K2K and MINOS: test flavor change with a distance you choose

A beam replaces an inferred path with known endpoints

An accelerator offered a different way to test the explanation drawn from atmospheric neutrinos. Super-Kamiokande had linked a shortfall of muon flavor to direction and travel distance. Cosmic rays supplied that natural source, but their production altitude, energy spectrum and flavor mix all had to be modeled.

A pulsed beam fixes when neutrinos are produced, and a survey fixes the line from source to detector. A detector near the source measures interactions before most of the long-baseline loss. A model then carries this reference to the distant detector.

What the experimenters control is mainly the baseline and the source, not the energy of each neutrino. The far measurement still mixes flux, interaction physics, event selection and detector response. The accelerator test is valuable because its uncertainties differ from those of the atmospheric test, not because they vanish.

Sources: [1] · [2] · [3] · [4]

K2K sends a beam from KEK to Kamioka

K2K made a beam of mostly muon neutrinos at KEK in Japan and aimed it at Super-Kamiokande, about 250 kilometers away. The near detectors included tracking systems and a water Cherenkov detector, whose response could be compared with the far detector’s.

The final 2006 paper reported 112 beam events in the Super-K fiducial volume, against about 158 expected without oscillation. That expectation had its own uncertainty, so the deficit had to be judged together with the near-to-far beam extrapolation, the exposure and the backgrounds, not by simple subtraction.

The energy-shape test used 58 of those events: selected single-ring, muon-like events with reconstructed energies. They are a subset of the 112, not 58 more, and the shape test does not cover all 112. Combining rate and shape, the paper rejected the no-oscillation expectation at 4.3 standard deviations.

Sources: [2] · [5]

MINOS tests a longer path with a different detector

MINOS used the NuMI beam from Fermilab, with one detector about one kilometer from the source and another about 735 kilometers away at Soudan, Minnesota. Magnetized steel and scintillator sampled each interaction. Reconstructing muon tracks this way is a different method from reading a water Cherenkov ring.

The first published analysis, in 2006, found 215 selected far events below thirty GeV, where 336 were expected without oscillation. Those numbers belong to that exposure and selection. A count from an earlier announcement or with a different energy cut may not be the same sample.

The magnetic field helps measure the muon’s momentum and charge sign, and shower energy helps reconstruct the incoming neutrino energy. Near and far data are each interpreted through their own detector response. Using the same technology at both sites reduces some uncertainties, but it does not make acceptance, pileup or geometry identical.

Sources: [3] · [4]

A deficit at particular energies says more than an overall shortfall

If the overall beam intensity were misjudged, counts would drop across the whole spectrum. Oscillation instead depends on distance divided by energy. A deficit concentrated in one energy range therefore tests a pattern that a single overall scaling cannot reproduce.

Reconstruction blurs the measured shape. K2K estimated the energy of selected events from the lepton’s kinematics plus an assumption about the interaction. MINOS added track and hadronic energy in a calorimetric estimate. Nuclear motion, unseen energy and final-state interactions can affect the two methods differently.

An analysis has to build these effects into the expectation and let the relevant nuisance parameters vary. Agreement with an oscillation pattern after that step means more than agreement with an ideal curve. K2K and MINOS gave complementary accelerator tests of the scale that atmospheric neutrinos had already suggested.

Sources: [2] · [3] · [6] · [7]

What disappearance shows, and what it leaves open

Disappearance means fewer selected muon-flavor charged-current interactions than predicted. No one follows a missing neutrino to see what it became. Showing that another flavor appears takes its own event selection, its own backgrounds and enough energy to pass that flavor’s production threshold.

A later antineutrino measurement is a separate dataset with its own beam composition and charge selection, so its numbers cannot stand in for the 2006 neutrino result. The same care applies to successive papers from one collaboration, because more exposure and better reconstruction change what can be inferred.

The lasting result is that the same physical pattern turned up again under controlled conditions, with different sources and instruments. Atmospheric neutrinos suggested flavor change, reactors probed complementary parameters, and beams tested a journey the experimenters chose. Together they make the case for flavor change in flight more convincing than any single deficit stripped of its experimental controls.

Sources: [2] · [3] · [6] · [8]

Try it in the Lab

Primary sources & revision

1. Super-Kamiokande Collaboration · Evidence for Oscillation of Atmospheric Neutrinos (1998)
2. K2K Collaboration · Measurement of Neutrino Oscillation by the K2K Experiment (2006; v3)
3. MINOS Collaboration · Muon-neutrino disappearance with MINOS and NuMI (2006)
4. Adamson et al. · The NuMI Neutrino Beam (2015/2016)
5. K2K Collaboration · Publications
6. Particle Data Group · Neutrino Masses, Mixing, and Oscillations (2024)
7. T2K Collaboration · The T2K Experiment (2011)
8. KamLAND Collaboration · First Results: Evidence for Reactor Anti-Neutrino Disappearance (2002/2003)

First published 2026-10-10; last revised 2026-10-13. 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.