Bridge from Classical Electrodynamics to QED

This is the last article of the track. Behind you: vector calculus sharpened into Maxwell's four equations, statics unified by relativity, light derived, radiation priced by Larmor and Liénard, and — in this final chapter — the theory driven honestly to its edges: the self-force acting before its cause, the electron mass refusing to be one number, the line shift that diverges. Those were not loose ends. They were the theory telling you, in its own language, where it ends. This article walks to that edge and looks across. On the far side is quantum electrodynamics — and the remarkable thing, the thing this article exists to show, is how much of the far side is visible from here. The free electromagnetic field, expanded in modes, is exactly a collection of harmonic oscillators; that is a classical theorem, proved with nothing but the Maxwell equations you own. Quantize each oscillator — one postulate — and photons, Planck's spectrum, spontaneous emission, and the Casimir force follow with almost embarrassing economy. The classical radiation formulas do not die in the crossing: they reappear, whole, as the large-quantum-number limit, with the Larmor rate hiding inside the Einstein A coefficient and Maxwell's equations themselves surviving as the wave equation of a single photon. Two sibling articles in the mechanics track made this crossing from other shores — Hamilton–Jacobi becoming Schrödinger, the action becoming the path integral. This article carries the field across: a bridge you are now, after two hundred and thirty-eight articles, equipped to walk.

Key takeaways

Classical electrodynamics fails quantitatively and honestly: the Rayleigh–Jeans spectrum overshoots measured blackbody radiation by a factor of 28 in mid-visible sunlight (and diverges beyond), the photoelectric effect starts nanoseconds after illumination where classical energy accumulation predicts hours, Compton scattering shifts wavelengths by $h/m_ec=2.43\,\mathrm{pm}$ , and the classical atom collapses in 16 ps.
The bridge is a classical theorem plus one postulate: the free field decomposes into normal modes each obeying $\ddot q+\omega^2q=0$ ; quantizing each oscillator gives photon ladders $\hbar\omega(n+\tfrac12)$ .
Detailed balance against Planck's law (Einstein, 1917) forces spontaneous emission with $A=\omega^3|\mathbf d|^2/3\pi\varepsilon_0\hbar c^3$ — numerically the classical damping rate $\Gamma_{\rm cl}$ weighted by the oscillator strength $f$ : for Lyman-α, $f\,\Gamma_{\rm cl}=6.27\times10^8\,\mathrm{s^{-1}}$ , the measured rate.
The fine-structure constant $\alpha=1/137.036$ organizes the handoff as a ladder of lengths: $r_e=\alpha^2 a_0=2.82\,\mathrm{fm}$ , $\lambda\!\!\!/_C=\alpha a_0=386\,\mathrm{fm}$ , $a_0=52.9\,\mathrm{pm}$ — each rung a boundary between theories.
QED repays the classical debts: the linear self-energy divergence softens to a logarithm, the line shift becomes the measured 1058 MHz Lamb shift, and $g-2=\alpha/\pi$ to first order — the most precisely verified prediction in physics.
KEY IDEA: the electromagnetic field is an infinite set of harmonic oscillators; quantizing them turns classical electrodynamics into QED, with every classical formula surviving as the many-photon limit.

The core insight

The free electromagnetic field is an infinite collection of harmonic oscillators; quantization replaces each oscillator's continuous amplitude with a quantum ladder whose excitations are photons. Classical electrodynamics is the coherent, many-photon limit of QED — and the places where the classical theory failed are exactly the places where one quantum matters.

Where this lives today

QED is the working language of precision physics: atomic clocks compare Lamb-shifted levels at the nineteenth digit; the electron's $g-2$ is at once the most stringent test of any theory and the finest source of $\alpha$ ; cavity and circuit QED engineer Derivation A's "springs" one at a time, switching spontaneous emission on and off by rebuilding the vacuum's mode count (the Purcell effect). The bridge itself is now an experimental subject: in 2018, two experiments at the Gemini laser (Cole et al., Poder et al.) collided GeV electron bunches with petawatt pulses and observed quantum radiation reaction — recoil-dominated emission in the regime where the classical Landau–Lifshitz description measurably fails — closing, in the laboratory, the pathology story this chapter's Runaway Solutions article opened. Even the plasma frequency of the previous articles crosses over: $\hbar\omega_p$ in copper is $10.8\,\mathrm{eV}$ , and electron energy-loss spectrometers count it in quanta — plasmons, the collective cousin of the photon. And the track's own geography closes a triptych. The mechanics track built two bridges to the quantum world: Connection to the Schrodinger Equation carried the particle across (Hamilton–Jacobi phase surfaces becoming wavefronts of $\psi$ ), and The Action Principle as Bridge to Quantum Field Theory carried the formalism across ( $e^{iS/\hbar}$ summed over histories). This article carried the field across — mode by mode, oscillator by oscillator. The three crossings are one crossing: Dirac's 1927 paper, which quantized the field's oscillators, opens by noting that the radiation field's Hamiltonian is that of the mechanics it presupposes. Maxwell's equations survive on the far side twice over — as the operator equations of the quantum field, and as the exact wave equation of a single photon's amplitude. The track ends where QED begins; nothing in it was wasted.

Origins · 1 minPaul Adrien Maurice Dirac and how this was discovered

Paul Adrien Maurice Dirac

1902–1984 · the 1927 paper that quantized the electromagnetic field and founded QED

In February 1927, at 24, Dirac submitted "The Quantum Theory of the Emission and Absorption of Radiation" — the paper generally counted as the birth of quantum electrodynamics. It does exactly what this article does: expand the field in oscillators, quantize each, and compute emission and absorption — obtaining Einstein's $A$ and $B$ coefficients from first principles and showing spontaneous emission to be stimulation by the vacuum. (His later, separate 1938 assault on the classical electron's self-force belongs to this chapter's runaway-solutions story — same man, opposite side of the bridge.) Dirac's taste was famously austere — "physical laws should have mathematical beauty" — and QED tested it: he never accepted renormalization, calling the subtraction of infinities "not sensible mathematics," even as the theory's predictions marched to twelve digits. The equation for the relativistic electron, the positron's prediction, Fermi–Dirac statistics, and the bra–ket language itself are his; the 1933 Nobel Prize (shared with Schrödinger) barely covers the ledger.

History

1900

— Planck quantizes the energy exchange of blackbody oscillators: E = n h nu, the first quantum

1905

— Einstein's light-quantum paper explains the photoelectric effect's thresholds and instant onset

1917

— Einstein's A and B coefficients: detailed balance against Planck's law forces spontaneous emission

1923

— Compton measures the wavelength shift of scattered X-rays: photons carry momentum h k

1927

— Dirac quantizes the radiation field: QED begins

1947

— Lamb and Retherford measure the 2s-2p splitting; Bethe computes it with a renormalized mass subtraction

1948

— Schwinger computes g-2 = alpha/pi; Feynman, Schwinger, Tomonaga and Dyson assemble modern renormalized QED

The physical picture

Picture the radiation field as a vast array of springs, one for each mode $(\mathbf k,\lambda)$ — each wavevector, each polarization. Classically a spring can hold any energy and swing with any amplitude; that continuity is precisely what fails. Quantization says each spring holds energy only in steps of $\hbar\omega$ above a ground floor of $\tfrac12\hbar\omega$ ; the step count is the number of photons in the mode. A bright laser beam is one spring crammed with $10^{16}$ quanta, and at that occupancy the graininess is invisible — which is why Maxwell served flawlessly for a century. A single atom deciding whether to emit is a spring absorbing its first quantum, and there the classical picture is not approximately wrong but qualitatively so. Where exactly did it break? At four measured places, each already touched by this track:

Figure

The failure ledger — four measured cracks in the classical wall, and what each one taught

Each row is quantitative. The blackbody failure is not "classical physics gets it a bit wrong": at $T=5800\,\mathrm{K}$ (the solar photosphere) and $\lambda=500\,\mathrm{nm}$ , equipartition among the field's modes — a perfectly correct count of the oscillators, as Derivation A will show — predicts a spectral energy density $28.6$ times the measured one, and by $100\,\mathrm{nm}$ the overshoot is a factor of $2\times10^{9}$ , on its way to a divergent total. The photoelectric failure is a timing catastrophe: spread a milliwatt per square metre of light over atoms and classical continuous absorption needs hours to bank one electron's escape energy, yet photoelectrons appear within nanoseconds. Compton's kink — the wavelength of scattered X-rays shifting by exactly $h/m_ec$ — cashes the validity ledger that the Thomson Scattering article drew up: a classical wave re-radiated by a free electron keeps its frequency, period. And the 16-picosecond death spiral of a classically radiating hydrogen atom, computed in the E9 Larmor Formula article, needs no re-deriving here: it is the shortest reductio in physics. The miracle of economy this article turns on: all four failures — plus this chapter's self-force pathologies — are cured by one structural change, made at a point classical theory itself prepared. You already know (from the Electromagnetic Wave Equation onward) that the free field in a box is a sum of independent plane-wave modes. Each mode's amplitude obeys the harmonic-oscillator equation — that is Maxwell, not Planck. The whole quantum crossing consists of treating each of those oscillators the way quantum mechanics treats every oscillator. Nothing about electromagnetism is re-invented; the field's own classical skeleton is simply read at single-quantum resolution. The crossing comes with a ruler: $\alpha=e^2/4\pi\varepsilon_0\hbar c\approx1/137$ measures the charge–field coupling in quantum units, and stratifies the electron's neighborhood into a ladder of lengths — each rung a factor 137 — telling you which theory governs which scale:

Figure

The alpha ladder — three lengths, each a factor 137 apart, each a boundary between theories

The ladder is this chapter's verdict written as geometry. The Classical Electron Theory and Self-Energy article found the classical theory devouring itself at $r_e$ ; the ladder shows why the disaster is never reached: a full rung before it, at the Compton wavelength, localizing an electron costs enough energy to create electron–positron pairs, the single-particle picture dissolves, and QED — not classical electrodynamics — writes the small-distance rules. The classical divergences are real defects of an extrapolation into territory the theory never owned.

The math: one theorem, one postulate, one constant

The field is oscillators — a classical theorem

Everything rests on a fact provable entirely inside this track. Work in the Coulomb gauge ( $\nabla\cdot\mathbf A=0$ , $\phi=0$ for the free field), where the two transverse polarizations of $\mathbf A$ are the field's only dynamical degrees of freedom — the gauge bookkeeping settled in the E9 potentials chapter.

Derivation AThe free field as a sum of harmonic oscillators

The goal is exact and classical: show that the free electromagnetic field in a box of volume $V$ is dynamically identical to an infinite family of independent harmonic oscillators, one per mode $(\mathbf k,\lambda)$ , with the field energy $\tfrac12\int(\varepsilon_0E^2+B^2/\mu_0)\,\mathrm d^3r$ becoming exactly $\sum\tfrac12(p^2+\omega^2q^2)$ . The strategy is a normal-mode expansion — the same manoeuvre that diagonalizes coupled springs — using the orthogonality of plane waves to kill every cross term. Watch two things: where transversality (the photon's two polarizations) enters, and the fact that no approximation is made anywhere — this is a theorem, not a model.

Step 1. Expand the potential in box modes. With periodic boundary conditions the allowed wavevectors are discrete, and

\mathbf A(\mathbf r,t)=\sum_{\mathbf k,\lambda}\boldsymbol\epsilon_{\mathbf k\lambda}\,\frac{1}{\sqrt{V}}\Big(q_{\mathbf k\lambda}(t)\,e^{i\mathbf k\cdot\mathbf r}+q^*_{\mathbf k\lambda}(t)\,e^{-i\mathbf k\cdot\mathbf r}\Big),

where $\boldsymbol\epsilon_{\mathbf k\lambda}$ ( $\lambda=1,2$ ) are unit polarization vectors with $\mathbf k\cdot\boldsymbol\epsilon_{\mathbf k\lambda}=0$ — the Coulomb-gauge condition $\nabla\cdot\mathbf A=0$ made algebraic. Two polarizations per $\mathbf k$ , not three: transversality has already removed the longitudinal mode, which is why the photon will have two spin states and no rest-frame third.

Step 2. The wave equation decouples the amplitudes. The free-field equation $\nabla^2\mathbf A=c^{-2}\partial_t^2\mathbf A$ acts on each plane wave independently ( $\nabla^2\to-k^2$ ), so every amplitude obeys

\ddot q_{\mathbf k\lambda}+\omega_k^2\,q_{\mathbf k\lambda}=0,\qquad \omega_k=c|\mathbf k| .

Why this is the entire dynamical content: the modes never talk to each other — linearity of Maxwell's equations in vacuum is the statement that the springs are uncoupled. Interactions (charges, or the nonlinear vacuum of strong-field QED) are precisely what couple them.

Step 3. The energy diagonalizes. Insert $\mathbf E=-\partial_t\mathbf A$ and $\mathbf B=\nabla\times\mathbf A$ into $H=\tfrac12\int(\varepsilon_0E^2+B^2/\mu_0)\,\mathrm d^3r$ . Orthogonality of the plane waves ( $\int e^{i(\mathbf k-\mathbf k')\cdot\mathbf r}\mathrm d^3r=V\delta_{\mathbf k\mathbf k'}$ ) annihilates all cross terms, and with real coordinates $Q,P$ built from $q,q^*$ each mode contributes a standard oscillator energy:

H=\sum_{\mathbf k,\lambda}\tfrac12\big(P_{\mathbf k\lambda}^2+\omega_k^2Q_{\mathbf k\lambda}^2\big).

This is still Maxwell — a rewriting, exact, of the field you have used for the whole track.

Step 4. Now the single postulate. Quantum mechanics has one non-negotiable statement about any system with energy $\tfrac12(P^2+\omega^2Q^2)$ : promote $Q,P$ to operators with $[Q,P]=i\hbar$ . Ladder operators $a=(\omega Q+iP)/\sqrt{2\hbar\omega}$ then give, mode by mode,

H=\sum_{\mathbf k,\lambda}\hbar\omega_k\Big(a^\dagger_{\mathbf k\lambda}a_{\mathbf k\lambda}+\tfrac12\Big),\qquad \big[a_{\mathbf k\lambda},a^\dagger_{\mathbf k'\lambda'}\big]=\delta_{\mathbf k\mathbf k'}\delta_{\lambda\lambda'} .

The integer eigenvalue $n_{\mathbf k\lambda}$ of $a^\dagger a$ counts photons: excitation quanta of mode $(\mathbf k,\lambda)$ , each carrying energy $\hbar\omega_k$ and momentum $\hbar\mathbf k$ . A photon is not a pellet flying through the field; it is one quantum of one of the field's own normal modes.

Step 5. Planck's law drops out — the first debt repaid. A mode in thermal equilibrium at temperature $T$ has mean occupation $\bar n=(e^{\hbar\omega/k_BT}-1)^{-1}$ (a two-line Boltzmann sum over the ladder). Multiply by the mode count per volume per frequency, $\omega^2/\pi^2c^3$ — the same counting that produced Rayleigh–Jeans — and the spectral energy density is

u(\omega)=\frac{\omega^2}{\pi^2c^3}\,\frac{\hbar\omega}{e^{\hbar\omega/k_BT}-1},

which is Planck's law: finite total energy, Rayleigh–Jeans recovered when $k_BT\gg\hbar\omega$ (set $\bar n\hbar\omega\to k_BT$ ), exponential cutoff where one quantum is expensive. The ultraviolet catastrophe was never a counting error — the mode count stands — it was equipartition assuming each spring can accept arbitrarily little energy. Verified: at $T=5800\,\mathrm K$ , $\lambda=500\,\mathrm{nm}$ : $\hbar\omega/k_BT=4.96$ , and the Rayleigh–Jeans/Planck ratio is $x^{-1}(e^x-1)|_{x=4.96}=28.6$ — the ledger's first number, recomputed. At $100\,\mathrm{nm}$ the ratio is $2.4\times10^9$ .

Step 4 bought more than Planck. Each oscillator's ground state carries $\tfrac12\hbar\omega$ — the field cannot be turned off, and that zero-point energy is not bookkeeping: confine the field between two conducting plates a distance $d$ apart and the allowed mode set changes, so the zero-point total changes with $d$ , producing a real, measured attraction — the Casimir pressure $P=\pi^2\hbar c/240d^4$ , which evaluates to $1.3\,\mathrm{mPa}$ at $d=1\,\mathrm{\mu m}$ : feeble, but measured to percent precision in modern torsion-balance and MEMS experiments. The vacuum of QED pushes on things. Keep that in mind for the Lamb shift below.

Emission — the correspondence made quantitative

An excited atom in a stationary state has, classically, no oscillating charge and therefore no reason to radiate; yet it does. The quantized field explains why (the vacuum's zero-point fluctuations tip it over), but remarkably, the rate was derived by Einstein in 1917 — before QED existed — from thermodynamic consistency alone, and the result is welded to this track's classical formulas.

Go deeper · Derivation B · 2 minSpontaneous emission by detailed balance — Einstein's 1917 route

Derivation BSpontaneous emission by detailed balance — Einstein's 1917 route

The goal is the spontaneous emission rate $A$ — the inverse lifetime of an excited atomic state — obtained without quantum field theory, from the requirement that atoms bathed in Planck radiation reach thermal equilibrium. The strategy is Einstein's: postulate the three elementary processes (absorption, stimulated emission, spontaneous emission), impose equilibrium against the Planck spectrum derived above, and watch consistency fix every coefficient. The step to watch is the last, where the classical Larmor physics of this track re-enters through the front door.

Step 1. The three processes. For a pair of levels $1,2$ ( $E_2-E_1=\hbar\omega$ , populations $N_1,N_2$ ) in a radiation bath of spectral density $u(\omega)$ : upward transitions at rate $N_1B_{12}u(\omega)$ ; downward at rate $N_2[A+B_{21}u(\omega)]$ — the $A$ term, emission with no radiation present, is the new postulate. Classically $A$ would be zero for a stationary state.

Step 2. Equilibrium forces the ratios. In equilibrium the rates balance and the populations are Boltzmann, $N_2/N_1=e^{-\hbar\omega/k_BT}$ (take nondegenerate levels; degeneracies decorate $B_{12}$ without changing the logic). Solving the balance for $u(\omega)$ :

u(\omega)=\frac{A/B_{21}}{(B_{12}/B_{21})\,e^{\hbar\omega/k_BT}-1} .

This must equal Planck's law at every temperature — one function matching another across a continuum of $T$ . That forces two identities:

B_{12}=B_{21}\equiv B,\qquad A=\frac{\hbar\omega^3}{\pi^2c^3}\,B .

Why this is astonishing: the ratio $A/B$ is fixed purely by the field's mode count $\omega^2/\pi^2c^3$ times $\hbar\omega$ — spontaneous emission is exactly "stimulated emission by one phantom photon per mode." The modern derivation says the same thing: the $\sqrt{n+1}$ in $a^\dagger|n\rangle$ splits into $n$ (stimulated) plus $1$ (vacuum).

Step 3. Anchor the scale classically. Quantum mechanics computes $B$ from the dipole matrix element, giving the textbook $A=\omega^3|\langle f|e\mathbf r|i\rangle|^2/3\pi\varepsilon_0\hbar c^3$ — the same $\omega^3$ , the same $\tfrac{2}{3\cdot4\pi\varepsilon_0 c^3}$ radiation constant as the Larmor formula, with the classical dipole amplitude replaced by a matrix element. The cleanest statement of the correspondence uses the oscillator strength $f$ , the dimensionless number this track's line-width article introduced: the quantum rate is the classical radiative damping rate $\Gamma_{\rm cl}=\omega^2\tau_e$ (with $\tau_e=2r_e/3c$ ) weighted by $f$ , with the level-degeneracy bookkeeping folded into $f$ 's standard conventions.

Step 4. Cash it on hydrogen. For Lyman-α ( $2p\to1s$ , $\lambda=121.57\,\mathrm{nm}$ ): $\Gamma_{\rm cl}=\omega^2\tau_e=1.50\times10^{9}\,\mathrm{s^{-1}}$ , and the transition's absorption oscillator strength is $f=0.4164$ . Then

A_{2p\to1s}\;=\;f\,\Gamma_{\rm cl}\;=\;0.4164\times1.50\times10^{9}\,\mathrm{s^{-1}}\;=\;6.27\times10^{8}\ \mathrm{s^{-1}} ,

a lifetime $\tau=1.60\,\mathrm{ns}$ — the measured value, identical to the full matrix-element calculation. Verified: both routes recomputed numerically — the matrix-element route ( $|\langle1s|\mathbf r|2p\rangle|^2=2^{15}a_0^2/3^{10}$ , summed over the $2p$ multiplet) gives $6.26\times10^8\,\mathrm{s^{-1}}$ ; the $f\Gamma_{\rm cl}$ route gives $6.27\times10^8$ ; they agree to 0.2%, and the Thomas–Reiche–Kuhn sum rule $\sum f=1$ (one electron) explains why the classical oscillator model of spectral lines worked at all: nature distributes exactly one classical electron's worth of response across each atom's spectrum.

This is the correspondence in its sharpest form: the classical formulas are conserved across the bridge. Larmor's power, the dipole pattern, the Lorentzian line shape of the Radiative Frequency Shift and Line Width article — all reappear in QED with matrix elements in place of amplitudes, exact in the limit $\bar n\gg1$ and reliable to within factors of $f$ even at $\bar n=0$ . The deep reason is the coherent state: the quantum state $|\alpha\rangle$ with Poissonian photon statistics whose field expectation value is a classical Maxwell wave, with relative quantum fluctuations $1/\sqrt{\bar n}$ . A radio transmitter at $\bar n\sim10^{20}$ photons per mode is describable by this track's classical theory to twenty digits of field amplitude — not as an approximation of convenience but as a controlled limit.

The handoff — what QED repays

Go deeper · Derivation C · 2 minThe alpha ladder and the fate of the classical divergences

Derivation CThe alpha ladder and the fate of the classical divergences

The goal is to derive the ladder of lengths that organizes the classical-quantum handoff — $r_e=\alpha^2a_0$ and $\lambda\!\!\!/_C=\alpha a_0$ — and then to read off, scale by scale, what becomes of each divergence this chapter exposed. The strategy is pure dimensional bookkeeping on three physical definitions, followed by the physics of which effect intervenes first as you shrink. Nothing here needs renormalization machinery; the point is precisely that the geography of the problem is visible classically.

Step 1. Three lengths from three definitions. The classical electron radius sets electrostatic self-energy equal to rest energy: $e^2/4\pi\varepsilon_0r_e=m_ec^2\Rightarrow r_e=e^2/4\pi\varepsilon_0m_ec^2=2.818\,\mathrm{fm}$ . The reduced Compton wavelength is where a photon's momentum kick matches the electron's rest scale: $\lambda\!\!\!/_C=\hbar/m_ec=386.2\,\mathrm{fm}$ . The Bohr radius balances Coulomb attraction against quantum kinetic pressure: $a_0=4\pi\varepsilon_0\hbar^2/m_ee^2=52.92\,\mathrm{pm}$ .

Step 2. Chain them with $\alpha$ . Divide:

\frac{r_e}{\lambda\!\!\!/_C}=\frac{e^2}{4\pi\varepsilon_0\hbar c}=\alpha, \qquad \frac{\lambda\!\!\!/_C}{a_0}=\frac{e^2}{4\pi\varepsilon_0\hbar c}=\alpha .

Both rungs are the same dimensionless number: $r_e=\alpha\,\lambda\!\!\!/_C=\alpha^2a_0$ , numerically $2.818\,\mathrm{fm}\times137.04=386.2\,\mathrm{fm}$ , $\times137.04$ again $=52.92\,\mathrm{pm}$ . The fine-structure constant is the conversion rate between the classical, relativistic-quantum, and atomic floors of the theory.

Step 3. Read the geography. At $a_0$ : atoms, chemistry, and the semiclassical radiation theory of Derivation B — quantum matter, effectively classical field. At $\lambda\!\!\!/_C$ : localizing the electron this tightly costs $\Delta p\,c\sim m_ec^2$ — pair creation energy — so the single-electron picture fails here, two orders of magnitude before the classical crisis scale. At $r_e$ : the classical self-energy catastrophe of the Classical Electron Theory article — a scale QED never permits the classical theory to reach, because the theory changed hands one rung up.

Step 4. The divergence ledger, repaid. Classically, $\delta m\,c^2\sim e^2/4\pi\varepsilon_0a\to\infty$ linearly as the radius $a\to0$ — brutal, and cutoff-dominated. In QED the electron is pointlike but dressed by virtual pairs beyond $\lambda\!\!\!/_C$ , which screen the singularity: the self-energy diverges only logarithmically, $\delta m/m\simeq(3\alpha/2\pi)\ln(\Lambda/m_ec^2)$ — for a cutoff at the Planck scale that is a $\sim17\%$ correction, not infinity-times anything. The renormalization move — absorb $\delta m$ into the measured mass — is the very manoeuvre the classical shell story of E9 pioneered; QED inherits the move and, thanks to the logarithm, gets away with it to any accuracy. The pathologies of this chapter resolve likewise: no runaways or preacceleration survive (quantum radiation reaction is photon recoil, discrete and causal — the Runaway Solutions article's closure), and the divergent classical line shift becomes a finite, calculable level shift.

Step 5. The triumph numbers. That finite level shift is the Lamb shift: the vacuum fluctuations of Derivation A jiggle the electron and lift $2s_{1/2}$ above $2p_{1/2}$ by the measured $1058\,\mathrm{MHz}$ (Lamb–Retherford 1947; Bethe's nonrelativistic estimate that same summer used exactly the classical-inspired mass subtraction of Step 4). And the electron's magnetic moment: Schwinger's 1948 one-loop QED gives $g-2=\alpha/\pi$ , i.e. $a_e=\alpha/2\pi=0.0011614$ , against today's measured $a_e=0.0011597$ — agreeing to $0.15\%$ at one loop, and to twelve significant figures when the full perturbation series meets the best experiments. It is the most precisely verified prediction in all of science, and it grew out of taming the very divergence this chapter met classically. Verified: all ladder numbers recomputed from CODATA constants ( $r_e=2.8179\,\mathrm{fm}$ , $\lambda\!\!\!/_C=386.16\,\mathrm{fm}$ , $a_0=52.918\,\mathrm{pm}$ ; ratios $137.036$ exactly); $\alpha/2\pi=0.001161$ ; Casimir pressure at $1\,\mathrm{\mu m}$ recomputed as $1.30\,\mathrm{mPa}$ .

Limiting cases

case	result	meaning
$\bar n=\|\alpha\|^2\gg1$ (coherent state)	classical Maxwell field	Field expectation values obey Maxwell exactly; relative fluctuations $1/\sqrt{\bar n}\to0$ . This track's entire subject, recovered as a limit.
$k_BT\gg\hbar\omega$	Planck $\to$ Rayleigh–Jeans	Equipartition is the many-quanta limit of the Planck ladder; the classical spectrum is right where quanta are cheap.
$\hbar\omega\ll E_{\rm particle}$	classical radiation formulas hold	Larmor/synchrotron emission with negligible recoil; quantum corrections enter as powers of $\hbar\omega/E$ .
$\hbar\omega\sim\Delta E_{\rm atom}$	one photon at a time	Discrete emission, selection rules, spontaneous decay — the classical continuum fails qualitatively.
$n=0$ (vacuum)	zero-point effects only	Spontaneous emission, Lamb shift, Casimir force: phenomena with no classical counterpart at any amplitude.
length scale $\to\lambda\!\!\!/_C$	pair creation; QED takes over	The classical theory's $r_e$ -scale crisis is never reached; the handoff happens one $\alpha$ -rung earlier.

Common trapThe photon is not a small classical bullet — and spontaneous emission is not a classical dipole radiating

Two habits from classical intuition fail across this bridge. First: a photon is one quantum of a delocalized field mode — it has no trajectory, no position operator in the classical sense, and a single photon can interfere with itself across both arms of an interferometer. Treating it as a tiny billiard ball predicts the wrong statistics everywhere it matters (bunching, antibunching, Hong–Ou–Mandel). Second: an excited stationary state has zero oscillating charge density, so the classical "its dipole oscillates and radiates" story is false at the starting line; emission is triggered by vacuum fluctuations, and the classical rate survives only because the vacuum acts like exactly one phantom photon per mode. The correspondence is quantitative — $P_{\rm classical}=\hbar\omega\,\Gamma$ with matched dipoles — but the mechanism is not the classical one. Trust the formulas across the bridge; do not trust the imagery.

Worked examples

1 · The photoelectric stopwatch (confidence builder)

Faint UV light of intensity $I=1\,\mathrm{mW/m^2}$ falls on potassium (work function $W=2.30\,\mathrm{eV}$ ). Classically, how long until the first electron can escape — and what is observed?

Step 1. Classical energy accumulation. A surface atom presents a collecting area of roughly its cross-section, $\sigma\approx\pi r_a^2$ with $r_a\approx1\,\mathrm{\text{Å}}$ : $\sigma=3.1\times10^{-20}\,\mathrm{m^2}$ . Power collected: $P=I\sigma=3.1\times10^{-23}\,\mathrm W$ .

Step 2. Time to bank the work function: $t=W/P=(2.30\times1.602\times10^{-19}\,\mathrm J)/(3.1\times10^{-23}\,\mathrm W)=1.2\times10^{4}\,\mathrm s\approx3.3$ hours. Even granting the atom a generous $2\,\mathrm{\text{Å}}$ radius only cuts it to about $49$ minutes.

Step 3. Observation: photocurrent begins in under a nanosecond, provided only $\hbar\omega>W$ — and no delay appears however faint the light. The classical estimate is wrong by thirteen orders of magnitude in time. One quantum of $300\,\mathrm{nm}$ light carries $\hbar\omega=4.1\,\mathrm{eV}>2.30\,\mathrm{eV}$ : the energy arrives pre-packaged, and the first photon absorbed anywhere liberates an electron there. Timing, not energetics, is the classical theory's unfixable failure here.

2 · One line, two languages — the lifetime of Lyman-alpha (exam level)

Compute the $2p\to1s$ decay rate of hydrogen from the classical anchor, and check the photon energetics.

Step 1. Classical damping rate at the transition frequency. $\lambda=121.57\,\mathrm{nm}\Rightarrow\omega=2\pi c/\lambda=1.549\times10^{16}\,\mathrm{s^{-1}}$ . With $\tau_e=2r_e/3c=6.27\times10^{-24}\,\mathrm s$ : $\Gamma_{\rm cl}=\omega^2\tau_e=1.50\times10^{9}\,\mathrm{s^{-1}}$ .

Step 2. Weight by the oscillator strength. Lyman-α carries $f=0.4164$ of the atom's unit sum-rule budget: $A=f\,\Gamma_{\rm cl}=6.27\times10^{8}\,\mathrm{s^{-1}}$ , so $\tau=1/A=\mathbf{1.60\,ns}$ — the measured lifetime.

Step 3. Sanity-check the quanta. Each decay emits one photon of $\hbar\omega=10.2\,\mathrm{eV}$ ; the emitted power of a single decaying atom, $\hbar\omega A=1.0\times10^{-9}\,\mathrm W$ momentarily, matches the classical dipole power for the corresponding classical amplitude — the $P_{\rm classical}=\hbar\omega\,\Gamma$ correspondence in numbers.

Step 4. Where the width went. The same $A$ is the line's natural width: $\Delta\nu=A/2\pi=100\,\mathrm{MHz}$ — a Lorentzian exactly as the classical Radiative Frequency Shift and Line Width article derived, with $\Gamma_{\rm cl}\to f\Gamma_{\rm cl}$ the only edit. The classical line-shape theory survives the bridge intact; only its amplitude bookkeeping is re-priced, line by line, by $f$ — and the sum rule $\sum f=1$ guarantees the books balance atom by atom.

Going further

References. Dirac, Proc. R. Soc. A 114, 243 (1927) — the founding paper · Sakurai, Advanced Quantum Mechanics, Ch. 2 · Cohen-Tannoudji, Dupont-Roc & Grynberg, Photons and Atoms · Jackson, Classical Electrodynamics, 3rd ed., Ch. 9 and §16 (classical radiation and its limits) · Griffiths, Introduction to Quantum Mechanics, time-dependent perturbation theory chapter · Milonni, The Quantum Vacuum.

Provenance. Mode expansion, commutators, and Hamiltonian follow standard canonical field quantization (Sakurai; Cohen-Tannoudji). The Einstein A-coefficient $A=\omega^3|\mathbf d|^2/(3\pi\varepsilon_0\hbar c^3)$ is the textbook result; a verifier should reproduce it from Fermi's golden rule and confirm the $2p\to1s$ hydrogen lifetime $\approx1.6\,\mathrm{ns}$ . Classical dipole power from Jackson Ch. 9 / Griffiths Ch. 11. Coherent-state Poisson statistics from Glauber's standard treatment. No nonstandard numbers used. Lane addendum: all ledger and ladder numbers (Rayleigh–Jeans ratios, photoelectric times, Compton shift, $r_e/\lambda\!\!\!/_C/a_0$ chain, Casimir pressure, $f\Gamma_{\rm cl}$ vs matrix-element rates, $\alpha/2\pi$ ) were recomputed from CODATA constants for this edition; the hydrogen collapse time is recalled by name from the E9 Larmor article, and the Compton validity ledger from the Thomson Scattering article.

Prerequisites. Multipole Expansion of Radiation Fields · Electric Dipole Radiation: Exact Fields and Zone Structure · Larmor Formula from the Liénard–Wiechert Fields · Liénard–Wiechert Potentials: Covariant Formulation · Vector Spherical Harmonics.

Leads to. Angular Distribution of Radiation from Accelerated Charges · Synchrotron Radiation · Magnetic Dipole and Electric Quadrupole Radiation — and, beyond this track: quantum optics, and QED proper (Feynman rules, loop corrections, renormalization).

If classical electrodynamics fails at the quantum scale, why do antennas, radar, and all of electrical engineering work perfectly with Maxwell alone?

Count the quanta. A 1 kW transmitter at 1 GHz emits $\sim10^{27}$ photons per second into modes whose occupation numbers are astronomical; the field state is a coherent state with $\bar n$ so large that quantum fluctuations sit twenty digits below the signal. Classical electrodynamics is the exact dynamics of coherent-state expectation values — not an approximation that happens to be good, but a controlled limit with corrections of order $1/\sqrt{\bar n}$ . Engineering lives in that limit. The failures in the ledger all involve either one quantum at a time (photoelectric, Compton, atomic decay) or the vacuum itself (Lamb, Casimir), where $\bar n\sim1$ or $0$ .

Is the photon just a normal mode of the field, or a particle? Which picture is right?

Both are precise if kept in their lanes. A photon is one quantum of one field mode — its definition, which explains why photons are identical, Bose-statistical, and non-conserved in number. "Particle" behavior (clicks in detectors, Compton kinematics with $E=\hbar\omega$ , $\mathbf p=\hbar\mathbf k$ ) is what quantized exchange of energy and momentum looks like when a mode interacts with localized matter. What the photon is not: a small classical object with a trajectory. When you want intuition, think "quantum of a mode"; when you want kinematics, use the four-momentum.

Where does the zero-point energy of every mode come from, and is it physically real given that it seems infinite?

Uncertainty applied to each oscillator: $Q$ and $P$ cannot both vanish, so the ground state carries $\tfrac12\hbar\omega$ . Summed over all modes the total is formally infinite — but only energy differences are measurable, and those are finite and measured: reshape the mode set with conducting boundaries and the zero-point total shifts, giving the Casimir pressure ( $1.3\,\mathrm{mPa}$ at $1\,\mathrm{\mu m}$ , verified to percent level); let the vacuum jiggle a bound electron and levels shift (Lamb, $1058$ MHz). QED handles the formal infinity the same way it handles the mass divergence: subtract the unobservable constant, keep the measurable differences.

The Einstein A coefficient came from thermodynamics in 1917, a decade before QED. Why did that argument work so early?

Detailed balance is merciless: the rates must reproduce Planck's law at every temperature, and that over-determination fixes $B_{12}=B_{21}$ and $A/B=\hbar\omega^3/\pi^2c^3$ with no field theory at all — only Planck's spectrum and Boltzmann statistics. What Einstein could not supply was the absolute scale (the matrix element inside $B$ ) and the mechanism; Dirac's 1927 quantization delivered both. Thermodynamic consistency sees structure early; that is its trade.

Does anything classical survive at exactly zero photons, or is the vacuum entirely quantum territory?

The equations survive: the operator fields obey Maxwell's equations exactly, and a single photon's amplitude propagates by the classical wave equation — which is why classical optics (interference, diffraction, polarization) predicts single-photon interference patterns perfectly. What has no classical counterpart is the state: correlations and fluctuations of the vacuum (spontaneous emission triggers, Casimir stresses, the Lamb jitter). A useful slogan with precise content: Maxwell survives as the theory's kinematics; the quantum enters through what states are allowed to occupy it.

How exactly does QED cure the runaway and preacceleration pathologies of the Abraham-Lorentz force?

By changing what "radiation reaction" is. Classically, the self-force is a continuous field acting back on a point charge, and demanding that of a third-derivative equation of motion produces runaways or acausal preacceleration — the Runaway Solutions article's subject. In QED, radiation reaction is the recoil from emitting discrete photons: a stochastic, strictly causal sequence of kicks, each conserving energy-momentum at a vertex. In the classical limit the average recoil reproduces the Landau–Lifshitz force (the legal effective form), while the pathological solutions correspond to nothing in the quantum theory — they were artifacts of extrapolating a point-charge self-interaction below the Compton scale, where the classical theory had already lost jurisdiction. The 2018 laser-electron experiments now probe exactly the regime where the quantum, stochastic character shows.

Is the fine-structure constant's value 1/137 explained by QED, and what would change if it were different?

QED takes $\alpha$ as input — no accepted theory derives its value, and its origin is among the sharpest open questions in physics ( $\alpha$ even runs with energy scale, reaching $\approx1/127$ at the Z mass). But its role is fully understood: it is the probability weight of one photon exchange, the expansion parameter of QED perturbation theory, and the spacing of this article's length ladder. Shift it modestly and chemistry re-prices (fine structure $\propto\alpha^2$ , stellar burning exponentially sensitive); make it of order 1 and the perturbative ladder collapses along with this article's separation of scales.

After this track, what should I read to actually do QED — and what will feel familiar when I get there?

Time-dependent perturbation theory and the quantized-field chapters of Sakurai or Cohen-Tannoudji (Photons and Atoms), then a QFT text for Feynman rules, loops, and renormalization. Familiar on arrival: gauge fixing (E9/E10 were the classical half of that story), retarded Green functions (the covariant wave equation article is the tree-level photon propagator in disguise), the multipole ladder (it becomes selection rules), and every renormalization argument (the classical electron chapter was the dress rehearsal). Genuinely new: antiparticles, loop integrals, running couplings. You are most of the way there.