Runaway Solutions and Preacceleration

Two articles in this track have already put the Abraham–Lorentz equation on trial. The strong-UG pair — Radiation Reaction Force and Abraham–Lorentz Equation and Its Pathologies — derived the self-force from energy bookkeeping, exhibited the runaway mode with an integrating factor, solved the step force exactly, and executed the Landau–Lifshitz reduction. Next door, Abraham-Lorentz Radiation Reaction Force assembles the covariant equation and proves the validity-domain theorem. This article repeats none of it: it is the pathology report — why one term produces exactly these two diseases, why you must choose one, and what acquits electrodynamics. The instrument is the response function. In the frequency domain the Abraham–Lorentz equation is one line, $\chi(\omega)=1/m(1+i\omega\tau)$ , and its entire pathology is one geometric fact: the pole sits at $\omega=+i/\tau$ , in the upper half-plane — precisely where causality forbids poles to live. From that misplaced pole we derive the trade theorem (causal solutions run away; bounded ones respond before the force), run the future-sided kernel through its starkest exhibits — the delta kick answered entirely in advance, the sinusoid answered with a phase lead — and price the fine-tuning that runaway suppression demands: sixty-nine million decimal digits to survive one femtosecond. Then the acquittal: give the electron any structure bigger than $c\tau=1.88\ \mathrm{fm}$ and the self-force becomes a delay, the misplaced pole migrates to safety, and both diseases vanish. The quantum electron, spread over a Compton wavelength two hundred times larger, was never in danger.

Key takeaways

In Fourier space $m(a-\tau\dot a)=F_{\text{ext}}$ has response $\chi(\omega)=1/m(1+i\omega\tau)$ , pole at $\omega=+i/\tau$ : the upper half-plane, where Titchmarsh forbids poles of any causal, stable response.
The misplaced pole forces a choice of Green's function: the causal one grows as $e^{t/\tau}$ (runaway), the bounded one has support before the force (preacceleration). Causality XOR boundedness — the trade theorem; there is no third option inside the equation.
The bounded branch is the future-sided kernel $ma(t)=\int_0^\infty e^{-s}F(t+\tau s)\,ds$ : acceleration now = force averaged over the coming few $\tau$ . A delta kick is answered entirely before it lands; a sinusoid is answered with a phase lead $\arctan(\Omega\tau)$ .
Expanding the kernel in moments gives $ma=F+\tau\dot F+\tau^2\ddot F+\cdots$ — an asymptotic ladder whose first rung is exactly the Landau–Lifshitz reduction, explaining why order reduction is the correct effective-theory move.
Runaway suppression is infinitely fine-tuned: an initial-acceleration error $\varepsilon$ grows to $\varepsilon\,e^{t/\tau}$ , so surviving 1 fs demands tuning to one part in $10^{69{,}000{,}000}$ .
A charged shell's self-force is a velocity difference across the delay $2a/c$ ; no runaway root exists when $a>c\tau=\tfrac{2}{3}r_e$ — exactly when the bare mass is positive. The quantum electron's effective size $\lambda\!\!\!/_C=206\,c\tau$ settles the case.
Key idea: the runaway and preacceleration are one pole in the wrong half-plane — artifacts of taking the point limit past the theory's own smallest scale, removed by any honest cutoff and by the quantum theory that replaces it.

The core insight

The radiation-reaction term raises Newton's equation to third order and puts the response function's pole at $\omega=+i/\tau$ , in the causal-forbidden upper half-plane; every pathology — exponential runaways, preacceleration, infinite fine-tuning — is that one misplaced pole, and any physical cutoff larger than $c\tau$ moves it back to safety.

Where this lives today

The practical legacy is a rule burned into every code that pushes charges: never integrate the third-order equation. Particle-in-cell and beam-dynamics packages implement radiation reaction through the Landau–Lifshitz form or its quantum-corrected extensions, because a naive Abraham–Lorentz integrator excites the $e^{t/\tau}$ mode from floating-point noise within a handful of steps — Example 1's ledger enforced by machine epsilon. The frontier stress-testing these solvers is the strong-field programme covered next door in Abraham-Lorentz Radiation Reaction Force. The deeper legacy is methodological. The gravitational self-force programme — extreme-mass-ratio inspirals for LISA — faces the same structure: a point particle's field diverging on its own worldline, a regularized equation of motion (MiSaTaQuWa) with higher-derivative terms, the same runaway threat, and the same cure: order reduction, justified by the effective-field-theory logic Derivation B's moments ladder makes explicit. The Abraham–Lorentz pathology is now the canonical statement of a standard lesson: when a small parameter multiplies your highest derivative, the extra solutions are lies told by the truncation.

Origins · 1 minJohn Archibald Wheeler and how this was discovered

John Archibald Wheeler

1911–2008 · the physicist who took the advanced solutions seriously — and made radiation reaction a two-way conversation with the universe

Wheeler treated scandals as clues. Where others saw the advanced solutions of Maxwell's equations — waves converging from the future — as debris, Wheeler and his student Feynman built the 1945 absorber theory on them: each charge interacts time-symmetrically, half retarded plus half advanced, with every other charge in the universe, and the absorber's advanced response conspires to produce both the observed retarded radiation and the exact Abraham–Lorentz self-force with no self-interaction at all. Radiation reaction becomes the universe answering back; preacceleration is the advanced whisper of that reply. Superseded by QED (its quantization troubles, Feynman said, drove him to path integrals), it remains the sharpest statement that the arrow of radiation lives in boundary conditions, not laws. Wheeler went on to name black holes and mentor a lineage from Feynman to Thorne; absorber theory shows the trait his students revered — the nerve to push respectable formalism into disreputable territory and read the answer honestly.

History

1904

— Abraham and Lorentz assemble the self-force of the deformable/rigid electron; the third-derivative term and its strange solutions surface immediately

1938

— Dirac derives the covariant equation from half-retarded-minus-advanced fields with mass renormalization, exhibiting the runaway and the future-integral remedy

1945

— Wheeler and Feynman publish absorber theory: radiation reaction as the time-symmetric response of the rest of the universe, reproducing the self-force without self-interaction

1948

— Eliezer proves the hydrogen-like atom admits no physical Abraham–Lorentz–Dirac solutions falling into the nucleus — the equation cannot be taken literally

1977

— Levine, Moniz and Sharp complete the extended-electron analysis: structure larger than cτ damps every root; the quantum electron is runaway-free with two orders to spare

The physical picture

Start with the knife edge. The general solution — constructed by integrating factor in the strong-UG Abraham–Lorentz Equation and Its Pathologies, recalled not re-derived — is one particular response plus a free multiple of $e^{t/\tau}$ . Picture the space of solutions as a sheaf of trajectories labelled by the initial acceleration: one single, infinitely thin thread stays bounded, and every neighbour peels away from it exponentially, doubling every $4.3\times10^{-24}$ s. This is not an unstable equilibrium of the motion but an instability of the law itself, present for a free particle in empty space: as an initial-value problem, the runaway is the generic fate, and the bounded thread demands literally unphysical precision. The second picture explains the first. Ask of the equation what one asks of any linear system: where are the poles of the response? A damped oscillator keeps its poles in the lower half of the complex $\omega$ plane — that placement is causality, letting the response vanish before the stimulus and decay after it. The Abraham–Lorentz response has its lone pole at $\omega=+i/\tau$ , upstairs. A pole upstairs offers only two contour routes, and they are the two diseases: demand causal support and the pole's residue grows as $e^{t/\tau}$ (runaway); demand boundedness and the response function acquires support at negative times (preacceleration). One pole, two symptoms, zero escapes — within the equation. The third picture is the escape hatch. The point-charge limit is what parked the pole upstairs. A charge of finite size $a$ answers its own field only after the light-crossing delay $\sim 2a/c$ , and a delay in the time domain multiplies the characteristic equation by $e^{-2as/c}$ in the frequency domain — which, for $a>c\tau$ , removes every root from the right half-plane. The pathologies live only when the electron is smaller than $1.88$ fm; quantum mechanics smears it over $386$ fm. The diseases are real properties of an idealization, not of nature.

Figure

The knife edge — solutions of the force-free Abraham-Lorentz equation for initial accelerations differing by parts in a thousand; the bounded thread is the measure-zero exception

Figure

The misplaced pole — the response function of the Abraham-Lorentz equation has its pole in the upper half of the complex frequency plane, and the two contour choices are the two diseases

The math: one pole, two diseases

Everything below runs on the nonrelativistic Abraham–Lorentz equation, quoted by name from the strong-UG pair that derived and first dissected it — Radiation Reaction Force (the energy-bookkeeping construction and the Lorentz-shell mass renormalization) and Abraham–Lorentz Equation and Its Pathologies (the integrating-factor solution, the exact step force, the Landau–Lifshitz reduction):

m\big(a-\tau\,\dot a\big)=F_{\text{ext}}(t),\qquad \tau=\frac{\mu_0 q^2}{6\pi m c}=\frac{2}{3}\frac{r_e}{c}=6.27\times10^{-24}\ \mathrm{s}\ \text{(electron)}, \qquad (1)

where: $a$ is the acceleration along one Cartesian component (the vector equation decouples), $\tau$ is the Larmor characteristic time built in Larmor Formula from the Liénard–Wiechert Fields, and the covariant completion with its validity-domain theorem is the neighbouring Abraham-Lorentz Radiation Reaction Force. The strong-UG twin proved that (1)'s unique bounded solution is the future-weighted integral $a(t)=\frac{1}{m\tau}\int_t^{\infty}e^{-(t'-t)/\tau}F_{\text{ext}}(t')\,dt'$ ; this article treats that object — and the runaway it exorcises — as a system, in the frequency domain where its anatomy is one pole.

Derivation AThe response function and the trade theorem

The goal is to compress both pathologies into one statement about analyticity: compute the response function, locate its pole, and show that its position — upstairs in the complex frequency plane — makes "causal" and "bounded" mutually exclusive. The strategy is standard linear response: Fourier transform, read off $\chi(\omega)$ , then construct the two admissible Green's functions by the jump condition and watch each contract a different disease. Watch the direction of the logic: we do not choose a pathology — the pole forces the pair, and Titchmarsh's theorem certifies that no cleverer contour exists.

Step 1. Fourier transform. With the physics convention $a(t)=\int\frac{d\omega}{2\pi}\,\tilde a(\omega)e^{-i\omega t}$ , each time derivative brings $-i\omega$ , so $\dot a\to-i\omega\tilde a$ and (1) becomes algebraic:

m\,\tilde a(\omega)\,\big(1+i\omega\tau\big)=\tilde F(\omega) \qquad\Longrightarrow\qquad \tilde a(\omega)=\chi(\omega)\,\tilde F(\omega),\quad \chi(\omega)=\frac{1}{m\,(1+i\omega\tau)} .

where: $\chi$ is the acceleration response to force — the whole dynamical content of (1) in one rational function. Its single pole sits where $1+i\omega\tau=0$ , i.e. at $\omega_\star=+i/\tau$ : on the positive imaginary axis, magnitude $1/\tau=1.60\times10^{23}\ \mathrm{s^{-1}}$ . Why the location is a verdict before any calculation: for the $e^{-i\omega t}$ convention, a response that is causal (vanishes before the stimulus) and stable (bounded after it) must be analytic everywhere in the upper half-plane — that is Titchmarsh's theorem, the same analyticity that generates the Kramers–Kronig relations obeyed by the dielectric response in the absorption-and-dispersion article. A physical $\chi$ may have poles only downstairs. Ours has its only pole upstairs. Something must give.

Step 2. Construct the Green's functions. $G(t)$ solves $m(G-\tau\dot G)=\delta(t)$ , so that $a(t)=\int G(t-t')F(t')dt'$ . Away from $t=0$ the homogeneous equation forces $G\propto e^{t/\tau}$ on each side; integrating the equation across $t=0$ gives the jump $-m\tau\,[G(0^+)-G(0^-)]=1$ . Exactly two one-sided assignments exist:

G_{\text{causal}}(t)=-\frac{1}{m\tau}\,e^{t/\tau}\,\theta(t), \qquad\qquad G_{\text{bounded}}(t)=+\frac{1}{m\tau}\,e^{t/\tau}\,\theta(-t).

where: $\theta$ is the unit step. Both satisfy the defining equation and the jump; they differ by the homogeneous runaway mode. In contour language: $G(t)=\int\frac{d\omega}{2\pi}\frac{e^{-i\omega t}}{m(1+i\omega\tau)}$ closes upward for $t<0$ and downward for $t>0$ ; putting the pole inside the upper closure yields $G_{\text{bounded}}$ , deforming it out yields $G_{\text{causal}}$ .

Step 3. Watch each contract its disease. $G_{\text{causal}}$ has support only after the stimulus — causality holds — but grows as $e^{t/\tau}$ forever: a step force answered through it gives $a(t)=(F/m)(1-e^{t/\tau})$ , which even starts in the wrong direction ( $\dot a(0)=-F/m\tau$ ) before running away. (Check against (1): $a-\tau\dot a=(F/m)(1-e^{t/\tau})+(F/m)e^{t/\tau}=F/m$ , exactly.) $G_{\text{bounded}}$ stays finite for all time — but its support is entirely at $t<0$ : the response precedes the force. Convolving it reproduces the future-weighted integral the strong-UG twin derived by boundary condition: "the non-runaway solution" and "the acausal Green's function" are one object in two domains.

Step 4. The trade theorem. Were some Green's function of (1) both causal and bounded, its transform would be analytic and bounded upstairs (Titchmarsh); but every Green's function of (1) transforms to $\chi(\omega)$ up to the homogeneous mode, and $\chi$ has its pole at $+i/\tau$ . Contradiction: causality and stability are jointly unsatisfiable — runaway XOR preacceleration, decided by a contour, no third setting. Why this framing earns its keep: two separate scandals become one theorem with one hypothesis — "the pole is upstairs" — which tells any cure exactly what to do: move the pole. Derivation C shows finite charge structure does precisely that. *Verified: $G_{\text{bounded}}$ convolved with $F\theta(t)$ reproduces the strong-UG twin's exact step solution ( $a=(F/m)e^{t/\tau}$ before, $F/m$ after); the causal branch's backward start $\dot a(0^+)=-F/m\tau$ checks directly from (1) with $a(0)=0$ ; pole magnitude $\hbar/\tau=105.0\ \mathrm{MeV}=\tfrac{3}{2}mc^2/\alpha$ recomputed from CODATA constants.*

The pole's magnitude deserves its own sentence. The runaway frequency scale is $1/\tau$ , and in energy units

\frac{\hbar}{\tau}=\frac{3}{2}\,\frac{mc^2}{\alpha}=105\ \mathrm{MeV}, \qquad (2)

where: $\alpha\approx1/137$ is the fine-structure constant. The identity is exact — $\tau=\tfrac{2}{3}r_e/c$ and $r_e=\alpha^2 a_0$ do the work — and it is the theory pronouncing sentence on itself: the pathology lives at an energy scale two hundred times the electron's rest mass and a hundred times the electron–positron pair threshold, deep inside territory where classical electrodynamics has already abdicated. The Bridge from Classical Electrodynamics to QED article collects this and the track's other self-indictments.

Go deeper · Derivation B · 3 minThe future kernel at work — delta kick, phase lead, and the moments ladder

Derivation BThe future kernel at work — delta kick, phase lead, and the moments ladder

The goal is to make the bounded branch quantitative: recast it as the future-sided kernel, run it through the two starkest exhibits — an impulsive kick and a steady sinusoid — then expand in moments to find that its first honest truncation is exactly the Landau–Lifshitz reduction. The strategy: substitute once to reach the dimensionless form, evaluate the kernel in closed form for each drive, and audit every result against the differential equation. Watch the kernel turn from scandalous (sharp drives) to respectable (forces slow on the scale $\tau$ ) — that crossover is the entire practical content of radiation reaction.

Step 1. The kernel in its natural variables. In the strong-UG twin's bounded solution substitute $t'=t+\tau s$ :

m\,a(t)=\int_0^{\infty}e^{-s}\,F_{\text{ext}}(t+\tau s)\,ds .

where: $s$ counts e-folds of the weighting. Read it as an expectation value: the acceleration now equals the force averaged over the future, sampled with weight $e^{-s}$ over the coming few $\tau$ . Setting $F=\text{const}$ gives $ma=F$ (the weight integrates to one) — Newton is embedded exactly, and all deviations are properties of the force's future variation.

Step 2. The delta kick — response entirely before the stimulus. Take $F(t)=\Delta p\,\delta(t)$ , an impulse delivering momentum $\Delta p$ . The kernel picks up the spike only while $t+\tau s=0$ has a solution with $s>0$ , i.e. only for $t<0$ :

a(t)=\frac{\Delta p}{m\tau}\,e^{t/\tau}\ \ (t<0), \qquad a(t)=0\ \ (t>0).

The particle does all of its accelerating before the kick lands and coasts afterward. The books still balance: $\int_{-\infty}^{0}a\,dt=\Delta p/m$ exactly — momentum conservation is untouched; only its timing is scandalous. This is preacceleration with the step's residual Newtonian phase stripped away: the pure anticipatory mode, nothing else. Why the exhibit matters: it kills the misreading of preacceleration as small "early leakage" of a normal response — for a sharp enough drive the response is one hundred percent early. The sharper the force, the worse the anticipation.

Step 3. The sinusoid — acausality as a phase lead. Take $F=F_0\cos\Omega t$ . The kernel evaluates in closed form ( $\int_0^\infty e^{-s}\cos(\Omega t+\Omega\tau s)\,ds=\mathrm{Re}\,[e^{i\Omega t}/(1-i\Omega\tau)]$ ):

a(t)=\frac{F_0/m}{\sqrt{1+\Omega^2\tau^2}}\,\cos\!\big(\Omega t+\varphi\big), \qquad \tan\varphi=\Omega\tau .

where: the amplitude is reduced by $(1+\Omega^2\tau^2)^{-1/2}$ — the radiated toll — and $\varphi$ is a lead: acceleration peaks before force, by $\varphi/\Omega\approx\tau$ . Every passive causal system (the driven dielectric of the absorption article, any RC circuit) responds with a lag; a lead is the frequency-domain fingerprint of the future-sided kernel. For an 800 nm laser on a free electron, $\Omega\tau=1.5\times10^{-8}$ rad of illegal anticipation and an amplitude correction of order $10^{-16}$ : the crime is real, the loot unmeasurable.

Step 4. The moments ladder, and why Landau–Lifshitz is the honest truncation. Taylor-expand the force inside the kernel, $F(t+\tau s)=\sum_n \frac{(\tau s)^n}{n!}F^{(n)}(t)$ , and use $\int_0^\infty s^n e^{-s}ds=n!$ :

m\,a(t)=F(t)+\tau\,\dot F(t)+\tau^2\,\ddot F(t)+\cdots

where: each rung costs one more factor of $\tau\times$ (force's rate of change) — the series is asymptotic, useful exactly when the force varies slowly on the scale $\tau$ , which is every situation classical electrodynamics may legitimately describe. Truncating after the second term is precisely the Landau–Lifshitz equation, executed at strong-UG level in Abraham–Lorentz Equation and Its Pathologies and in covariant form next door in Abraham-Lorentz Radiation Reaction Force — recalled here, not re-derived. The ladder explains its legitimacy: LL is not an ad-hoc repair but the leading term of the bounded branch's own slow-force expansion, with the acausal window $\sim\tau$ collapsed to a point. Runaways never appear because the expansion never resurrects the homogeneous mode; preacceleration never appears because anticipation by $\tau$ is invisible at any order in a slow-force expansion. Verified: the sinusoid solution substituted into (1) leaves residual below $10^{-8}$ of the drive in a direct numerical check; the delta-kick response integrates to $\Delta p/m$ to five digits on a fine grid; the moments identity $\int_0^\infty s^ne^{-s}ds=n!$ is exact; $\Omega\tau$ at 800 nm recomputed as $1.48\times10^{-8}$ .

Figure

The purest exhibit — a delta kick at t = 0 is answered entirely before it arrives; the exponential anticipation carries exactly the full impulse

Go deeper · Derivation C · 2 minThe extended electron — a light-crossing delay moves the pole and acquits the theory

Derivation CThe extended electron — a light-crossing delay moves the pole and acquits the theory

The goal is the acquittal: any charge structure of size $a>c\tau$ removes the runaway root entirely — and the criterion is identical to a positive bare mass. The strategy: quote the closed form of the rigid shell's self-force — a pure velocity difference across the light-crossing delay — audit that it reproduces both the electromagnetic-mass and Abraham–Lorentz terms in the point limit, then run the characteristic equation on force-free motion. The subtlety to watch: the shell's electromagnetic mass carries the notorious 4/3 factor; its dissection belongs to Classical Electron Theory and Self-Energy next door and is deliberately not performed here — for the runaway question only the sign of the bare mass matters.

Step 1. The shell's self-force. For a rigid spherical shell of charge $q$ and radius $a$ in nonrelativistic motion, the exact self-force (Sommerfeld 1904, Page 1918; the modern treatment is Levine–Moniz–Sharp) is a velocity difference across the delay $\delta_a=2a/c$ :

F_{\text{self}}(t)=\frac{q^2}{12\pi\varepsilon_0 a^2 c}\,\Big[v(t-\delta_a)-v(t)\Big] .

where: $\delta_a$ is the light-crossing time of the shell's diameter — the soonest one side can learn what the other did. The self-interaction is retarded bookkeeping, nothing more.

Step 2. Audit the point limit. Expand $v(t-\delta_a)=v-\delta_a\dot v+\tfrac{1}{2}\delta_a^2\ddot v-\cdots$ :

F_{\text{self}}=-\frac{q^2}{6\pi\varepsilon_0 a c^2}\,\dot v\;+\;\frac{q^2}{6\pi\varepsilon_0 c^3}\,\ddot v\;+\;O(a) .

The first term is the electromagnetic-mass drag $-m_{\text{em}}\dot v$ with $m_{\text{em}}=q^2/6\pi\varepsilon_0ac^2$ (the shell-renormalization story of Radiation Reaction Force, recalled); the second is exactly $m\tau\,\ddot v$ — the Abraham–Lorentz term, with the correct coefficient and no free parameters, and every $a$ -dependence gone. The ALD force is the first structure-independent whisper of the delay. Why this audit is the hinge: it certifies that (1) is the $a\to0$ limit of a perfectly causal, delay-based self-interaction. The pathologies must therefore be born in the limit itself — and the characteristic equation shows exactly where.

Step 3. The characteristic equation. Force-free motion of the shell obeys $m_0\dot v(t)=F_{\text{self}}(t)$ with $m_0=m-m_{\text{em}}$ the bare (non-electromagnetic) mass. Insert $v=e^{st}$ :

m_0\,s=\frac{q^2}{12\pi\varepsilon_0a^2c}\,\big(e^{-s\delta_a}-1\big) .

where: runaways = roots with $\operatorname{Re}s>0$ . Test the real axis: for real $s>0$ the left side is positive when $m_0>0$ , the right side strictly negative ( $e^{-s\delta_a}<1$ ): no positive real root exists — the runaway is gone — precisely when the bare mass is positive. The full complex-root analysis (Levine–Moniz–Sharp 1977) confirms all roots then have $\operatorname{Re}s<0$ : genuine damping, no anticipation.

Step 4. Translate the criterion. $m_0>0$ means $m>m_{\text{em}}=q^2/6\pi\varepsilon_0ac^2$ , i.e.

a\;>\;\frac{q^2}{6\pi\varepsilon_0 mc^2}=\frac{2}{3}\,r_e\;=\;c\,\tau\;=\;1.88\ \mathrm{fm}.

Quantum mechanics acquits the electron before the case is heard: its wavefunction cannot be confined below $\lambda\!\!\!/_C=\hbar/mc=386\ \mathrm{fm}$ , exceeding $c\tau$ by $3/2\alpha\approx206$ ; the Moniz–Sharp nonrelativistic-QED calculation confirms that the Compton-spread electron has neither runaways nor preacceleration, with two orders of magnitude to spare. Conversely, $a<c\tau$ means $m_0<0$ — and that unphysical object, a negative bare mass stabilized by self-repulsion, is what runs away. Verified: the expansion coefficients in Step 2 recomputed symbolically (the $\delta_a$ and $\delta_a^2/2$ terms reproduce $m_{\text{em}}$ and $\mu_0q^2/6\pi c$ exactly); $c\tau=1.879\ \mathrm{fm}=\tfrac{2}{3}r_e$ and $\lambda\!\!\!/_C/c\tau=3/2\alpha=205.6$ recomputed from CODATA; the sign argument at real $s>0$ checked at both limits $s\to0^+$ (slope comparison: $m_0$ vs $-q^2\delta_a/12\pi\varepsilon_0a^2c$ ) and $s\to\infty$ .

Limiting cases

case	result	meaning
Slow force, $\tau\dot F\ll F$	$ma\simeq F+\tau\dot F$	the moments ladder's first rung = Landau–Lifshitz; radiation reaction is a small, causal-looking correction — the physical domain
Constant force	$ma=F$ exactly	the kernel's weight integrates to one; Newton embedded with no correction at all
Impulsive force $\Delta p\,\delta(t)$	$a=(\Delta p/m\tau)e^{t/\tau}\theta(-t)$	pure preacceleration: the entire response precedes the stimulus, impulse exactly conserved
Sinusoid, $\Omega\tau\ll1$	amplitude $\times(1-\tfrac{1}{2}\Omega^2\tau^2)$ , phase lead $\Omega\tau$	acausality as an unmeasurably small phase lead — $1.5\times10^{-8}$ rad at 800 nm
$\Omega\tau\sim1$	anticipation order unity	but $\hbar\Omega\sim105$ MeV: a hundred pair thresholds past classical validity — the regime does not exist for the classical theory
Shell radius $a>c\tau$	all characteristic roots damped	positive bare mass ⟺ no runaway ⟺ causal response; the point limit is the sole culprit

Common trapThe pathologies are predictions of classical electrodynamics that experiments could hunt for

They are not, on two counts. First, magnitude: probing the acausal window $\tau\sim10^{-24}$ s needs $\Omega\sim1/\tau$ , i.e. quanta of $\hbar/\tau=105$ MeV — particle physics, not classical electrodynamics; inside the classical domain the largest illegal signal is a $10^{-8}$ rad phase lead under a $10^{-16}$ amplitude correction. Second, provenance: Derivation C shows the pathologies evaporate for any charge structure larger than $1.88$ fm, while the physical electron is quantum-spread over $386$ fm. The diseases belong to an idealization — the strictly point charge with positive observed mass, hence negative bare mass — that nature never instantiates. The correct posture is the effective-theory one: use the Landau–Lifshitz reduction inside the validity domain (the neighbouring covariant article proves how good it is), and read the runaway/preacceleration pair as the classical theory politely announcing its own ultraviolet boundary.

Worked examples

1 · The tuning ledger — how sharp is the knife edge (confidence builder)

Setup. Treat the force-free Abraham–Lorentz equation as an initial-value problem for an electron, preparing the initial acceleration within $\varepsilon$ of the bounded value. How small must $\varepsilon$ be for the trajectory to stay sane for one attosecond? One femtosecond?

Step 1. The error mode. Any preparation error rides the homogeneous solution: $\delta a(t)=\varepsilon\,e^{t/\tau}$ . "Sane" means, generously, $\delta a$ smaller than some laboratory scale $A$ — the answer will be so extreme that the choice of $A$ is irrelevant.

Step 2. Count the e-folds. With $\tau=6.27\times10^{-24}$ s, one attosecond is $t/\tau=10^{-18}/6.27\times10^{-24}=1.60\times10^{5}$ e-folds, and one femtosecond is $1.60\times10^{8}$ e-folds. Converting to decimal digits ( $\div\ln10=2.303$ ): $6.9\times10^{4}$ digits for the attosecond, $6.9\times10^{7}$ digits for the femtosecond.

Step 3. Confront the numbers. Keeping an electron on the bounded thread for one femtosecond means specifying its initial acceleration to sixty-nine million decimal digits — while 62 digits already distinguishes one Planck volume in the observable universe. No sense exists in which such initial data exists.

The lesson: the bounded solution cannot be reached, only selected by decree — and the decree's price is preacceleration. This is the trade theorem experienced numerically, and why every serious solver abandons (1) for Landau–Lifshitz.

2 · The border audit — pricing the crime at the theory's edge (exam level)

Setup. An 800 nm Ti:sapphire laser drives a free electron. Quantify every acausal signature of the bounded branch, then find where anticipation would become order unity and what physics actually lives there.

Step 1. Phase lead. $\Omega=2\pi c/\lambda=2.36\times10^{15}\ \mathrm{s^{-1}}$ , so $\varphi\simeq\Omega\tau=(2.36\times10^{15})(6.27\times10^{-24})=\mathbf{1.5\times10^{-8}\ rad}$ . The acceleration peaks $\varphi/\Omega\approx\tau=6.3\times10^{-24}$ s before the field does — six yoctoseconds of anticipation.

Step 2. Amplitude toll. The reduction factor is $(1+\Omega^2\tau^2)^{-1/2}\approx1-\tfrac{1}{2}\Omega^2\tau^2$ , a fractional change of $\mathbf{1.1\times10^{-16}}$ — sixteen orders below the drive, far beneath any intensity measurement.

Step 3. Where anticipation becomes order unity. Set $\Omega\tau=1$ : $\Omega=1.60\times10^{23}\ \mathrm{s^{-1}}$ , photon energy $\hbar\Omega=\hbar/\tau=\mathbf{105\ MeV}$ — the identity (2), $\tfrac{3}{2}mc^2/\alpha$ . Compare the milestones passed on the way: quantum recoil matters from $\hbar\Omega\sim mc^2=0.511$ MeV (Compton scattering — the Bridge from Classical Electrodynamics to QED takes over); electron–positron pairs appear from $1.022$ MeV; the sinister frequency sits a factor 103 beyond even that.

Step 4. The timing comparison. The anticipation window is $\tau$ ; the electron's intrinsic quantum time is the Compton time $\hbar/mc^2=1.29\times10^{-21}$ s. Their ratio is $\tau\,mc^2/\hbar=\tfrac{2}{3}\alpha=\mathbf{4.9\times10^{-3}}$ : the classical crime is committed inside a window two hundred times finer than the quantum blur of the criminal. No classical observation can be localized in time finely enough to witness it.

The lesson: every route to observing the pathology — frequency, amplitude, timing — crosses into quantum electrodynamics long before the pathology becomes visible. The acausality is mathematically exact and physically unreachable: an artifact stationed safely beyond the border it marks.

Going further

Provenance. Verify $\tau=\mu_0 q^2/(6\pi m c)=2q^2/(3\cdot4\pi\varepsilon_0 mc^3)=6.26\times10^{-24}\,\mathrm{s}$ for the electron against CODATA constants and Griffiths Eq. 11.79; check $r_e=2.82\times10^{-15}\,\mathrm{m}$ and $\tau=(2/3)r_e/c$ . The Abraham-Lorentz equation, runaway $e^{t/\tau}$ , and the preacceleration integral solution are in Griffiths §11.2.2-11.2.3 and Jackson §16.3 (Eqs. 16.9-16.10). The covariant ALD equation and Schott term are Jackson Eq. 16.42; the Landau-Lifshitz reduction is in Landau & Lifshitz Classical Theory of Fields §76. A verifier should recompute the integrating-factor solution and confirm continuity of $a(t)$ at the step.

Beyond this article. The frequency-domain and extended-electron treatments follow Levine, Moniz & Sharp (Am. J. Phys. 45, 75 (1977)) and Rohrlich's Classical Charged Particles; Dirac's 1938 paper remains startlingly readable; Spohn's Dynamics of Charged Particles and Their Radiation Field proves the order-reduction statements rigorously.

Builds on — Larmor Formula (the τ object) · Relativistic Generalization of Larmor's Formula · Liénard-Wiechert Potentials · Fields of a Relativistic Moving Charge. Leads to — Radiative Frequency Shift and Line Width (the self-force's good sector: the damped oscillator) · Classical Electron Theory and Self-Energy (the 4/3 problem next door) · Bridge from Classical Electrodynamics to QED (the quantum closure of every loose end counted here).

If the future kernel violates causality, why is it still called the physical solution of the Abraham-Lorentz equation?

Because within the equation it is the least unphysical option — the trade theorem leaves only runaway or anticipation, and exponential blow-up of a free particle is infinitely worse than $10^{-24}$ s of early response. But "physical solution" is a courtesy title: the bounded branch is the correct solution of an idealization, agreeing with the causal extended-charge theory everywhere outside a window $\sim\tau$ that classical physics cannot resolve. Practical computations use the Landau–Lifshitz reduction — the bounded branch with the acausal window collapsed to zero: all of the physics, none of the crime.

Does preacceleration let me send a signal into the past?

No. The anticipation is a property of the solution map of a deterministic equation, not a channel: exploiting it would mean deciding the force at $t=0$ after observing the response at $t<0$ , but the force function is given on the whole time axis — the future boundary condition has already consumed it. And any apparatus sharp enough to test $\tau$ -scale timing needs $\hbar\Omega\sim105$ MeV, where pair creation has replaced classical electrodynamics. The acausality is real inside the mathematics and unreachable from the laboratory — what it means for an effective theory to fail gracefully.

The causal Green's function gives a step response that initially accelerates opposite to the force. Is that a typo?

No — a genuine perversity. With $a(0)=0$ the equation itself gives $\dot a(0)=[a(0)-F/m]/\tau=-F/m\tau$ : the first move is backwards. The causal branch can only fit $a(0)=0$ by riding the homogeneous mode with a negative coefficient, so $(F/m)(1-e^{t/\tau})$ dips negative on its way to $-\infty$ and never passes through the sensible value at all. Remember this when someone proposes to "just integrate forward carefully": the causal initial-value problem is not approximately right for a while — it is qualitatively wrong from the first instant.

Why does the moments ladder stop being trustworthy after the first rung, if I can compute every term?

Because it is an asymptotic expansion of an integral whose kernel knows the future. The term ratio is $\sim\tau\dot F/F$ : tiny for slow forces, so the first rung (Landau–Lifshitz) is exponentially accurate — but higher rungs buy nothing, since the neglected remainder is non-perturbative in $\tau$ (it carries the $e^{t/\tau}$ structure) and no finite number of terms sees it. Worse, higher truncations reintroduce higher derivatives and with them exactly the spurious solutions the reduction was built to kill. The effective-theory rule: truncate where the small parameter enters once; everything beyond is noise the idealization cannot honestly claim.

Wheeler-Feynman absorber theory also produces preacceleration. Does that mean the acausality is real physics after all?

Absorber theory relocates the mystery rather than deepening it: the electron has no self-interaction at all, radiation reaction is the advanced response of all other charges, and the half-advanced fields carry a whisper of the future into the present — preacceleration reappears with the same magnitude $\tau$ and the same invisibility. The lesson is that the pathology is not tied to point self-energy specifically but to demanding strictly local, instantaneous dynamics from a system that is nonlocal on the scale $c\tau$ . Every honest cure — finite size, quantum smearing, absorber boundary conditions — introduces its nonlocality openly; the Abraham–Lorentz equation's sin was hiding it in a third derivative.

If the pathologies live outside classical validity, why did Dirac bother constructing the relativistic equation so carefully in 1938?

Because Dirac was after structure, not laboratory numbers — and his construction earned three things that outlived its pathologies. Covariance: the relativistic equation with its Schott term (assembled next door) is the unique consistent completion of the Larmor energy budget, needed wherever radiation reaction is not negligible. Mass renormalization: Dirac's separation of the divergent self-field into an absorbed mass term and a kept finite force is the direct classical ancestor of QED renormalization — the Bridge from Classical Electrodynamics to QED picks up this thread. Honesty: by exhibiting the runaway and the future-integral remedy explicitly, Dirac made the pathology a theorem rather than folklore. Knowing precisely how a theory fails is knowledge.

Runaway Solutions and Preacceleration

Where this lives today

The physical picture

The math: one pole, two diseases

Limiting cases

Worked examples

1 · The tuning ledger — how sharp is the knife edge (confidence builder)

2 · The border audit — pricing the crime at the theory's edge (exam level)

Going further

Before this

Where this leads