5.1. The J‑PARC Neutrino Beam

The Japanese Proton Accelerator Research Complex (J‑PARC) in Tokai on the East coast of Japan is a new facility built on the site where Japan developed their first nuclear reactors (Figure 5.2). The accelerator chain begins with a linear accelerator (LINAC) which boosts H^- ions up to a kinetic energy of 181 MeV before passing through charge-stripping foils. The resulting protons are then injected into a 3 GeV Rapid-Cycling Synchrotron (RCS) which in turn feeds a 30 GeV Main Ring (MR) from which the protons destined for the neutrino beamline are extracted.

Fast kicker magnets extract protons from the MR so that they can be bent inside the ring and along the neutrino beamline. As they approach the target the protons are focused and directed downwards by 3.6 ° with respect to the horizontal. Although the direction to Super‑Kamiokande is actually only 1.3 ° downwards, this additional vertical offset forms part of the 2.5 ° off-axis angle of Super‑Kamiokande with respect to the beam.

The protons then collide with the target which is a graphite rod 914 mm long and 26 mm in diameter, with a density of 1.8 g cm^-3. It is cooled with helium gas and is designed to operate at a beam power of 750 kW, though to date the maximum delivered has been 200 kW.

The target is located inside the first of three magnetic horns which select and focus the secondary particles (mostly π⁺) produced in the target (Figure 5.3). The first horn is primarily responsible for “collecting” the pions, while the second and third horns further downstream focus them down a 96 m decay volume.

Positive pions are the most common neutrino-producing particles which enter the decay volume though there are also a significant number of positive kaons. In addition, a small number of wrong-sign pions and kaons are unavoidable and it is also possible for μ⁺ produced from π⁺/K⁺ decays to also decay and contribute to ν_μ and ν_e contaminations (Table 5.1).

The decay volume ends 117.5 m from the target with two muon monitoring detectors which provide valuable data for validating the beam simulation.

One of the notable features of the T2K neutrino beam is that it is the world's first off-axis neutrino beam. As was mentioned earlier, the beam is directed such that the Super‑Kamiokande detector is situated 2.5 ° off the beam's primary axis. Directly along the beam's axis, a detector is exposed to a broad range of decay kinematics giving a correspondingly broad neutrino energy spectrum. At fixed angles off-axis however, only a more select range of decay kinematics contribute resulting in a beam with a much narrower energy spectrum (Figure 5.5). There are two good reasons for doing this. The first is that the peak of an off-axis spectrum is lower than for an on-axis beam, allowing T2K to push the peak energy down towards the oscillation maximum energy of ∼600 MeV - giving higher statistics where it matters most. The second advantage comes from the removal of a large flux of higher energy neutrinos. These neutrinos do not contribute to the oscillation signal at Super‑Kamiokande but will generate background interactions - in particular NC interactions - which can be reconstructed in the signal region. For oscillation analyses then, an off-axis beam delivers higher statistics with lower backgrounds.

At the end of T2K Run 4, in May 2013, the beam had delivered a total of 2.57 × 10²⁰ POT (Table 5.2). This is delivered in ∼5 µs long “spills” of 8 (6 in T2K Run 1) “bunches” each of which is approximately 15.0 ns wide (Figure 5.6).

Although designed to take a currents of 320 kA, thus far all T2K physics data has been taken while operating the horns at 250 kA, with the exception of T2K Run 3b when they were operated at 205 kA. The effect of this reduction in horn current is mainly a reduction of around 10 % in the total flux at the peak (Figure 5.7)·

5.2. Super‑Kamiokande

Super‑Kamiokande is perhaps the most famous experiment in the history of neutrino physics and played an important part in establishing evidence for both solar and atmospheric neutrinos oscillating. It also holds the world's leading limit on proton decay [3]. Now in it's fourth running configuration, Super‑Kamiokande is being re-utilised as the far detector for T2K.

The device itself is a large water-Cherenkov detector within a vertical cylindrical tank. A stainless steel frame forming a smaller cylinder within the tank separates the volume into two regions known as the inner and outer detectors. The entire volume is filled with 50 kT of ultra-pure water though, in practice, the fiducial mass is somewhat smaller.

The inner detector is the target region, and it is instrumented on all sides by 11129 inward-facing photo-multiplier tubes (PMTs) which cover ∼40 % of the surface area. Particles produced in the inner detector which are above the Cherenkov threshold produce rings of light which can be detected by these PMTs and from which the event can be reconstructed. The position of the interaction vertex can be determined from the position, shape and diameter of the ring. The thickness of the light ring is a measure of the distance the particle travelled, with exiting particles generating a completely filled circle. Finally, primitive particle identification for separating showering and non-showering particles (effectively electron-muon separation) can be done using the diffuseness of the outer part of the ring (Figure 5.9). It is the fact that this particle identification cannot discriminate between a showering electron or photon that makes NC π⁰ production such an important background for ν_e selection.

The outer detector was designed as a veto for incoming (mostly cosmic-ray) particles and is essentially a 2 m thick region surrounding the inner detector. The outwards facing wall is also instrumented with PMTs albeit much more sparsely - the tubes are fewer and smaller. In T2K however, because the direction and timing requirements are already quite effective at selecting beam events, interactions in the outer detector can also contribute to the oscillation analysis.

• Inner Detector:
  • 11129 × 0.5 m diameter PMTs
  • Cylindrical volume 33.8 m diameter, 36.2  high
• Outer Detector:
  • 1885 × 0.2 m diameter outward-facing PMTs on inner wall
  • Cylindrical tube, outer diameter 39 m, inner diameter ∼34.3 m

5.3. INGRID

The Interactive Neutrino Grid (INGRID) [4] is the on-axis near detector primarily designed to constrain the beam normalisation and direction. It achieves this by placing identical modules at a variety of off-axis angles in a plane perpendicular to the beam. The direction can then be determined from the beam's profile across this plane, and the normalisation from the rate of interactions.

The baseline design consists of 16 identical modules (Figure 5.10). 14 of these are placed in a 10 m × 10 m cross-shape, centred on the beam's axis. This is formed of a vertical stack of 7 modules, which sits in front of a horizontal row of a further 7 modules. In between these two sets are two modules located above the horizontal row and on either side of the vertical stack. These off-axis modules assist in characterising any asymmetries in the beam shape - which is important for an off-axis beam.

Each standard module is primarily constructed of a central tracking region surrounded by veto planes (Figure 5.11). The tracking region is approximately 1.2 m square and is comprised of 9 × 65 mm thick iron plates, separating the first 10 of 11 planes of plastic scintillator. Each of these planes contains 2 layers of 10 mm × 50 mm bars arranged in an x-y configuration. The central tracking region is surrounded above, below and on the sides by veto planes of 22 plastic scintillator bars aligned perpendicular to the beam.

Because of the large amount of iron dead-material and low tracking granularity, INGRID modules are unable to reconstruct much more than a primary μ^-, though this is sufficient for its task of profiling the beam. After the construction of INGRID, additional components were assembled in to a scintillator-only module which is referred to as the “proton module” - so named because its totally active design allows it, for example, to also reconstruct a proton from a CC QE interaction. This module was placed centrally and between the vertical and horizontal arms (inline with the off-axis modules).

During T2K Runs 1 and 2 INGRID consistently recorded a mean event rate of 1.5 / 10¹⁴ POT, agreeing with the predicted rate from simulation within uncertainty: data / simulation = 1.06 ± 0.04. Fitting the number of events in each module as a function of off-axis angle allowed it to measured the direction of the T2K beam to an accuracy of around 0.4 mrad, well below the target resolution of 1.0 mrad (Figure 5.12).

5.4. ND280

ND280 is the experiment's off-axis near-detector, located in the near-detector facility 280m downstream of the target, and slightly downstream of INGRID.

As discussed earlier, the primary requirements of the detector are:

• Measurement of NC π⁰ production on water
• Studies of inclusive and exclusive CC ν_μ interactions, on nuclear targets and specifically on water
• A measurement of the ν_e contamination of the initial beam

The first goal requires a target of water on which neutrinos can interact to produce π⁰s. Almost immediately after creation these pions will decay via π⁰ → γγ requiring the presence of high density material to encourage the photons to shower if they are to be detected.

Meanwhile, the study of CC ν_μ interactions requires precise measurements of the primary lepton's angle and momentum, in addition to the ability to observe secondary particles emerging from the nucleus. The former drives the design towards lower-densities and the latter requires an active target.

In an ideal world a neutrino detector should have a single target region but, unfortunately for ND280, the dual requirements of efficiently converting photons while maintaining good position resolution cannot be simultaneously satisfied. The same is true for the requirements of having a water target, and an active target. As a result the detector is a compromise of multiple sub-systems.

Another significant influence on the design of ND280 is the need for a magnet. Due to their penetrating nature, the momentum of muons cannot practically be measured calorimetrically in such a confined space but must instead be done via curvature in a magnetic field. Such a field also brings the added capability to distinguish μ^- from μ⁺ - valuable when ν_μ make up ∼6% of the total flux and the majority of cosmic-ray muons are μ⁺.

The design conceived to meet all these requirements (Figure 5.13) placed the primary detectors in a central region, known as the “basket”, and surrounded them on all but the upstream-side with electro-magnetic calorimeters. Upstream-most in the basket is the Pi-Zero Detector (P0D) which combines water, metal foils and plastic scintillator to provide the target, photon conversion and reconstruction required for the study of NC π⁰s. Downstream of the P0D is the “tracker”, in which three Time Projection Chambers (TPCs) sandwich two plastic scintillator detectors referred to as Fine Grained Detectors (FGDs). The FGDs act as the target for ND280's interaction studies - offering sufficient mass for a reasonable interaction rate in addition to an active target to access some of the activity near an interaction vertex. The argon-gas TPCs give excellent tracking, for direction and momentum reconstruction, and particle identification from dE/dx. They serve both the FGDs and the P0D, tracking any forward-escaping particles.

The surrounding electromagnetic calorimeters (ECals) consist of two types with very different remits. The more substantial Tracker ECals provide particle identification and energy measurements for final-state particles leaving the tracker. The much more limited P0D ECals which surround the P0D are there only to catch muons and photons escaping at high angles, and to veto incoming backgrounds.

All these systems are enclosed within a magnet which generates a 0.2 T magnetic field and houses the Side Muon Ranging Detector (SMRD) - planes of scintillator between the plates of the iron yoke - that acts primarily as a cosmic trigger. However encasing the detector within the former UA1 and NOMAD magnet which was available substantially restricts the available space to a box approximately 7 m long × 3.6 m wide × 3.5 m high.

5.4.1. The Pi-Zero Detector

The P0D [5] is the upstream-most detector in the ND280 basket with a central water target region sandwiched by two ECal regions (Figure 5.14). The water targets alternate with plastic scintillator modules and brass foils to encourage the photons to convert. The ECal regions meanwhile alternate lead sheets between the plastic scintillator modules to maximise photon conversion and containment. As a final feature, the water targets of the P0D can be filled and drained during running to aid in separating the water contribution to interactions in the fiducial volume. In its most recent technical note [6] the P0D reported a π⁰ reconstruction efficiency of 3.6 % (though updated analyses are reporting improvements upon this).

5.4.3. The Time Projection Chambers

The three Time Projection Chambers (TPCs) [9] are placed immediately downstream of each of the three target sub-detectors (P0D, FGD 1 and FGD 2) to provide tracking of the particles which emerge. For all but the highest-angle tracks they achieve position resolutions < 1.0 mm, allowing them to accurately determine the charge and momentum of particles from their curvature in the magnetic field (with transverse momentum resolution σ_p/p² ∼ 0.1 %).

They also provide particle identification by vertically segmenting tracks, and calculating the mean rate of energy loss (dE/dx) from the path-length corrected energy deposited in each segment (after truncating off the highest 30 %). This is then compared with the expected value from simulation to identify the particle (Figure 5.15). The resolution on the mean dE/dx is better than 8 % for MIPs, making the TPCs the ND280's primary particle identification tool [9].

The TPCs are all constructed identically (Figure 5.16). A central cathode separates two drift chambers containing an argon-based gas, and readout electronics are placed on the lateral sides. Each drift chamber has twelve 342 mm × 359 mm “micromegas” readout tiles, each of which provides a grid of 48 × 36 pads.

5.5. Common Components

The P0D, FGDs, ECals, SMRD and INGRID all contain plastic scintillator as their active components. The extruded plastic scintillator was produced at Fermilab and is composed of polystyrene doped with wavelength shifting scintillators PPO (1 %) and POPOP (0.03 %) which results in blue light emission (420 nm). The plastic was co-extruded with a 0.25 mm thick reflective layer of TiO₂ to keep scintillation light in and external light out.

This scintillation light is then collected and transported by 1 mm diameter optical fibres which are strung down 3 mm diameter extruded holes in the centre of the bars (or laid in a groove in the SMRD planes). These fibres, manufactured by Kuraray, are also wavelength shifting and convert the blue light spectrum to a green (476 nm) that is well matched to the light sensors attached to the ends.

These light sensors are Multi-Pixel Photon Counters (MPPC) [10] manufactured by Hamamatsu which, unlike photomultiplier tubes, can still be operated in a magnetic field. They consist of 667 pixels arranged accross a 1.3 mm × 1.3 mm active area, and boast a photon detection efficiency of ∼20 % for the light received from the optical fibres.

5.6. Simulations

As with all complex experiments it is necessary for T2K to run simulations in order to understand how the measurements recorded by its detectors correspond to the physics which occurred.

The J‑PARC neutrino beam is simulated for the purpose of providing flux predictions [2] to the near and far detectors using software collectively known as “JNUBEAM”. The simulated beamline geometry includes the target, horns, decay volume and beam monitors. 30 GeV protons are directed onto the graphite target, and the resulting particles are tracked until they decay to neutrinos or stop. The simulation of primary interactions inside the target are based on data from the NA61/SHINE [11] experiment, secondary interactions are simulated with FLUKA [12], and interactions outside the target use GEANT3/GCALOR [13].

The simulation of the near detector geometries is done in GEANT4 [14] as is the passage of final-state particles emerging from neutrino-nucleus interactions. The neutrino interactions themselves can be simulated with either the GENIE or NEUT generators (see Chapter 3), which propagate neutrinos from the beam flux simulation through the geometry exported from the GEANT4 simulation. After the particles have been propagated through the near detectors a custom software package, called “ELECSIM”, simulates the response of the detector and its electronics to the energy deposited. In the scintillator sub-detectors this includes simulating the light emitted in response to energy deposition, the light's transport through the bar and down the optical fibres, the response of the MPPCs to that light and the electronics chain thereafter. For the TPCs ELECSIM simulates the electron drift, response of the micromegas and, again, the electronics chain which follows.

Super‑Kamiokande's “SKDETSIM” is a fortran-based software package which is responsible for the geometry and final-state particle simulations. Again, the neutrino interactions can be generated with GENIE or NEUT using the neutrino flux from the beam simulation.

5.7. ND280 Software & Reconstruction

The first processing stage in the ND280 software chain is calibration, and is the last stage at which simulation and data are treated differently. For data the goal is to correct for the various effects introduced by the scintillator bars, optical fibres and readout electronics. For simulation the goal is to replicate the output from this process. In both cases the result is a collection of calibrated “hits” with a position, time and information on the energy deposited (often referred to as “charge”).

The ND280 reconstruction software is essentially a two-stage process as a result of the very different time/space/charge information provided by the different sub-detector systems, which makes a unified reconstruction algorithm unsuitable. Instead, each sub-detector has its own dedicated reconstruction algorithm which groups associated hits together, fits the resulting objects, and calculates properties for use by analysers (such as the TPC particle identification, or FGD vertex activity). The last of these to be run is the TPC reconstruction, which needs to find matching objects in neighbouring detectors in order to determine the time at which the track was created.

The results from the individual sub-detector reconstructions are then passed to a “global” reconstruction, which matches objects with compatible start/end points and directions between detectors. These combined objects are re-fitted using a Kalman filter from the Recpack [15] toolkit, utilising geometric and magnetic field models, to create final “reconstructed objects” with position, direction and momentum measurements..