5 MeV Reactor Antineutrino Deviation Evidence

• 5 MeV spectral deviation is a phenomenon where reactor antineutrino measurements reveal a localized excess near 5 MeV in the prompt positron-energy spectrum.
• It is characterized by a nearly Gaussian response and a statistically significant bump, confirmed by likelihood analyses comparing experimental data with standard reactor-flux predictions.
• The consistency of this deviation across different detectors suggests it is an intrinsic effect in the inverse beta decay event spectrum rather than a detector artifact.

The approximately 5 MeV spectral deviation in reactor antineutrino measurements denotes an excess in the prompt positron-energy spectrum associated with inverse beta decay (IBD), relative to standard reactor-flux predictions. In "Evidence for a 5 MeV Spectral Deviation in the Goesgen Reactor Neutrino Oscillation Experiment," it was argued that this feature was already present in the mid-1980s Gösgen reactor experiment at Switzerland’s 2.8 GWth pressurized-water reactor, and that a log-likelihood analysis disfavors a no-anomaly interpretation at the level of $3.8\,\sigma$ (Zacek et al., 2018).

1. Experimental setting and detector architecture

The Gösgen experiment operated at three baselines from the reactor core, $G_1=37.9$ m, $G_2=45.9$ m, and $G_3=64.7$ m. Its detector was a segmented $1\,\mathrm{m}^3$ apparatus containing 377 L of Lucite-walled liquid scintillator divided into 30 cells arranged as 5 planes of 6 cells each. Each cell was instrumented with two photomultipliers, enabling good energy resolution and pulse-shape discrimination for rejection of fast-neutron backgrounds. Neutrons produced in IBD were thermalized in the scintillator and then captured in adjacent $^3$He wire chambers (Zacek et al., 2018).

The detection channel was

$\bar\nu_e + p \to e^+ + n,$

with threshold $E_{\mathrm{th}}=1.8$ MeV. A defining instrumental feature was that only the kinetic energy of the positron was recorded, because most $511$ keV annihilation $\gamma$-rays escaped. This differs from later unsegmented liquid-scintillator detectors, which more nearly record the full prompt energy. Monte Carlo simulations were therefore used to model bremsstrahlung, annihilation-$G_1=37.9$0 escape, and wall effects through response functions $G_1=37.9$1. For a 5 MeV monoenergetic positron, the response was described as a nearly Gaussian peak centered at $G_1=37.9$2 MeV with $G_1=37.9$3 MeV, plus a low-energy tail.

2. Spectrum construction and reactor-model comparison

The comparison baseline for the measured spectrum was the "ILL+Vogel" model, which combines integral $G_1=37.9$4 spectra measured at ILL for $G_1=37.9$5 and $G_1=37.9$6 with theoretical spectra for $G_1=37.9$7 and $G_1=37.9$8. To build a single Gösgen spectrum, the $G_1=37.9$9 and $G_2=45.9$0 datasets were corrected for slight burn-up differences of $G_2=45.9$1 and $G_2=45.9$2, respectively, rescaled to $G_2=45.9$3, and merged bin-by-bin in $G_2=45.9$4 MeV positron-energy intervals. The resulting average fissile-fraction mix was $G_2=45.9$5, $G_2=45.9$6, $G_2=45.9$7, and $G_2=45.9$8, a composition stated to be close to modern near-detector conditions (Zacek et al., 2018).

Because the Gösgen apparatus recorded positron kinetic energy rather than full prompt energy, an energy-scale shift of $G_2=45.9$9 MeV was applied when comparing to spectra from later unsegmented detectors. The measured-to-predicted ratio was defined as

$G_3=64.7$0

Under this convention, the combined spectrum exhibited a clear bump peaking at positron energy $G_3=64.7$1 MeV, corresponding to neutrino energy $G_3=64.7$2 MeV, with $G_3=64.7$3. The paper reports this shape as being in excellent agreement with RENO near-detector data.

3. Localization and characterization of the spectral deviation

The Gösgen excess was not presented as a broad normalization mismatch but as a localized spectral structure. Its characteristic positron-energy position was estimated as

$G_3=64.7$4

which maps to $G_3=64.7$5 MeV. Its width was quoted as approximately

$G_3=64.7$6

with $G_3=64.7$7 MeV from the detector response, and its relative amplitude as

$G_3=64.7$8

Integrated over the spectrum, this corresponds to approximately $G_3=64.7$9 of total IBD events (Zacek et al., 2018).

These numbers are important because they describe the anomaly in the same phenomenological terms later used for the modern "5 MeV bump": position, width, and amplitude. The observation that the Gösgen detector was segmented and predominantly recorded positron kinetic energy is central to the article’s interpretation. The paper argues that this weakens explanations based purely on detector artifacts and also disfavors exotic $1\,\mathrm{m}^3$0 inelastic processes as the primary cause of the effect.

4. Likelihood-ratio test and statistical significance

The statistical analysis was formulated as a comparison between two hypotheses. The null hypothesis $1\,\mathrm{m}^3$1 was a flat ratio, meaning that the combined Gösgen spectrum was consistent with unity within reactor-model and statistical errors. The alternative $1\,\mathrm{m}^3$2 was that the Gösgen ratio followed the bump shape measured by RENO. The two test statistics were

$1\,\mathrm{m}^3$3

with $1\,\mathrm{m}^3$4 the statistical uncertainty in bin $1\,\mathrm{m}^3$5. The corresponding log-likelihoods, up to additive constants, were

$1\,\mathrm{m}^3$6

The likelihood-ratio statistic was then

$1\,\mathrm{m}^3$7

The paper also notes the asymptotic conversion

$1\,\mathrm{m}^3$8

under Wilks’ theorem for large datasets, while emphasizing that the significance was obtained more precisely from Monte Carlo (Zacek et al., 2018).

The Monte Carlo procedure generated $1\,\mathrm{m}^3$9 pseudo-datasets under $^3$0, using Gösgen statistical errors, computed $^3$1 for each realization, and evaluated the fraction $^3$2 with $^3$3. The reported result was

$^3$4

On this basis, the no-anomaly hypothesis was stated to be disfavored at $^3$5. Burn-up corrections in the range $^3$6 and possible residual energy-response mismatches were tested, and their effect on $^3$7 was found to be negligible compared with statistical errors.

5. Relation to later reactor-antineutrino results

The article situates the Gösgen result within the broader reactor-antineutrino literature by comparing its spectral feature to the bump later reported by Daya Bay, RENO, Double Chooz, and NEOS. According to the paper, the position, width, and amplitude of the Gösgen structure agree within errors with those later observations (Zacek et al., 2018).

This comparison is methodologically significant because the Gösgen apparatus differed substantially from modern near detectors. Its segmented geometry and incomplete annihilation-$^3$8 containment imply a distinct prompt-energy response. A plausible implication is that a common spectral feature surviving across such different detector realizations is more naturally interpreted as an effect in the underlying IBD event spectrum than as an experiment-specific reconstruction artifact. The paper states this conclusion more specifically by arguing that the Gösgen observation strengthens an interpretation in terms of an IBD-rate effect.

6. Historical significance and unresolved origin

The paper’s principal historical claim is that early evidence for the 5 MeV spectral feature existed well before the modern reactor-neutrino anomaly program. In that sense, Gösgen is presented not as an isolated precursor result but as a dataset whose reanalysis foreshadowed a now well-established spectral feature (Zacek et al., 2018).

At the same time, the result did not identify the origin of the deviation. The article’s outlook is therefore cautious. It notes that future very-short-baseline measurements and studies at high-enrichment $^3$9 cores may clarify the source of the bump. That outlook reflects the paper’s underlying distinction between two questions that should not be conflated: whether the spectral deviation exists, and what reactor-physics or detection mechanism produces it. The Gösgen analysis addresses the first question by arguing that the feature is statistically present in historical data; it leaves the second question open.