KamLAND: When 53 Reactors Confirmed What the Sun Had Been Telling Us

The Sudbury Neutrino Observatory had measured solar neutrinos directly. By 2002, the deficit was no longer mysterious — two thirds of the electron neutrinos produced in the Sun’s core were arriving at Earth as muon or tau neutrinos. The case for oscillation was strong.

But solar neutrino arguments rest on solar physics. To make the case airtight, you want a terrestrial source where you control the production rate, you know the baseline, and you can independently confirm the same oscillation parameters. The reactor antineutrino is the natural choice. Reactors burn ²³⁵U and ²³⁹Pu and produce a steady, well-measured flux of ${\bar{\nu }}_e$ at MeV energies. If solar oscillation is real, with the parameters SNO inferred, then ${\bar{\nu }}_e$ from a sufficiently distant reactor should also disappear. Not just disappear — disappear with the energy spectrum that maps the solar mass splitting.

KamLAND was designed to do exactly that. A kiloton liquid-scintillator detector under a Japanese mountain, with 53 commercial reactors scattered around it at distances mostly between 130 and 240 km, and a single specific question: do reactor antineutrinos oscillate the way SNO predicts?

The first published result, in 2002, said yes. By 2008, the experiment had become the most precise reactor-based measurement of the solar mass splitting. KamLAND-Zen, its successor program, is now the world’s tightest constraint on neutrinoless double beta decay. The detector lasts; the physics it has produced lasts longer.

Choosing the right baseline

Reactor antineutrinos from a single core peak around 4 MeV. At that energy, the first oscillation maximum for the solar splitting ($\Delta m_{21}^2\approx 7.4\times 10^{-5}$ eV²) sits at $L\approx 60$ km. The second maximum is at 180 km. Choosing where to put your detector is a real question.

A short baseline (10 km or so) sees the smaller, faster oscillation tied to $\Delta m_{31}^2$ — that’s where Daya Bay and RENO operate, hunting θ₁₃. A long baseline like 180 km cuts directly to the solar oscillation regime, where θ₁₂ and $\Delta m_{21}^2$ govern the survival probability.

The Kamioka mine had the right combination: an existing underground site with rock overburden equivalent to 2,700 metres of water, enough to suppress cosmic-ray muons by six orders of magnitude. The Japanese reactor fleet — at the time of construction, 53 commercial cores spread across the four main islands — gave a natural average baseline of approximately 180 km when weighted by thermal power. The detector saw a flux of about $10^6$ ${\bar{\nu }}_e$ per cm² per second, of which roughly 1 per day would interact in a kiloton target.

The trade-off was statistics. KamLAND would see a few hundred to a few thousand ${\bar{\nu }}_e$ events over years of running — small numbers compared to short-baseline reactor experiments closer to single cores. But the L/E coverage was unique.

Building the detector

KamLAND occupied the cavern that had previously housed Kamiokande, the predecessor to Super-Kamiokande. The original 3-kt water tank was drained and replaced with a 13-metre-diameter spherical balloon containing 1,000 tonnes of liquid scintillator. The scintillator — a mix of pseudocumene and dodecane with the fluorescent dye PPO — produces ~8,000 photoelectrons per MeV in the surrounding photomultiplier array.

Around the balloon, a buffer layer of mineral oil shielded the scintillator from radioactivity in the photomultipliers themselves. The 1,879 inward-facing PMTs (60 cm diameter) provided 34% of the inner surface coverage, slightly less than Super-K. An outer water tank served as both an active cosmic-ray veto and shielding.

The radiopurity requirements were severe. Trace contamination of uranium and thorium in the scintillator, the balloon film, the mineral oil, and even the photomultiplier glass produces backgrounds in the energy region of interest. Multi-stage distillation, water extraction, and selective sourcing brought the contaminations down to roughly $10^{-15}$ g of U/Th per gram of scintillator — a level previously not achieved anywhere except Borexino, which was being built in parallel at Gran Sasso.

The detector recorded its first inverse-beta-decay events in early 2002.

The first result

The detection signature is the same one Cowan and Reines used in 1956: ${\bar{\nu }}_e+p\to e^{+}+n$ The positron annihilates promptly. The neutron thermalises and is captured on hydrogen with a 2.2-MeV gamma ray, with a capture time of about 200 microseconds. The delayed coincidence between the prompt positron and the delayed gamma is unmistakable.

The first KamLAND paper, published December 2002, reported on 145 days of livetime. Expected number of unoscillated events from the predicted reactor flux: 86.8. Observed: 54. The deficit was real at 99.95% confidence — about a 4σ effect. The energy spectrum showed a distortion characteristic of oscillation, not simply a flat reduction in rate.

The extracted oscillation parameters were consistent with the solar measurements: $\Delta m_{21}^2=(6.9\pm 0.4)\times 10^{-5}{\text{eV}}^2$ ${\sin }^2(2{\theta }_{12})=0.83\pm 0.05$

In the LMA-MSW solution that SNO had favoured, the values agreed within uncertainties. The two completely different sources — the Sun and Japanese power reactors — were giving the same answer.

This was the moment that closed the oscillation argument. Solar neutrinos disappear. Reactor antineutrinos disappear. The disappearance happens with the same parameters at vastly different sources and energies. Pontecorvo’s 1957 idea was now, definitively, an empirical fact.

Refining the picture

By 2008, KamLAND had accumulated 2,135 events on a 1.49-kiloton fiducial mass. The energy spectrum showed clear oscillation features — peaks and dips at characteristic L/E values. A direct fit to the spectrum extracted oscillation parameters with several-percent precision: $\Delta m_{21}^2=(7.59\pm 0.21)\times 10^{-5}{\text{eV}}^2$ ${\tan }^2{\theta }_{12}=0.47_{-0.05}^{+0.06}$

The energy spectrum itself — modulated by the L/E-dependent oscillation phase — was the most distinctive evidence. Pure flux deficit (no oscillation) would give a flat reduction; oscillation gives a specific spectral shape with maxima and minima. KamLAND saw the predicted shape clearly.

Combined with solar-neutrino data, the global fit gave the most precise determination of the solar-sector parameters available before JUNO. Many of the values cited in PDG tables today still trace primarily to KamLAND-plus-solar.

A side benefit: geoneutrinos

KamLAND’s energy threshold extended down to about 1.8 MeV (the inverse-beta-decay threshold). Above this, reactor antineutrinos dominate. But just above the threshold, in the 1.8–3.4 MeV window, sits the antineutrino spectrum from naturally occurring uranium and thorium in the Earth’s crust and mantle.

These geoneutrinos had been proposed since the 1960s but never directly detected. By 2005, KamLAND had accumulated enough exposure to claim a 4.5σ detection — the first observation of antineutrinos from radioactive decay inside the Earth. The result, refined over subsequent years, fed into a quantitative measurement of Earth’s radiogenic heat production.

KamLAND still records geoneutrinos. Together with Borexino’s later measurements, these have become the two principal data points constraining Earth’s interior radiogenic heat budget — currently around 20 TW out of the 47 TW total Earth heat flow.

The 2011 transition

By 2010, KamLAND’s reactor-antineutrino programme had reached the precision floor where additional running gave diminishing returns. The Daya Bay measurement of θ₁₃ in 2012 also redirected attention in reactor physics toward shorter baselines and CP-related questions. KamLAND’s strategic shift came in 2011.

The detector was modified by adding a smaller inner balloon containing ${}^{136}$Xe-loaded liquid scintillator — about 320 kg of enriched xenon dissolved in the inner volume. The new configuration, called KamLAND-Zen, became a search for neutrinoless double-beta decay of ${}^{136}$Xe.

The transition was elegant: the same detector, same PMTs, same overall infrastructure, but now optimised for the $0\nu \beta \beta$ signal at 2.458 MeV rather than the broad reactor-antineutrino spectrum. The reactor program continued in reduced form, mostly for monitoring and geoneutrino measurements, but the primary science focus shifted.

KamLAND-Zen has set increasingly stringent limits on the $0\nu \beta \beta$ half-life of ${}^{136}$Xe. The most recent published bound is approximately $T_{1/2}>2.3\times 10^{26}$ years, corresponding to an effective Majorana mass below ~50 meV in favorable nuclear-matrix-element scenarios. This is the most stringent existing bound on neutrinoless double beta decay.

Why the experiment mattered

Three lasting contributions.

It closed the solar oscillation argument with a terrestrial source. Before KamLAND, the case for solar oscillation rested entirely on inferences from a single source (the Sun) interpreted through a model (the Standard Solar Model). KamLAND replaced the inference with a direct measurement, with a known source and a controlled detector.

It established reactor antineutrinos as a precision neutrino-physics tool. The technique — high-purity scintillator, multi-reactor flux model, oscillation-pattern fit — became the template that Daya Bay, RENO, JUNO, and others have refined. The current precision-neutrino programme would not exist without KamLAND’s demonstration that reactor physics could deliver percent-level oscillation parameter measurements.

It demonstrated that one detector can serve multiple programmes. KamLAND ran reactor physics, geoneutrino physics, and (later) double-beta-decay physics with essentially the same hardware. The pattern — a multi-mode underground detector reused across a decade or more — is now the standard for major neutrino-physics facilities.

The Kamioka mine continues to host neutrino physics. The original Kamiokande site became KamLAND, and is now KamLAND-Zen. Super-Kamiokande operates at full capacity. Hyper-Kamiokande is being built nearby. Three generations of detectors in one location, each producing foundational physics, each leaving its successor a richer experimental landscape.