Hyper-Kamiokande at Sheffield

Hyper-Kamiokande is the proposed successor to Super-Kamiokande. It will act both as the far detector for an upgraded T2K beam (with the same energy and the same off-axis geometry, but higher intensity) and as a detector for solar, supernova and atmospheric neutrinos and other non-accelerator physics


The current Hyper-Kamiokande design calls for two water tanks, each with diameter 74 m and height 60 m (compared to 39 m diameter and 42 m height for Super-K). Because of surface-area-to-volume scaling, although the total water mass of one Hyper-K tank is only 5 times that of Super-K, the fiducial mass is over 8 times larger. 

Improved photomultiplier tubes with better resolution and higher quantum efficiency should lead to improved performance compared to Super-K, especially for faint signals.

Hyper-Kamiokande construction will be staged, with one tank currently scheduled to start operation in 2026 and the other a few years later. 

It is not necessary that the two tanks be located in the same place, and recently there has been an interesting proposal to locate the second tank in Korea, where it would have a longer baseline and sit on the second oscillation peak for the T2K beam. This has the potential to reduce systematic errors in oscillation measurements.

Hyper-Kamiokande physics

The physics goals of Hyper-Kamiokande are as follows:

Neutrino oscillation measurements

Using a combination of accelerator neutrinos from the upgraded T2K beam and atmospheric neutrinos, Hyper-K aims to determine the mass hierarchy (is mass state ν3 the lightest or the heaviest eigenstate?), to measure CP violation in the lepton sector—a possible explanation for the matter-antimatter asymmetry of our universe—and to make high-precision measurements of oscillation parameters (Δm2 values and mixing angles).

In addition, neutrino oscillation measurements provide sensitivity to exotic phenomena such as sterile neutrinos and violation of Lorentz invariance.

Solar neutrino measurements

​​​​​​​There is currently some tension in the θ12 sector between solar and reactor neutrino measurements.

With a greater fiducial mass and higher-efficiency PMTs than Super-K, Hyper-K could investigate this by measuring day-night asymmetry in solar neutrino flux caused by matter-enhanced oscillations in the Earth.

Other new avenues that could be explored in solar neutrino physics include monitoring the Sun's fusion reactions by observing 8B neutrinos, and a first observation of the higher-energy hep neutrino flux.

Nucleon decays

The original Kamiokande experiment was designed to search for proton decays (the acronym originally stood for Kamioka Nucleon Decay Experiment not Neutrino Detection Experiment!), and this has continued to be a physics topic of interest for Super-K. 

Hyper-K should be able to improve significantly on Super-K limits, especially as the increased efficiency of Hyper-K PMTs should lead to improved performance, especially in detecting the 2.2 MeV γ ray from neutron capture on hydrogen (rejecting events containing neutrons is an important method of reducing neutrino backgrounds in proton decay searches).

Supernova burst neutrinos

​​​​​​​30 years ago, Kamiokande-II was one of the two experiments to detect a significant neutrino signal from Supernova 1987A in the Large Magellanic Cloud.  The estimated rate of core-collapse supernovae in our Galaxy is of the order of one every 50 years, so we have been slightly unfortunate not to have a Galactic supernova occur in the 20 years that Super-Kamiokande has been active.

Hyper-Kamiokande would detect approximately 60000 neutrinos per tank from a supernova at a typical Galactic distance of 10 kpc (about 30000 light years).  This would permit detailed study of the explosion.

It is worth noting here that Hyper-K and DUNE are almost perfectly complementary in their sensitivity to supernova neutrinos, with DUNE detecting νe (via scattering off argon) and Hyper-K ν̅e (via inverse beta decay).

Supernova relic neutrinos

Detecting neutrinos from a core-collapse supernova in our Galaxy is easy (60000 neutrinos arriving in a ten-second window would be hard to miss), but relies on cooperation from the Galactic population of massive stars: no supernova, no neutrino burst.

However, neutrinos are stable particles, so the neutrinos from every core-collapse supernova that has exploded in the history of the universe are still travelling through space.

Super-Kamiokande has searched for these supernova relic neutrinos without success, because of a background from below-Cherenkov-threshold muons decaying in the detector (producing an apparently isolated Michel electron whose energy is unfortunately comparable to that expected from a supernova neutrino signal).

This is one of the motivations for introducing gadolinium into Super-K, as the supernova relic neutrino interactions should produce a neutron (they are inverse beta decay, ν̅e + p → e+ + n) while the muon decays obviously do not.

However, Hyper-K could complement the Super-K measurement by concentrating on the higher energy region, 20–30 MeV (the Super-K measurement will be most sensitive to 10–20 MeV neutrinos).

Dark matter searches

If the dark matter is made up of supersymmetric neutralinos, regions of high dark matter density (the core of the Sun, the Galactic centre) should experience significant rates of neutralino annihilation. 

Many of the possible final states, e.g. WW, ZZ, bb̅, have decay modes that include neutrinos, making neutrino telescopes a possible means of searching for dark matter.

The signal should be detectable over the large atmospheric neutrino background because of its directionality. Hyper-K can improve on limits set by Super-K by a factor of a few, and has sensitivity to low-mass neutralinos which are difficult to detect in conventional direct-detection experiments (and where there are a number of controversial claims of potential signals).

Hyper-Kamiokande in the UK

The HyperK-UK consortium consists of the Universities of Edinburgh, Imperial College, Lancaster, Liverpool, Oxford, Queen Mary, Royal Holloway, Sheffield and Warwick together with RAL and the Institute of Particle Physics Phenomenology (IPPP) at Durham. The intended UK contributions to Hyper-K include:

  • sensitivity studies and software infrastructure;
  • detector R&D;
  • a gadolinium-loaded water Cherenkov near detector;
  • an improved DAQ system;
  • development of a calibration system;
  • secondary beamline R&D.

These build directly on the UK's contribution to the T2K experiment and will give the UK a high profile in Hyper-K. The Sheffield group is contributing primarily to calibration, software infrastructure, and sensitivity studies.

For a detailed overview of the Hyper-Kamiokande experiment, download the Hyper-K design report (PDF - 282 pages).



The physics goals of Hyper-Kamiokande require a considerable reduction in the systematic error budget compared to T2K. 

One contribution to the systematic error is the uncertainty in our understanding of the detector response: in a recent T2K paper on νμ disappearance, the Super-Kamiokande detector systematics total 3.8% for neutrinos and 3.3% for antineutrinos. The aim is to reduce this systematic error to below 2% for Hyper-Kamiokande.

Calibration of a large water Cherenkov experiment is a complex business—see Abe et al. (2013) for a full description of the Super-Kamiokande calibration procedures. 

The UK contributions to Hyper-Kamiokande include the development of an optical calibration system based on LEDs or laser diodes, the construction of a "pseudo-muon light source" (PMLS) designed to mimic the light output from a muon traversing the detector, calibration of the neutron yield from gadolinium loading of the water, and radiopurity assay of samples of gadolinium sulphate from commercial suppliers. 

The latter two activities are clearly relevant to our participation in Super-Kamiokande, and it is also proposed to install the prototype Hyper-Kamiokande optical calibration system in Super-Kamiokande for field trials.

Optical calibration system

The optical calibration system needs to be able to calibrate the time response and gain of the photomultiplier tubes and also to monitor the water quality, especially the scattering and attenuation lengths, in the detector. 

Solid state devices such as LEDs or laser diodes can be configured to deliver a short pulse of known intensity, duration and wavelength—as in Super-K, a range of different wavelengths would be used, since both the water properties and the PMT response are wavelength-dependent. 

For ease of accessibility, it is planned to mount the electronics outside the detector and propagate the signals through optical fibres to injection points inside the tank, as in the Super-K system shown above.

Optical diffusers are required to control the light injected into the detector for calibration. Different tasks require different light profiles: for example, a measurement of the scattering length requires a narrow beam of light so that scattered light can be identified, while the PMT calibrations such as timing, gain and relative efficiency are better served by wider profiles so that a significant number of PMTs are illuminated by each fibre, reducing the number of channels needed. 

It is likely that several different types of diffuser will be required to achieve the necessary flexibility, and the UK working group led by Liverpool and Sheffield intends to develop and test a number of options, such as the prototype diffuser ball shown in the figure. 

Following successful tests, we will develop a calibration device to be deployed in Super-Kamiokande consisting of a fibre bundle and several interchangeable diffusers to allow testing of these devices in situ in the detector.  The results from these tests will guide the final design of the system for Hyper-K.

Pseudo-muon light source

The ideal calibration source to test reconstruction algorithms is something that looks like a real track but has known and reproducible properties.

Snell's law, n1 sin θ1 = n2 sin θ2, where θ is the angle to the normal, indicates that if we deploy a thin hollow tube filled with air (n1 ≈ 1) in the tank, and illuminate it from one end so that the light is incident on the side of the tube almost tangentially (θ1 ≈ 90º), then the light exits the tube as a cone of half-angle θ such that cos θ = sin  θ2 ≈ 1/n2, just like a Cherenkov cone. 

Such a calibration source would provide an interesting alternative to the usual technique of using cosmic-ray muons as reconstruction test samples.

The concept of the device is an acrylic tube illuminated by a pulsed light source consisting of several LEDs at different wavelengths from 380 to 600 nm tuned to provide the expected light yield from a muon of 390 photons per cm. 

The length of the cylinder represents the range, and therefore the energy, of the muon; in principle, several cylinders of different lengths could be deployed to mimic a range of muon energies. The initial plan is to construct a prototype device which could be tested in a small water tank and then shipped to Kamioka.


The advantages of loading the water of a water Cherenkov detector with gadolinium to provide neutron tagging have been discussed elsewhere. Two issues need to be considered:

What is the neutron yield in a Gd-loaded water Cherenkov detector? Is commercially available gadolinium sulphate sufficiently radiopure to maintain the water purity in Super-K/Hyper-K?

Sheffield is well placed to address both these issues.  The main goal of the ANNIE experiment is to measure the neutron yield, and we have low-background facilities at Boulby Mine to conduct radiopurity assays. 

In addition, the UK groups plan to evaluate the neutron response in Super-K using an encapsulated 252Cf neutron source, and to evaluate the feasibility of deploying neutristors—miniature solid-state neutron generators—within Super-K to provide D-D neutrons on demand.

Supernova neutrinos


Stars with masses above about 10 M⊙ will end their lives by forming an iron core. Fusing iron consumes energy, so when the iron core becomes too massive to support itself against its own gravity, the subsequent collapse is not halted by the onset of a new fusion reaction.

Instead, the collapse continues, first disintegrating all nuclei into their component protons and neutrons and subsequently combining the protons with free electrons to form more neutrons, p + e– → n + νe.

The most likely outcome of this is a core-collapse supernova, in which the innermost 1.5 M⊙ or so of the star becomes a ball of neutrons—a neutron star—supported by degeneracy pressure (i.e. the Pauli exclusion principle, forbidding spin-½ particles like neutrons from occupying the same quantum state at the same time), while the rest of the star is explosively ejected in a supernova explosion.

Although supernovae are some of the brightest events in the Universe—at its peak, a bright supernova can be comparable in brightness to its entire host galaxy—in fact, 99% of the power of a core-collapse supernova is emitted in neutrinos.

Besides the neutronisation spike of electron neutrinos produced when the protons and electrons combine, there is a large thermal flux of neutrinos produced by a variety of reactions, both charged-current (conversion of a charged lepton to its equivalent neutrino via W emission) and neutral-current (Z → νν̅). 

Charged-current reactions produce mainly νe and ν̅e, because electrons and positrons are by far the most common charged leptons in the supernova core; neutral-current reactions produce all types. The energies of the produced neutrinos are typically ~10–30 MeV: more energetic than solar or reactor neutrinos, but nowhere near the GeV energies of accelerator neutrinos.

Famously, neutrinos have already been detected from one supernova. Supernova 1987A in the Large Magellanic Cloud yielded about 20 neutrino events in the Kamiokande-II and IMB detectors. This was enough to confirm predictions of the energy output and the timescale of neutron star cooling, but not very much more than that. 

In contrast, a core-collapse supernova in the Milky Way (which would be around 5 times closer than SN 1987A) would produce tens of thousands of events in Hyper-Kamiokande, mostly through the inverse beta decay reaction ν̅e + p → e+ + n, with a smaller number from elastic scattering, νe + e– → νe + e–. 

The elastic scattering sample is important because it is directional—the scattered electron is travelling in approximately the same direction as the incoming neutrino—whereas the positron in inverse beta decay is almost isotropic.

Since the neutrino signal typically arrives some hours before the light (because the interior of the exploding star is much more transparent to neutrinos than it is to photons), directional information is very important in helping astronomers identify the optical supernova.

Are we likely to observe a Galactic supernova?

The last supernova visually observed in our Galaxy was Kepler's, in 1604.  If the last Galactic supernova was over 400 years ago, is there any point in studying the expected neutrino signal?

Fortunately, there is good evidence that not all Galactic supernovae appear bright at visible wavelengths. The well-known supernova remnant Cassiopeia A, which is a bright radio and X-ray source, is the remnant of a Type IIb core collapse supernova which should have been seen around 1680, but wasn't—presumably because its light was absorbed by dust in the line of sight. 

The supernova remnant shown here, G1.9+0.3, is even younger: measurements of its rate of expansion suggest that the explosion should have been seen around 1900, give or take a decade. Therefore, it is clear that the rate of optically visible supernovae underestimates the true rate. 

Observations of the radioactive isotope 26Al, which is produced by massive stars before and during their explosion as core collapse supernovae, suggest that the rate of such supernovae is about 2±1 per century, so it is not unreasonable to hope that one will happen during the operational lifetime of Hyper-Kamiokande (indeed, it might be argued that we have been mildly unlucky not to observe one in Super-Kamiokande's 20 years of operation to date).

It should be noted that not all supernovae will produce an intense neutrino burst. Type Ia supernovae, which are heavily used in cosmology because of their nearly-identical light curves, do not arise from the core collapse of a massive star and are not expected to produce a neutrino burst. 

It so happens that Kepler's supernova was probably of this type (and the one before it, Tycho's supernova of 1572, definitely was), and the lack of a bright central X-ray source suggests that the progenitor of G1.9+0.3 was too, though Cas A was a core collapse. 

This is atypical: estimates suggest that, in a galaxy like the Milky Way, core-collapse supernovae should be more common than Type Ia supernovae.

What could we learn from a Galactic supernova neutrino burst?

The neutrino burst from a Galactic supernova could potentially provide information in two areas: the physics of supernova explosions, and the physics of neutrinos. 

The physics of supernova explosions

Although the main principles are clear, the details of core-collapse supernova explosions are not well understood. A physically realistic simulation of a core collapse supernova would require:

  • a full 3D treatment of the supernova—existing simulations show that the explosion is likely to be far from spherically symmetric, and observations tend to support this;   
  • a relativistic treatment of gravity—since the gravitational field near the surface of a neutron star is large enough for Newtonian theory to be a poor approximation;
  • detailed simulation of the production and transport of neutrinos—since it is generally believed that it is the neutrinos which actually cause the explosion: without their energy input, the shock-wave created when the outer layers of the star bounce off the newly-formed neutron star stalls, and the whole thing collapses into a black hole.

The physics of neutrinos

The impact of a Galactic supernova on our understanding of neutrino physics would of course depend on its timing: if it were to occur in the relatively near future, it would provide important information on the mass hierarchy: in particular, because of matter-enhanced oscillations, the neutronisation spike is essentially oscillated away in the normal hierarchy (note that because of the low energy of supernova neutrinos, νμ and ντ are effectively unobservable in both water Cherenkov and liquid argon detectors), but is still present in the inverted hierarchy. 

However, both Hyper-K and DUNE aim to resolve the mass hierarchy anyway, so if the supernova came after a few years of operation this information would probably already be known.

However, the supernova signal would still provide information about neutrino flavour conversion in extreme environments which are impossible to replicate anywhere else. This might include "spectral splits" where neutrino-neutrino interactions force a wholesale conversion of one flavour to another above a particular energy.

Unfortunately, it is not practical to achieve this with current computing power.  State-of-the-art simulations such as those by the Garching group have 3D modelling, but simplified neutrino transport and an approximation to general-relativistic gravity.

Better neutrino transport models are available, but only in unrealistic spherically-symmetric supernova models.  Neutrino oscillation phenomena in supernova explosions are extremely complex—the neutrino densities are so large that, believe it or not, neutrino-neutrino interactions play an important role—and so far are only studied by taking the neutrino data from models without oscillation and retrospectively applying oscillation corrections.

A high-statistics sample of neutrinos from a Galactic core-collapse supernova would obviously provide important input into our understanding of stellar explosions. 

Among the phenomena predicted by supernova simulations are rapid variations in the neutrino flux caused by Standing Accretion Shock Instability (SASI) oscillations and a dipole asymmetry between neutrino and antineutrino fluxes (Lepton-Emission Self-Sustained Asymmetry or LESA), which might be tested by comparing results from Hyper-K (dominated by antineutrinos because of inverse beta decay) and DUNE (dominated by neutrinos through scattering on argon).

If the proto-neutron star collapses into a black hole, this would be signalled by the sudden cessation of neutrino emission when the event horizon forms; otherwise, the neutrino energies would provide some information on the equation of state of neutron star matter.

A supernova neutrino burst in Hyper-Kamiokande

Observing a supernova neutrino burst in Hyper-K is not trivial.  Although the expected number of interactions is very large, the neutrino energies are low compared to the neutrinos in the J-PARC beam.

Therefore, Hyper-K triggering and event reconstruction has to be designed with these low-energy events in mind.  Spectral information is especially important, so we need to be able to reconstruct the neutrino energy, not simply tag the event. 

This possibility is an important advantage of Hyper-K over the IceCube experiment, which should see millions of neutrinos (in the form of a sudden gigantic increase in "background noise") but will be unable to reconstruct them since it is designed for very high energy astrophysical neutrinos.

Sheffield PhD student Jost Migenda is currently working on this problem.  Jost also plans to investigate the prospects for detecting a Galactic supernova in DUNE: as noted above, Hyper-K and DUNE are extremely complementary as regards supernova neutrinos because the former sees mainly antineutrinos and the latter mainly neutrinos.

Fossil supernovae: the Diffuse Supernova Neutrino Background

Neutrinos are, as far as we know, absolutely stable. This means that the neutrinos from every core-collapse supernova that has ever happened are still zipping around the cosmos, albeit with their energies appropriately redshifted due to the expansion of the universe. 

These relics of past supernovae are usually referred to as the Diffuse Supernova Neutrino Background (DSNB), or sometimes Supernova Relic Neutrinos (SRN), and could provide significant information on the rate of massive star formation in past epochs.

They also have the advantage that they must exist: we are not reliant on the vagaries of small-number statistics in providing us (or failing to provide us) with a Galactic supernova.

Predictions of the DSNB flux suggest, as shown in the figure, that it could be observable at energies of 20–30 MeV: at lower energies it is drowned out by reactor and solar neutrinos, and at higher energies by atmospheric neutrinos. 

To date, the problem with this scenario has been that very low energy muons, below Cherenkov threshold, can decay in your detector, producing an electron which is above Cherenkov threshold and will appear to come from a neutrino interaction (since the parent muon is not visible).

Such Michel electrons have a well-understood spectrum, which unfortunately lies in the range 20–60 MeV and thus completely covers the signal region. This problem would be considerably alleviated by loading the water in Hyper-K with gadolinium to tag neutrons. 

The supernova relic neutrinos would interact via inverse beta decay, which has a neutron in the final state, whereas electrons produced from invisible muon decays would not involve a neutron. One of the motivations for introducing gadolinium into Super-Kamiokande is to search for this signal.

For much more on supernova neutrinos, consult the recent review by Mirizzi et al. (2015).