {\mathcal{R}^3} ,\qquad \vec{Z}_2\left(\vec{r},t\right)=\frac {\vec{Z}\left(t_\mathcal{R}\right) \times \vec{r}} {\mathcal{R}^2} $$ Finally leading us to: $$ \vec{S}\left(\vec{r},t\right)\cdot \hat{r} = \frac {1} {16 \pi^2 \epsilon_0 \mathcal{R}^2} \left( \frac {\vec{Y}\left(t_\mathcal{R}\right) \times \hat{r}} {\mathcal{R}} + \frac {\vec{Z}\left(t_\mathcal{R}\right) \times \hat{r}} {c} \right) \cdot \left( \frac {\vec{W}\left(t_\mathcal{R}\right) \times \hat{r}} {\mathcal{R}^2} + \frac {\vec{X}\left(t_\mathcal{R}\right) \times \hat{r}} {\mathcal{R} c} - \frac {\vec{Z}\left(t_\mathcal{R}\right) \times \hat{r}} {c^2} \right) $$ So while $P_{\delta}\left(\mathcal{R},t\right)$ is still very well defined, and physically I still expect that $\int_{t_1}^{t_2} P_{\delta}\left(\mathcal{R},t\right) ~dt = 0$ for the definition of $t_1$ and $t_2$ given in the question, I cannot take the mathematics forward.

As to the axis with a zero power flow, it also appears from the expression above that if we consider the medium distance and near fields, there won't be one in general.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.