Status of the CMS electromagnetic calorimeter

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Status of the CMS electromagnetic calorimeter J. Fay


Introduction
We present here a general overview of the CMS electromagnetic calorimeter. For more details, please refer to the more specific presentations included in these proceedings [1][2][3][4][5][6][7] .

Objectives
The main goal of the CMS electromagnetic calorimeter (ECAL) is to measure precisely the energy and the position of e, γ and π°. The position and energy resolution benchmark comes from a low mass (115 < m H < 100 GeV/c 2 ) Higgs boson decay to two photons.
The energy resolution can be parameterized as follows • a is the stochastic term which includes contributions from the shower fluctuations, the photodetector statistics and the transverse shower leakage. The aim is to obtain a ~ 2.7% GeV 1/2 in the barrel. • b is the constant term which is dominant at high energies and includes the effect of mis-calibrations, non-uniformities of the detector, instabilities in the temperature and high voltage applied to the photodetectors and the rear and front shower leakage. Here the aim is b ~ 0.55% • c is the noise term due to the electronic noise and incremental dark current from the photodetectors under irradiation. It is also sensitive to pile-up in the detector. The aim is to obtain in the barrel c ~ 155 (210) MeV under low (high) luminosity conditions. To achieve such requirements, the CMS collaboration has chosen to build a homogeneous electromagnetic calorimeter made of lead tungstate (PbWO 4 ) crystals. This crystal is unique to work in the very demanding LHC environment where: • the LHC bunch crossing rate is 40 MHz and lead tungstate scintillation is fast (80% in 25ns). • there is a severe radiation environment and, as the result of R&D, lead tungstate is sufficiently radiation hard. • compactness (CMS manages to put its entire calorimetry inside its 4T solenoid) and granularity are required, and lead tungstate has a radiation length X 0 = 0.89 cm and Moliere radius R M = 2.2 cm.
Unfortunately there are also some drawbacks: • lead tungstate presents a relatively low light yield hence the photodetector must have some gain. • its light yield shows a strong dependence on temperature and a very precise cooling system is required.
The photodetectors must be able to work in a magnetic field B = 4T, amplify the signal (required by the low light yield from PbWO 4 ) and be radiation hard.

Production and data quality control of the crystals
The ECAL barrel is made of 2 x 18 so called Super-Modules (SM), plus 1 spare SM not intended to be put in the detector, each containing 85 (η) x 20 (φ) crystals. The crystals have a tapered shape, which varies slightly with η (17 types). The mean dimensions are 2.2 x 2.2 x 23 cm 3 .
The crystals are manufactured by the Bogoroditsk Techno-Chemical Plant (BTCP) in Russia. Up to now, 24600 crystals have been delivered out of a total required number of 61200 (not including the 1700 crystals for the 37 th SM).

The Avalanche Photodiodes (APD)
The CMS APDs are manufactured by Hamamatsu Photonics, Japan. Their main properties are summarized in Table 1. They are received at PSI and undergo a set of tests in order to check their radiation hardness: they are irradiated with a Co-60 γ source to 5 kGy, at a dose rate of 2.5 kGy/hour. After 24 hours of relaxation time, breakdown voltage Vb and dark current I dark at a gain of 50 are measured. They are sent to CERN, where the noise is measured and they are annealed in an oven for 4 weeks at T=80°C. Then Vb and I dark are re-measured to check for any changes.
Potentially unreliable APDs are rejected if they show a significant shift of Vb after irradiation, high I dark or noise ( Figure 2) The rejection rate is approximately 5%, the main reason being high noise after irradiation (3-3.5%), the other criteria accounting for less then 0.6% each.
The APD production is nearly finished: • All ordered APDs (130 000) have been received • ~ 20 000 APDs are still on the screening line • 103 000 have already been sent to Lyon for capsule production • The last APDs are foreseen to be sent by July 2004 In Lyon, a pair of APDs is assembled in parallel and a kapton cable is soldered to form a capsule. On one of every 10, a temperature sensor with its own kapton cable is added. The following quality checks are performed: gain curve, I dark , breakdown voltage noise and pulse shape are measured and the results are recorded in the CRISTAL database. The rejection rate (mainly due to high noise) is less then 1 %.
The capsules for 24 Super Modules have already been constructed and tested. It is foreseen to test the last capsule by June 2005.

"Bare" Super Module assembly
In the 2 regional centers (CERN and Rome), the capsules are glued on the crystals at a rate of 50 per day and Modules are assembled (400 or 500 crystals). Four Modules are then grouped together at CERN to form the bare Super Module. Up to now, 50 Modules (35%) and 10 Super Modules (28%) have been assembled.

The end caps
The crystals of the end caps are all identical, slightly shorter and wider than those of the barrel (3 x 3 x 22 cm 3) . Arrays of 5 x 5 crystals are grouped in identical modules called 'Super Crystals' which in turn make four identical 'Dees' with 3662 crystals each.
Due to the high radiation levels in the End cap region, APDs cannot be used. We use instead Vacuum Photo Triodes (VPT), which are in fact singlestage photomultipliers with a gain close to 10. They are manufactured by Research Institute Electron, St Petersburg, Russia. Out of total 15500, 7600 VPTs have been delivered.

The Preshower
In front of the end caps, a fine granularity preshower detector improves γ − π° rejection. It is a sampling calorimeter made of two lead absorbers (2 and 1 X 0 respectively), each followed by a layer of silicon microstrip detectors. The two layers of detectors have their strips orthogonal to each other. More than 2100 (out of a total of 5000) silicon sensors have been received from the four production centers (Russia, India, Taiwan and Greece)

The light monitoring
The optical transmission of PbWO 4 crystals is reduced by gamma irradiation doses through the formation of color centers. Saturation is reached after a small dose and partial damage recovery can be observed in a few hours. This loss in extracted light, which depends on the dose rate (~ 3 -5 % for 0.15 Gy/h), must be monitored over the full ECAL with a light monitoring system.
To follow damage and recovery during LHC cycles, we will use two laser systems with four wavelengths: blue/green (440/495 nm) where the transmission damage is significant and red (700/800 nm) where there is very little damage in order to separate out other effects in the readout system. The light is injected to each crystal using quartz fibres. Laser fluctuations are measured by very stable (0.1 %) reference silicon PN diodes from Hamamatsu.

The cooling system
To reach the targeted energy resolution, cooling is a key issue since both the light yield (LY) of the crystals and the APD gain M depend on the temperature (dLY/dT = -1.9 % / °C and 1/M dM/dT = -2.4 % / °C). This implies that the temperature of the calorimeter in the APD and crystal region must be maintained at 18 ± 0.05 °C while each electronics channel dissipates 2.5 W. To be able to achieve the required performance, heat must be removed directly from the chips and the heat flow to the APD and the crystal minimized.
The cooling circuit has two functions: to regulate the temperature in the crystal region with a precision ∆T of typically a few 1/100 °C by flowing water with a very precise temperature and to remove the dissipated heat by creating a good thermal coupling between the electronics and the water. This is accomplished by maintaining the electronics components in contact through gap filler and gap pad with stainless steel pipes carrying the cooling water embedded in aluminium cooling bars. Figure 3  In 2003, the preamplifier and ADC combination was redesigned.
The previous design was based on analogue multiplexing with the FPPA (Floating Point Pre-Amplifier) which is a low noise preamplifier with 4 intermediate gain stages (33, 9, 5, 1) and analogue multiplexing of the highest non saturating signal. It used a commercially available ADC: AD9042 from Analog Devices. Slow-control measurements (temperature, leakage current) were available directly on-chip.
The new design uses digital multiplexing and consists of a MGPA (MultiGain Pre-Amplifier) with 3 gain stages (12,6,1) and 3 ADCs. The highest non-saturating signal is chosen after digitization. The slow control measurements are performed by a dedicated service chip (DCU). Both chips are produced in IBM 0.25 µm technology that permits lower cost and a faster turnaround.
This solution gives better performance but necessitates new chips to be developed.

The new ASICs
The first version of the MGPA has been tested in beam in summer 2003. No major problem was found.
A new ADC has been developed: ADC41240 ( (quad-12 bit ADC with onboard selection logic). Version 1 of these chips has been also used in test beams in 2003. A minor problem on Vref layout routings has been identified and version 2 chips were received in March. The tests performed on them show that this problem is cured.
The FENIX chips (situated on the FE board) have two main functions: to generate the Trigger Primitives (Trigger Tower transverse energy, bunch crossing assignment) and to read out the data when a positive trigger decision is received. There are 7 chips on each FE board, all of the same type but running in 3 different operation modes (Strip mode to receive the data from the VFEs, TCP mode to generate the trigger primitives and DAQ to send the data).
Finally a new interface chip between the ADC and the FENIX (LVDS to LVCMOS level converter) had to be developed.
A number of chips, equivalent to that needed to equip at least 3 SMs, has been fabricated and will be ready by mid 2004. If they work as expected, all needed chips for the entire calorimeter will be sent to fabrication by the end of 2004.

VFE boards
All of the VFE boards will undergo three consecutive tests: • Pre-production in the factory in order to validate the production and obtain a very high production yield (>99%). • Burn-in to get rid of "infant mortality" of the weak components. • Calibration, using a calibrated charge generator and a digital readout system.
The results from the measurements are stored in the construction Data Base.

Precalibration
Due to the delay in fabrication of the Super Modules and the fact that there will be no beam available at CERN in 2005, we will not be able to precalibrate all the SM in beam. Nevertheless, we have shown we can obtain a first intercalibration from the data stored in the construction Data Base (the principal information is the light yield measured in the Regional Centers, corrected for capsule and electronics gains). This procedure has been checked ( Figure 4) by comparing these results (Lab LY) with test beam ones (Test Beam LY):

Figure 4 Comparison between the results measured during the detector construction and test beam
This comparison shows that intercalibration coefficients can be obtained from the construction data with a precision of ~4 %. Work is still going on in order to improve these results. .
At the start of LHC operation, a fast intercalibration based on φ symmetry in minimum bias events will give a precision close to 2% in a few hours. With the electron pair invariant mass reconstruction of events coming from Z → e + e -, , it is possible to provide the intercalibration of calorimeter regions and to set the absolute energy scale. With the help of the tracker measurements, isolated electrons from W eν will permit the target resolution of 0.5% to be reached.

Energy resolution
The data show low frequency noise, the result of which is a pulse baseline that varies from event to event. This effect is present in all channels with a uniform (15 -20 %) spatial correlation. The baseline variation can be subtracted, eventby-event, by using an appropriate set of weights to compute the signal amplitude. After correction, the noise for the sum of nine crystals is 129 MeV, corresponding to a noise per channel of about 43 MeV. Figure 5 shows the energy resolution for the sum of nine crystals. These results are obtained with events selected when the incoming electron is located in a 4x4 mm 2 window at the center of the crystal. They confirm that the design performance can be reached.

Conclusions
The CMS electromagnetic calorimeter collaboration has already achieved encouraging results on crystal and photodetector quality control, electronics, monitoring and cooling systems, 2003 test beam results with new electronics. The detector construction now enters a new challenging phase: • Production of the crystals • Production and integration of the electronics • Precalibration in the beam of as many Super Modules as possible