ARICH in a Nutshell
This page gives a general description of the Aerogel RICH detector and provides promotional images to be used in the Belle2 presentation.
The ARICH detector provides particle identification system for the Belle2 in the front-end endcap region. It's main objective is to provide good separation (
) between kaons and pions in the full kinematical region of the experiment (about 0.5-4.0 GeV
). This is of great importance not only for the reconstruction of decay modes but also for the efficiency of flavor tagging algorithms.
Principle of operation
ARICH relies on the relation between the emission angle of Cherenkov photons in a transparent medium and the charged particle velocity,
(where n is the refractive index of the medium). Aerogel tiles are used as a radiator. In aerogel emitted photons then propagate through ~20 cm of an expansion volume and hit the photon detectors (HAPDs). For pions and kaons of equal momenta this is sketched bellow left. On the right, the Cherenkov angle vs. particle momentum is shown for pions and kaons. At 3.5 GeV
the difference between the pion and kaon Cherenkov angle is ~30 mrad.
Novel Focusing configuration
To improve performance we use a little trick. From one point of view it is desirable to have as thick aerogel layer as possible, as this increases the number of emitted photons. On the other hand, thicker aerogel increases the uncertainty in the photon emission point, leading to worse reconstructed angle resolution. To overcome this difficulty we use two aerogel layers with different refraction indices, which are chosen so that the two rings, from the first and the second layer, overlap on the detector plane. In this way we almost double the number of photons, without significantly affecting angle resolution. The concept is sketched bellow, along with its confirmation from the beamtest data (reconstructed Cherenkov angle distribution is shown for the case of a single aerogel layer and for the case of two layers in focusing configuration. The number of photons is ~ the same, while the angle resolution is significantly improved in the second case).
As a radiator a silica aerogel is used. Aerogel is an amorphous, highly porous solid of silicon dioxide (SiO2
), that is widely used as a Cherenkov radiator because of its tunable, intermediate refractive index and good optical transparency. In the ARICH the 3.5 m2 radiator plane is covered by two layers of wedge-shaped aerogel tiles of size 17 x 17 x 2 cm, where the refractive index of tiles in the first and the second layer is n_1 = 1.045 and n_2 = 1.055, respectively. The transmission length of n_1 tiles is ~45 mm and of n_2 tiles about 35 mm. All of tiles to be used were already produced and tested for optical properties. Below are some tile example photos, including the aluminum support frame mockup.
As a photon detector a newly developed Hybrid Avalanche Photo Detector (HAPD) will be used (developed in joint effort of the Belle II collaboration and Hamamatsu Photonics). The principle of its operation is shown below. Incident photon is converted into photo-electron by a bi-alkali photo-cathode with peak quantum efficiency of ~30% at 400 nm.The electron is then accelerated in a vacuum tube with high electric field towards the segmented avalanche photo-diode (APD) with 144 pads of size 5.1 × 5.1 mm. APD is a photo-diode with the applied reverse bias voltage (~300 V), resulting in an internal amplification (avalanche gain) due to the impact ionization. In the HAPD the avalanche gain of the photo-diode is about 40 and the bombardment gain (due to the electron acceleration) is about 1700. Therefore, a detection of a signle photon results in an avalanche of about 60000 electrons. For the readout a dedicated high gain and low noise electronics was developed: a digitizer ASIC which consists of a preamplifier, a shaper and a comparator (SA02) is followed by an FPGA (Xilinx Spartan-6 XC6SLX45
), where the hit information is recorded and communicated to further stages of the experiment data acquisition. On the right photo, HAPD with the attached front-end electronics board is shown.
In total ARICH consists of 420 HAPD modules arranged in seven concentric rings (r_in = 56cm, r_out = 114 cm) and of 2 × 124 aerogel tiles of wedge shape, as shown bellow. In order to maintain good performance also on the outer edge of the detector, where Cherenkov photons would miss the photo-sensitive area, 18 planar mirror plates are placed as shown and sketched bellow right.
Below left photo shows the current status of the detector construction. Very recently, one sixth of all HAPDs (70) were mounted on the detector aluminum support structure. The left two pictures show full ARICH geometry, implemented in the Geant4 detector simulation. On the right, simulation of 2 pion tracks (in blue) with 3.5 GeV
hitting ARICH is shown (with green lines being Cherenkov photons).
Performance in beamtests
To test the performance of the designed ARICH we constructed a small prototype, with two consecutive aerogel tiles and six HAPD modules, arranged as in a part of actual detector layout.
In the recent few years we have performed three beam tests, at KEK in 2009 (3 GeV
electron beam), at CERN in 2011 (120 GeV
hadron beam), and at DESY in 2013 (4-5 GeV
/c electron beam).
Below the photo of the setup from 2013 beamtest is shown. The middle figure shows the accumulated distribution of reconstructed cherenkov angle, and its projection on "theta" angle is shown in the left plot. The single photon angle resolution is about 13 mrad and on average 9 photons per track are detected.
Performance in simulation (Geant4)
ARICH detector geometry and event reconstruction algorithm are implemented in the basf2 framework. Using the full Belle2 simulation we study the performance of ARICH in the full kinematic region by shooting 0.5 - 4.0 GeV
pions and kaons from the interaction point. The obtained Cherenkov angle distributions of single photons from tracks with 3.5 GeV
are shown below left. On average we detect 12 photons per track (middle plot) and their angle distribution is then used to calculate the likelihood of track belonging to a pion or kaon. The distribution of pion/kaon likelihood difference for pion and kaon tracks is shown on the right plot. As desired, two well separated peaks are seen.
More technically, for each track hitting ARICH we calculate the likelihood for each charged particle hypothesis
is the expected number of detected photons for id particle hypothesis, the sum runs over all pads (channels) that were hit in an event, and
is the number of expected hits on the i-th pad with a hit, assuming the id-th particle hypothesis. Although it sounds complicated, it is just a comparison of the observed response (number of hits and their spatial distribution) with the response expected for a given hypothesis.
From the likelihood distributions at different momenta we produce the bottom left plot which shows the kaon indentification efficiency vs. kaon momentum, where at each momentum we allow only 2% of pion fake rate (2% of pions are identified as kaons). The bottom middle plot shows the separation between pion and kaon likelihood peaks in units of effective peak sigmas (assuming Gaussian peak shape) as a function of momentum. Next we also test the performance by simulation of physics events containing a pair of B mesons. For the pions and kaons from the
decays that hit ARICH we obtain the separation as shown on the bottom right plot. Allowing 2% of pions to "fake" kaons we correctly identify 98% of kaons.
Simulation vs. beamtest comparison
Finally we compare the distributions of Cherenkov angle and number of photons/track as obtained form simulation and from beamtest data. This is shown bellow. Note that here the number of photons/track is much lower as nominally, as we exclude 2 edge rows of HAPD channels from the analysis (55% of HAPD surface). In the beamtest the distribution of photons on these edge channels is highly distorted due to distorted internal electric field in the HAPD. This effect is not present when the HAPD is operated in magnetic field, as will be in Belle II.
-- Main.luka - 26 Jul 2016