The Rosenbluth formula

The Rosenbluth formula is the equation that turned the proton from a point into an object with a size. Fire an electron at a proton and, at low enough energy, it bounces off elastically — the proton recoils but stays whole. If the proton were a structureless Dirac particle, one line of QED would predict the entire angular distribution. It is not, and the deviation of the measured cross section from that point prediction is the proton's structure made visible. The Rosenbluth formula packages that deviation into exactly two functions of one variable: the electric form factor $G_E(Q^2)$ , tied to how the proton's charge is spread out, and the magnetic form factor $G_M(Q^2)$ , tied to how its magnetisation is spread out. Everything the single-photon exchange can tell you about the proton's shape is in those two numbers at each momentum transfer $Q^2$ . The formula's deepest feature is a piece of experimental leverage hidden in its algebra. The electric and magnetic contributions ride on different powers of the scattering angle — the charge term on $\cos^2(\theta/2)$ , the magnetic term on $\sin^2(\theta/2)$ — so at a fixed $Q^2$ you can sweep the beam energy and angle together, hold $Q^2$ constant, and watch the two terms trade places. Plotted the right way the data fall on a straight line whose slope gives $G_M$ and whose intercept gives $G_E$ . That trick, the Rosenbluth separation, is how the proton's charge radius of about 0.84 fm was first measured, and the steep fall of both form factors with $Q^2$ is precisely what drove physicists to push to higher energy — where the proton stops recoiling as a whole and its quarks are struck one at a time.

Key takeaways

Elastic $e^-p \to e^-p$ scattering through one virtual photon; the proton's finite size enters as two form factors, the electric $G_E(Q^2)$ and the magnetic $G_M(Q^2)$ .
The formula extends the point-charge Mott cross section by a recoil factor $E_3/E_1$ and a magnetic spin term, all controlled by $\tau = Q^2/4m_p^2$ .
Elastic two-body kinematics has only one free variable: measuring the electron angle $\theta$ fixes both the recoil energy $E_3$ and $Q^2 = 4E_1E_3\sin^2(\theta/2)$ .
The electric term rides $\cos^2(\theta/2)$ and the magnetic term $\sin^2(\theta/2)$ , so a Rosenbluth separation — a straight-line fit at fixed $Q^2$ — pulls $G_E$ and $G_M$ apart.
Normalisations $G_E(0)=1$ and $G_M(0)=+2.79$ encode the proton charge and its anomalous magnetic moment; both fall like the empirical dipole, giving a charge radius $\approx 0.84$ fm.
Key idea: Replace a point proton's QED cross section with two $Q^2$ -dependent form factors that encode its charge and magnetic structure; their distinct angular dependence ( $\cos^2$ vs $\sin^2$ ) lets you disentangle them from data at fixed $Q^2$ but varying angle.

The core insight

Replace a point proton's QED cross section with two $Q^2$ -dependent form factors that encode its internal charge and magnetic structure; because they multiply different powers of the scattering angle, data taken at fixed $Q^2$ but varying angle separate them cleanly.

Where this lives today

The Rosenbluth formula is not a settled museum piece — it sits at the centre of two live controversies. The first is the proton radius puzzle. Electron-scattering and hydrogen-spectroscopy measurements had converged on a charge radius near 0.877 fm, but in 2010 a measurement using muonic hydrogen (where a muon orbits the proton and probes its size far more sharply) gave 0.841 fm — a seven-sigma discrepancy that launched a decade of re-measurement. The PRad experiment at Jefferson Lab (2019), pushing elastic $e^-p$ scattering to very low $Q^2$ , found 0.831 fm, and the field has now largely converged near the smaller muonic value — showing how much rides on the extrapolation of $G_E(Q^2)$ to $Q^2=0$ . The second controversy is deeper: the Rosenbluth separation and the newer polarisation-transfer technique disagree about $G_E/G_M$ at high $Q^2$ . The resolution is that single-photon exchange is not exact — two-photon exchange, a small radiative correction, distorts the Rosenbluth slope more than the polarisation observables, and experiments comparing $e^+p$ with $e^-p$ scattering have now measured it directly. Looking outward, the same elastic formalism, extended to inelastic final states where the proton breaks up, becomes the structure-function framework of deep inelastic scattering — the tool that revealed quarks. The MUSE experiment at PSI is currently scattering electrons and muons off the same protons to test lepton universality in this most classic of reactions.

The physical picture

An electron does not touch the proton; it exchanges a single virtual photon with it. That photon carries a spacelike four-momentum $q$ , and the resolving power of the probe is set by $Q^2 \equiv -q^2 > 0$ : a larger $Q^2$ means a shorter wavelength and a sharper look inside. If the proton were a point, the photon would couple to it through the bare Dirac vertex $\gamma^\mu$ and the cross section would be fixed. Because the proton is an extended, magnetised blob of charge, the coupling is dressed — the bare vertex is replaced by a structure function that depends on how hard the photon probes. The first figure is that story drawn as its one diagram, with the kinematic relations that make elastic scattering a one-variable problem.

Figure

Elastic electron–proton scattering through single-photon exchange: the electron radiates a virtual photon of momentum transfer q with Q² = −q² = 4E₁E₃ sin²(θ/2), which is absorbed at the dressed proton vertex Γᵘ carrying the form factors. Two-body kinematics fixes the recoil energy E₃ = E₁ mₚ ÷ (mₚ + E₁(1 − cos θ)), so the electron angle θ alone determines the whole event

The angular distribution is where the two form factors part company. The charge of the proton and its magnetic moment couple to the photon in different ways: the charge term is largest when the electron barely deflects (forward angles), while the magnetic spin-flip term grows toward backward angles. Written out, the electric piece rides a factor $\cos^2(\theta/2)$ and the magnetic piece a factor $\sin^2(\theta/2)$ , and the two are weighted by the dimensionless recoil parameter $\tau = Q^2/4m_p^2$ . Hold $Q^2$ fixed — which you can do by adjusting beam energy and angle in tandem — and the ratio of the two terms slides smoothly with angle. Dividing the measured cross section by the point-charge (Mott) rate and plotting against $\tan^2(\theta/2)$ turns that slide into a straight line: intercept sets the electric combination, slope sets $G_M$ . The second figure is that line, drawn with real dipole values at $Q^2 = 0.5\ \mathrm{GeV}^2$ .

Figure

The Rosenbluth separation at fixed Q² = 0.5 GeV² (τ = 0.142): the reduced cross section Σ = (dσ/dΩ) ÷ (σ_Mott · E₃/E₁) plotted against tan²(θ/2) is a straight line. Its intercept (G_E² + τG_M²)/(1 + τ) = 0.219 fixes the electric term and its slope 2τG_M² = 0.262 fixes the magnetic term, giving G_M = 0.96 and G_E = 0.34 at this momentum transfer

The core equations

The central result is the Rosenbluth formula itself: the most general Lorentz-invariant differential cross section for elastic $e^-p$ scattering through single-photon exchange. It is the Mott cross section for a point charge, multiplied by the recoil factor $E_3/E_1$ , with the point vertex replaced by the two form factors weighted by their angular factors.

\frac{d\sigma}{d\Omega}=\frac{\alpha^2}{4E_1^2\sin^4(\theta/2)}\,\frac{E_3}{E_1}\left(\frac{G_E^2+\tau G_M^2}{1+\tau}\cos^2\frac{\theta}{2}+2\tau G_M^2\sin^2\frac{\theta}{2}\right) \qquad (1)

where: $\alpha=e^2/4\pi\approx 1/137$ — the fine-structure constant; $E_1,E_3$ — incoming and scattered electron energies; $\theta$ — electron scattering angle; $G_E(Q^2),G_M(Q^2)$ — electric and magnetic form factors; $\tau=Q^2/4m_p^2$ — the recoil/magnetic weighting parameter; $m_p=938.3\ \mathrm{MeV}$ — the proton mass. The first prefactor $\alpha^2/4E_1^2\sin^4(\theta/2)$ is the relativistic point-charge (Mott) cross section; $E_3/E_1<1$ is the target-recoil suppression; the bracket is where all the structure lives. Elastic two-body scattering off a target at rest has only one independent kinematic variable, so the electron angle fixes everything. Four-momentum conservation gives the recoil energy and the momentum transfer directly.

E_3=\frac{E_1\,m_p}{m_p+E_1(1-\cos\theta)},\qquad Q^2=2E_1E_3(1-\cos\theta)=4E_1E_3\sin^2\frac{\theta}{2} \qquad (2)

where: the electron is treated as ultra-relativistic ( $m_e\to 0$ ), so $|\mathbf p|\approx E$ . Measure $\theta$ and $E_1$ and you know $E_3$ , and hence $Q^2$ ; measure $E_3$ independently and its agreement with this formula is the experimental proof that the scattering was elastic and the proton survived intact. The parameter that appears throughout is $\tau=Q^2/4m_p^2$ , small at low $Q^2$ and large when the probe is hard. The proton's structure enters through its electromagnetic vertex. Lorentz invariance and current conservation permit exactly two independent functions of $Q^2$ : the Dirac form factor $F_1$ (which survives for a point charge) and the Pauli form factor $F_2$ (the anomalous part), combined into the physically clean Sachs form factors $G_E$ and $G_M$ .

\Gamma^\mu=F_1(Q^2)\,\gamma^\mu+\frac{i\sigma^{\mu\nu}q_\nu}{2m_p}\,F_2(Q^2),\qquad G_E=F_1-\tau F_2,\quad G_M=F_1+F_2 \qquad (3)

where: $\sigma^{\mu\nu}=\tfrac{i}{2}[\gamma^\mu,\gamma^\nu]$ ; $F_1(0)=1$ (the proton's unit charge) and $F_2(0)=\kappa_p=1.79$ (its anomalous magnetic moment in nuclear magnetons). The Sachs combinations are chosen because it is $G_E$ and $G_M$ , not $F_1$ and $F_2$ , that appear squared and uncrossed in the cross section — there is no $G_EG_M$ interference term, which is exactly what makes the separation possible. The normalisations at $Q^2=0$ are set by static properties of the proton, and the measured $Q^2$ -dependence is captured to first approximation by a single empirical dipole shape shared by both form factors.

G_E(0)=1,\quad G_M(0)=\mu_p=+2.79,\qquad G_E(Q^2)\approx\frac{G_M(Q^2)}{\mu_p}\approx\left(1+\frac{Q^2}{0.71\ \mathrm{GeV}^2}\right)^{-2} \qquad (4)

where: $\mu_p=1+\kappa_p=2.79$ nuclear magnetons is the full proton magnetic moment; the fact that $\mu_p\neq 1$ was itself the first sign the proton is not a point Dirac particle. The empirical relation $G_M\approx\mu_p G_E$ (called form-factor scaling) holds well at low $Q^2$ but is not a theorem, and it breaks down above a few $\mathrm{GeV}^2$ . To pull the two form factors apart, divide the measured cross section by the Mott-times-recoil factor to form a reduced cross section, which is linear in $\tan^2(\theta/2)$ at fixed $Q^2$ .

\Sigma\equiv\frac{d\sigma/d\Omega}{\sigma_{\text{Mott}}\,(E_3/E_1)}=\frac{G_E^2+\tau G_M^2}{1+\tau}+2\tau G_M^2\tan^2\frac{\theta}{2},\qquad \sigma_{\text{Mott}}=\frac{\alpha^2\cos^2(\theta/2)}{4E_1^2\sin^4(\theta/2)} \qquad (5)

where: at fixed $Q^2$ (hence fixed $\tau$ ) the only variable on the right is $\tan^2(\theta/2)$ . Plotting $\Sigma$ against $\tan^2(\theta/2)$ gives a straight line: the slope is $2\tau G_M^2$ and the intercept is $(G_E^2+\tau G_M^2)/(1+\tau)$ . From the slope you get $G_M$ , and feeding that back into the intercept gives $G_E$ — the Rosenbluth separation of the second figure. Finally, in the low- $Q^2$ limit the form factors are Fourier transforms of the charge and magnetisation distributions, and the slope of $G_E$ at the origin defines the proton's mean-square charge radius.

\langle r^2\rangle=-6\,\frac{dG_E}{dQ^2}\bigg|_{Q^2=0},\qquad \sqrt{\langle r^2\rangle}\approx 0.84\ \mathrm{fm} \qquad (6)

where: the factor $-6$ comes from expanding $G_E(Q^2)=1-\tfrac16\langle r^2\rangle Q^2+\cdots$ (a Fourier transform of $\rho(r)$ evaluated at small $Q^2$ , where $Q^2\approx\mathbf q^2$ ). The dipole form gives $\langle r^2\rangle=12/0.71\ \mathrm{GeV}^{-2}$ , i.e. $\sqrt{\langle r^2\rangle}=0.81$ fm; modern extractions land near 0.84 fm.

Derivation AWhy the proton needs exactly two form factors

The goal of this derivation is to show that once you demand Lorentz invariance and electromagnetic current conservation, the most general single-photon coupling of a spin-half proton is fixed to contain precisely two independent functions of $Q^2$ — no more, no fewer — and to identify them with the electric and magnetic Sachs form factors that appear, squared and without a cross term, in the Rosenbluth cross section. This is the structural backbone of the whole subject: it explains why elastic $e^-p$ scattering measures two numbers at each $Q^2$ rather than one or three.

Step 1. Write the proton's matrix element as $\langle p'|J^\mu|p\rangle=\bar u(p')\,\Gamma^\mu\,u(p)$ , and ask what tensors $\Gamma^\mu$ can be built from the available vectors $p^\mu$ , $p'^\mu$ (or equivalently $q^\mu=p'^\mu-p^\mu$ and $P^\mu=p^\mu+p'^\mu$ ) and the gamma matrix $\gamma^\mu$ , with coefficients that can only be scalar functions of the one invariant $Q^2$ . Why: between two on-shell spinors the only Lorentz-covariant building blocks are $\gamma^\mu$ , $P^\mu$ , and $q^\mu$ , so any vertex is a combination of these three with $Q^2$ -dependent coefficients.

Step 2. Impose current conservation $q_\mu\langle p'|J^\mu|p\rangle=0$ . Acting with $q_\mu$ and using the Dirac equation $\slashed p u(p)=m_p u(p)$ kills any term proportional to $q^\mu$ , eliminating one of the three structures. Why: charge conservation is not optional — it removes the longitudinal piece and reduces three candidate form factors to two.

Step 3. Use the Gordon identity, $\bar u(p')\gamma^\mu u(p)=\bar u(p')\big[\tfrac{P^\mu}{2m_p}+\tfrac{i\sigma^{\mu\nu}q_\nu}{2m_p}\big]u(p)$ , to trade the $P^\mu$ structure for $\gamma^\mu$ and $\sigma^{\mu\nu}q_\nu$ . The vertex becomes $\Gamma^\mu=F_1(Q^2)\gamma^\mu+\tfrac{i\sigma^{\mu\nu}q_\nu}{2m_p}F_2(Q^2)$ , the Dirac and Pauli form factors. Why: this is the canonical basis — $F_1$ is the charge-like coupling that a point Dirac particle would have alone, and $F_2$ is the extra magnetic coupling that only an extended, anomalously-magnetic object carries.

Step 4. Repackage into Sachs form factors $G_E=F_1-\tau F_2$ and $G_M=F_1+F_2$ . When the cross section is computed, these appear as $G_E^2$ and $G_M^2$ with no $G_EG_M$ cross term, and with the clean angular split of equation (1). Why: the Sachs combination diagonalises the cross section, which is what makes the electric and magnetic contributions separately measurable — and gives $G_E$ and $G_M$ their interpretation as charge and magnetisation Fourier transforms at low $Q^2$ . *Verified: $F_1(0)=1$ , $F_2(0)=\kappa_p=1.79$ give $G_E(0)=1$ and $G_M(0)=1+1.79=2.79=\mu_p$ ; the absence of a $G_EG_M$ interference term is confirmed by expanding equation (1). Matches Thomson §7.5 and Halzen & Martin §8.3.*

Go deeper · Derivation B · 1 minElastic kinematics: one angle fixes the whole event

Derivation BElastic kinematics: one angle fixes the whole event

The aim here is to derive the recoil relation $E_3=E_1m_p/[m_p+E_1(1-\cos\theta)]$ and the momentum transfer $Q^2=4E_1E_3\sin^2(\theta/2)$ from nothing but four-momentum conservation, and to make explicit why elastic $e^-p$ scattering off a target at rest is a one-variable problem. This is what lets an experiment fix $Q^2$ by choosing a beam energy and an angle, and what lets it confirm elasticity simply by checking that the scattered electron energy came out where the formula says it should.

Step 1. Label four-momenta: electron in $p_1=(E_1,\mathbf p_1)$ , out $p_3=(E_3,\mathbf p_3)$ , proton in $p_2=(m_p,\mathbf 0)$ , out $p_4$ . Conservation gives $p_4=p_1+p_2-p_3$ . Square it: $p_4^2=m_p^2$ (elastic), so $(p_1+p_2-p_3)^2=m_p^2$ . Why: squaring the recoiling proton's momentum and using its on-shell mass is the trick that eliminates the unmeasured proton entirely, leaving a relation among electron quantities only.

Step 2. Expand the square with $p_1^2=p_3^2=0$ (massless electron) and $p_2^2=m_p^2$ : $m_p^2+2p_1\!\cdot\!p_2-2p_2\!\cdot\!p_3-2p_1\!\cdot\!p_3=m_p^2$ . The proton-at-rest dot products are $p_1\!\cdot\!p_2=E_1m_p$ and $p_2\!\cdot\!p_3=E_3m_p$ , while $p_1\!\cdot\!p_3=E_1E_3(1-\cos\theta)$ . Why: choosing the lab frame where the proton is at rest turns the invariants into laboratory energies and the one measurable angle.

Step 3. Substitute and cancel $m_p^2$ : $E_1m_p-E_3m_p-E_1E_3(1-\cos\theta)=0$ . Solve for $E_3$ : $E_3=E_1m_p/[m_p+E_1(1-\cos\theta)]$ , equation (2). Why: this is the recoil formula — the scattered energy is uniquely determined by the incident energy and angle, the hallmark of elastic two-body kinematics.

Step 4. Form the momentum transfer $Q^2=-q^2=-(p_1-p_3)^2=2p_1\!\cdot\!p_3=2E_1E_3(1-\cos\theta)=4E_1E_3\sin^2(\theta/2)$ . Combined with Step 3, both $E_3$ and $Q^2$ are functions of $\theta$ alone at fixed $E_1$ . Why: this closes the loop — one angle fixes $E_3$ , which fixes $Q^2$ , so an experiment scans $Q^2$ by scanning $\theta$ , and any spread of measured $E_3$ away from the predicted value flags inelasticity. *Verified numerically: $E_1=0.5$ GeV, $\theta=90°$ give $E_3=0.326$ GeV, $Q^2=0.326\ \mathrm{GeV}^2$ , $\tau=0.093$ (Python check). Consistent with Thomson Eq. 7.30.*

Go deeper · Derivation C · 2 minThe Rosenbluth separation of G_E and G_M

Derivation CThe Rosenbluth separation of G_E and G_M

This derivation shows how the cross section of equation (1), which mixes the two form factors, is manipulated into a form that measures them one at a time. The purpose is to expose the piece of experimental cleverness the formula is named for: by dividing out the calculable point-charge rate and plotting the remainder against $\tan^2(\theta/2)$ at fixed $Q^2$ , the data collapse onto a straight line whose slope and intercept are simple combinations of $G_M$ and $G_E$ . This is how every early proton form-factor measurement was made.

Step 1. Factor $\cos^2(\theta/2)$ out of the bracket in equation (1), using $\sin^2(\theta/2)=\cos^2(\theta/2)\tan^2(\theta/2)$ . The cross section becomes $d\sigma/d\Omega=\sigma_{\text{Mott}}(E_3/E_1)\big[(G_E^2+\tau G_M^2)/(1+\tau)+2\tau G_M^2\tan^2(\theta/2)\big]$ , where $\sigma_{\text{Mott}}=\alpha^2\cos^2(\theta/2)/4E_1^2\sin^4(\theta/2)$ . Why: pulling out the Mott factor isolates the pure-structure content into a bracket that is manifestly linear in $\tan^2(\theta/2)$ .

Step 2. Define the reduced cross section $\Sigma\equiv(d\sigma/d\Omega)/[\sigma_{\text{Mott}}(E_3/E_1)]$ , so $\Sigma=A(Q^2)+B(Q^2)\tan^2(\theta/2)$ with intercept $A=(G_E^2+\tau G_M^2)/(1+\tau)$ and slope $B=2\tau G_M^2$ . Why: dividing by the known kinematic prefactor removes all the trivial angle and energy dependence, leaving a clean line whose two parameters are exactly what we want.

Step 3. Fix $Q^2$ and vary the angle. Because $Q^2$ is held constant, $\tau$ is constant, so as the beam energy and angle are changed together to keep $Q^2$ fixed, each measurement is a point $(\tan^2(\theta/2),\Sigma)$ on one straight line. A linear fit returns $A$ and $B$ . Why: holding $Q^2$ fixed while sweeping $\theta$ is the whole experimental strategy — it is the only way to keep the form factors constant while the angular weighting changes.

Step 4. Extract the form factors: from the slope, $G_M^2=B/2\tau$ ; from the intercept, $G_E^2=A(1+\tau)-\tau G_M^2$ . At $Q^2=0.5\ \mathrm{GeV}^2$ ( $\tau=0.142$ ) a slope $B=0.262$ and intercept $A=0.219$ give $G_M=0.96$ and $G_E=0.34$ . Why: this is the payoff — two numbers read off one line at each $Q^2$ , and repeating at many $Q^2$ maps out both form factors and hence the proton's charge and magnetisation profiles. *Verified: with dipole inputs at $Q^2=0.5\ \mathrm{GeV}^2$ , $G_E=0.344$ , $G_M=0.961$ give $A=0.2185$ , $B=0.262$ ; inverting slope and intercept recovers $G_M=0.961$ , $G_E=0.344$ (Python check). Matches Thomson §7.5.1.*

Limiting cases

Case	Result	Meaning
$Q^2\to 0$	$G_E\to 1$ , $G_M\to 2.79$ ; cross section $\to$ Mott + magnetic	the static proton — unit charge and full magnetic moment $\mu_p$
Point Dirac proton	$F_2=0$ , so $G_E=G_M=1$ ; no anomalous term	structureless spin-half target; the measured $\mu_p\neq 1$ already rules this out
Low $Q^2$ ( $\tau\ll 1$ )	$G_E,G_M$ are Fourier transforms of $\rho(r),\mu(r)$ ; $Q^2\approx\mathbf q^2$	form factors read out the charge and magnetisation distributions directly
Forward angle ( $\theta\to 0$ )	bracket $\to (G_E^2+\tau G_M^2)/(1+\tau)$	the electric combination dominates; best sensitivity to $G_E$
Backward angle ( $\theta\to 180°$ )	$\tan^2(\theta/2)\to\infty$ ; magnetic term $2\tau G_M^2$ wins	isolates $G_M$ — the spin-flip magnetic scattering
High $Q^2$ ( $\tau\gg 1$ )	cross section $\propto G_M^2$ ; both form factors $\sim Q^{-4}$ (dipole)	the steep fall-off; elastic rate collapses and drives the move to deep inelastic scattering

Common trapForm factors are not simply the Fourier transforms of charge and current

The seductive but only-half-true picture is that $G_E(Q^2)$ and $G_M(Q^2)$ are the Fourier transforms of the proton's charge density $\rho(r)$ and magnetisation $\mu(r)$ . This holds cleanly only in the low- $Q^2$ , non-relativistic limit, where the four-momentum transfer $Q^2$ and the three-momentum transfer $\mathbf q^2$ coincide (they are related by $Q^2(1+Q^2/4m_p^2)=\mathbf q^2$ ). At larger $Q^2$ the proton recoils relativistically, the "rest frame" in which a spatial density would be defined is itself moving, and the interpretation as a static Fourier transform breaks down — the form factors are genuinely functions of the invariant $Q^2$ , not of a spatial distribution. A second, related trap: do not assume $G_M=2.79\,G_E$ as a matter of principle. The observed form-factor scaling is an empirical low- $Q^2$ accident, with no theoretical guarantee; both must be measured independently, and at high $Q^2$ polarisation-transfer experiments show the ratio $\mu_p G_E/G_M$ falling well below one. Treat "proton radius" as shorthand for the slope of $G_E$ at the origin, not as the literal edge of a ball of charge.

Worked examples

1 · Kinematics and form factors at a fixed setting (fully worked)

A 500 MeV electron beam scatters elastically off a hydrogen target and electrons are detected at $\theta=90°$ . Find the scattered energy $E_3$ , the momentum transfer $Q^2$ , the parameter $\tau$ , the dipole values of $G_E$ and $G_M$ , and the reduced cross section $\Sigma$ . Take $m_p=938.3$ MeV.

Step 1. Recoil energy from equation (2): with $\cos 90°=0$ , $1-\cos\theta=1$ , so $E_3=E_1m_p/[m_p+E_1]=(500)(938.3)/(938.3+500)=469150/1438.3=\mathbf{326.2\ MeV}$ . The proton has taken a real kick — the electron lost a third of its energy to recoil, so $E_3/E_1=0.652$ .

Step 2. Momentum transfer: $Q^2=4E_1E_3\sin^2(45°)=4(500)(326.2)(0.5)=326200\ \mathrm{MeV}^2=\mathbf{0.326\ GeV^2}$ . Then $\tau=Q^2/4m_p^2=0.326/(4\times 0.880)=\mathbf{0.0926}$ — small, so the recoil is mild and the electric term still matters.

Step 3. Dipole form factors from equation (4): $G_E=(1+0.326/0.71)^{-2}=(1.459)^{-2}=\mathbf{0.470}$ , and $G_M=\mu_p G_E=2.79\times 0.470=\mathbf{1.31}$ . Both are already well below their $Q^2=0$ values of 1 and 2.79 — the finite proton size is unmistakable even at this modest $Q^2$ .

Step 4. Reduced cross section from equation (5) with $\tan^2(45°)=1$ : $\Sigma=(G_E^2+\tau G_M^2)/(1+\tau)+2\tau G_M^2=(0.221+0.0926\times1.716)/1.0926+2\times0.0926\times1.716=0.347+0.318=\mathbf{0.665}$ . Feeding this back through $\sigma_{\text{Mott}}(E_3/E_1)$ gives a differential cross section of about $\mathbf{18\ nb/sr}$ — the number a spectrometer would actually register. $E_3=326$ MeV, $Q^2=0.33\ \mathrm{GeV}^2$ , $G_E=0.47$ , $G_M=1.31$ , $\Sigma=0.67$ .

2 · Reading a Rosenbluth line

At fixed $Q^2=0.50\ \mathrm{GeV}^2$ , a set of measurements at different beam energies and angles yields a straight line of reduced cross section versus $\tan^2(\theta/2)$ with intercept $A=0.219$ and slope $B=0.262$ . Extract $G_M$ and $G_E$ .

Step 1. Compute $\tau$ : $\tau=Q^2/4m_p^2=0.50/(4\times0.880)=\mathbf{0.142}$ . This is the only kinematic input the separation needs, and it is fixed once $Q^2$ is fixed.

Step 2. From the slope, $B=2\tau G_M^2$ , so $G_M^2=B/2\tau=0.262/(2\times0.142)=0.923$ , giving $G_M=\mathbf{0.96}$ . The slope alone delivers the magnetic form factor — this is why backward-angle data (large $\tan^2$ ) are prized: they lengthen the lever arm on $B$ .

Step 3. From the intercept, $A=(G_E^2+\tau G_M^2)/(1+\tau)$ , so $G_E^2=A(1+\tau)-\tau G_M^2=0.219\times1.142-0.142\times0.923=0.250-0.131=0.119$ , giving $G_E=\mathbf{0.34}$ .

Step 4. Sanity check against the dipole: at $Q^2=0.5\ \mathrm{GeV}^2$ , $G_E^{\text{dipole}}=(1+0.5/0.71)^{-2}=0.344$ and $G_M^{\text{dipole}}=2.79\times0.344=0.96$ — the extracted values match, confirming the fit. $G_M=0.96$ from the slope, $G_E=0.34$ from the intercept.

3 · The proton charge radius from the dipole

Use the dipole form factor to estimate the proton's RMS charge radius, and compare with the modern value.

Step 1. Expand the dipole at small $Q^2$ : $G_E=(1+Q^2/0.71)^{-2}\approx 1-2Q^2/0.71+\cdots$ , so $dG_E/dQ^2|_0=-2/0.71=-2.82\ \mathrm{GeV}^{-2}$ .

Step 2. Apply equation (6): $\langle r^2\rangle=-6\,dG_E/dQ^2|_0=6\times2.82=16.9\ \mathrm{GeV}^{-2}$ . This is a squared length in natural units.

Step 3. Convert with $\hbar c=0.1973\ \mathrm{GeV\cdot fm}$ : $\sqrt{\langle r^2\rangle}=\sqrt{16.9}\times0.1973=4.11\times0.1973=\mathbf{0.81\ fm}$ . The proton is about eight-tenths of a femtometre across — a real, measurable size.

Step 4. The dipole is only an approximation; precise electron-scattering and muonic-hydrogen measurements now give $\sqrt{\langle r^2\rangle}\approx\mathbf{0.841\ fm}$ , close to the dipole estimate but with a famous few-percent controversy — the proton radius puzzle — in the small gap between methods. The dipole gives $\approx 0.81$ fm; the accepted value is $\approx 0.84$ fm.

Physics of the portrait

The formula carries the name of a graduate student who wrote it down in 1950, before any accelerator could test it — and who then spent the rest of his life on an entirely different problem, becoming one of the founders of fusion plasma physics.

Origins · 1 minMarshall Rosenbluth and how this was discovered

Marshall Rosenbluth

1927–2003 · the man who gave the proton a cross section

Marshall Nicholas Rosenbluth was born in Albany, New York, and came to the University of Chicago after the war to study under Enrico Fermi and Edward Teller — arriving in one of the great concentrations of physics talent in history. His 1950 paper, "High Energy Elastic Scattering of Electrons on Protons" (Physical Review 79, 615), was essentially a homework problem taken to its logical end: what is the most general cross section for an electron scattering elastically off a spin-half proton if the proton is allowed to have structure? He found that Lorentz invariance and current conservation permit exactly two form factors, and he wrote the cross section — the equation that now bears his name — years before Robert Hofstadter's Stanford electron-scattering experiments would use it to measure the proton's finite size and win a Nobel Prize. Remarkably, the Rosenbluth formula was a youthful sideline. Rosenbluth's enduring career was in plasma physics and controlled thermonuclear fusion, where his contributions were so foundational that he was called "the Pope of Plasma Physics"; he worked out the stability theory of confined plasmas that underlies every tokamak. He is also, through a separate 1953 collaboration at Los Alamos, a co-author of the Metropolis algorithm — the Monte Carlo method that reshaped computational science — making him one of the very few people whose name is attached to both a cornerstone of experimental particle physics and a cornerstone of statistical computation. The proton's cross section was the first thing he is remembered for and, in a sense, the least of what he did.

History

1911

— Rutherford's alpha-scattering reveals a tiny, point-like nuclear charge, the baseline for all later probes

1933

— Otto Stern measures the proton's magnetic moment near 2.5–2.8 nuclear magnetons, far from the Dirac value of 1

1950

— Marshall Rosenbluth derives the general elastic electron–proton cross section with two form factors

1955

— Hofstadter and McAllister at Stanford measure the proton's finite size, a charge radius near 0.7–0.8 fm

2000

— Jefferson Lab polarisation-transfer data reveal that G_E/G_M falls with Q², exposing two-photon-exchange effects

Going further

Texts. Thomson, Modern Particle Physics, §7.5 (The Rosenbluth formula), Eqs. 7.31–7.37, and §7.5.1–7.5.2 · Halzen & Martin, Quarks and Leptons, §8.3 (elastic electron–proton scattering) · Griffiths, Introduction to Elementary Particles, 2nd ed., §8.3 · Peskin & Schroeder, An Introduction to QFT, §6.2 (form factors) · Perkins, Introduction to High Energy Physics, 4th ed., §6.5.

Provenance. The point-charge starting point (the Mott cross section) is recalled by name from aq-pp-rutherford-mott-scattering; the general form-factor vertex from aq-pp-form-factors; the underlying $e^-p$ amplitude from aq-pp-relativistic-ep-scattering; the QED vertex and propagator from aq-pp-feynman-rules-qed; and the invariant $Q^2$ from aq-pp-mandelstam-variables. A verifier should confirm: (1) the two-form-factor structure follows from Lorentz invariance plus current conservation with $G_E(0)=1$ , $G_M(0)=2.79$ ; (2) the kinematics $E_3=E_1m_p/[m_p+E_1(1-\cos\theta)]$ and $Q^2=4E_1E_3\sin^2(\theta/2)$ ; (3) the reduced cross section is linear in $\tan^2(\theta/2)$ with slope $2\tau G_M^2$ and intercept $(G_E^2+\tau G_M^2)/(1+\tau)$ ; (4) the dipole radius $\sqrt{\langle r^2\rangle}=0.81$ fm.

Prerequisites. Rutherford and Mott Scattering · Form Factors · Relativistic Electron–Proton Scattering · Feynman Rules of QED · Mandelstam Variables. Leads to. Deep Inelastic Scattering · Probing Proton Structure · The Quark–Parton Model · The Anomalous Magnetic Moment.

Ask the tutor: Ask me to work a full Rosenbluth separation from a real data table at several angles, or to trace how the elastic form factors give way to the structure functions of deep inelastic scattering.

Why does the proton need two form factors when a spinless target needs only one?

Because the proton has spin, and a spin-half particle couples to the photon in two independent ways. A spinless target (like a pion, to good approximation) has only a charge distribution, so a single form factor $F(Q^2)$ — the Fourier transform of its charge — captures everything. A spin-half proton additionally carries a magnetic moment, and the photon can flip the proton's spin by coupling to that magnetisation. Lorentz invariance and current conservation, worked through carefully, show these are exactly the two allowed structures: an electric (charge, non-spin-flip) coupling and a magnetic (spin-flip) coupling. That is why the Rosenbluth cross section has two functions $G_E$ and $G_M$ , and why you need two measurements — the slope and the intercept of the Rosenbluth line — to pin them both down at each $Q^2$ .

What exactly is τ, and why does it multiply the magnetic term?

$\tau=Q^2/4m_p^2$ is the momentum transfer measured in units of the proton mass — a dimensionless gauge of how relativistic the proton's recoil is. At low $Q^2$ the proton barely moves and $\tau\ll 1$ ; at high $Q^2$ it recoils hard and $\tau\gg 1$ . It multiplies the magnetic form factor because magnetic (spin-flip) scattering requires the photon to carry momentum that reorients the proton's spin — an effect that grows with the momentum transfer. So the magnetic term $2\tau G_M^2\sin^2(\theta/2)$ is doubly enhanced at high $Q^2$ and backward angles, which is exactly why elastic scattering there becomes almost purely magnetic, and why the total elastic cross section eventually behaves like $G_M^2$ times kinematic factors.

How is Q² held fixed while the angle is varied, in practice?

Because $Q^2=4E_1E_3\sin^2(\theta/2)$ and $E_3$ itself depends on $E_1$ and $\theta$ , you have two knobs — beam energy $E_1$ and angle $\theta$ — and one constraint. Fixing $Q^2$ leaves a one-parameter family of $(E_1,\theta)$ settings. To reach a large $\tan^2(\theta/2)$ (backward angle) at the same $Q^2$ , you lower the beam energy; to reach forward angles you raise it. An experiment therefore takes data at several $(E_1,\theta)$ combinations all corresponding to the same $Q^2$ , each giving one point on the Rosenbluth line. The line is only as good as the range of angles you can span, which is why facilities that can run at many beam energies produce the cleanest separations — a long lever arm in $\tan^2(\theta/2)$ tightens both the slope and the intercept.

What is the proton radius puzzle, and did the Rosenbluth method get it wrong?

The proton charge radius is defined by the slope of $G_E$ at $Q^2=0$ , so measuring it means extrapolating form-factor data to zero momentum transfer. For decades, electron scattering and ordinary hydrogen spectroscopy agreed on about 0.877 fm. Then in 2010 muonic hydrogen — where a muon, being 200 times heavier, orbits 200 times closer and probes the proton's size far more sharply — gave 0.841 fm, a striking disagreement. The Rosenbluth-based scattering value was not obviously "wrong," but its extrapolation to $Q^2=0$ was more delicate than appreciated, and the low- $Q^2$ PRad experiment, designed to minimise that extrapolation, later found 0.831 fm. The consensus now sits near the smaller muonic value. The lesson is not that the formula failed but that extracting a radius from a slope demands exquisite control of the lowest- $Q^2$ data and of radiative corrections.

Why do the Rosenbluth and polarisation-transfer methods disagree at high Q²?

Both measure the ratio $G_E/G_M$ , but in different ways, and at high $Q^2$ they give different answers — Rosenbluth separation says the ratio stays roughly flat, polarisation transfer says it falls steadily. The resolution is two-photon exchange. The Rosenbluth formula assumes the electron exchanges exactly one photon with the proton; in reality it occasionally exchanges two, a small correction of order $\alpha$ . That correction happens to distort the Rosenbluth slope — which is where $G_E$ hides at high $Q^2$ , since the electric term is small there — much more than it distorts the polarisation-transfer observables, which measure a ratio directly. So the polarisation result is the more reliable one at high $Q^2$ , and the discrepancy is now understood as a real (if small) breakdown of the single-photon approximation, confirmed by comparing electron and positron scattering.

Are G_E and G_M ever literally the charge and magnetisation distributions?

Only in the low- $Q^2$ , non-relativistic corner. There $Q^2\approx\mathbf q^2$ , the proton's recoil is negligible, and you can meaningfully speak of a rest-frame charge density $\rho(r)$ and magnetisation $\mu(r)$ whose Fourier transforms are $G_E(\mathbf q^2)$ and $G_M(\mathbf q^2)$ . As $Q^2$ grows the proton recoils relativistically, so the "rest frame" in which you would define a spatial distribution is itself in motion, and the transform picture loses its footing — the object that is well-defined is $G(Q^2)$ , a function of the invariant momentum transfer, not of a static shape. Modern work reframes the spatial interpretation using light-front variables or the Breit frame, but the honest statement is that "the proton's charge radius" means "the slope of $G_E$ at the origin," and everything beyond that is model-dependent.