An harmonic oscillating source will emit fields of the form

\begin{align} &\mathbf{H}=\frac{ck^2}{4\pi}(\mathbf{n}\times\mathbf{p})\frac{e^{ikr}}{r}\left( 1-\frac{1}{ikr} \right) \tag{1}\\ &\mathbf{E}=\frac{1}{4\pi\varepsilon_0}\left\{ k^2(\mathbf{n}\times\mathbf{p})\times\mathbf{n}\frac{e^{ikr}}{r}+[3\mathbf{n}(\mathbf{n}\cdot\mathbf{p})-\mathbf{p}]\left( \frac{1}{r^3}-\frac{ik}{r^2}\right)e^{ikr} \right\} \tag{2} \end{align}

where $\mathbf{n}=\frac{\mathbf r}{r}$ is the unit vector that points in the radiation direction and $\mathbf{p}$ is the electric dipole vector.

As Jackson says (Classical Electrodynamics, chapter 9) it's possible to find equation (2) from equation (1) using the formula

$$ \mathbf{E}=\frac{iZ_0}{k}\nabla\times\mathbf{H} \tag{3}$$

Applying equation (3) to equation (1) I obtain

$$ \mathbf{E}=\frac{ik}{4\pi\varepsilon_0}\nabla\times\left[\frac{e^{ikr}}{r}\left( 1-\frac{1}{ikr}e^{ikr} \right)(\mathbf{n}\times\mathbf{p})\right]$$


$$ [\nabla\times(f\mathbf v\times\mathbf w)]_k=\varepsilon_{ijk}\partial_i\varepsilon_{abj}fv_aw_b=\varepsilon_{ijk}\partial_if\varepsilon_{abj}v_aw_b=[\nabla f\times(\mathbf v\times\mathbf w)]_k$$

Identifying $f=\frac{e^{ikr}}{r}\left( 1-\frac{1}{ikr}e^{ikr} \right)$ I calculate

$$\nabla\left[\frac{e^{ikr}}{r}\left( 1-\frac{1}{ikr}e^{ikr} \right)\right]=\frac{e^{ikr}}{r}\left( ik-\frac{2}{r}+\frac{2}{ikr^2} \right)\mathbf n$$

Then i get

$$ \mathbf{E}=\frac{ik}{4\pi\varepsilon_0}\frac{e^{ikr}}{r}\left( ik-\frac{2}{r}+\frac{2}{ikr^2} \right)\mathbf{n}\times(\mathbf{n}\times\mathbf{p}) \tag{4}$$

That is definitely different from equation (2). So where am I wrong? How can I obtain the correct equation (2)?


1 Answer 1


Note that $\mathbf{n}$ is not a constant vector but dependent on position. You have not taken the derivative of this, which is the source of error.

In particular, you say $$\nabla \times (f\mathbf{v}\times\mathbf{w}) = \nabla f \times (\mathbf{v}\times\mathbf{w}).$$

This is incorrect, the correct expression would be $$\nabla \times (f\mathbf{v}\times\mathbf{w}) = \nabla f \times (\mathbf{v}\times\mathbf{w})+ f\,\nabla \times (\mathbf{v}\times\mathbf{w}).$$

  • $\begingroup$ Well, I can always choose $\mathbf{n}$ to be parallel to, for example, the x axis, then $\mathbf{n}=(1,0,0)$. $\endgroup$ Mar 13, 2017 at 11:10
  • $\begingroup$ Still, when you apply the gradient, you should consider how $\mathbf{n}$ changes. If you move from $(1,0,0)$ to $(1,dy,0)$, $d \mathbf{n}/dy = 1$ which cannot be neglected. $\endgroup$ Mar 13, 2017 at 11:17
  • $\begingroup$ Yes, of course. I realized now that it was a quite dumb question. Thank you, anyway! $\endgroup$ Mar 13, 2017 at 14:05

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.