The Breakthrough Prize in Fundamental Physics for 2026 went to the Muon g-2 collaborations, and the headline version of the story, that a famous anomaly was finally put to rest, is close to right but skips the interesting part. The final Fermilab measurement is the most precise ever made of how a muon’s internal magnet behaves, and the latest Standard Model prediction now sits within about 0.6 standard deviations of it. What changed was the theory, not the experiment.

The prize was presented on April 18, 2026, at a ceremony in Santa Monica, California. It carries $3 million and recognizes three generations of experiments at CERN, Brookhaven and Fermilab. This article walks through what the muon g-2 number is, how the ring measures it, why theory and experiment seemed to disagree, and what is still unsettled.

What a muon’s g-factor actually is

A muon is a heavier cousin of the electron, roughly 200 times more massive, and like the electron it has spin. That spin acts as a tiny bar magnet. The strength of the magnet relative to the spin is set by a number called the g-factor, which a simple relativistic quantum calculation puts at exactly 2.

It is not exactly 2. The vacuum around the muon is never empty. Virtual photons, electrons, quarks and heavier particles flicker in and out of existence, and each one nudges the magnet slightly. Physicists track the leftover as the magnetic anomaly, aμ = (g − 2)/2, which is why the whole program is called g-2.

Here is the appeal. The Standard Model can add up every known particle’s contribution and predict aμ. If an unknown particle or force exists, it would add its own small push, and the measurement would drift away from the prediction. A g-2 measurement is therefore a test of everything the theory knows, all at once.

Three generations of experiments

The campaign is older than most of the people now working on it. The Brookhaven announcement summarizing it describes a lineage that begins at CERN, where three experiments between 1959 and 1979 found a g-factor slightly above 2 in agreement with theory, and continues at Brookhaven National Laboratory on Long Island from 2001 to 2004.

Stage Where What it showed
CERN experiments (1959 to 1979) Geneva g slightly above 2, matching theory
Brookhaven E821 (2001 to 2004) Long Island, New York An anomaly larger than predicted, short of a discovery
Fermilab E989 (first result 2021) Batavia, Illinois Confirmed Brookhaven with better precision
Fermilab final result (2025) Batavia, Illinois Most precise measurement, 127 parts per billion

Fermilab did not build a new magnet. It reused the 50-foot-diameter superconducting storage ring from Brookhaven, which was moved in 2013 from Long Island to Illinois over a journey of about 3,200 miles by land and sea that took 35 days. Data taking at Fermilab began on May 31, 2017.

How the ring turns spin into a number

The principle is neat. Muons are injected into a storage ring with a very uniform magnetic field, and their spins precess as they circle. If g were exactly 2, the spin direction and the momentum direction would rotate in lockstep. Because g is slightly larger, the spin gradually pulls ahead, and the rate at which it pulls ahead is proportional to aμ.

Muons decay, and the decay products carry the answer out. Detectors around the ring register the decay particles, and the rate of detections rises and falls at the slip frequency. Fit that oscillation precisely, and you have the first half of the measurement.

The second half is the magnetic field. Frequency alone is useless unless you know how strong the field was, and the Brookhaven write-up cites a field known to about 0.1 parts per million, or about 100 parts per billion, for that earlier generation. A measurement at the level Fermilab reached depends as much on mapping that field as on counting decays.

  • Frequency: how quickly the spin gets ahead of the momentum, read from decay patterns.
  • Field: how strong and how uniform the magnet is over the muon orbit.
  • Combination: the two together give aμ.

The final number

On June 3, 2025, the collaboration announced its third and last result. It uses data from 2021 to 2023, combined with the earlier datasets, and more than tripled the dataset behind the 2023 result. The Fermilab summary says it draws on the best-quality data, taken after late-run improvements to the muon beam.

The value is aμ = 0.001 165 920 705, with a statistical uncertainty of 0.000 000 000 114 and a systematic uncertainty of 0.000 000 000 091. The overall precision is 127 parts per billion, better than the original design goal of 140 parts per billion.

The result agrees with the 2021 and 2023 Fermilab results. It was submitted to Physical Review Letters. About 176 scientists from 34 institutions in seven countries are listed on the collaboration, according to the Department of Energy summary.

Why the anomaly seemed to be there, and then was not

For years the headline was a gap between measurement and prediction. The gap came from the prediction side. The hardest term to calculate is the hadronic vacuum polarization, the effect of virtual quarks and gluons, which cannot be computed from first principles with simple methods.

In 2020 the international Theory Initiative published a prediction that used a data-driven technique, building the hadronic piece from other experiments that measure electron-positron collisions. That prediction sat well below the experimental value. In 2021 a second approach, lattice QCD, which puts quarks and gluons on a computational grid, produced a number closer to the measurement and narrowed the gap.

The 2025 white paper, posted as arXiv:2505.21476, made the lattice approach central. A new CMD-3 measurement of electron-positron collisions into two pions had increased tensions among the data-driven evaluations, so the authors adopted a lattice average, about 0.9 percent precise, for that term. Hadronic light-by-light uncertainty was roughly halved too. The resulting prediction is aμ = 116 592 033(62) × 10−11, a relative uncertainty of 530 parts per billion.

Doing the comparison yourself

Lining the experimental value up with the 2025 prediction shows how close they now are. In units of 10−11, the Fermilab central value is about 116 592 070.5, while the prediction is 116 592 033. The gap is roughly 38 units. The white paper, using the combined Brookhaven and Fermilab average, quotes 38(63), which is a difference of about 0.6 of its own uncertainty. Its authors state there is no tension between the Standard Model and experiment at the current precision.

Quantity Value Relative uncertainty
Experiment, Fermilab final 116 592 070.5 × 10−11 (central value, from the published aμ) 127 ppb
Standard Model, 2025 white paper 116 592 033(62) × 10−11 530 ppb
Difference (experiment average minus theory) 38(63) × 10−11 about 0.6 of its uncertainty

One observation stands out. The experiment is now about four times more precise than the theory it is being compared with. The uncertainty on the difference is dominated by the prediction, not the measurement. In that sense the experimenters finished the job and handed the problem back to the theorists.

What is not settled

This is not a clean victory for the Standard Model. The data-driven and lattice methods still disagree with each other, and the Fermilab announcement said that discrepancy is still being studied. If the data-driven number had been kept, the old gap would still be there. Choosing lattice QCD was a judgment about which method to trust, made after the CMD-3 result complicated the data-driven picture.

  • Method disagreement: the data-driven electron-positron evaluations differ among themselves and with lattice QCD.
  • Precision target: the white paper treats the 127 parts per billion experimental precision as the goal for theory, and says resolving the tension among data-driven evaluations is key.
  • Future experiment: a J-PARC experiment in Japan is expected to measure aμ in the early 2030s, initially at lower precision than Fermilab.
  • Other analyses: the Fermilab team plans to use its data to look for a muon electric dipole moment and to test CPT symmetry.

What the prize recognizes

The prize goes to all living co-authors of the key result papers, so hundreds of people, and four collaboration members accepted it on stage: Chris Polly of Fermilab, Lee Roberts of Boston University, William Morse of Brookhaven, and David Hertzog of the University of Washington. It is a rare award for a null-ish result, one that tested a theory and did not break it.

That is worth saying plainly. A measurement that removes a hint of new physics is still a result, and arguably a more useful one than a hint that fades slowly. It tells theorists where the Standard Model is solid, and it frees searches elsewhere. For a sense of how the community hunts for the next surprise, see our look at the High-Luminosity LHC upgrade, where the collision rate rises in the coming years.

Sources