Cherenkov Radiation

Every radiation result in this chapter so far has run on one engine: acceleration. The Larmor ledger bills $a^2$ , the relativistic lobes bill $\dot{\boldsymbol\beta}$ , and a charge in uniform motion radiates nothing — that is a theorem, proved more than once upstream. This article is about the loophole in the fine print. The theorem's proof lives in vacuum, where light always outruns the charge and the retarded-time equation has exactly one solution. Put the charge in water, where light crawls at $c/n = 0.75c$ , and let the charge cruise at $0.99c$ : now the charge outruns its own field, the retarded construction picks up a second solution — the loophole flagged by name in Liénard–Wiechert Potentials: Covariant Formulation — and the wake of polarization the charge drags through the medium piles up into a coherent conical shock of real light. No acceleration anywhere. The geometry is pure Huygens: $\cos\theta_c = 1/\beta n$ , a protractor reading of the particle's speed. The energetics is the Frank–Tamm formula, derived here by the frequency-domain route: per unit path and per unit frequency the charge loses $\propto \omega\,(1 - 1/\beta^2n^2)$ — a spectrum that rises with frequency until the medium's own dispersion, owned by Absorption and Dispersion, shuts it off. That rising ramp is the blue glow of reactor pools; the cutoff is why there is no X-ray Cherenkov light. Converted to photons the formula becomes the design equation of a detector industry: a few hundred photons per centimeter, on a cone whose angle names the particle. Super-Kamiokande, IceCube, and every ring-imaging particle identifier at the LHC are this article, built at kiloton scale.

Key takeaways

Cherenkov radiation is emitted by a charge in uniform motion through a medium whenever $v > c/n$ : the charge outruns its own wavelets and the retarded-time equation gains a second root — the loophole of Liénard–Wiechert Potentials: Covariant Formulation, executed here.
The emission is a Huygens envelope: coherence survives only on a cone with $\cos\theta_c = 1/\beta n$ ; the threshold $\beta n > 1$ is the condition for the cone to exist.
The Frank–Tamm formula gives the spectral energy loss $\dfrac{d^2E}{dx\,d\omega} = \dfrac{q^2}{4\pi\varepsilon_0 c^2}\,\omega\left(1 - \dfrac{1}{\beta^2 n^2(\omega)}\right)$ — rising with $\omega$ , cut off where dispersion drags $n(\omega)$ below $1/\beta$ .
Photon bookkeeping: $\dfrac{d^2N}{dx\,d\lambda} = \dfrac{2\pi\alpha z^2}{\lambda^2}\sin^2\theta_c$ — about $330\ \sin^2\theta_c^{-1}$ -normalized photons per cm in water over 300–600 nm; the energy cost ( $\sim$ 1 keV/cm) is a $10^{-3}$ perturbation on ionization loss.
Water numbers: threshold kinetic energy 264 keV (electron) / 54.6 MeV (muon); saturated angle $\theta_c = 41.2°$ ; the cone angle plus an independent momentum measurement identifies the particle.
The blue glow is a spectral ramp, not a resonance: flat in $\omega$ per photon, $1/\lambda^2$ per wavelength.
Key idea: a charge radiates coherently whenever it outruns light in the medium — the cone $\cos\theta_c = 1/\beta n$ encodes the velocity, and the Frank–Tamm spectrum grows with frequency until $n(\omega)$ drops below threshold.

The core insight

A charge radiates coherently whenever it outruns light in the medium: the emission cone $\cos\theta_c = 1/(\beta n)$ encodes the velocity, and the Frank–Tamm formula sets the spectral intensity, which grows with frequency until $n(\omega)$ drops below threshold.

Where this lives today

Cherenkov light is the workhorse photon of astroparticle physics. Water and ice instruments read neutrinos by the rings of their charged secondaries — Super-Kamiokande's 11,000 photomultipliers caught SN1987A's burst and settled solar-neutrino oscillations; IceCube instruments a cubic kilometer of Antarctic ice ( $n \approx 1.31$ , $\theta_c \approx 40°$ ) and has traced PeV neutrinos to blazars; Hyper-Kamiokande and KM3NeT scale the same cosine up again. RICH systems at LHCb and Belle II run worked example 2's logic for flavor physics, and imaging atmospheric Cherenkov telescopes (H.E.S.S., MAGIC, the Cherenkov Telescope Array) photograph the ~1° air-Cherenkov flashes of TeV gamma-ray showers — the atmosphere itself as radiator. Beyond detectors, the effect is a probe and a tool: reactor operators and IAEA safeguards inspectors monitor spent fuel by its blue glow; radiotherapy researchers image Cherenkov light emitted in patients' tissue during electron-beam treatment to map dose in real time; and the Askaryan effect — coherent radio-band Cherenkov emission from the net charge of showers in ice — powers radio arrays hunting EeV neutrinos. A vacuum theorem's loophole, working at every scale from fuel pond to ice cap.

Origins · 1 minPavel Alekseyevich Cherenkov and how this was discovered

Pavel Alekseyevich Cherenkov

1904–1990 · Lebedev Institute, Moscow — the graduate student who trusted a faint blue anomaly

Cherenkov was set a deliberately dull thesis problem by his supervisor Sergei Vavilov in 1932: study the luminescence of uranyl-salt solutions under gamma rays. He noticed the faint blue light refused to behave like luminescence — pure solvents glowed too, quenching agents did not kill it, and cooling did not slow it. Where others would have filed "background," Cherenkov spent two years of near-darkness experiments — dark-adapting his eyes for an hour before each session, using his own vision as the detector — proving the glow was universal, directional, and polarized. Vavilov correctly attributed it to fast electrons; the decisive signature — emission on a forward cone — Cherenkov measured in 1936. Ilya Frank and Igor Tamm supplied the electrodynamics next door in 1937, deriving the cone and the spectrum from Maxwell's equations in a medium (reportedly astonished the answer needed no acceleration). The three shared the 1958 Nobel Prize — Vavilov, who died in 1951, missing it — and the Soviet literature still calls the light Vavilov–Cherenkov radiation. Cherenkov lived to see his thesis "background" become the operating principle of neutrino astronomy.

History

1888

— Heaviside computes the conical field of a hypothetical superluminal charge — and is forgotten

1904

— Sommerfeld analyzes a vacuum charge faster than light, months before relativity forbids it

1934

— Cherenkov and Vavilov report anomalous blue emission from irradiated liquids

1936

— Cherenkov measures the emission's forward cone — the geometric fingerprint

1937

— Frank and Tamm derive the cone and spectrum from Maxwell's equations in a medium

1955

— the antiproton is confirmed at Berkeley with a velocity-selecting Cherenkov counter

1958

— Nobel Prize to Cherenkov, Frank, and Tamm

1987

— Kamiokande-II and IMB catch SN1987A's neutrinos as Cherenkov rings in water

2013

— IceCube opens PeV neutrino astronomy with Cherenkov light in Antarctic ice

The physical picture

Think of a speedboat. Moving slower than the water waves it makes, it sits inside its own expanding ripples and no wake forms. Moving faster, it continually escapes them: the ripples it left behind pile up along two lines, and the familiar V-shaped bow wave appears. Cherenkov radiation is the electromagnetic version, with $c/n$ playing the role of the ripple speed and the V rotated into a cone. The medium is an essential actor, not a backdrop. A charge gliding through a dielectric polarizes the molecules around it — each one becomes a momentary dipole, stretched and released as the charge passes. Below threshold this polarization wake is symmetric fore and aft: every molecular dipole's radiation is cancelled by its neighbors', the wake relaxes silently, and the net field is just the dressed Coulomb field riding along with the charge. Above threshold the symmetry breaks. The charge has moved on before the light from its wake can catch up, the cancellation fails, and there is exactly one family of directions where the wavelets from every point of the track arrive in phase: the surface of the Mach cone. Coherence over a macroscopic track length is what makes the light bright — the emission is not a sum of independent molecular scatters but one giant in-phase antenna the length of the track, the $N^2$ logic of the coherence theorem in Larmor Formula from the Liénard–Wiechert Fields wearing a trench coat. And the color: the blue is not a resonance. The Frank–Tamm spectrum rises linearly in $\omega$ — equal photon counts per unit frequency, which repackaged per unit wavelength gives $1/\lambda^2$ — so within the transparent window the short wavelengths always win. The ramp would climb forever were it not for the medium itself: past the ultraviolet resonances the refractive index dips below $1/\beta$ (the X-ray regime where $n < 1$ , owned by Absorption and Dispersion), the threshold condition fails, and the spectrum shuts off. The glow is blue because the medium stops playing just past violet.

The loophole, executed

The vacuum no-radiation theorem deserves one paragraph of respect before we break it. A charge in uniform motion in vacuum is a boosted electrostatics problem — Fields of a Relativistic Moving Charge exhibits its field as the velocity field alone, $1/r^2$ at infinity, carrying no energy to the far zone. Equivalently: the retarded-time equation $|\mathbf{r} - \mathbf{w}(t_r)| = c(t - t_r)$ has exactly one root when the worldline is slower than the signals, so the potential is a smooth, radiationless functional of the single retarded position. Liénard–Wiechert Potentials: Covariant Formulation proved that uniqueness and flagged, by name, where it fails: in a medium, the signals travel at $c/n$ , and a charge with $v > c/n$ can be inside the light cones of two different past positions at once. Two roots is not a technicality — it is the radiation. Where retarded contributions from two emission times superpose, the invariant denominator $u \cdot R$ (the covariant $\kappa\mathscr{r}$ factor of the potentials article) passes through zero on the cone's surface: the potential diverges where the wavelets bunch. Dispersion softens the front in reality — different colors' cones stack at slightly different angles — but the divergence's location is exact, and it is the whole geometry of what follows. Everything in this article unpacks two consequences: where the coherent surface lies (Derivation A), and how much energy crosses it (Derivation B).

The cone

Derivation ACone angle and threshold from coherence along the track

The goal is the Cherenkov relation $\cos\theta_c = 1/\beta n$ and the threshold $\beta n > 1$ , derived as an interference statement: we treat every point of the track as a Huygens emitter and ask in which direction the contributions from all emission times add in phase. This is stronger than drawing the tangent envelope — it shows the cone is the only direction that survives cancellation, which is why the radiation is monodirectional per frequency rather than merely peaked. Watch the roles: the particle clock sets emission positions, the medium speed $c/n$ sets propagation, and coherence does the rest.

Step 1. Two emission events. Let the charge move along $z$ at speed $v$ , emitting wavelets continuously. Compare the phases, arriving at a distant observer in direction $\theta$ from the track, of wavelets emitted at $z' = 0$ and at $z' = v t'$ . The second wavelet is emitted later by $t'$ , but its path toward the observer is shorter by the projection $v t' \cos\theta$ . Propagating at $c/n$ , its arrival is delayed relative to the first by

\Delta t_{\rm arr} = t' - \frac{v t' \cos\theta}{c/n}.

Why the projection suffices: the observer is many wavelengths away, so the two paths are parallel to leading order — the standard far-zone reduction, the same one every antenna integral in this track uses.

Step 2. The coherence condition. In-phase arrival for every pair of emission times requires $\Delta t_{\rm arr} = 0$ — not merely stationary, but zero, because $t'$ is a continuous parameter and any linear drift of phase with $t'$ integrates the contributions to nothing over a long track. Setting the bracket to zero:

1 = \frac{v\cos\theta}{c/n} \qquad\Longrightarrow\qquad \boxed{\;\cos\theta_c = \frac{c/n}{v} = \frac{1}{\beta n}\;} \qquad (1)

where: $\theta_c$ — angle of the emitted wavevector from the velocity; $\beta = v/c$ ; $n$ — refractive index at the emitted frequency, so the cone is really a stack of cones $\theta_c(\omega)$ in a dispersive medium. The particle's speed has become an angle: this single line is the operating principle of every ring-imaging detector.

Step 3. Threshold as existence. Equation (1) has a real solution only when $\beta n \ge 1$ : the particle must outrun the light of the medium,

v > \frac{c}{n} \qquad\Longleftrightarrow\qquad \gamma > \gamma_{\rm thr} = \frac{n}{\sqrt{n^2 - 1}}. \qquad (2)

At threshold the cone closes forward ( $\theta_c = 0$ ) and — Derivation B will price this — the intensity vanishes with it. Just above threshold a detector sees a pencil of forward light; at $\beta \to 1$ the angle saturates at $\cos\theta_{c,\max} = 1/n$ , the medium's signature angle ( $41.2°$ in water).

Step 4. The wake reading. The same condition, viewed as geometry: in time $t$ the charge advances $\beta c\,t$ while the wavelet emitted at the start has grown to radius $(c/n)t$ ; the common tangent line from the charge's current position touches every wavelet precisely when $\beta n > 1$ , at the complementary angle. The radiated $\hat{\mathbf{k}}$ is perpendicular to that tangent surface — at $\theta_c$ from the track — and the polarization lies in the $(\hat{\mathbf{k}}, \mathbf{v})$ plane, because the source is the radial stretching of the medium's dipoles off the track (no azimuthal component exists by symmetry). Note the Mach-cone reading: the shock front sits at $\sin^{-1}(1/\beta n)$ from the track, and the wavevector is at the complement — consistent with (1). In water at $\beta \to 1$ : $\theta_c = \arccos(1/1.33) = 41.25°$ ; from (2), $\gamma_{\rm thr}(\text{water}) = 1.517$ , giving kinetic threshold energies of 264 keV for electrons and 54.6 MeV for muons — the design numbers for Super-Kamiokande.

Figure

The detector designer's chart, computed from cos theta-c equals one over beta n — cone angle versus beta for water, aerogel and C4F10 gas; each radiator wakes up at its own threshold and saturates at its own angle

The energy bill — Frank and Tamm's formula

The cone says where; it does not say how much. The energy accounting is a boundary-value problem in the frequency domain, and its structure is worth seeing honestly because the same computation with $\beta n < 1$ returns exactly zero — the theorem and the loophole in one formula.

Go deeper · Derivation B · 2 minFrank–Tamm spectral energy loss from the cylindrical energy flow

Derivation BFrank–Tamm spectral energy loss from the cylindrical energy flow

The goal is the Frank–Tamm formula: the energy radiated per unit path length and per unit frequency by a charge in uniform motion with $\beta n(\omega) > 1$ . The route is the frequency-domain one — Fourier the source along the track, solve the in-medium wave equation for each $\omega$ , and audit the Poynting flux through a distant cylinder wrapped around the trajectory. The physics decision happens in one sign: the transverse behavior of the field flips from evanescent to propagating exactly at threshold. We display the structure fully and quote the final Bessel-function bookkeeping at working level, with the sanity audits that make the coefficient trustworthy.

Step 1. The source has one frequency per angle. A charge in uniform motion, $\rho = q\,\delta(x)\delta(y)\delta(z - vt)$ , is a steady pattern in the co-moving sense: Fourier-transforming in $t$ at fixed $z$ ties every frequency to one longitudinal wavenumber, $k_z = \omega/v$ . The current's frequency component is a line source $\propto e^{i\omega z/v}$ riding the track — no knob of its own, everything fixed by $v$ .

Step 2. The in-medium wave equation and the sign that decides. Each Fourier component of the vector potential obeys $(\nabla_\perp^2 + k_\perp^2)\tilde{A}_z = -\mu_0 \tilde{J}_z$ with the transverse wavenumber forced by the dispersion relation $k^2 = n^2\omega^2/c^2$ of Electromagnetic Waves in Linear Dielectric Media:

k_\perp^2 = \frac{n^2\omega^2}{c^2} - \frac{\omega^2}{v^2} = \frac{\omega^2}{v^2}\left(\beta^2 n^2 - 1\right). \qquad (3)

where: everything is per-frequency, $n = n(\omega)$ . For $\beta n < 1$ this is negative: the transverse profile is a decaying $K_0(|k_\perp|\rho)$ , an evanescent sheath glued to the track — the dressed Coulomb field, no energy escaping, the vacuum theorem surviving its move into matter. For $\beta n > 1$ it is positive: the solution is an outgoing Hankel function $H_0^{(1)}(k_\perp\rho)$ , a real cylindrical wave. The threshold is not imposed; it is the sign of (3).

Step 3. The direction audit. The outgoing wave has wavevector components $(k_z, k_\perp) = (\omega/v,\ (\omega/v)\sqrt{\beta^2n^2 - 1})$ , so its propagation angle from the track satisfies

\cos\theta = \frac{k_z}{n\omega/c} = \frac{1}{\beta n},

— Derivation A's cone, re-derived from the field solution with no Huygens language. One phenomenon, two proofs.

Step 4. The flux through the cylinder. The radiated energy per unit track length is the Poynting flux through a cylinder of radius $\rho$ : $dE/dx = -\oint (\mathbf{E}\times\mathbf{H})\cdot\hat{\boldsymbol\rho}\, \rho\,d\phi\,dt$ . Inserting the large- $\rho$ asymptotics of the Hankel functions ( $\propto e^{ik_\perp\rho}/\sqrt{\rho}$ — flux independent of $\rho$ , the escape certificate), the fields' cross products assemble — this is Jackson §13.4's page of Bessel bookkeeping, executed at working level — into

\boxed{\;\frac{d^2E}{dx\,d\omega} = \frac{q^2}{4\pi\varepsilon_0 c^2}\,\omega\left(1 - \frac{1}{\beta^2 n^2(\omega)}\right)\;} \qquad (4)

where: the bracket is exactly $\sin^2\theta_c(\omega)$ — the spectrum carries the cone's geometry point by point; the prefactor $q^2\omega/4\pi\varepsilon_0 c^2$ is the coherent line-antenna scale. Per octave the loss rises linearly in $\omega$ : high frequencies dominate up to wherever the medium stops cooperating.

Step 5. What turns it off. The integral of (4) over all frequencies would diverge for constant $n$ — the shock front would be a true singularity. Dispersion rescues finiteness: past the UV resonances $n(\omega)$ dives below 1 (the X-ray phase-race regime of Absorption and Dispersion), $\beta n < 1$ fails for any $\beta < 1$ , and the emission window closes. The total is an integral over the transparent, above-threshold window only — which for water means, in practice, the visible and near-UV. As a check: the dimensions of (4) are J·m⁻¹·(rad/s)⁻¹ ✓; the bracket $\sin^2\theta_c$ vanishes at threshold, confirming that intensity goes to zero as the cone closes (matching Derivation A Step 3); and integrating (4) over the 300–700 nm window in water gives 1.12 keV/cm — consistent with Derivation C's 380 photons/cm at a mean photon energy of 2.9 eV.

From energy to photons — the detector designer's formula

Go deeper · Derivation C · 2 minPhoton yield and the engineering numbers

Derivation CPhoton yield and the engineering numbers

The goal is the working currency of Cherenkov detectors: how many photons per centimeter, in which band, at what angle. The route is bookkeeping — divide the Frank–Tamm spectrum by the energy per photon, $\hbar\omega$ , and watch a remarkable simplification: the yield is flat in frequency, and the fine-structure constant walks in the door uninvited. One honest correction to the base text's coefficient is made and flagged.

Step 1. Photons per frequency. Dividing (4) by $\hbar\omega$ :

\frac{d^2N}{dx\,d\omega} = \frac{q^2}{4\pi\varepsilon_0\hbar c^2}\sin^2\theta_c = \frac{\alpha z^2}{c}\,\sin^2\theta_c, \qquad (5)

writing $q = ze$ and recognizing $\alpha = e^2/4\pi\varepsilon_0\hbar c$ . The spectrum of photon counts is flat in $\omega$ : equal numbers per unit bandwidth, right up to the cutoff. That the yield of a classical formula is governed by the fine-structure constant is the article's quiet joke: $\hbar$ entered only through the price of a photon. Why $\alpha$ makes the light faint but countable: per centimeter of water the charge crosses $\sim 5\times10^4$ optical wavelengths; multiplied by $\alpha\sin^2\theta_c \sim 3\times10^{-3}$ that yields hundreds, not billions, of photons — bright enough to image, dim enough that a dark-adapted eye near a reactor core sees a glow, not a glare.

Step 2. Photons per wavelength — the blue bias made explicit. Changing variables ( $\omega = 2\pi c/\lambda$ , $|d\omega| = 2\pi c\,d\lambda/\lambda^2$ ):

\frac{d^2N}{dx\,d\lambda} = \frac{2\pi\alpha z^2}{\lambda^2}\,\sin^2\theta_c. \qquad (6)

Flat in frequency means $1/\lambda^2$ in wavelength: twice as many photons per nanometer at 300 nm as at 425 nm. The "blue" glow is really this Jacobian plus the eye's response plus water's transmission window.

Step 3. Band totals. Integrating (6) over a detection band $[\lambda_1, \lambda_2]$ :

\frac{dN}{dx} = 2\pi\alpha z^2 \sin^2\theta_c\left(\frac{1}{\lambda_1} - \frac{1}{\lambda_2}\right). \qquad (7)

For water at $\beta \approx 1$ ( $\sin^2\theta_c = 0.435$ ): 332 photons/cm over 300–600 nm (a bialkali photomultiplier's window), 380 photons/cm over 300–700 nm. The equivalent PDG bookkeeping is per photon energy: $d^2N/(dx\,dE) = 370\,z^2\sin^2\theta_c$ photons per eV per cm — and multiplying by the 2.36 eV width of the 300–700 nm band reproduces the same 380/cm (computed both routes).

Step 4. The energy honesty. The total energy cost, from Derivation B's integral: about 1.1 keV/cm in water — versus roughly 2 MeV/cm of ionization loss for a minimum-ionizing particle. Cherenkov light is a $\sim5\times10^{-4}$ perturbation on the particle's energy budget: it steers nothing dynamically (no analogue of radiation reaction is needed here), while being the entire observable for a photodetector. The best signals are cheap. Two independent routes agree: directly from (7), $2\pi\alpha \times 0.4347 \times (1/300 - 1/600)\,\mathrm{nm}^{-1} = 332\ \mathrm{cm}^{-1}$ over 300–600 nm; via the per-eV PDG coefficient, $370\ \mathrm{eV^{-1}\,cm^{-1}} \times 2.362\ \mathrm{eV} = 874 \times \sin^2\theta_c = 380\ \mathrm{cm}^{-1}$ over 300–700 nm — consistent. A common error is to quote the PDG coefficient of 370 as a band total in photons/cm; it is actually a per-eV-per-cm rate, and the band total requires multiplying by the band width in eV.

Figure

The photon spectrum per unit wavelength computed from the one-over-lambda-squared law for water at beta near one — the blue ramp, the detector band, and the ultraviolet cutoff where dispersion closes the threshold

Limiting cases

case	result	meaning
threshold $\beta n \to 1^+$	$\theta_c \to 0$ , $d^2E/dx\,d\omega \to 0$	cone closes forward, intensity dies with it — threshold counters exploit the sharp turn-on
ultrarelativistic $\beta \to 1$	$\cos\theta_c \to 1/n$ , yield saturates	the medium's signature angle: 41.2° water, 13.9° aerogel — rings stop growing, PID switches to momentum
sub-threshold or vacuum $\beta n < 1$	$k_\perp^2 < 0$ : evanescent sheath, zero flux	the uniform-motion theorem, recovered inside the same formula
high frequency $n(\omega) \to 1^-$	window closes	dispersion (owned by Absorption and Dispersion) makes the total finite; no X-ray Cherenkov
heavy projectile, same $\beta$	identical yield ( $z^2$ only)	Cherenkov reads velocity and charge, never mass — the complement of momentum in PID

Common trapNo acceleration is required — and the phase velocity being beaten is not a signal outrunning light

Two standing confusions. First: Cherenkov radiation comes from a charge in uniform motion; nothing here contradicts "accelerated charges radiate" because that slogan's converse — only accelerated charges radiate — was a vacuum theorem, proved with a single retarded root. The medium supplies the second root and the energy flows from the particle's kinetic energy through the work the wake's field does back on the charge (the in-medium drag force), about 1 keV/cm in water: real, tiny, and fully accounted by Frank–Tamm. Second: $v > c/n$ beats the phase velocity of light in the medium, not causality. Signal fronts in a dispersive medium travel at the front velocity — exactly $c$ , as the front theorem of Phase and Group Velocity certifies — and the particle never outruns that. Nothing superluminal happens except bookkeeping. (A third trap: "the glow is blue" is not a resonance line — it is a $1/\lambda^2$ ramp truncated by transparency and photocathode.)

Worked examples

1 · Super-Kamiokande's ring, end to end (confidence builder)

Setup. A muon of momentum 1 GeV/c crosses a water Cherenkov detector ( $n = 1.33$ ). Find the threshold it had to beat, its cone angle, the ring it paints on a wall 10 m away, and the light budget of a 3 m track segment (band 300–600 nm).

Step 1. Threshold. $\gamma_{\rm thr} = n/\sqrt{n^2-1} = 1.517$ : kinetic thresholds $m c^2(\gamma_{\rm thr}-1)$ = 264 keV for electrons, 54.6 MeV for muons. Our muon ( $E = \sqrt{p^2 + m^2} = 1005.6$ MeV, $\beta = 0.9945$ ) is far above.

Step 2. The angle. $\beta n = 1.3226$ , so $\cos\theta_c = 0.7561$ , $\theta_c = \mathbf{40.9°}$ — within half a degree of the saturated 41.25°: at 1 GeV/c the ring radius has stopped measuring speed (that is exactly why Super-K distinguishes $e/\mu$ by ring sharpness, not size — an electron showers and fuzzes the ring, a muon draws it crisply).

Step 3. The ring. On a wall 10 m downstream the cone paints a circle of radius $r = 10\tan(40.9°) = \mathbf{8.7}$ m — rings in Super-K are enormous, routinely wrapping the cylinder's wall, and the fitted ring center + radius give the track direction and vertex.

Step 4. The light budget. Yield at $\beta = 0.9945$ : $\sin^2\theta_c = 1 - 1/(\beta n)^2 = 0.4284$ , so $dN/dx = 2\pi\alpha \times 0.4284 \times (1/300 - 1/600)\ \mathrm{nm^{-1}} = \mathbf{327}$ photons/cm. Over 3 m: $\approx \mathbf{9.8\times10^4}$ photons in-band. With $\sim$ 40% photocathode coverage and $\sim$ 20% quantum efficiency, roughly $8\times10^3$ photoelectrons — the actual scale of a Super-K muon event display.

The lesson: one cosine and one Jacobian run a 50-kiloton instrument: the angle names the geometry, equation (7) pays for the pixels.

2 · Particle identification by ring radius — aerogel at 2.5 GeV/c (exam level)

Setup. A beamline delivers pions, kaons, and protons, all at $p = 2.5$ GeV/c, through a 2 cm silica-aerogel radiator ( $n = 1.03$ ). Same momentum, different masses — different $\beta$ . What does the ring camera see (band 300–600 nm)?

Step 1. Velocities. $\beta = p/\sqrt{p^2 + m^2}$ : pion (139.6 MeV) $\beta_\pi = 0.99845$ ; kaon (493.7 MeV) $\beta_K = 0.98105$ ; proton (938.3 MeV) $\beta_p = 0.93623$ .

Step 2. Against threshold. $\beta_{\rm thr} = 1/1.03 = 0.97087$ . The proton is below threshold — dark. Both mesons are above.

Step 3. The two rings. $\cos\theta_c = 1/\beta n$ : pion $\theta_c = \mathbf{13.5°}$ , kaon $\theta_c = \mathbf{8.3°}$ — a 5.2° gap, enormous against a milliradian-scale ring-fitting resolution. Photon counts from (7) with the respective $\sin^2\theta_c$ (0.0545 and 0.0208): $\mathbf{83}$ photons (pion) vs $\mathbf{32}$ (kaon) in 2 cm — both comfortably imageable.

Step 4. The design logic. At higher momenta all $\beta \to 1$ and the aerogel rings converge — so spectrometers stack gas radiators behind it: in C4F10 ( $n = 1.0014$ ) the thresholds sit at $p_{\rm thr} = m/\sqrt{n^2-1}$ = 2.6 / 9.3 / 17.7 GeV/c for $\pi/K/p$ — the window slides up. Choosing $n$ is choosing the momentum range: the second figure, read backwards, is RICH design.

The lesson: Cherenkov light measures $\beta$ and only $\beta$ ; paired with a magnet's $p$ , the mass falls out — the cleanest particle scale in physics.

Going further

References. Jackson, Classical Electrodynamics, 3rd ed., §13.4 · Zangwill, Modern Electrodynamics, §23 (radiation in media) · Frank & Tamm, Dokl. Akad. Nauk SSSR 14, 109 (1937) · Particle Data Group, Review of Particle Physics — Cherenkov detectors · Jelley, Cherenkov Radiation and its Applications (1958) A note on photon-yield bookkeeping: the PDG coefficient $370\ z^2\sin^2\theta_c$ photons per eV per cm is a spectral density (per unit photon energy), not a band total. Multiplying by the 2.36 eV width of the 300–700 nm band gives the band total $\approx 380$ photons/cm in water at $\beta \approx 1$ ( $\sin^2\theta_c = 0.435$ ), equivalently 332/cm over the 300–600 nm photomultiplier band — both obtained from $2\pi\alpha\sin^2\theta_c(1/\lambda_1 - 1/\lambda_2)$ and confirmed by the per-eV route. For water ( $n = 1.33$ ): $\theta_{c,\max} \approx 41°$ , electron threshold KE $\approx 0.26$ MeV. The energy cost is $1.12$ keV/cm (computed) versus $\sim$ 2 MeV/cm ionization loss. Worked detector numbers: Super-K at 1 GeV/c muon — $\theta_c = 40.9°$ , ring radius 8.7 m at 10 m, 327 photons/cm, $9.8\times10^4$ photons over 3 m; aerogel PID at 2.5 GeV/c — $13.5°/8.3°$ /dark for $\pi$ /K/p, 83/32 photons in 2 cm; C4F10 thresholds 2.6/9.3/17.7 GeV/c for $\pi$ /K/p.

Builds on: Liénard-Wiechert Potentials · Fields of a Relativistic Moving Charge · Relativistic Generalization of Larmor's Formula Leads to: Synchrotron Radiation · Angular Distribution of Radiation from Accelerated Charges

Where does the radiated energy come from if the particle never accelerates?

From its kinetic energy, through a perfectly ordinary force: the polarization wake the charge creates is asymmetric above threshold (piled up behind the shock), and the wake's own electric field at the charge's position points backward along the track. The work done against that drag force per unit path is exactly the Frank–Tamm $dE/dx$ — about 1 keV/cm in water. So the particle does decelerate, imperceptibly: a 1 GeV muon loses a part in $10^6$ of its energy per centimeter this way, three orders of magnitude below its ionization loss. Field energy out equals mechanical work in, with the medium as intermediary; remove the medium and drag and light vanish together.

Doesn't the particle outrunning light violate special relativity?

No — three speeds are in play and only one is protected. The particle moves at $v < c$ always. The phase velocity $c/n$ it outruns is the speed of steady sinusoidal crests in the medium, a bookkeeping speed that carries no information (it exceeds $c$ itself in the X-ray regime, per Absorption and Dispersion, with no drama). The protected speed is the signal-front velocity, which in every causal medium equals exactly $c$ — the front theorem of Phase and Group Velocity — and nothing in this article approaches it. The correct analogy is the sonic boom: a jet outrunning sound violates no law of acoustics; a muon outrunning water-light violates no law of relativity.

Why is the light emitted on a razor-thin cone rather than in a broad lobe like every other radiation pattern in this chapter?

Because the emitter is macroscopically extended and perfectly phase-locked. A point dipole radiates a fat $\sin^2\theta$ lobe — one emitter, no interference. Here every millimeter of a centimeters-long track emits, and the relative phases are rigidly set by the uniform motion; off the cone direction, contributions from different track segments arrive with a phase that drifts linearly along the track and cancel almost perfectly. It is the physics that turns a long uniform antenna array into a knife-edge beam — the array-factor logic of the antenna articles — with the track as a continuous array. The cone's width is diffraction-limited by the track length, $\Delta\theta \sim \lambda/(L\sin\theta_c)$ : milliradians for centimeter tracks, which is why RICH rings look drawn with a compass.

The Frank–Tamm spectrum rises with frequency — what stops the total from diverging, and how sharp is the shock front really?

The two questions are the same question. For a strictly constant $n$ , integrating $\omega\,d\omega$ to infinity diverges, and the corresponding field singularity would be an infinitely sharp conical shock (the vanishing of the Liénard–Wiechert denominator on the cone). Real media disperse: $n(\omega)$ falls through $1/\beta$ somewhere in the ultraviolet, the above-threshold window is finite, and the integrated loss lands near a keV per centimeter. The same cutoff smears the shock: superposing cones of slightly different $\theta_c(\omega)$ over the finite window turns the singular front into a thin but finite-thickness pulse. There's a pleasing echo of the radiation-reaction story here: a classical formula pushed to arbitrarily high frequency produces an infinity, and the physical resolution is that matter's response function refuses to follow — a preview of how Classical Electron Theory and Self-Energy will diagnose a deeper divergence.

Why do neutrino detectors use Cherenkov light instead of the much larger ionization signal?

Because Cherenkov light travels and carries geometry. The 2 MeV/cm of ionization is deposited locally — reading it requires instrumenting the entire volume (as liquid-argon TPCs do, at great cost per ton). The 300-photons/cm Cherenkov signal propagates tens of meters through transparent water or ice to sparse photomultipliers on the boundary: you instrument a surface to read a volume, which is the only economical way to build at the kiloton–gigaton scale neutrino physics needs. And the light is not merely present but shaped: the cone angle gives the track direction, the ring topology distinguishes electrons from muons, timing across the array gives the vertex. Ionization tells you how much; Cherenkov tells you which way — and for a neutrino telescope pointing is everything.

Is there Cherenkov radiation from neutral particles, or in the radio band?

Not from a truly neutral point particle — the yield carries $z^2$ , and a neutrino at any speed emits nothing. But ensembles can radiate coherently through a net-charge imbalance: a high-energy shower in ice develops a $\sim$ 20% electron excess (Compton scattering feeds electrons into the shower; positrons annihilate), and at wavelengths longer than the shower's transverse size — the radio band — that excess radiates as a single coherent charge, with power scaling as the square of the excess. This is the Askaryan effect, accelerator-confirmed in 2001 and now the basis of radio neutrino arrays (RNO-G and kin) targeting EeV neutrinos where optical arrays are too small.

What sets the difference between a threshold counter, a RICH, and a time-of-flight system using the same physics?

Which observable of the cone they read. A threshold counter reads existence: choose $n$ so species A radiates and species B doesn't at your momentum (worked example 2's dark proton) — one bit per particle, cheap. A RICH reads the angle: fit the ring, get $\beta$ to a part in $10^3$ or better — many bits, works where both species radiate. Both saturate as all $\beta \to 1$ : past tens of GeV even gases stop separating and PID hands over to transition radiation or calorimetry. Time-of-flight is the non-Cherenkov read of the same $\beta$ , competitive at low momenta. Real spectrometers stack several: aerogel + gas RICH covering 2–100 GeV/c is the standard ladder, read directly off this article's second figure.