I would like to derive the common (and important - to me atleast) equation:

$\epsilon = \epsilon_a + j\frac{\sigma}{\omega\epsilon}$

I have seen some resources calling this a definition (in which case, it may not have a derivation), but it looks like it definitely should and it should fall out of Maxwell's equations, but I cannot do it!

So, starting with,

$\nabla \times \vec{H} = j\omega\vec{D}+\vec{J}$

and using Ohm's law and a constitutive relation, which for convenience is:

$\vec{J} = \sigma \vec{E}$ and $\vec{D} = \epsilon\vec{E}$

One can arrive at:

$\nabla \times \vec{H} = j\omega\epsilon\vec{E}+\sigma\vec{E}$ (Eq. 1)

Keeping the above in mind for a moment and working on this:

$\nabla \times \vec{E} = -j\omega\mu\vec{H}$

one can take the curl of both sides:

$\nabla \times (\nabla \times \vec{E}) = -j\omega\mu\nabla \times \vec{H}$

Using the vector identity $\nabla \times (\nabla \times \vec{A}) = (\nabla(\nabla\cdot\vec{A})-\nabla^2 \vec{A})$, recalling Eq. 1 and noting that for a region without a source $\nabla\cdot\vec{E} = 0$, the above becomes:



This wave equation for a lossy material is as far as I can get...

  • $\begingroup$ magnetic permeability is often taken $\mu=1$ in these situations of non-magnetic materials $\endgroup$
    – Simon
    Commented Jan 17, 2020 at 15:30
  • $\begingroup$ @Simon thank you. That solves one problem... but the rest is still puzzling me. $\endgroup$
    – smollma
    Commented Jan 17, 2020 at 17:06

2 Answers 2


It is actually just the rhs of one of the Maxwell's equations $$ \nabla\times\mathbf{H}=\mathbf{J}+\frac{\partial \mathbf{D}}{\partial t}. $$ If we perform Fourier transform in time (or simply take $\mathbf{H},\mathbf{J},\mathbf{D},\mathbf{E}\propto e^{j\omega t}$) then $$ \nabla\times\mathbf{H}(\omega)=\mathbf{J}(\omega)-j\omega \mathbf{D}(\omega) = \sigma \mathbf{E}(\omega)-j\omega \epsilon\mathbf{E}(\omega)=\\-j\omega\left(\frac{\sigma}{-j\omega}+\epsilon\right)\mathbf{E}(\omega)=-j\omega\epsilon_{eff}\mathbf{E}(\omega)=-j\omega\mathbf{D}_{eff}(\omega), $$ where the effective permittivity is introduced in analogy with $\mathbf{D}=\epsilon\mathbf{E}$.

It could be also be generalized to frequency dependent conductance and permittivity, taking the correct real-time relation between the quantities: $$ \mathbf{J}(\omega)=\sigma(\omega)\mathbf{E}(\omega)\Leftrightarrow \mathbf{J}(t)=\int_{-\infty}^tdt'\sigma(t-t')\mathbf{E}(t'),\\ \mathbf{D}(\omega)=\epsilon(\omega)\mathbf{E}(\omega)\Leftrightarrow \mathbf{D}(t)=\int_{-\infty}^tdt'\epsilon(t-t')\mathbf{E}(t') $$


Let's define a new wave equation that electric field will be subject to. In phasor notation we have $\frac{\partial}{\partial t}\to j\omega$, which gives us $$\nabla^2\vec{E}+\omega^2\frac{1}{v^2}\vec{E}=0$$

We can compare this to your last equation which read $$\nabla^2\vec{E}+\omega^2\epsilon\mu\vec{E}(1-j\frac{\sigma}{\omega\epsilon})=0$$ $$\nabla^2\vec{E}+\omega^2\epsilon\mu\vec{E}(1-j\frac{\sigma}{\omega\epsilon})=0=\nabla^2\vec{E}+\omega^2\frac{1}{v^2}\vec{E}$$ $$\implies \omega^2\epsilon\mu\vec{E}(1-j\frac{\sigma}{\omega\epsilon})-\omega^2\frac{1}{v^2}\vec{E}=0$$ $$\implies \omega^2 (\epsilon\mu-j\frac{\sigma\mu}{\omega}-\frac{1}{v^2}) \vec E=0 \implies \epsilon\mu-j\frac{\sigma\mu}{\omega}-\frac{1}{v^2}=0$$

Normally, we would have $v^2=\frac{1}{\epsilon\mu}$. We can still define effective dielectric and magnetic constants to reflect the characteristics of this system. As it is usual, we set $\mu_{eff}=\mu$ since magnetic effects are usually ignored in these settings, $$\epsilon\mu-j\frac{\sigma\mu}{\omega}=\epsilon_{eff}\mu_{eff}=\epsilon_{eff}\mu$$ $$\implies \boxed{\epsilon_{eff} = \epsilon-j\frac{\sigma}{\omega}}$$

First, this doesn't match exactly your formula in the beginning, however the Wikipedia page actually has this result.

I actually don't know if in free space $\sigma\ne0$ results in any other behavior than the "lossy" EM wave propagation. But if that's all, this assumption accurately models the system (since the wave equations are of correct shape).

  • $\begingroup$ Thanks for your reply. This isn't really what I was looking for. This isn't a derivation. Would it be possible to add some more steps? Also, it looks like you've squared $v$ to get $\epsilon$ and $\mu_a$ in the numerator. You then cancel $\mu_a$ with $\mu_a$ in the denominator. If squaring $v$ makes $\epsilon$, not $\epsilon_a$, why does it make $\mu_a$ and not $\mu$?! The final form of your derivation is not the same as the equation I wanted to arrive at. $\endgroup$
    – smollma
    Commented Jan 17, 2020 at 17:22
  • $\begingroup$ Yeah there are a few mistakes I am trying to resolve. Though the "derivation" depends on defining a effective refractive index to the material so that it still follows Maxwell's equations in a reasonable way. I'll try to improve this. $\endgroup$
    – acarturk
    Commented Jan 17, 2020 at 17:25
  • $\begingroup$ Thank you! I appreciate that. $\endgroup$
    – smollma
    Commented Jan 17, 2020 at 17:26
  • $\begingroup$ There’s an issue in the answer relating the second derivative in time to $j\omega$. It should be the first derivative in time corresponding to $j\omega$. $\endgroup$
    – Newbie
    Commented Jan 30, 2022 at 4:25
  • 1
    $\begingroup$ @Newbie That's a (thankfully) simple typo and doesn't affect the rest of the answer. I'll fix it now. $\endgroup$
    – acarturk
    Commented Jan 30, 2022 at 18:35

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.