Super-Kamiokande

In the previous phases, the ID-PMTs processed signals by custom electronics modules called analog timing modules (ATMs). Charge-to-analog converters (QAC) and time-to-analog converters (TAC) are contained in these modules that had dynamic range from 0 to 450 picocoulombs (pC) with 0.2 pC resolution for charge and from −300 to 1000 ns with 0.4 ns resolution for time. There were two pairs of QAC/TAC for each PMT input signal, this prevented dead time and allowed the readout of multiple sequential hits that may arise, e.g. from electrons that are decay products of stopping muons.[15]

The SK system was upgraded in September 2008 in order to maintain the stability in the next decade and improve the throughput of the data acquisition systems, QTC-based electronics with Ethernet (QBEE).[16] The QBEE provides high-speed signal processing by combining pipelined components. These components are a newly developed custom charge-to-time converter (QTC) in the form of an application-specific integrated circuit (ASIC), a multi-hit time-to-digital converter (TDC), and field-programmable gate array (FPGA).[17] Each QTC input has three gain ranges  "Small", "Medium" and "Large" – the resolutions for each are shown in Table.[15]

For each range, analog to digital conversion is conducted separately, but the only range used is that with the highest resolution that is not being saturated. The overall charge dynamic range of the QTC is 0.2–2500 pC, five times larger than the old . The charge and timing resolution of the QBEE at the single photoelectron level is 0.1 photoelectrons and 0.3 ns respectively, both are better than the intrinsic resolution of the 20-in. PMTs used in SK. The QBEE achieves good charge linearity over a wide dynamic range. The integrated charge linearity of the electronics is better than 1%. The thresholds of the discriminators in the QTC are set to −0.69 mV (equivalent to 0.25 photoelectron, which is the same as for SK-III). This threshold was chosen to replicate the behavior of the detector during its previous ATM-based phases.[15]

Gadolinium was introduced into the Super-Kamiokande water tank in 2020 in order to distinguish neutrinos from antineutrinos that arise from supernova explosions.[14][18] This is known as the SK-Gd project (other names include SuperKGd, SUPERK-GD, and similar names).[19] In the first phase of the project, 1.3 tons of a Gd salt (gadolinium sulfate octahydrate, Gd(SO₄)₃⋅(H₂O)₈) were added to the ultrapure water in 2020, giving 0.02% (by mass) of the salt. This amount is about a tenth of the planned final target concentration.[14][18]

Nuclear fusion in the Sun and other stars turns protons into neutrons with the emission of neutrinos. Beta decay in the Earth and in supernovas turns neutrons into protons with the emission of anti-neutrinos. The Super-Kamiokande detects electrons knocked off a water molecule producing a flash of blue Cherenkov light, and these are produced both by neutrinos and antineutrinos. A rarer instance is when an antineutrino interacts with a proton in water to produce a neutron and a positron.[20]

Gadolinium has an affinity for neutrons and produces a bright flash of gamma rays when it absorbs one. Adding gadolinium to the Super-Kamiokande allows it to distinguish between neutrinos and antineutrinos. Antineutrinos produce a double flash of light about 30 microseconds apart, first when the neutrino hits a proton and second when gadolinium absorbs a neutron.[18] The brightness of the first flash allows physicists to distinguish between low energy antineutrinos from the Earth and high energy antineutrinos from supernovas. In addition to observing neutrinos from distant supernovas, the Super-Kamiokande will be able to set off an alarm to inform astronomers around the world of the presence of a supernova in the Milky Way within one second of it occurring.

The biggest challenge was whether the detector's water could be continuously filtered to remove impurities without removing the gadolinium at the same time. A 200-ton prototype called EGADS with added gadolinium sulfate was installed in the Kamioka mine and operated for years. It finished operation in 2018 and showed that the new water purification system would remove impurities while keeping the gadolinium concentration stable. It also showed that gadolinium sulfate would not significantly impair the transparency of the otherwise ultrapure water, or cause corrosion or deposition on existing equipment or on the new valves that will later be installed in the Hyper-Kamiokande.[19][20]

The outer shell of the water tank is a cylindrical stainless-steel tank 39 m in diameter and 42 m in height. The tank is self-supporting, with concrete backfilled against the rough-hewn stone walls to counteract water pressure when the tank is filled. The capacity of the tank exceeds 50 kilotonnes of water.[10]

The basic unit for the ID PMTs is a "supermodule", a frame which supports a 3×4 array of PMTs. Supermodule frames are 2.1 m in height, 2.8 m in width and 0.55 m in thickness. These frames are connected to each other in both the vertical and horizontal directions. Then the whole support structure is connected to the bottom of the tank and to the top structure. In addition to serving as rigid structural elements, supermodules simplified the initial assembly of the ID. Each supermodule was assembled on the tank floor and then hoisted into its final position. Thus the ID is in effect tiled with supermodules. During installation, ID PMTs were pre-assembled in units of three for easy installation. Each supermodule has two OD PMTs attached on its back side. The support structure for the bottom PMTs is attached to the bottom of the stainless-steel tank by one vertical beam per supermodule frame. The support structure for the top of the tank is also used as the support structure for the top PMTs.

Cables from each group of 3 PMTs are bundled together. All cables run up the outer surface of the PMT support structure, i.e., on the OD PMT plane, pass through cable ports at the top of the tank, and are then routed into the electronics huts.

The thickness of the OD varies slightly, but is on average about 2.6 m on top and bottom, and 2.7 m on the barrel wall, giving the OD a total mass of 18 kilotons. OD PMTs were distributed with 302 on the top layer, 308 on the bottom, and 1275 on the barrel wall.

To protect against low energy background radiation from radon decay products in the air, the roof of the cavity and the access tunnels were sealed with a coating called Mineguard. Mineguard is a spray-applied polyurethane membrane developed for use as a rock support system and radon gas barrier in the mining industry.[10]

The average geomagnetic field is about 450 mG and is inclined by about 45° with respect to the horizon at the detector site. This presents a problem for the large and very sensitive PMTs which prefer a much lower ambient field. The strength and uniform direction of the geomagnetic field could systematically bias photoelectron trajectories and timing in the PMTs. To counteract this 26 sets of horizontal and vertical Helmholtz coils are arranged around the inner surfaces of the tank. With these in operation the average field in the detector is reduced to about 50 mG. The magnetic field at various PMT locations were measured before the tank was filled with water.[10]

A standard fiducial volume of approximately 22.5 kilotonnes is defined as the region inside a surface drawn 2.00 m from the ID wall to minimize the anomalous response caused by natural radioactivity in the surrounding rock.

An online monitor computer located in the control room reads data from the DAQ host computer via an FDDI link. It provides shift operators with a flexible tool for selecting event display features, makes online and recent-history histograms to monitor detector performance, and performs a variety of additional tasks needed to efficiently monitor status and diagnose detector and DAQ problems. Events in the data stream can be skimmed off and elementary analysis tools can be applied to check data quality during calibrations or after changes in hardware or online software.[10]

To detect and identify such bursts as efficiently and promptly as possible Super-Kamiokande is equipped with an online supernova monitor system. About 10,000 total events are expected in Super-Kamiokande for a supernova explosion at the center of the Milky Way Galaxy. Super-Kamiokande can measure a burst with no dead-time, up to 30,000 events within the first second of a burst. Theoretical calculations of supernova explosions suggest that neutrinos are emitted over a total time-scale of tens of seconds with about a half of them emitted during the first one or two seconds. The Super-K will search for event clusters in specified time windows of 0.5, 2 and 10 s.[10] Data are transmitted to realtime SN-watch analysis process every 2 min and analysis is completed typically in 1 min. When supernova (SN) event candidates are found, ${\displaystyle {R_{\text{mean}}}}$ is calculated if the event multiplicity is larger than 16, where ${\displaystyle {R_{\text{mean}}}}$ is defined as the average spatial distance between events, i.e.

${\displaystyle {R_{\text{mean}}}={\frac {\sum _{i=1}^{{N_{\text{multi}}}-1}\sum _{j=i+1}^{N_{\text{multi}}}|{r_{\text{i}}}-{r_{\text{j}}}|}{{N_{\text{multi}}}{C_{\text{2}}}}}}$

Neutrinos from supernovae interact with free protons, producing positrons which are distributed so uniformly in the detector that ${\displaystyle {R_{\text{mean}}}}$ for SN events should be significantly larger than for ordinary spatial clusters of events. In the Super-Kamiokande detector, Rmean for uniformly distributed Monte Carlo events shows that no tail exists below ${\displaystyle {R_{\text{mean}}}}$⩽1000 cm. For the "alarm" class of burst, the events are required to have ${\displaystyle {R_{\text{mean}}}}$⩾900 cm for 25⩽${\displaystyle {N_{\text{multi}}}}$⩽40 or ${\displaystyle {R_{\text{mean}}}}$⩾750 cm for ${\displaystyle {N_{\text{multi}}}}$>40. These thresholds were determined by extrapolation from SN1987A data.[10][21] The system will run special processes to check for spallation muons when burst candidates meeting "alarm" criteria and make a primary decision for further process. If the burst candidate passes these checks, the data will be reanalyzed using an offline process and a final decision will be made within a few hours. During the Super-Kamiokande I running, this never occurred. One of the important capabilities for [Super-Kamiokande] is to reconstruct the direction to supernova. By neutrino–electron scattering, ${\displaystyle \nu _{\text{x}}+e^{-}\to \nu _{\text{x}}+e^{-}}$, a total of 100–150 events are expected in case of a supernova at the center of the Milky Way Galaxy.[10] The direction to supernova can be measured with angular resolution

${\displaystyle \delta \theta \sim {30^{\circ } \over {\sqrt {N}}}}$

where N is the number of events produced by the ν–e scattering. The angular resolution, therefore, can be as good as δθ∼3° for a supernova at the center of the Milky Way Galaxy.[10] In this case, not only time profile and the energy spectrum of a neutrino burst, but also the information on direction of supernova can be provided.

There is a process called the "slow control" monitor, as part of the online monitoring system, watches the status of the HV systems, the temperatures of electronics crates and the status of the compensating coils used to cancel the geomagnetic field. When any deviation from norms is detected, it will alert physicists to prompt to investigate, take appropriate action, or notify experts.[10]

To monitor and control the offline processes that analyze and transfer data, a set of software was sophisticatedly developed. This monitor allows non-expert shift physicists to identify and repair common problems to minimize down time, and the software package was a significant contribution to the smooth operation of the experiment and its overall high lifetime efficiency for data taking.[10]

The energy of the Sun comes from the nuclear fusion in its core where a helium atom and an electron neutrino are generated by 4 protons. These neutrinos emitted from this reaction are called solar neutrinos. Photons, created by the nuclear fusion in the center of the Sun, take millions of years to reach the surface; on the other hand, solar neutrinos arrive at the earth in eight minutes due to their lack of interactions with matter. Hence, solar neutrinos make it possible for us to observe the inner Sun in "real-time" that takes millions of years for visible light.[22]

In 1999, the Super-Kamiokande detected strong evidence of neutrino oscillation that successfully explained the solar neutrino problem. The Sun and about 80% of the visible stars produce their energy by the conversion of hydrogen to helium via

${\displaystyle 4p\to {}^{4}\!He+2e^{+}+2\nu _{e}+26.73}$ MeV

Consequently, stars are a source of neutrinos, including the Sun. These neutrinos primarily come through the p-p chain in lower masses, and for cooler stars, primarily through the CNO cycle of heavier masses.

The left frame shows the three principal cycles comprising the p-p chain (p-p I, p-p II and p-p III), and the neutrinos sources associated with these cycles. The right frame shows the CNO-I cycle.

In the early 1990s, particularly with the uncertainties that accompanied the initial results from Kamioka II and the Ga experiments, no individual experiment required a non-astrophysical solution of the solar neutrino problem. But in aggregate, the Cl, Kamioka II, and Ga experiments indicated a pattern of neutrino fluxes that was not compatible with any adjustment of the SSM. This in turn helped motivate a new generation of spectacularly capable active detectors. These experiments are Super-Kamiokande, the Sudbury Neutrino Observatory (SNO), and Borexino. Super-Kamiokande was able to detect elastic scattering (ES) events

${\displaystyle \nu _{x}+e^{-}\to \nu _{x}+e^{-}}$

which, due to the charged-current contribution to ${\displaystyle \nu _{e}}$ scattering, has a relative sensitivity to ${\displaystyle \nu _{e}}$ s and heavy-flavor neutrinos of ∼7:1.[23] Since the direction of the recoil electron is constrained to be very forward, the direction of the neutrinos are kept in the direction of recoil electrons. Here, ${\displaystyle \cos \theta _{Sun}}$ is provided where ${\displaystyle \theta _{Sun}}$ is the angle between the direction of recoil electrons and the Sun's position. This shows that the ${\displaystyle {}^{8}\!B}$ solar neutrino flux can be calculated to be ${\displaystyle 2.40\pm 0.03(stat.){}_{-0.07}^{+0.08}\!(sys.)\times 10^{6}cm^{-2}s^{-1}}$. Comparing to the SSM, the ratio is ${\displaystyle {Data \over SSM_{BP98}}=0.465\pm 0.005(stat.){}_{-0.013}^{+0.015}\!(sys.)}$.[24] The result clearly indicates the deficit of solar neutrinos.

Atmospheric neutrinos are secondary cosmic rays produced by the decay of particles resulting from interactions of primary cosmic rays (mostly protons) with Earth atmosphere. The observed atmospheric neutrino events fall into four categories. Fully contained (FC) events have all their tracks in the inner detector, while partially contained (PC) events have escaping tracks from the inner detector. Upward through-going muons (UTM) are produced in the rock beneath the detector and go through the inner detector. Upward stopping muons (USM) are also produced in the rock beneath the detector, but stop in the inner detector.

The number of observed number of neutrinos is predicted uniformly regardless of the zenith angle. However, Super-Kamiokande found that the number of upward going muon neutrinos (generated on the other side of the Earth) is half of the number of downward going muon neutrinos in 1998. This can be explained by the neutrinos changing or oscillating into some other neutrinos that are not detected. This is called neutrino oscillation; this discovery indicates the finite mass of neutrinos and suggests an extension of the Standard Model. Neutrinos oscillate in three flavors, and all neutrinos have their rest mass. Later analysis in 2004 suggested a sinusoidal dependence of the event rate as a function of "Length/Energy", which confirmed the neutrino oscillations.[25]

The K2K experiment was a neutrino experiment from June 1999 to November 2004. This experiment was designed to verify oscillations observed by Super-Kamiokande through muon neutrinos. It gives first positive measurement of neutrino oscillations in conditions that both source and detector are under control. The Super-Kamiokande detector plays an important role in the experiment as the far detector. Later experiment T2K experiment continued as the second generation follow up to the K2K experiment.

T2K (Tokai to Kamioka) experiment is a neutrino experiment collaborated by several countries including Japan, United States and others. The goal of T2K is to gain deeper understanding of parameters of neutrino oscillation. T2K has made a search for oscillations from muon neutrinos to electron neutrinos, and announced the first experimental indications for them in June 2011.[26] The Super-Kamiokande detector plays as the "far detector". The Super-K detector will record the Cherenkov radiation of muons and electrons created by interactions between high energy neutrinos and water.

The proton is assumed to be absolutely stable in the Standard Model. However, the Grand Unified Theories (GUTs) predict that protons can decay into lighter energetic charged particles such as electrons, muons, pions, or others which can be observed. Kamiokande helps to rule out some of these theories. Super-Kamiokande is currently the largest detector for observation of proton decay.

Water purification system schematic

The 50 kilotons of pure water is continually reprocessed at rate about 30 tons/hour in a closed system since early 2002. Now, raw mine water is recycled through the first step (particle filters and RO) for some time before other processes, which involve expensive expendables, are imposed. Initially, water from the Super-Kamiokande tank is passed through nominal 1 μm mesh filters to remove dust and particles, which reduce the transparency of the water for Cherenkov photons and provide a possible radon source inside the Super-Kamiokande detector. A heat exchanger is used to cool down the water in order to reduce the PMT dark noise level as well as suppress the growth of bacteria. Surviving bacteria are killed by a UV sterilizer stage. A cartridge polisher (CP) eliminates heavy ions, which also reduce water transparency and include radioactive species. The CP module increases the typical resistivity of recirculating water from 11 MΩ cm to 18.24 MΩ cm, approaching chemical limit.[10] Originally, an ion-exchanger (IE) was included in system, but it was removed when IE resin was found to be a significant radon source. The RO step that removes additional particulates, and the introduction of Rn-reduced air into the water that increases radon removal efficiency in the vacuum degasifier (VD) stage which follows were installed in 1999. After that, a VD removes dissolved gases in the water. These dissolved gases in water are a serious background event source for solar neutrinos in the MeV energy range and the dissolved oxygen encourages the growth of bacteria. The removal efficiency is about 96%. Then, the ultra filter (UF) is introduced to remove particles whose minimum size corresponds to molecular weight approximately 10,000 (or about 10 nm diameter) thanks to hollow fiber membrane filters. Finally, a membrane degasifier (MD) removes radon dissolved in water, and the measured removal efficiency for radon is about 83%. The concentration of radon gases is miniaturized by realtime detectors. In June 2001, typical radon concentrations in water coming into the purification system from the Super-Kamiokande tank were less than 2 mBq m⁻³, and in water output by the system, 0.4±0.2 mBq m⁻³.[10]

Air purification system schematic

Purified air is supplied in the gap between the water surface and the top of the Super-Kamiokande tank. The air purification system contains three compressors, a buffer tank, dryers, filters, and activated charcoal filters. A total of 8 m³ of activated charcoal is used. The last 50 L of charcoal is cooled to −40 °C to increase removal efficiency for radon. Typical flow rates, dew point, and residual radon concentration are 18 m³/h, −65 °C (@+1 kg/cm²), and a few mBq m⁻³, respectively. Typical radon concentration in the dome air is measured to be 40 Bq m⁻³. Radon levels in the mine tunnel air, near the tank cavity dome, typically reach 2000–3000 Bq m⁻³ during the warm season, from May until October, while from November to April the radon level is approximately 100–300 Bq m⁻³. This variation is due to the chimney effect in the ventilation pattern of the mine tunnel system; in cold seasons, fresh air flows into the Atotsu tunnel entrance that is a relatively short path through exposed rock before reaching the experimental area, while in the summer, air flows out the tunnel, drawing radon-rich air from deep within the mine past the experimental area.[10]

In order to keep radon levels in the dome area and water purification system below 100 Bq m⁻³, fresh air is continually pumped at approximately 10 m³/min from outside the mine which generates a slight over-pressure in the Super-Kamiokande experimental area to minimize the entry of ambient mine air. A "Radon Hut" (Rn Hut) was constructed near the Atotsu tunnel entrance to house equipment for the dome air system: a 40 hp air pump with 10 m³ min⁻¹ /15 PSI pump capacity, air dehumidifier, carbon filter tanks, and control electronics. In autumn 1997, an extended intake air pipe was installed at a location approximately 25 m above the Atotsu tunnel entrance. This low level satisfies that goals of air quality so that carbon filter regeneration operations would no longer be required.[10]

The offline data processing system is located in Kenkyuto and is connected to Super-Kamiokande detector with 4 km FDDI optical fiber link. Data flow from online system is 450 kbytes s⁻¹ on average, corresponding to 40 Gbytes day⁻¹ or 14 Tbytes yr⁻¹. Magnetic tapes are used in offline system to store data and most of the analysis is accomplished here. The offline processing system is designed platform-independent because different computer architectures are used for data analysis. Because of this, the data structures are based on ZEBRA bank system developed in CERN as well as the ZEBRA exchange system.[10]

Event data from Super-Kamiokande online DAQ system basically contains a list of number of hit PMT, TDC and ADC counts, GPS time-stamps and other housekeeping data. For solar neutrino analysis, lowering the energy threshold is a constant goal, so it is a continual effort to improve the efficiency of reduction algorithms; however, changes in calibrations or reduction methods require reprocessing of earlier data. Typically, 10 Tbytes of raw data is processed every month so that a large amount of CPU power and high-speed I/O access to the raw data. In addition, extensive Monte Carlo simulation processing is also necessary.[10]

Offline system was designed to meet demand of all these: tape storage of a large database (14 Tbytes yr−1), stable semi-realtime processing, nearly continuous re-processing and Monte Carlo simulation. The computer system consists of 3 major sub-systems: the data server, the CPU farm and the network at the end of Run I.[10]

A system dedicated to offsite offline data processing was set up at the Stony Brook University in Stony Brook, NY to process raw data sent from Kamioka. Most of the reformatted raw data is copied from system facility in Kamioka. At Stony Brook, a system was set up for analysis and further processing. At Stony Brook the raw data were processed with a multi-tape DLT drive. The first stage data reduction processes were done for the high energy analysis and for the low energy analysis. The data reduction for the high energy analysis was mainly for atmospheric neutrino events and proton decay search while the low energy analysis was mainly for the solar neutrino events. The reduced data for the high energy analysis was further filtered by other reduction processes and the resulting data were stored on disks. The reduced data for the low energy were stored on DLT tapes and sent to University of California, Irvine for further processing.

This offset analysis system continued for three years until their analysis chains were proved to produce equivalent results. Thus, in order to limit manpower, collaborations were concentrated to a single combined analysis[27]