Synchrotron Radiation

A proton at the top of the LHC's ramp emits about 480 photons every time it goes round — soft ultraviolet, roughly $12$ eV apiece. Per proton that is nothing: five keV out of six point eight TeV, one part in a billion per turn. Multiply by $3\times10^{14}$ protons and it becomes $3.4$ kilowatts of ultraviolet light, poured continuously into a pipe held at $1.9$ kelvin. It is the dominant heat load on the LHC's beam screens, and the machine carries hardware whose only job is to absorb it. Now run the same ring with electrons at a mere $104.5$ GeV — one sixty-fifth of the energy — and each particle throws away $\mathbf{3.49\ GeV\ per\ turn}$ : three and a third percent of everything it owns, every $89$ microseconds. That was LEP, and that is why LEP is gone and the LHC lives in its tunnel. Sixty-five times less energy. Six hundred thousand times more radiation. The whole of that ratio is relativistic kinematics, and this article is the ledger that balances it.

Key takeaways

Signature $(+,-,-,-)$ : $\mathrm{d}s^2 = c^2\mathrm{d}t^2 - \mathrm{d}x^2$ , $p\!\cdot\!p = +m^2c^2$ , timelike positive.
Scope, up front. The radiated power (Liénard, the invariance theorem, the covariant form) is owned by `aq-em-relativistic-larmor-formula`. The radiated spectrum ( $\omega_c = \tfrac32\gamma^3 c/\rho$ , the $1/\gamma$ cone, the universal shape) is owned by `aq-em-synchrotron-radiation`. Both are used here and neither is re-derived. This article owns what neither does: the count.
The γ-ledger — the owned result. Power goes as $\gamma^4$ ; characteristic photon energy as $\gamma^3$ . Photon number is a Lorentz invariant; energy and rate are not. So the counting factorises — $\gamma^4 = \gamma^1 \times \gamma^3$ — and the photon emission rate grows only as $\gamma^1$ .
Made exact: a bent relativistic charge emits $\boxed{N_\gamma = \tfrac{5}{2\sqrt{3}}\,\alpha\,\gamma \approx 0.010533\,\gamma \text{ photons per radian}}$ of bend — and per turn, $\tfrac{5\pi}{\sqrt{3}}\alpha\gamma \approx 0.066180\,\gamma$ . Independent of $\rho$ , of $B$ , of the machine. Only $\gamma$ and the fine-structure constant.
With the mean photon energy $\langle u\rangle = \tfrac{8}{15\sqrt3}u_c = 0.30792\,u_c$ , the ledger closes exactly: $N_\gamma \langle u\rangle = U_0$ . LEP2: $13\,534$ photons/turn $\times\ 258$ keV $= \mathbf{3.49\ GeV}$ . LHC: $480$ photons/turn $\times\ 12.4$ eV $= \mathbf{5.93\ keV}$ .
The tunnel theorem. LEP and the LHC occupy the same $26.7$ km ring. The LHC carries $\mathbf{65\times}$ the energy per particle and radiates $\mathbf{5.9\times10^{5}}$ times less, because at fixed energy and fixed radius $P \propto 1/m^4$ and $(m_p/m_e)^4 = 1.14\times10^{13}$ .
Key idea: the radiation is not a fluid. It is a countable number of quanta — about one hundredth of γ per radian — and that number, not the power, is what an accelerator's designer ultimately lives with.

The core insight

Ask a bent electron three questions and get three different powers of $\gamma$ . Energy per second? $\gamma^4$ . How hard is each photon? $\gamma^3$ . How many photons per second? $\gamma^1$ . These are not three facts but one fact seen from three sides, because photon number is a frame-independent count while energy and arrival rate are not. Divide the power by the photon energy and the $\gamma$ 's cancel down to a single one. That surviving $\gamma^1$ , times nothing more exotic than the fine-structure constant, is the entire radiative character of every circular machine ever built.

Where this lives today

The tax became the product. Dozens of storage rings now exist for no purpose but to bend electrons and collect what falls off. The ESRF's Extremely Brilliant Source runs $6$ GeV electrons on a $23.3$ m bending radius, putting the critical energy at $21$ keV — hard X-rays, tunable, collimated into a $1/\gamma$ cone of $85$ microradians. Structural biology rests on equation (2) being large.

And the count, not the power, limits them. Brilliance — photons per second per unit phase-space volume — wants $N_\gamma$ large and the emittance small, and those fight: the same discrete recoils that make $N_\gamma$ countable also pump the beam's oscillations, so the equilibrium emittance is the balance of quantum excitation against radiation damping. Fourth-generation light sources are engineered right up against that floor — a floor that exists only because $\hbar \ne 0$ .

And the tax still decides what gets built. No circular electron collider will reach the energy frontier again: at fixed $\rho$ , $U_0 \propto E^4$ , so FCC-ee needs a $91$ km ring for $182$ GeV per beam — and stops there, because $E^4$ always wins. The only ways out are a linear machine (no bend, no radiation) or a heavier particle. Both roads leave through the same equation.

The physical picture

Two results are taken as given, from live canon, and used without proof.

The power. A charge in circular motion of radius $\rho$ radiates $P_{\text{circ}} = \dfrac{\mu_0 q^2 c^3}{6\pi}\dfrac{\gamma^4\beta^4}{\rho^2}$ — owned by `aq-em-relativistic-larmor-formula`.

The spectrum. That power is not delivered at the orbital frequency. Beaming squeezes the emission into a cone of half-angle $\sim1/\gamma$ , the observer sees a flash lasting only $\sim\rho/\gamma^3c$ , and the reciprocal is the critical frequency $\omega_c = \tfrac32\gamma^3c/\rho$ — owned by `aq-em-synchrotron-radiation`. Now put them side by side, and notice what the two owners never have cause to say to each other. The power carries $\gamma^4$ ; the photon energy carries $\gamma^3$ ; and a power divided by an energy is a rate:

\text{photons per second} \;\sim\; \frac{P}{\hbar\omega_c} \;\propto\; \frac{\gamma^4}{\gamma^3} \;=\; \boldsymbol{\gamma}.

Only the first power. That is not an accident of algebra — it is a statement about what survives a boost. Photon number is a count, and a count is the same in every frame. Energy is not: it Doppler-shifts. Rate is not: it time-dilates. Transform from the electron's instantaneous rest frame — where it is a plain dipole emitting a modest number of soft photons — into the lab, and the energies are boosted and the rate is boosted, and their product is the $\gamma^4$ . But no photon is created by the boost. You cannot Lorentz-transform a photon into existence. Strip both boosts out of $\gamma^4$ and exactly $\gamma^1$ remains. That $\gamma^1$ is what this article owns, and it is the number a machine designer actually lives with.

Figure

THE γ-LEDGER — the owned result. Three questions, three different powers of γ, and they are one fact. TOP: the radiated POWER goes as γ⁴ (red) — owned by aq-em-relativistic-larmor-formula. MIDDLE: the critical photon ENERGY goes as γ³ (gold) — owned by aq-em-synchrotron-radiation. BOTTOM: divide one by the other and the photon RATE goes only as γ¹ (blue) — and THAT is what nobody else owns, because it is a statement about Lorentz invariance. Photon NUMBER is a frame-independent count. Energy Doppler-shifts and rate time-dilates, and their product is the γ⁴ in the power — but the count passes through the boost untouched. So γ⁴ = γ¹ × γ³, exactly, and the surviving γ¹ carries nothing but the fine-structure constant: N = (5/2√3)·α·γ = 0.010533γ photons per radian of bend, independent of ρ, of B, of the machine. The ledger closes: N·⟨u⟩ = U₀, verified at LEP2 (13,534 photons × 258 keV = 3.49 GeV per turn) and at the LHC (480 photons × 12.4 eV = 5.93 keV per turn)

The core equations

The radiated power for circular motion — quoted from `aq-em-relativistic-larmor-formula`, not derived here:

P_{\text{circ}} = \frac{\mu_0 q^2 c^3}{6\pi}\,\frac{\gamma^4\beta^4}{\rho^2} \qquad (1)

where: $\rho$ is the radius of curvature. Physical reading: the $\gamma^4$ is the signature of perpendicular acceleration. The base prints this with $c^1$ instead of $c^3$ — short by $c^2$ . Energy radiated per turn, obtained by multiplying (1) by the revolution period $T = 2\pi\rho/\beta c$ :

\boxed{\;U_0 = \frac{\mu_0 q^2 c^2}{3}\,\frac{\gamma^4\beta^3}{\rho} \;\equiv\; \frac{q^2\gamma^4\beta^3}{3\varepsilon_0\rho}\;} \qquad\xrightarrow{\;\text{electrons}\;}\qquad U_0[\mathrm{keV}] = 88.46\,\frac{E^4[\mathrm{GeV}]}{\rho[\mathrm{m}]} \qquad (2)

where: the two forms are identical, since $\mu_0c^2 = 1/\varepsilon_0$ . Physical reading: $E^4$ , and the $1/m^4$ hiding inside it — at fixed energy $\gamma = E/mc^2 \propto 1/m$ . This is the equation that decides whether your ring holds electrons or protons. The base prints $c^1$ where it needs $c^2$ , turning $88.5$ keV into $2.95\times10^{-7}$ keV. Critical frequency and beaming cone — quoted from `aq-em-synchrotron-radiation`, not derived here:

\omega_c = \frac{3}{2}\,\frac{\gamma^3 c}{\rho} = \frac{3}{2}\,\gamma^3\omega_0,\qquad\qquad \theta_{\text{beam}} \sim \frac{1}{\gamma} \qquad (3)

where: $\omega_0 = \beta c/\rho$ is the orbital frequency. Physical reading: $\gamma^3$ , not $\gamma^2$ . The base's own second equality says $\tfrac32\gamma^2\omega_0$ and is wrong, contradicting its own first equality. One power of $\gamma$ from the cone sweeping past; two more from the source chasing its own light. Now the article's own equations. The mean photon energy, in units of $u_c = \hbar\omega_c$ :

\langle u\rangle = \frac{8}{15\sqrt{3}}\;u_c \;=\; 0.30792\,u_c \qquad (4)

where: a property of the universal spectrum, whose shape belongs to `aq-em-synchrotron-radiation`. Physical reading: $u_c$ is not the typical photon energy — it is the median of the radiated power. The average photon is three-tenths of it. And the owned result — the photon count:

\boxed{\;\frac{\mathrm{d}N_\gamma}{\mathrm{d}\theta} = \frac{5}{2\sqrt{3}}\,\alpha\,\gamma \;=\; 0.010533\,\gamma\;}, \qquad\qquad \boxed{\;N_\gamma^{\text{turn}} = \frac{5\pi}{\sqrt{3}}\,\alpha\,\gamma \;=\; 0.066180\,\gamma\;} \qquad (5)

where: $\theta$ is the angle turned through, $\alpha = 1/137.036$ . Physical reading: look at what is absent. No $\rho$ . No $B$ . No mass, no charge, no $c$ . The photon count per radian depends on nothing but $\gamma$ and $\alpha$ — an electron and a proton at the same $\gamma$ emit the same number, though the electron's are $(m_p/m_e)^3$ times harder. This is as close as accelerator physics comes to a pure number.

Why the exponents are 4, 3 and 1

Derivation AThe gamma-ledger — photon number is invariant, so the counting factorises

The two owners of this physics each hold one exponent. The power carries $\gamma^4$ ; the characteristic photon energy carries $\gamma^3$ . Nothing in either article requires the two to be compared, and so they never are. But a power divided by an energy is a rate, and the moment you perform that division you are asserting something about what survives a Lorentz transformation and what does not. The goal here is to do the division honestly, to see that the surviving exponent is exactly $1$ , and to turn that into a closed form with no free parameters — the number of photons a bent charge emits per radian of its bend.

Step 1. State the invariance that makes the argument work. Photon number is a Lorentz invariant. Boost from the particle's rest frame to the lab and every photon is shifted, every emission time dilated, every direction aberrated — but no photon is created and none destroyed. Why: the load-bearing step, so state it precisely. $N$ is a count of events on a worldline, and counts do not transform. Energy transforms (Doppler). Rate transforms (time dilation). Their product is the power, and the power is what carries $\gamma^4$ . The count is the residue left when both boosts are divided out.

Step 2. Do the division at the level of scalings. In the lab,

\underbrace{P}_{\propto\;\gamma^4} \;=\; \underbrace{\left(\frac{\mathrm{d}N_\gamma}{\mathrm{d}t}\right)}_{\propto\;\gamma^{\,?}} \times \underbrace{\langle u\rangle}_{\propto\;\gamma^3} \qquad\Longrightarrow\qquad \frac{\mathrm{d}N_\gamma}{\mathrm{d}t} \;\propto\; \gamma^{4-3} = \boldsymbol{\gamma^1}.

Why: $\gamma^4 = \gamma^1 \times \gamma^3$ , and the arithmetic is the physics. The two exponents everyone quotes are not independent facts to be memorised — they are a rate and an energy whose product is a power, and once you know two of the three you know the third. The reason nobody writes down the third is that neither owner has occasion to divide.

Step 3. Make it exact, with (2) and (4). Per unit angle turned, the energy radiated is $\mathrm{d}U/\mathrm{d}\theta = U_0/2\pi$ , and each photon carries $\langle u\rangle = \tfrac{8}{15\sqrt3}u_c$ on average, so

\frac{\mathrm{d}N_\gamma}{\mathrm{d}\theta} = \frac{U_0/2\pi}{\langle u \rangle} = \frac{\mu_0q^2c^2\gamma^4/(6\pi\rho)}{\dfrac{8}{15\sqrt3}\cdot\dfrac{3\hbar c\gamma^3}{2\rho}} .

Why: watch the $\rho$ . It appears once in the numerator (from $U_0 \propto 1/\rho$ ) and once in the denominator (from $u_c \propto 1/\rho$ ). They cancel. The bending radius — the single most expensive parameter in an accelerator — drops out of the photon count entirely.

Step 4. Cancel and recognise the constant. The $\rho$ 's go, three powers of $\gamma$ go, and what remains is

\frac{\mathrm{d}N_\gamma}{\mathrm{d}\theta} = \frac{5\sqrt3}{24}\cdot\frac{\mu_0 q^2 c}{\hbar}\,\gamma .

But $\alpha \equiv \dfrac{q^2}{4\pi\varepsilon_0\hbar c} = \dfrac{\mu_0 q^2 c}{4\pi\hbar}$ , so $\dfrac{\mu_0q^2c}{\hbar} = 4\pi\alpha$ , and

\boxed{\;\frac{\mathrm{d}N_\gamma}{\mathrm{d}\theta} = \frac{5}{2\sqrt{3}}\,\alpha\,\gamma = 0.010533\,\gamma\;},\qquad\qquad N_\gamma^{\text{turn}} = 2\pi\cdot\frac{5}{2\sqrt3}\alpha\gamma = \frac{5\pi}{\sqrt3}\,\alpha\,\gamma = 0.066180\,\gamma .

Why: the fine-structure constant walked in of its own accord. It had to: a photon count is dimensionless, so once $\rho$ , $c$ , $\hbar$ , $q$ and $\mu_0$ have finished cancelling, the only dimensionless things left in the problem are $\alpha$ and $\gamma$ . The mass has vanished too — an electron and a proton at the same $\gamma$ emit the same number of photons per radian, though the electron's are $(m_p/m_e)^3 = 6.2\times10^{9}$ times harder.

Step 5. Check that the ledger closes. $N_\gamma\langle u\rangle = U_0$ is an identity by construction, so it is not a test — the test is whether both sides, computed from independent closed forms, agree numerically. At LEP2 ( $104.5$ GeV, $\rho = 3026$ m, $\gamma = 204\,501$ ): $u_c = 837$ keV, $\langle u\rangle = 258$ keV, $N_\gamma = \mathbf{13\,534}$ per turn, and $13\,534 \times 258\ \mathrm{keV} = \mathbf{3.49\ GeV}$ — exactly what $88.46\,E^4/\rho$ gives independently. Why: the ledger balances by two routes that share no algebra, which is the only kind of confirmation worth having. And read the LEP number again: thirteen thousand hard X-ray photons, from one electron, on one lap.

Step 6. See what the count buys that the power does not. Because $N_\gamma$ is finite and countable — not a smooth fluid — the energy loss is grainy. Each of those $13\,534$ photons kicks the electron back with a discrete recoil at a random moment, and the random walk of those recoils pumps energy into the beam's oscillations. That is quantum excitation, and balanced against radiation damping it sets an electron ring's irreducible equilibrium emittance. Why: the emittance floor of every light source on Earth exists because $N_\gamma$ is a number and not infinity. If the radiation were a continuous fluid — $N_\gamma \to \infty$ , $\langle u\rangle \to 0$ , $U_0$ fixed — the recoils would average away and the floor would be zero. It is not, because the count is $0.0662\gamma$ and no more. That calculation needs $\hbar$ ; this article stops at the door, having supplied the number the door is made of. *Verified in verify_aq-sr-synchrotron-radiation.py: the coefficients $5/(2\sqrt3)\alpha = 0.010533$ , $5\pi/\sqrt3\,\alpha = 0.066180$ and $8/(15\sqrt3) = 0.30792$ are computed at 50 dps; the ledger $N_\gamma\langle u\rangle = U_0$ is confirmed against the independent closed form $U_0 = (4\pi/3)(r_0/\rho)\gamma^4mc^2$ at both LEP2 and the LHC; and the $\rho$ -independence is asserted inversely — the count must NOT move when $\rho$ changes, while both of its parents ( $U_0$ and $u_c$ ) MUST.*

Go deeper · Derivation B · 2 minThe tunnel theorem — the same ring, sixty-five times the energy, six hundred thousand times less light

Derivation BThe tunnel theorem — the same ring, sixty-five times the energy, six hundred thousand times less light

LEP and the LHC are the same hole in the ground: a $26.7$ km tunnel under the Franco-Swiss border, bored once, used twice. That coincidence makes an unusually clean experiment out of equation (2), because it holds the bending radius almost fixed and changes only the particle. The goal is to compute what each machine throws away per turn, and to show that the entire, enormous ratio between them is a single power law with an exponent of four.

Step 1. Extract the mass dependence. In (2), $U_0 \propto \gamma^4/\rho$ , and at a fixed energy $E$ the Lorentz factor is $\gamma = E/mc^2 \propto 1/m$ . Therefore, at fixed $E$ and fixed $\rho$ ,

\boxed{\;U_0 \;\propto\; \frac{1}{m^4}\;}

Why: the $1/m^4$ is not in the formula — you have to put the energy in to see it. This is why the mass-dependence of synchrotron radiation is so often misquoted: written in $\gamma$ it looks mass-free, and it is only mass-free at fixed $\gamma$ , which is not how anyone builds a machine.

Step 2. Evaluate the ratio for the two species. $(m_p/m_e)^4 = (1836.15)^4 = \mathbf{1.14\times10^{13}}$ . Why: at the same energy in the same ring, a proton radiates ten trillion times less than an electron. That single number is the reason the LHC exists in the shape it does.

Step 3. Do the real machines, which are not at the same energy — the LHC is at $65$ times LEP's. LEP2 ( $104.5$ GeV, $\rho = 3026$ m, $\gamma = 204\,501$ ): $U_0 = 88.46(104.5)^4/3026$ keV $= \mathbf{3.49\ GeV\ per\ turn}$ , or $3.34\%$ of the beam energy every $89\ \mu$ s. LHC ( $6.8$ TeV, $\rho = 2804$ m, $\gamma = 7247$ ): $U_0 = \mathbf{5.93\ keV\ per\ turn}$ , a fraction $8.7\times10^{-10}$ . Why: the LHC's protons sit at sixty-five times the energy of LEP's electrons in essentially the same ring. Naively that should radiate $65^4 = 1.8\times10^{7}$ times more. Instead it radiates $\mathbf{5.9\times10^{5}}$ times less — the mass ratio's fourth power beats the energy ratio's by seven orders of magnitude.

Step 4. Read the count, which is this article's own contribution to the story. From (5): the LHC proton emits $0.066180 \times 7247 = \mathbf{480}$ photons per turn of mean energy $\mathbf{12.4\ eV}$ (vacuum UV, $100$ nm); the LEP2 electron emits $\mathbf{13\,534}$ per turn at $\mathbf{258\ keV}$ (hard X-ray). Why: the counts differ by a factor of only $28$ — it is the photon energies that differ by $21\,000$ . That is the ledger doing its work. The rate is a $\gamma^1$ effect and is nearly the same in both machines; all the brutality is in the $\gamma^3$ of the photon energy. The LHC does not radiate less because it emits fewer photons. It radiates less because its photons are soft. *Verified in verify_aq-sr-synchrotron-radiation.py: $U_0$ is computed for both machines by two independent routes — the practical $88.46E^4/\rho$ and the first-principles $(4\pi/3)(r_0/\rho)\gamma^4mc^2$ — and asserted equal; $(m_p/m_e)^4 = 1.14\times10^{13}$ , the $5.9\times10^{5}$ loss ratio, the $3.4$ kW beam-screen load and the $11\,245$ Hz revolution frequency are each recomputed at 50 dps and grepped against this article.*

Figure

THE TUNNEL THEOREM. LEP and the LHC are the same hole in the ground — a 26.7 km ring, bored once, used twice — so this is an unusually clean experiment on U₀ ∝ γ⁴/ρ. LEFT: LEP2's electrons at 104.5 GeV (γ = 204 501) threw away 3.49 GeV every single turn — 3.34% of everything they owned, every 89 microseconds. RIGHT: the LHC's protons at 6.8 TeV (γ = 7247) carry SIXTY-FIVE TIMES the energy in essentially the same ring, and lose 5.93 keV per turn — 5.9×10⁵ times LESS. The entire ratio is one power law: at fixed energy and fixed radius U₀ ∝ 1/m⁴, and (m_p/m_e)⁴ = 1.14×10¹³ beats the energy ratio's 65⁴ by seven orders of magnitude. BELOW: the ledger. The photon COUNTS differ by only a factor of 28 — that is the γ¹ rate law, and it is nearly the same in both machines. It is the photon ENERGIES that differ by a factor of 21 000 — that is the γ³ law. The LHC does not radiate less because it emits fewer photons. It radiates less because its photons are soft

Limiting cases

Case	Result	Meaning
$\gamma \to 1$	$P \to \mu_0q^2a^2/6\pi c$	ordinary Larmor. Broad dipole pattern, single frequency $\omega_0$ : cyclotron, not synchrotron
$\gamma \gg 1$	$P \propto \gamma^4$ , $\omega_c \propto \gamma^3$ , $N_\gamma \propto \gamma$	the ledger. Three exponents, one fact
$\gamma \to \infty$ at fixed $\rho$	$N_\gamma/\theta \to \infty$ linearly	photons come faster, but only linearly — the violence is in their energy, not their number
$c \to \infty$	$\gamma \to 1$ , $\omega_c \to \tfrac32\omega_0$ , $P \to$ Larmor	no beaming, no compression, no $\gamma^4$ . The whole effect is relativistic
$\hbar \to 0$	$N_\gamma \to \infty$ , $\langle u\rangle \to 0$ , $U_0$ fixed	the radiation becomes a smooth fluid. Quantum excitation vanishes and the equilibrium emittance would be zero
electron $\to$ proton at fixed $E$ , $\rho$	$U_0 \times (m_e/m_p)^4 = 8.8\times10^{-14}$	the tunnel theorem. This is why the LHC is a proton machine

Common trapThree of this base's equations are wrong, and its neighbour convicts them

The base carries a family of factor-of- $c$ errors, and the proof is not subtle: the neighbouring base `aq-sr-relativistic-beam-dynamics` prints the standard $U_0[\mathrm{keV}] = 88.5\,E^4[\mathrm{GeV}]/\rho[\mathrm{m}]$ , and it is right (first principles gives $\mathbf{88.46}$ ). The two bases contradict each other in print.

eq[1], $P_{\text{circ}} = \mu_0q^2c\,\gamma^4\beta^4/6\pi\rho^2$ : short by $c^2$ . Correct: $\mu_0q^2\mathbf{c^3}\gamma^4\beta^4/6\pi\rho^2$ .
eq[4], $\Delta E = \mu_0q^2c\,\gamma^4\beta^3/3\rho$ : short by $c$ . Correct: $\mu_0q^2\mathbf{c^2}\gamma^4\beta^3/3\rho \equiv q^2\gamma^4\beta^3/3\varepsilon_0\rho$ . The base's form gives $2.95\times10^{-7}$ keV where the truth is $88.5$ — wrong by $3\times10^{8}$ .
eq[3], $\omega_c = \tfrac32\gamma^3c/\rho = \tfrac32\gamma^2\omega_0$ : the second equality has the wrong exponent — with $\omega_0 = \beta c/\rho$ it must be $\tfrac32\gamma^{\mathbf{3}}\omega_0$ . The base's own first equality says $\gamma^3$ ; only the restatement says $\gamma^2$ . Right form, wrong restatement, on the same line — the most dangerous error there is, because it survives every check you would think to run on the equation you actually read.

Common trapThe Liénard prefactor depends on which bracket you pair it with

The relativistic Larmor power is written two ways and they do not share a prefactor. In terms of the acceleration $\mathbf{a} = \mathrm{d}\mathbf{v}/\mathrm{d}t$ (Griffiths 11.73):

P = \frac{\mu_0q^2\gamma^6}{6\pi c}\left(a^2 - \frac{|\mathbf{v}\times\mathbf{a}|^2}{c^2}\right) .

In terms of $\dot{\boldsymbol\beta} = \mathrm{d}\boldsymbol\beta/\mathrm{d}t = \mathbf{a}/c$ (Jackson 14.26, converted from Gaussian):

P = \frac{\mu_0q^2\mathbf{c}}{6\pi}\,\gamma^6\left[\dot{\boldsymbol\beta}^2 - (\boldsymbol\beta\times\dot{\boldsymbol\beta})^2\right] .

The two brackets differ by exactly $c^2$ , so the two prefactors must too. Same physics; both reduce to $\mu_0q^2a^2/6\pi c$ at $\gamma = 1$ . The endemic error is to take $\mu_0q^2/6\pi c$ from the first and staple it to the bracket of the second. That is exactly what this article's base does — it writes the $\dot{\boldsymbol\beta}$ bracket, defines $\dot{\boldsymbol\beta}$ in its own symbol table as "coordinate acceleration over $c$ ", and puts $\mu_0q^2/6\pi c$ in front. Short by $c^2 = 9\times10^{16}$ . Always name the bracket variable. The prefactor is not a constant of nature — it is a bookkeeping choice, and it is the one people get wrong.

Worked examples

1 · Count the photons an LEP electron threw away on one lap

LEP2 ran electrons at $104.5$ GeV on a bending radius of $3026$ m. How many photons did one electron emit per turn, how hard were they, and does the ledger balance against the machine's known $3.5$ GeV-per-turn loss?

Step 1 — identify. This is the owned result, equation (5), plus the critical energy from (3) and the mean-energy fraction from (4). Everything is fixed by $\gamma$ alone.

Step 2 — the Lorentz factor. $\gamma = 104.5\ \mathrm{GeV}/0.511\ \mathrm{MeV} = \mathbf{204\,501}$ . LEP's electrons were the most relativistic objects ever placed under human control.

Step 3 — the count. From (5), and without needing $\rho$ at all:

N_\gamma = 0.066180 \times 204\,501 = \mathbf{13\,534\ photons\ per\ turn} .

Step 4 — the hardness. The critical energy $u_c = \tfrac32\hbar c\gamma^3/\rho$ ; with $\hbar c = 1.9733\times10^{-7}$ eV·m,

u_c = \frac{3}{2}\cdot\frac{(1.9733\times10^{-7})(204\,501)^3}{3026} = \mathbf{837\ keV},\qquad \langle u\rangle = 0.30792\,u_c = \mathbf{258\ keV} .

Hard X-rays — each photon carrying half an electron rest mass.

Step 5 — the boundary number, and the check. Multiply:

N_\gamma\,\langle u\rangle = 13\,534 \times 258\ \mathrm{keV} = \mathbf{3.49\ GeV} ,

against $88.46 \times (104.5)^4/3026 = \mathbf{3.49\ GeV}$ from the wholly independent practical formula. The ledger closes. And the physical picture is worth holding: a single electron, once per lap, sprays out thirteen thousand hard X-ray photons and loses $3.3\%$ of everything it has. LEP's RF plant existed to hand that back, $11\,245$ times a second, and by the end it could not hand back enough. That is how a $27$ km accelerator dies.

2 · Why the LHC needs a beam screen

The LHC runs protons at $6.8$ TeV, $\rho = 2804$ m, with $3.2\times10^{14}$ protons per beam. Protons famously do not radiate. Compute the loss anyway.

Step 1 — identify. Equation (2) for the per-turn loss, (5) for the count, and then the step that turns "negligible" into "engineering": multiply by the number of protons.

Step 2 — per proton. $\gamma = 6800/0.9383 = \mathbf{7247}$ , and (2) gives $U_0 = \mathbf{5.93\ keV\ per\ turn}$ — a fraction $8.7\times10^{-10}$ of the beam energy. Per proton this is genuinely nothing.

Step 3 — the count and the colour. From (5), $N_\gamma = 0.066180 \times 7247 = \mathbf{480\ photons\ per\ turn}$ , critical energy $u_c = 40.2$ eV, mean energy $\langle u\rangle = \mathbf{12.4\ eV}$ — a wavelength of $100$ nm. The LHC's protons glow in the vacuum ultraviolet.

Step 4 — multiply by the beam. The revolution frequency is $c/26\,659\ \mathrm{m} = 11\,245$ Hz, so

P_{\text{beam}} = (5934\ \mathrm{eV})(1.602\times10^{-19})(11\,245)(3.2\times10^{14}) = \mathbf{3.4\ kW\ per\ beam} .

Step 5 — the boundary number. $3.4$ kW into a bore at $1.9$ kelvin. Removing a watt at liquid-helium temperature costs of order a kilowatt of plug power, so this is the LHC's dominant beam-induced heat load — and it is why the cold bore carries a beam screen: a perforated, actively cooled liner at $5$ – $20$ K whose only job is to catch this ultraviolet light before it lands on a $1.9$ K magnet. The particle that "does not radiate" still radiates kilowatts. Equation (2) is about one proton. Accelerators are not.

Physics of the portrait

Origins · 2 minSynchrotron radiation and how this was discovered

Synchrotron radiation

1898–1949 · Liénard writes the relativistic power · Schwinger builds the theory · the light is seen, by accident, through a mirror at General Electric

It was predicted long before anyone could make it. Alfred Liénard, working in the pre-relativistic electrodynamics of 1898 — seven years before Einstein — derived the radiated power of a charge in arbitrary motion and published it in L'Éclairage Électrique, an electrical-engineering trade journal. His formula is exactly right at all speeds. It had to be: Maxwell's equations were relativistic all along, and Liénard simply solved them without knowing what he had. Then it was seen, and nobody was looking for it. On 24 April 1947, at the General Electric Research Laboratory in Schenectady, a technician peering through a mirror into the vacuum chamber of a $70$ -MeV synchrotron noticed a bright arc of blue-white light. It had been predicted; in that moment it was an accident. Elder, Gurewitsch, Langmuir and Pollock published Radiation from Electrons in a Synchrotron (Phys. Rev. 71, 829 (1947)), and the effect took the name of a machine built to do something else. Two years later it was understood. Schwinger's On the Classical Radiation of Accelerated Electrons (Phys. Rev. 75, 1912 (1949)) produced the full spectrum — critical frequency, Bessel-function shape, polarization — and turned a nuisance into an instrument. That reversal is this article's whole story. To an accelerator builder, synchrotron radiation is the tax that killed LEP. To everyone else it is the brightest X-ray source in existence, and there are now dozens of machines whose only purpose is to bend electrons in order to throw the light away on purpose.

History

1898

— Alfred Liénard publishes the radiated power of a charge in arbitrary motion (L'Eclairage Electrique 16, 5), correct at all speeds, seven years before special relativity.

1944

— Ivanenko and Pomeranchuk predict that radiation will set an energy ceiling on circular electron accelerators.

1947

— Elder, Gurewitsch, Langmuir and Pollock see the light directly, through a mirror, on the General Electric 70 MeV synchrotron (Phys. Rev. 71, 829); first observation, on 24 April.

1949

— Julian Schwinger derives the full classical spectrum, including the critical frequency and the polarization (Phys. Rev. 75, 1912).

1956

— Tomboulian and Hartman make the first deliberate use of synchrotron light for spectroscopy, at Cornell — the beginning of the light-source industry.

1989

— LEP begins operation; at 104.5 GeV in its final years each electron loses 3.49 GeV per turn, and the RF plant needed to replace it defines the machine's energy ceiling.

2008

— The LHC is commissioned in LEP's tunnel with protons, whose (m_p/m_e)^4 suppression of 1.14e13 makes 6.8 TeV possible where 105 GeV had been the limit.

Going further

Texts. Jackson, Classical Electrodynamics, 3rd ed., §14.6 · Rybicki & Lightman, Radiative Processes in Astrophysics, ch. 6 · Sands, The Physics of Electron Storage Rings, SLAC-121 (1970) — the source of the photon-count and mean-energy coefficients · Hofmann, The Physics of Synchrotron Radiation · Schwinger, Phys. Rev. 75, 1912 (1949) · Elder, Gurewitsch, Langmuir & Pollock, Phys. Rev. 71, 829 (1947).

Provenance. Signature $(+,-,-,-)$ . Every number is recomputed at 50 dps in verify_aq-sr-synchrotron-radiation.py, which reads this file and asserts each printed value against what it computed. Base defects repaired, all re-proved here and none taken on trust: (i) eq[1] $P_{\text{circ}}$ short by $c^2$ ; (ii) eq[4] $\Delta E$ /turn short by $c$ — it yields $2.95\times10^{-7}$ keV where the truth is $88.46$ , and the neighbouring base `aq-sr-relativistic-beam-dynamics`, which prints the correct $88.5$ , is what convicts it; (iii) eq[3]'s second equality prints $\tfrac32\gamma^2\omega_0$ where it must be $\tfrac32\gamma^3\omega_0$ ; (iv) the Liénard prefactor is mis-paired with its bracket (second warning) — the fix is to name the bracket variable, not to change the number.

Scope, and the de-overlap. `aq-em-relativistic-larmor-formula` owns Liénard, the invariance of the radiated power, the covariant form and the $\gamma^6$ -vs- $\gamma^4$ split; `aq-em-synchrotron-radiation` owns the spectrum — critical frequency, $1/\gamma$ cone, universal shape, polarization. Both are quoted here and neither is re-derived. This article owns the γ-ledger: photon number is a Lorentz invariant while energy and rate are not, so $\gamma^4 = \gamma^1 \times \gamma^3$ and the emission rate is only $\tfrac{5}{2\sqrt3}\alpha\gamma$ per radian. The classical fields underneath are `aq-sr-lienard-wiechert-potentials`. Quantum excitation needs $\hbar$ and is signposted, never derived.

Builds on. Relativistic Beam Dynamics · The Relativistic Larmor Formula · Relativistic Beaming. Leads to. Muon Storage-Ring Time Dilation · Collider vs Fixed-Target Energetics.

If photon number is a Lorentz invariant, why does the power depend on the frame at all?

Because power is a number of photons times an energy, divided by a time, and both of those transform. Boost from the electron's rest frame to the lab: every photon in the forward cone is blueshifted (energy up), and the emission interval is contracted for a forward observer (rate up). Multiply the two boosts and you get the $\gamma^4$ ; divide them back out and you recover the count, which never moved. The invariance of $N_\gamma$ is not in tension with the frame-dependence of $P$ — it explains it, by telling you the $\gamma^4$ must factorise into exactly two pieces, and which piece is which.

Why does the bending radius drop out of the photon count?

Because it enters numerator and denominator identically. $U_0 \propto \gamma^4/\rho$ : a tighter bend radiates more. $u_c \propto \gamma^3/\rho$ : a tighter bend also makes each photon harder, by the same factor. Divide to get a count and the $\rho$ 's cancel exactly. A tighter bend does not give you more photons. It gives you the same number of harder ones — a design fact invisible in either parent formula alone.

Are the γ⁴ and the γ³ derived here?

No, deliberately. The $\gamma^4$ power law belongs to `aq-em-relativistic-larmor-formula`; the $\gamma^3$ critical frequency, the $1/\gamma$ cone and the universal spectrum belong to `aq-em-synchrotron-radiation`. Both are live canon; re-deriving them here would duplicate two good articles. This article takes both as given and asks the one question neither has occasion to ask — what happens when you divide one by the other. That the answer is $\gamma^1$ , carrying nothing but $\alpha$ , is why the article exists.

The base says the critical frequency is (3/2)γ²ω₀. Which is right?

$\tfrac32\gamma^{\mathbf{3}}\omega_0$ , with $\omega_0 = \beta c/\rho$ the orbital frequency. The base's own first equality $\omega_c = \tfrac32\gamma^3c/\rho$ is correct, and since $\omega_0 \approx c/\rho$ the two are the same statement — so the base's restatement contradicts the line it restates. But beware a live-canon lookalike that is NOT an error: `aq-em-synchrotron-radiation` writes $\omega_c = \tfrac32\gamma^2\,eB/m$ , which is correct, because $eB/m = \gamma\omega_0$ (the relativistic cyclotron frequency is $\omega_0 = eB/\gamma m$ ). Same $\gamma^3$ , different bookkeeping. An exponent means nothing until you say what it multiplies.

Protons "do not radiate." Why does the LHC have a beam screen, then?

Because "does not radiate" is a claim about one proton, and an accelerator is not one proton. Per turn an LHC proton loses $5.93$ keV out of $6.8$ TeV — one part in a billion, genuinely negligible. But there are $3.2\times10^{14}$ of them orbiting $11\,245$ times a second, and the product is $3.4$ kW of vacuum-ultraviolet light poured continuously into a bore at $1.9$ K. Removing a watt at liquid-helium temperature costs of order a kilowatt at the wall, so this is the dominant beam-induced heat load, and the beam screen — a perforated liner at $5$ – $20$ K inside the cold bore — exists to intercept it. The negligible, multiplied by enough, is an engineering department.

What sets the floor on how small an electron beam can be made?

The fact that $N_\gamma$ is finite. Each of the $13\,534$ photons an LEP electron emitted per turn carried momentum, and each recoil kicked the electron at a random moment in its oscillations. That random walk — quantum excitation — pumps energy into the beam while radiation damping drains it, and the balance fixes the ring's irreducible emittance. Were the radiation a continuous fluid ( $N_\gamma \to \infty$ , $\langle u\rangle \to 0$ , $U_0$ fixed), the recoils would average to nothing and the floor would be zero. It is not, because the count is $0.0662\gamma$ and no more. This article's owned result is the input to that calculation — which needs $\hbar$ , so it is signposted here, never derived.