SK-Gd: How Gadolinium Turned Super-Kamiokande Into an Antineutrino Telescope

Super-Kamiokande has been the world’s most prolific neutrino detector since 1996. Its 50,000 tons of ultra-pure water and 11,000-plus photomultiplier tubes have delivered the discovery of atmospheric neutrino oscillation, the most precise measurement of solar neutrinos, and a remarkable series of constraints on rare processes. But the detector has a long-standing limitation: it cannot tell electrons from positrons. The Cherenkov rings the two particles emit are essentially indistinguishable. For many of Super-Kamiokande’s physics goals this is fine — solar neutrinos scatter electrons and the sign does not matter — but for a class of measurements that depend on identifying electron antineutrinos, it has been a frustrating blind spot.

In 2020 the collaboration began an experimental program to fix this. Gadolinium sulfate, $\mathrm{G}{\mathrm{d}}_2(\mathrm{S}{\mathrm{O}}_4{)}_3$, has been dissolved into the water at a small but carefully controlled concentration. Gadolinium nuclei have an enormous neutron-capture cross-section, and the captured neutron releases an 8 MeV gamma-ray cascade that the photomultipliers can detect. Because the inverse beta-decay reaction that dominates electron-antineutrino detection produces both a positron and a neutron, the neutron capture provides a delayed coincidence tag: a prompt positron followed by a delayed cascade within the same volume marks the event as an antineutrino. The technique transforms Super-Kamiokande from a detector that sees neutrinos in bulk into one that can identify antineutrinos event by event.

This post is about how the gadolinium upgrade works, why it took so long to deploy, and what it now opens up — most importantly, the long-pursued diffuse supernova neutrino background.

The detection problem

The dominant antineutrino reaction in water at MeV energies is inverse beta decay on a free proton in a hydrogen nucleus:

${\bar{\nu }}_e+p\to e^{+}+n$

The positron carries away most of the energy and is detected immediately through its Cherenkov ring, providing a prompt signal. The neutron, in contrast, thermalizes through scattering on protons over a few tens of microseconds and is then captured. In pure water, the only capture target is the hydrogen of the water molecules:

$n+p\to d+\gamma (2.2\text{ MeV})$

The deuteron’s binding energy is only 2.2 MeV, and a single 2.2 MeV gamma ray is far too faint to detect in Super-Kamiokande against the background of natural radioactivity and dark current. So in pure water the neutron simply vanishes, and the inverse beta-decay reaction looks just like an isolated positron — indistinguishable from many backgrounds.

Gadolinium changes the capture process. Natural gadolinium contains two isotopes with extraordinary thermal-neutron capture cross-sections: ${}^{155}\mathrm{Gd}$ at 60,900 barns and ${}^{157}\mathrm{Gd}$ at 254,000 barns. A few thousand-fold larger than hydrogen’s cross-section of 0.33 barns. Even a tiny mass fraction of gadolinium captures essentially all of the neutrons before they reach hydrogen, and the capture releases not 2.2 MeV but a cascade of gammas totalling about 8 MeV:

$n+{}^{157}\mathrm{Gd}\to {}^{158}\mathrm{G}{\mathrm{d}}^{\ast }\to {}^{158}\mathrm{Gd}+\sum \gamma (\approx 8\text{ MeV})$

Eight MeV is well above the dark-rate noise floor and is clearly visible in Super-Kamiokande’s photomultiplier array. The cascade arrives with the characteristic delay set by neutron thermalization and the capture mean free path — about 30 microseconds at the operating concentration. A prompt positron followed by a delayed 8 MeV cascade within tens of microseconds in the same fiducial volume is an essentially clean antineutrino tag.

Why it took two decades

The idea is straightforward; the implementation was not. Adding gadolinium to a 50,000-ton ultra-pure water detector raises several distinct challenges.

The first is water transparency. Super-Kamiokande’s photomultipliers are 30 to 40 metres away from much of the active volume, and the absorption length of pure water at the relevant wavelengths must remain on the order of 80 metres for the detector to work at full sensitivity. Gadolinium sulfate, if added at the wrong concentration or with the wrong purity, can degrade the water’s optical properties enough to compromise the detector. Years of R&D went into a dedicated 200-ton EGADS test facility at Kamioka, which demonstrated that ultra-pure gadolinium sulfate, run through purification systems compatible with the dissolved load, leaves transparency essentially unchanged.

The second is radiopurity. Gadolinium ores naturally contain small amounts of uranium and thorium. Even part-per-billion contamination of the gadolinium sulfate would inject radioactive isotopes into the detector and overwhelm the very signals the upgrade is supposed to enable. The collaboration developed a multi-stage purification process that brings the uranium and thorium content of the gadolinium sulfate down to parts per trillion. Without this work the upgrade would have made the background problem worse, not better.

The third is water-system compatibility. Super-Kamiokande’s water purification system runs continuously to maintain optical clarity. Adding gadolinium meant designing a new purification chain — a “molecular-sieve” approach — that removes contaminants while leaving the dissolved gadolinium sulfate intact. The upgraded water system has been running stably since 2020.

The fourth is scaling the concentration. The initial loading in 2020 used 0.011 per cent gadolinium sulfate by mass, capturing about half the neutrons on gadolinium and the rest on hydrogen. Subsequent campaigns have increased the concentration, and at 0.03 per cent — reached in 2022 — roughly 90 per cent of neutrons are captured on gadolinium. The efficiency gain directly improves the antineutrino tagging power.

What it opens up

The headline target is the diffuse supernova neutrino background, sometimes called the supernova relic neutrino flux. Every core-collapse supernova in the observable universe contributes a few electron antineutrinos to a diffuse, isotropic flux that has been accumulating since the first stars formed. The integrated flux is small — a few events per year above Super-Kamiokande’s analysis threshold — but the spectrum carries unique information about the average supernova energy spectrum, the cosmological history of star formation, and the failed-supernova fraction.

Detecting this signal has been a community goal for thirty years. Without antineutrino tagging it is essentially impossible: the prompt positron from inverse beta decay sits beneath a much larger background of atmospheric and solar neutrinos producing single isolated events. With the delayed gadolinium cascade, the background drops by orders of magnitude, and the diffuse signal becomes plausibly visible. The first SK-Gd analysis of the diffuse flux was released in 2023 and pushed the upper limit substantially closer to the predicted range. With continued running at full concentration and improved analysis, the first detection — or a definitive exclusion of the lower-end models — is expected within the coming years.

Beyond the diffuse flux, gadolinium loading also improves Super-Kamiokande’s response to a galactic supernova burst. A nearby core-collapse supernova would deliver thousands of inverse beta-decay events in the detector. With gadolinium tagging, these are cleanly separated from the smaller number of charged-current electron-neutrino events, allowing flavor decomposition of the neutrino burst that constrains the underlying explosion mechanism and collective oscillation physics. The technique also benefits reactor-antineutrino observations and proton-decay searches by clarifying the identity of events that would otherwise be ambiguous.

The technique propagates

The success of SK-Gd has prompted other water-based detectors to consider similar loading. WATCHMAN, a reactor-monitoring concept, builds on the same delayed-coincidence approach. THEIA, a proposed 25-50 kton water-based scintillator detector, would combine gadolinium-style tagging with the much higher photon yield of liquid scintillator, opening up a still-broader physics program. The technique has matured from a clever idea — first proposed by John Beacom and Mark Vagins in 2004 — into a workable, production-grade upgrade that may eventually be applied to multi-kiloton detectors that have yet to be built.

Summary

By dissolving a small amount of gadolinium sulfate into its 50,000 tons of ultra-pure water, Super-Kamiokande has acquired a delayed-coincidence tag for electron antineutrinos. The 8 MeV gamma cascade from gadolinium neutron capture, arriving tens of microseconds after the prompt positron from inverse beta decay, is bright enough to be detected and rare enough as a background to provide a clean event-by-event antineutrino identification. The upgrade has been deployed in stages since 2020 and now operates at the design concentration of about 0.03 per cent. Its primary target is the diffuse supernova neutrino background, a long-anticipated signal that finally appears within reach, along with improved galactic-supernova characterisation, reactor monitoring and proton-decay searches. SK-Gd has redefined what a water Cherenkov detector can do, and the technique is shaping the design of the next generation of large neutrino detectors.