Bjorken scaling & structure functions
Fire an electron hard enough at a proton and the proton does not merely recoil — it shatters. The scattered electron emerges having dumped an energy ν and a momentum-transfer-squared Q² into a spray of hadrons X, and the reaction e⁻p → e⁻X is no longer elastic. Naively this should be hopeless. A composite, extended, strongly-interacting blob smeared over a femtometre ought to soak up the probe, so that the harder you hit it the more the cross section collapses — exactly what the elastic form factors do, falling like Q⁻⁸. The astonishment of the late-1960s SLAC–MIT experiments was that the inelastic cross section did not collapse. Scaled to the Mott point-charge cross section, the deep inelastic yield barely depended on Q² at all. A cross section that ignores the resolving power of the probe is the fingerprint of scattering off something with no size of its own: a hard, point-like constituent buried inside the proton. This article turns that single fact into the discovery of quarks. Two functions, the structure functions F₁(x,Q²) and F₂(x,Q²), carry everything an unpolarized single-photon experiment can learn about proton structure — two of them because inelastic scattering needs two independent variables, not one. Their measured near-independence of Q² at fixed x is Bjorken scaling, and it is precisely what point-like constituents predict. Feynman's parton model then identifies the scaling variable x with the fraction of the proton's momentum carried by the struck constituent, and the Callan–Gross relation F₂ = 2xF₁ pins those constituents down further: they are spin-½ Dirac particles. The proton is a bag of quarks.
- Deep inelastic e⁻p → e⁻X cross sections fall only weakly with Q², unlike the steeply-falling elastic cross section — the signature of scattering from tiny, hard, point-like objects.
- The general single-photon inelastic cross section needs two structure functions F₁(x,Q²) and F₂(x,Q²), because inelastic scattering has two independent kinematic variables (x and Q²).
- Bjorken scaling: at large Q² the structure functions depend on x alone, F_i(x,Q²) → F_i(x) — a scale-free response that only point-like constituents give.
- In Feynman's parton model the Bjorken variable x = Q²/(2m_pν) is the fraction of the proton's momentum carried by the struck parton, and F₂(x) = Σ e_i² x q_i(x).
- The Callan–Gross relation F₂(x) = 2xF₁(x) holds in data, proving the constituents are spin-½ (spin-0 partons would give 2xF₁/F₂ = 0).
- Key idea: a soft extended proton would make the cross section plunge with Q²; instead the structure functions barely move, so the probe is striking hard, point-like, spin-½ constituents — the quarks — each carrying a momentum fraction equal to Bjorken x.
If the proton were a soft, extended blob the cross section would fall fast with Q²; instead the structure functions are nearly independent of Q² at fixed x, meaning the electron scatters elastically off hard, point-like, spin-½ constituents — the quarks — and the Bjorken variable x is the fraction of the proton's momentum each struck quark carries.
Where this lives today
Bjorken scaling was only ever approximate, and its small, calculable violations became the next great triumph. At the HERA collider at DESY, the H1 and ZEUS experiments collided 27.5 GeV electrons and positrons with 920 GeV protons, reaching Q² up to ~10⁴ GeV² and x down to ~10⁻⁴ — a vast extension of the SLAC reach. There F₂ is seen to drift slowly with Q²: it rises steeply at small x and falls gently at large x as Q² grows, exactly the logarithmic scaling violation that QCD predicts through the DGLAP evolution equations, in which quarks continually radiate gluons that split into quark–antiquark sea. Measuring that drift determines the strong coupling α_s and the gluon distribution, the direct link to the running of α_s and to asymptotic freedom. The parton distribution functions extracted from this global effort — in modern fits such as CT18, MSHT20 and NNPDF — are now indispensable inputs to every LHC prediction, since a proton–proton collision is at bottom a collision of the partons those DIS experiments mapped. The programme continues: the Electron-Ion Collider now under construction at Brookhaven will image the proton's quark and gluon structure, its spin, and its transverse tomography with a precision that turns Bjorken's flat lines into a full three-dimensional portrait of the nucleon.
The physical picture
Boost the proton to enormous momentum and relativistic time dilation freezes its internal churning: for the brief instant the virtual photon is overhead, the proton is a frozen swarm of nearly free, point-like partons, each carrying some fraction of the total momentum. A photon of virtuality Q² resolves transverse distances of order ħ/√Q², so at large Q² it is a fine enough scalpel to strike a single parton rather than the proton as a whole. That parton absorbs the photon and recoils on its own; the rest of the proton sails on as spectators. The first figure fixes this picture and, with it, the one relation that makes deep inelastic scattering intelligible — that the Bjorken variable x is nothing but the struck parton's share of the proton momentum.
The second figure is the discovery itself. Elastic scattering off the whole extended proton is governed by the dipole form factor, whose square drives the cross section down like (1 + Q²/0.71)⁻⁴ — more than three orders of magnitude lost between Q² = 1 and 8 GeV². If deep inelastic scattering were the same kind of scattering off the same extended charge cloud, F₂ would plummet the same way. Instead, plotted against Q² at fixed x, F₂ lies flat. That flatness is Bjorken scaling, and it is the plot that told Friedman, Kendall and Taylor they were watching electrons bounce off something point-like inside the proton.
The core equations
Inelastic scattering needs two variables where elastic needed one. The natural covariant pair is the photon virtuality Q² and the energy loss ν, from which the dimensionless Bjorken variable x and the inelasticity y are built. In the lab frame, with the proton at rest and the electron mass neglected, all four are read off the measured scattered energy E₃ and angle θ.
where: Q² — positive four-momentum transfer squared; ν — energy transferred to the proton; x — Bjorken scaling variable, 0 ≤ x ≤ 1; y — inelasticity (fractional energy loss), 0 ≤ y ≤ 1; m_p = 0.938 GeV. Reading it: x = 1 is the elastic point (the whole proton absorbs the photon), while x < 1 measures how much smaller than the proton the struck object is. The invariant mass of the hadronic debris X distinguishes elastic scattering, the resonance bumps, and the deep inelastic continuum. It follows from squaring the total hadronic four-momentum p + q.
where: W — invariant mass of the final hadronic system X. Elastic scattering fixes W = m_p (so x = 1); the Δ(1232) and higher resonances sit at fixed W just above; the deep inelastic region is W ≳ 2 GeV, where the proton is genuinely broken up. The most general parity-conserving, single-photon-exchange cross section for e⁻p → e⁻X is fixed by Lorentz invariance and current conservation to contain exactly two functions. In the SLAC lab form the electric-type response rides cos²(θ/2) and the magnetic-type response rides sin²(θ/2), just as in elastic scattering — but now each coefficient is a function of two variables.
where: F₁, F₂ — dimensionless structure functions; the prefactor is the Mott/Rutherford envelope; the cos² term is the electric (charge) contribution and the sin² term the magnetic (spin) contribution. Varying the beam energy at fixed (x, Q²) changes the balance of the two terms and lets an experiment separate F₁ from F₂. Bjorken scaling is the empirical statement that, in the deep inelastic limit, these two functions stop depending on the resolving scale Q² and collapse onto curves in x alone.
where: the limit is taken at fixed x — the Bjorken limit. A form factor that does not depend on Q² is a Fourier transform of a point: scaling is the direct statement that the constituents have no intrinsic size. The parton model computes the scaling functions explicitly. If the proton contains partons of charge e_i with number density q_i(x) (the probability of finding parton i carrying momentum fraction x), then incoherently summing elastic electron–parton scattering gives both structure functions at once.
where: e_i — parton charge in units of e (up-type +2/3, down-type −1/3); q_i(x) — parton distribution function. Dividing the two lines gives F₂ = 2xF₁ automatically — the Callan–Gross relation — provided the partons are spin-½. That relation is the sharp spin diagnostic, isolating the fixed ratio between the magnetic and electric terms that a Dirac point particle must obey.
where: F_L — the longitudinal structure function. Spin-½ partons give F_L = 0 (the whole cross section is transverse); spin-0 partons would instead give F₁ = 0 and hence 2xF₁/F₂ = 0. The measured ratio 2xF₁/F₂ ≈ 1 settles the question: the quarks are spin-½.
The goal is to build the invariant toolkit of deep inelastic scattering — the virtuality Q², the energy loss ν, the Bjorken variable x, and the hadronic mass W — from four-momentum conservation alone, and to show that x is confined to the interval from 0 to 1 with x = 1 marking the elastic boundary. Everything rests on labelling the four-momenta and squaring the right combinations; the electron mass is dropped throughout because every experiment of interest has E₁ ≫ m_e = 0.511 MeV.
Step 1. Label the electron momenta k (in) and k′ (out), the proton p = (m_p, 0) at rest, and the exchanged photon q = k − k′. The virtuality is Q² ≡ −q² = −(k−k′)² = 2k·k′ ≈ 2E₁E₃(1 − cos θ) = 4E₁E₃ sin²(θ/2), using the half-angle identity. Why: Q² is a Lorentz invariant, so it may be evaluated in the lab where it becomes a product of the two things the spectrometer records — outgoing energy and angle.
Step 2. Define the invariant energy transfer ν ≡ p·q/m_p, which in the lab frame is just ν = E₁ − E₃, the energy the electron gives up. Form the dimensionless ratio x ≡ Q²/(2p·q) = Q²/(2m_pν). This is the Bjorken variable; being a ratio of invariants it is itself frame-independent. Why: two independent invariants are the irreducible content of inelastic scattering, and (Q², x) is the pair in which scaling is cleanest.
Step 3. Compute the hadronic mass by squaring the final hadronic momentum p + q: W² = (p+q)² = m_p² + 2p·q + q² = m_p² + 2m_pν − Q². Elastic scattering keeps the proton intact, W = m_p, which forces 2m_pν = Q², i.e. x = 1. Why: this exhibits x = 1 as the elastic endpoint — the whole proton recoils as one — so any x < 1 is scattering off a sub-object.
Step 4. Bound x. Since W² ≥ m_p² (the final state is at least a proton) we need 2m_pν − Q² ≥ 0, giving x ≤ 1; and Q², ν > 0 give x > 0. Hence 0 < x ≤ 1, with the deep inelastic continuum filling W ≳ 2 GeV, safely above the Δ(1232) resonance. Why: the physical range 0 < x ≤ 1 is exactly what the momentum-fraction interpretation of the next derivation demands — a fraction cannot exceed one.
Go deeper · Derivation B · 2 minThe parton model — why x is the struck constituent's momentum fraction
The aim is to prove Feynman's identification: the Bjorken variable x equals the fraction ξ of the proton's momentum carried by the parton that absorbs the photon, and to show that this instantly explains Bjorken scaling. The whole argument lives in the infinite-momentum frame, where the proton moves so fast that internal transverse momenta and masses are negligible and the partons are, for the duration of the interaction, free and on-shell. This is the frame in which "the photon hits one parton" becomes exact.
Step 1. Give the struck parton a fraction ξ of the proton four-momentum, so its momentum is ξp, neglecting its mass and transverse momentum against the huge longitudinal boost. It absorbs the photon q and emerges as a real, massless parton, so its final momentum squares to zero: (ξp + q)² = 0. Why: putting the final parton on-shell is the single dynamical assumption — it is the statement that the photon scatters elastically off a free constituent, not off the bound proton as a whole.
Step 2. Expand the on-shell condition: ξ²p² + 2ξ p·q + q² = 0. In the infinite-momentum frame the term ξ²p² = ξ²m_p² is negligible, and q² = −Q², so 2ξ p·q − Q² = 0. Why: dropping the m_p² term is exactly the Bjorken limit Q², ν → ∞ — the frame where the constituent picture becomes clean is the same limit in which scaling is observed.
Step 3. Solve: ξ = Q²/(2p·q) = Q²/(2m_pν) = x. The momentum fraction of the struck parton is the Bjorken variable. A measurement of x for each event is therefore a measurement of which parton was hit. Why: this is the conceptual heart of the parton model — an abstract kinematic ratio becomes a physical property of a constituent, turning the structure function into a momentum-fraction census.
Step 4. Deduce scaling. A point-like parton has no internal length scale, so its elastic response to the photon is independent of Q²; the only x-dependence left is q_i(x), the probability of finding a parton at fraction x. Summing incoherently over partons weighted by e_i² gives F₂(x) = Σ e_i² x q_i(x) — a function of x alone. Why: the absence of any scale in a point constituent is precisely why F₂ cannot depend on Q² — Bjorken scaling and point-like partons are the same statement.
Go deeper · Derivation C · 2 minThe Callan–Gross relation and the spin of the partons
The purpose is to derive F₂(x) = 2xF₁(x) and to show it is a direct fingerprint of spin-½ constituents, so that its experimental confirmation is a measurement of quark spin. The strategy is to compare the two structure functions produced by elastic scattering off a single charged Dirac parton with those produced by a spinless one, and read off how the ratio 2xF₁/F₂ discriminates between them. This is the DIS analogue of the electric-versus-magnetic split already familiar from elastic electron–proton scattering.
Step 1. Recall that for a point spin-½ target the cross section carries a fixed magnetic (spin-flip) term alongside the electric term, with their relative size set entirely by the Dirac magnetic moment g = 2 — there is no freedom to adjust it. Elastic electron–quark scattering therefore has a definite ratio of its sin²(θ/2) to its cos²(θ/2) coefficient. Why: a Dirac particle's magnetic scattering is not an independent parameter but is locked to its charge scattering, and this lock is what becomes the Callan–Gross relation.
Step 2. Translate that fixed ratio into structure functions. Matching elastic electron–quark scattering to equation (3), the magnetic coefficient becomes 2F₁/m_p and the electric coefficient F₂/ν, and the Dirac lock enforces F₂(x) = 2x F₁(x). Equivalently the longitudinal structure function F_L = F₂ − 2xF₁ vanishes: a spin-½ parton cannot absorb a longitudinally-polarized photon. Why: F_L = 0 is the crisp, model-independent form of the statement — the entire deep inelastic cross section off spin-½ quarks is transverse.
Step 3. Contrast spin-0. A scalar parton has no magnetic moment, so its sin²(θ/2) coefficient vanishes: F₁ = 0, and the ratio 2xF₁/F₂ = 0. The two spin hypotheses thus predict opposite extremes — 1 for spin-½, 0 for spin-0. Why: the observable 2xF₁/F₂ is engineered to separate the two cases maximally, making it a decisive spin measurement rather than a consistency check.
Step 4. Confront data. The SLAC measurements gave 2xF₁/F₂ ≈ 1 across the scaling region, consistent with spin-½ and flatly inconsistent with spin-0. Combined with scaling itself, this fixes the constituents as point-like spin-½ Dirac particles carrying fractional charge — the quarks. Why: this is the payoff — a ratio of cross-section pieces measured to be one is a direct readout of the quark's spin, independent of any detailed proton model.
Limiting cases
| Case | Result | Meaning |
|---|---|---|
| Elastic endpoint, x → 1 | W = m_p, single δ-function | the whole proton absorbs the photon; recovers elastic form-factor scattering |
| Bjorken limit (Q², ν → ∞, x fixed) | F_i(x, Q²) → F_i(x) | scaling; the constituents are point-like and scale-free |
| Spin-½ partons | F₂ = 2xF₁, F_L = 0 | magnetic scattering locked to charge scattering by the Dirac moment |
| Spin-0 partons | F₁ = 0, 2xF₁/F₂ = 0 | no magnetic term — ruled out by the measured ratio ≈ 1 |
| Resonance / low-Q² region | F_i strongly Q²-dependent | extended proton; form factors fall, no scaling — not yet deep inelastic |
| Real QCD | F_i(x, Q²), slow logarithmic Q² drift | scaling violations from gluon radiation (DGLAP evolution), not a failure of the picture |
Worked examples
1 · Full deep inelastic kinematics from one measured event (fully worked)
A 20.0 GeV electron scatters off a proton (m_p = 0.938 GeV) and is detected at θ = 6.0° with energy E₃ = 10.0 GeV. Find Q², ν, the Bjorken x, the inelasticity y, and the hadronic mass W, and confirm the event is deep inelastic.
Step 1. Momentum transfer from equation (1). With sin(3.0°) = 0.05234, sin²(3.0°) = 0.002739: Q² = 4 × 20.0 × 10.0 × 0.002739 = 2.19 GeV².
Step 2. Energy loss: ν = E₁ − E₃ = 20.0 − 10.0 = 10.0 GeV. Inelasticity: y = ν/E₁ = 10.0/20.0 = 0.500 — the electron surrenders half its energy.
Step 3. Bjorken variable: x = Q²/(2m_pν) = 2.19/(2 × 0.938 × 10.0) = 2.19/18.76 = 0.117. In the parton picture the electron struck a quark carrying about 12 % of the proton's momentum.
Step 4. Hadronic mass from equation (2): W² = m_p² + 2m_pν − Q² = 0.880 + 18.76 − 2.19 = 17.45 GeV², so W = 4.18 GeV. This is far above the resonance region (W ≳ 2 GeV), so the proton is genuinely shattered.
Step 5. Interpret. A single event — one energy and one angle — delivers the full invariant set (Q², x, y, W). Repeating over many events at fixed x while varying Q² traces out F₂(x, Q²); the observed flatness is Bjorken scaling. Q² = 2.19 GeV², ν = 10.0 GeV, x = 0.117, y = 0.500, W = 4.18 GeV — a clean deep inelastic event off a quark carrying ≈ 12 % of the proton momentum.
2 · The momentum sum rule and the missing half
Using the parton-model structure function F₂(x) = Σ e_i² x q_i(x), and the measured valence-quark momentum integrals ∫₀¹ x[u(x)+ū(x)] dx ≈ 0.36 and ∫₀¹ x[d(x)+d̄(x)] dx ≈ 0.18 for the proton, compute the total quark momentum fraction and ∫₀¹ F₂^{ep}(x) dx. What do the results reveal?
Step 1. Total momentum carried by quarks and antiquarks: add the two integrals, ∫ x[u+ū] + ∫ x[d+d̄] = 0.36 + 0.18 = 0.54. Including the smaller strange-sea contribution the accepted value settles near 0.5.
Step 2. The photon couples to charge, so F₂^{ep}(x) = x[4/9(u+ū) + 1/9(d+d̄) + ...]. Integrate over x using the given moments: ∫ F₂^{ep} dx = 4/9(0.36) + 1/9(0.18) = 0.160 + 0.020 = 0.18.
Step 3. Read the physics. If quarks carried all the proton's momentum, Step 1 would give 1.0; instead it gives ≈ 0.5. Roughly half the proton's momentum is carried by constituents that do not couple to the photon — electrically neutral gluons, invisible to deep inelastic scattering but demanded by the momentum bookkeeping.
Step 4. Cross-check the charge weighting with the large-x neutron/proton ratio. Isospin gives F₂^{en} = x[4/9(d+d̄) + 1/9(u+ū)], so as x → 1 (valence up-quark dominance, d/u → 0) the ratio F₂^{en}/F₂^{ep} → (1/9)/(4/9) = 1/4, a clean prediction of fractional quark charges borne out by data. Quarks carry only ≈ 50 % of the proton momentum — the discovery of gluons hides in the missing half — and the charge-weighted structure functions confirm the +2/3, −1/3 quark charges.
Physics of the portrait
Bjorken scaling carries its author's name because the flatness of the structure functions was a prediction, wrung from the abstract machinery of current algebra before anyone drew a single parton. The man who made it was James Bjorken, and his 1969 result told the SLAC experimenters exactly what to look for.
Origins · 1 minJames Bjorken and how this was discovered
James Daniel Bjorken — "BJ" to a generation of theorists — was among the first to take seriously the idea that the proton's deep inelastic response should simplify at high energy. Working at SLAC in the late 1960s with the tools of current algebra and sum rules, he argued in 1969 that in the limit of large Q² and large ν, taken with x = Q²/(2m_pν) held fixed, the structure functions must lose their separate dependence on Q² and ν and become functions of the single variable x alone. This was Bjorken scaling, and it was a bold claim about a regime no one had cleanly measured. When the SLAC–MIT team of Friedman, Kendall, Taylor and collaborators plotted their inelastic e⁻p data, the structure functions were indeed flat in Q², precisely as predicted. Richard Feynman, visiting SLAC, supplied the picture that made the result physical — the parton model, in which the proton is a swarm of point-like constituents and Bjorken's x is the fraction of momentum each carries — and Callan and Gross showed that the relation F₂ = 2xF₁ fixed those partons as spin-½. Bjorken's insight thus sits at the hinge of the story: it turned a pile of cross-section data into the first hard evidence that quarks, introduced by Gell-Mann and Zweig as bookkeeping devices, are real dynamical objects inside the proton. He went on to seminal work on jets, on CP violation, and on the light-cone structure of field theory, but scaling remains the idea that carries his name into every particle physics course.
Going further
Texts. Thomson, Modern Particle Physics, §8.2–8.2.1 (deep inelastic scattering, structure functions, Bjorken scaling and the Callan–Gross relation) — the primary source for the notation and lab-frame cross section used here · Halzen & Martin, Quarks and Leptons, §8 (the parton model, x as momentum fraction, F₂ = 2xF₁) · Griffiths, Introduction to Elementary Particles, 2nd ed., §8.3 (deep inelastic scattering and partons) · Peskin & Schroeder, An Introduction to Quantum Field Theory, §17.3–17.5 (the parton distributions and scaling violations) · Schwartz, Quantum Field Theory and the Standard Model, §32 (DIS and the parton model). The key results to keep in mind: the kinematic relations Q² = 4E₁E₃sin²(θ/2), x = Q²/2m_pν, W² = m_p² + 2m_pν − Q², and the range 0 < x ≤ 1; the parton-model identification ξ = x from (ξp + q)² = 0 in the infinite-momentum frame; the structure functions F₂ = Σ e_i² x q_i(x), F₁ = ½ Σ e_i² q_i(x) and hence F₂ = 2xF₁; the worked event E₁ = 20 GeV, θ = 6°, E₃ = 10 GeV giving Q² = 2.19 GeV², x = 0.117, W = 4.18 GeV; and the momentum sum ≈ 0.5 with ∫F₂^{ep}dx ≈ 0.18 from the quoted valence moments.
Prerequisites. Deep Inelastic Scattering · The Rosenbluth Formula · Form Factors and Charge Distributions. Leads to. The Quark–Parton Model · Parton Distribution Functions · The Running of α_s.
Ask the tutor: Ask me to run the full deep inelastic kinematics for any beam energy, angle, and scattered energy, or to walk the parton model from Bjorken's x all the way to F₂ = 2xF₁ and the momentum sum rule.
Is the Bjorken variable x really the fraction of the proton's momentum carried by the struck quark?
Yes, and that identification is the whole point of the parton model. Kinematically x = Q²/(2m_pν) is just a ratio of measured invariants, defined for any inelastic event. But if you boost the proton to very high momentum and demand that the struck parton be free and on-shell after absorbing the photon — (ξp + q)² = 0 — the algebra forces ξ = x. So the abstract variable and the physical momentum fraction coincide. This is why plotting F₂ against x is so powerful: you are effectively histogramming the proton's constituents by how much momentum they carry. The interpretation is exact only in the Bjorken limit; at finite Q² there are calculable corrections, but the picture is remarkably robust.
What is the difference between structure functions and form factors?
Form factors G_E(Q²) and G_M(Q²) describe elastic scattering, where the proton stays intact, and they genuinely are Fourier transforms of its static charge and magnetization distributions — one variable, Q², and a characteristic fall-off that measures the proton's size. Structure functions F₁(x, Q²) and F₂(x, Q²) describe inelastic scattering, where the proton breaks up, and they depend on two variables. They are not spatial Fourier transforms at all; they are a dynamic census of the proton's momentum, telling you the probability of finding a constituent carrying a fraction x of the momentum. The tell-tale difference is the Q² behaviour: form factors fall steeply as the probe sharpens, while structure functions barely move — which is exactly how you know you have stopped resolving the whole proton and started resolving its parts.
Why does Q²-independence prove the constituents are point-like?
A form factor is Q²-independent only if the object it describes has no size. Any charge distribution of finite extent produces a form factor that falls off once the probe's wavelength ħ/√Q² becomes smaller than the object — the sharper the probe, the more the extended structure "washes out" the coherent scattering. So a structure function that stays flat as Q² climbs is telling you the scattering centre remains unresolved no matter how fine the probe: it has no internal length scale. That is the operational definition of point-like. The elastic proton form factor, which plunges by orders of magnitude over the same Q² range, is the control experiment — it shows what an extended object looks like, and the flat structure functions look nothing like it.
What exactly does the Callan–Gross relation tell us?
It tells us the quarks are spin-½. The relation F₂ = 2xF₁ says the magnetic (spin) part of the deep inelastic cross section is locked in a fixed ratio to the electric (charge) part — the ratio that a Dirac particle with g = 2 is obliged to have. A spin-0 constituent has no magnetic moment, so it would give F₁ = 0 and the ratio 2xF₁/F₂ = 0; a spin-½ constituent gives 2xF₁/F₂ = 1. The measured value sits at about 1, so the constituents are spin-½. Equivalently, the longitudinal structure function F_L = F₂ − 2xF₁ vanishes for spin-½ partons: a spin-½ quark cannot absorb a longitudinally polarized virtual photon. Callan–Gross is therefore a clean, model-independent spin measurement dressed as a relation between structure functions.
If quarks carry the proton's charge, why does the momentum add up to only half?
Because the photon only sees electrically charged constituents, and the quarks are not the whole story. Summing the measured quark and antiquark momentum fractions gives about 0.5, not 1.0 — half the proton's momentum is unaccounted for. That missing half is carried by gluons, the electrically neutral carriers of the strong force, which the virtual photon cannot couple to and which therefore leave no direct trace in F₂. The momentum sum rule was one of the first quantitative hints that the proton contains more than quarks; the neutral partons demanded by the bookkeeping are exactly the gluons of QCD. So the "missing" momentum is not an error but a discovery.
If scaling is only approximate, does that undermine the parton picture?
No — the small violations are its greatest success. Exact scaling would hold if the partons were truly free, but in QCD they interact: a quark can radiate a gluon, and a gluon can split into a quark–antiquark pair, and both processes depend logarithmically on the resolving scale Q². As Q² rises the probe resolves more of this radiation, so the structure functions drift slowly — F₂ rises at small x and falls at large x. This drift is not random; it is predicted precisely by the DGLAP evolution equations, and measuring it determines the strong coupling α_s and the gluon distribution. The parton model survives intact as the leading picture; QCD supplies the calculable corrections. Scaling violations are how deep inelastic scattering became a precision test of QCD rather than just evidence for quarks.
Why are there two structure functions and not one?
Because inelastic scattering has two independent kinematic variables, and Lorentz invariance plus current conservation for single-photon exchange allow exactly two independent functions to multiply them. Physically, one function (loosely F₂) captures the electric-charge scattering that rides cos²(θ/2), and the other (F₁) captures the magnetic spin-flip scattering that rides sin²(θ/2) — the same electric/magnetic split familiar from elastic electron–proton scattering, now promoted to functions of two variables. An experiment separates them by varying the beam energy at fixed (x, Q²), which changes the relative weight of the two angular terms. The Callan–Gross relation then reveals that the two are not independent after all for spin-½ partons: F₂ = 2xF₁ collapses them to one, which is itself the deepest statement about what is inside the proton.