Radiation from accelerating charges — Larmor formula

Intuition [Beginner]

A charged particle at rest produces a static electric field. A charged particle moving at constant velocity produces electric and magnetic fields that move with it, but carries no energy outward to infinity. Only when the charge accelerates does it launch electromagnetic waves that detach from the source and carry energy away. This is the physical content of the Larmor formula.

The total power radiated by a non-relativistic point charge is

$$P=\frac{q^2{a}^2}{6\pi {\epsilon }_0{c}^3},$$

where $q$ is the charge, $a$ is the magnitude of its acceleration, ${\epsilon }_0$ is the permittivity of free space, and $c$ is the speed of light. Double the acceleration, quadruple the radiated power: $P$ scales as $a^2$.

The radiation is not emitted equally in all directions. It is strongest perpendicular to the acceleration and zero along the acceleration direction. The pattern is a doughnut centred on the acceleration vector.

Visual [Beginner]

Picture a charge at the origin, accelerating to the right. Before the acceleration began, the field lines were radial. After the acceleration starts, a kink in each field line propagates outward at speed $c$. This kink is the radiation field: a transverse disturbance in the field lines that carries energy away.

The angular pattern looks like a torus. If the acceleration points along the $z$-axis, the radiated power per unit solid angle is proportional to ${\sin }^2\theta$, where $\theta$ is the polar angle. Maximum radiation at the equator ($\theta =90^{\circ }$), zero at the poles ($\theta =0^{\circ }$, $180^{\circ }$).

This ${\sin }^2\theta$ pattern is why synchrotrons — where charges move in circles — are such intense radiation sources. Circular motion means continuous centripetal acceleration, and the radiation is emitted tangentially.

Worked example [Beginner]

An electron in an X-ray tube has charge $q=1.6\times 10^{-19}$ C and is accelerated at $a=10^{18}$ m/s${}^2$. Find the radiated power.

Step 1. Write out the Larmor formula:

$$P=\frac{q^2{a}^2}{6\pi {\epsilon }_0{c}^3}.$$

Step 2. Substitute the numbers. The numerator is $q^2{a}^2=(1.6\times 10^{-19}{)}^2\times (10^{18}{)}^2=2.56\times 10^{-38}\times 10^{36}=2.56\times 10^{-2}$. The denominator is $6\pi \times 8.85\times 10^{-12}\times (3\times 10^8{)}^3=6\pi \times 8.85\times 10^{-12}\times 2.7\times 10^{25}\approx 4.5\times 10^{15}$.

Step 3. Divide: $P\approx 2.56\times 10^{-2}/4.5\times 10^{15}\approx 5.7\times 10^{-18}$ W per electron. Small per particle, but a beam of $10^{15}$ electrons radiates kilowatts — this is the synchrotron radiation that limits circular accelerator performance.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Retarded potentials

The inhomogeneous wave equation for the potentials in Lorenz gauge is

$$\square \varphi =-\frac{\rho }{{\epsilon }_0},\square \mathbf{A}=-{\mu }_0\mathbf{J},$$

where $\square ={\nabla }^2-c^{-2}{\partial }^2/\partial t^2$ is the d'Alembertian. The retarded Green's function for the wave equation in three dimensions is

$$G_{\mathrm{ret}}(\mathbf{r},t)=\frac{\delta (t-r/c)}{4\pi r},$$

which vanishes for $t<0$ (causality). The retarded potentials are

$$\varphi (\mathbf{r},t)=\frac{1}{4\pi {\epsilon }_0}\int \frac{\rho ({\mathbf{r}}^{\prime },t_{\mathrm{ret}})}{|\mathbf{r}-{\mathbf{r}}^{\prime }|}{d}^3{\mathbf{r}}^{\prime },$$ $$\mathbf{A}(\mathbf{r},t)=\frac{{\mu }_0}{4\pi }\int \frac{\mathbf{J}({\mathbf{r}}^{\prime },t_{\mathrm{ret}})}{|\mathbf{r}-{\mathbf{r}}^{\prime }|}{d}^3{\mathbf{r}}^{\prime },$$

where $t_{\mathrm{ret}}=t-|\mathbf{r}-{\mathbf{r}}^{\prime }|/c$ is the retarded time. Information from the source at ${\mathbf{r}}^{\prime }$ arrives at the field point $\mathbf{r}$ after a light-travel-time delay $|\mathbf{r}-{\mathbf{r}}^{\prime }|/c$.

Lienard-Wiechert potentials

For a point charge $q$ on a trajectory ${\mathbf{r}}_0(t)$ with velocity $\mathbf{v}(t)={\dot{\mathbf{r}}}_0(t)$, the retarded-potential integrals collapse to the Lienard-Wiechert potentials:

$$\varphi (\mathbf{r},t)=\frac{q}{4\pi {\epsilon }_0}\frac{1}{(1-\hat{\mathbf{n}}\cdot \mathbf{\beta })R}{|}_{t_{\mathrm{ret}}},$$ $$\mathbf{A}(\mathbf{r},t)=\frac{q}{4\pi {\epsilon }_{0}c}\frac{\mathbf{\beta }}{(1-\hat{\mathbf{n}}\cdot \mathbf{\beta })R}{|}_{t_{\mathrm{ret}}},$$

where $\mathbf{R}=\mathbf{r}-{\mathbf{r}}_0(t_{\mathrm{ret}})$, $R=|\mathbf{R}|$, $\hat{\mathbf{n}}=\mathbf{R}/R$, and $\mathbf{\beta }=\mathbf{v}/c$, all evaluated at the retarded time $t_{\mathrm{ret}}$ determined implicitly by $t_{\mathrm{ret}}=t-R(t_{\mathrm{ret}})/c$. The factor $(1-\hat{\mathbf{n}}\cdot \mathbf{\beta }{)}^{-1}$ accounts for the Doppler-like compression of the emitted field when the source moves toward the observer.

Radiation fields

Computing $\mathbf{E}=-\nabla \varphi -\partial \mathbf{A}/\partial t$ from the Lienard-Wiechert potentials yields two pieces. The velocity field (Coulomb-like) falls as $1/R^2$ and carries no energy to infinity. The acceleration (radiation) field falls as $1/R$:

$${\mathbf{E}}_{\mathrm{rad}}(\mathbf{r},t)=\frac{q}{4\pi {\epsilon }_0{c}^2}\frac{\hat{\mathbf{n}}\times [(\hat{\mathbf{n}}-\mathbf{\beta })\times \dot{\mathbf{\beta }}]}{(1-\hat{\mathbf{n}}\cdot \mathbf{\beta }{)}^{3}R}{|}_{t_{\mathrm{ret}}}.$$

In the non-relativistic limit $\beta \to 0$ this reduces to

$${\mathbf{E}}_{\mathrm{rad}}=\frac{q}{4\pi {\epsilon }_0{c}^{2}R}\hat{\mathbf{n}}\times (\hat{\mathbf{n}}\times \mathbf{a})=\frac{q}{4\pi {\epsilon }_0{c}^{2}R}[\hat{\mathbf{n}}(\hat{\mathbf{n}}\cdot \mathbf{a})-\mathbf{a}].$$

The radiation field is transverse (${\mathbf{E}}_{\mathrm{rad}}\cdot \hat{\mathbf{n}}=0$), falls as $1/R$, and is proportional to the acceleration. The magnetic field is ${\mathbf{B}}_{\mathrm{rad}}=\hat{\mathbf{n}}\times {\mathbf{E}}_{\mathrm{rad}}/c$.

The Larmor formula

The energy flux carried by the radiation field is given by the Poynting vector $\mathbf{S}={\mathbf{E}}_{\mathrm{rad}}\times {\mathbf{B}}_{\mathrm{rad}}/{\mu }_0=c{\epsilon }_0|{\mathbf{E}}_{\mathrm{rad}}{|}^2\hat{\mathbf{n}}$. In the non-relativistic limit, the power radiated per unit solid angle is

$$\frac{dP}{d\Omega }=R^2\mathbf{S}\cdot \hat{\mathbf{n}}=\frac{q^2{a}^2}{16{\pi }^2{\epsilon }_0{c}^3}{\sin }^2\theta ,$$

where $\theta$ is the angle between $\hat{\mathbf{n}}$ and $\mathbf{a}$. Integrating over solid angle using $\int {\sin }^2\theta d\Omega =8\pi /3$:

$$\boxed{P=\frac{q^2{a}^2}{6\pi {\epsilon }_0{c}^3}}$$

This is the Larmor formula. In Gaussian units the prefactor simplifies: $P=2q^2{a}^2/(3c^3)$.

The covariant (relativistic) generalisation is the Lienard formula:

$$P=\frac{q^2}{6\pi {\epsilon }_{0}c}{\gamma }^6[a^2-|\mathbf{\beta }\times \mathbf{a}{|}^2],$$

where $\gamma =(1-{\beta }^2{)}^{-1/2}$. For acceleration parallel to velocity, $P\propto {\gamma }^6{a}^2$; for perpendicular acceleration (circular motion), $P\propto {\gamma }^4{a}^2$.

Counterexamples to common slips

• The Larmor formula gives the power radiated in the instantaneous rest frame when the particle is non-relativistic. For relativistic particles the total radiated power is not obtained by replacing $a$ with the relativistic acceleration; one must use the Lienard formula or work covariantly.
• The radiation field at angle $\theta =0$ or $\pi$ (along the acceleration) is zero, not just small: ${\sin }^2\theta =0$ exactly. The pattern is genuinely toroidal.
• The $1/R$ fall-off of the radiation field is what makes energy transport to infinity possible. The total power through a sphere of radius $R$ is $R^2$ times the Poynting flux, and since the flux is $\propto 1/R^2$ the result is independent of $R$.
• The Larmor formula does not apply to a uniformly moving charge ($a=0$). The field of a uniformly moving charge is a compressed Coulomb field, not a radiation field.

Key theorem with proof [Intermediate+]

Theorem (Larmor formula from the Poynting vector). A non-relativistic point charge $q$ with instantaneous acceleration $\mathbf{a}$ radiates total power

$$P=\frac{q^2{a}^2}{6\pi {\epsilon }_0{c}^3}.$$

Proof. In the non-relativistic limit the radiation electric field at position $\mathbf{r}$ (with $r=|\mathbf{r}|$ large) from a charge at the origin with acceleration $\mathbf{a}$ is

$${\mathbf{E}}_{\mathrm{rad}}=\frac{q}{4\pi {\epsilon }_0{c}^{2}r}\hat{\mathbf{n}}\times (\hat{\mathbf{n}}\times \mathbf{a}),$$

where $\hat{\mathbf{n}}=\mathbf{r}/r$. The magnitude is $|{\mathbf{E}}_{\mathrm{rad}}|=qa\sin \theta /(4\pi {\epsilon }_0{c}^{2}r)$ where $\theta$ is the angle between $\hat{\mathbf{n}}$ and $\mathbf{a}$.

The Poynting vector is $\mathbf{S}=c{\epsilon }_0|{\mathbf{E}}_{\mathrm{rad}}{|}^2\hat{\mathbf{n}}$. The power through a solid angle $d\Omega$ at radius $r$ is

$$\frac{dP}{d\Omega }=r^2\mathbf{S}\cdot \hat{\mathbf{n}}=\frac{q^2{a}^2{\sin }^2\theta }{16{\pi }^2{\epsilon }_0{c}^3}.$$

Integrate over the full sphere:

$$P=\frac{q^2{a}^2}{16{\pi }^2{\epsilon }_0{c}^3}\int {\sin }^2\theta d\Omega .$$

With $d\Omega =\sin \theta d\theta d\phi$ and $\phi \in [0,2\pi )$, $\theta \in [0,\pi ]$:

$${\int }_0^{2\pi }d\phi {\int }_0^{\pi }{\sin }^3\theta d\theta =2\pi \cdot \frac{4}{3}=\frac{8\pi }{3}.$$

Therefore

$$P=\frac{q^2{a}^2}{16{\pi }^2{\epsilon }_0{c}^3}\cdot \frac{8\pi }{3}=\frac{q^2{a}^2}{6\pi {\epsilon }_0{c}^3}.■$$

Corollary (Lienard formula). The relativistic generalisation for a charge with velocity $\mathbf{v}=c\mathbf{\beta }$ and acceleration $\mathbf{a}$ is

$$P=\frac{q^2}{6\pi {\epsilon }_{0}c}{\gamma }^6[a^2-|\mathbf{\beta }\times \mathbf{a}{|}^2]=\frac{q^{2}c}{6\pi {\epsilon }_0}{\gamma }^6[(\mathbf{\beta }\cdot \dot{\mathbf{\beta }}{)}^2-{\dot{\mathbf{\beta }}}^2],$$

obtained by evaluating the covariant expression $-(q^2/6\pi {\epsilon }_0{c}^3){\alpha }_{\mu }{\alpha }^{\mu }$ where ${\alpha }^{\mu }=d{U}^{\mu }/d\tau$ is the four-acceleration and $U^{\mu }$ the four-velocity, and projecting onto the lab frame. For linear acceleration ($\mathbf{a}\parallel \mathbf{v}$) this gives $P=q^2{\gamma }^6{a}^2/(6\pi {\epsilon }_0{c}^3)$; for circular motion ($\mathbf{a}\perp \mathbf{v}$) it gives $P=q^2{\gamma }^4{a}^2/(6\pi {\epsilon }_0{c}^3)$.

Worked example: synchrotron radiation power

An electron ($q=e$, $m=9.11\times 10^{-31}$ kg) moves in a circle of radius $R=10$ m at $\gamma =1000$ (energy $\approx 511$ MeV). The acceleration is $a=v^2/R\approx c^2/R=9\times 10^{15}$ m/s${}^2$. Using the Lienard formula for perpendicular acceleration:

$$P=\frac{e^2{\gamma }^4{a}^2}{6\pi {\epsilon }_0{c}^3}=\frac{(1.6\times 10^{-19}{)}^2\times 10^{12}\times (9\times 10^{15}{)}^2}{6\pi \times 8.85\times 10^{-12}\times (3\times 10^8{)}^3}.$$

The numerator is $2.56\times 10^{-38}\times 10^{12}\times 8.1\times 10^{31}=2.07\times 10^6$. The denominator is $\approx 4.5\times 10^{15}$. So $P\approx 4.6\times 10^{-10}$ W per electron. With $10^{12}$ stored electrons the total is $\sim 460$ kW — this is why modern synchrotron light sources need powerful RF systems to replace the radiated energy.

Exercises [Intermediate+]

The full Lienard formula and relativistic beaming [Master]

The non-relativistic Larmor formula suffices for $\beta \ll 1$. At relativistic energies the covariant expression is essential. The key results are:

Lienard formula (total power):

$$P=\frac{q^{2}c}{6\pi {\epsilon }_0}{\gamma }^6[{\dot{\mathbf{\beta }}}^2-(\mathbf{\beta }\times \dot{\mathbf{\beta }}{)}^2].$$

For linear acceleration ($\dot{\mathbf{\beta }}\parallel \mathbf{\beta }$): $P=q^2{\gamma }^6{\dot{\beta }}^{2}c/(6\pi {\epsilon }_0)$. For circular motion ($\dot{\mathbf{\beta }}\perp \mathbf{\beta }$): $P=q^2{\gamma }^4{\dot{\beta }}^{2}c/(6\pi {\epsilon }_0)$. The extra factor of ${\gamma }^2$ for linear acceleration makes the difference: at the same acceleration, a linearly accelerated charge radiates ${\gamma }^2$ times more than a circularly accelerated one. But in practice, circular accelerators radiate far more because maintaining a given energy in a circle requires continuous centripetal acceleration.

Angular distribution (relativistic):

$$\frac{dP}{d\Omega }=\frac{q^2}{16{\pi }^3{\epsilon }_{0}c}\frac{|\hat{\mathbf{n}}\times [(\hat{\mathbf{n}}-\mathbf{\beta })\times \dot{\mathbf{\beta }}]{|}^2}{(1-\hat{\mathbf{n}}\cdot \mathbf{\beta }{)}^5}.$$

The $(1-\hat{\mathbf{n}}\cdot \mathbf{\beta }{)}^{-5}$ factor produces extreme forward beaming for $\gamma \gg 1$: the radiation is concentrated in a cone of half-angle $\sim 1/\gamma$ around the instantaneous velocity. This is why synchrotron light sources produce intense, highly collimated beams.

Synchrotron radiation spectrum. The critical frequency for synchrotron radiation is ${\omega }_c=(3/2){\gamma }^{3}c/R$, where $R$ is the bending radius. The spectrum extends from low frequencies up to $\sim {\omega }_c$ and falls exponentially above. The total power is

$$P_{\mathrm{sync}}=\frac{q^{2}c{\gamma }^4}{6\pi {\epsilon }_0{R}^2}.$$

For the CERN Large Hadron Collider ($\gamma \approx 7500$ for 7 TeV protons, $R\approx 2800$ m), the synchrotron power is $\sim 3.6$ kW per proton beam — manageable for protons but ruinous for electrons at comparable energies, which is why electron-positron colliders above $\sim 100$ GeV must be linear.

Thomson scattering, Rayleigh scattering, and bremsstrahlung [Master]

Thomson scattering is the elastic scattering of electromagnetic radiation by a free charge. The incident wave drives oscillation $x=q{E}_0\cos \omega t/(m{\omega }^2)$ for $\omega$ well below any resonance. The oscillating charge radiates (Larmor) and the ratio of scattered to incident power defines the Thomson cross-section ${\sigma }_T=(8\pi /3)r_0^2$ where $r_0=q^2/(4\pi {\epsilon }_{0}m{c}^2)\approx 2.82\times 10^{-15}$ m. The quantum generalisation (Compton scattering) reduces to Thomson for $\hslash \omega \ll m{c}^2$.

Rayleigh scattering is Thomson scattering from a bound charge distribution smaller than the wavelength. For a polarisable atom with static dipole polarisability ${\alpha }_0$, the cross-section is $\sigma \propto {\omega }^4{\alpha }_0^2$. The ${\omega }^4$ dependence is why the sky is blue and sunsets are red: shorter wavelengths scatter much more.

Bremsstrahlung (braking radiation) is emitted when a charged particle is deflected by another charge (typically an electron by a nucleus). The acceleration $a\sim Z{e}^2/(4\pi {\epsilon }_{0}m{r}^2)$ at impact parameter $r$ gives a radiated energy $\sim P\cdot \Delta t\sim q^2{a}^2\Delta t/(6\pi {\epsilon }_0{c}^3)$, which integrates over impact parameters to give the bremsstrahlung cross-section. The Bethe-Heitler formula gives the full quantum result. In a plasma with ion number density $n_i$, the bremsstrahlung power per unit volume is $P/V\propto Z^2{n}_i{n}_e\sqrt{k_{B}T}$, which dominates cooling in hot ($T\gtrsim 10^7$ K) astrophysical plasmas.

Radiation reaction: Abraham-Lorentz-Dirac equation [Master]

The self-force on a radiating charge has a troubled history. The sequence of results:

Abraham-Lorentz force (1904–1909):

$$m\dot{\mathbf{v}}={\mathbf{F}}_{\mathrm{ext}}+\frac{q^2}{6\pi {\epsilon }_0{c}^3}\ddot{\mathbf{v}}.$$

This is third-order in time, admitting runaway solutions ($\mathbf{v}\propto e^{t/{\tau }_0}$ with ${\tau }_0=q^2/(6\pi {\epsilon }_{0}m{c}^3)\approx 6.3\times 10^{-24}$ s for the electron) and pre-acceleration.

Dirac (1938): Dirac derived the covariant equation

$$m\frac{d{U}^{\mu }}{d\tau }=F_{\mathrm{ext}}^{\mu }+\frac{q^2}{6\pi {\epsilon }_0{c}^3}(\frac{d^2{U}^{\mu }}{d{\tau }^2}+\frac{1}{c^2}{U}^{\mu }\frac{d{U}_{\nu }}{d\tau }\frac{d{U}^{\nu }}{d\tau }).$$

The second term in the radiation-reaction bracket ensures the constraint $U_{\mu }{U}^{\mu }=c^2$ is preserved. The Dirac equation still admits runaway solutions; imposing the boundary condition $d{U}^{\mu }/d\tau \to 0$ as $\tau \to +\infty$ eliminates them at the cost of pre-acceleration on timescale ${\tau }_0$.

Landau-Lifshitz reduction (1962): Treat the radiation reaction as a perturbation. Substitute the zeroth-order equation $md{U}^{\mu }/d\tau =F_{\mathrm{ext}}^{\mu }$ into the radiation-reaction term, obtaining a second-order equation:

$$m\frac{d{U}^{\mu }}{d\tau }=F_{\mathrm{ext}}^{\mu }+\frac{q^2}{6\pi {\epsilon }_0{c}^3}[\frac{d{F}_{\mathrm{ext}}^{\mu }}{d\tau }+\frac{1}{m{c}^2}{U}^{\mu }{F}_{\mathrm{ext}}^{\nu }{F}_{\mathrm{ext},\nu }].$$

This is a regular second-order ODE with no runaway solutions. It is the standard equation used in particle-accelerator simulations and in plasma physics. The regime of validity is ${\tau }_0\omega \ll 1$ (radiation reaction is a small correction to the Lorentz force), which holds for all practical accelerator and astrophysical applications.

Limits of classical radiation theory. The classical theory breaks down when $\hslash \omega \gtrsim m{c}^2$ (quantum radiation regime) or when $\gamma$ is so large that the formation length exceeds the quantum coherence length. The quantum radiation-reaction parameter $\chi =\gamma E/E_{\mathrm{crit}}$ (where $E_{\mathrm{crit}}=m^2{c}^3/(q\hslash )\approx 1.3\times 10^{18}$ V/m for electrons) governs the transition: for $\chi \ll 1$ the classical Larmor/Lienard formulae apply; for $\chi \sim 1$ quantum electrodynamics (QED) must be used, and radiation is emitted stochastically rather than continuously.

Connections [Master]

• Gravitational radiation (unit 13.07.01): The quadrupole formula for gravitational-wave power $P_{\mathrm{GW}}=(G/5c^5)⟨{\stackrel{...}{Q}}_{ij}{\stackrel{...}{Q}}^{ij}⟩$ is the gravitational analogue of the Larmor formula. Both express radiated power as a squared time-derivative of a source moment: the dipole moment $q\mathbf{r}$ for EM, the quadrupole moment $Q_{ij}$ for gravity. The gravitational formula has no dipole term because mass is positive and the mass dipole is just the centre of mass (conserved).

• Spectroscopy (unit 14.12.01): Atomic spectra arise from quantum transitions that are the discrete analogues of classical radiation. The Einstein A coefficient for spontaneous emission is the quantum generalisation of the Larmor power: $A_{21}={\omega }^3|d_{21}{|}^2/(3\pi {\epsilon }_0\hslash c^3)$ where $d_{21}$ is the transition dipole matrix element. The classical Larmor power for an oscillating dipole $d=q{x}_0\cos \omega t$ gives $P={\omega }^4|d{|}^2/(12\pi {\epsilon }_0{c}^3)$, and the correspondence $P=\hslash \omega \cdot A$ connects the two.

• Antenna theory: The Larmor formula generalises to arbitrary current distributions via the multipole expansion. The electric-dipole radiation from a current $\mathbf{J}$ is $P={\omega }^4|\mathbf{p}{|}^2/(12\pi {\epsilon }_0{c}^3)$ where $\mathbf{p}=\int {\mathbf{r}}^{\prime }\rho d^3{r}^{\prime }$. This is the starting point for all antenna design.

• Covariant electrodynamics (unit 10.06.01): The Lienard formula is the contraction $P=-(q^2/6\pi {\epsilon }_0{c}^3){\alpha }_{\mu }{\alpha }^{\mu }$ of the four-acceleration with itself, making it a Lorentz scalar. The derivation from the Faraday tensor $F^{\mu \nu }$ and the retarded Green's function on Minkowski spacetime is the cleanest route.

• Plasma physics: Synchrotron and bremsstrahlung radiation are the dominant cooling mechanisms in hot plasmas. The Larmor power sets the energy loss timescale ${\tau }_{\mathrm{cool}}=E/P$ for charged particles in magnetic confinement devices and in astrophysical environments (pulsar magnetospheres, accretion disks).

Historical and philosophical context [Master]

The recognition that accelerating charges radiate developed over three decades at the turn of the twentieth century.

Larmor (1897): Joseph Larmor derived the power formula $P=2q^2{a}^2/(3c^3)$ in Gaussian units, published in Philosophical Magazine 44, 503–512 (1897). His derivation used the Hertzian dipole radiation formula and the correspondence between a point charge and an oscillating dipole. Larmor was working within the ether-theoretic framework; the result is independent of that interpretation.

Lienard (1898) and Wiechert (1900): Independently derived the retarded potentials for a moving point charge. Lienard's paper appeared in L'eclairage electrique 16, 5–14, 53–59, 106–112 (1898). Wiechert's in Archives Neerlandaises 5, 549–563 (1900). The Lienard-Wiechert potentials remain the standard starting point for radiation calculations.

Abraham (1904): Max Abraham derived the radiation-reaction force $\mathbf{F}=(2q^2/3c^3)\ddot{\mathbf{v}}$ (Gaussian) by computing the self-force of the extended electron model. His derivation assumed a rigid spherical shell of charge and identified the $4/3$ problem in electromagnetic mass.

Lorentz (1909): Hendrik Lorentz gave a more systematic treatment in The Theory of Electrons, deriving the radiation-damping force and noting the runaway problem.

Dirac (1938): Paul Dirac rederived the radiation-reaction equation from energy-momentum conservation across the charge's light cone, obtaining the covariant Abraham-Lorentz-Dirac equation. His paper (Proceedings of the Royal Society A167, 148–169) introduced the boundary-condition resolution of the runaway problem.

Landau and Lifshitz (1962): The perturbative reduction of the Abraham-Lorentz-Dirac equation to a regular second-order ODE appeared in The Classical Theory of Fields (Course of Theoretical Physics Vol. 2, 2nd ed., Pergamon, 1962), §76. This is now the standard approach in accelerator physics.

The philosophical significance: the radiation-reaction problem was the first indication that classical point-charge electrodynamics is not a self-consistent theory. The third-order equation, runaway solutions, and pre-acceleration signal that at distances $\sim r_0=q^2/(4\pi {\epsilon }_{0}m{c}^2)\approx 2.8\times 10^{-15}$ m (the classical electron radius) the classical theory must be replaced by QED. The classical electron radius is precisely the scale at which the electrostatic self-energy $q^2/(4\pi {\epsilon }_0{r}_0)$ equals $m{c}^2$.

Bibliography [Master]

• Larmor, J. (1897). "On the theory of the magnetic influence on spectra; and on the radiation from moving ions." Philosophical Magazine 44, 503–512. The original derivation of the Larmor power formula.

• Lienard, A. (1898). "Champ electrique et magnetique produit par une charge electrique concentree en un point et animee d'un mouvement quelconque." L'eclairage electrique 16, 5–14, 53–59, 106–112. The retarded potential for a moving point charge.

• Wiechert, E. (1900). "Elektrodynamische Elementargesetze." Archives Neerlandaises 5, 549–563. Independent derivation of the retarded potential.

• Abraham, M. (1904). "Theorie der Elektrizitat," Vol. 2. Teubner, Leipzig. The radiation-reaction force from the extended electron model.

• Lorentz, H. A. (1909). The Theory of Electrons. Teubner, Leipzig. Systematic treatment of radiation damping and the self-force.

• Dirac, P. A. M. (1938). "Classical theory of radiating electrons." Proceedings of the Royal Society A167, 148–169. The covariant equation and boundary-condition resolution.

• Landau, L. D. and Lifshitz, E. M. (1975). The Classical Theory of Fields, 4th ed. Pergamon. §66–67 (radiation from moving charges), §75–76 (radiation reaction and the Landau-Lifshitz reduction). The standard reference for the perturbative approach.

• Jackson, J. D. (1999). Classical Electrodynamics, 3rd ed. Wiley. Ch. 14 (radiation by moving charges), Ch. 16 (radiation damping). The standard graduate textbook.

• Griffiths, D. J. (2017). Introduction to Electrodynamics, 4th ed. Cambridge. Ch. 11 (radiation). Accessible intermediate-level treatment with worked examples.

• Zangwill, A. (2013). Modern Electrodynamics. Cambridge. Ch. 20 (radiation), Ch. 23 (radiation reaction). Comprehensive modern reference.

• Susskind, L. and Friedman, A. (2017). Special Relativity and Classical Field Theory. Basic Books. Lecture 11. Beginner-accessible treatment of radiation from accelerating charges.