Overview of PbWO4 calorimeter in CMS

Abstract The status is given of the lead tungstate (PbWO 4 ) electromagnetic crystal calorimeter for the CMS detector at LHC. This includes the progress on the engineering design of the calorimeter and its readout system, latest results on PbWO 4 crystals and photodetector development, with particular attention to radiation hardness issues, and results from test measurements in particle beams.


INTRODUCTION
The CMS collaboration is developping, for its detector, a high-precision electromagnetic calorimeter (ECAL). Its conception is based on the need to optimize the discovery potential of a light Higgs, with a mass m H < 150 GeV, decaying into two photons. Since the natural width of the Higgs there is expected to be very narrow, and the background from QCD jets huge, the detector resolution is the crucial factor for this kind of signature, which is the only promising one to explore the low mass range in the search for a Standard Model Higgs particle.

PbWO 4 calorimeter design
Beyond the benefit of a homogeneous calorimeter, Lead Tungstate has been chosen for several reasons. It has a fast scintillation emission, with a light decay time which matches the LHC bunch crossing time of 25 ns. It has small radiation length (X 0 = 0 : 89 cm) and Molière radius (R M = 2 : 2 cm), so that a very compact and highly granular calorimeter can be built [1]. However, its light output has a temperature dependence of 2%/ 0 C, which puts some constraints on the detector construction.
The ECAL barrel electromagnetic calorimeter is designed to have an inner radius of 1.29 m and a granularity ' = 0 : 01750:0175 using 23 cm (26 X 0 ) long crystals with a front face of 2222 mm 2 for a total crystal volume of 8:14 m 3 . Avalanche Photodiodes shall be used as photodetectors. The mechanical design has to take into account several constraints, given by the fragility of the crystals, the need for temperature regulation of crystals and photodetectors, the need for hermeticity and its large dimensions. The design is based on modularity and on the use of high-strength, low-Z materials, where alveolar submodules of 2 5 fiberglass cells containing individual crystals are the smallest subunits.
These submodules will carry a reflective inner coating, yielding intercrystal gaps of < 500 m. Fifty respectively 40 such submodules (depending on ) are foreseen to be assembled into baskets made out of Carbon-fibre composite, which will take the cantilever of the submodules. Supermodules consisting of 4 baskets held by a U-shaped spine at the outer ECAL radius shall be used for calibration in the test beam and installation. Hermeticity has been optimized by tilting the crystal axis by 3 0 with respect to the direction pointing towards the centre of the detector. The barrel ECAL covers the pseudorapidity region j j< 1:48.
The ECAL end-cap calorimeter is designed along a similar scheme, however using identical "supercrystals" containing 6 6 identical crystals, 22 cm long (owing to the presence of a 3 X 0 thick preshower detector in front) and with 24:7 24:7 mm 2 front face dimensions. The front face of the calorimeter shall be at 3170 mm distance from the detector centre along the beam axis, and the crystals will be oriented towards a point located 1300 mm beyond the detector centre for hermeticity. An alveloar fiber- glass support structure is foreseen here as well, with the supermodules cantilevered off the front side of "Dee"-shaped, 5 cm thick Aluminium support plates. All readout electronics, from the preamplifier to the fiberoptics readout, is to be sited on the rear of the support plate. The endcap ECAL covers the pseudorapidity region 1:48 j j< 3:0.  the calorimeter is almost entirely due to neutrons, and it is expected to reach 2 10 13 n =cm 2 in the ECAL barrel for an integrated luminosity of 5 10 5 pb 1 (equivalent to 10 years of running at LHC).
However, no damage by neutron irradiation has been seen sofar [2], while a sample dependent effect is observed with photon irradiation. It has been shown that the scintillation mechanisms are not damaged, nor is the scintillation emission spectrum changed [3]. The degradation of light output is namely due to radiation-induced absorption, i.e. the formation of colour centres, which reduce the crystal transparency. A believed cause for the formation of colour centres is the presence of Oxygen vacancies. It will be thus possible to monitor the loss in transmission due to irradiation using a light injection system in the calorimeter, and to apply corrections for it.
To compensate for such defects, two different approches are pursued by the crystal growers. One, for chinese crystal produced via Bridgman method, is a systematically optimized post-growth annealing process. In a first round of developments, such a treatment at temperatures close to the melting point of PbWO 4 in different gaseous atmospheres has been applied to crystals cut in pairs from the same "father" crystal; oxygen-annealed crystals appear considerably more radiation-hard than air-annealed crystals, as can be seen in the comparison between crystals 115-1 and 115-2 of Fig. 2(a), both cut from the same ingot. Further, the best oxygen-annealing cycle has been determined from the highest radiation resistant sample out of two obtained from the same ingot. The reproducibility of such crystals was demonstrated through the delivery of several oxygen-annealed 50 mm long samples, whose behavior under irradiation also appears in Fig. 2(a). Finally, some samples were tested using a radiation dose corresponding to the levels expected at large in the ECAL end-caps, or, respectively, to the integrated dose after one year of high luminosity running at LHC in the barrel. From the example shown in Fig. 2(b) it can be seen that excellent radiation resistance can be achieved even for such high levels of radiation. For russian crystals grown via the Czochralski method, the crystal quality was optimized using specific dopants, either pentavalent on the tungsten site (using niobium) or trivalent on the lead side (using lanthanum, yttrium or lutetium). A successful compensation of defects was obtained, in that these crystals show an improved transparency, as well as a considerable improvement of radiation hardness, whose progress can be appreciated in Fig. 3.

Photodetectors
For the detection of the PbWO 4 scintillation light in the ECAL barrel, Silicon Avalanche Photodiodes (APDs) have been chosen, since they are insensitive to magnetic fields, allowing thus operation in the 4 Tesla CMS solenoidal field, and since they also have some internal gain, as needed with the modest light yield produced by the crystals. Additionally, their quantum efficiency matches well the PbWO 4 emission spectrum. However, several APD parameters had to be optimized, and this has been the object of extensive R&D efforts pursued by manufacturers like EG&G and Hamamatsu, in collaboration with several institutions (U. of Minnesota, PSI, ETH Zürich, RWTH Aachen, RAL, Saclay, U. of Rome I and INFN). The crucial parameters there are the APD capacitance (C), the so-called Excess Noise Factor (F) which quantifies the fluctuations in the multiplication process, the gain (M) stability with respect to temperature and voltage, and the response to ionizing radiation hitting the diode (expressed as an effective thickness d ef f . A summary of the achieved performance can be found in Table 1. Concerning ] New prototypes (M=50) T' 20 0 C APD passi-C(pF) requests Si 3 N 4 < 100 < 500 few nA 2 big small < 2 < 2 > 0:5 Table 1: Characteristics of Avalanche Photodiodes the resistance to radiation, it is now established that the APD behaviour corresponds to that of Silicon. The bulk current (I B ) is observed to increase linearly with neutron flux (Fig. 4(a)). For a flux of e.g. After a long-term exposure to radiation, the dominant contribution from APDs to the electronic noise is given by [5]:   Table 1 it can be seen that characteristics for both manufacturers are not far from the optimum requested for the CMS ECAL. The APD surface area shall be increased to 0:5 c m 2 . Likely, two squared APDs per crystal will be used. This shall allow to keep the noise increase with irradiation at an acceptable level for an optimum ECAL performance. Operation at 14 o C 18 o C could also help reducing the dark current and is being considered.
Vacuum phototriodes are the best candidates as end-cap photodetectors, since they work well at small angles to the magnetic field, and offer a sufficient resistance to the high radiation environment expected at large values.

Calibration ad monitoring
(a) For a non-radiation hard crystal, LASER monitoring response and electron shower response versus radiation dose (left), and correlation between LASER monitoring response and electron shower response (right).  A precalibration of all the crystals in a high-energy electron beam at two energies is foreseen. Each supermodule (or Dee) will be already equipped with a monitoring system using injected LASER light. Since it has been established that radiation only affects the crystal transparency, light injection shall allow to monitor changes in calorimeter response. The correlation between response to beam particles and monitoring system will be established at that level. In situ, an absolute calibration using physics events (mainly Z ! ee) will be performed, depending on luminosity, every 1 to 5 weeks to achieve the ultimate detector performance ( 0:5 constant term).
Studies have been performed in beam tests, using a LASER at two different wavelengths (one emitting red the other green light). A particularly radiation-soft crystal has been irradiated, and its change in response to electromagnetic showers has been compared to the LASER monitoring response. One observes on the left part of Fig. 5(a) the decrease of the signal with radiation dose, as well as the decrease of signal due to the LASER light. On the right side of Fig. 5(a) the correlation is plotted between the cystal response to electromagnetic showers and its response to the monitoring light. Particularly in the range of high values (0.95 to 1.) expected during operation of the calorimeter the correlation is nicely linear and can thus be exploited for an accurate tracking of the effect of radiation on the crystals.
6 Performance studies on crystal matrices in 1997

Beam tests on a 7 7 crystals matrix
Monitoring studies as well as studies of intrinsic properties were performed in 1997 on a 77 crystal matrix, using one APD per crystal, and conventional read-out electronics, with preamplifiers + line drivers bringing the signals along 80 m cables into Lecroy fastbus ADCs. The crystal dimensions corresponded to the projective shape at = 0 of the barrel, including the 3 o tilt with respect to the pointing direction.
Temperature and voltage were stabilized and the crystal calibration was monitored with LASER light injected via optical fibres from the front face of the crystals. The tests made use of the SPS H4 beam line of the CERN North area. Besides the mentioned montoring studies, the crystal energy resolution was studied up to an unprecedented energy of 280 GeV. The corresponding energy deposition spectrum in a sum of 3 3 crystals is shown in Fig. 5(b). No high-energy tails are observed, and thus no sign of rear leakage, which would cause direct signals in the APDs. Taking into account the fluctuations in beam momentum due to synchrotron radiation, 0:2 < =p < 0:32, one obtains a energy resolution compatible with the prediction, =E= 0 : 38, from a fit to the data at lower beam energies.  The distribution of energy resolutions at 120 GeV for 21 different sums of 33 crystals yields an average =E= 0 : 53 ( Fig. 6(a)). A fit of the energy resolution values as a function of beam energy yields for the same 21 sums of 3 3 crystal arrays a distribution of constant terms centered around 0:34 (Fig. 6(c)) and of stochastic terms with an average of 4:2 ( Fig. 6(b)). This last number will be reduced further through the use of 2 APDs per crystal.

Prototype '97 (6 6 crystals)
An ECAL pre-prototype has been built, consisting of 36 crystals ( = 0 geometry) in an Al+fiberglass alveolar structure with nominal dimensions, arranged in 2 6 cooled subunits. Temperature and voltage are stabilized, and the crystal calibration monitored with light injection via optical fibres. The aims are to test the entire electromechanical arrangement and to evaluate the front-end performance. The prototype includes capsules with one APD per crystal and preamplifiers with linear, multiple gains. A circuit called Floating Point Unit (FPU) selects the appropriate gain range for each 25 ns sample and multiplexes them into a single ADC input. A commercial 12-bit 40 MHz voltage sampling ADC is used (Analog Device AD9042). The resulting digital word consists of a 12-bit mantissa and a two-bit code indicating which gain stage was used. A digital, optical link follows towards the higher level read-out and data-acquisition.
A full coverage of the dynamic range has already been demonstrated, and an electronic noise has been measured of 4500 e for a 120 pF APD capacitance. The ADC linearity was measured to be better than 0.1%, and it has proven to be radiation hard beyond 3:3 10 13 n=cm 2 + 2 : 2 Mrad. This system is now being evaluated in all its aspects.

Conclusions
The CMS ECAL has reached its final R & D phase, and its target performance has been demonstrated in beam tests. It has been shown that the LASER monitoring system can track crystal response changes of 5% and beyond. A complete electronics chain with the required performance has been produced, and a system test of 36 channels is being performed. A sufficient radiation resistance has been reached for all detector elements (crystals, APDs and all the readout elements). In 1998, preproduction runs for crystals will be performed. The production characteristics for the photodetectors will soon be defined. The Technical Design Report is being written, and is due end of 1997.