Form factors & charge distributions

Rutherford and Mott built their scattering formulae on a fiction that works astonishingly well: that the target is a mathematical point. Fire an electron at a proton, though, and the prediction fails — the measured cross section falls below the Mott curve, and faster the harder you hit. The reason is that the proton is not a point. Its charge is smeared over a region roughly a femtometre across, and an electron scattering from it is really scattering from every part of that cloud at once. The wavelets returning from different parts of the distribution arrive with different phases, and when those phases spread across more than a wavelength they interfere destructively and drain the amplitude. The single number that packages this interference is the form factor $F(\mathbf{q}^2)$ . The beautiful result of this article is that the form factor is not a messy correction bolted onto the point-like answer — it is the answer, cleanly factorised. In the Born approximation the amplitude off an extended charge $Q\rho(\mathbf{r})$ separates exactly into the point-charge amplitude times $F(\mathbf{q}^2)=\int\rho(\mathbf{r})\,e^{i\mathbf{q}\cdot\mathbf{r}}\,d^3r$ , the three-dimensional Fourier transform of the charge shape. Measure how the cross section deviates from Mott across a range of momentum transfers and you are literally measuring the Fourier components of the proton's charge cloud — the experiment that won Robert Hofstadter the Nobel Prize and first put a size on the proton. This article derives that factorisation, extracts the charge radius from the low- $\mathbf{q}$ slope, and follows it to the two relativistic form factors the modern proton demands.

Key takeaways

Point-like Mott and Rutherford cross sections fail for extended targets; a form factor $F(\mathbf{q}^2)$ multiplies the point result for finite size.
$F(\mathbf{q}^2)=\int\rho(\mathbf{r})\,e^{i\mathbf{q}\cdot\mathbf{r}}\,d^3r$ is the Fourier transform of the normalised charge density, with $F(0)=1$ ; the cross section is $\dfrac{d\sigma}{d\Omega}=\left(\dfrac{d\sigma}{d\Omega}\right)_{\text{Mott}}\big|F(\mathbf{q}^2)\big|^2$ , the extra factor being pure interference between wavelets from different points of the cloud.
The slope at zero gives the size: $F\approx1-\tfrac16\mathbf{q}^2\langle r^2\rangle$ , so $\langle r^2\rangle=-6\,dF/d\mathbf{q}^2\big|_0$ — the charge radius comes straight out of the data.
An exponential cloud gives the dipole $F=(1+\mathbf{q}^2a^2)^{-2}$ ; the proton follows $G_E(Q^2)=(1+Q^2/0.71\,\text{GeV}^2)^{-2}$ , so $r_p\approx0.81$ fm — but the relativistic proton really needs two invariant form factors $G_E,G_M$ (Rosenbluth).
Key idea: an extended charge distribution scatters like a point charge multiplied by the Fourier transform of its shape — so measuring how the cross section deviates from the point-like prediction directly reveals the spatial size and structure of the target.

Where this lives today

Form factors are the working language of hadron structure at every electron-scattering laboratory. At Jefferson Lab, precision measurements of $G_E$ and $G_M$ by polarisation transfer — cleaner than the classic Rosenbluth separation — revealed in the early 2000s that $G_E/G_M$ falls with $Q^2$ far faster than everyone had assumed, reshaping models of the proton's charge and magnetisation currents. The charge radius, defined as the form-factor slope, sits at the centre of the proton radius puzzle: the 2010 muonic-hydrogen result ( $r_p=0.841$ fm) disagreed by several standard deviations with the then-accepted electron value near $0.877$ fm, and reconciling the two — through new scattering data, hydrogen spectroscopy, and low- $Q^2$ experiments like PRad — has driven a decade of work now converging near $0.84$ fm. Pushed to large momentum transfer and inelastic final states, the same technique becomes deep inelastic scattering, where elastic form factors give way to structure functions and the proton's point-like quarks come into view — the direct descendant of Hofstadter's ruler.

The physical picture

An electron does not scatter off "the proton" as a lump; it scatters off every volume element of the charge cloud, and the amplitude is the coherent sum of all those little scatterings. Two elements separated by $\mathbf{r}$ send wavelets to the detector along paths differing by $\mathbf{q}\cdot\mathbf{r}$ , so a wavelet from position $\mathbf{r}$ carries phase $e^{i\mathbf{q}\cdot\mathbf{r}}$ . When the momentum transfer is small — a soft, glancing collision — the phase barely varies across the cloud and every wavelet adds in step, so the target looks point-like and $F\to1$ . When $\mathbf{q}\cdot\mathbf{r}$ swings through a radian or more across the distribution, the wavelets dephase, interfere destructively, and suppress the amplitude. The first figure fixes that geometry.

Figure

Coherent scattering from an extended charge cloud ρ(r): wavelets leaving two volume elements separated by r pick up a relative phase q·r, where the momentum transfer is q = k_i − k_f with magnitude q = 2k sin(θ/2). Summing e^{iq·r} weighted by ρ over the whole cloud gives the form factor F(q²) = ∫ ρ(r) e^{iq·r} d³r

The payoff is that this abstract interference factor is a curve you can measure and read like a ruler. Sweep the electron energy and angle across a range of momentum transfers, divide by Mott, and square-root: you have $|F(\mathbf{q}^2)|$ point by point. The gentle initial droop encodes the size of the charge cloud — a steeper droop means a bigger proton — and its detailed shape the radial profile. The second figure is what a nuclear physicist actually plots: the elastic form factor of the proton, whose smooth dipole fall-off is the fingerprint of an exponential cloud about $0.8$ fm across.

Figure

The measured proton electric form factor follows the dipole shape G_E(Q²) = (1 + Q²/0.71 GeV²)⁻², falling from 1 at Q² = 0 to 0.25 at Q² = 0.71 GeV². The tangent at the origin has slope dG_E/dQ² = −2.82 GeV⁻², and the relation ⟨r²⟩ = −6 dG_E/dQ² at Q² = 0 turns that slope into the charge radius r_p ≈ 0.81 fm

The core equations

The definition is the whole idea in one line: the form factor is the Fourier transform of the normalised charge density $\rho(\mathbf{r})$ with respect to $\mathbf{q}$ . Normalising $\rho$ to unit charge fixes $F(0)=1$ , so all the finite-size physics is in how $F$ falls away from one.

F(\mathbf{q}^2)=\int\rho(\mathbf{r})\,e^{i\mathbf{q}\cdot\mathbf{r}}\,d^3r,\qquad \int\rho(\mathbf{r})\,d^3r=1 \qquad (1)

where: $\rho(\mathbf{r})$ — charge density normalised to unit integral; $\mathbf{q}=\mathbf{k}_i-\mathbf{k}_f$ — three-momentum transferred; $\mathbf{q}^2=|\mathbf{q}|^2$ . Reading it: $F$ is the phase-weighted sum of $e^{i\mathbf{q}\cdot\mathbf{r}}$ over the cloud; at $\mathbf{q}=0$ all phases are one and $F=1$ , the total charge. Because the amplitude factorises (Derivation A), the cross section is the point-like Mott result times the squared modulus of the form factor — all the finite-size information in one factor.

\frac{d\sigma}{d\Omega}=\left(\frac{d\sigma}{d\Omega}\right)_{\text{Mott}}\big|F(\mathbf{q}^2)\big|^2 \qquad (2)

where: $(d\sigma/d\Omega)_{\text{Mott}}$ — the point-charge cross section recalled from aq-pp-rutherford-mott-scattering; $|F(\mathbf{q}^2)|^2$ — dimensionless suppression, unity for a point. Reading it: dividing measured data by Mott and square-rooting isolates $|F|$ — the route from data to charge structure. At the amplitude level the statement is cleaner still: the Born matrix element for the extended target is the point-target one times $F$ , with the momentum transfer set by the beam wavenumber and angle for elastic (no-recoil) kinematics.

M_{fi}=M_{fi}^{\text{pt}}\,F(\mathbf{q}^2),\qquad |\mathbf{q}|=2k\sin\frac{\theta}{2} \qquad (3)

where: $M_{fi}^{\text{pt}}$ — point-charge matrix element (Rutherford/Mott); $k=|\mathbf{k}_i|=|\mathbf{k}_f|$ — electron wavenumber; $\theta$ — scattering angle. Reading it: higher angle and energy both raise $|\mathbf{q}|$ , so probing small structure demands high momentum transfer — resolving power $\sim1/|\mathbf{q}|$ . Expanding the exponential for small momentum transfer turns the leading behaviour of $F$ into a direct measurement of the mean-square charge radius — the key consequence of equation (1).

F(\mathbf{q}^2)=1-\frac{1}{6}\mathbf{q}^2\langle r^2\rangle+\cdots,\qquad \langle r^2\rangle=-6\,\frac{dF}{d\mathbf{q}^2}\bigg|_{\mathbf{q}^2=0} \qquad (4)

where: $\langle r^2\rangle=\int r^2\rho(\mathbf{r})\,d^3r$ — mean-square charge radius; the linear term vanishes for spherical $\rho$ , so $\mathbf{q}^2$ is the leading correction. Reading it: the initial slope of $F$ , extrapolated to zero, is $-\tfrac16\langle r^2\rangle$ — the radius is read off the tangent at the origin, no model required. A realistic exponential charge cloud produces the dipole form factor, and the measured proton follows it closely — the empirical starting point of nucleon structure.

\rho(r)\propto e^{-r/a}\ \Longrightarrow\ F(\mathbf{q}^2)=\frac{1}{(1+\mathbf{q}^2a^2)^2},\qquad G_E(Q^2)=\left(1+\frac{Q^2}{0.71\,\text{GeV}^2}\right)^{-2} \qquad (5)

where: $a$ — cloud scale length; $Q^2=-q^2>0$ — invariant squared momentum transfer, $=|\mathbf{q}|^2$ in the no-recoil limit; $M_V^2=0.71\,\text{GeV}^2$ — empirical dipole mass-squared. Reading it: one parameter fixes the whole curve, and $\langle r^2\rangle=12a^2$ gives $r_p\approx0.81$ fm (Worked example 1). The single- $F$ picture breaks down once the proton recoils relativistically and its magnetic moment enters: the full elastic $e\text{-}p$ amplitude needs two invariant Sachs form factors, developed in the Rosenbluth article.

G_E(0)=1,\qquad G_M(0)=\mu_p=2.793,\qquad G_{E,M}(Q^2)\ \text{separated by Rosenbluth} \qquad (6)

where: $G_E,G_M$ — electric and magnetic form factors, functions of $Q^2$ alone; $\mu_p$ — proton magnetic moment in nuclear magnetons. Reading it: $G_E(0)=1$ counts the proton's unit charge, $G_M(0)=2.793$ its anomalous magnetic moment — two numbers no point Dirac particle would carry.

Derivation AFactorising the Born amplitude into point scattering times a shape factor

The goal here is to prove the central claim rigorously: in first-order perturbation theory the amplitude off an extended charge distribution is exactly the point-charge amplitude times the Fourier transform of the charge shape, with no approximation beyond the Born one itself. We start from the potential of a smeared charge, insert it into the Born matrix element between plane-wave states, and show a single change of variables cleaves the double integral into two independent pieces — one reconstructing the point scattering, one the form factor.

Step 1. Write the potential as a superposition of Coulomb pieces from each element of the target charge $Q\rho(\mathbf{r}')$ : $V(\mathbf{r})=\int\frac{Q\rho(\mathbf{r}')}{4\pi|\mathbf{r}-\mathbf{r}'|}\,d^3r'$ , with $\int\rho\,d^3r'=1$ . Why: electrostatics is linear, so an extended charge's potential is the sum of its elements' Coulomb potentials — superposition, no new physics.

Step 2. Form the first-order Born matrix element between plane waves, $M_{fi}=\int e^{-i\mathbf{k}_f\cdot\mathbf{r}}\,V(\mathbf{r})\,e^{i\mathbf{k}_i\cdot\mathbf{r}}\,d^3r=\int V(\mathbf{r})\,e^{i\mathbf{q}\cdot\mathbf{r}}\,d^3r$ , with $\mathbf{q}=\mathbf{k}_i-\mathbf{k}_f$ . Why: the Born amplitude is the Fourier transform of the potential at the momentum transfer, so all the angular dependence is carried by $\mathbf{q}$ .

Step 3. Change variables to $\mathbf{R}=\mathbf{r}-\mathbf{r}'$ ( $d^3r=d^3R$ ), so the exponential splits as $e^{i\mathbf{q}\cdot\mathbf{r}}=e^{i\mathbf{q}\cdot\mathbf{R}}\,e^{i\mathbf{q}\cdot\mathbf{r}'}$ and the double integral becomes a product $M_{fi}=\left[\int\frac{Q\,e^{i\mathbf{q}\cdot\mathbf{R}}}{4\pi R}\,d^3R\right]\left[\int\rho(\mathbf{r}')\,e^{i\mathbf{q}\cdot\mathbf{r}'}\,d^3r'\right]$ . Why: because the Coulomb kernel depends only on the difference, the shift decouples the electron-to-cloud geometry ( $\mathbf{R}$ ) from the cloud's internal structure ( $\mathbf{r}'$ ) — the heart of the factorisation.

Step 4. The first bracket is exactly the Born amplitude for a point charge $Q$ (the transform of $1/4\pi R$ is $1/\mathbf{q}^2$ ); the second is the form factor $F(\mathbf{q}^2)=\int\rho(\mathbf{r}')e^{i\mathbf{q}\cdot\mathbf{r}'}d^3r'$ . Hence $M_{fi}=M_{fi}^{\text{pt}}\,F(\mathbf{q}^2)$ , and squaring gives equation (2). Why: point-charge scattering and shape information were never entangled, so the finite-size effect is a clean multiplicative form factor, as claimed.

Go deeper · Derivation B · 1 minThe low-q expansion and the mean-square charge radius

Derivation BThe low-q expansion and the mean-square charge radius

This derivation extracts the most useful number in the subject — the charge radius — from the form factor near zero momentum transfer, showing why it appears as the slope of $F$ at the origin. The strategy is to Taylor-expand the exponential in $F$ , angular-average for a spherical cloud so the odd term drops out, and read off the coefficient of $\mathbf{q}^2$ as one-sixth of the mean-square radius — a model-independent extraction needing only low- $\mathbf{q}$ data.

Step 1. Expand $e^{i\mathbf{q}\cdot\mathbf{r}}=1+i\,\mathbf{q}\cdot\mathbf{r}-\tfrac12(\mathbf{q}\cdot\mathbf{r})^2+\cdots$ inside $F(\mathbf{q}^2)=\int\rho(\mathbf{r})\,e^{i\mathbf{q}\cdot\mathbf{r}}\,d^3r$ . Why: at small $\mathbf{q}$ only the coarsest features — total charge and overall size — survive, so a few terms of the expansion capture everything measurable.

Step 2. Average over angles for spherical $\rho(r)$ : the zeroth term gives $\int\rho\,d^3r=1$ ; the linear term vanishes since $\rho$ is even in $\mathbf{r}$ ; and the angular average of $(\mathbf{q}\cdot\mathbf{r})^2$ is $\tfrac13\mathbf{q}^2r^2$ . Why: a spherical cloud has no preferred direction, so the linear term vanishes and the first correction must be quadratic and isotropic.

Step 3. Collect the survivors: $F(\mathbf{q}^2)=1-\tfrac16\,\mathbf{q}^2\langle r^2\rangle+\cdots$ , where $\langle r^2\rangle\equiv\int r^2\rho\,d^3r$ is the mean-square charge radius. Why: this is equation (4) — the leading deviation of $F$ from unity is set by a single moment of the charge distribution, its mean-square radius.

Step 4. Differentiate to express the radius as a slope, $\langle r^2\rangle=-6\,dF/d\mathbf{q}^2\big|_{\mathbf{q}^2=0}$ : fit $F$ at small $\mathbf{q}^2$ , read the tangent's slope, multiply by $-6$ . Why: this turns an abstract Fourier statement into a lab recipe — no assumption about the shape of $\rho$ , only data low enough in $\mathbf{q}^2$ to see the slope.

Go deeper · Derivation C · 1 minFrom an exponential charge cloud to the dipole form factor

Derivation CFrom an exponential charge cloud to the dipole form factor

Here we compute a form factor end-to-end for the physically realistic case — an exponentially falling charge density — and show it produces the celebrated dipole shape $F=(1+\mathbf{q}^2a^2)^{-2}$ the proton obeys, then check its radius against Derivation B's slope formula. The route reduces the 3D transform to a radial integral, evaluates a standard exponential integral, and simplifies.

Step 1. Take $\rho(r)=\frac{1}{8\pi a^3}e^{-r/a}$ (normalised, since $\int_0^\infty r^2e^{-r/a}\,dr=2a^3$ ). For spherical $\rho$ the transform reduces to $F(\mathbf{q}^2)=\frac{4\pi}{q}\int_0^\infty r\,\rho(r)\sin(qr)\,dr$ , $q=|\mathbf{q}|$ . Why: the angular part of $e^{i\mathbf{q}\cdot\mathbf{r}}$ integrates to $\frac{\sin qr}{qr}$ for a spherical cloud, collapsing a volume integral to a one-dimensional radial one.

Step 2. Insert $\rho$ : $F=\frac{1}{2a^3q}\int_0^\infty r\,e^{-r/a}\sin(qr)\,dr$ , and the standard result $\int_0^\infty r\,e^{-br}\sin(qr)\,dr=\frac{2bq}{(b^2+q^2)^2}$ with $b=1/a$ gives $F=\frac{1}{2a^3q}\cdot\frac{2q/a}{(1/a^2+q^2)^2}$ . Why: this is a textbook Laplace-type integral — the exponential cloud is chosen precisely because it integrates in closed form.

Step 3. Multiply through by $a^4$ to get $F(\mathbf{q}^2)=\frac{1}{(1+q^2a^2)^2}$ , the dipole form factor. Why: the squared Lorentzian is the signature of an exponential charge distribution, and its slow $\sim q^{-4}$ tail is the gentle fall-off seen in the proton data.

Step 4. Cross-check: expanding $(1+q^2a^2)^{-2}=1-2a^2q^2+\cdots$ against $F=1-\tfrac16\mathbf{q}^2\langle r^2\rangle$ gives $\langle r^2\rangle=12a^2$ , which also follows from $\int r^2\rho\,d^3r=\frac{1}{2a^3}\int_0^\infty r^4e^{-r/a}\,dr=12a^2$ . Why: two independent routes to $\langle r^2\rangle=12a^2$ agree, confirming the dipole result and Derivation B's slope formula at once.

Limiting cases

Case	Result	Meaning
Long wavelength, $\mathbf{q}\to0$	$F\to1$	probe cannot resolve structure; recovers point-like Mott
Low $\mathbf{q}^2$	$F\approx1-\tfrac16\mathbf{q}^2\langle r^2\rangle$	initial slope measures the charge radius, model-independently
Point charge, $\rho=\delta^3(\mathbf{r})$	$F=1$ for all $\mathbf{q}$	electron and quark stay structureless to the highest $\mathbf{q}^2$ probed
Exponential cloud (proton)	$F=(1+\mathbf{q}^2a^2)^{-2}$	dipole fall-off $\sim q^{-4}$ ; smooth, no zeros — the proton shape
Uniform sphere (nucleus)	$F$ has zeros at $qR=4.49,\,7.73,\dots$	diffraction minima locate the sharp nuclear edge
Relativistic $e\text{-}p$	two form factors $G_E,G_M$	proton spin and magnetic moment force Rosenbluth separation

Common trapReading F(q²) as a fundamental, Lorentz-invariant property of the proton

It is tempting to treat the form factor as an intrinsic, frame-independent attribute of the target, but the $F(\mathbf{q}^2)$ derived here is nothing so grand: it is explicitly the non-invariant three-dimensional Fourier transform of a static charge distribution, valid only in the Born approximation and only when the target recoils negligibly. Two things spoil the naive picture for a real proton. First, the proton recoils relativistically, so there is no single frame in which $\rho(\mathbf{r})$ is simply "the charge density" — the clean Fourier interpretation survives only in the special Breit (brick-wall) frame. Second, the proton has spin and a large anomalous magnetic moment, so one scalar function cannot capture the elastic amplitude; you need two invariant Sachs form factors $G_E(Q^2)$ and $G_M(Q^2)$ , functions of $Q^2=-q^2$ , disentangled by Rosenbluth. The static $F(\mathbf{q}^2)$ is the correct first step, but a low-energy shadow of the invariant form factors, not the final word.

Worked examples

1 · The proton charge radius from the dipole form factor (fully worked)

The proton's electric form factor is well described by the dipole $G_E(Q^2)=(1+Q^2/M_V^2)^{-2}$ with $M_V^2=0.71\,\text{GeV}^2$ . Find the root-mean-square charge radius in femtometres.

Step 1. Get the low- $Q^2$ slope by expanding the dipole, $G_E=1-\dfrac{2Q^2}{M_V^2}+\cdots$ , so $\dfrac{dG_E}{dQ^2}\Big|_{0}=-\dfrac{2}{0.71}=\mathbf{-2.82\ GeV^{-2}}$ — the tangent on the second figure.

Step 2. Apply the slope-to-radius formula (4): $\langle r^2\rangle=-6\,\dfrac{dG_E}{dQ^2}\Big|_0=\dfrac{12}{M_V^2}=\dfrac{12}{0.71}=\mathbf{16.9\ GeV^{-2}}$ .

Step 3. Convert with $(\hbar c)^2=0.03894\,\text{GeV}^2\text{fm}^2$ : $\langle r^2\rangle=16.9\times0.03894=\mathbf{0.658\ fm^2}$ .

Step 4. Square-root: $r_p=\sqrt{0.658}=\mathbf{0.811\ fm}$ . This is the classic dipole value, within a few percent of modern precision measurements ( $r_p=0.841$ fm from muonic hydrogen) — the small gap being exactly the detail the dipole approximation cannot capture. The dipole form factor implies $r_p\approx0.81$ fm — the proton is about $0.8$ femtometres across.

2 · Momentum transfer, resolution, and how much the finite size bites

A $500\ \text{MeV}$ electron scatters elastically off a proton at $\theta=60^\circ$ . Find $|\mathbf{q}|$ , the spatial resolution, and how much the finite proton size suppresses the cross section relative to Mott.

Step 1. The electron is ultra-relativistic, so $k\approx E=0.5\ \text{GeV}$ and $|\mathbf{q}|=2k\sin(\theta/2)=2(0.5)\sin30^\circ=\mathbf{0.50\ GeV}$ ; the invariant is $Q^2\approx|\mathbf{q}|^2=\mathbf{0.25\ GeV^2}$ in the no-recoil limit.

Step 2. The resolving power is $\sim\hbar c/|\mathbf{q}|=0.1973/0.50=\mathbf{0.39\ fm}$ — comparable to the proton radius, so this collision sees inside the charge cloud.

Step 3. Evaluate: $G_E(0.25)=(1+0.25/0.71)^{-2}=(1.352)^{-2}=\mathbf{0.547}$ .

Step 4. The cross section carries $|G_E|^2=(0.547)^2=\mathbf{0.30}$ : the measured rate is only about $30\%$ of the point-like Mott prediction. At $Q^2=0.25\ \text{GeV}^2$ the finite size has already knocked the elastic cross section down to a third of the point value — the interference is not a small correction.

3 · Diffraction minima for a nucleus (why the proton has none)

A uniform charge sphere of radius $R$ has $F(q)=\dfrac{3}{(qR)^3}\big[\sin qR-qR\cos qR\big]$ , vanishing at $qR=4.49,7.73,\dots$ Estimate the first minimum for a medium nucleus with $R\approx4.1\ \text{fm}$ .

Step 1. Set $qR=4.49$ : $q=4.49/4.1\,\text{fm}=\mathbf{1.10\ fm^{-1}}$ .

Step 2. Convert: $|\mathbf{q}|=1.10\,\text{fm}^{-1}\times0.1973\,\text{GeV}\,\cdot\,\text{fm}=\mathbf{0.22\ GeV}$ . A sharp dip near this momentum transfer is exactly what Hofstadter saw for nuclei like $^{12}$ C and $^{40}$ Ca, its position pinning down the nuclear radius. Nuclei show diffraction minima because they have a fairly sharp edge; the proton's soft exponential cloud has none — its dipole $F$ never crosses zero, itself evidence of no hard surface.

Physics of the portrait

The form factor turned an abstract Fourier transform into a Nobel-winning ruler, wielded by the man who first gave the proton a size.

Origins · 1 minRobert Hofstadter and how this was discovered

Robert Hofstadter

1915–1990 · electron scattering and the structure of the nucleon

Robert Hofstadter took Rutherford's point-nucleus fiction and demolished it with electrons. Working at Stanford in the early 1950s, he realised the new high-energy electron linac (the precursor of SLAC) was the ideal probe of nuclear charge: an electron interacts only electromagnetically, through the well-understood photon, so any deviation from the Mott point-charge prediction must come from the shape of the target's charge. By measuring cross sections across a range of angles and energies — across momentum transfer — and dividing out the Mott factor, he read off $F(\mathbf{q}^2)$ and Fourier-inverted it to map the charge distributions of nuclei from helium to lead, with their radii and surface thicknesses. In 1955–56, with Robert McAllister, he turned the beam on hydrogen and found the proton itself is not a point: its cross section fell below Mott just as a spread-out charge demands, giving a charge radius of about $0.7$ – $0.8$ fm and, for the first time, a number for the size of the proton. Hofstadter shared the 1961 Nobel Prize for this work; his method, pushed higher in energy, would a decade later reveal the point-like partons inside the proton and launch the quark era.

History

1911

— Rutherford's point-nucleus model fits alpha scattering, defining the point-like baseline

1926

— Born's approximation links the scattering amplitude to the Fourier transform of the potential

1956

— McAllister and Hofstadter measure the proton's finite charge radius, about 0.7 fm

1961

— Hofstadter awarded the Nobel Prize for elastic electron scattering off nuclei and nucleons

2010

— muonic-hydrogen spectroscopy gives r_p = 0.841 fm, opening the proton-radius puzzle

Going further

Texts. Thomson, Modern Particle Physics, §7.3 (form factors) and §7.2 (Mott scattering), the primary source here · Halzen & Martin, Quarks and Leptons, §8.1–8.2 (elastic $e$ - $p$ scattering, Rosenbluth) · Griffiths, Introduction to Elementary Particles, 2nd ed., §8 · Perkins, Introduction to High Energy Physics, 4th ed., §6 · Povh et al., Particles and Nuclei, Ch. 5–6.

Provenance. The Mott cross section that $|F|^2$ multiplies is recalled from aq-pp-rutherford-mott-scattering; the Born matrix element and its Fourier structure from aq-pp-perturbation-theory-exchange and aq-pp-nonrel-qm-review. The invariant Sachs form factors $G_E(Q^2)$ , $G_M(Q^2)$ and their separation are owned by aq-pp-rosenbluth-formula and aq-pp-relativistic-ep-scattering; the inelastic continuation to structure functions by aq-pp-probing-proton-structure. A verifier should confirm: (1) the factorisation $M_{fi}=M_{fi}^{\text{pt}}F$ via the $\mathbf{R}=\mathbf{r}-\mathbf{r}'$ substitution; (2) $\langle r^2\rangle=-6\,dF/d\mathbf{q}^2|_0$ ; (3) $F=(1+\mathbf{q}^2a^2)^{-2}$ with $\langle r^2\rangle=12a^2$ ; (4) $r_p=\sqrt{12/M_V^2}\,\hbar c=0.811$ fm with $M_V^2=0.71$ GeV²; (5) worked example 2: $|\mathbf{q}|=0.50$ GeV, $G_E=0.547$ , $|G_E|^2=0.30$ at $E=500$ MeV, $\theta=60^\circ$ .

Prerequisites. Rutherford and Mott Scattering · Perturbation Theory and Particle Exchange · Non-relativistic QM Review. Leads to. Relativistic $e$ - $p$ Scattering · The Rosenbluth Formula · Probing Proton Structure.

Ask the tutor: Ask me to Fourier-transform a Gaussian or uniform-sphere cloud, walk through the Rosenbluth separation of $G_E$ and $G_M$ , or extract a charge radius from a real dataset.

Why should the form factor be the Fourier transform of the charge density?

Because scattering in the Born approximation is a Fourier transform. The Born amplitude equals the transform of the potential at the momentum transfer $\mathbf{q}$ , and a static charge sources its potential by superposition. When you carry the integrals through (Derivation A), the electron-to-target geometry factors off as the point-charge amplitude, leaving the transform of the charge density alone. Physically, each volume element scatters a wavelet with phase $e^{i\mathbf{q}\cdot\mathbf{r}}$ , and summing those over the cloud is what a Fourier transform does — not an analogy but the mathematical content of coherent scattering.

What is the difference between the F(q²) here and the proton's G_E and G_M?

The $F(\mathbf{q}^2)$ here is a single scalar function, the non-relativistic static-limit Fourier transform of a charge density, exact only when the target recoils negligibly and has no spin structure. The real proton recoils relativistically and carries spin-½ with a large anomalous magnetic moment, so its elastic amplitude needs two invariant functions of $Q^2=-q^2$ : the electric $G_E(Q^2)$ , which reduces to our $F$ in the low- $Q^2$ Breit-frame limit, and the magnetic $G_M(Q^2)$ for the smeared magnetic moment. Our $F$ is the shadow $G_E$ casts at low energy.

How do you actually get the charge radius out of scattering data?

Measure the elastic cross section at several small angles and energies, all low in $\mathbf{q}^2$ . Divide each by Mott and square-root to get $|F(\mathbf{q}^2)|$ . Plot against $\mathbf{q}^2$ and fit the initial slope as $\mathbf{q}^2\to0$ ; by equation (4) that slope is $-\tfrac16\langle r^2\rangle$ , so multiply by $-6$ and square-root for the rms radius. The subtlety — and the source of the proton-radius puzzle — is that you must extrapolate to exactly zero momentum transfer, where there is no data, so the answer depends on how you model the low- $\mathbf{q}^2$ shape.

Why does the proton follow a dipole shape specifically?

The dipole $G_E=(1+Q^2/0.71\,\text{GeV}^2)^{-2}$ is the Fourier transform of an exponentially falling density $\rho\propto e^{-r/a}$ (Derivation C). Empirically the proton's charge really does fall off roughly exponentially from the centre with no sharp edge — unlike a heavy nucleus with its well-defined surface. There is no deep QCD derivation of the exponential shape; the dipole is a phenomenological fit good to perhaps $10\%$ over a wide $Q^2$ range. Modern data show clear deviations, which is why the field now measures $G_E$ and $G_M$ point by point rather than assuming a shape.

What are the diffraction minima that nuclei show but the proton does not?

A charge distribution with a fairly sharp edge — a medium or heavy nucleus, well approximated by a uniform sphere — has a form factor that oscillates through zero at specific momentum transfers ( $qR=4.49,7.73,\dots$ ), producing sharp dips in the cross section. These are literally diffraction minima, the same mathematics as light through a circular aperture, and their positions give the nuclear radius directly. The proton shows none: its exponential cloud has no edge, so its dipole $F$ decreases monotonically and never reaches zero. That absence is itself information — the proton is a soft, edgeless object, not a ball with a surface.

Is the electron's form factor exactly 1?

To every measurement ever made, yes — the electron scatters as a perfect point, its form factor indistinguishable from $F=1$ up to the highest momentum transfers explored, probing distances well below $10^{-18}$ m, a thousand times smaller than the proton. This is one of the sharpest facts in particle physics: the electron, like the quarks and other leptons, has no measurable size or internal charge structure. If it were composite, $F$ would droop below 1 at high $\mathbf{q}^2$ ; the persistent flatness is a stringent limit on substructure and a cornerstone of treating leptons and quarks as fundamental point particles.

What is the proton-radius puzzle, and is it settled?

For years the proton radius from electron scattering and ordinary-hydrogen spectroscopy sat near $0.877$ fm. In 2010 a precise measurement using muonic hydrogen — where a muon orbits so close it feels the proton's size acutely — gave $0.841$ fm, disagreeing by about seven standard deviations. Because the radius is the form-factor slope at zero momentum transfer, the discrepancy put the whole extraction under scrutiny: was it the electron extrapolation to $\mathbf{q}^2=0$ , the muonic analysis, or new physics coupling to muons? A decade of work — the PRad low- $Q^2$ experiment, re-measured hydrogen transitions, and reanalyses — has largely converged near $0.84$ fm, so most now consider it a measurement/extrapolation issue rather than new physics, though a few tensions linger.