The 5-MeV Bump: A Reactor-Neutrino Anomaly Still Unexplained

By 2014, the reactor antineutrino programme had matured to the point where the leading short-baseline experiments — Daya Bay, RENO, and Double Chooz — were measuring the ν̄_e spectrum to per-cent-level precision. The expectation was that the spectrum would match the Hubert-Mueller and Schreckenbach-Klapdor flux predictions developed since the 1990s and validated against earlier measurements at lower precision.

It did not. All three experiments observed an unexpected excess of events at a positron energy of approximately 4–6 MeV, corresponding to antineutrino energies of 5–7 MeV. The excess was about 10% above prediction, sharply localised in this energy band, and reproducible across all three independent reactor experiments. It became known as the “5-MeV bump” or “5-MeV excess”.

A decade later, the bump remains one of the standing puzzles of reactor-neutrino physics. It is not a sterile-neutrino signature; that has been definitively excluded. The most likely explanation is an error in the reactor antineutrino flux predictions — specifically in how individual fission-product beta-decay branches contribute to the integrated antineutrino spectrum. But the precise origin of the discrepancy has resisted unambiguous identification.

This post walks through the observation, the elimination of various explanations, and the current state of the puzzle.

The observation

Reactor antineutrino experiments detect ν̄_e through inverse beta decay: ${\bar{\nu }}_e+p\to e^{+}+n$ The positron carries most of the antineutrino energy, with a small offset from the kinematics and the threshold. The positron annihilates promptly, producing a “prompt” scintillation signal; the neutron thermalises and is captured (on hydrogen, gadolinium, or another nucleus) on a timescale of microseconds, producing a “delayed” signal. The delayed coincidence cleanly identifies inverse-beta-decay events.

The positron energy spectrum, after corrections for detector response and oscillation, gives the antineutrino spectrum directly. Pre-2014 measurements at single-detector experiments (CHOOZ, KamLAND, Bugey) were consistent with the predicted Hubert-Mueller spectrum within their statistical uncertainties.

Daya Bay’s 2014 paper, with much larger statistics, was the first to clearly identify a structured deviation:

• An excess of events around 5 MeV positron energy at $4\sigma$ significance
• The integrated flux measurement remained consistent with a 5% deficit relative to prediction (the “reactor antineutrino anomaly”, separately a topic of investigation)

RENO confirmed the bump independently in 2015 with similar significance. Double Chooz reported the same feature in 2016.

Excluding the sterile-neutrino interpretation

The first hypothesis was that the bump might be related to the broader reactor antineutrino anomaly — a 5-6% flux deficit observed at short baselines. If both effects were due to oscillation into a sterile neutrino with $\Delta m^2\sim 1$ eV², they should be related: the oscillation should produce both a flux deficit and a spectrum distortion.

But the spectrum distortion predicted by sterile-neutrino oscillation is smooth across energies, with the deficit increasing or decreasing monotonically with energy. The observed bump is localised in a narrow energy band. The two patterns are mathematically incompatible.

Subsequent measurements at very short baselines (PROSPECT, STEREO, NEOS, DANSS) further excluded sterile neutrinos at the LSND parameters. By 2020, the consensus had moved to: the broader reactor anomaly is likely an error in the flux prediction, and the bump is a separate, more localised manifestation of similar issues.

The flux-prediction angle

Reactor antineutrinos are produced by beta decay of fission products. To predict the spectrum, you must:

1. Calculate the inventory of all fission products in the reactor at any given time
2. For each, compute the beta-decay endpoint energy and branching fractions
3. Combine all the contributions weighted by their relative abundance and decay rate
4. Account for time evolution as the reactor burns

Two principal approaches exist:

Cumulative method (Schreckenbach-Klapdor, 1980s): Use measured cumulative beta spectra of the four major fissionable isotopes (²³⁵U, ²³⁸U, ²³⁹Pu, ²⁴¹Pu), then convert from electron spectra to antineutrino spectra using known kinematic relationships. This was the dominant method until the 2010s.

Summation method (Mueller-Lhuillier-Letourneau-Onillon, 2011 onwards): Build the spectrum from databases of individual fission-product beta-decay branches, summed and weighted appropriately. This requires accurate decay data for thousands of individual nuclei.

Both methods produce essentially the same predicted spectrum at low energies. They differ slightly at high energies where individual high-Q-value beta decays contribute substantially.

The 5-MeV bump appears in both predictions but is underestimated in both. Specifically, there is some structure in the 5–7 MeV region of the antineutrino spectrum that the canonical predictions don’t capture.

What might cause the bump

Three current candidate explanations:

1. Underestimated branching for high-Q-value decays. Some nuclear-data evaluations may have underestimated the rate of specific fission-product beta decays with endpoints in the 5-7 MeV range. Improved nuclear-data evaluations from the 2010s (notably the IAEA and ENSDF databases) have revised some of these branches upward.

2. Forbidden-decay corrections. Many fission-product beta decays are “forbidden” (require parity- or angular-momentum-changing matrix elements) rather than “allowed”. Forbidden decays have non-trivial spectral shapes that depend on nuclear matrix elements; the canonical 1980s evaluations may have used inaccurate forbidden-decay shapes for specific isotopes contributing in the 5-MeV region.

3. Specific isotope contributions. A few specific fission products may dominate the bump. Candidates include ${}^{96}$Y, ${}^{142}$Cs, ${}^{96}$Sr, and others. If their decay branches were measured directly, the bump could be either confirmed (if the prediction is updated to match) or unmasked as a real anomaly (if it remains after the update).

The PROSPECT experiment at Oak Ridge, in 2017, directly measured the ²³⁵U contribution to the reactor antineutrino spectrum by running near a research reactor that uses essentially pure ²³⁵U fuel. The result showed deviations from the Hubert-Mueller prediction in the 5-MeV region — consistent with the bump being primarily a ²³⁵U-related effect. This is a major piece of evidence for the flux-prediction explanation.

Implications

If the bump is fully explained by reactor-flux prediction errors, then:

Implications for the broader reactor anomaly: The 5-6% flux deficit may also be a flux-prediction error rather than a sterile-neutrino signal. This strengthens the case that the reactor anomaly is not new physics — though does not by itself eliminate the LSND/MiniBooNE problems.

Implications for ongoing reactor experiments: Daya Bay, RENO, and Double Chooz must use empirical (data-driven) corrections to the flux prediction in the 5-MeV region, rather than relying on the canonical models. The corrections are now standard in current oscillation analyses.

Implications for next-generation experiments: JUNO, scheduled to come online in 2026, is designed for sub-percent precision on its mass-ordering measurement. It is designed assuming the bump persists and applies appropriate corrections.

If, alternatively, the bump turns out to be a real effect — for example, indicating new physics in fission-product beta decays — the implications would be much broader. The most credible analyses say this is unlikely but not yet conclusively excluded.

What we don’t know

Despite a decade of investigation, several key questions remain:

• Why is the bump localised in such a narrow energy window? If it’s caused by a single dominant fission product, which one?
• Why is the magnitude consistent across all three reactor experiments (Daya Bay, RENO, Double Chooz) despite different reactor cores and experimental setups?
• Does the bump remain after the most recent flux-prediction updates? Some updated predictions (Estienne et al. 2019, Mueller et al. 2024) include data-driven corrections to specific fission products in the 5-MeV region; these reduce but do not entirely eliminate the discrepancy with measured spectra.

A definitive resolution will likely require either: (a) a direct, isotope-by-isotope measurement of the antineutrino spectrum from individual fission products at the 5-MeV scale (challenging but possible at TANL-type facilities); or (b) a sufficiently precise measurement at multiple reactor experiments to fit out the dominant contributions.

The 5-MeV bump remains, in 2026, the most concrete and well-characterised reactor-physics puzzle. It is not new fundamental physics. It is a testament to how much remains to be understood even in the “well-known” parts of nuclear physics. And it is a direct constraint on next-generation experiments that depend on reactor antineutrino spectra for their primary measurements.