Upgrade of LHCb-RICH detectors

The two RICH detectors installed in LHCb have performed successfully during the 2010–2012 data taking period. The data from these detectors were essential to most of the physics results published by LHCb. During this period LHCb collected data corresponding to 3 fb 1 of integrated luminosity. From 2019 onwards LHCb plans to collect data corresponding to 5 fb 1 of integrated luminosity per year in order to improve the statistical precision of the experiment and to search for very rare B-decays and D-decays. This would increase the readout rate and occupancies for the RICH detectors. To cope with these challenges, the Hybrid Photon Detectors (HPDs) are being replaced by MaPMTs and the optics of the upstream RICH is being modified. Simulations show that these measures recover the particle identification performance that would have been otherwise lost at the high luminosity. & 2015 CERN for the benefit of the Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The LHCb experiment [1] at CERN aims to study the CP violation and rare processes in the decays of hadrons with b and c quarks in order to search for physics beyond the standard model (SM). From the measurements done by LHCb and other experiments so far, it appears that flavor changing processes are consistent with the CKM mechanism in SM and that large sources of flavor symmetry breaking are excluded at the TeV scale. Hence in order to measure small deviations from SM, very high precision measurements are needed. The LHCb upgrade aims for a significant increase in the statistical precision of its measurements.
The beam conditions in LHCb during different run periods and the corresponding detector configurations are listed in Table 1.
As indicated in this table, from 2019 onwards, the luminosity in LHCb is expected to reach up to 2 Â 10 33 cm À 2 s À 1 . But there are two main bottlenecks for collecting and processing the data produced at this high luminosity: The maximum readout rate of the front end readout electronics is 1 MHz.
For hadronic channels, the LHCb level 0 hardware trigger yield saturates at high luminosity and therefore its discrimination power is limited.
In order to solve these problems, it is proposed [2] to read out all the detectors at 40 MHz and use software triggers to select events. This would increase the rate of data stored to tape from 5 kHz that was achieved in 2012 to 20 kHz. In 2012 LHCb collected data corresponding to an integrated luminosity of 2 fb À 1 . It is envisaged that this would increase to 5 fb À 1 per year with the upgraded LHCb from 2019 onwards, for another 10 years.

Design of RICH detector upgrade
The current RICH system contains two RICH detectors, the first one (RICH 1) is installed upstream of the LHCb magnet and the second one (RICH 2) is installed downstream of the LHCb magnet. The RICH 1 uses an aerogel radiator for particle identification (PID) in the momentum range 1.5-10 GeV/c and C 4 F 10 gas for momenta up to 40 GeV/c. The RICH 2 uses the CF 4 radiator for momenta up to 100 GeV/c. A detailed description of the current RICH system and its performance can be found in Ref. [1,3].
The overall geometry structure for the upgraded RICH is the same as that for the current RICH system, as described in the LHCb re-optimization TDR [4]. However, several components will be changed as described below.
Photodetectors. The current RICH system uses HPDs [5] as photodetectors and each HPD has an encapsulated readout chip, whose maximum readout rate is 1 MHz. As a result, in order to read out at 40 MHz during the upgrade period, the HPDs would need to be replaced by a new photodetector. The Hamamatsu 1 R11265 MaPMT has been proposed to be a candidate for this. This is to be read out with a custom-made chip named CLARO [6]. Each MaPMT has an 8 Â 8 array of pixels. For the MaPMT prototypes, the photocathode is made of super-bialkali and the entrance window is made of borosilicate.
Testing the prototypes of R11265 MaPMTs is underway and arrays of such MaPMTs are used in simulation studies. In these studies the quantum efficiency (QE) used is taken from the manufacturer's specifications [7]. In order to provide magneticfield shielding, each set of 4 Â 4 MaPMTs is kept inside a module with walls made of a high magnetic permeability material.
Radiator. The PID using the aerogel radiator is found to be difficult even at the current luminosity due to its low photon yield and worse resolution [3] compared to those from the gas radiators. Therefore it is expected that this radiator will be removed for the upgrade. In order to cover the PID in the momentum range 1-10 GeV/c a time-of-flight detector [8] is being proposed and its development is ongoing. But this activity may take more than 6 years and as a consequence such a detector is not expected to be installed by 2018.
In the simulation studies for upgrade, the aerogel is removed. For the tracks with momenta below the kaon threshold of 9.3 GeV/c in C 4 F 10 , the PID algorithm uses the fact that any signal hits produced by such tracks must be from pions or muons or electrons.
Optical arrangement. The increase in luminosity from 2019 onwards results in an increase in the number of charged tracks and this leads to the number of hits in the RICH detector being augmented. As a result, the peak occupancy in the RICH 1 detector was found to exceed 35%. This can be seen in Fig. 1 where the distribution of the hit coordinates on the detector plane at the high luminosity, labeled as Lumi20 in Table 2, is plotted.
This increase in occupancy degrades the PID performance and the optics was redesigned to recover the performance. The main constraints on the redesign were the following: The new RICH has to be built by 2018. The overall size of the RICH 1 had to remain constant in order to complete the installation within the 18 months planned for the LHC shutdown and to avoid interference with the tracking detectors adjacent to RICH 1.
The result of the optics redesign was that the radius of curvature of the spherical mirror was increased from 2710 mm to 3800 mm and it was moved downstream by 86 mm. The removal of the aerogel also increases the radiator length of C 4 F 10 . The position and orientation of the flat mirror and the detector plane were adjusted accordingly. This allowed us to reduce the occupancy of the hits in the detector plane and to gain more radiator length. The distribution of the hit coordinates on the detector plane after all these modifications is shown in Fig. 2.
For RICH 2 the peak occupancy was below 7% even at Lumi20 and therefore there was negligible degradation in the PID performance from this. As a result the RICH 2 optics was not proposed to be modified. Table 1 Beam conditions and detector configurations for LHCb in different years. The conditions for the future runs are those proposed in Ref. [2]. In the second column, the label 'L' stands for average luminosity in 10 32 cm À 2 s À 1 .
Period ( Table 2 Beam configurations used in RICH upgrade simulation. Here ν is the average number of pp interactions per bunch crossing. Luminosity label Luminosity in LHCb (cm À 2 s À 1 ) Collision energy (TeV) ν

Simulation studies
The goals of the simulation studies were to study the RICH detector occupancies indicated in the previous section and to evaluate the PID performance of the upgraded RICH system.

Simulation software
The simulations for RICH upgrade are performed using the LHCb simulation framework [9]. For this, the pp collisions are generated using PYTHIA [10] with the beam configurations which existed in 2012 and with those expected from 2019 onwards, as listed in Table 2. Decays of hadronic particles are described by EVTGEN [11] in which final-state radiation is generated using PHOTOS [12]. The interaction of generated particles with the detector and its response is implemented using the GEANT4 toolkit [13]. The simulation of the RICH detector in this framework is described in Ref. [14].
For the MaPMTs in the upgraded RICH, a readout efficiency of 90% is assumed. The charged particle tracks reconstructed from the tracking detectors and the hits from the RICH are input into a log-likelihood algorithm [15] for particle identification. The algorithm produces likelihood values for each track using every one of the particle hypotheses among pion, kaon, proton, muon and electron. These likelihoods are used for assigning the particle type for each track. In practice, the difference in the logarithm of the likelihood (DLL) between a pair of particle hypotheses is used for particle type assignment.

Resolutions and yields
The single photon Cherenkov angle resolutions and photon yields obtained from the simulations are listed in Table 3 for RICH 1 and in Table 4 for RICH 2. The various components of the resolutions are also listed in these tables. The chromatic contribution is from the variation of the refractive index of the radiator, the emission point contribution is mainly from the tilt of the mirrors and the pixel contribution is from the granularity of the MaPMT pixels. The point spread function (PSF) contribution in the HPD comes from the spread of the photoelectron directions from each point on the cathode. For the RICH 1 the improved optics contributes to an improved resolution compared to the current system. The QE specifications used for MaPMT result in a smaller chromatic error and higher photon yield compared to those of the HPD. The removal of the aerogel and the shift of the spherical mirror contribute to the improvement of the photon yield in the upgraded RICH 1 compared to that in the current system.

PID performance
The PID performance is defined in terms of the efficiency of identifying a true kaon as a kaon and the misidentification probability of a true pion to be a kaon or heavier particle. These quantities are plotted one versus the other for different values of the cut on the DLL between kaon and pion hypotheses obtained from the PID algorithm. In Fig. 3, the PID performance is plotted for the current geometry at the three luminosities, and compared to the upgrade geometry running at Lumi20. As expected, with the current geometry, there is a steady loss in the PID performance as the luminosity is increased from the present to the upgraded running conditions. However, this performance is mostly recovered when adopting the new RICH 1 geometry.

Photodetectors and their readout
Several tests have been performed on prototypes of the MaPMT and the CLARO chip that is to be connected to the MaPMT. One of these tests is the effect of the magnetic field on the MaPMT performance. The maximum magnetic field in the region of the photodetectors in RICH 1 is expected to be 3 mT and that in RICH 2 is expected to be 1.5 mT. The tests [16] done with MaPMTs inside a solenoid indicate that they would need to be shielded individually to avoid loss of gain and detection efficiency. The tests done with the prototype of the CLARO chip are described in Ref. [16]. These indicate that the chip has a recovery time of less than 25 ns as required. Detailed studies on the response from this chip are  Black : Lumi4 current geometry Blue : Lumi10 current geometry Red : Lumi20 current geometry Green : Lumi20 upgraded geometry Black : Lumi4 current geometry Blue : Lumi10 current geometry Red : Lumi20 current geometry Green : Lumi20 upgraded geometry Black : Lumi4 current geometry Blue : Lumi10 current geometry Red : Lumi20 current geometry Green : Lumi20 upgraded geometry Black : Lumi4 current geometry Blue : Lumi10 current geometry Red : Lumi20 current geometry Green : Lumi20 upgraded geometry Fig. 3. PID performance of the current geometrical layout for Lumi4, Lumi10 and Lumi20. The performance from the upgraded geometry at Lumi20 is superimposed. underway and the results from these will in turn be used in full simulations.
The data from the CLARO chip is transmitted to a reprogrammable FPGA [16] which in turn sends the data using fiber optical links to a readout unit connected to the LHCb data acquisition system. The FPGA boards connected to the low occupancy regions of the detector will perform zero suppression of the data and those connected to the high occupancy regions will send the data without zero suppression.
The MaPMT assembly on the RICH detectors are to be done in such a way that a set of 4 Â 4 MaPMTs will be plugged into a module which will provide magnetic shielding, heat dissipation into a cooling system, interface to the readout and detector controls. The walls of the module will be made of a high magnetic permeability material which can provide the shielding. Fig. 4 shows the schematic picture of a module assembly.

RICH mechanics design
Several components of the RICH 1 system will be modified as a result of the optics re-design of RICH 1. New spherical and flat mirrors will be built with the same technologies used in the current RICH [17]. The RICH 1 spherical mirrors are made of a light-weight carbon-fiber material. Since the location and orientation of all the mirrors in the RICH 1 are changed in the new design, all the mirror mounts will be replaced with new ones. The dimensions of the enclosure which contains the RICH 1 radiator gas and the quartz windows where the photons exit from this enclosure are all changed and as a result new ones will be built. For both RICH detectors, new photon detector assembly will be re-built.

Summary
The LHCb upgrade requires that for the RICH system, the HPDs be replaced by MaPMTs with the external readout so that it can acquire data at 40 MHz. The increase in the luminosity from 2019 onwards would have resulted in a degraded PID performance with the current RICH system. The usage of MaPMTs and the modification of the optics of RICH 1 would recover this. Simulations showed that the PID performance of the new RICH system is superior to that obtained from the current system.
Several tests on the prototypes of the photodetectors and their readout have been performed and new design of the mechanics of the photodetector assembly is underway. Most of the components of the RICH 1 will be rebuilt. A technical design report [16] is submitted and the engineering design is underway. It is expected that the new system will be installed in LHCb in 2018.