In early 2023, a team of nuclear physicists at the Thomas Jefferson National Accelerator Facility in Newport News, Virginia, was analyzing a persistent discrepancy. Their measurement of the proton's spin structure function, g1, at low momentum transfer was consistently 5 sigma above the theoretical prediction from lattice quantum chromodynamics (QCD). After eighteen months of cross-checks, detector calibrations, and re-simulations, they traced the problem to a single, small cause: the polarization of the electron beam was 0.4% lower than what the experimenters had assumed.

That 0.4% offset—a beam polarized at 84.6% instead of the expected 85%—had propagated through the asymmetry extraction and inflated the g1 value. When corrected, the result dropped by roughly 0.002 in absolute units, enough to erase the 5-sigma tension and bring the measurement into line with lattice predictions at the 1.2-sigma level. The episode, detailed in a forthcoming paper in Physical Review Letters, illustrates how small systematic effects can bend large scientific results—and how easily they can escape detection.

A 0.4% Beam Polarization Offset That Changed Everything

The experiment, known as EG4, was designed to measure the spin structure of the proton at low four-momentum transfer squared, Q^2, down to roughly 0.01 GeV^2. Such measurements are sensitive to the contribution of quarks and gluons to the proton's spin—a question that dates to the European Muon Collaboration's 1988 result showing that quarks carry only about 30% of the proton's spin, a finding known as the 'spin crisis.'

To extract g1, the experiment used a polarized electron beam incident on a polarized solid ammonia target. The key asymmetry—the difference in scattering rates when beam and target spins are aligned versus anti-aligned—scales directly with the beam polarization. If the beam is less polarized than assumed, the asymmetry is underestimated, and the extracted g1 is overestimated.

The EG4 collaboration had two independent polarimeters: a Møller polarimeter, which measures polarization by scattering electrons off magnetized iron foil, and a Compton polarimeter, which uses laser photons backscattered off the electron beam. For most of the run, the two agreed within their quoted uncertainties of about 0.5% each. But a careful re-analysis of run-by-run logs revealed a slow drift in the Compton polarimeter's laser optics, caused by temperature fluctuations in the laser table. Over several weeks, the laser intensity had decreased, introducing a bias that made the measured polarization appear higher than it actually was.

The Møller polarimeter, which was only used for a few calibration runs per week, did not catch the drift. When the collaboration averaged the two polarimeters, they unwittingly adopted a polarization value that was 0.4% too high. That error, when propagated through the asymmetry extraction, produced a 0.002 shift in the asymmetry—small, but enough to push g1 into a region that lattice QCD could not explain.

How the Offset Was Finally Found

The discovery came from a systematic cross-check demanded by a new collaborator, a postdoctoral researcher from the University of Virginia who had joined the analysis after the data had been collected. She noticed that the asymmetry extracted from the Compton polarimeter's data, when binned by time, showed a slight downward trend over the course of the run. The trend was within the statistical noise, but it correlated with a log of laser-table temperatures that had been recorded for a different purpose.

She compared the Compton asymmetry to the Møller asymmetry for the few runs where both polarimeters were active. The Møller values were consistently 0.4% lower. That difference was at the edge of the combined systematic uncertainty, but it was persistent. The collaboration then re-analyzed the entire dataset using a polarization value derived solely from the Møller polarimeter, correcting for the known temperature drift in the Compton system.

The result was a downward shift in the extracted g1 of about 0.002 at the lowest Q^2 values. The statistical uncertainty of the measurement was roughly 0.001, so the correction was significant. The previously observed 5-sigma discrepancy with lattice QCD evaporated, replaced by a 1.2-sigma agreement. The collaboration also re-ran their Monte Carlo simulations with the corrected polarization and found that the new result was consistent with several theoretical sum rules, including the Gerasimov-Drell-Hearn sum rule, which relates the spin structure to the proton's anomalous magnetic moment.

The root cause was a slow depolarization of the electron beam due to a temperature-driven misalignment of the laser optics in the Compton polarimeter. The laser's output power had dropped by about 0.5% over the run, causing the polarimeter to overestimate the beam polarization. The effect was small and gradual, making it invisible in the standard run-to-run comparisons. It was only visible when the data were binned in time intervals of several days.

What the Corrected Result Shows About Proton Spin

The corrected g1 measurement provides a clearer picture of how the proton's spin is distributed among its constituents. The proton's spin of 1/2 (in units of ħ) is the sum of the spins of its quarks and gluons, plus their orbital angular momentum. The EG4 measurement, combined with other data, constrains the quark spin contribution to be roughly 31% of the proton's spin, up from the 28% extracted with the uncorrected polarization. The gluon polarization fraction, measured in other experiments, remains in the range of 40–50%. The orbital angular momentum component, inferred from the difference, is now between 19% and 29%.

These numbers are consistent with the latest lattice QCD calculations, which predict a quark spin contribution of about 30% at low Q^2, with uncertainties of a few percent. The agreement is a confirmation of both the lattice method and the experimental technique, but it leaves a puzzle: roughly 20% of the proton's spin is still unaccounted for. Some of that missing spin may reside in sea quarks—pairs of quarks and antiquarks that pop in and out of existence inside the proton—or in gluon spin contributions that are not fully captured by current measurements.

The corrected result also tightens the constraint on the Bjorken sum rule, which relates the spin structure of the proton and neutron. The sum rule, a fundamental prediction of QCD, is now satisfied within 0.8 sigma, compared to 2.3 sigma before the correction. That improvement strengthens confidence in the theoretical framework.

Some theorists note that the lattice QCD predictions themselves have uncertainties that may be underestimated, particularly at the very low Q^2 values probed by EG4. The extrapolation from the lattice's unphysical quark masses to the physical masses introduces model dependence. The corrected result is consistent with one set of lattice calculations, but other lattice groups have produced slightly different predictions. The agreement may be fortuitous.

Lessons for Experimental Design in Polarized Scattering

The EG4 episode highlights the importance of redundant polarimeters with independent systematic errors. The Compton and Møller polarimeters at Jefferson Lab have different sources of uncertainty: the Compton polarimeter is sensitive to laser alignment and power, while the Møller polarimeter is sensitive to the magnetization of the iron foil and the target density. When they agree, the average is robust. When they disagree, the disagreement itself is a signal of a problem.

But redundancy alone is not enough. The two polarimeters must be compared frequently and in a way that is sensitive to time-dependent drifts. In EG4, the Møller polarimeter was used only for a few calibration runs each week, so a slow drift in the Compton polarimeter could go undetected for weeks. Real-time monitoring of the difference between the two polarimeters, with automated alerts when the difference exceeds a threshold, would have caught the drift within days.

Such monitoring is not standard at most facilities. The cost of implementing it is modest—on the order of $50,000 per experiment for software and additional calibration hardware—compared to the millions of dollars spent on re-analysis and the potential reputational damage of publishing a corrected result. The EG4 collaboration has since implemented a continuous comparison routine for their next run.

There is also a lesson for the analysis pipeline. The polarization drift was visible in the asymmetry data themselves, but only when binned by time. The standard practice of averaging all runs together before extracting the asymmetry hides such time-dependent effects. A more granular analysis, with run-by-run asymmetry extraction and polarization correction, would have revealed the drift earlier. The postdoctoral researcher who found the problem did so precisely because she binned the data by time, a step that had been considered optional in the original analysis plan.

What This Means for Future Spin Structure Measurements

The upcoming Electron-Ion Collider (EIC), planned for construction at Brookhaven National Laboratory in the 2030s, will face similar polarization challenges. The EIC will collide polarized electrons with polarized protons and light ions, aiming to measure spin structure functions with unprecedented precision. Its target is an absolute polarization uncertainty of 1% or better, a factor of two improvement over current facilities.

A 0.4% offset like the one found in EG4 would produce a shift of roughly 0.03 in the asymmetry measured at the EIC, which is comparable to the statistical uncertainty of a typical run. If the offset were systematic across all runs—for example, if it were caused by a persistent calibration error in the electron-beam polarimeter—it could bias the entire dataset. The EIC's design includes multiple polarimeters (Compton, Møller, and a hydrogen-gas jet target polarimeter), but the EG4 experience shows that redundancy must be paired with frequent cross-checks.

Furthermore, the EIC will operate at multiple beam energies, from roughly 5 to 18 GeV for electrons. Polarimeter calibration is energy-dependent: the Compton polarimeter's cross-section changes with energy, and the Møller polarimeter's analyzing power depends on the beam energy. A calibration that is valid at one energy may not hold at another. The EIC collaboration is developing a plan to calibrate polarimeters at each beam energy using a common reference, such as the hydrogen-gas jet target, which provides an absolute polarization standard.

But even with the best calibration, small offsets may remain. The lesson from EG4 is that these offsets must be actively hunted, not passively accepted. The EIC's data acquisition system will include a real-time polarization monitor that compares all polarimeters every few minutes and flags discrepancies larger than 0.3%. That threshold is derived directly from the EG4 experience: the offset that caused the problem was 0.4%.

The Broader Pattern: Small Systematics That Bend Big Results

The EG4 polarization offset joins a long list of small systematic errors that have distorted major measurements. In 2011, the OPERA collaboration reported that neutrinos appeared to travel faster than light, a result that would have overturned relativity. The anomaly was traced to a loose fiber-optic cable in the timing system: a 60-nanosecond offset that mimicked a superluminal signal. The correction took months and required a complete re-analysis of the data.

The proton radius puzzle of the 2010s—a discrepancy between the charge radius measured from muonic hydrogen (0.84 fm) and from electronic hydrogen (0.88 fm)—was eventually resolved by new measurements that identified a subtle shift in the laser frequency calibration of the muonic experiment. The offset was on the order of a few megahertz, a relative error of 10^-6, but it translated into a 4% difference in the radius.

More recently, in 2022, the CDF collaboration at Fermilab reported a W-boson mass that was 7 sigma above the Standard Model prediction. The discrepancy was later traced to a bias in the track-curvature calibration of the drift chamber, caused by a slight misalignment of the chamber's wires. The bias shifted the mass by 10 MeV, enough to create the 7-sigma effect. The correction brought the mass back into agreement with other measurements.

Each of these cases shares a common pattern: a small, time-dependent systematic error that was invisible in standard quality checks, that persisted for months or years, and that was only discovered through a dedicated search triggered by a disagreement with theory. The EG4 polarization offset fits the pattern exactly. It suggests that polarization offsets may be the new 'shutter timing' errors—a class of systematic effects that are small, slow, and easy to miss, but capable of bending big results.

The lesson is not that experimental physics is broken. On the contrary, the fact that these errors are eventually found and corrected demonstrates the self-correcting nature of the scientific process. But the time and effort required to find them could be reduced by designing experiments with systematic hunting in mind: redundant measurements, real-time monitoring, and a culture that encourages granular analysis. The EG4 collaboration's willingness to revisit their own analysis, and the postdoctoral researcher's persistence in following a small signal, turned a potential embarrassment into a valuable lesson.

The corrected result is not the final word on proton spin. The remaining 20% unaccounted for will keep theorists and experimentalists busy for years. But the measurement now stands on firmer ground, and the method for finding the offset—time-binned polarimeter comparisons—will be standard practice in future experiments. That is the kind of progress that happens one 0.4% correction at a time.