Bremsstrahlung

Aim an electron beam at a metal plate and X-rays come out — the oldest experiment of modern physics, run within weeks of Röntgen's 1895 discovery, and still the working heart of every hospital X-ray tube. The mechanism needs nothing beyond this chapter's machinery: an electron swinging past a nucleus is accelerated by the Coulomb field, and an accelerated charge radiates. The radiation is called bremsstrahlung — braking radiation — and its distinctive personality comes entirely from the statistics of Coulomb encounters: how hard the kick is (impact parameter), how brief (collision time), and how many nuclei per centimeter stand in the way. The spectral signature is the opposite of a line: a single sudden kick radiates a flat spectrum, white up to a cutoff set by the inverse collision time — the sharpest possible application of the Parseval spectral ledger built by Larmor Formula from the Liénard–Wiechert Fields. Summing kicks over all impact parameters multiplies the white plateau by the celebrated Coulomb logarithm, and the bookkeeping delivers the two scalings that organize everything: radiated energy $\propto Z^2$ (heavy targets glow harder) and $\propto 1/m^2$ (only light projectiles radiate — an electron outshines a proton by six and a half orders of magnitude). The consequences sprawl. The flat spectrum with its quantum Duane–Hunt edge is the X-ray tube's continuum; the $Z^2$ energy loss growing linearly with energy defines the radiation length and the electromagnetic showers of every calorimeter; and a Maxwellian plasma's summed kicks — thermal bremsstrahlung — make galaxy clusters glow in X-rays, turning the flat spectrum's exponential cutoff into a cosmic thermometer. One collision, integrated three ways.

Key takeaways

A charge deflected in a nuclear Coulomb field radiates a pulse; the pulse's Fourier content is flat up to $\omega \sim 1/\tau \sim v/b$ — the universal white spectrum of a sudden kick, with $dW/d\omega = q^2|\Delta\mathbf{v}|^2/6\pi^2\varepsilon_0c^3$ at low frequency.
Radiated energy scales as $Z^2/m^2$ : double the nuclear charge, four times the light; swap the electron for a proton, $3.4\times10^6$ times darker. Bremsstrahlung is an electron phenomenon.
Integrating over impact parameters turns the plateau into the Kramers spectrum $\propto Z^2 n_i \ln(b_{\max}/b_{\min})/m^2v^2$ — the Coulomb logarithm's cutoffs are physics (screening above, quantum de Broglie or strong-deflection below).
Quantum mechanics caps the spectrum hard: no photon above the electron's kinetic energy — the Duane–Hunt edge $\lambda_{\min} = hc/eV$ , target-independent.
At high energy, radiative loss grows $\propto E$ and overtakes ionization: $E(x) = E_0e^{-x/X_0}$ defines the radiation length ( $X_0 \approx 0.56$ cm in lead) and electromagnetic showers.
A thermal plasma radiates free–free emission $\propto Z^2n_en_iT^{-1/2}e^{-\hbar\omega/k_BT}$ : galaxy clusters are 10-keV bremsstrahlung lamps, and the cutoff measures their temperature.
Key idea: bremsstrahlung is the Larmor formula driven by Coulomb statistics — a white kick spectrum, weighted $Z^2/m^2$ , summed over impact parameters into a Coulomb logarithm, cut off by quantum energy conservation.

The core insight

A charge deflected in a Coulomb field radiates because it accelerates; the radiated energy per encounter is governed by the magnitude and duration of that acceleration, giving a roughly white spectrum cut off at $\omega \sim v/b$ and total emission scaling as $Z^2/m^2$ .

Where this lives today

Bremsstrahlung is infrastructure. Medical imaging and radiotherapy run on tube and linac continua shaped by the $Z^2$ and Duane–Hunt rules; radiation-protection codes tabulate exactly this article's forward-beamed lobes when they size linac bunker walls. In particle physics the radiation length is the meter-stick of detector design — tracker budgets are quoted in percent of $X_0$ , electromagnetic calorimeters in tens of $X_0$ — and the same $E/X_0$ law governs the cosmic-ray air showers that extensive-array observatories read from the ground. Muon tomography of volcanoes and cargo works because the $(m_e/m_\mu)^2$ suppression lets muons survive where electrons shower. In astrophysics, thermal bremsstrahlung is the standard candle of hot diffuse plasma: cluster cosmology (gas fractions, scaling relations, the dark-matter ledger of worked example 2), the warm-hot intergalactic medium, H II region radio continua, and accretion-flow X-rays all lean on the $n_e^2\sqrt{T}e^{-\hbar\omega/kT}$ form. And at the exotic frontier the same soft-photon theorem that opened this article — the flat $|\Delta v|^2$ plateau of any sudden kick — reappears as the low-energy limit of QED radiative corrections and as "beamstrahlung" in collider design: whenever charges are kicked, the white spectrum follows, and the quantum theory keeps the classical bookkeeping as its zeroth order (the bridge articles of this chapter's close, by name).

Origins · 1 minHendrik Anthony Kramers and how this was discovered

Hendrik Anthony Kramers

1894–1952 · Copenhagen and Leiden — who priced the continuum before the quantum theory could

Kramers appears in this track once already, silently: the Kramers–Kronig machinery of the absorption-and-dispersion article carries half his name. This article honors the earlier band: his 1923 Philosophical Magazine paper on continuous X-ray emission, written at Bohr's institute in the awkward twilight between classical electrodynamics and the still-unbuilt quantum mechanics. Kramers computed bremsstrahlung exactly as this article does — classical Coulomb orbits, Fourier-analyzed pulses, impact-parameter sweeps — and then, via the correspondence principle, massaged the classical spectrum into quantum emission rates, producing the "Kramers formula" for X-ray continua and, in its stellar application, the Kramers opacity law $\kappa \propto \rho T^{-3.5}$ that governed stellar-structure theory for half a century. When the real quantum theory arrived (Bethe–Heitler, 1934), its result was the Kramers spectrum times a slowly varying correction — and that correction has carried the name Gaunt factor ever since John Gaunt's 1930 quantum evaluation of Kramers's formula. Pricing the right answer with the wrong theory, to within a factor of order one, is the correspondence principle's finest habit; Kramers was its most disciplined practitioner.

History

1895

— Röntgen discovers X-rays from a discharge tube — bremsstrahlung observed before it is named

1909

— Sommerfeld computes the radiation of a decelerated charge, coining Bremsstrahlung

1915

— Duane and Hunt measure the sharp short-wavelength edge: the continuum stops at hν = eV

1923

— Kramers derives the classical continuum spectrum and its stellar opacity law

1930

— Gaunt evaluates the quantum correction to Kramers — the Gaunt factor is born

1934

— Bethe and Heitler publish the relativistic quantum cross-section; radiation length follows

1937

— shower theory (Bhabha–Heitler, Carlson–Oppenheimer) explains cosmic-ray cascades

1966

— X-ray astronomy finds hot cluster gas: thermal bremsstrahlung becomes a cosmological probe

The physical picture

Watch one encounter. An electron at speed $v$ passes a nucleus of charge $Ze$ at impact parameter $b$ . The Coulomb force yanks it sideways for a time $\tau \sim b/v$ — the interval during which the nucleus is within $\sim b$ of closest approach — and during the yank the electron's acceleration peaks at $a \sim Ze^2/4\pi\varepsilon_0mb^2$ . The Larmor machinery (the E9 flagship, recalled by name) says this transverse jolt broadcasts a pulse of radiation. Closer pass: stronger, briefer, bluer. Farther pass: weaker, longer, redder. The whole subject is this one pulse, replayed over the statistical ensemble of passes a beam makes through matter. The frequency story is the deepest part, and it costs nothing. A pulse of duration $\tau$ has a Fourier transform that is constant below $\omega \sim 1/\tau$ — at low frequency the transform integrates the whole acceleration history and sees only the net velocity kick $\Delta\mathbf{v}$ , blind to the pulse's shape. So every collision, whatever its geometry, radiates the same white spectrum at low frequency, with amplitude set by $|\Delta\mathbf{v}|^2$ alone. Sharp lines are the fingerprint of bound oscillators; flat continua are the fingerprint of collisions. An X-ray tube emits both at once — the bremsstrahlung continuum from the deflections, plus the element's characteristic lines from knocked-out bound electrons — and telling them apart is diagnostic bread and butter. The amplitude story is one fraction: $a = F/m$ . The force depends on the target ( $\propto Z$ ), the acceleration on the projectile ( $\propto 1/m$ ), and the power on the square of both — $Z^2/m^2$ . That single fraction decides that tungsten anodes beat copper ones fourfold per unit charge squared, that protons pass through matter radiating essentially nothing, and that muons — 207 electron masses — penetrate kilometers of rock where electrons shower and die in centimeters.

Figure

One Coulomb pass: the transverse acceleration pulse of width b over v, and its Fourier ledger — flat up to the inverse pulse width, exponentially dark beyond

The white spectrum of a sudden kick

The tool is imported, not built: Larmor Formula from the Liénard–Wiechert Fields established the Parseval ledger — the radiated energy per unit frequency of any nonrelativistic acceleration history $a(t)$ is set by its Fourier transform $\tilde{\mathbf{a}}(\omega) = \int a(t)e^{i\omega t}dt$ :

\frac{dW}{d\omega} = \frac{q^2\,|\tilde{\mathbf{a}}(\omega)|^2}{6\pi^2\varepsilon_0 c^3}. \qquad (1)

where: the $6\pi^2$ is the Larmor $6\pi$ with the extra $\pi$ from folding negative frequencies into the physical range $\omega > 0$ . That article built the ledger; this one spends it on the most important special case in the theory: the collision. For any pulse of duration $\tau$ , the transform at $\omega \ll 1/\tau$ forgets everything but the area under the pulse — the net velocity change:

\tilde{\mathbf{a}}(\omega) \xrightarrow{\omega\tau \ll 1} \int \mathbf{a}(t)\,dt = \Delta\mathbf{v} \qquad\Longrightarrow\qquad \boxed{\;\frac{dW}{d\omega}\bigg|_{\omega \ll 1/\tau} = \frac{q^2\,|\Delta\mathbf{v}|^2}{6\pi^2\varepsilon_0 c^3}\;} \qquad (2)

where: $\Delta\mathbf{v}$ — the total kick of the encounter. Equation (2) is the low-frequency theorem: flat, shape-independent, universal for every collision in physics (its quantum descendants govern soft-photon emission in QED). All the collision's individuality — its duration, the shape of $a(t)$ — lives in where the plateau ends, at $\omega \sim 1/\tau$ , beyond which $\tilde{a}$ and the spectrum die exponentially. For the Coulomb pulse the exact shape is computable and plotted in the hero figure: $|\tilde{a}|^2 \propto [xK_1(x)]^2$ with $x = \omega b/v$ — flat to within 4% below $x = 0.3$ , down a hundredfold by $x = 4$ (computed for this edition).

Derivation AOne Coulomb pass — the radiated energy and its Z squared over m squared scaling

The goal is the energy radiated in a single small-angle encounter — projectile charge $-e$ , mass $m$ , speed $v$ ; nucleus $Ze$ fixed; impact parameter $b$ — done exactly in the impulse approximation: $W \propto Z^2e^6/m^2vb^3$ , the number whose scalings organize the entire subject. The strategy: straight-line trajectory, transverse Coulomb force, Larmor power integrated through the pass. The watch-out is validity: the result assumes the deflection is small and the motion nonrelativistic; both edges are priced afterward.

Step 1. Trajectory and force. On the straight path $\mathbf{r}(t) = (vt, b, 0)$ , the distance to the nucleus is $r(t) = \sqrt{v^2t^2 + b^2}$ and the transverse force component is

F_\perp(t) = \frac{Ze^2}{4\pi\varepsilon_0}\,\frac{b}{(v^2t^2 + b^2)^{3/2}},

the longitudinal component integrating to zero by the symmetry of the pass (it decelerates on approach, accelerates on retreat). Why keep only the transverse kick: at lowest order in the deflection angle the longitudinal history cancels in $\Delta\mathbf{v}$ but not instant-by-instant in $a^2$ — its contribution to the radiated energy is smaller by the deflection angle squared, consistently negligible in the impulse regime.

Step 2. Integrate Larmor through the pass. With $a(t) = F_\perp(t)/m$ and $P(t) = e^2a^2(t)/6\pi\varepsilon_0c^3$ :

W = \int_{-\infty}^{\infty} P\,dt = \frac{e^2}{6\pi\varepsilon_0 c^3 m^2}\left(\frac{Ze^2}{4\pi\varepsilon_0}\right)^{2} b^2 \int_{-\infty}^{\infty}\frac{dt}{(v^2t^2+b^2)^{3}}.

The integral is elementary (substitute $vt = b\tan\theta$ ): $\int dt/(v^2t^2+b^2)^3 = 3\pi/8vb^5$ — recomputed by quadrature for this edition, exact to machine precision.

Step 3. Assemble. Collecting the constants ( $e^2/6\pi\varepsilon_0 = \tfrac{2}{3}\,e^2/4\pi\varepsilon_0$ , then everything into $(4\pi\varepsilon_0)^3$ ):

W = \frac{e^2}{6\pi\varepsilon_0 c^3 m^2}\left(\frac{Ze^2}{4\pi\varepsilon_0}\right)^{2}\frac{3\pi}{8\,v\,b^3} = \boxed{\;\frac{\pi}{4}\,\frac{Z^2e^6}{(4\pi\varepsilon_0)^3\,c^3\,m^2\,v\,b^3}\;}

— the coefficient audited numerically against the direct Larmor integral (agreement to 15 digits; the source text's boxed coefficient $\pi/12$ with a stray extra $\varepsilon_0$ is dimensionally broken — see the provenance addendum).

Step 4. Read the scalings. $Z^2$ : the force is $\propto Z$ , the power its square — high- $Z$ anodes and high- $Z$ shielding both follow. $1/m^2$ : same force, lighter particle, bigger jolt — a proton radiates $(m_e/m_p)^2 = 3.0\times10^{-7}$ as much as an electron in the same encounter, a muon $(m_e/m_\mu)^2 = 2.3\times10^{-5}$ . $1/b^3$ : close passes dominate the energy (though not the soft-photon count — the impact-parameter sweep of Derivation B redistributes the emphasis logarithmically). $1/v$ : slower passage means a longer kick and more radiated energy per encounter — but a lower frequency ceiling $v/b$ . Verified: time integral $3\pi/8vb^5$ by scipy quadrature (exact); worked number $W = 0.090$ eV for a 100 keV electron passing tungsten at $b = 10$ pm, with $\Delta v/v = 0.14$ confirming the impulse regime; scalings against Jackson §15.1 and the source text's boxed form.

From one pass to a beam in matter — the Coulomb logarithm

A beam of electrons in a target of nuclear density $n_i$ does not choose one $b$ ; it samples all of them, with the ring $2\pi b\,db$ per unit path length weighting distant passes heavily. The single-pass plateau (2) carries $|\Delta\mathbf{v}|^2 \propto 1/b^2$ (the impulse $\Delta v = 2Ze^2/4\pi\varepsilon_0mvb$ , the same integral as Step 1), and the competition between $1/b^2$ per event and $b\,db$ events is the signature arithmetic of Coulomb physics:

\frac{d^2W}{d\omega\,dx} = n_i\int_{b_{\min}}^{b_{\max}} \frac{dW}{d\omega}(b)\;2\pi b\,db \;\propto\; \int \frac{db}{b} = \ln\frac{b_{\max}}{b_{\min}}. \qquad (3)

where: the logarithm is exactly the structure that plasma physicists meet in collision rates and particle physicists in multiple scattering — scale-free $1/b$ physics, regulated at both ends by cutoffs that are themselves physics, not bookkeeping. The next derivation supplies them.

Go deeper · Derivation B · 2 minThe Kramers spectrum and where the logarithm's cutoffs come from

Derivation BThe Kramers spectrum and where the logarithm's cutoffs come from

The goal is the classical bremsstrahlung spectrum per unit path — the Kramers form — with its Coulomb logarithm derived rather than gestured at, and the cutoffs identified with their physical owners: frequency itself above, quantum mechanics or strong deflection below. The strategy: insert the kick spectrum into the impact-parameter sweep, watch the logarithm appear, then interrogate both ends.

Step 1. The kick per pass. Integrating the transverse force of Derivation A over the pass: $\Delta v(b) = \tfrac{1}{m}\int F_\perp dt = \dfrac{2Ze^2}{4\pi\varepsilon_0\,m v b}$ — the classic factor-of-2 Coulomb impulse ( $\int dt\,b/(v^2t^2+b^2)^{3/2} = 2/bv$ ).

Step 2. Sweep the rings. Insert (2) with this $\Delta v$ into (3):

\frac{d^2W}{d\omega\,dx} = n_i\,\frac{e^2}{6\pi^2\varepsilon_0c^3}\left(\frac{2Ze^2}{4\pi\varepsilon_0 mv}\right)^{2} 2\pi\int_{b_{\min}}^{b_{\max}}\frac{db}{b} = \frac{16}{3}\,\frac{Z^2e^6\,n_i}{(4\pi\varepsilon_0)^3\,c^3m^2v^2}\,\ln\frac{b_{\max}}{b_{\min}},

the Kramers form — prefactor algebra verified symbolically for this edition ( $\tfrac{e^2}{6\pi^2\varepsilon_0}\cdot\tfrac{4}{(4\pi\varepsilon_0)^2}\cdot2\pi = \tfrac{16}{3}\cdot\tfrac{1}{(4\pi\varepsilon_0)^3}$ exactly).

Step 3. The upper cutoff belongs to the frequency. A pass at impact parameter $b$ only populates frequencies below its own ceiling $v/b$ ; asking for spectral density at $\omega$ silently discards passes with $b \gtrsim v/\omega$ — their pulses are too slow to contain that frequency. So $b_{\max} = v/\omega$ : soft photons collect contributions from enormous distances, hard photons only from close passes, and the Coulomb log grows toward low frequency (the spectrum is white only up to its slowly varying logarithm). In dense targets, atomic screening imposes its own ceiling at the Thomas–Fermi radius; whichever is smaller wins.

Step 4. The lower cutoff is a confession. Classically, the impulse approximation fails when the deflection is large: $\Delta v \sim v$ at $b_{\min}^{\rm cl} = 2Ze^2/4\pi\varepsilon_0mv^2$ (twice the closest-approach scale). Quantum mechanically, the electron is not localizable better than its de Broglie wavelength: $b_{\min}^{\rm qm} = \hbar/mv$ . Nature uses the larger; the ratio $b_{\min}^{\rm cl}/b_{\min}^{\rm qm} = 2Z\alpha\,c/v$ decides the regime — slow projectiles on high- $Z$ nuclei are classical, fast ones quantum. For a 100 keV electron on tungsten, $2Z\alpha c/v = 1.97$ : the tube that hangs in every clinic operates almost exactly at the classical–quantum crossover, and honest work there uses the quantum (Bethe–Heitler) cross-section with the classical picture as scaffolding. The Gaunt factor is precisely the community's name for "the exact quantum logarithm divided by a reference classical one" — order unity, tabulated, and the last word on every cutoff dispute (deferred by name to the quantum treatment: this article's classical honesty budget ends at the logarithm). Verified: prefactor identity symbolic; $2Z\alpha c/v = 1.97$ and $b_{\min}^{\rm cl} \approx 1.4$ pm vs $b_{\min}^{\rm qm} = 0.70$ pm recomputed for 100 keV on tungsten; the $b_{\max} = v/\omega$ frequency ceiling against the exact $[xK_1(x)]^2$ pulse spectrum (cutoff at $x \approx 1$ , computed).

The integrated bills — stopping, showers, and hot plasmas

Go deeper · Derivation C · 2 minRadiative stopping, the radiation length, and the thermal bremsstrahlung of a plasma

Derivation CRadiative stopping, the radiation length, and the thermal bremsstrahlung of a plasma

The goal is the two frequency-integrated consequences that make bremsstrahlung consequential: the high-energy stopping law $-dE/dx = E/X_0$ that defines the radiation length and electromagnetic showers, and the thermal (free–free) emissivity of a Maxwellian plasma with its exponential thermometer. Both are assemblies of the Kramers machinery — the first integrated over frequency at fixed high energy, the second averaged over a thermal velocity distribution — and both are stated with their quantum honesty attached.

Step 1. Integrate the spectrum at high energy. The plateau (3) runs up to the quantum ceiling $\hbar\omega_{\max} = E_{\rm kin}$ (Duane–Hunt: one photon cannot carry more than the projectile's kinetic energy). The radiated energy per unit path is then plateau × bandwidth: $-dE/dx \sim (d^2W/d\omega dx)\cdot(E/\hbar)$ . Every factor in (3) except the bandwidth is energy-independent in the relativistic regime (with $v \to c$ and the relativistic beaming of Angular Distribution of Radiation from Accelerated Charges folding the emission forward without changing the total), so

-\frac{dE}{dx}\bigg|_{\rm rad} \propto Z^2 n_a\,E \qquad\Longrightarrow\qquad E(x) = E_0\,e^{-x/X_0}, \qquad \frac{1}{X_0} = 4\alpha\,r_e^2\,n_a\,Z(Z+1)\ln\!\frac{183}{Z^{1/3}},

the loss proportional to the energy itself — the defining feature of a radiative process, against ionization's near-constant bill. The boxed $1/X_0$ is the Bethe–Heitler result quoted with its full quantum-screening pedigree (the $Z(Z+1)$ counts nuclear plus electron targets; the $183/Z^{1/3}$ is Thomas–Fermi screening fixing $b_{\max}$ ): classical physics supplies every scaling in it, quantum mechanics the exact logarithm. Benchmark values (Particle Data Group, quoted): $X_0 = 0.5612$ cm in lead, 36.1 cm in water, 304 m in air.

Step 2. The consequences of $\propto E$ . Where radiative loss crosses ionization loss defines the critical energy $E_c \approx 610\ \mathrm{MeV}/(Z + 1.24)$ — 7.3 MeV in lead, 87 MeV in air. Above $E_c$ , an electron's fate is geometric: radiate a photon within $\sim X_0$ , the photon pair-produces within $\sim \tfrac{9}{7}X_0$ , both offspring repeat — an electromagnetic shower doubling every radiation length until the population's mean energy falls to $E_c$ , whence $\sim\ln_2(E_0/E_c)$ generations. Calorimeter design is this arithmetic cast in tungsten and scintillator.

Step 3. The thermal average. In an ionized plasma the "beam" is the electrons' own thermal motion. Averaging the Kramers emission over a Maxwellian at temperature $T$ : the emissivity per unit frequency picks up (i) $n_en_i Z^2$ from encounter counting, (ii) one power of $\langle 1/v\rangle$ -weighted flux against the $1/v^2$ per-pass spectrum, netting $T^{-1/2}$ , and (iii) the Boltzmann fact that only electrons with $\tfrac12mv^2 > \hbar\omega$ can emit $\hbar\omega$ :

\varepsilon^{ff}_\omega \;\propto\; Z^2\,n_e\,n_i\,T^{-1/2}\,e^{-\hbar\omega/k_BT}\,\bar{g}_{ff},

with $\bar{g}_{ff}$ the velocity-averaged Gaunt factor ( $\sim1$ –1.5). Integrated over frequency the bandwidth $\propto k_BT$ flips the scaling to $\varepsilon^{ff} \propto Z^2n_en_i\sqrt{T}$ . The exponential is the thermometer: fit the X-ray spectrum's cutoff and read $k_BT$ directly; the normalization then hands you $n_e^2 \times$ volume — the density of matter you will never sample in a lab. Verified: $E_c$ (Pb) $= 610/83.24 = 7.3$ MeV; shower generation count $\ln_2(E_0/E_c)$ dimensional; $T^{-1/2}$ vs $\sqrt{T}$ bookkeeping against Rybicki & Lightman Ch. 5 (their eq. 5.14b/5.15b); cluster worked example below recomputed end-to-end. Quantum honesty: everything past the classical scalings — exact logs, screening constants, $\bar{g}_{ff}$ — is quoted Bethe–Heitler/tabulated physics, flagged as such.

Figure

The impact-parameter sweep assembled — per-pass white plateaus with cutoffs at v over b stack into the Kramers spectrum, whose Coulomb logarithm grows toward soft frequencies; the quantum Duane–Hunt edge terminates everything

Limiting cases

case	result	meaning
$\omega \ll v/b$ , one pass	$dW/d\omega = q^2\|\Delta v\|^2/6\pi^2\varepsilon_0c^3$ , flat	the low-frequency theorem — shape-blind white spectrum
$\omega \gtrsim v/b$	exponential cutoff, exact shape $[xK_1(x)]^2$	the pulse is too slow to carry the frequency
swept over $b$	$\propto Z^2n_i\ln(b_{\max}/b_{\min})/m^2v^2$	Kramers spectrum; log grows softward
$\hbar\omega \to E_{\rm kin}$	hard zero (Duane–Hunt)	quantum conservation truncates the classical white plateau
$E \gg E_c$	$-dE/dx = E/X_0$ , showers	radiative regime; $X_0 \propto 1/Z^2n_a$
Maxwellian plasma	$\varepsilon_\omega \propto Z^2n_en_iT^{-1/2}e^{-\hbar\omega/k_BT}$	free–free glow; the cutoff is a thermometer
projectile mass $m \to m_p$	suppression $(m_e/m_p)^2 = 3\times10^{-7}$	bremsstrahlung is an electron phenomenon

Common trapThe continuum is not the lines, the edge is not classical, and protons do not radiate their share

Three standing confusions. First, an X-ray tube's output superposes two unrelated spectra: the bremsstrahlung continuum (free electrons deflecting — this article) and the characteristic lines (bound-state transitions after ionization — atomic physics); attributing the lines to "braking" or the continuum to "transitions" scrambles both subjects. Second, the sharp short-wavelength edge $\lambda_{\min} = hc/eV$ is pure quantum bookkeeping — a photon cannot outcarry the electron — and no classical argument produces it: classically the white plateau would extend to the inverse closest-approach time and fade smoothly. The edge's target-independence (tungsten and copper anodes share it at fixed voltage) is the tell that it belongs to $\hbar$ , not to $Z$ . Third, the $1/m^2$ is unforgiving: quoting bremsstrahlung numbers for protons or ions without the $(m_e/m)^2$ suppression inflates them by six to seven orders of magnitude — the reason "radiative stopping" is absent from proton-therapy dose planning while it dominates electron-beam design.

Worked examples

1 · The 100 kV tungsten tube on the wall (confidence builder)

Setup. A clinical X-ray tube: electrons accelerated through $V = 100$ kV strike a tungsten anode ( $Z = 74$ ). Anatomize its output with this article's tools.

Step 1. Kinematics. $\gamma = 1 + eV/m_ec^2 = 1.196$ , $\beta = 0.548$ — mildly relativistic; the classical machinery applies with modest beaming corrections.

Step 2. The edge. Duane–Hunt: $\lambda_{\min} = hc/eV = \mathbf{12.4}$ pm ( $\hbar\omega_{\max} = 100$ keV exactly) — measured to be independent of the anode metal, historically one of the cleanest early determinations of $h$ .

Step 3. One representative pass, $b = 10$ pm: kick $\Delta v/v = 0.14$ (impulse regime holds), radiated energy $W = \mathbf{0.090}$ eV spread flat up to $\hbar v/b = \mathbf{10.8}$ keV — one pass populates the whole diagnostic X-ray band with $\sim10^{-5}$ photons. Millions of passes per electron do the rest.

Step 4. The regime check and the bill. $2Z\alpha c/v = 1.97$ : the classical and quantum $b_{\min}$ nearly coincide — the tube sits at the crossover, and precision spectra need the Bethe–Heitler/Gaunt machinery. Total conversion efficiency (empirical rule $\eta \approx 10^{-9}\,ZV$ , its $ZV$ scaling exactly this article's $Z^2E/\ldots$ bookkeeping): $\eta \approx 10^{-9}\times74\times10^5 = \mathbf{0.7\%}$ . The other 99.3% of the beam power is heat — which is why anodes are tungsten (melting point), rotating (heat spreading), and water-cooled (everything else).

The lesson: the tube is the whole article in one device — white kick spectra, $Z^2$ target choice, quantum edge, and the $1/m^2$ reason it must be an electron gun.

2 · Weighing a galaxy cluster by its glow (exam level)

Setup. The intracluster medium of a rich cluster: $T = 10^8$ K, $n_e \approx n_i = 10^{-3}\ \mathrm{cm^{-3}}$ , radius 1 Mpc. What does thermal bremsstrahlung predict?

Step 1. The thermometer. $k_BT = \mathbf{8.6}$ keV — the free–free exponential cuts off in the X-ray band; fitting the observed spectral rollover is the temperature measurement (no in-situ probe will ever exist).

Step 2. The luminosity. Frequency-integrated emissivity $\varepsilon^{ff} \approx 1.4\times10^{-27}\sqrt{T}\,n_en_i\bar{g}\ \mathrm{erg\,cm^{-3}s^{-1}}$ (cgs, $\bar{g} = 1.2$ ): $\varepsilon^{ff} = 1.7\times10^{-29}\ \mathrm{erg\,cm^{-3}s^{-1}}$ . Over the volume $\tfrac43\pi R^3 = 1.2\times10^{74}\ \mathrm{cm^3}$ : $L_X \approx \mathbf{2\times10^{45}}\ \mathrm{erg\,s^{-1}}$ — $5\times10^{11}$ solar luminosities of pure free–free light, matching observed rich clusters.

Step 3. The stability check. Thermal energy density over emissivity gives the cooling time $t_{\rm cool} = 3n_ek_BT/\varepsilon^{ff} \approx \mathbf{78}$ Gyr — several Hubble times. The glow is bright in absolute terms yet dynamically negligible: the cluster atmosphere is stable, except in dense cores where $n_e$ rises a hundredfold and $t_{\rm cool} \propto 1/n_e$ collapses below a Gyr — the "cooling-flow" problem that active galactic nuclei are drafted to solve.

Step 4. The payoff. $L_X \propto n_e^2V\sqrt{T}$ plus the spectral $T$ hands over $n_e$ and the gas mass ( $\sim10^{14}M_\odot$ , outweighing the stars); hydrostatic balance at temperature $T$ then weighs the total mass — the classic route by which X-ray bremsstrahlung certified that most cluster matter is dark.

The lesson: one exponential and one $n_e^2\sqrt{T}$ normalization convert a smudge of X-rays into temperature, gas mass, and a dark-matter scale.

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., §15.1–15.4 · Griffiths, Introduction to Electrodynamics, 4th ed., §11.2 · Zangwill, Modern Electrodynamics, Ch. 23 · Heitler, The Quantum Theory of Radiation, 3rd ed. · Rybicki & Lightman, Radiative Processes in Astrophysics, Ch. 5

Provenance. Equations cross-checked against Jackson (Classical Electrodynamics, 3rd ed., §15.1–15.4) for the single-collision radiation, virtual-quanta method, and Kramers spectrum; Griffiths §11.2 for the Larmor formula; Rybicki & Lightman Ch. 5 for the thermal free–free emissivity and Gaunt factor. Numerical claims to verify: mass-scaling suppression $(m_e/m_p)^2\approx3\times10^{-7}$ (so protons radiate ~3.4×10^6 times less); Duane–Hunt relation $\lambda_{\min}=hc/eV$ ; lead radiation length $X_0\approx0.56\,\mathrm{cm}$ (Particle Data Group) and critical energy ~7–10 MeV; the time integral $\int(v^2t^2+b^2)^{-3}\mathrm dt=3\pi/(8vb^5)$ . Relativistic beaming half-angle $1/\gamma$ from the Liénard angular distribution. Addendum for this edition: the Parseval ledger (1) is owned by Larmor Formula from the Liénard–Wiechert Fields and spent here; the beaming lobe shapes are Angular Distribution of Radiation from Accelerated Charges's (its parallel-acceleration anatomy is the bremsstrahlung lobe) — both recalled by name, not re-derived. Four source-text corrections, verified: the source's single-collision box wrote $W \approx (\pi/12)Z^2e^6/(4\pi\varepsilon_0)^3\varepsilon_0c^3m^2vb^3$ — the stray $\varepsilon_0$ breaks dimensions and the coefficient disagrees with its own Step-4/5 assembly; the correct closed form, matching the direct integral to machine precision, is $(\pi/4)Z^2e^6/(4\pi\varepsilon_0)^3c^3m^2vb^3$ ; the source's relativistic-Liénard eqCard wrote $\dot{\boldsymbol\beta}^2/6\pi\varepsilon_0c^3$ -type bookkeeping short by $c^2$ (dimensionless $\dot\beta$ needs $/c$ , not $/c^3$ ); its applications text quoted $X_0(\mathrm{Pb}) \approx 0.35$ cm against the PDG 0.5612 cm its own provenance cites (0.56 cm used here); its radiation-length eqCard tex carried a garbled (empty-denominator) fraction — the standard $1/X_0 = 4\alpha r_e^2n_aZ(Z+1)\ln(183Z^{-1/3})$ is restored. This edition's recomputations: time integral $3\pi/8vb^5$ (quadrature, exact); $W = 0.090$ eV and $\Delta v/v = 0.14$ at $b = 10$ pm, 100 keV on W; crossover $2Z\alpha c/v = 1.97$ ; $\lambda_{\min} = 12.40$ pm; $E_c$ (Pb) $= 7.3$ MeV; cluster $L_X \approx 2\times10^{45}\ \mathrm{erg/s}$ , $t_{\rm cool} \approx 78$ Gyr; exact pulse spectrum $[xK_1(x)]^2$ for the figures.

Builds on: Larmor Formula from the Liénard–Wiechert Fields · Liénard–Wiechert Potentials: Covariant Formulation · Relativistic Generalization of Larmor's Formula · Angular Distribution of Radiation from Accelerated Charges Leads to: Synchrotron Radiation · Fields of a Relativistic Moving Charge · Electric Dipole Radiation: Exact Fields and Zone Structure

If each collision radiates a flat spectrum, why do real X-ray tube spectra rise and then fall with photon energy?

Three multiplicative effects sculpt the plateau. Toward high energy, the Coulomb log shrinks ( $b_{\max} = v/\omega$ admits fewer passes) and the Duane–Hunt edge terminates the spectrum outright — so the thin-target spectrum tilts gently down to a cliff. Real anodes are thick: electrons slow as they penetrate, and the superposition of ever-softer thin-target spectra gives the classic linear-in-energy decline (Kramers's thick-target law $I \propto Z(\hbar\omega_{\max} - \hbar\omega)$ ). Toward low energy, photons generated microns deep are photoelectrically absorbed by the anode itself and by the tube window — the steep low-energy rise is filtered away. What survives is the familiar hump: filtration below, thick-target decline above, cliff at $eV$ . The flat single-collision plateau is still in there — it is the zeroth brick, visible directly only in thin-target experiments.

Is the electron really on a straight line while it radiates? The nucleus bends it.

The impulse approximation is a controlled first order, not a fib. The exact classical problem — radiation from a hyperbolic Kepler orbit — was solved by Kramers, and the straight-line result is its leading term in the deflection angle $\theta_d \sim \Delta v/v$ . The correction to the radiated energy enters at $O(\theta_d^2)$ : for the worked tube example ( $\Delta v/v = 0.14$ ) a couple of percent. The approximation self-reports its failure: when $\Delta v \sim v$ , i.e. $b \lesssim b_{\min}^{\rm cl} = 2Ze^2/4\pi\varepsilon_0mv^2$ , the trajectory is strongly bent, the pulse asymmetric, and the flat-plateau bookkeeping wrong — which is exactly why $b_{\min}^{\rm cl}$ is one of the two candidate lower cutoffs of the Coulomb logarithm. Below it, either the full hyperbolic orbit (classical regime) or Bethe–Heitler (quantum regime) takes over; the logarithm quarantines our ignorance into an additive constant.

Why does the electron do the radiating and not the nucleus — momentum conservation kicks both, after all?

Both are kicked with equal and opposite forces, so the accelerations stand in ratio $m_{\rm nuc}/m_e \sim 10^5$ and the radiated powers in ratio $10^{10}$ : the nucleus's share is gone before any other consideration. (Its charge $Ze$ helps it only as $Z^2$ , nowhere near $10^{10}$ .) There is a subtler point hiding here: for a close pair of comoving equal charges the dipole moments would cancel and quadrupole radiation would lead — but electron and nucleus carry opposite charges and opposite kicks, so their dipole contributions add in principle. They simply add in the ratio of accelerations, leaving the electron's term alone in practice. The same reasoning explains electron–electron bremsstrahlung's weakness at low energy: equal masses and equal charges make the pair's dipole moment $\propto \sum q_i\mathbf{r}_i$ rigorously constant, killing dipole radiation entirely and demoting e–e brems to a quadrupole effect.

How does the same process give a flat spectrum in a tube but an exponential one in a galaxy cluster?

Same brick, different wall. One electron at fixed speed $v$ radiates the flat plateau up to its personal ceiling. A thermal plasma is an ensemble of speeds: an emitted photon $\hbar\omega$ requires an electron with $\tfrac12mv^2 \geq \hbar\omega$ , and the Maxwellian supplies such electrons with probability $\propto e^{-\hbar\omega/k_BT}$ . The observed spectrum is the flat plateaus stacked and weighted by that Boltzmann census — flat where nearly all electrons qualify ( $\hbar\omega \ll k_BT$ ), exponentially falling where only the tail does. That is why the free–free cutoff measures $T$ so cleanly: the exponential's slope is the temperature, no geometry or density required. The tube avoids this only because its "distribution" is a monoenergetic beam; make the beam thermal and the Duane–Hunt cliff smears into precisely the cluster's exponential.

Does the Gaunt factor mean the classical calculation was wrong?

It means the classical calculation was right about everything except a logarithm's argument. The full quantum (Bethe–Heitler) cross-section reproduces the classical structure — the $Z^2/m^2v^2$ prefactor, the near-flat spectrum, the logarithmic factor — and differs in the exact argument and additive constant of the log, plus threshold behavior near the Duane–Hunt edge. The Gaunt factor $g(\omega, v)$ is defined as quantum-over-Kramers, and it is a number of order unity varying slowly across huge swaths of parameter space (velocity-averaged values 1–1.5 for thermal plasmas). This is the correspondence principle doing exactly what it promises: classical electrodynamics prices the process, quantum mechanics adjusts the last ten percent — and the community keeps the classical formula as the definition, honoring the price tag by name.

Why is bremsstrahlung beamed forward at high energy when the kick is sideways?

Because beaming belongs to the velocity, not to the force. The instantaneous emission pattern of any accelerated charge is the rest-frame donut processed through relativistic aberration — the whole content of Angular Distribution of Radiation from Accelerated Charges — and aberration crowds every emission direction into the forward cone $\theta \sim 1/\gamma$ about $\mathbf{v}$ , regardless of which way $\mathbf{a}$ points. For bremsstrahlung the acceleration is mostly transverse, so the relevant lobe is that article's perpendicular-geometry pattern: a filled forward cone (with its tangent nulls), identical in structure to synchrotron emission. At 1 GeV, $1/\gamma \approx 0.5$ mrad: the photons travel with the beam, which is why converter targets feed collimated photon lines and why linac shielding is thickest straight ahead.

Thermal bremsstrahlung is emission from free electrons — can the plasma also absorb by the inverse process, and does it matter?

Yes — free–free absorption, the exact time-reverse (an electron in a nucleus's field absorbs a passing photon), tied to the emissivity by Kirchhoff's law: $\alpha_\omega = \varepsilon_\omega/B_\omega(T)$ for a thermal medium. Its strength scales as $n_e n_i T^{-1/2}\omega^{-2}$ (with the Rayleigh–Jeans factor folded in, effectively $\omega^{-2}$ at radio frequencies), so it matters exactly where the emissivity's $\omega^{-0}$ flatness meets long wavelengths and high densities: compact H II regions turn over from flat to $\nu^2$ blackbody-like spectra at gigahertz frequencies (self-absorption), stellar interiors get the Kramers opacity $\propto \rho T^{-3.5}$ that structured stellar models for decades, and laser-plasma experiments use inverse bremsstrahlung as their principal heating channel. One matrix element, run in both directions, with the Boltzmann factor adjudicating the balance.