Short reactor baselines: look for a change with distance

Turn a suggested explanation into a distance prediction

A sterile flavor would lack the ordinary active weak interactions. Mixing with an additional mass state could let an initially electron-flavor sample lose detectable electron flavor during propagation. This is a hypothesis to test, not a particle established by the reactor rate discrepancy.

The simplest common model adds one state to the three known active flavors. At sufficiently short baselines, the new, larger mass-squared gap dominates the oscillatory phase while the known smaller-gap effects are negligible. Its amplitude and frequency specify a pattern across both distance and energy.

The expression below uses natural units and neglects matter effects. It makes the experimental opportunity clear: compare different travel distances at the same energy. A source-spectrum error shared by all positions can survive in absolute counts while canceling substantially in that relative test.

Sources: [1] · [2] · [3] · [4]

STEREO and PROSPECT compare parts of one apparatus

STEREO operated near the compact research reactor at the Institut Laue–Langevin in Grenoble, France. Its six target cells sampled different distances around ten meters. A reactor dominated by uranium-235 fission simplified the source mixture while the cells supplied simultaneous spectral comparisons.

PROSPECT-I used a segmented, lithium-loaded liquid-scintillator detector at the High Flux Isotope Reactor in Oak Ridge, United States. Its final oscillation analysis grouped events into six distance ranges between roughly seven and nine meters. The detector’s segmentation and neutron signature help control background in a near-surface environment.

Simultaneous cells reduce reliance on matching reactor histories from separate runs, but each cell still has a response and efficiency. Calibration must prevent an energy-scale difference from becoming a false distance effect. Segmentation is a useful control only when the parts are understood well enough to compare.

Sources: [1] · [2] · [5]

DANSS moves the same detector beneath a power reactor

DANSS used a highly segmented plastic-scintillator detector beneath a core at the Kalinin nuclear power plant in Russia. The 2018 analysis compared positions about 10.7 to 12.7 meters from the core. Moving the same detector changes the baseline while preserving much of its detection machinery.

A movable apparatus trades one advantage for another challenge. Positions are sampled at different times, so fuel and power changes must be tracked or averaged appropriately. A commercial core is also larger than a compact research core; events arise from a distribution of production points, not one point at its center.

Those geometric spreads wash out rapid oscillations. Finite energy resolution can do the same. A model may produce a striking ideal curve yet a much weaker detector-folded variation. Comparing experiments requires their actual core dimensions, baseline distributions and responses, not just the shortest quoted distance.

Sources: [3] · [1] · [2]

What the cited null tests exclude

STEREO’s final result, published in 2023, found its cell spectra compatible without the proposed extra oscillation and excluded the favored reactor-anomaly region over its sensitive range. It also measured a source spectrum that disagreed with the reference. A persistent source anomaly and a rejected oscillation explanation can therefore coexist.

The final PROSPECT-I analysis, submitted in 2024 and published in 2025, used the full 2018 sample with an improved multi-period selection. Its baseline spectra were consistent with one another. It extended exclusions into additional regions and rejected the reported Neutrino-4 best-fit point; that point-specific comparison is not a blanket test of all new states.

DANSS’s cited 2018 result likewise obtained exclusion regions from spectral ratios at different positions. These are dated analyses, not a claim that every collaboration’s current limit has been combined here. Each confidence region belongs to a model, dataset and statistical construction that must accompany it.

Sources: [1] · [2] · [3]

An empty preferred region is not an empty theory space

If an oscillation is very slow over the sampled distances, or very fast relative to the geometric and energy spreads, its relative signature can be weak. Small mixing is difficult to test too. An exclusion contour records where the actual experiment can discriminate, rather than declaring uniform sensitivity everywhere.

“Model-independent” in a relative-spectrum discussion usually means reduced dependence on an absolute source spectrum. It does not mean independence from propagation assumptions, geometry, response or statistical choices. An additional-state fit still specifies what signal it is testing.

The common lesson is methodological. Change distance while controlling the source and instrument, then seek a correlated pattern rather than a missing total alone. These reactors test one concrete explanation of anomalies; solar calibration, accelerator and cosmological questions require their own records and assumptions.

Sources: [3] · [2] · [1] · [6]

Try it in the Lab

Primary sources & revision

1. STEREO Collaboration · Uranium-235 spectrum and sterile-neutrino test (Nature, 2023; arXiv v3, 2024)
2. PROSPECT Collaboration · Final PROSPECT-I Short-Baseline Oscillation Search (2024 submission; 2025 publication)
3. DANSS Collaboration · Search for sterile neutrinos at the DANSS experiment (2018)
4. Particle Data Group · Neutrino Masses, Mixing, and Oscillations (2024)
5. PROSPECT Collaboration · Experimental science and detector design
6. Particle Data Group · Statistics (2025)

First published 2026-10-10; last revised 2026-10-10. Original explanatory text and diagrams by Sandbox Physics. Illustrations are schematic; no experimental event records are reproduced here. This is an independent educational publication, not an official Nobel or experiment collaboration publication.

Continue the story