I am taking an undergraduate course in E&M following Griffiths.

I was wondering if there is a good way to embed the information of a dipole into the charge distribution (and if it would be of any use) as the limit of some function similar to a Dirac delta function, or modeling it in some other manner. I know that you can write a pure dipole as a limit where the distance between two point charges tends to $0$ but the dipole moment of the system does not.

My Thoughts and Ideas so far (Limited to 1D):

Using two delta functions and a limit for the distance. This solution is awkward and feels like it involves too many working parts to be a satisfactory solution - I would prefer just a single limit as opposed to three of them.

So, I was experimenting with functions like $nxe^{-nx^2}$ which have a total charge of $0$. However, this does not work since the dipole moment still tends to $0$ when $n$ tends to $\infty$.

My Question:   Is there anywhere in physics, where a charge distribution with a dipole would be used? If so, is there an agreed upon definition for such a function? (Like the Dirac delta with the Integral of Product rule and unit area property?)

If not, then is there any other way of defining/modeling a pure dipole more concretely (directly?) rather than through the multipole expansion.


1 Answer 1


The (distributional) derivative of the delta function does what you want. By definition, $$ \int \mathrm dx \ \delta'(x)f(x) = -f'(0)$$ which is motivated by the standard integration by parts formula. As a result, a charge distribution $\rho(\mathbf r) = \alpha \delta (x) \delta(y) \delta'(z)$ produces a net dipole moment $$\mathbf p = \int \mathrm dx \mathrm dy \mathrm dz \ \mathbf r \ \rho(\mathbf r) = -\alpha \hat z$$ and no other multipole moments. More generally, a dipole $\mathbf p = p \hat n$ (where $\hat n$ is a unit vector) is generated by the charge distribution $\rho(\mathbf r ) = - p \big(\hat n \cdot \nabla\big)\delta^3(\mathbf r) = - \mathbf p \cdot \nabla\big[\delta(x)\delta(y)\delta(z)\big] $


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.