# Divergent self-energy of point charges in Classical Electrodynamics

Assuming the electron to be a classical point particle, if one calculates the self-energy one finds $$U=\frac{e^2}{8\pi\epsilon_0r}$$ which diverges as $$r\rightarrow 0$$. Therefore, the measured mass of the electron should be $$m_{0e}+U/c^2=m_e.$$

Now, $$m_{0e}$$ is the mass of the electron in absence of its electric field (which is therefore unobservable because its electric field cannot be switched off) and $$m_e$$ is the measured electron mass.

Why is it a problem that in classical electrodynamics, the self-energy of a point electron diverge? The divergence in $$U$$ may be absorbed in $$m_{0e}$$ which is unobservable as often done in renormalization.

This is presented as a problem in textbooks because often one equates $$m_ec^2=U$$, and forget about $$m_{e0}$$. But I don't see a reason why $$m_{e0}$$ should be neglected. See page 28 here.

• Divergences are always a problem, whether we can reabsorb them into unobservable parameters or not. Just because we can hide the divergence into a counter-term doesn't mean that there is not a problem. In classical mechanics, and in quantum mechanics, a divergence always make the theory ill-defined. (note that "rigorous" treatments of QFT always study the regularised theory, say, in dimensional regulatisation, in which case there are no divergences at all, and the theory is "well" defined, at least in a perturbative sense). Dec 25, 2016 at 19:43
• @AccidentalFourierTransform Assuming a cut-off in length (by assuming new physics exists at short distance), one can take a similar regularization scheme here, too. Right?
– SRS
Dec 25, 2016 at 20:08
• @SRS Yes but that kind of regularization would not be Lorentz (ie Relativistic) invariant.
– user139175
Dec 27, 2016 at 9:03

Assuming the electron to be a classical point particle, if one calculates the self-energy one finds $$U=\frac{e^2}{8\pi\epsilon_0r}$$ which diverges as $$r\rightarrow 0$$.

This formula is the result of calculation of electrostatic potential energy of a system of charged particles distributed at different points of space. The usual distribution employed for this calculation is a uniform surface density on a sphere of radius $$r$$, or uniform volume density in a ball of radius $$r$$. It makes no sense to do this calculation for a single charged particle existing at a point. So your claim above is incorrect; for classical point particle, we have no way to calculate its self-energy and arrive at this formula. The calculation has to be done for some system of particles, whose size is described by $$r$$.

Therefore, the measured mass of the electron should be $$m_{0e}+U/c^2=m_e.$$

Yes, this is the so-called electromagnetic mass effect; EM energy associated with internal EM interactions in a charged system (electrostatic potential energy) manifests as modification of its effective inertial mass. This can be positive or negative, in case the system parts have charge of same sign, it is positive.

Why is it a problem that in classical electrodynamics, the self-energy of a point electron diverge? The divergence in $$U$$ may be absorbed in $$m_{0e}$$ which is unobservable as often done in renormalization.

"Self-energy of a point electron diverges" is a result of unwarranted application of either 1) the Poynting energy formula to calculate EM interaction energy inside a point particle, or 2) trying to calculate the limit $$\lim_{r\to 0} U$$ and assuming this limit (which is infinity) is a valid result for the point particle. Both of these methods give infinite energy. But this merely says that either

1. the Poynting energy formula value for field of point charge is infinite, or
2. compressing spatially distributed charge to a point takes infinite energy.

But this does not imply all point charges have to be associated with infinite real energy (in the sense of stored previously done work, or extractable energy available for doing work later). Maybe point charge energy is not given by the Poynting formula, and maybe point charge is not a result of compression of spatially distributed charge into a point.

Is "Self-energy of a point electron diverges" a problem? It depends on other assumptions. If you regard Poynting energy formula as invalid for point particles, then this just the mathematical property of the Poynting energy of point particle, devoid of any physical relevance, so this is not a problem; electron energy is simply not given by the Poynting formula. This is closely related to the view that point particles do not act on themselves = no self-interaction in point particles. This viewpoint has been analyzed and developed into a consistent theory of charged point particles by many people in the past, such as Tetrode, Fokker, Frenkel, and the Feynman-Wheeler collaboration.

If you believe this infinite Poynting energy is real energy that had to be used in the past to assemble the point particle, or energy that can be released somehow, then the fact this energy is infinite is problematic, because it predicts potential infinitely strong bomb in every point particle. Also, there is no consistent theory of point particles with infinite EM self-energy; such theory has to involve non-EM energy, hence non-EM forces and also has to explain radiation energy as loss of self-energy through self-interaction. Such self-interaction of point particle cannot be consistently described in the framework of Maxwell equations and Lorentz force (so that local energy conservation and equations of motions are satisfied). One would have to modify the theory so that things would work much differently and EM self-interaction was somehow finite and consistent with radiation energy, while EM self-energy is infinite. This was tried many times but compelling solution was never found.

This is presented as a problem in textbooks because often one equates $$m_ec^2=U$$, and forget about $$m_{e0}$$. But I don't see a reason why $$m_{e0}$$ should be neglected. See page 28 here.

You are correct, the fact that non-EM forces have to be present to hold the charge together (and cancel the positive infinity of energy by negative infinite of energy) is sometimes omitted.

EM textbooks aren't usually very good on this topic. Probably because this problem with infinite EM energy and self-interaction of point charged particles has been sort-of solved long ago, but the solution (no self-interaction and Poynting formulae are not applicable to point particles) is so different from what people expected (consistent theory of self-interaction of point particles). Also no big revelation came from it. So the solution is not well known and accepted part of physics. This subject is still considered a murky mine-field area.

This is indeed not a fundamental problem, because classical electrodynamics is only an approximate description of electrodynamics valid in some appropriate scaling limit. If one then attempts to formulate such an approximate theory completely divorced from the real fundamental laws of Nature, then one typically encounters problems when naively taking limits to scales where the theory should not be valid.

A non-trivial problem in classical electrodynamics where renormalization is needed is the rigorous treatment of electromagnetic radiation. While one can calculate the emitted radiation of accelerated charges without problem, the backreaction of the emitted radiation on the charges is typically ignored. E.g. the OP's book contains a calculation of the emitted radiation by pulsars, it mentions that these pulsars slow down due to conservation of energy, but it's not pointed out that this implies that the Lorentz force equation as mentioned in the book must therefore be wrong as it doesn't contain any terms that could exert a torque on a magnet that is freely rotating in empty space.

What needs to be included, therefore, is the force exerted on a charge due to its own electromagnetic field, but this is divergent for point charges. A rigorous treatment of the self-force was only given recently in this article.

In general, analogues of the renormalization group methods as used in QFT will have to be invoked in all physics disciplines where the degrees of freedom reside in arbitrarily small length scales (e.g. fluid dynamics), as soon as one is rigorous enough. The fact that we typically don't see such methods applied in textbooks is because they typically don't treat the subject in a rigorous way.

• Angular momentum density is given by $\vec{r}\times \vec{N}/c$, where $\vec{N}$ is the Poynting vector. Torque would be the rate of change of that, which is net non-zero when magnetic and rotation axes are misaligned. Dec 24, 2019 at 11:52