MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

This question is related to my previous question about the regularity of the Maxwell equations.

Assume we are working on a space where there are only electric point charges, $(q_i)$, and a blob of matter $M$ which is a bounded, closed and connected set in $\mathbf R^3$. With respect to the electric field, the only difference is the permittivity $\varepsilon$. That is, $\varepsilon$ is a function which is dependent on the space variables $(x, y, z)$. In this particular $M$ is homogeneous so $\varepsilon$ is constant almost everywhere except on the boundary of $M$, where it has a jump.

If we are given the potential $V$ of this system, then the electric field $\mathbf E = -\nabla V$ will likewise make a jump. My question is now related to expressions of the form. $$\int_{\mathbf R^3} |\mathbf E|^2 \ \nabla \varepsilon \ \mathrm{d}V.\tag{*}$$

Hence, to my understanding we have returned to the regularity of the solutions of the Maxwell equations. $\nabla \varepsilon$ will contain a $\delta$-distribution and in its support $\mathbf E$, $(*)$ is not so well-defined as it would seem. The result would strongly depend on the chosen approximating sequence to $\delta$. This was the problem and the Maxwell equations are one particular example where I have seen them occur.

But, of course, if we instead of having $\varepsilon$ have a jump, we could have a smooth -but fast- transition. I could have a sequence $(\varepsilon_n)_n$ of those functions that converge distributionally to the Heaviside distribution. However, $\varepsilon$ determines the solution $\mathbf E$ so, $\mathbf E$ will be different too.

My questions:

  1. Can expressions of the form $(*)$ where we are actually evaluation a function on a null-set which is not well-defined on that point? In PDE we would use traces for this. In particular, we map the boundary to a space where it has positive measure and work there. Is there a general theme for such expression as the one here?
  2. The expression above can have an approximation as $\nabla \varepsilon_n$ somehow. We know that $\mathbf E$ is the solution for the limit $\lim_n \nabla \varepsilon_n$, can we somehow reduce the question to computing (for instance) $$\lim_{n \to \infty} \int_{\mathbf R^3} |\mathbf E|^2 \ \nabla \varepsilon_n \ \mathrm{d}V,$$ or will this never be well-defined? The problem here of course is that $\varepsilon_n$ is not the "correct" one that matches $\mathbf E$.
  3. How is the question related to regularity? Naively I would expect that even though the distribution $(*)$ is not well-defined as an integral, that $\lim_n |\mathbf E_n|^2 \nabla \varepsilon_n = |\mathbf E|^2 \nabla \varepsilon$ distributionally.
  4. In hyperbolic equations, if we change some of the parameters locally, does the global solution "feel" this? If a wave is propagating and I change some of the material it will go through way before the wave arrived there, the current solution should not change. Is this a general property? (Err... I realise this is basically the definition, but the question is thus is if this makes the above property easier to prove. I would suspect so.).

And, as before, I know little to nothing about hyperbolic equations. As per suggestion the last time, I do have the book "Multi-dimensional hyperbolic partial differential equations" by Sylvie Benzoni-Gavage and Denis Serre (but so much material and takes some time to digest). However, I do know a bit about the elliptic theory and now also the Strichartz estimates.

share|cite|improve this question
up vote 3 down vote accepted

Yes, there is a way of interpreting this integral. Decompose E into a component $E_t$ tangential to level sets of $\epsilon$ and a component $E_n$ normal to them. Moreover, let us set $D=\epsilon E$, and decompose it in the same way. You can write your integrand as $E_t^2\nabla \epsilon-D_n^2\nabla(1/\epsilon)$. Now if you take the limit of a discontinuous material boundary, then both $E_t$ and $D_n$ are continuous, so there is no problem definining the integral in this form.

share|cite|improve this answer
That's a neat trick! I still do wonder if there is a 'bigger hammer' if you do not find this trick. – Jonas Teuwen Nov 29 '12 at 7:55
It is not a 'trick': the quantity $D$ is a physical quantity, namely it is the electric displacement field. – username Oct 7 '13 at 6:52

Your Answer


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.