Advanced Technology and Particle Physics

125

depending on module orientation-'. Deformations due to thermal stresses are a concern. The option to clamp firmly only one of the modules ends was chosen for both SVXII and ISL. Let 6a be the angle to which wafers strips are not parrallel to the CDFII +z axis. In SVXII and ISL* after ladder alignment and assembly (656 ladders, with lengths up to 27.82 cm), wafer strips deflect from an ideal line by da < 10.0 /zrad. Module mounting onto bulkheads and ledges or wafers gluing onto CF structure introduced further deviations. Measured numbers for LOO and SVXII once completed are Sam < 26 /xrad and Sasvxu < 34 /xrad. In addition, the SVXII barrel center (r, ip) offsets were (< 40 /xm , < 2.1 mrad) and slope differences contained in less then 70 ^irad. For ISL we have only an indirect evaluation. Assuming that modules clamped onto SF ledges inherit the direction defined by their mating-ledge slots then Sotisi < 220 /irad. The average axis used for the three sub-detectors alignment, re-mesured once SVX was completed, have 6asvx < 150 ^rad (SF length is 197.16 cm); in this reference frame, the maximum recorded discrepancy from a nominal value was 300 /xm. All these quantities are summarized in Table 3. 3.3

Collision data

Preliminary results return unbiased residual distributions with a < 13 /urn (for all SVXII layers in each barrel) and barrel-to-barrel differences in slopes within < 200 /irad. We are finding efficient track patterns with attached ISL hits and we have indication that SVXII-ISL layers are parallel within less then 178/xrad. 4

Conclusions

A description of the SVX, the silicon tracking system of CDFII, and some details related to its internal alignment were reported. SVX will become "tracker" only once the necessary software that efficently performs SVX-COT track linking is optimized. The material budget in the tracking volume is worse than expected but the good quality of the overall SVX assembly is already allowing fast progress of the alignment corrections study. We also welcome the recent news proving that the major concern, the ISL cooling, is on its way to being fixed. •* Ledge n = 0 is at ip = 0 in layer 6C,7F/B and tp = - 7 . 5 ° in 6 F / B ; n increase anti-clockwise. *As pointed out, we cannot really speak of ladders for LOO. However the following limit (5Q;OO is valid for three longitudinally subsequent LOO wafers.

126

Survey before assembly Ladders internal alignment (SVXII,ISL) da < 10 /urad < 170 /xm ISL module ~ 50 cm long C.C.O. [odd] [even] < 145 /xm < 65 /mi SVXII module ~ 30 cm long C.C. < 55 /jm Stability (SF, ST under full load) Survey after assembly
References 1. M.Bishai for the CDF coll., The silicon data acquisition system and front-end electronics for CDF Run II, These Proceedings. 2. S.D'Auria for the CDF coll., Commissioning and Operation of the CDF SVX detector, These Proceedings. 3. I.Fiori for the CDF coll., CDF online Silicon Vertex Tracker, These Proceedings. 4. A.Affolder et al., Status report of the ISL detector at CDFII, Proceedings 5th Int. Conf. Large Scale Applications and Radiation Hardness of Semiconductor Detectors, Firenze, Italy, 2001. 5. A.Basti et al., The Intermediate Silicon Layers Space Frame, Proceedings 4th Int. Conf. Large Scale Applications and Radiation Hardness of Semiconductor Detectors, Firenze, Italy, 1999. 6. D.Stuart, CDF note 5268, 2000. 7. S.Blusk et al., Characterization of a SVXII Bulkhead, CDF note 4473, 1998. 8. A.Basti et al., The Intermediate Silicon Layers Cooling System, CDF note 5034, 1999. 9. D.Glenzinski (CDF coll.) private communication. 10. CDFII coll., The CDFII detector Technical Design Report, Fermilab-Pub. 390-E, 1996. 11. R.McNulty, T.Shears, A.Skiba A procedure for the software alignment of the CDF Silicon System, CDF note 5700, 2001. 12. E.Sanders, D.Saltzberg Proposal for a Rasnik Relative Alignment Monitor for the CDFII SVX-ISL-COT, CDF internal note 4510, February 1998.

T H E C D F ONLINE SILICON V E R T E X T R A C K E R

A. BARDI, A. BELLONI, R. CAROSI, G. CHLACHIDZE, M. DELL'ORSO, S. DONATI, S. GALEOTTI, P. GIANNETTI, V. GLAGOLEV, E. MESCHI, F. MORSANI, D. PASSUELLO, G. PUNZI, L. RISTORI, A. SEMENOV, F. SPINELLA INFN, University and Scuola Normale Superiore of Pisa, 1-56100 Pisa, Italy A. BARCHIESI, M. RESCIGNO, S. SARKAR, L. ZANELLO INFN sezione di Roma I and University La Sapienza, 1-00173 Roma, Italy M. BARI, S. BELFORTE, A.M. ZANETTI INFN sezione di Trieste 1-34012 Trieste, Italy I. FIORI INFN and University of Padova 1-35031 Padova, Italy (corresponding author. E-mail address: [email protected]) B. ASHMANSKAS, M. BAUMGART, J. BERRYHILL, M. BOGDAN, R. CULBERTSON, H. FRISCH, T. NAKAYA, H. SANDERS, M. SHOCHET, U. YANG Enrico Fermi Institute, University of Chicago, Chicago, IL 60637, USA A. CERRI Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA Y. LIU, L. MONETA, T. SPEER, X. WU University of Geneve, GH-1211 Geneve 4, Switzerland The Silicon Vertex Tracker (SVT) is the new trigger processor which reconstructs 2-D tracks with high speed and accuracy at the level 2 trigger of the CDFII experiment. SVT allows tagging events with secondary vertices and therefore enhances the CDFII B-physics capability. SVT has been fully assembled and operational since the beginning of Tevatron Runll in April 2001. In this paper we briefly review the SVT design and physics motivation and then describe its performance during the early phase of CDF Runll.

1

Introduction

The Collider Detector at Fermilab (CDF) is a general purpose detector for the study of high energy pp interactions produced at the Tevatron Collider. CDF completed its first period of data taking (RunI) in 1996. The Teva-

127

128

tron has recently completed major upgrades to achieve higher energy (from 800 to 990 GeV per beam) and instantaneous luminosity (from 2 x 10 31 to 1032 cm~2s~1). CDF has been upgraded to cope with the reduced bunch spacing condition (from 3.5 /xs to 396 ns and then further down to 132 ns during the phase "b" of Runll), with the higher number of interactions per bunch crossing, and to extend its physics reach. 1 The detector upgrades relevant for this paper are the tracker and the trigger. The new silicon vertex detector (SVXII and LayerOO)2'3 covers the region from 1.5 to 10 cm from the beamline and about 2 units in pseudorapidity. SVXII is made of 5 layers of double-sided (r — (j> and r — z readout) microstrip sensors arranged in a 12-fold azimuthal geometry and segmented in 6 longitudinal barrels (3 mechanical units, each read out at both ends) along the beam line (CDF z-axis). LayerOO is made of one layer of radiation hard silicon microstrips (r — <j) readout) placed just outside the beam pipe (r = 1.5 cm). The trigger is completely new and consists of three levels. Its challenging task is to reduce the 5 MHz level 1 input rate down to a level 3 output rate of 50 Hz for data storage and analisys. Levels 1 and 2 are completely hardware implemented while level 3 is software implemented and runs on a PC farm. The most relevant level 1 device is the eXtremely Fast Track finder processor (XFT) 4 which reconstructs 2-D tracks (in the r — plane, transverse to the beam line) in the central drift chamber (COT). The SVT is part of the level 2 trigger. It receives the list of COT tracks (Pt and (f> coordinates) from the XFT processor and the digitized pulse heights on the silicon layers (~ 105 channels) and reconstructs tracks with offline-like quality. The expected resolution for SVT tracks is: 6~l mrad, 6Pt ~ 0.003 • Pt2 GeV/c, Sd ~ 35 /xm (d is the track impact parameter i.e. the distance of closest approach of the particle trajectory helix to the z-axis of the CDF reference system). By providing a precision measurement of the impact parameter SVT allows triggering on events containing long lived particles. B-hadrons in particular have a decay length of the order of 500 /xm and tracks which come out of the B decay vertices have an impact parameter on average greater than 100 \xm. This novel feature, available for the first time at a hadron collider experiment, greatly enriches the CDF B physics program. In particular it provides access to rare purely hadronic B decays like B° -» TTTV and Bs -> Dsn -> hadrons, which are extremely interesting respectively for CP violation and Bs mixing. 2

T h e S V T working principle

The SVT has a very short time to perform its task (on average ~ 20 /xs per event) to keep up with the 50 KHz level 1 accept rate. In order to speed up

129

operations SVT has a parallelized design: it is made of 12 identical azimuthal slices ("wedges") which work in parallel. Also tracks are reconstructed in 2-D only (stereo info from SVXII is dropped) and only with Pt above 2 GeV/c. Track reconstruction is performed in two steps: in the pattern recognition step candidate tracks ("roads") are searched among a list of precalculated low resolution patterns; in the subsequent track fitting step the full resolution fit of the hit coordinates found within the roads is performed using a linearized algorithm. The pattern recognition step is performed in a completely parallel way by the Associative Memory system which uses full custom VLSI chips (AMchips).5 The AM system compares the SVT input data with the set of precalculated patterns. A pattern is defined as a combination of five bins ("SuperStrips"): four SuperStrips correspond to the position coordinates of the particle trajectory on four silicon layers which can be chosen among the five SVXII layers and LayerOO, the fifth SuperStrip corresponds to the azimuthal angle of the particle trajectory at a distance r = 12 cm from the beam line (this distance was found to maximize average pattern coverage). The output of the AM system is the list of patterns ("roads") for which at least one hit has been found on each SuperStrip. Each SVT wedge uses ~ 32K patterns which cover more than 95% of the phase space for Pt > 2 GeV/c. Simulation studies have shown that SVT performance is optimized (in terms of processing time and final resolution) by choosing a SuperStrip size of about 250 fj,m on the silicon layers and 5° for the <j> angle measured by XFT. The track fitting method is based on linear approximations and principal component analysis. 6 The analytical relationship between the track parameters and the 6 measured hit coordinates (hit positions on 4 silicon layers, curvature, and azimuthal angle of XFT track) can be expressed in terms of 6 equations: Pj=Pj(x)

= Fj-x

+ Qj

(1)

where, x is the vector of hit coordinates and P = (Pi,....,P 6 ) = (d,c,<j),Xi,X2,X3) i s a vector consisting of: the impact parameter (d), the curvature (c), the azimuthal angle ((/>) at the point of closest approach to the 2-axis, and three independent constraints (xi, X2 and X3) which all real tracks must satisfy within detector resolution effects. Fj and Qj are constants which depend only on the detector geometry and on the magnetic field. Once pattern recognition has been performed, each track candidate is confined within a road and hit coordinates can be referred to the SuperStrips edge (xo); equation 1 becomes: P0j + SPj = Fj • (x 0 + (5x) + Qj

(2)

where, PQJ — Fj • x 0 + Qj is a constant which depends on the road, while 6Pj

130

is the correction which depends on the precise hit positions inside the road. The PQJ coefficients which correspond to the track lying exactly on the road edge, are calculated in advance and stored in RAMs. Therefore the track fitting task reduces to a fast computation of simple scalar products (done by FPGA chips): SPj = Fj • Sx.

(3)

The output list of high precision tracks is sent to the global level 2 processor for the final trigger decision. SVT is made by over one hundred VME boards housed in 8 crates. 7 The installation has been completed and the system has been fully operational since the beginning of 2001. In the following section we report on the performance achieved by the system in the early phase of Runll. 3

S V T performance

The first evidence that SVT finds good tracks is the plot of the correlation between the impact parameter d and the azimuthal angle <j> in a sample of candidate tracks (figure 1, left). If the position of the interaction vertex in the transverse plane is (x0,yo), displaced from the nominal one (0,0), the relationship between d and <j> for primary tracks is: d = —xosin{<j>) + yocos((p).

(4)

A fit of the d — <j> scatter plot measures the (xo, yo) coordinates with an accuracy of few microns. A beam displacement of few millimetres is the present typical running condition for CDF. This is a potential problem because unphysical large impact parameters erroneously affect the level 2 trigger decision. In practice the problem has been solved in the following way: a process running on one of the SVT VME crate controllers performs the fit of the d — cj> correlation of the SVT tracks independently for tracks in each one of the six SVXII ^-barrels. The beam offset is subtracted online: d! = d + xosin()

(5)

and physical impact parameters (measured with respect to the actual beam position) are actually output by the SVT. Figure 1 illustrates the effect of beam offset subtraction (left) and the online d' distribution (right). The gaussian shape and the width of the d' distribution originate from the convolution of the actual transverse beam profile with the impact parameter (i.p.) resolution. A gaussian fit gives a ~ 69 \xm.

131

Figure 1. Left: d — <j> correlation for SVT tracks before {top) and after (bottom) correcting for the beam oflset. The measured beam position in the transverse plane is XQ = 0.0995 cm and s?o = -0.3895 an with a precision of ~ 3 (im. The regions without points around 0 = 2.2 and $ = 4.2 rad are due to missing SVXII wedges. The scale on the x axes is radians, the scale on the y axes is cm. Right: Distribution of the online impact parameter (in cm) corrected for the beam offset.

A big effort has been put in understanding the various factors contributing to the impact parameter resolution with the aim of possibly improving it. In fact, as the resolution degrades the trigger rate increases and the background contamination in the data samples selected by impact parameter cuts is larger. The potentially most relevant contribution to the impact parameter resolution arises from the z-tilt of the beam with respect to the detector. The effect of a ^-misalignment (m x = SxjSz, mv = Sy/6z) shows up as a residual 4> modulation in the impact parameter after correcting it for the beam offset: d' — — mxzosin()

(6)

where ZQ is the z-coordinate of the point of closest approach of the particle trajectory to the z-axis, Since SVT does not measure ZQ, a beam tilt along z results in a irreducible widening of the impact parameter distribution. In order to make this spread small compared to the natural beam width, SVT requires the detectors and the beam line to be all parallel within 100 /irad. Assembly of the SVXII barrels met (even exceeded) this specification,3 the z-alignment of the beam orbit is unfortunately more challenging. During the April-October 2001 data taking the beam slope was found to be significantly large: m , ~ 600 /irad, my ~ 150 fir&d, well beyond the SVT specification.

132

The effect of the z-misalignment on the impact parameter distribution was estimated using data of a special run taken with an approximately null beam tilt. A gaussian fit to the online impact parameter distribution gives a ~ 59 \im. In addition to the beam tilt, there are two more major contributions to the impact parameter resolution. One is the relative misalignment of the SVXII wedges. This can be easily corrected by performing the beam offset fit and subtraction independently in each wedge. The size of this correction to i.p. resolution is approximately 6 pm. The second major contribution to i.p. resolution is a consequence of the linear approximation used in the SVT track fitting method, which assumes a first order power expansion centred on the nominal beam position. Since during the April-October 2001 data taking the beam was very far from its nominal position (~ 4mm away) the effect of non linearity was significantly large: it degrades i.p. resolution by approximately 5 \im. This effect can be corrected in two steps: first the fit constants in equation 2 are recalculated centred on the measured beam position; second, the beam position fit is done on each wedge separately using a linear d—<j> relationship. With these corrections applied the impact parameter distribution was found to have a gaussian shape with a sigma of 48 fim in a run taken with the beam aligned in z.a All these corrections can be implemented in the SVT and applied online. But the best SVT performance relies on the accelerator capability to provide an aligned beam. Using a sample of events in which at least two good (x2 < 10) SVT tracks were found we have been able to calculate the true beam transverse size (O\B). The result of this study (described in detail in ref.6) for the run with negligible beam tilt is as = 33 ± 1 \im. Deconvolving this beam width from the above 48 firn we obtain ad ~ 35 /xm for the SVT i.p. resolution, in agreement with the design value. In October 2001 the first trigger tests using SVT have been done. A level 2 trigger was implemented requiring at least 2 SVT tracks with \ 2 < 25, Pt > 2 GeV/c, \d\ < 50 fim and a level 1 prerequisite of at least 2 XFT tracks. Figure 2 (left) shows the distribution of the second largest impact parameter in the event Using this data (corresponding to ~ 15 n& _1 of integrated luminosity) a small signal of D° —> Kir was reconstructed (figure 2, right).6 "Additional, lower order, corrections can be applied: like correcting for a residual non linearity in d and <j>, and for the misalignment of silicon layers within a wedge. However their online implementation in the online is less straightforward because it requires reprogramming of some SVT boards and adjusting of the SVT map which describes the detector geometry. The effect of these additional correction has been studied, and found to reduce the sigma of the i.p. distribution from 48 to 45 /im.

133 CDF Trigger oa Impact

D° -» KM signalfromtrigger tracks

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

2.1

2.2

(KreXGeV/c2) Figure 2. Plots done with d a t a selected by the level 2 trigger requests described in the main text. Left: Impact parameter of the track with the second-highest impact parameter in each event. Right: Invariant mass distribution for track pairs assuming the two tracks to be n and K, and applying cuts optimized to select a D° —»• Kn signal.

4

Conclusions

The SVT is the new level 2 trigger processor dedicated to the reconstruction of tracks within the tracking chamber and the silicon vertex detector of the CDPII experiment. The device has been thoroughly tested during the early phase of Runll data taking (April-October 2001). The performance is already close to design. Jut.© S © i @ EtC €s§3

1. J. Spalding, Run~II upgrades and physics prospects, these proceedings. 2. S. D'Auria, Commissioning and Operation of the CDF Silicon Detector, these proceedings. 3. F. Palmonari, The CDF-II Silicon Tracking System, these proceedings. 4. C. Ciobanu et al, IEEE Trans. Nucl. Sci. 46, 933-939 (1999). 5. S.R. Amendolia et al, IEEE Trans. Nucl. Sci. 39, 795-797 (1992). 6. W. Ashmanskas et al, Performance of the CDF Online Silicon Vertex Tracker to be published in IEEE Trans. Nucl. Sci. (2001). 7. S. Donati et al, Nuovo Cimento 112A n . l l , 1239-1243 (1999).

COMMISSIONING A N D OPERATION OF T H E CDF SILICON DETECTOR S. D'AURIA ON BEHALF OF THE CDF COLLABORATION Dept. of Physics and Astronomy, University of Glasgow, G12-8QQ Glasgow, U.K. e-mail [email protected] The CDF-II silicon detector has been partially commissioned and used for taking preliminary physics data. This paper is a report on commissioning and initial operations of the 5.8m2 silicon detector. This experience can be useful to the large silicon systems that are presently under construction.

Introduction The collider detector at Fermilab (CDF) is designed to study protonantiproton collisions at a centre of mass energy of 1.98 TeV. The experimental apparatus had a major upgrade recently, in order to take advantage of the upgraded luminosity of the Tevatron accelerator. A general description of the apparatus can be found elsewhere in there proceedings 1,z and in the Technical Design Report 3 . The Silicon Detector is a high-precision microstrip tracker that covers the pseudorapidity region |^71 < 2, with an outer radius of 28 cm and a coverage of \z\ < 50cm along the beam axis, around the nominal interaction point. The detector consists of 3 subsystems: "LOO" is the innermost, single-sided strip layer at 1.35 cm from the beam line; "SVX" is a 5-layer, double-sided silicon microstrip tracker, made of 3 identical barrels placed along z, each with a 12-fold geometry along the polar angle (f>; "ISL" is the outer part 4 , consisting of one-layer, double-sided central barrel and two double-layer barrels, also with double-sided silicon microstrips, in the forward and backward regions. The three subsystems use the same front-end chip 5 ' 6 , the same control cards 7 and the same readout system 8 . They differ in the sensor geometry, the read-out hybrids and power supplies. The total area of the double-sided silicon is 5.8m2, and features 722500 strips read out by 5644 front-end chips. The chip is the heart of the Silicon Detector. It consists of a charge amplifier, a double correlated sample and hold circuit, an analog pipeline, a comparator and ADC and a threshold logic for sparsification of 128 strips. The analog pipeline is 42 cells deep and works at 7 MHz. The chip allows for Dynamical Common mode noise rejection (DCMNR) and dead-tilme-less

134

135

operation. Each chip is programmable with a 197 bit word to set bandwidth, signal polarity, DCMNR, threshold, calibration mask. The interconnection system includes 1.75 million wire bonds to single strips, 310 thousand wire bonds to chips, 10 thousand wire bonds to control lines and 816 connections to receiver/transmitter circuits that control more than one module (5 modules for SVX, 10 for ISL). Given the size of the detector and its flexibility in terms of parameters to set, the commissioning was a formidable task. 1

Commissioning

The Silicon Detector was extensively tested at the fabrication facility throughout the construction process. Each part had passed tests with very stringent parameters e.g. less than 2% of disconnected channels, no readout errors 9 . Functional tests were made at each step. All ladders had passed the tests before insertion in the barrel, but after barrel assembly 11 modules out of 360 have developed anomalously high noise on single channels or clusters of channels. We believe this problem is due to buildup of surface charge on the interface with the oxide layer. In fact it affects only a small fraction of detectors and only in layer-2 and layer-4. These are small-angle stereo detectors, their fabrication structure is different from the one of the remaining layers and they come from a different manufacturer. The detector was repeatedly tested during assembly and before shipment to the experimental hall, which is located a few kilometers away from the fabrication facility. The system grounding was always reasonably good and, provided that all the ground straps of the ladders were connected to the bulkhead, the noise performance was, for the majority of the devices, the same as measured on single devices before assembly. We could not test the Silicon Detector in the assembly hall because the electronics and power supply crates are mounted on the walls of the collision hall, while the cables had been installed on the main CDF detector. So the complete chain of readout, power supply and controls had to be tested all at the same time when the detector was rolled in. We initially cabled only a part of the detector consisting of 50 ladders. We finished the cabling during a one month shutdown of the accelerator. We tested parts of detectors as long as they wer cabled. The cables, the power supplies and the data acquisition components had been previously tested separately with one standard test stand and had passed the specification requirements. We repeated a detailed test of each component when it was in place in the collision hall. The power supplies and their cables were first tested

136

with a resistive load and voltages, currents, protection circuits and interlocks were checked. Then functional tests of the DAQ-related part were performed making use of a portable test device, that we called "wedge in a box". It consists of 5 SVX hybrids and one portcard, with solid-state cooling. This device was known to read-out correctly within a defined range of conditions so that functional test and debugging was performed on the DAQ and on its cables. This method was also essential to identify 3 low voltage cable bundles that had passed the passive load test but had a dispersion on the ground connection, due to mechanical damage. Once the "outer" part of the system was fully debugged, the real detector segment was connected to it and completely tested. The cabling and testing operation was long in time and required a large effort for a variety of reasons: the space available in the bore and the clearance for plugging the cables were extremely limited; the failure rate of component that were previously was higher than expected. It required a large, organized effort by 30 people. The main operating difficulty was due to the tuning of the interlock system that protects the system and was being commissioned at the same time. It had to allow the system to be powered under non normal conditions, especially high humidity, and keep the detector safe. A false alarm on the temperature was, initially, difficult to recognize and could block the cabling crew for several hours, inhibiting any power to the detector. Cables are assembled in bundles consisting of one voltage cable, one sense, one command, one bias voltage cable and 5 optical cables. A small number of cables had to be replaced: 4 voltage cables, 3 voltage sense cables, 6 data (optical) cables, 3 command cables out of 114 bundles. Also 3 FIBs (Fiber Interface Board) had to be replaced. The largest difficulty was due to the optical link between the front-end (portcard) and the Read-out (Optical Fiber Transition Module). The high failure rate was due to the light level mismatch between transmitter and receiver. A 9-channel optical cable is driven by a custom-made monolithic DOIM GaAs laser 10 . Five transmitters, (45 channels) share the same power voltage. They have been selected to have about the same characteristics, but the same was not done for the receivers. Some difference in light level produced a considerable error rate. In addition some receiver modules had flaky pin connections to the VME board. Re-insertion after contact cleaning was necessary for a large number of receivers. Another source of difficulty was due to the power supplies n : 35 out of 102 developed problems and had to be repaired. We had not received the PS modules in time for detector commissioning and were only able to operate a partial system. We have commissioned and operated 70% of SVX and 35% of

137

ISL. The LOO is only partially commissioned due to the late arrival of power supplies and also to wait for stable operation of the Tevatron beam. The ISL had a cooling blockage problem 2 , but the totality of the detector has been functionally tested, but only for a very short period to avoid overheating. The overall result of this test was that 3 SVX wedges could not be readout. For one of them the problem has been identified in a short between two signals inside the detector. This was not present before shipment. As a lesson learned, at level of system design, we should have allowed the use of standard protections for vital wire bond connections, at the cost of a more difficult procedure for test and re-work. Also minimizing the transport of the sealed detector, if at all possible, would be desirable. Two other wedges are being investigated during access in October 2001. 9 wedges could not be powered due to lack of tested power supplies, 48 ladders had readout problems related to optical power mismatch and could not be operated. 2

Integration

After test in the collision hall, the ladders were re-tested one by one with a stand-alone version of the DAQ program, checking for readout errors under "standard" conditions. Ladders that passed this test were integrated in the CDF DAQ. We operated 70% of the SVX and 35% of ISL. We had some difficulties also in this phase. Communication errors with the Power Supplies gave rise to spontaneous turning on and off of apparently random channels. This problem was solved by improving the timing of this communication, by decreasing the number of power supplies served by the same serial line and with software checks. In particular the PS Users Interface program was charged to add a number of tests that made it considerably slower than foreseen. As the light output of the DOIMS depends on their operating temperature, we had new cases of mismatch due to the increased light power when operating at - 6 ° C . We had to develop tools to synchronize the power supply to the daq system and to constantly monitor the operation in order to respond efficiently to any error message. The number of ladders integrated vs. time is shown in fig. 1 (a). The steady linear increase was due to the testing procedure, that had to negotiate time with "physics" data taking of the rest of CDF, so that we could operate the detector only in a fraction of time. In addition, we required the Silicon to be in off status during beam injection and unstable beam conditions. The availability of DAQ time was the main factor limiting a rapid increase of the number of integrated ladders. The setback around

138

Figure 1. (a) Number of Silicon modules integrated with CDF vs. time, (b) Timing pulse height vs. relative phase of the chip clock with respect to the beam crossing.

day 260 was due to a VME power supply failure and momentary inability to operate a part of the system. 3

Operations

We have collected to date 4.5 ph~l of Physics quality data on tape; this excludes detector studies and special runs to check the trigger rate. In order to operate the Silicon in the most stable way we decided to make full use of the chip capability and operate it in DCMNR-on mode. A fixed threshold of 5 ADC counts, about 16% of the most probable m.i.p. signal, gave a reasonable readout time, and occupancy completely acceptable at the present low trigger rate. We have not optimized the chip parameters yet, although we are already using the best compromise between having a uniform standard set of parameters and good performances. The first variable to set in the Silicon system was the timing, i.e. the relative phase between the bunch crossing and the issue of LI trigger to the chip. Failing to syncronize correctly would result in a loss of charge and "spillage" of charge in the neighbouring beam crossing packets. Before plug-in we have measured the delay of all command cables. They were all the same, within 2 ns. Using the the data from the first beam collisions we did a coarse and fine time scan, as shown in fig. 1 (b), and verified that all the detectors show a maximum at the same delay, as expected. The noise performance with and without beam are as expected and we are not experiencing any measurable pick-up from the beam or from the outer part of the detector. Some ladders have a 50% increased noise in those channels located above the support rails. This sensitivity to the infrastructure is

139

probably due to loosened ground connections and the problem is being-addressed during access. Otherwise the noise is the same as measured at the fabrication facility. Measuring the pedestal and noise is essential to operate correctly the detector. Two calibration methods have been implemented. Firstly, in "Datamode" , data are collected with free running trigger in read-all mode. Then noise and pedestals are calculated off-line and the parameters written to the calibration database. This method is intrinsically slow, because the calibration constants are available a few hours after the run is finished. Especially in the initial commissioning phase, when detector parameters and configuration were changing continuously, we experienced difficulties in monitoring the data quality due to the time delay between data taking and calibration data ready. The "X-mode" calibration will allow for a fast turnaround. All

la N

M m

\

M»

Layer 0

IS

Signal Peak 22.8

m

Strip Noise Width 1.7

I JL I «^fc. »

m m

>

" »

:

' " ' < * ' • •

m

m

CA»C

so

ioo

iso

a»

300

Figure 2. (a) Signal to noise for clusters on tracks, as measured with one ladder that was read in read-all mode. (b)Correlation between the cluster charge on z side and cluster charge measured on the <j> side, for all clusters. No tracking is applied. The gain of the amplifiers are well equalized.

the calculations of pedestal and noise are performed by the VME cpu in the collision hall. The final result is then loaded in the database in a matter of minutes. The FIB module can also subtract on-line any residual pedestal, so that data will be ready for clustering and for taking part to the second level trigger (SVT) 12.

140

10

15 20 2S m 3S 40 43 SO m&t Prsba&te duster Chsrge $0G counts*

Figure 3. Distribution of the peak of the Landau distributions. There is an entry in this plot for each ladder that was read-out. The dashed histogram refers to the <£-side. The gain equalization was applied to the small-angle-stereo layers. Entries with low charge correspond to ladders that were not biased, while entries with high charge are due to histograms with low tracking statistics.

The on-line data monitoring is performed on-line by analyzing a percentage of the data with off-line code. Cluster charge for all clusters and for those belonging to a track are monitored, as well as occupancy and track parameters. Off-line data quality is monitored with a larger statistics and makes use of the correct calibration constants. The cluster charge on tracks and the shape of the Landau distributions, corrected for the path lenght in Silicon, are checked both visually and with an automatic set of test. These include a Kolmogorov test on the Landau distribution, with respect to a reference histogram, as a powerful automated test that iags which ladders to look at. Other useful variables are the ratio between the number of clusters on tracks and the total number of clusters, the ratio between the number of clusters on $-side and the number of clusters on z-side. An example of pulse height distribution from SVX is shown in fig. 2-(a). The S/N was measured to be 13.3 and is in good agreement with the design value (11 to 17) 3 , even before optimization of chip parameters. By changing these it is possible to equalize the gain, betwen and z-side, as shown in fig. 2 (b), where the charge cor-

141

relation between p and n side is shown. The distribution of peak position of the Landau distributions is shown in fig. 3, that is an overview of the detector pulse height performance. After the end of access we shall optimize the chip parameters to maximize the performances of the detector, but also important is that the detector configuration will be stabilized, in order to extract physics signals with constant efficiency. We have already used the part of Silicon that we have commissioned to reduce the background in the J/V> dimuon mass peak: in fig. 4 the distribution with Silicon in is compared with the one obtained using only the information from the central tracking chamber and from the muon central chambers: the width is only 20% better, but the S/N improved by a factor of 2.5. The Silicon information has also been used in the second level tracking trigger 12 .

•)/>•> M V ,

-

r

|

-•

|

i

••••

1

'

l]

S

i I

1

• i l i ' i I '— •i Si J-V 434*24 tfWflfs. 3^2]±0.01 M*jVVC"

-

'I

-

£ 2000 _ -

|

"

1

f

1

1

V

.

.„„«-"- 1

I

150

83 g

100

*s 1

3.2

,

1

3.*

,

1

.

3.6

1 3.S

^•v****

.

4.0

2.70

2.90

3.10 3.30 HH Mass GeV/'c*

Figure 4. The J/ip peak as measured requiring two muons in the central region. On the right it is shown the spectrum for the subset of events that have at least one muon as a Silicon track.

Conclusions The CDF Silicon Tracker is on is way to be fully commissioned. 70% and 35% of the SVX and ISL subsystems respectively have been taking data with beam collisions and have been used for extracting the first Physics signals. After an access to the detector that is occurring during this conference we

142

plan to be able to run 97% of SVX wedges and recover as much ISL cooling as possible. The detector will be fully operational for physics-quality data taking by January 2002. Acknowledgments We thank the Fermilab staff and the technical staff of the participating institutions. This work was supported in part by Particle Physics and Astronomy Research Council, the U.S Department of Energy, Istituto Nazionale di Fisica Nucleare, The Ministry of Science Culture and Education of Japan and Academia Sinica, Republic of China. References 1. J. Spalding, Run-H upgrades and physics prospects, these proceedings. 2. F . Palmonari [CDF Collaboration], The CDF-II Silicon Tracking System, these proceedings. 3. CDF Collaboration, The CDF-II Detector Technical Desiggn Report, FERMILABPub-96/390-E. 4. A. Affolder et al. [CDF Collaboration], Intermediate Silicon Layers Detector For The Cdf Experiment, Nucl. Instrum. Methods A 4 5 3 , 84 (2000). 5. M. Garcia-Sciveres et al., The SVX3d Integrated Circuit For Dead-Timeless Silicon Strip Readout, Nucl. Instrum. Methods A 4 3 5 , 58 (1999). 6. T. Zimmerman et al., SVX3: A deadtimeless readout chip for silicon strip detectors, Nucl. Instrum. Methods A 4 0 9 , 369 (1998). 7. J. Andersen et al. The portcard for the Silicon Vertex Detector Upgrade of the Collider Detector at Fermilab IEEE Trans. Nucl. Sci. 4 8 , 504 (2001). 8. M. Bishai [CDF Collaboration], The CDF Silicon data acquisition system for Run-H, these proceedings. 9. G. Bolla [CDF Collaboration], Testing And Quality Insurance During The Construction Of The Svxii Silicon Detector Nucl. Instrum. Methods A 4 7 3 , 53 (2001). 10. M. Chou et al.Dense Optical Interface Module (DOIM)Fermilab Internal document, Mar. 1996 11. Custom made power supplies for the SY527 system are made by CAEN. 12. I. Fiori [CDF Collaboration], The CDF on-line silicon vertex trigger, these proceedings.

T H E ASSEMBLY OF T H E A M S SILICON T R A C K E R , VERSION 1 A N D 2 C. CECCHI University

of Perugia

and INFN, Via A. Pascoli, 06100 Perugia, E-mail: [email protected]

ITALY

The AMS (Alpha Magnetic Spectrometer) experiment is a detector designed to search for antimatter and dark matter. A first version, AMS1, has flown on June 1998, on board of the Shuttle Discovery, during the STS91 mission. The complete detector, AMS2, will be installed on the International Space Station in 2004 and it is foreseen to operate for a period of three years.

1

Introduction

The AMS experiment is a space born detector which will search for antimatter and dark matter by measuring with the highest accuracy the Cosmic Rays composition, thanks to its large acceptance (~ 0.5 m2sr) and long observation time (three years). In this paper I will give a short overview of the AMS experiment 3 ; first I will describe the construction of the tracker of the AMS01 detector and I will then present the status of the construction of the new silicon tracker for AMS02. The performances achieved with the old detector, and the expected results with the new one, will be also discussed.

2 2.1

AMS01 The AMS-01 detector

The detector 4 is composed by a permanent magnet equipped with a tracker, which consists of 6 planes (two outside the magnet and four inside) of 300 /jm thick double sided silicon microstrip detectors, a time of flight system, based on scintillation counters, to measure the velocity of the particles, and a threshold cerenkov counter to discriminate between low energy hadrons, electrons and positrons. The magnet is made of blocks of Nd-Fe-B and gives a dipolar field of 0.14 T. The total acceptance of the detector is 0.5 m2sr, for an analysing power of 0.14 Tm2. The spatial resolution of the tracker is 10/xm in the bending plane and about 30 ^m in the non bending plane 5 .

143

144

'

..

.MffFLDINGSUrPOIU (AliOX)

HYBjiUDEOXdOP) 10HG X M T O N

CF1SHYBEBD L«J>DlRHSIHIOH.-::.Mi!.w l

>..«" HYBRID BOX (BOTTOM)

Figure 1. View of the components of a ladder (left). View of one ladder, silicon side (right).

2.2

The assembly of the AMS01 tracker

The basic component of the AMS tracker is the so called "ladder" (a detailed description of the ladder assembly procedure can be found in *). The ladder is made of several silicon sensors (from 7 to 15) aligned and glued together. A foam reinforcement is glued on both sides of the silicon. On one side of the ladder is glued a upilex fanout to reroute the strips signal to the readout electronics, which is placed at the end of the ladder. A view of the components used to assemble a ladder is shown in Figure 1 (left). The ladders are then mounted on Carbon Aluminium honeycomb. A picture of the ladder in the bending side is shown in Figure 1 (right). To obtain a high quality detector, two aspects have been particularly considered: stringent requirements on the quality of single components and mechanical precision of the assembly procedure. For the first one, acceptance criteria have been applied on silicon sensors and on the ladder itself in order to fullfiU the final specifications. The most important parameters used to accept or to reject sensors and ladders are shown in Table 1. In total 65 ladders have been produced, 6 of them have been rejected because classified as bad. 21 ladders out of 65 were classified as marginal, meaning that they had one or more parameters close to the edge of the acceptance criteria. Ladders can be declared marginal or bad because of different

145 Table 1. Acceptance criteria applied on silicon sensors and on assembled ladder for AMSOl and AMS02. Silicon acceptance criteria Total leakage current Hot strips s-side Hot strips n-side Ladder acceptance criteria Total leakage current Hot strips s-side Hot strips n-side

AMSOl

AMS02

6/JA

2nh

8 15 AMSOl 30/zA 3% 5%

6 4 AMS02 10M 2% 2%

reasons, assembly problems, damaged sensors, gluing problems (insufficient glue, overspread), bonding problems (damages to substrates, failed bonds), high leakage current or bad electronics. At the end of the construction of the AMSOl silicon tracker the number of bad channels, on all the six planes was of the order of 9%. The second important point of the construction of the tracker is the accurate positioning of the sensors in the ladder. The alignment relies on a very precise cut of the wafers. Therefore the first step before the assembly is the check of the sensor cut. The distance between the reference crosses of the wafers and the edge of the sensor has been measured for all the wafer and a precision of the order of 3-4 /xm has been found. The performance of the AMSOl silicon tracker, in terms of the momentum resolution, which is strictly related to the precision in the alignment, is shown in Figure 2.

3 3.1

AMS02 The AMS02 detector

The future AMS02 detector will consist of a superconducting magnet equipped with a tracker, which consists of 8 planes (two outside the magnet and six inside) of double sided silicon microstrip detectors, of 300 ^m thickness, for a total of 7m2 of silicon detector. A time of flight system is present, as in the prevoius version, to measure the velocity of the particles. The apparatus is completed by a transition radiation detector to separate electrons from protons up to 300 GeV, a ring imaging Cerenkov detector to study heavy nuclei, and the an electromagnetic calorimeter to measure electrons and photons up to 1 TeV. The superconducting magnet gives an analysing power of 0.9 Tm2, a factor of seven more than AMSOl.

146

I

• Test Beam Data I STS91 Data

•

[ D

Figure 2. Resolution of the tracker of the AMS01 experiment as function of the rigidity as measured in the test beam(dots) and during the flight(flight).

3.2

The construction of the AMS02 silicon tracker

The AMS02 silicon tracker is made of 192 ladders of different lengths. The components for the assembly of the tracker are at an advnced phase of production: Silicon sensors Qualification and preproduction phases have been terminated. The production yield is of about 70%, taking into account the specifications required on number of hot strips on p-side and on n-side and on the total leakage current. Upilex fanout The first preproduction is finished, the second one is on progress. A total yield of 50% on K5 (long cables) and of 85% on K6 (short cables) has been produced, the yield been limited mainly due to bonding failures. Front End electronics 6 Production of capacitors and front-end hybrids is in progress, with a yield of about 80 % on both of them. The final goal is to produce 192 ladders, plus 30% spares, totally equipped and functional. The AMS02 silicon tracker is being assembled in three assembly centers, Pe-

147 160 140 120 100 80 60 40 20 0

s

£

1

60 50

ovn.«

TOO

^

40 30

L 2

20 10 UVJ . . . . L.. 4 6

. . I ... .

0

8

L

r

250

i w u . ,M ., 2

4

...., .

6

noise distribution k

noise distribution S

L™

140

-

£

"S,

120

200

100 150

80 60

100

40 50 0

20

,,

.

i

20 40 w noise distribution S

.

0

1.,

.

1

.

20 40 law noise distribution k

Figure 3. Noise measured in one of the produced ladders. The bottom plots show the noise measured without subtracting the common noise.

rugia, Geneva and an italian industrial research center G&A Engineering 2 . Due to the relatively large scale of the construction, an industrial approach has been chosen. This has several implications; in particular a stricter documentation is necessary as well as a strong interaction between physicists and the company. A large effort has been put in the technology transfer to the industry. The industrial research center G&A Engineering, will assemble half of the ladders, the rest will be shared between the University of Perugia and the University of Geneva. A total of 27 ladders have already been assembled at the rate of 1.5 ladder/week. The results from the first 6 ladders are available. An example is shown in Figure 3, were the noise measured in one of these ladders in shown. The results of these tests suggest that the quality of the AMS02 ladders is excellent. The future plans for the AMS02 tracker construction are organised in order to complete the ladder assembly by the end of 2002; after that the assembled ladders will be integrated in the tracker, and finally the complete test of the detector will follow. The flight for the AMS02 detector on the International Space Station is foreseen for end 2004, beginning 2005.

148

4

Conclusions

The AMS01 silicon tracker comprised 57 ladders, for a total of 2m 2 of silicon microstrip detectors. A lot of experience has been gained in assembly procedure and a good quality has been obtained in assembly accuracy and in module quality. The AMS02 silicon tracker is now in construction, and an industrial approach has been choosen, due to the dimension of the detector. The assembly is in progress in three different assembly lines. A total of 27 ladders have been assembled and the end of the assembly is foreseen for December 2002. References 1. 2. 3. 4. 5.

Talk of M. Pauluzzi at Vertex 2000. G&A Engineering s.r.l., Localita Miole 100, 67063 Oricola (AQ)-Italy AMS Collaboration, J. Alcaraz et al., Phys.Lett.B461 (1999) 387-396 G.M. Viertel and M. Capell, Nucl. Inst. Meth. A419 (1998) 295-299 J.Alcaraz et al. Nuovo Cim. 112A 1325 (1999) W. J. Burger, Nucl. Inst. Meth. A435 (1999) 202 6. G. Ambrosi, Nucl. Inst. Meth. A435 (1999) 215

T H E AMS I N F R A R E D T R A C K E R A L I G N M E N T SYSTEM FROM STS91 TO ISS W. WALLRAFF AND V. VETTERLE /. Physikalisches Institut, RWTH-Aachen, Germany E-mail: [email protected] J. VANDENHIRTZ LemnaTec GmbH, Schumanstrafie 18 Wurselen D52146 E-mail: [email protected]

Germany

We report on AMS tracker alignment control in space using artificial laser produced straight tracks (flight data AMS-01, laboratory tests AMS-02) as well as precisely measured high momentum cosmics tracks.

1 1.1

AMS experiment AMS-01

The large acceptance Antimatter Spectrometer (AMS) experiment l 2 has been operated successfully on the NASA STS91 shuttle flight (02-June-98 12-June-98, AMS-01). It will be redeployed, including major upgrades, for a 1000 day data taking mission (AMS-02) on the International Space Station late in 2004; see talk by R. Battiston at this conference.

1.2

AMS Si-tracker Tracker Alignment System TAS

AMS particle tracking is based on 8/6 (AMS-02/01) planes of double-sided Si detectors providing a maximum detectable rigidity (MDR) of 3000(500) GV by measuring the sagitta of the tracks in a 0.9 T superconducting (0.12 T permanent NdFeB) magnet. The sagitta can be determined with an accuracy of 22(25) fim. In AMS the position stability of the tracking elements is controlled using nearly straight tracks. Fig. la shows the laser beams and their measured profiles (recorded in space and transmitted to ground on June 4th 1998) in the AMS-01 configuration. From an analysis of the residuals for > 4 GV tracks individual ladder displacements have been derived 3 4 (for principle see fig. lb, results fig. 4).

149

150

AMS Laser & Cosmlcs alignment

Figure 1. a) AMS 01 Si tracker and t h e Tracker Alignment System. T h e insert shows laser profiles observed while AMS was in orbit, b) ladder displacement measurement with cosmic tracks (curvature greatly exaggerated, 10 GV sagitta 0.5/3 m m for AMS-01/02).

1.3

TAS technical aspects

Artificial tracks are produced by 1082 nm Laser radiation. Si is highly transparent at 1082 nm, provided the natural reflectivity (nsi = 3.3) can be reduced and shadowing by the metallization of the readout strips can be kept small. AMS alignment sensors are antirelective coated and use 10 /im wide readout strips in the Laser impact areas. Thus single layer transparency can be as high as 50%. It has been shown (AMS-01) that a Laser ray can be recorded in 6 Si layers in sequence 4 . The AMS-02 tracker (8 planes Si, SC magnet) will be equipped with 2 sets of 10 laser rays each, that traverse the Si in 2 opposite directions (fig. 2b) and do overlap in the central planes. These rays are detected by generating electron hole pairs in the fully depleted Si particle detectors 4 6 . Signals from the alignment rays are recorded exactly like the charged particle tracks. 1082 nm Laser radiation is generated with high efficiency in DBR-Laser diodes coupled to monomode optical fibres that deliver - via miniature projection optics - low divergence circular rays into the tracker (fig. 2a). At the photon intensities readily available from Laser diodes (> 108 / pulse) signals exceeding that of 1000 mips can be produced in the Si layer (thickness 300 /im) close to the projection optics. At adequate Laser intensities this approach allows high precision (< 2 fan) tracker stability tests in very short time (< I s ) . The fully operational system (20 beams) weighs less than 5 kg.

151

AMS-02 S i - t r a c k e r &. l a s e r alignment rays 5 pairs chmck {x,y} in tke

cmntra.1 area {300 as 3.00 mm2}

mZTnm

of a l l 8 detector plaaas

Figure 2. AMS 02 Tracker Alignment System, a) laser radiation generation, b) basic configuration of 2 x 5 laser ray pairs.

2 2.1

TAS a n d A M S t r a c k e r stability STS91 flight

Overall the AMS-01 tracker 7 has been extraordinarily stable. Over the whole flight - including lift-off and landing - all tracking elements were found at their expected positions within ± 15 fun. Laser measurements (once per 3 orbits (on manual command)) were confirmed by observation with stiff (almost straight) tracks comparing extrapolated tracker hits with actually measured ones (fig. 4). Correcting for the time evolution of the small but finite displacements results in an approximately 20corrections are important only for the high rigidity tail of the cosmic ray spectra observable by AMS (p > 80 GV). The excursions observed are probably due thermal effects because they correlate with changes in flight attitude hence heating by the sun.

152

Figure 3. a) AMS-02 Si sensor transparency, standard and with antirefiective coating, b) high quality coatings eliminate front back interferences and distortions of the laser beam while passing through a sensor.

2.2

prospects

Based on AMS-Q1 experience a tracker stability verification along 10 lines and with better than 4 jura accuracy can be expected from short (< 10 s) runs 4 times per orbit. This measurement over an area of 300 x 100 mm 2 in the center of the acceptance is complemented by minimizing for high rigidity tracks over the full acceptance the pulls in the redundant trackfit with 8 points through the rather smooth and very stable AMS-02 B-field. The method of position control of Si trackers with artificial laser generated straight tracks has not only applications for space experiments. A similar system has recently been studied for implementation in the large area Si tracker of the CMS experiment 5 to be installed at LHC.

153

Figure 4. AMS 01 tracker stability during the STS 91 space flight; a) time line of y displacements (JLB), from stiff cosmic tracks; squares indicate Laser data. Frequency distributions of observed displacements in the AMS-01 (Laser) alignment ladders before b) and after correction c) observed for high momentum cosmic rays during the STS91 spaceflight; details are given in the references 3 and 4.

Acknowledgments NASA, DoE and DLR have generously supported this work. We like to thank the AMS collaboration and the Si tracker team for their cooperation. The meeting at Villa Olmo has proven again to be highly useful, many thanks to the organizers. References 1. 2. 3. 4.

U. Becker, ICRC XXVI, ice 1574, Salt Lake City (1999). W. Wallraff, JHEP-PREP-hep2001/211, Budapest (2001). W. Wallraff et al., ICEC XXVII OG 110, 2197, Hamburg (2001). J. Vandenhirtz Ein Infrarot Laser Positions Kontroll System fur das AMS Experiment, PhD thesis RWTH-Aachen (July 2001). 5. B. Wittmer The Laser Alignment System for the CMS Silicon Microstrip Tracker, PhD thesis RWTH-Aachen (November 2001). 6. Weihua Gu Characterization of the CMS Pixel Detectors, PhD thesis RWTH-Aachen (October 2001). 7. J. Alcaraz, et al.; A Silicon microstrip tracker in space; Experience with the AMS Silicon tracker on STS-91, Nuovo Cimento 112A, 1325 (1999).

P E R F O R M A N C E OF T H E BABAR TRACKER

SILICON V E R T E X

V. RE INFN-Pavia and Universita di Bergamo C. BOREAN, C. BOZZI, V. CARASSITI, A. COTTA RAMUSINO, L. PIEMONTESE INFN-Ferrara and Universita di Ferrara A.B. BREON, D. BROWN, A.R. CLARK, F. GOOZEN, C. HERNIKL, L.T. KERTH, A. GRITSAN, G. LYNCH, A. PERAZZO, N.A. ROE, G. ZIZKA Lawrence Berkeley National Laboratory D. ROBERTS, J. SCHIECK University of Maryland E. BRENNA, M. CITTERIO, F. LANNI, F. PALOMBO INFN-Milano and Universita di Milano L. RATTI, P.F. MANFREDI, INFN-Pavia and Universita di Pavia C. ANGELINI, G. BATIGNANI, S. BETTARINI, M. BONDIOLI, F. BOSI, F. BUCCI, G. CALDERINI, M. CARPINELLI, M. CECCANTI, F. FORTI, D. GAGLIARDI, M.A. GIORGI, A. LUSIANI, P. MAMMINI, M. MORGANTI, F. MORSANI, N. NERI, E. PAOLONI, A. PROFETI, M. RAMA, G. RIZZO, F. SANDRELLI, G. SIMI, G. TRIGGIANI, J. WALSH INFN-Pisa, Universita di Pisa, Scuola Normale Superiore di Pisa P. BURCHAT, C. CHENG, D. KIRKBY, T.I. MEYER, C. ROAT Stanford University M. BONA, F. BIANCHI, D. GAMBA, P. TRAPANI INFN-Torino and Universita di Torino L. BOSISIO, G. DELLA RICCA, S. DITTONGO, L. LANCERI, A. POMPILI, P. POROPAT, I. RASHEVSKAIA, G. VUAGNIN INFN-Trieste and Universita di Trieste S. BURKE, D. CALLAHAN, C. CAMPAGNARI, B. DAHMES, D. HALE, P. HART, N. KUZNETSOVA, S. KYRE, S. LEVY, O. LONG, J. MAY,

154

155 M, MAZUR, J. RICHMAN, W. VERKERKE, M. WITHERELL University of California, Santa Barbara J. BERINGER, A.M. EISNER, A. FREY, A.A. GRILLO, M. GROTHE, R.R JOHNSON, W. KROEGER, W.S. LOCKMAN, T. PULLIAM, W. ROWE, R.E. SCHMITZ, A. SEIDEN, E.N. SPENCER, M. TURRI, W. WALKOWIAK, M. WILDER, M. WILSON University of California, Santa Cruz E. CHARLES, P. ELMER, J. NIELSEN, W. OREJUDOS, I. SCOTT, H. ZOBERNIG University of Wisconsin, Madison The BABAR Silicon Vertex Tracker (SVT) consists of five layers of double sided, AC coupled silicon strip detectors. The detectors are readout with a custom IC, capable of simultaneous acquisition, digitization and reduction of data. The SVT is an essential part BABAR, and is able to reconstruct B meson decay vertices with a precision sufficient to measure time-dependent CP violating asymmetries at the PEP-II asymmetric e+e _ collider. The BABAR SVT has been taking colliding beam data since May 1999. This report will give an overview of the SVT, with emphasis on its running performance.

1

Introduction

The SVT design requirements and features are described in detail elsewhere 2 ' 3 . The SVT has been designed to provide precise reconstruction of charged particle trajectories and decay vertices near the interaction region. In conjunction with the BABAR drift chamber, the SVT is responsible for particle tracking. The design has been driven primarily by physics requirements, with constraints imposed by the PEP-II interaction region and the BABAR experiment. The PEP-II e+e~ asymmetric storage ring produces B-mesons couples at the T(45) peak with a boost j3j — 0.55 along the beam direction. The resulting average separation of B decay vertices along the beam direction is Az w 250/jm. To avoid significant impact on the CP asymmetry measurement, the mean spatial resolution on each B decay vertex along the z-axis must be better than 80 /mi 1 . To adequately recostruct B, r and charm decays, a resolution of order 100 /xm in the plane perpendicular to the beam line is needed. It is desirable that the SVT provide a tracking efficiency of 70% or more for tracks with a transverse momentum in the range 50 — 120MeV/c. This feature is fundamental for the identification of slow pions from D*-meson decays. The SVT is required to be able to withstand 2 Mrad of ionizing radiation. Forthermore, the accelerator environment demands a radiation monitor-

156

"- Beam Pipe

Figure 1. Schematic view of SVT: longitudinal section. The roman numerals label the six different types of sensors.

ing system capable of aborting the beams when detecting excessive radiation. Finally, the SVT must readout physics events at a LI trigger rate of 2000 Hz. Requirements and constraints have led to the choice of a barrel-shaped SVT made of five layers of double-sided silicon strip sensors. The modules of the inner three layers are straight, while the modules of layers 4 and 5 are arc/j-shaped (Fig. 1), to minimize the amount of silicon required to cover the solid angle, while increasing the crossing angle for particles near the edges of acceptance. To fulfill the physics requirements, the spatial resolution for perpendicular tracks must be 10-15 fim in the three inner layers and about 40 ^m in the two outer layers 2 . The inner three layers perform the impact parameter measurements, while the outer layers are necessary for pattern recognition and low pr tracking. 2

Performance

Due to a series of minor mishaps incurred during the installation of the SVT, nine out of 208 readout sections (each corresponding to one of two sides of a half-module) were damaged and are currently not functioning. There has been no module failure due to radiation damage. The SVT hit efficiency is measured by finding out if there are SVT hits corresponding to the traversed silicon sensors for reconstructed tracks. A global efficiency of about 97% is measured on the half-modules connected to functioning readout sections (Fig. 2). This includes inefficiencies from software reconstruction, dead channels, broken AC coupling capacitors, dead channels on front-end electronics and so on. The

157 1

1

& C

v.- ."""-r"" -

i

* • • ' •

i

1

' '

•

• "V

**-»** ;

+-*"-

--—.-„

- —_ ~

' • - . —

1

''"'"I

0.9

0.9

'o °-8 - 1

h

2

3

4

. 1

5

0.6

4

3

2

5

+

,

ti 0.7 I

i

.

1 . . . .

I !

20

30

+

,

,

i

i _ ^

,

.

: ,

i

• _•

20

40

forward half-module number

0.8

+

•

•

'

• '

30

•

, •

0.7 0.6

backward half-module number

Figure 2. SVT hit reconstruction efficiency, as measured on n+(i events. The horizontal axis corresponds to different modules, vertical lines separate different layers.

— 40

E 3 20 -50 0 50 angle (degrees)

-50

0

50 -50 angle (degrees)

0

50

-50

o

50 -50 angle (degrees)

0

50

Figure 3. Single hit resolution for (j> (left) and z (right) strips.

spatial resolution has been measured on tracks reconstructed by the SVT alone. The resolution (Fig. 3) depends on readout pitch, number of floating strips, noise, and is measured to be about 15/um and 30 — 40/xm for inner 1-3 and outer 4-5 layers, respectively. The reconstruction efficiency of slow pions from D* -¥ Dn decays has been estimated by comparing real and simulated data and found to be larger than 70% for pion momenta larger than 50 MeV/c (see Fig. 4). 3

Radiation Damage

A system of 12 PIN diodes is located near the first SVT layer to monitor continuously the radiation exposure of SVT and to protect the SVT from excessive radiation due to beam instabilities. The radiation dose strongly depends on the azimuthal angle: the diodes situated in the horizontal plane see about 10 times the radiation dose as the out-of-plane diodes. The highest measured dose at the time of writing (October 2001) is 920krad, to be compared with the radiation of budget at this time, 880 krad. Test detectors

158 .y 8000 — a )

. &

0.0

OJ. 02 03 0.4 Tiansveise M om entum G e V £ )

0i

a

01 02 03 04 Transverse M om entum PeVyfc)

F i g u r e 4. Slow 7r r e c o n s t r u c t i o n efficiency.

have been irradiated up to about 4Mrad and bulk type inversion has been observed at about 3 Mrad. The electrical properties after inversion have been measured and might allow regular charge collection;4 further investigations are ongoing to reach conclusive evidence. The front-end chips have also been irradiated up to about 4 Mrad: gain and noise degradation of about —3%/ Mrad and + 9 % / Mrad have been recorded, 5 respectively. According to current estimates, horizontal inner modules will accumulate a 2-3 Mrad radiation dose by the end of 2004, exhausting their planned lifetime radiation budget. 4

Conclusions

SVT has been operating efficiently since its installation in BABAR. The basic design goals have been fulfilled. Improved understanding on radiation damage on detectors and front-end electronics has been reached. References 1. The BABAR Collaboration, Letter of Intent for the Study of CP Violation and Heavy Flavor Physics at PEP-II, SLAC-443 (1994). 2. BABAR Technical Design Report, SLAC-R-457 (1995). 3. C. Bozzi et al., Nucl. Instrum. Methods A 447, 15 (2000). 4. I. Rachevskaia, Radiation damage to silicon by GeV electrons, talk given at the 5th International Conference on Large Scale Applications and Radiation Hardness of Semiconductor Detectors, July 4-6, 2001, Firenze. 5. A. Perazzo, private communication.

CHARGED PARTICLE TRACKING WITH THE HERA-B DETECTOR CARSTEN KRAUSS FOR THE HERA-B COLLABORATION Physikalisch.es Institut Universitat Heidelberg, Philosophenweg 12, 69112 Heidelberg The HERA-B experiment at DESY is a large acceptance fixed-target spectrometer using a silicon vertex detector, an inner GEM MSGC detector and an outer large volume honeycomb drift chamber for track reconstruction. The detectors are operated in a radiation environment comparable to LHC conditions. The tracking detectors had been finished at the beginning of year 2000 and have been successfully operated. They represent the worlds largest operated GEM MSGC system and the so far largest drift chamber system for high-rate application. We report on the detector operation, and summarize the performances achieved. We present the performance of the track finding algorithm and report on the reconstruction performance for the year 2000 data.

1

T h e H E R A - B Tracking S y s t e m

T h e H E R A - B detector was designed t o reconstruct decays of particles containing a b-quark with high accuracy. These particles are produced in proton (920GeV) nucleon interactions at t h e H E R A proton storage ring. T h e detector (see Fig. 1) is a forward wire target spectrometer with a silicon vertex detector, a large tracking system built of gaseous detectors, and particle identification detectors. T h e H E R A - B detector was operated until Aug. 2000, then repaired and upgraded during 2001 and will be ready for operation with

the HERA restart in 2002. T h e vertex detector system (VDS) is mounted in a vacuum vessel together with t h e target wires. It consists of 8 super-layers (each consisting of up t o 4 views) of silicon strip detectors with 50/xm readout pitch. T h e first 7 of these super-layers are mounted on R o m a n pots and can be moved to allow machine operation with increased aperture during injection. T h e particle rates in t h e VDS can reach u p t o 3 x l 0 7 s - 1 c m - 2 . In t h e center of mass system t h e coverage of t h e detector is larger t h a n 90% of t h e solid angle. T h e main tracking system of H E R A - B is divided into two p a r t s because the track density varies like 1/r 2 with distance r from the beam pipe. In the inner region (6-25cm distance from t h e center of t h e beam pipe) t h e tracks are measured with the inner tracker ( I T R ) . In this region the particle flux is u p t o 10 7 s - 1 c m ~ 2 . T h e inner tracker in t o t a l covers an area of 17m 2 . T h e outer region (20-600 cm radial distance from t h e beam-pipe) is covered by t h e outer tracker ( O T R ) . In this detector particle densities of u p t o

159

160

Figure 1. Top view of the HERA-B detector. The main tracking detectors stretch from the end of the vertex vessel to the electromagnetic calorimeter. The regions of the tracking system are labeled

10 s s' 1 c m - 2 are measured. The total active area covered by the OTR is 1000m2. The main tracking system starts behind the vertex detector. Several inner and outer tracker stations are installed in the magnet, where the chambers have to be able to work in a magnetic field of up to 0.85T. The tracking chambers are arranged in 0°, +5° and -5° stereo angles. 2

The Vertex Detector System

Charged particles produce more than 7 hits in the vertex detector which is sufficient for a stand-alone reconstruction. The tracks in the VDS are used to reconstruct primary and secondary vertex positions. This information can be used already on the second trigger level to cut on the distance of secondary vertices from the primary vertex to enrich the data sample with long-lived particles. The silicon detectors have a typical signal to noise ratio of 20-25 on the n-side and 15-18 on the p-side. The single hit resolution of the vertex detector is 3-4/Km. The vertex resolution has been measured to be around 40/jm. The very complicated alignment of the movable detector modules is stable and the positions of the system are known to a level of 2-7/mi in the direction perpendicular to the beam-pipe and 5G-250^m along the beam-pipe. This system is fully commissioned and reaches design levels.

161

Figure 2. a) Side view of the vertex detector vessel. The moving mechanism for the upper and lower pots can be seen. The protons enter the vessel from the right b) Side view of a single pot. The aluminum cover to separate the detector module from the primary vacuum is partially removed.

3

T h e I n n e r Tracker

The detector used in the inner tracker of HERA-B is a GEM MSGC. These chambers are a combination of MSGC 1 (micro strip gas chamber) and GEM 2 (gas electron multiplier), as shown in Fig. 3. Both devices produce gas amplification. At the GEM a gas gain of 20-50 and at the MSGC a gain of « 200 is reached. The division of gas amplification is necessary in the HERAB environment, because each device alone can not be stably operated in a hadronic beam with sufficient gas gain. The gain needed for the HELIX 128 readout chip 3 and a strip length of up to 25cm is at least w5000. The inner tracker system is built of 184 chambers with more than 140,000 MSGC strips. The gas used is a mixture of 70% Ar and 30% CO2. The chambers were designed for a maximal radiation exposure of 1 Mrad/y. The operation of gaseous micro pattern detectors in a hadronic environment is very difficult. It could be established that a careful conditioning of the GEM MSGCs in the beam is mandatory for a stable operation. Even after the training of the chambers, a slow switching-on procedure of the high voltage has to be strictly followed. In addition to the careful handling of the chambers, a fast monitoring system controls the high voltage and switches the system off in case of problems.

162

Figure 3. Schematic display of the GEM MSGC. The layout of an inner tracker chamber is shown on the right.

The most common operational problem caused by the conditions in HERA-B is a spark between the upper and the lower side of the GEM. These sparks can develop into discharges on the MSGC wafer or into shorts within the GEM itself. Both of the latter cases have to be avoided, because they permanently damage the chamber. The ITR has a GEM spark detection system integrated in the high voltage distribution system, which reduces the voltage applied at the GEM for « 1 minute after a GEM spark was detected. This protects the GEM from developing a permanent short from a spark. With all these measures it is possible to operate the GEM MSGC chambers in the high rate hadronic environment of HERA-B. On average the 150 inner tracker chambers installed in 2000 were operated for «1Q80 hours each. The inner tracker chambers had an average GEM spark rate of 1.2 sparks per chamber and 24h of operation. The GEM spark rate increases roughly exponentially with the applied voltage. The level of 1.2-8 sparks per operation day at voltages between 420 and 460V can be tolerated. The GEM voltages of all chambers were individually adjusted (420-460V) to reach a similar efficiencies. The individual adjustment was necessary to compensate the observed gain variations in the GEMs, which are most likely production induced. The efficiency level reached is shown in Fig. 4, efficiencies between 91% and 98% were measured, the design efficiency is 95%.

163 JL

•

' ~J

-

r

T—

"'"' ^^^^*^;;:-::;1"'

--''''

•

/'"'

„

~

±^

* — * MS91-+S < •-« MSM-+2 " »-•"-MS01-+3 ~ > - - * MS01-+4 "

'/

<tpm symbols i Udrift= -1A kV, Death. • .515 V

"~

H1W syratmls: UdrlS m -1J kV, Ucotfc » -SIS V

-

• — • MSM++I _ • — • MS81++2 .

/A'

7

[

,

1

.

1

,

1

,

1

,

1

GEMv«»ltsge[V]

Figure 4. Measured efficiency of some inner tracker chambers as a function of the applied GEM voltage.

4

The Outer Tracker

»)

K^E^^^33

Figure 5. a) Schematic view of the honeycomb chambers used in the outer tracker, b) Outer tracker efficiency for the chambers in the pattern recognition area with different T D C thresholds.

The outer tracker uses honeycomb drift chambers (Fig. 5a). These chambers are built of Pokalon Cfoil (a carbon loaded polycarbonate). This foil is coated with copper followed by a layer of gold to increase the surface conductivity. The gas mixture used in these chambers is 65% Ar, 30% CF4 and 5% CO2. The gas gain in the chambers is typically 30,000. Occupancies of up to

164 20% in the central region are measured. The module dimensions are typically 20cm width and up to 4.5m in length. The complete outer tracker is divided into 13 stations, which are instrumented with more than 1000 detector modules. During the R&D for the outer tracker severe aging problems had to be overcome. In the 2000 running period the system was fully installed and routinely operated, but production quality problems kept the efficiency lower than possible with these chambers. This can be seen in Fig. 5b, where some chambers with a high TDC threshold could only reach an efficiency of about 85%. Other - usually non trigger chambers - reached efficiencies of up to 95%. It will be possible to reach a higher level in 2002 because the problems have been identified and largely solved. 5

Track Reconstruction

The main tracking system of HERA-B is divided into three parts: The magnet area, the pattern recognition area and the trigger area (See Fig. 1). The track reconstruction starts in the pattern region. The reconstructed track segments are then extrapolated into the magnet and into the trigger area using a Kalman filter-based algorithm. The tracks are then matched to VDS tracks and a combined re-fit is performed to determine the momentum. During the commissioning phase in 2000 the detector performance was lower than expected and the reconstruction algorithm had to become more robust against missing layers. A new pattern recognition algorithm based on space-points was therefore developed in addition to the existing one based on triplet seeds. The space-points calculated from hits in several layers are matched and combined using quality criteria. This algorithm proved to be very robust and fast. In a Monte Carlo study with a very pessimistic assumption about the detector performance (85% efficiency for the ITR and 90% for the OTR and reduced resolutions for both detectors), the reconstruction efficiency for a /i-track from a J/ip decay is 97%. In Fig. 6 a typical reconstructed event in the pattern recognition region can be seen. The hits in the detectors are marked by crosses, the reconstructed tracks and track segments are marked by the interconnecting lines. 6

Summary

In HERA-B, the first tracking system for LHC-like conditions has been built. The high track density poses high requirements both on the detector technology and on the reconstruction. The vertex detector system is fully com-

165

missioned and works satisfactory. The GEM MSGC technology used in the inner tracker is difficult to operate in the given environment, but it is capable of Milling the requirements. The honeycomb chambers of the outer tracker are a suitable means to cover large areas in a high rate environment. The reconstruction is working with improved speed and efficiency. The commissioning of the HERA-B tracking system was largely completed in 2000. Due to improvements and repairs that have been completed in the meanwhile, the tracking system should reach design performance for the 2002 data taking.

Figure 6. A typical event in the inner part of the pattern recognition area of the main tracker.

References 1. A. Oed, Position Sensitive Detectors with Microstrip Anode for Electron Multiplication with Gases, Nucl Instrum. Methods A 263, 351-359 (1988) 2. F. Sauli, A new concept for electron amplification in gas detectors, Nucl Instrum. Methods A 386, 531 (1997) 3. W. Fallot-Burghardt et al., Helixl28-x User Manual, HD-ASIC-33-0697, http://wwwasic.kip.umheidelberg.de/~feuersta/projects/Helix/helix/helix.html, 1999

THE ZEUS MICRO VERTEX DETECTOR A. POLINI FOR THE ZEUS MICROVERTEX GROUP DESY, Notkestrasse 85, 22607 Hamburg, Germany, E-mail: [email protected] For the luminosity upgrade at the HERA ep collider, the ZEUS experiment has designed, constructed and recently installed a Silicon Microvertex Detector. The design of the detector and the performance of prototypes are discussed. The readout chain, the adopted online and control solutions as well as the integration within the existing trigger and acquisition systems of the experiment are presented. Tests with the full detector prior to installation and the first experience using cosmic rays in the final environment are reported.

1

Introduction

The ZEUS detector at DESY is designed to study high energy interaction produced at the HERA e±p collider. In the year 2000-2001, the 920 GeV proton - 27 GeV electron collider has undergone a substantial upgrade aiming at an increase by a factor 5 in the peak luminosity, corresponding to 200 p b _ 1 integrated luminosity per year. During the upgrade shutdown, ZEUS has been equipped with a silicon vertex detector which, besides a general improvement and extension of the track reconstruction, will enhance the identification of short lived particles. 2

Detector Layout

The Microvertex Detector (MVD) consists of a barrel section with three double layers of sensors surrounding the beampipe and four wheels in the forward, outgoing proton, direction. Longitudinal and transversal views of the detector with respect to the beam line are shown in Fig. 1. The sensors are single sided and made of high-resistivity (3 — 6 kti cm) 320 fj.ro. thick n-type silicon into which p+ strips, 12 /un wide and with a 20 fim pitch, are implanted. The signal is read out via AC coupling of 14 /mi strips placed at a pitch of 120 /jm. The rear side consists of a thick n+ diffusion. Test beam results have shown that, using capacitive charge sharing, a resolution up to 8 fjm can be obtained for tracks perpendicular to the sensor *. In the barrel region two consecutive sensors of square shape (60 x 60 mm), with orthogonal strips, are glued and electrically connected together via a copper trace etched on 50 pm thick Upilex foil. The connection of the

166

167 (a)

W

Forward MVD

0

I

D

Barrel MVD

[

=

(*>)

=

Figure 1. Transversal view with respect to the beam line of the forward (a) and the barrel sections (c) and longitudinal view (b) of the complete detector.

sensor assembly to the read-out hybrid is also done via a Upilex foil. This structure with a mirror one having perpendicular strip orientation forms a barrel module with 2048 read-out strips or 1024 channels. Five modules are mounted on a carbon fiber ladder that provides the required stiffness and support for the cooling pipes, cabling and slow control sensors. The forward section consists of four wheels, each made of two parallel layers of 14 silicon sensors of same type as the barrel section but with a trapezoidal shape and 480 read-out channels. Two sensors mounted behind each other form a forward segment and provide a two coordinate measurement via strips tilted by 180°/14 in opposite directions. A more detailed description of the detector layout and of the silicon sensors can be found in 2 . The read-out of the 207,360 channels is done via the HELIX 128-3.0 frontend chip 3 , specifically designed for the HERA environment. This 0.8 /xm CMOS chip is fully programmable and equipped with a 128 channel read-out system and a 136 step analog pipeline. The noise performance of the HELIX depends on the input capacitance (C) and is 400 + 40- C [pF] equivalent noise charge (ENC). Irradiation tests of the HELIX have been presented at this confererence 4 . During normal operating conditions the power dissipation of the HELIX is 2mW per channel while the power dissipation in the silicon sensors is negligible. Eight chips belonging to the same barrel module or the same forward segment are connected together in a programmable failsafe token ring and read out via a single digitization channel.. The analog serialized data are sent through passive copper links to dedicated 10 bit ADC modules 5 . The ADC system, after common noise and pedestal subtraction, provides strip clustering and two output data streams, the first based on cluster data for triggering purposes and the second for complete read-out of accepted events. The read-out is performed via VME Power PC boards running LynxOS and a dedicated software library 6 . The detector is operated at controlled humidity and temperature with dry air flow and water cooling infrastructure.

168

3

Final Assembly and Standalone test with Cosmic Rays

After assembly completion, the MVD was set up in a test environment and connected to the final read-out electronics and control infrastructure including the final set of power and signal cables. An extensive test program aiming at the complete understanding of the detector prior to the final installation was performed including a scope test of the front-end electronics and cabling, tests and development of the data acquisition and slow control system, study of signal to noise and detector performance under nominal conditions. Finally a cosmic trigger using scintillator layers surrounding the whole detector was set up and 2.5 million of cosmic events were collected over a period of 3 weeks of continuous running. During the test two barrel modules (out of 206) and a single front-end hybrid were found to be faulty. A preliminary analysis on the cosmic events has been performed requiring at least two hits in both projections in the modules of the outer layer. The input for these calculations was the designed geometrical position of the detectors. A resolution of 70 /im in r — (j> and of 80 fim in the r — z coordinate was obtained, dominated by the systematic uncertainty in the position of the sensors. After alignment corrections the design goal of a resolution of less than 20 fim looks feasible. Results from the cosmic run were also an average efficiency close to the geometrical acceptance of the silicon sensors (i.e. 0.93), a total number of bad channels lower than 2 % and a signal to noise ratio, including the full read-out chain, of at least 13. The noise and pedestal levels have shown to be stable at the level of 1-2 ADC counts although attention has to be payed to environmental parameters like temperature and humidity. Low humidity has turned out to be also important for low and stable currents when operating the sensors at depletion voltage. 4

Integration within the ZEUS experiment

In April 2001 the detector and the associated electronics infrastructure was moved to the final location in the ZEUS experiment. Quick tests on the whole detector have been performed as well as the integration in the ZEUS data acquisition and in the slow control environment. Before the HERA luminosity running, cosmic runs including the MVD and the other ZEUS detector components, were successfully achieved. 4-1

The MVD Data Acquisition and the new Global Tracking Trigger

The ZEUS data acquisition system is based on a three level trigger. Because of the HERA bunch crossing rate of 10.4 MHz, i.e. 96 ns between consecutive interactions, the experiments have required a pipelined read-out design. A

169 First Level Trigger, based on a reduced set of information from the detector components, is issued after 46 bunch crossings and reduces the trigger rate to aproximately 500 Hz. Detector data, stored in deadtime-free pipelines, is subsequently digitized, buffered and used for a Second Level Trigger (SLT) evaluation which is performed first by the single detector components, within 10 ms, and whose combined result is used to lower the rate to 100 Hz. For accepted events the complete detector information is read out and sent to the Third Level Trigger computer farm where event reconstruction and final online selection are performed. As a trigger component, the MVD has stimulated the realization of a new Global Tracking Trigger (GTT), based on the combined information of the Microvertex Detector together with the existing Central Tracking Detector (CTD) and the newly installed forward Straw Tube Tracker. The GTT will provide efficient trigger and reconstruction capability already at the SLT. For the hardware, the strategy has been to use, whenever possible, commercial off the shelf equipment easily scalable and mantainable. After investigation on the performance achievable in terms of data throughput and process latency, a solution based on a farm of standard PCs connected via a Fast and Gigabit ethernet network has been chosen. In the final system, data belonging to the same event and coming from the different tracking detectors is sent, according to a dynamic list of idle processors, to one computing node where the combined information is decoded and a complete tracking algorithm is run. After processing the GTT results are sent to the Global Second Level Trigger and the list of the idle processors is updated. All data transfers are done using standard TCP protocol. For development and performance tests a playback capability has been provided: upon a First Level Trigger the component front-end electronics (currently CTD and MVD) is read out and Monte Carlo or previously saved events stored in memory are injected into the GTT trigger chain at the component VME interfaces, and sent through the system exactly as for regular data. Preliminary measurements using a high transverse energy dijet photoproduction sample with high track multiplicity as well as previously recorded cosmic events, have shown no appreciable increase in the system deadtime or trigger latency, being consistent with the performance of the existing ZEUS DAQ and trigger systems during data taking at the end of 2000. Concerning the algorithm performance, the same simulations have shown that the track resolution is significantly better than the existing CTD SLT 8 , and either approaches or exceeds the current offline tracking resolution for tracks with only CTD information. The vertex resolution of the GTT system, even using only the CTD, is significantly better (10 mm) than the existing

170

CTD SLT (80 mm) due to the use of additional detector data not possible before because of the large processing time required. When including the MVD a nominal resolution of approximately 500 /j,m is reached. The efficiency of the GTT reconstruction to find the event vertex is similar to the present offline reconstruction, and approaches 100 % for events with 5 or more tracks. 4-2

Radiation Monitor

The MVD has been equipped with a composite radiation monitoring system to enable instantaneous and integrated dose measurements at several detector points and an automatic beam dump trigger. The designed lifetime of 5 years of operation in the HERA environment sets a limit on the integrated radiation dose that can be tolerated by the detector per year. A dose of about 250 Gy/yr can be seen as a safe limit for low signal to noise degradation partially recoverable by tuning of the front-end parameters. The system has proven to be an invaluable monitoring tool during the upgraded HERA machine commissioning. 5

Conclusions

The status of the new Microvertex Detector recently installed in the ZEUS detector has been reported. First tests in the complete environment and results based on cosmic data are encouraging. The Global Tracking Trigger concept looks promising and will significantly enhance the online tracking resolution. The luminosity run is expected to start early in 2002. Acknowledgements I would like to thank my colleagues from the ZEUS MVD group and in particular R. Carlin, U. Kotz, and C. Youngman for the useful discussions. References 1. 2. 3. 4. 5. 6. 7. 8.

M. Milite, DESY Thesis 2001-050 (2001). E. N. Koffeman, Nucl. Instrum. Methods A 473, 26 (2001). M. Feuerstack, Nucl. Instrum. Methods A 447, 89 (2000). J.J. Velthuis, these proceedings. T. Fusayasu, K. Tokushuku, Nucl. Instrum. Methods A 436, 281 (1999). A. Polini, ZEUS Note 99-071 (1999), unpublished. M. Sutton, C. Youngman ZEUS Note 99-074 (1999), unpublished. A. Quadt et al., Nucl. Instrum. Methods A 438, 472 (1999).

T H E R U N IIB U P G R A D E OF T H E CDF SILICON DETECTORS

S. CABRERA FOR THE CDF COLLABORATION Duke University, Physics Department, Durham, North Carolina 27708, E-mail:

[email protected]

A substantial portion of the Runlla silicon detectors of the CDF experiment will not perform adequately for the duration of Run l i b (15 f b - 1 ) because of radiation damage. The Silicon Vertex Detector (SVX-II) and Layer 00 will be fully replaced at the end of Run Ha. The Run lib silicon tracker has a baseline design that safely achieves the required radiation tolerance by using single sided sensors that are actively cooled. The new Run lib castellated layout contains more silicon surface area and has a more uniform radial distribution. It minimizes the number of hybrid and sensor varieties providing quick construction and assembly. The total mass in the tracking volume is reduced by eliminating unnecessary the passive material from the CDF volume.

1

Introduction

A successful Run II engineering run in 2000 established the 36x36 bunch pp operation at a center-of-mass energy of 1.96 TeV at the Fermilab Tevatron and produced (58 ± 17) nb~ of integrated luminosity for the comissioning of the CDF detector. Run II officially began in March 2001, and luminosity is currently being delivered to both the CDF and Dj? experiments which have started to carry out their ambitious physics programs 1 . Run II is divided into two distinct stages: Run Ha will collide proton on antiproton bunches with a bunch spacing of 396 ns switching to 132 ns bunch spacing at 2-1032cm~2 s - 1 of instantaneous luminosity, and Run lib will maintain the bunch spacing at 132 ns. The number of superimposed interactions per beam crossing a must be kept low enough to allow a comprehensive event reconstruction by the collider detectors. The latest prospects of the Run Ha luminosity levels are an integrated luminosity of 2 f b - 1 and a peak value in the instantaneous luminosity of 2-10 32 cm- 2 s- 1 . The transition from Run Ha to lib in the begining of 2005 will require a shutdown of approximately 6 months, primarily to replace the radiation dama T h i s magnitude obeys a Poisson distribution whose mean is a linear function of the instantaneous luminosity with a positive slope dependent on the number of bunches in the proton and p beam

171

172

aged components of the silicon detectors of the CDF and DJ? b experiments. The goal for the three year Run lib period (2005-2007) together with Run Ha is to accumulate 15 fb~ by increasing the instantaneous luminosity to 5-10 32 cm -2 s _ 1 . One of the major efforts is focussed on increasing the number of antiprotons per bunch in the collider by a factor of 2-3 over the Run Ha value of 3.0-1010 in 2 to 3 years without a major interruption to the Run Ha program. The upgrades in the accelerator system for p production, collection, handling and accumulation are well described elsewhere 3 . 1.1

Radiation damage in the Runlla silicon systems

The capabilities of the CDF experiment for the total integrated luminosity expected in Run II are limited by the radiation damage to the Run Ha silicon tracker c , which is estimated to survive to approximately 5 fb _ 1 . This radiation damage will affect both the silicon sensors and the readout electronics. The silicon sensors of the Run Ha silicon tracker will suffer a deterioration in performance primarily due to radiation damage to the bulk silicon, through displacements of silicon or impurity atoms from their lattice sites. One effect is an increase in the leakage current that degrades the ratio of signal to noise. A second effect is a change in the dopant concentration of the silicon which leads to an increase in the depletion voltage. The SVX-II sensors are double sided and must be operated fully depleted, otherwise the strips on the ohmic side would remain effectively shorted together. These sensors are inherently limited in the bias voltage that they can sustain: the electronics on both sides of the sensors are referenced to a common ground and the applied bias voltage must be held off by coupling capacitors that can withstand ~ 100 volts. The increased depletion voltage will be the dominant mechanism leading to the demise of the SVX-II layers with double sided sensors that combine axial and small angle stereo manufactured by Micron (layers 2 and 4). The same mechanism will be dominant in the case of the single sided Layer 00 sensors. The SVX-II layers with double sided axial and 90° stereo sensors manufactured by Hamamatsu (layers 0, 1 and 3) will die as a result of the combination of both bulk damage effects. The integrated luminosities that can not be exceeded in order to maintain reasonable detector performance were evaluated for the different components of the Run Ha silicon tracker 5 . These limits were under 15 fb _ 1 for Layer 00 and Layers 0,1,2, and 4 of the SVX-II. Other SVX-II components like b

See 2 for a complete description of the Run lib upgrade of the D 0 silicon tracker. See reference 4 for a complete description of the Runlla silicon tracker of the CDF experiment.

c

173

the Port-Cards 6 ' 4 ' 7 , that generate the control signals for the SVX3 readout chip, and the electronic components of the DOIMS (Dense Optical Interface Modules) connecting the Port-Cards with the front end electronics, are not expected to survive 5 fb _ 1 . All the layers of the ISL (Intermediate Silicon Layers) 4 will survive Run lib and will not need to be replaced. For mechanical reasons and the tight schedule for this upgrade programme a full replacement scenario of the SVX-II and Layer 00 systems was approved. 2 2.1

The baseline design for the R u n l i b replacement detector Comparison of the Run Ha and Run lib silicon systems

The baseline design of the replacement detector 8 ' 9 has been developed to accomplish several goals. In order to achieve the required radiation tolerance, single-sided highbias-voltage sensors will be used instead of the double sided sensors used in SVX-II detector. Therefore to retain or improve the tracking capabilities of the Run Ha silicon tracker at least twice as many sensors as in the current Run Ha detector will be needed. The current SVX3 chip will be replaced by a new rad-hard SVX4 chip. The new Run lib castellated layout (see figure 1) takes advantage of the entire volume between the beam pipe and the ISL space frame from 2 to 18.5 cm in radius. There is more silicon detector area given that the outer instrumented layers are located at larger radii in the Run lib design (see table 1). The mass distribution is better because the radial gap in the Run Ha detector between the ISL and the SVX-II which was occupied by passive material such as cables and Port-Cards (see section 2.3), has been eliminated. The wedge structure of the Run Ha detector, that maintained the same 12-fold <j> segmentation for all the layers, is abandoned in Run lib in favour of minimizing the number of hybrid and sensor varieties. This is required for the quick construction and assembly, and reduces the overall cost. Layers 2 through 6 (90% of the detector) have a uniform stave design and only 3 sensor types while SVX-II had 5 sensor types. There will be only 4 types of hybrids while SVX-II and Layer 00 have 12 hybrid types. The layers 0 and 1 together have 2 types of sensors, in comparison with the 2 types of sensors for Layer 00 alone in Run Ha. All the layers will use intermediate strips between the readout strips allowing smaller pitch and better hit resolution while keeping the channel count low. The total increase in the number of sensors from Ha to lib is ~250%

174 Power and Control Cables

Figure 1. Run l i b silicon tracker r-0 inner layout versus end view of SVX-II detector at the same scale. ISL detector is not included.

Table 1. The Run lib silicon tracker baseline design in comparison with the Run Ila SVX-II and Layer 00 detectors.

lib Layer L0 inner L0 outer LI L2 L3 inner L3 outer L4 inner L4 outer L5 inner L5 outer L6 inner L6 outer

lib axial R(cm) 1.95 2.35 3.35 4.55 6.45 7.70 9.50 10.6 12.5 13.6 15.5 16.6

lib stereo R(cm)

3.00 (90°) 4.90 (90°) 6.10 (90°) 7.35 (90°) 9.15 (2.5°) 10.25 (2.5°) 12.15 (2.5°) 13.25 (2.5°) 15.15 (90°) 16.25 (90°)

lib Ila
Ila R(cm) R(cm) 1.3 1.85 2.45 (90°) 2.99 (90°) 4.12 (90°) 4.57 (90°) 6.52 (1.2°) 7.02 (1.2°) 8.22 (90°) 8.72 (90°) 10.09 (1.2°) 10.64 (1.2°)

Ila segm. 12 12 12 12 12 12 12 12 12 12 12 12

while the increase in the number of chips and readout channels increases by only ~24%.

175

2.2

The silicon detectors

The silicon sensors for the Run lib detector have been chosen with regard to the existing and tested radiation hard technologies needed for the LHC: single sided sensors designed to be operated with high bias voltages to cover an increase in the depletion voltage due to bulk damage in the silicon. The sensors will be actively cooled to be operated at a lower temperature in order to survive higher radiation doses. The single sided sensors are also easier to manufacture, test and handle, avoiding problems during the detector construction. For the innermost layer the sensors will be identical to the Layer 00 sensors. For all the outer axial layers and outer 90° stereo layers the strip (readout) pitch is 44(88) (im. The outer 2.5° stereo layers has a strip (readout) pitch of 45.75(91.5) /zm. The axial layers 0 and 1 have a strip(readout) pitch of 25(50) /j.m. The 90° stereo layer 1 sensors have a strip(readout) pitch of 47.5(95) /xm. The 90° stereo sensors use well established "double metal" technology 10 to simplify readout by providing axial readout strips orthogonal to the implants. Consequently the level of noise will be higher in the 90° sensors, but still within the 40 pF limit for good performance with the SVX4 chip. 2.3

The data acquisition system

The stave is the basic structural unit of layers 1-6. The 156 staves of the outer 5 layers use a uniform design. Each stave is 60 cm long and it is supported at z=0 and ±60 cm. The z segmentation of a given layer consists of two staves, 6 readout modules and 12 silicon sensors. The stave for layer 1 is similar in concept to the stave of the outer 5 layers but both narrower and shorter. Each stave has built-in electrical bus cables and cooling tubes which are sandwiched between six axial sensors in the upper face and six stereo sensors in the bottom face. The electrical bus is a copper-kapton flex cable which is laminated to the carbon fiber surfaces of the stave. The sensors are glued on top of the cable. The core of the stave is fabricated of carbon fiber and rohacell, with integrated 2x6 mm PEEK cooling tubes to cool the sensors, the hybrids and the Mini Port-Card. The polyetheretherketone plastic (PEEK) was selected for radiation hardness and mechanical stability. A readout module is made up of two sensors wirebonded together and a readout hybrid, circuit board holding the SVX4 readout chips and related components. The hybrids are glued onto the silicon surface at one end of the silicon

176

sensor in all layers except the innermost (see section 2.4). Four-chip hybrids are used for the outer 5 layers, three-chip hybrids for layer 1 and two-chip hybrids for layer 0. They are wirebonded to the bus cables to provide a connection to the cables coming from the Mini Port-Card. The SVX4 readout chip integrates and digitizes the silicon signals and replaces the SVX3 chip used in the Run Ila silicon detector. It will provide a lower noise and faster rise-time amplification, which allows for larger detector capacitances. It uses standard 0.25 fxm deep sub-micron technology, which is very tolerant to high radiation. The signal-to-noise ratio is expected to be 30% better. Some radiation tests in progress suggest this chip will survive more than 30 fb _ 1 . The Mini Port-Card (MPC) is a simplified version of the Run Ila PortCard (PC) d that uses five Run Ila transceiver chips to regenerate the control signals for the SVX4 chips, but eliminates the complication of optical readout. The MPC will be glued to the end of a stave and electrically connected to the end of the stave bus with wire bonds. The rest of the components of the Run Ila PC not present in the Run lib MPC have been moved to the new Junction Port-Card (JPC) e in order to reduce the mass and the required cooling. The JPC will be moved outside the tracking volume to the face of the central calorimeter location where the cooling required by the active components is available.

2.4

The innermost layer

The layer 0 configuration is axially 12-fold symmetric. The carbon fiber support structure with integrated cooling tubes will be mounted on the beampipe to support the silicon. A readout module in layer 0 consists of 2 sensors glued and bonded together. All Run lib layer 0 sensors are 2 chips wide while Layer 00 has alternating 1 and 2 chip wide sensors. There is only one type of sensor and hybrid which are almost identical to the Layer 00 n two-chip hybrids. Lightweight cables connect the sensors to the hybrids which are located outside the tracking region (|z| >50cm) avoiding the problem of fitting hybrids on narrow inner sensors and reducing significantly the material and cooling requirements. The use of these cables causes a degradation in the signal to noise ratio by the noise pickup and a higher readout capacitance, and thus they are only used on layer 0. d e

See reference 6 ' 7 for more details about the Run Ila Port-Card. See 8 for more details about the Run lib Junction-Port-Card.

177

3

Conclusions

The CDF Run lib silicon tracker, which will replace the Run Ila detector because of radiation damage, has a baseline design which is comparable in performance and has many structural advantages. We have discussed this basic design and the components from which it will be built. The design includes 90° layers for precision tracking in 3 dimensions. We look forward to the physics this device will provide in Run lib. Acknowledgments I am indebted to the CDF Run lib Silicon design group, in particular I would like to thank P.Azzi-Bachetta, N.Bachetta, D.Benjamin, B. Flaugher, J.Goldstein, C.Hill, J.Incandela, M. Kruse, P. Maksimovic, T.Nelson, D. Stuart, K. Yamamoto and W-M. Yao for a very enjoyable collaboration. References 1. J. Spalding for the CDF coll., "The Run II upgrades and physics prospects'' , These proceedings. 2. A. Bean for the Dj? coll., "Design of un Upgraded Dj? Silicon Microstrip Tracker for Fermilab Run2", These proceedings. 3. M. Church, "Substantial upgrades to Tevatron luminosity", hepex/0105041. 4. F.Palmonari for the CDF coll., "The CDF II silicon tracking system", These proceedings. 5. N. Bacchetta et al., Run2b Silicon Working Group Report, CDF internal note 5425, September 2000. 6. M.Bishai for the CDF coll., "The silicon data acquisition system and front-end electronics for CDF Run II, These proceedings. 7. S.Dauria for the CDF coll., "Comissioning and Operation of the CDF SVX detector", These proceedings. 8. The CDF coll., "Run 2b Technical Design Report ", November 2001. 9. B. Flaugher and N. Bachetta (CDF coll.) internal talks: http://wwwcdf.fnal.gov/internal/run2b/Run2b-silicon.html 10. G. Bolla et al (1998-1999) "Silicon microstrip detectors on 6-inch technology", Nucl.Instrum.Meth. A435, 51-57. 11. T.K. Nelson, "The CDF Layer 00 Detector", Pub. Proc. The Meeting of the Division of Particles and Fields (DPF 2000) of the American Physical Society, Columbus, OH, Aug. 2000. FERMILAB-CONF-01/357-E

T H E B T E V PIXEL D E T E C T O R S Y S T E M L.MORONI * INFN Sez. di Milano, Via Celoria 16, 20133 Milano, Italy E-mail: luigi. moroniQmi. infn. it BTeV is a collider experiment approved to run at the Tevatron at Fermilab. The experiment will conduct precision studies of heavy flavor decays, with particular emphasis on CP violation, flavor oscillations and rare decays. One of its unique features is a state of the art pixel detector system, designed to provide accurate measurements of the decay vertices of heavy flavored hadrons that can be used in the first level trigger. This will insure the ability to perform precision study of a variety of final states and to search for rare phenomena with very high efficiency. The main design features of the pixel vertex detector are reviewed.

1

Introduction

BTeV is an experiment expected to be running in the new Tevatron CO interaction region at Fermilab. Its physics goals encompass precision measurements of all the CKM phases that can be extracted from CP violation observables in b decays, charm and beauty rare decays and flavor oscillations 1. This experiment exploits two important advantages of the "forward" region: the correlation in the direction of the b and b produced that improves the flavor tagging efficiency, and the boost that allows the implementation of an efficient trigger algorithm based upon the identifications of detached charm and beauty secondary vertices. The key element in this triggering approach is the pixel vertex detector. The pixel technology has been chosen instead of silicon strips because of the superior pattern recognition capabilities and stronger resilience to radiation damage. The geometry and readout scheme chosen are optimized for our physics goals through detailed Monte Carlo studies. A recent extensive beam run 2 that took place at Fermilab in 1999 validates the accuracy of our predictions. 2

BTeV pixel system overview

The pixel vertex detector is located at the center of BTeV, inside a 1.6 T dipole magnet. The detector system is located in a secondary vacuum, separated from the beam vacuum by a thin Al membrane ( « 150/um thick) that fulfills •REPRESENTING T H E BTEV PIXEL GROUP

178

179

Figure 1. Schematic view of a pixel half plane. Clearly visible are the shingled support structure with embedded cooling tubes and the hybrid pixel detector modules. The top layer is the high density cable routing signals and bias voltages.

also the purpose of shielding the sensor and associated electronics from RF pick-up. It encompasses 30 tracking stations, uniformly distributed over 1.28 m. The telescope length is matched to the expected length of the luminous region (az « 30 cm). Each station is composed of two pixel planes. Each plane has about 500,000 pixels in a 10 cm x 10 cm area, excluding a small central hole in the beam region. Fig. 1 shows one of the two L-shaped structures that make up a detector plane. These structures are retractable, as we are planning to move them away from the beam during injection. The design of a system satisfying these requirements without degrading the intrinsic hit resolution achievable is very challenging and the key aspects of our approach will be discussed below. The pixel devices are bump bonded to the readout electronics being developed at Fermilab. The hybrid sensor-readout assembly will be hosted on "shingled" carbon composite structures that allow close to 100% coverage over the active area. These structures will include integrated cooling channels. The unit pixel cell is 50 /mm x 400 jim. The technology chosen is n+/n/p+, namely the readout elements are located on the ohmic side of the device. This insures that the device is still operational after type inversion, even if it needs to be biased below full depletion voltage. In this approach the key design issue is the technology adopted to insure that adequate inter-pixel insulation

180

is achieved throughout the sensor lifetime. More details on our strategy will be given below. The final version of the electronics will feature analog readout including a 3 bit flash ADC in each pixel cell. The technology chosen to achieve the desired radiation hardness is 0.25 fim. CMOS process, implemented by two commercial foundries. Prototypes featuring a few cells have demonstrated that even after being exposed to 32 MRad of radiation, the analog performance of this front end still satisfies the experiment requirements. The readout electronics is close to the beam. Its control signal and output data are routed to the periphery through two flex circuits, a high density one, with very minute feature sizes, carrying the signals to the periphery of the pixel plane, connected to more conventional flex cables that fan the signals out of the vacuum box. The hybrid pixel detector is attached to a low mass support structure. The hybrid detector is going to be attached to the substrate at room temperature and then maintained at a temperature of about « —5°C to — 10° C. The substrate includes integrated cooling channels. Prototyping work is under way in collaboration with ESLI Technologies. The total material in a detector plane is 0.9% of a radiation length, mostly consisting of the silicon in the sensor and readout chip. 3

The hybrid silicon pixel detector

This paper will focus on our recent results on the sensor development and the bump bonding R&D 3.1

The pixel sensor

We have performed extensive test bench studies on sensors with p-stop interpixel insulation, which were manufactured for us by SINTEF, Oslo. Sensors were irradiated at the Indiana University Cyclotron Facility. Fig. 2 shows an example of the current-voltage data measured from a p-stop sensor including 12x92 cells and 10 guard rings. The measurement is performed by applying the reverse bias to the backplane and grounding all the pixel cells. We studied both oxygenated and non-oxygenated wafers. We found no significant difference between these two kind of wafers. In most of our measurements the breakdown voltage exceeded 400 V and did not change appreciably upon irradiation. The depletion voltage does not exhibit a steep increase as a function of the dose after type inversion. For example, at a fluence of 4 x 1014 p/cm 2 , sensors with an initial depletion voltage of 200 V showed a depletion voltage of

181

200

300

400

500

Voltage (V) Figure 2. Current-voltage characteristics for a SINTEF sensor before irradiation and after 4 x 10 1 4 p / c m 2 fiuence.

about 120 V. Irradiated and non-irradiated detectors will be studied with test beam data to be acquired in the summer of 2002. The static measurements indicate that the p-stop approach may be adequate for our application.

3.2

Bump bonding studies

We have identified two vendors for our initial study: Advanced Interconnect Technology, LTD (AIT) of Hong Kong, to explore the indium bump approach, and MCNC in North Carolina, USA, to explore the solder bump approach. To establish the reliability of each technology, as well as its radiation resilience, we have fabricated dummy detectors with rows of daisy chained bumps. The study of the electrical properties of these strips allows us to identify open or high resistance bumps. The resistance between neighboring strips allows to identify shorts. The indium bump detectors were on an even finer pitch than required (30 /xm). We monitored the quality of the bumps

182 Table 1. Rate of occurrence of problems (per bump) for dummy hybrid detectors manufactured with indium bumps (AIT) or solder bumps (MCNC). The cooling and heating cycles are described in the text. Problem bad contact after 1 yr bad contact after cooling bad contact after heating

In bumps 2.1 x 1 0 - 4 2.2 x 10~ 5 2.1 x 1 0 - 4

Solder bumps 4.0 x l O - 4 1.4 x l O " 4 6.3 x l O " 4

over the period of 1 year. We performed thermal cycling (alternating exposure to -10°C for 144 hours and +100° C for 48 hours in vacuum) and irradiated the dummy detectors with a 137 Cs source up to a dose of 13 MRad. Table 1 shows the rate of bump faults, manifesting themselves as high resistance developed through the daisy chained bumps. It can be seen that both techniques are highly reliable. Upon irradiation, the indium bumped detectors developed high resistance in a systematic pattern, almost in every first channel in a group of four. The grouping is a pattern characteristic of the specific sensor design. This effect is attributed to indium diffusion, with subsequent oxidation accelerated by radiation. Oxidized indium exhibited higher resistance. This effect was more prominent in the first cell of the group, as the edge cells have more exposure to Oxygen. Solder bumped hybrid detectors were affected mostly by a deterioration of the Al traces, that frequently became extensively flaky and bubbly upon irradiation. Thus the bump failure rates are extremely small in both technologies provided that we take enough precautions to avoid exposure to oxygen, as accelerated oxidation is an expected consequence of radiation exposure. 4

Acknowledgements

I would like to thank Gabriele Chiodini, Maria R. Coluccia, Simon Kwan and all the Pixel R&D group of Fermilab for useful discussion. I am particularly indebited with Marina Artuso for her precious suggestions in preparing this manuscript. References 1. A. Kulyavtsev, et al. BTeV proposal, Fermilab, May 2000. 2. J.A. Appel et al, hep-ex/0108014; to be published in Nucl. Instr. and Meth. A.

MONOLITHIC CMOS PIXELS FOR C H A R G E D PARTICLE TRACKING YU.GORNUSHKIN, G.DEPTUCH, M.WINTER IReS, IN2P3-CNRS/ULP, BP20, 67037 Strasbourg cedex, France E-mail: [email protected] G.CLAUS, W.DULINSKI LEPSI,

IN2P3/ULP, 23 rue du Loess, BP 67037 Strasbourg cedex 02, France

20,

The increasing need of high performance flavour tagging capabilities in particle physics experiments has triggered the development of a novel - fully integrated silicon pixel detector, called Monolithic Active Pixel Sensor. The first MAPS prototypes adapted to the detection of minimum ionising particles (mip) were designed and fabricated in standard CMOS technology. Their first tests demonstrate that the sensors detect mips with high signal-to-noise ratio, detection efficiency close to 100% and provide excellent spatial resolution. The main aspects of these results are summarised in this paper, together with preliminary results on the radiation hardness of MAPS. An outlook on the R&D started to adapt and integrate the sensors for future vertex detectors is also provided.

1

Principle of operation and detector description

Monolithic Active Pixel Sensors (MAPS) constitute a novel technique for silicon position sensitive detectors which allows integrating on the same substrate the detector element and the read-out electronics. The principle of operation of MAPS optimised for charged particle detection is based on a concept similar to that used in CMOS cameras - a rapidly growing up alternative to the CCD technology in digital photography and video applications 1. A thin epitaxial layer of low-resistivity silicon (doped to the level 1015 c m - 3 ) is used as a sensitive detector volume. The charge generated by the impinging particle at a rate of about 80 electron-hole pairs per 1 /xm is collected by a n-well/p-epi diode, created by n-well implantation into the epitaxial layer. The electrons liberated in epitaxial layer diffuse towards the diode within a typical time of a few tens of nanoseconds 2 . Because of the three orders of magnitude difference between the doping levels of the p-epitaxial layer and of the neighbouring p + + wells and substrate, potential barriers are created at the region boundaries, that act like mirrors for the excess electrons. Such a detector can be fabricated through a standard and therefore cost effective and easily available

183

184

_ ^

Pi Pi Pi =. Pi

Figure 1. Schematic diagram of MIMOSA read-out electronics.

CMOS process. In order to validate the ideas presented and to investigate the potential of the technology, four prototype chips with pixel arrays of slightly different design were fabricated. The first two prototypes, called MIMOSA" -1 and -2, were fabricated in two different CMOS processes: MIMOSA-1 in a 0.6-/xm process featuring an epitaxial layer of about 14 /on, MIMOSA-2 in a 0.35-/xm process with less than 5 /mi epitaxial layer thickness. MIMOSA-1 (resp. MIMOSA-2) contain four (resp. six) independent matrices of active elements having slightly different design. Each matrix consists of 64x64, 20 /xm wide, pixels. The individual pixel is comprised of three MOS transistors and a floating diffusion diode. The diode was implemented in the center of each pixel in most matrices. For one MIMOSA-1 matrix however, each pixel hosted four diodes connected in parallel, and one MIMOSA-2 matrix hosted two diodes per pixel. This was supposed to reduce the charge dispersion and collection time, at the expense of increased noise reflecting the higher node capacitance and the smaller charge-to-voltage conversion gain. Compared to MIMOSA-1, MIMOSA-2's design incorporates few new ideas, including features improving the read-out speed and the radiation hardness. The thinner epitaxial layer of MIMOSA-2 was expected to translate into less signal than MIMOSA-1. This feature was compensated by a lower noise level consecutive to smaller node capacitance and very small diode leakage current. Basic prototype parameters, such as the total conversion gain and the pixel equivalent noise charge (ENC), were determined with a 5.9 keV X-ray source of 55 Fe. Scematic design of MIMOSA chips is presented in Fig.l. More details on the chip architecture, principle of operation, results of the device simulation and of the tests with the X-ray source can be found in 2 ' 4 . "standing for Minimum Ionising M O S Active pixel sensor.

185

Figure 2. Left: Collected charge (most probable value) as a function of the cluster multiplicity. Right: Residue distribution of a track position measured by 1 diode/pixel MIMOSA-1 sensors (corrected centre-of-gravity method)

2

Test b e a m s t u d i e s

The first two MIMOSA chips were tested with a pion beam of 120 GeV/c at the CERN SPS. A beam telescope 3 of 8 planes of high precision silicon strip detectors, grouped in pairs of planes providing two orthogonal coordinates, was used to define the beam particle trajectory and intersection point with the MIMOSA plane with an accuracy of ~ 1 [im. The individual pixels were read out serially and an external 12-bit flash ADC unit digitised raw information from each pixel. The read-out clock cycle frequency used was 2.5 MHz for MIMOSA-1, and up to 10 MHz for MIMOSA-2. The off-line signal processing started with correlated double sampling (CDS) 5 to eliminate some dominant noise components. The remaining noise was evaluated from the first 250 events of every run and used to compute the pedestals subtracted from the CDS outcome as well as the noise entering the signal-to-noise (S/N) calculation. A cluster finding algorithm, using pixels with S/N > 5 as a seed, reconstructed clusters matching the beam telescope prediction. The charge collected in the cluster is displayed on Fig.2 (left) as a function of the cluster multiplicity. The values found are in agreement with the simulation results 6 . Typically, about 90% of the total charge was concentrated within a cluster of 3x3 pixels and the whole charge was contained within 18 pixels in most cases. As predicted, the 4 diodes/pixel option of MIMOSA-1 provided the most concentrated charge distribution. The charge collected with MIMOSA-2 (1 diode/pixel) was less spread than with MIMOSA-1, as expected from the thinner epitaxial layer. The coordinate of the track impact was measured using the charge distribution in the cluster. After a precise alignment of the prototypes w.r.t. the telescope, the v/idth of the distribution of the residue between the telescope extrapolation

186 Table 1. Major performances of the MIMOSA prototypes. Prototype Number of diodes/pixel Pixel noise (mean value) [e~ ENC] Signal-to-noise ratio (mean value) Detection Efficiency [%] Spatial resolution [/mi ]

MIMOSA-1 1 4 12 25 42 32 99.2 ± 0.2 99.5 ± 0.2 2.1 ± 0.1 1.4 ± 0.1

MIMOSA-2 1 9 22 98.5 ± 0.3 2.2 ± 0.1

and the MAPS reconstructed impact position varied from 1.7 to 2.5 /xm, depending on the chip design (Fig.2 right). The best resolution was achieved with MIMOSA-1, with 1 diode/pixel, the charge distribution being the widest and the S/N ratio being the highest. After subtraction of the track prediction resolution (~ 1 /xm), the intrinsic resolution of MIMOSA-1 came out to be better than 1.5 /xm. The performances of both prototypes are summarised in Table 1 and in 5 .

3

Preliminary results on radiation tolerance

The two prototypes were irradiated with 30 MeV/c protons and 10 keV Xrays. Irradiation effects were investigated by studying changes in the sensor response to 5.9 keV photons emitted by a 55Fe source. It was observed that 600 kRad X-rays provoke an increase of the diode leakage current by more than an order of magnitude. Similarly, a dose of 5 x 10 11 p/cm 2 induces a factor of 5 increase of the diode leakage current and a loss in the charge collected of about 40 %. On-going studies of sensors exposed to neutrons sources will clarify in which extend the observed sensitivity to protons is due to mechanical deformations of the epitaxy or if it originates from other degraded parts of the sensors (e.g. interfaces). Independently of the outcome of this study, present results on the radiation tolerance of CMOS sensors are encouraging, the first MIMOSA prototypes satisfying already the requirements of several applications (including those of a future e+e~ linear collider). Moreover, the influence of the temperature on the chip performances will be investigated in order to reduce its sensitivity and to recover (at least part of) the observed performance losses. Finally, forthcoming sensor prototypes will allow to continue exploring the radiation tolerance of the sensors and to design chips exploiting more and more efficiently the real potential of their technology.

187

4

Outlook: N e x t R & D steps

The validity of the CMOS sensor detection principle for charged particle being established, the present phase of development aims to explore the technology potential of the sensors (by testing alternative sensing devices, by studying and improving the sensor radiation tolerance, etc.) and to adapt them to various applications. Recently, a third prototype was fabricated in a standard 0.25 //m deep submicron IBM process with 8 x 8 fim2 wide pixels, and a fourth chip was manufactured in a 0.35 ^m AMS process with no epitaxy. These new arrays, which host various sensing devices, are currently being tested. Aiming to validate the sensors as building blocks of future vertex detectors, the design of the first real scale prototype (MIMOSA-5) was recently sent for fabrication in the same 0.6 /jm process as MIMOSA-1. The chips have the size of a CMOS reticle (i.e. about 19.4 x 17.4 mm 2 ) and consist of 4 independent matrices of 512x510 pixels with 17 fim pitch. They are stitched together along one direction across the wafer. Ladders of 5 or 7 chips are being achieved in this way. They will offer the first opportunity to test on real scale the performances observed with small scale prototypes, to investigate the chip production yield and the properties of stitching, and to experience the read-out of millions of pixels. A crucial issue of the next R&D steps will be the design of fast read-out micro-circuits integrated on the sensor substrate and including sparsification. 5

Conclusion

First prototypes of CMOS sensors designed for charged particle tracking show that this detection technique provides excellent performances (e.g. S/N ~20-=-40, detection efficiency >99%, spatial resolution down to ~ 1.5 /mi). Whether the latter can be reproduced on real scale will come out from the tests of the first large device, currently fabricated. The coming three years are expected to allow designing fast read-out micro-circuits with sparsification integrated on the chip as well as to explore extensively the real potential of this technology. The outcome of this programme will determine in which extent CMOS sensors can be used in future vertex detectors.

References 1. B.Diericks et al in Proc. IEEE CCD&AIS workshop, Brugge, Belgium (June 1997),p.Pl;

188

2. R.Turchetta et al., Nucl. Instrum. Methods A 458, 677 (2001); 3. C.Colledaniet al, Nucl. Instrum. Methods A 372, 379 (1996); 4. M.Winter et al., Proceedings of IEEE NSS Conference, October 2000, Lyon, France; 5. Yu.Gornushkin et al., Proceedings of the Vienna Conference on Instrumentation, February 2001, Vienna, Austria, and references therein; 6. G.Deptuch et al., Nucl. Instrum. Methods A 465, 92 (2001).

STATUS AND NEW LAYOUT OF THE ATLAS PIXEL DETECTOR P. NETCHAEVA INFN Genoa, Italy on behalf of the ATLAS Pixel collaboration [IJ

The ATLAS Fixe! detector is based on a set of radiation-hard electronics chips able to resist a dose of SOOkGy. The implementation of these chips in the DMILL technology did not give the expectedresults.Re-design of the radiation-haw! clips in DeepSubMicron technology is ongoing, but has implied a one and a half year delay in an already tight schedule. Major layout changes have therefore been necessary to allow installation of the ATLAS pixel detector at LHC start-up. This paper illustrates the status of the ATLAS pixel project, the motivations for the new layout, the way this should be implemented and the prototype fabrication and testing.

1. INTRODUCTION

The ATLAS Pixel detector has been already described elsewhere [1] and will be only briefly recalled here. This detector should be able to measure three space points in the pseudorapidity region up to |T|| = 2.5. It is made of two barrel layers, initially having 10.1 cm and 13.2 cm radius, plus so called b-layer of 4.3 cm radius. The main aim of the b-layer is to allow better impact parameter resolution for B-physics and for ©-tagging. The forward region is covered by set of 5+5 disks with internal and external radii of, respectively, 12.6 cm and 18.7 cm. The detector system is composed of modular units (fig.l).

Fig.l. ATLAS Pixel module.

Each module consists of a silicon sensor tile, sixteen front-end readout integrated circuits (FE chips) and a kapton flex hybrid. The flex hybrid distributes power and control signals to the FE chips and allows reading them out through a module control circuit (MCC). Passive components including termination resistors, decoupling capacitors and temperature sensor are also included. The sensitive area of a module, i.e. of a sensor tile, is 16.4 mm x 60.8 mm. The FE chips are connected to the pixel cells

189

19Q

through bump bonds. The size of one pixel is 50 am x 400 um and each FE chip serves 18 x 160 pixel cells. The flex hybrid is glued to the backside of the sensor tile; electrical connections from the MCC and the 16 FE chips to the flex hybrid are done through ultrasonic wedge bonding. 2. RADIATION HARDNESS

2.1 Sensors

Radiation damage in the severe LHC "environment can result in a voltage for full depletion of silicon detectors, which may exceed the maximum allowed operation voltage, and thus requires the detectors to be operated only in partial depletion [1]. The baseline design of ATLAS Pixel silicon sensors is characterized by: ® n + pixels on oxygenated n-bulk material (double-sided processing) to allow partially depleted operation [2], • moderated p-spray isolation to allow for high voltage breakdown after the type inversion, ® a bias grid to allow the sensor testing before module assembly. The sensors radiation hardness to 10 years of LHC operation has been proven [3]. One of the proofs recently obtained is the interpixel isolation test, which demonstrates no ionization-induced damage (fig.2). The sensors under test have been irradiated up to 500 kGy with 20 keV electrons; bias voltage of 500 V has been applied during the irradiation. The I-V curves show the pinch off at higher voltages for the irradiated sensor, but the interpixel isolation (about 100 MOhm at 100 V bias) is still sufficient.

upteSMkSy, with 500 V Was during Irffufatfon j

so

Fig.2. Interpixel isolation test

The preproduction of 150 good tiles has been done; careful tests have confirmed that the ATLAS specifications are met. The production will start in January 2002.

191

2.2 Electronics The four integrated circuits: FE chip (16 per module), MCC (event builder for 16 FE chips), VDC (VCSEL Driver Chip to drive data off-detector) and DORIC (to decode/encode clock and control signals) have been implemented in the DMILL radiation hard technology. Despite the large efforts of the Pixel collaboration this technology has not been able to produce the most complex chips (FE and MCC) with acceptable yield. Therefore the DMILL design work was stopped and maximum priority had been given to translate the various designs to the DeepSubMicron (DSM) technology. First digital DSM test chip had some limited functionality but worked roughly as expected. The analog test chip had been submitted to IBM and TSMC in February / March 2001. This chip contained the preliminary designs of analog blocks of 20 pixels and other analog circuits, as well as the final layout of the critical items. On-line results and first tests done after irradiation (at a dose of 610 kGy with 55 MeV protons) indicated little or no change in performance of the test chip. 3. INSERTABLE LAYOUT The failure of the DMILL design and the transition from DMILL to DSM has generated 1.5 years delay in the Pixel detector schedule and made impossible the "ready for installation" date initially foreseen (april 2004). This date was required to install the Pixel detector together with the Barrel Inner detector. The "Insertable" layout decouples the Pixel detector from the rest of the Inner detector and allows its independent installation later. This decoupling is obtained installing a 7 m long support tube together with the Barrel Inner detector. The Pixel system can then be slided in this support tube when the vacuum is broken and the forward section of beam pipe is out. The central section of beam pipe is integrated in the Pixel system and move with it. The Pixel system, services and beam pipe are prepared on surface, lowered using a temporary support and rolled into the support tube held by SCT barrel. The services for all but blayer should go out both sides. The b-layer services go out on one side to allow installation and dismounting together with the beam pipe in place. To fit the SCT bore the Pixel system had to be squeezed (the outer barrel radius has been changed from 14.2 cm to 12.2 cm) and shortened (fig.3). The number of disks is reduced (from 5+5 to 3+3); all disks have been made equal (8 sector each) and come closer to the interaction point. The inner radius of the Pixel system has increased from 4.1 cm to 5.0 cm because of the beam pipe radius increase. In total, the detector active area is reduced by 17 %. The Insertable layout has some reduced performance. In the plot for probability to have less than 3 hits (fig.4) we see some losses due to clearances in the barrel section.

192

Side C

111111 ifl 111111 M i l l Hit" HI H\

.—

<2>

2 -f

Side A

*<9

S 3 130 cm

SECTION A-A 1& Side A view 676 MODULES LMER181J2S

<mmoui.es LMEMR w.i IBS MODULES B-LnYERR SOS

Fig.3. The new Pixel detector layout.

The peak at r| = 1.5 -f- 2.2 is caused by inability of getting the first disk close enough to barrels, while the peak at T) = 2.5 arise because the minimum inner radius of the disks must be compatible with the b-layer insertion. The Pixel detector material is increased at high r\ and become more asymmetric since b-layer services exit one side (fig.5). The impact parameter resolution degrades at low momenta due to the b-layer radius increase. 0.)

».2S 0.2

&.15 0.1 0 05

^ T O ^ i i u,

jK

Fig.4. Probability for less than 3 layers hit versus pseudorapidity for the Insertable layout.

193

Mg.5. Material budget (X»). i i $ i t the previous layout, dark: Insertable layout.

4. LOCAL SUPPORTS It was necessary quite some ingenuity to preserve most of the mechanical design of ATLAS Pixel system already done, including both local supports: barrel staves and disk sectors. The shingled stave support (fig.6a) offers hermetic coverage over the full acceptance range of each barrel pixel layer.

Fig.6. Local supports: a) the bi-stave assembly: two staves with modules; b) tbe sector assembly with modules attached on one side.

Pixel modules are glued onto the carbon-carbon thermal management tiles (TMTs) to form individual mechanical modules. The TMTs have a machined groove, which accommodates a cooling tube, and makes contact to it through thermal grease. In the disk part the Pixel modules are placed on both sides of each disk to ensure continuous active coverage. This placement leads naturally to a sandwich construction for the disks with facings of high stiffness and high thermal conductivity on either side of a light core with an embedded coolant channel. A disk is divided into the angular regions, called sectors (fig.6b), of convenient size and cooling load. The set of quality control tests has been defined to estimate the local supports quality [4]. In order to achieve an acceptable lifetime of silicon detectors, the operating temperature must be maintained at or below 0°C. The temperature of the pixel silicon

194

sensors is determined by the heat load generated by the electronics attached to the sensors, the leakage currents in the sensors, the heat flow from the surrounding power and cooling services and heat flow from/to outside the pixel detector volume. Since the heat transfer efficiency to the surrounding environmental gas is very low, the temperature of the sensors is driven by the heat dissipation efficiency of the cooling structures. Therefore, the transverse thermo-conductivity of the local supports has to be measured. The carbon-carbon electrical resistance is used for the local support heating. The power applied to one stave shingle in the test is equal and simulates the typical power used by one detector module in operation. The cooling is being performed with turbulent water flux. The temperature control is done with a thermocamera. An example of a good stave thermo-image is shown on-fig. 7s.

FIf.7. Local supports tamo-conductivity test: a) the good stave; b) the "defective" stave.

The temperature structure on the image corresponds to the stave variable thickness and, therefore, the resistance. A stave internal structure defect has then been simulated: the thermal grease between the aluminium pipe and the carbon-carbon TMT has not been put for a length of ~1 cm. The temperature distribution along the defective stave (fig.7b) is significantly different from the regular one; the defect can be easily identified and the stave can be rejected at this early stage. 5. MODULE TESTS

The main quality factors of the pixel modules are: threshold and noise, both their mean values and their dispersion over the entire matrix, the number of dead pixels and the

195 stability of all these values during the operation. To check the operation and define the characteristics of each pixel module the following laboratory tests are performed: • Digital test: digital signals are injected after each pixel discriminator; this allows to verify the proper functioning of the read-out chain. • Analog test: analog signals are injected at each preamplifier input; this allows to measure the threshold and noise values. After this test the tuning of threshold values for every FE chip is done. • Test with source: 109Cd (22KeV y), 241 Am (60KeV y), 90Sr (|3) are used to verify the whole chain (sensor-readout) quality. The self-triggering capability of the electronics greatly simplifies this kind of test. About 20 functional pixel modules with slightly different electronics, flex and bump design have been assembled during the period 1998-2001. The threshold value at which it was possible to operate them varied between 3000e and 5000e, with sigma values of 150-300e. The noise value is in the range of 150e-300e. The modules were operated on the SPS (H8, 7t+ 180 GeV beam) at CERN, laboratory results have been confirmed including good stability. 6. CONCLUSIONS A new Pixel detector layout that allows installation independent from the rest of the ATLAS Inner detector has been procured. This layout has minor acceptance and resolution losses, but facilitates installation, repair and upgrades. The mechanical changes did not offset the global schedule. The local supports design did not need any change and their production will start soon. The sensors pre-production has been successfully completed. The laboratory and beam tests demonstrate the good performance and stable operational characteristics of pixel detector modules.

REFERENCES 1. ATLAS collaboration, ATLAS Pixel Detector TDR, CERN / LHCC / 98-13 (1998). L. Rossi, The ATLAS pixel detector. Nuclear Instruments and Methods in Physics Research Section A 435 (1999) n.l, 2, pp 80-90. F. Ragusa, Recent Developments in the ATLAS pixel detector. Nuclear Instruments and Methods in Physics Research Section A 447 (2000) n. 1, 2, pp 184-193. 2. G. Lindstrom et al, Radiation hard silicon detectors-developments by the RD48 (ROSE) collaboration. Nuclear Instruments and Methods in Physics Research Section A, 466 (2001), n. 2, pp 308-326. 3. R.Wunstorf for the ATLAS Pixel collaboration, Radiation tolerant sensors for the ATLAS Pixel detector. Nuclear Instruments and Methods in Physics Research Section A, 466 (2001), n. 2, pp 327-334. 4. ATLAS Pixel Local Supports Requirements, ATLAS Project Document ATL-IPEP-0005.

T H E ATLAS SILICON M I C R O S T R I P T R A C K E R C O N S T R U C T I O N STATUS D. F E R R E R E DPNC,

Universite

Representing

de Geneve, the ATLAS

Geneve, SCT

Switzerland

Collaboration

The Semiconductor Tracker (SCT) is based on a large area of silicon microstrip sensors inside t h e ATLAS inner detector system. About 4100 detector modules must be assembled and placed on 4 barrel cylinders and 18 forward disks. Most of the modules are composed of 4 single sided detectors mounted back to back with 40 mrad rotation allowing the reconstruction in the other direction. The sensors are read-out by a copper/Kapton multilayer hybrid which holds 12 binary read-out ASICs. The commands and data are transmitted via optical fibers. T h e SCT is designed to operate 10 years in the LHC and, all the active and passive module components must survive a high radiation level. A description and a current status of the silicon detectors and the SCT modules will be given. T h e electrical performances of the prototype modules will also be discussed.

1

Introduction

The Semiconductor Tracker (SCT) which is based on silicon microstrip sensors, will be a part of the inner tracking system of ATLAS 1 and is embedded in a 2 T field. The inner detector 2 is composed of 3 tracking systems: a semiconductor pixel detector, the SCT and the Transition Radiation Tracker (TRT). The design of the inner detector is motivated by the physics requirements of a good tracking performance for the reconstruction of secondary vertices, track impact parameters, track isolation and the measurement of the high momentum particles. The SCT detector consists of 4 barrel cylinders and 9 forward disks on each side surrounding the ATLAS interaction point. It allows the reconstruction of 4 space points inside a coverage of \TJ\ <2.5. The size of SCT is 5.6 meters times 1 meter diameter. The expected mean event rate is 23 interactions every 25 ns bunch-crossing for a design luminosity of 1034 cm~2s~1, with fluctuations about that value. The high bunch-crossing frequency requires a fast front-end electronics response and an efficient signal processing. Due to the high radiation environment, damage is induced into the electronics and the silicon detectors. They are required to operate with the specified performance over 10 years in the LHC. The maximum expected cumulative fluence is 2 x 10 14 1 MeV neutron equivalent per cm 2 .

196

197 2

The modules

The SCT module designs have been made to ensure hermeticity over all azimuth angles in the range \rj\ <2.5. Five varieties of module (Fig. 1) will be built: one barrel and 4 forward types (inner, middle, short-middle and outer). In addition to the good electrical functionality of the modules that is expected over 10 years of operation, all modules must satisfy the thermo-mechanical specifications. Up to 7 W on the electronics side plus 1 W on the detector side will be dissipated during operation. Since the detector leakage current is dependent on the temperature, the heat dissipation must be efficient enough to avoid thermal runaway 3 . A material of very good thermal conductivity (1700 W/mK) has been chosen for the detector baseboard or spine made of Thermo-Pyrolithic Graphite (TPG). Since this substrate is a soft material, components suchas BeO facing plates for the barrel and A1N bar-spines for the forward are used to have a sufficient module stiffness between the electronics and the detector parts.

Figure i. Pictures of a barrel (left) and an outer forward (right) module prototype.

The hybrid electronics has evolved independently for the barrel and the forward regions, allowing a specific design for each. The hybrids consist of a multilayer copper-kapton circuit that receives passive components and 6 ABCD readout chips on each side. A thermally conductive hybrid substrate (CarbonCarbon or carbon fiber material) is also used allowing the most important heat dissipation of the module. The cooling contact area of the barrel module is located on the hybrid connector side, whereas for the forward module it is shared between the hybrid and the detector part with an additional contact area at the far end detector side. The main electrical function of the hybrid is to read out the 768 strips on each side of the module. It provides various lines to the chips for the digital and the analogue power, the 40 MHz clock frequency, the commands and the data. In addition, the forward hybrid holds

198

Detector

Module type

Barrel W12 W21 W22 W31 W32

Barrel Inner Middle Middle Outer Outer

Length (mm) 64 61.06 65.085 54.435 65.54 57.515

Outer/Inner width (mm) 63.6 / 63.6 55.488 / 45.735 66.13 / 55.734 74.847 / 66.152 64.636 / 56.475 71.814 / 64.653

pitch 80 (/zm) 207 (/xrad) 207 (/urad) 207 (tirad) 161.5 (zxrad) 161.6 (lirad)

Table 1. The various dimensions of the silicon detectors.

the receiver and transmitter chips for the optical communication links. For the barrel module, the optical components are put on a separate circuit called the "dog-leg". The 4100 modules will be assembled at 11 centers with the aim of positioning the detectors with an internal alignment better than 5 /zm in-plane and 10 fim. for the front-to-back plane. This tight tolerance is essential to avoid the alignment correction within each module for the track reconstruction. 3

The silicon detectors

About 16000 silicon detectors (~60 m 2 ) will be used for the construction of SCT. The sensors are made of p + -implanted strips on an n-bulk 4 silicon substrate. The strips are AC coupled through a thin dielectric layer (~ 300 nm) and biased via polysilicon or implanted resistors of ~ 1 MQ. The detector edge region design depends on the manufacturer and allows high voltage operation up to 500V. The detector are on average 285 microns thick with a bulk resistivity specified to be in a range of 2 to 10 kfi.cm with a mean value that drives the initial depletion voltage around 60 to 80 Volts. Each detector is composed of 770 strips with the first and the last strip connected to ground. The detector dimensions, shapes and pitches vary depending on the module type to be assembled (see Table 1). In August 2000, the qualified manufacturers received authorisation from ATLAS for production release and they started with a production of pre-series detectors. The series production started in Spring 2001 with ~80% of the detectors made by Hamamatsu (Japan) and the remainder processed by CiS (Germany). About 8 months after the production started, more than 35% of the detectors were delivered. The detectors are fully tested by the manufacturers and all the characteristics are registered into the SCT production DB 5 . Once the detectors are delivered to the test centers, they are inspected visu-

199

ally under a microscope and the current behavior is systematically controlled. On a detector batch sample other specific tests suchas the full strip test (that allows strip defect identification) and the depletion voltage measurement are performed. The results are then compared to the manufacturer datasheet. The detector quality was found to be very satisfactory after some systematic tests in various SCT institutes. At this stage 1 to 3 % of the inspected detectors do not fulfill the SCT specifications mainly due to a visual defect or a leakage current excess. Post-irradiation behaviors The detector behavior is affected by high particle fluences. They cause ionization and bulk displacement damage. The effective doping concentration varies as the function of the fluence and the annealing. Above a certain particle fluence, there is a type inversion that occurs in the silicon6. This type inversion n —> p is typically observed around 5.xl0 1 3 proton/cm 2 for SCT silicon sensors with some dependence on the bulk resistivity. The excess of acceptor concentration is then unstable with time and temperature and above a certain annealing period it never stops increasing. Also the silicon detectors will operate at -7 °C to contain the reverse annealing effect. The maintenance of the SCT will impose a yearly warm-up that is equivalent to 21 days at 25 "C. The typical annealing that is performed on the irradiated detectors is 7 days at 25 °C which corresponds to ~ 50 V lower4 than at 21 days. The charge collection efficiency has been measured for several irradiated and annealed detectors (Fig. 2). 6 cm strip detectors were read out with SCT128A analogue chips running at 40 MHz. Using a Ru 106 (3—source, the plateau is above ~ 300 V.

Figure 2. Charge collection efficiency measured for Barrel, W31 and W32 detectors.

200

In addition to the change of the depletion voltage, the leakage current of the detectors is strongly deteriorated. The excess of leakage current is directly proportional to the particle fluence4. Also at the operating temperature of the silicon (-7 °C) and after 10 years of operation the leakage current is expected to be ~0.5 mA instead of several tens of nA at the start of SCT operation. 4

The module readout performance

The binary ABCD 7 readout chip is developed and manufactured in the radiation-hard DMILL process technology. The chip series production started in August 2001. The electrical performance of the chips has been largely studied on barrel and forward modules. Un-irradiated barrel modules showed an average Equivalent Noise Charge (ENC) of 1400e~ which is about 100e~ lower than the 12 cm strip forward modules. The inner forward module, due to its shorter strips, has a noise of 1250e _ . The calibration and noise are measured using internal charge injection. Also a threshold scan is performed at every injected charge and the median response of the "S-curve" (fitted with an error function) provides the mean response value. Some electrical instabilities at low threshold have been observed only on the forward modules and this has led to a revision of the hybrid design. Some of the modules have been irradiated at the CERN PS with 24 GeV/c protons up to 3 x l 0 1 4 p/cm 2 . Then the standard detector annealing is performed by leaving the module 7 days at 25 °C (section 3). On those irradiated barrel and forward modules, the measured ENC increased to ~ 1800-2000 e~. To complete the full electrical readout performance, the modules are also tested in the test beam and in the system test. Beam test Various module types have been evaluated in beam tests at CERN SPS H8 and at KEK PS 112. Both the August 2000 beam test at CERN and the December 2000 beam test at KEK showed satisfactory results 8 ' 9 in term of noise occupancy, median charge, efficiency and residuals. The modules were mounted inside a cooled box, which was placed between layers of a beam telescope made of 4 accurate XY silicon planes. A pair of scintillators triggers the readout of the ABCD chips after an appropriate pipeline delay. The modules are tested in several conditions, for example as a function of the detector bias, the detector angle and the value of the magnetic field. The indication of the median charge (Fig. 3), which is reconstructed by making threshold scans, shows that unirradiated modules reach the plateau above a detector bias voltage of 120V whereas for irradiated detectors the plateau

201

is obtained for all module types above 350V. This confirms the individual irradiated detector studies (section 3). At the plateau the signal-to-noise ratio is about 16:1 and 10:1 for the irradiated and non-irradiated modules. One of the most important design specification for the SCT module is the particle detection efficiency at 1 fC threshold that must be above 99% and where the noise occupancy must not exceed 5 x l 0 - 4 . The efficiency and the noise occupancy have largely been tested. Also at the bias detector plateau all of the module types work close to the specified values for non-irradiated and irradiated modules 10 .

0

100

200

300

400

500

Bias voltage [V]

Figure 3. Median charge versus bias voltage measured at perpendicular incidence in a 1.56T magnetic field. Measurements were made for unirradiated and irradiated barrel and forward modules (Module # 1 to # 6 ) .

System test Finally an electrical module must be stable when the readout is performed together with other modules in an environment that is close to the final assembled barrel and disk. Also a barrel sector and a disk sector have been equipped at CERN with prototype services like kapton power tapes, optical fibers, cooling pipes and blocks and an adequate mounting system. The noise performance and particularly the noise occupancy are tested for a group of mounted modules. The results are compared with those provided by the same modules previously tested on electrical stands. This test allowed optimizing the grounding scheme of the module and it has shown no particular degradation of the electrical performances. 5

Conclusions

Over the last few years, SCT qualified most of its active components. The series production of key items such as the silicon detectors and the binary readout chips has started. The silicon detector production and quality con-

trol are progressing well with 36% of the ordered detectors already delivered with a constant and good quality. The Barrel have been shown to have satisfactory electrical, thermal and mechanical performance. Ongoing forward module development is nearing this status. The unirradiated and irradiated modules up to the maximum expected particle fluence equivalent to 10 years of operation showed excellent results by doing individual test on a test stand, but also in the beam test and the system test. The module production inside various SCT institutes is about to start and the preparation of the assembly systems and the test systems are nearly complete. Acknowledgments The author acknowledges the support of the DPNC and the SCT collaboration in presenting a general status of ATLAS SCT at this conference. All his colleagues who contributed to the material that is presented here are particularly thanked. References 1. The ATLAS Technical Proposal, CERN/LHCC/94-43 (1993). 2. The ATLAS Inner Detector Technical Design Report, CERN/LHCC/9716 and CERN/LHCC/97-17 (1997). 3. C. Heush, A. Holodenko, H.G. Moser, Measurement of Silicon Detector Thermal Runaway, ATL-INDET-99-015. 4. P. Allport et al., ATLAS Irradiation studies of n-in-n and p-in-n silicon microstrip detectors, NIM A 435(1999) 74-79. 5. The SCT production data base, http://melb.unige.ch:3143/phyprdwww/sctprd/welcome.html 6. H.J. Ziock et al., IEEE Trans. Nucl. Sci. NS-40(4) (1993). 7. W. Dabrowski et al., Design and Performance of the ABCD Chip for the Binary Readout of Silicon StripDetectorsin the ATLAS Semiconductor Tracker, IEEE Trans.Nucl.Sci. Vol. 47, 1843-1850, 2000. 8. Y. Unno et al, Beamtest of Non-irradiated and Irradiated ATLAS SCT Microstrip Modules at KEK, IEEE 2001, procedings to be published. 9. M. Vos, proceedings of the 5th International Conference on Large Scale Applications and Radiation Hardness of Semiconductor Detectors, July 4-6, 2001, to be published. 10. G. Moorhead, proceedings of the 5th International Conference on Large Scale Applications and Radiation Hardness of Semiconductor Detectors, July 4-6, 2001, to be published.

T H E SILICON STRIP T R A C K E R OF T H E CMS EXPERIMENT M. BIASINI* Universitd degli Studi di Perugia, Dipartimento di Fisica, Via Pascoli 1-06127 Perugia, Italy E-mail: [email protected] The CMS experiment at the Large Hadron Collider will have a large Silicon Strip Tracker. The status of the project is here reviewed. The detector layout is presented, and some of the expected performances are discussed. The problem of radiation damage of silicon sensor is addressed, and the final choice for the silicon technology is shown. The construction of such a large scale detector requires an adequate organization, which is here discussed. Finally the present status of the Silicon Strip Tracker construction is presented.

1

Introduction

The wide range of physics accessible at the Large Hadron Collider (LHC) requires the presence of a powerful and robust tracking system. The CMS detector will have a Silicon Strip Tracker (SST) fully based on silicon detector technology 1 , z . The expectations for the tracking system are a good reconstruction efficiency for high transverse momentum particles, of the order of 95% for isolated tracks and 90% for tracks inside jets. The momentum resolution for high pt isolated tracks is expected to be of the order of few percent for tracks in the central region. In addition, other stringent requirements come from the necessity to operate in the harsh radiation environment of the LHC, from the request of minimum material in front of the calorimeters, and from the practical constraints of a large scale detector construction. 2

Detector Layout and Components

Starting from the inner part of the detector, the tracker consists of silicon pixel and silicon microstrip devices. Figure 1 shows a view of one fourth of the SST. The central region is covered by the Tracker Inner Barrel (TIB) and Tracker Outer Barrel (TOB), which aref complemented in the forwardbackward region by the Tracker Inner Disk (TID) and the Tracker EndCap (TEC). The detectors indicated in red are made of double sided modules, obtained from two single sided ones, mounted back-to-back. The modules *FOR THE CMS T R A C K E R COLLABORATION

203

204 0.1

0

200

0J

OJ

400

HA

600

0.5

MO

0.6

1000

0.7

0.8

1200

1400

0.9

1600

1

1800

1.1

2000

1.2

2200

1J

2400

14

2600

1.5

2800

Figure 1. Side view of one-quarter of the CMS Silicon Strip Tracker (dimensions in mm).

Table 1. Number of components for each subdetector of the Silicon Strip Tracker. Subdetector TIB TOB TID T E C (thin) T E C (thick)

Single/Double Modules 1188/768 3048/1080 240/288 1648/432 2448/720

Detectors (tot)

Wafers (tot)

N. of APV ch.

2724 5208 816 2512 3888

2724 10416 816 2512 7776

13968 24192 4416 30208

which are closer to the beam line (TIB, TIB and first four rings of TEC) are made of 320/xm (thin) sensors, while the others are equipped with 500/xm (thick) ones, in order to compensate for the increased noise due to the longer strip length. The front-end read-out chip is the APV25, a 128 channels chip, based on deep sub-micron technology. The number of modules, detectors, wafers and read-out channels are shown in Table 1 for each subdetector. 3

Expected Performances

Detailed simulations of the SST have been developed in order to estimate the properties and performances of the detector 3 . Great effort has been devoted in the choice of materials to reduce the material budget in front of the calorimeters. Figure 2 (left) shows the contributions of different parts of the detector to the total thickness in radiation lenghts. Different algorithms

205 iA

r AH Tracker

Figure 2. Left: Contribution of the different components of the SST to the material budget (Outside refers to material beyond the active volume of the tracker). Right: Transverse momentum resolution of the CMS TVacker for single muons.

have been developed to reconstruct the information from the SST; in Figure 2 (right), the expected resolution on transverse momentum for single muons is shown, confirming a few percent resolution for tracks up to 100 GeV. Several studies have been recently done on the CMS tracker capability to tag jets originating from b-quarks. Figure 3 shows the results of an algorithm based on a track-decay-length, which gives a very high discriminating power (left). By combining the information from all the tracks in a jet, an efficiency is obtained of 60% for 100 GeV b-jets in the barrel region, for a light quark jets rejection of about 100 (right). 4

T h e Choice of Silicon Sensor Technology

Radiation damage is one of the main concern in the investigation for the use of silicon strip detector at LHC; as an example, ten years of running at LHC will give a fluence in the Inner Tracker of 1.6 x 10 14 MeV equivalent neutrons/cm 2 . The effect of radiation damage is, on one side, related to surface effects, like the increase of interstrip capacitance which brings to increased noise, and, on the other one, to bulk effects, like the increase of the leakeage current and of the depletion voltage, and the decrease of charge collection. Extensive R&D

206

n.. ..

^S ;

\ ;

Decay

LDecay

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0. b-efflciency

Figure 3. B-Tagging performances of the CMS Silicon Strip Tracker: Track-decay-length significance for 100 GeV b-jets and u-jets in the barrel region (left), and u-jet rejection as a function of b-jet efficiency (right).

programs have been carried out in the past years 7 , in order to investigate the sensor type, the geometry and the substrate, by means of simulations, irradiations, electrical characterizations and beam-tests. The final choice for the SST sensor technology is a single-sided detector, with p-type strips on n-type substrate. The read-out strips have an integrated AC coupling, and the bias is done with polysilicon resistors. In order to reduce the effects of radiation damage, a low resistivity substrate (1.5-3.0 MQ cm) is used for the inner tracker, and the crystal orientation is <100>-type, in place of the usual <111>. In addition, a mask design has been chosen with overhanging of metal on top of the p-strips implant. Finally, this choice is compatible with an industrial production on 6" wafers. 5 5.1

\

The Silicon Strip Tracker construction Organization

The construction of such a large system (~ 220 m 2 ) based on silicon technology has never been experienced before. Since the 16000 silcon modules will be assembled and tested in different laboratories, strong effort is put to guarantee good quality, uniformity and stability in time for the module production.

207

Figure 4. Organization for the construction of the CMS Silicon Strip Tracker.

At the same time, extensive use will be done of deeply autornitized systems for sensor test, module assembly and testing. The following main steps are foreseen to happen in different labs after parts procurement 2 (see figure 4): 1) sensor quality assurance 4 ; 2) module assembly5; 3) bonding and testing; 4) integration on mechanics; 5) sub-detector assembly; 6) tracker assembly.

5.2

Status of the Construction

The construction of 200 modules of the CMS Silicon Strip Tracker is in progress (Milestone200)6. Silicon sensors from three companies have been received and sensor quality has been tested in the CMS laboratories, using the final procedure and specifications. Silicon modules are being assembled and bonded using the autornitized procedures. First modules have been completed (see figure 5), and preliminary tests both in the lab and on a test-beam indicate that they are inside specifications. Final tendering and contract signatures for the part production are underway, and the massive production of the CMS Silicon Strip Tracker will start soon.

208

Figure 5. Picture of a final module for the CMS Silicon Strip Tracker.

6

Conclusions

The tracking system of the CMS experiment will be fully based on silicon technology. The layout of the detector has been optimized in order to match the expected performances. The outstanding problem of the radiation damage at LHC has been extensively studied and the final choice of silicon technology ensures full functionality of the Silicon Strip Tracker for the lifetime of the experiment. A detailed organization has been developed for the construction of such a large scale detector, in order to guarantee a uniform quality and match the schedule. First final modules have been produced, and the massive construction of the Silicon Strip Tracker will start soon. References 1. CMS Tracker Technical Design Report, CERN/LHCC 98-6 CMS TDR 5 (1998). 2. Addendum to CMS Tracker TDR 5, CERN/LHCC 2000-016 (2000). 3. M. Lenzi, Nucl. Instrum. Methods A 473, 31-38 (2001). 4. F. Hartmann, submitted to Nucl. lustrum. Methods A , (2001). 5. M. Lenzi, these proceedings. 6. A. Dierlamm, these proceedings.

T H E CMS SILICON T R A C K E R A U T O M A T E D M O D U L E ASSEMBLY M. LENZI CERN, EP Division, 1211 Geneva 23, Switzerland On behalf of CMS Tracker Collaboration In December 1999 the CMS Tracker Collaboration decided to make its tracking system entirely based on silicon detectors . The new tracker layout requires the assembly of about 17000 silicon strip detector modules, each one composed of silicon sensors and readout hybrid glued onto a carbon fibre frame. To guarantee the assembly of such a large quantity of modules with high quality but reduced manpower, an automated system has been developed at CERN. The system setup is based on a robotic machine, Aerotech AGS 10000 Cartesian Gantry Positioning System, equipped with high-precision positioning motors. The design of such automated silicon module assembly system is described and the performance in terms of positioning precision and assembly rate is shown.

1

Introduction

The CMS Tracker is composed of about 17000 silicon detector modules to be assembled in a few years with high and reproducible quality. A silicon module is composed of silicon sensors and read-out hybrid glued onto a carbon fibre support frame. In the outer CMS Tracker each module is composed of 2 different sensors whose strips are daisy-chained together by means of ultrasonic wire bonding 2 . To guarantee the assembly of such a large module quantity with high precision but reduced manpower, an automated silicon module assembly system has been developed at CERN based on a high-precision robotic positioning machine manufactured by Aerotech 3 . The pilot project has successfully demonstrated the feasibility of such a high quality automated module assembly and consequently 6 assembly centres, 4 in Europe (Bari, Perugia, Lyon, Brussels) and 2 in USA (Fermilab), are now equipped with similar systems and are almost ready to start module construction. The job to be performed by the automated system consists of the application of glue on the carbon fibre frame followed by pick-up and placement of silicon sensors and read-out hybrids onto the frame. The module layout has been designed to be as simple as possible in order to allow a fully automatic assembly 2 .

209

210

The correct positioning of sensors and hybrids is ensured by fiducial marks on the individual module components and identified by means of a pattern recognition system using video images from a CCD camera. A photograph of the CERN gantry setup is shown in Fig. 1; the supply and

Figure 1. Closeup view of the gantry setup.

assembly platforms are also visible under the gantry working area. These platforms use vacuum to hold the module components in place during the assembly and they allow to build up to 3 modules at the same time. A glue dispensing system has been designed using air pressure to allow automatic application of glues. The vacuum and air pressure valves are interfaced through a custom designed logic circuit which is controlled by the robotic positioning machine.

2

Pilot project g a n t r y s e t u p

The gantry system setup is based on a robotic positioning machine, Aerotech AGS 10000 Cartesian Gantry Positioning System, equipped with highprecision positioning motors that allow movement along 4 coordinates (X, Y, Z and # rotation). The X and Y drives are linear motors with linear encoders which provide high speed, acceleration and accuracy. In order to achieve more precision and reproducibility, the X axis drive consists of two identical linear motors mounted on two parallel rails. The Y axis drive is based on a single rail that spans the two rail of the X axis. The $ axis drive is mounted on the Z axis which in turn is mounted on the Y axis. The Z and $ axis drives can move respectively

211

100 mm and 360° with low speed but high precision. In order to build silicon modules, the Aerotech gantry system has been equipped with several additional mechanical devices, most of which have been designed and constructed at CERN. In the 50cm x 50cm working area under the gantry, one supply and one assembly platform have been located. The former has individual built-in vacuum chucks to hold in place the silicon sensors prior to their survey-mark scan and pick-up. Similarly, the assembly platform supports the frames onto which the silicon sensors and read-out hybrids are mounted. It also holds the hybrids in place prior to assembly. All the module components are held in place by individually controlled vacuum chucks equipped with sensors to detect vacuum leaks during the assembly process. In addition, a tool support head was designed to pick up and hold the took needed to pick up sensors and hybrids and to dispense the glue. This head is mounted on the # axis stage and uses vacuum to pick up and hold tools in place. A vacuum feed-through system provides the vacuum to pick up objects with the tool. A closeup view of the tool head is shown in Fig. 2. Two different tools are used to pick-up and move respectively sensors and

Figure 2. Tool support head to pick up and hold the interchangeable pick-up and glue dispensing tools.

hybrids. The latter has vacuum through 3 suction pads which come in contact with the hybrids, while the proper handling of the very delicate sensors during the pick and place operations is obtained by means of a flat teflon-coated vac-

212

uum tool. The sensor and hybrid pick-up tools are shown in Fig. 3 together with the glue dispenser tool. During the pick-up procedure, the contact between the tool and the module

Figure 3. Sensor and hybrid pick-up tools and glue dispensing tool.

component to be moved is detected through built-in pressure sensors located in the tool head, thus preventing excessively large forces from being applied to the components. The glue dispensing tool is held by the same tool holder support and has a separate air pressure feed-through connection with its own vacuum line to hold it in place. This tool is designed to hold a standard plastic syringe which fits over an o-ring sealed nozzle and is locked in place with a screw type fitting. All the tools are stored in a tool rack when not used. The rack is located at the back of the gantry base plate behind the sensor platform and is moved forward and backward on air pressure pistons so that it can be displaced out of the way during gantry motions. Finally, the Z drive is equipped with a CCD camera to sight fiducial marks on the individual module components. The gantry assembly system should be located in a clean room of class 10000 and with good temperature control. This is needed for ensuring the precision accuracy of the gantry machine. Humidity control is also required during the glue curing phase. 3

P a t t e r n Recognition

In order to ensure a proper alignment of the module components on the carbon fiber frame an accurate measurement of their initial position is needed.

213

For this reason precise reference markers are designed both on sensors and hybrids; their positions are localised automatically and with excellent precision during the assembly process through a pattern recognition system. This system is also used to measure the position of the assembly platform and thus the precise location of the carbon fiber frames that are fixed with respect to the assembly platform. The same system is utilised to re-measure the locations of the sensor and hybrid marks after the module assembly is terminated and therefore allows for verification of a correct assembly process. The pattern recognition system consists of a MATROX CORONA LC frame grabber card, a colour CCD Camera and an high magnification microscope optics (200x). The LabView environment has been used to develop applications and implement an user interface and the algorithms implemented in the IMAQ Vision Library were used to define and find patterns. The pattern recognition strategy consists of searching for a pre-defined pattern (the marker) within the camera image; the results of the fit are the marker position and the matching score which indicates the quality of the fit. The measurement takes approximately 1.5 seconds for each marker. For our given optics most of the uncertainty is due to noise in the image from the video camera. Additional sources are dust or dirt on the surface and different illumination condition due to tilting of the surface. The RMS position resolution in one dimension achieved measuring sensor markers in realistic condition is about 0.4 /mi, as shown in Fig. 4.

Figure 4. Pattern recognition resolution for sensor markers.

214

4

Calibration

The gantry specifications claim an absolute position accuracy of a few microns. This goal can be achieved only by means of software corrections of the machine positioning which use a 2D calibration of the X and Y axis. To perform the 2D axis calibration a 50cm x 50cm plate made of glass with low thermal expansion coefficient has been built. A thin film with a grid of circular markers with a grid spacing of 2cm has been glued on the glass plate. The precise measurements of the positions of the markers have been first made at the metrology laboratory at CERN. Then the plate has been put on the gantry working area and aligned with the machine axis; the marker positions have then been measured using the pattern recognition program. The two set of measurements are then compared and the difference between the position coordinate of each marker on the plate has been used to determine the 2D calibration file. Finally, the marker positions have been measured using these corrections to verify the effectiveness of the calibration procedure. The difference between the "true" position obtained by the metrology measurements and the position measured at the gantry are shown in Fig. 5 respectively before and after the calibration procedure. The results are shown for the coordinate orthogonal to the strips, which is the more critical parameter for the sensor alignment issue. Before the calibration the gantry measurement

Figure 5. Difference between t h e marker position on the calibration platform measured at the metrology or at the gantry machine as a function of the real position. The measurement has been performed respectively before (up) and after (bottom) the calibration of the gantry machine.

could be off with respect to the true position by as much as 100 /j,m with a

215

gradient of up to 500/xm/cm, while after the calibration the absolute position accuracy is about 1/xm. 5

The assembly procedure

The Aerotech gantry system includes a software interface, called MMI (Man Machine Interface), and program libraries that allow control of the machine either manually or automatically with a custom program. In addition, the communication with external hardware is provided via digital I/O channels. In particular 56 digital inputs and 56 digital outputs can be set and read remotely by a MMI program. These features have made possible the full automatization of the module assembly procedure by implementing an MMI program that execute all the actions needed to build modules without requiring the intervention of the operator. In the assembly program the digital outputs are used to switch the vacuum valves and air pressure valves while the digital inputs are used to read the status of the vacuum and contact sensors. The assembly process consist of operator actions as well as automatic actions performed by the assembly program. The operator actions are needed to prepare and load the module components and assembly tools, in particular: • initialisation of equipment; • load carbon fiber frames and hybrids on assembly platform; load sensors on supply platform; • place platforms on the gantry working area and lock them into place; • prepare the glues, fill the syringes and load them on the tool rack; • run the automatic assembly program; • remove assembly platform while keeping on vacuum for glue curing. Once the preparatory work has been performed, the assembly program proceeds to the automatic building of the modules. The main assembly program actions are: • measure fiducial marks on assembly platform, sensors and hybrids using pattern recognition; • dispense conductive epoxy glue on the proper pads on the frame kapton foil circuit to supply the bias voltage to the sensor backplane;

216

• dispense silicone glue on the carbon fiber frame for sensor fixing and thermistor contact; • pick-up the sensor, rotate it by the proper angle to align to the carbon fiber frame and place it on the support frame; • dispense araldite glue on support frames to fix hybrids; • pick-up, rotate and place hybrids; • re-measure sensor and hybrid position using pattern recognition to verify the assembly process. All the relevant information about the assembly of each module (environment conditions such temperature and humidity, alignment result obtained by the final marker survey, faults occurring during the assembly) are registered during the module construction process and successively stored in the general tracker construction database. 6

Performance

The performance of the automated module assembly has been evaluated in terms of accuracy and reproducibility of component placement. The precision of X and Y movements has been measured by picking up a silicon sensor, moving it of a fixed distance and putting it down. The positions of the sensor markers before and after the movement are then compared; the results show a position resolution of about lfim. The more critical point in terms of precision turned out to be the sensor rotation, needed to align it to the carbon fibre support frame. A test has then been performed consisting in pick-up a sensor, rotate it of a fixed angle and put it down. Measuring the position of two sensor markers before and after the rotation shows that the rotation accuracy is ±0.002 degrees while the resolution is about 0.0005 degrees, as shown in Fig. 6. The precision of the full assembly process has been measured after several trial assemblies. The sensor to sensor alignment precision is about 3/xm; the results are shown in Fig. 7 for the coordinate orthogonal to the strips for which more accuracy is required, but similar results have been obtained for the other coordinate. The absolute positioning of each sensor with respect to the module frame positioning pins is within 3/zm, after the calibration procedure. Finally, for what concerns the relative position between the hybrid and the sensors, it turns out to be well within the requirement of 20/xm.

217 c

35

1

30

cr=0.51mdeg

25 20 15 10

I I

" jI \

1

'

Jf. \

-5 -4 -3 -2 -1

....

12 3 4 5 (mdeg)

Figure 6. Rotation angle resolution

a '

Mean RMS

-0.8000 3.040

•c « c ^

I

; iI 1

-20

-15

-10

10

15

20

Ax Figure 7. Relative displacement after the module assembly between the two sensor in the direction ortoghonal to the strips, which is the one where more precision is required.

For what concerns the assembly rate, the system allows to build up to 3 modules in 25 minutes. On average, about 25% of the time is spent to measure the positions of the component's fiducial markers by using the pattern recognition procedure both before the assembly, for a precise location of components, and in the final survey to evaluate the assembly quality. The dispensing of the three different glues takes approximately 45% of the total assembly time. The remaining 30% of the time is spent to pick-up and place

218

sensors or hybrids.

7

Conclusions

The decision of an all-silicon layout for the CMS Tracker implies to assemble a very large number of silicon modules in a short time. For this reason, a pilot project has been developed at CERN in order to automate the silicon module assembly process. Starting from an Aerotech high precision robotic positioning machine, this project consisted of designing and building several mechanical devices such as working platforms, pick-up tools, an automatic glue dispensing system and a vacuum system. In addition a pattern recognition program has been implemented in order to localise the position of components to be assembled and a software program has been written to fully automate the assembly process. The final system is able to build up to 3 modules in 25 minutes with a sensor to sensor relative displacement within 3^m. These results have successfully demonstrated the feasibility of automated module assembly with high precision but reduced manpower. The successful outcome of this project has lead to the decision to equip further 6 assembly centres, both in Europe and USA, with a similar setup to share the assembly tasks during the production phase of the CMS silicon tracker. The final goal is to produce 17000 modules in about 2 years. Acknowledgments I wish to thank Alan Honma and Jean-Claude Labbe, who made possible the successful achievement of the automated silicon module assembly project. References 1. CMS Tracker Collaboration, Addendum to the CMS Tracker TDR, CERN/LHCC 2000-016. 2. The Tracker Project, Technical Design Report, CERN/LHCC 98-6 CMS TDR 5, 15 April 1998. 3. Aerotech, Inc., 101 Zeta Drive Pittsburgh, PA 15238-2897

CMS SILICON T R A C K E R - MILESTONE 200 A. D I E R L A M M Institut

fur Experimentelle E-mail:

O N BEHALF OF THE CMS

Kernphysik, Universitat [email protected] SILICON T R A C K E R

Karlsruhe

(TH)

COLLABORATION

The tracker of C M S 1 ' 2 will fully consist of silicon micro-strip and pixel sensors. Building a detector with 210 m 2 sensor surface in about 3 years requires a tightly controlled construction schedule. All different aspects of the production are exercised within a pre-production of 200 modules (Milestone 200) to identify and eliminate possible bottlenecks and to test the complete electronic chain. The quality, process stability and radiation hardness of the silicon sensors will be permanently monitored. Automatic assembly procedure and industrial bonding machines will guarantee a fast and reliable construction. All modules will be tested for signal, noise and pedestals at room temperature and operation temperature of -10°C. Quality assurance of the Milestone 200 sensors and modules including irradiation and stability tests are presented.

1

Introduction

The CMS tracker implements 25000 silicon strip sensors (up to «10cmx 10cm) covering a total area of 210 m 2 . Close to ten million read-out channels are connected to 75000 read-out chips (APV25 with 128 channels each). This large detector system will be realized, applying quasi-industrialized methods in production and quality assurance. The basic modules consist of up to two sensors, a read-out hybrid and the support frame. For modules placed within a distance of 60 cm from the beam line 320 /xm thick sensors are used, which will be produced by Hamamatsu Photonics (Japan). Most of the sensors will be placed between 60 cm and 120 cm; they have longer strips and in order to improve S/N the thickness will be 500 /zm. They will be produced by ST Microelectronics (Italy). Both producers use 6"-technology. The production procedure 3 is foreseen as follows. The sensors are qualified by the companies and the good ones are sent to the Control & Distribution Center (CERN). There they are registered in the database and distributed to four Quality Test Centers (QTC) in Karlsruhe, Perugia, Pisa and Vienna. These centers perform optical inspections and electrical tests of the sensors. During Milestone 200 all sensors are qualified, but during production, tests will be done on 5-10 % of the sensors. The QTCs select few sensors and test structures to send them to the Irradiation Qualification Centers (IQC) in Karlsruhe or Louvain and to the Process Qualification Centers (PQC) in

219

220

Florence, Strasbourg or Vienna for deeper investigations. All sensors accepted by the QTC are shipped to seven Module Assembly Centers using fully automatized robots to assemble the modules. For bonding these modules are sent to 12 Bonding Centers equipped with industrial machines. These centers are also responsible for module quality assurance. Finally, burn-in tests of modules and sub-systems will be performed at several centers. During production of the first 200 modules the different test centers are calibrated by shipping the same sensors and modules around and comparing measurement results. Further purpose of the Milestone 200 is to verify the logistics as described above and to qualify the materials. After all one Tracker Outer Barrel (TOB) rod and one Tracker End Cap (TEC) segment will be built to qualify design and to check functionality during a full sub-system test.

1IM@300V |I„@4S0V a 30-^

100 200 300 400 500 600

2000 3000

Leakage Current [nA]

<0

•5 10 5

0-I— 450

,ES1 475

uvj

525

Coupling Capacitance [pF]

0

1

2 3 4 5 6 7 8 No. of Bad Strips / Sensor

9

10

Figure 1. Statistics on 125 HPK sensors measured in Karlsruhe and Vienna. All sensors are within the acceptance criteria: Iieak < 10 fiA, Rpoiy = (1.7±0.5) MCI, Cc = (500±100) pF, Bad strips < 1 %

221

1000

I

I

2.3 •

2.2 100

':

10

9

1

I

0.1 0.01

ff2.1

f

•

—o— Before Irradiation —•— After Irradiation and Annealing (Measurement at-10*C)

'

VoM/VaVAVnMA^A^ 10

15

20 25 30 Strip number

-1—•—i—•—i—<—i—

."%A<^~V«^*"V~^%».~,,s~.s»A~'>*A/"~

2.0

i I 1.9 I 1.8

f 1,

- Before Irradiation - Raw data after Irradiation and Annealing (Measurement at -10'C) -Corrected due to high leakage current

1.6 35

40

45

50

10 1.5

15 20 25 30 35 40 45 50 55 Strip number

Figure 2. Increase of leakage current after irradiation has an effect on the measured bias resistance, which must be corrected according to the formula Rpoly — —u— , • ft

2

+IUak

Quality Assurance

In Karlsruhe and Vienna all 160 HPK sensors for the Milestone 200 have been tested (see Figure 1) with the result that all sensors fulfill the acceptance criteria 4 ! There are only 16 pinhole in 160 sensors corresponding to a fraction of 0.02 % bad strips! Sensors from ST Microelectronics are being qualified at Perugia and Pisa. The IQC in Karlsruhe performed a proton irradiation of one HPK sensor, one ST sensor and some HPK test structures with biasing. They were kept at low humidity and temperatures below -10°C during irradiation and storage. Characterization of the irradiated sensors of ST and HPK showed no changes in the strip parameters such as: bias resistance, cou120 115 IT 110 f 105 1 100 1 95 3. 90 3 85 f 80 U 75 8 70 65 60

'^l - Before Irradiation - After Irradiation and Annealing (Measurement at -10'C) 10

15

20

25

30

Strip number

35

40

45

50

200

400

600

Bias Voltage [VJ

Figure 3. Coupling and inter-strip capacitances level off after irradiation with a fluence of 1.8 x 10 1 4 p(33MeV)/cm 2 (s»18 LHC years).

222

1x10" 2x10" 3x10 eq. Fluence [nflMeVycm2]

Figure 4. Full depletion voltages can be calculated by the so called "Hamburg model" 5 . The default parameters agree with our measurements, as shown on the left hand side. The HPK sensor got the fluence expected for 320 /im material! Therefore UFD ~ d2 will be scaled down to 410 V. On the right hand side a 3d-plot of the full depletion voltage as function of fluence and annealing time is shown for 500 fim and 4.4 kficm material.

pling capacitance and inter-strip capacitance (see Figures 2 and 3) after a fluence of 1.8 x 10 14 p(33MeV)/cm 2 and beneficial annealing (80 min, 60°C). Full depletion voltage can be calculated as function of fluence and annealing time using the "Hamburg model" 5 . The predicted values were found to be in good agreement with the measurements on the large sensors, both from ST and Hamamatsu, as shown in Figure 4. Further predictions can be derived by using the extracted parameters. Both sensors with a thickness of 500 fim

^ S Statistics of tongtirM currant n=66. msan=18.67nA

.

50

P«.

1

•*»!

,

,

,

,

, j ^ * *

100 150 200 250 300 350 Standard Deviation of l(t) [nA]

Figure 5. Long-term test: Measurement of leakage current vs. time at room temperature and below 30 % humidity applying 400 V. On the right side one can see a statistic plot of a(heak(t)) at 400 V.

223

have a break-down voltage above 1000 V after irradiation. The PQCs performed a stability test on non-irradiated sensors by measuring the leakage current over a longer period. At room temperature and humidity below 30 % a stable leakage current (a(I(t)) < 20 nA) for 99 % of the sensors was measured at 400 V over four days. Some of these curves are shown in Figure 5. All the other process acceptance tests (IV, CV, C;nt, Rinti Rpoiy, RAI, R-p+, Cc, Idiel, Vbreak) are performed on a fraction of test structures and showed good results. 3

Assembly and Qualification of Modules

Module assembly has just started using the full construction chain. Sensors and hybrids are glued and placed on the frame structure by automatized robots 6 . In the laboratory a fully equipped basic TOB module showed <2 ADC counts (RMS of 0.4) noise per strip (~1500e~) and pedestals of approximately 300 (RMS of 6). In a test beam with 200 GeV/c muons the modules showed signal to noise ratio of about 16 in deconvolution mode and ~28 in peak mode. 4

Outlook

The basic parts of the modules (sensors, APVs and hybrids) have good quality and we are confident to assemble good working final modules. This was confirmed by beam test results. In the near future some modules will be submitted to the IQCs for irradiation qualification. The next step is to perform a system test on a 6 module TOB rod and a 20 module TEC assembly. The Milestone 200 will be fulfilled by spring 2002. References 1. CMS Tracker TDR, CERN/LHCC 98-6 CMS TDR 5 (1998) 2. M. Lenzi, "Performance of the all-silicon CMS Tracker", Nucl. Instrum. Methods A 473, 31-38 (2001) 3. F. Hartmann,"The CMS all-silicon tracker, strategies to ensure a high quality and radiation hard silicon detector" accepted by Nucl. Instrum. Methods A , (2001) 4. Addendum to CMS Tracker TDR 5, CERN/LHCC 2000-016 (2000) 5. M. Moll, Ph.D. Thesis, DESY-THESIS-1999-040 (1999) 6. M. Lenzi, "Automatic module assembly of CMS silicon detectors", these proceedings

T E S T OF T H E CMS SILICON STRIP D E T E C T O R S I N T H E HADRON BEAM V. ZHUKOV* UIA Antwerpen T. BAUER, M. FR1EDL, J. HRUBEC, M. KRAMMER, M. PERNICKA HEPHY Vienna D. CREANZA, V. RADICCI INFN Bari A. DIERLAMM, G. DIRKES, E. GRIGORIEV, F. HARTMANN, S. HEIER, F. RODERER IEKP Universitat Karlsruhe (TH) D. BISELLO, A. KAMINSKY, G. MARSEGUERRA, D. PANTANO, I. STAVITSKI INFN Padova M. ANGARANO, M. BIASINI, G.M.BILEI, S. CECCHETTI, L. FANO, M. GIORGI, V. POSTOLACHE, L. SERVOLI INFN Perugia CMS silicon microstrip detectors of different types equipped with the APV readout chips have been exposed to a high intensity 350 MeV/c pion beam. We study the performance of irradiated and non-irradiated silicon sensors as well as the readout chip behavior. Maximum signal to noise for the irradiated oxygenated sensor has reached 15 in deconvolution mode.

1

Introduction

The Silicon Strip Tracker is an important part of the CMS experiment at the LHC collider 1 . More than 15000 detector modules provide precision tracking with at least eight hits for each high pt track. The track reconstruction efficiency depends on the signal-to-noise ratio (SNR) which should exceed eight to assure an efficiency above 98% 3 . Although this SNR is easily achieved for non-irradiated detectors it may significantly degrade while exposing detectors to the equivalent fluence of about 2. 10 14 n/cm 2 expected after 10 years of the LHC operation at the innermost radius of the Silicon Tracker. •ON LEAVE OF ABSENCE FROM MSU,MOSCOW

224

225 Table 1. Detector specifications.

det. PD26 BA1 BA2 PD27 VB25 PD1 PD4

sensors Micron CSEM CSEM Micron HPK Micron Micron

crystal <100> <100> <111 > <100> <100> <100> <100>

p(k£lcm) 2 2.5 6 2 6 6 1.4

Ox OX OX -

F.n/cm* 10 14 10 14 2 • 10 14

readout lxAPV25 lxAPV25 lxAPV25 lxAPV25 3xAPV25 2xAPV6 2xAPV6

We have studied the performance of silicon detectors produced by different technologies, irradiated and non-irradiated, equipped with APV6 and newer APV25 readout chips 4 exposed to a 350 MeV/c pion or proton beams at Paul Scherrer Institute cyclotron facility (PSI). The quasi-continuous pion beam with its 50 MHz structure is very close to the conditions expected in the LHC where bunch crossings occur at 40 MHz.

2 2.1

Detector modules Sensors

The sensor specifications are summarized in Table 1. The silicon wafers of n-type and a thickness of 300 fim were processed by Micron and CSEM with p + strips which are ~ 18 /jm wide, 60 mm long and have a 61 ^m pitch. Larger sensors(VB25) were produced at Hamamatsu on 320 ^m thick wafers with 35 fim wide, 80 mm long strips and 140 /xm pitch. All sensors were AC coupled and the bias was delivered through ~ 2 Mfi polysilicon resistors. The wafers had different crystal orientations < 111 > or < 100 > and a bulk resistivity between from 1.4 and 7kficm. Some sensors were produced using the oxygenation (OX) technique which allows a reduction of the operating voltage for sensors irradiated with charged particles 2 . The oxygenation was performed from a local oxygen layer grown into the bulk at 1200° C during 100 hours, resulting in an oxygen concentration of about 3 • 10 17 cm~ 3 . Selected detectors were pre-irradiated by 25 . . . 34 MeV protons to an equivalent fluence ofl...2-1014n(lMeV)/cm2.

226

2.2

Readout chips

Two successive generations of the APV chip were used to read out the sensors:!) APV6 manifactured in a 1.2//m radiation hard CMOS process and 2) APV25 made in commercial 0.25 pm deep sub-micron CMOS, which provides intrinsic radiation tolerance 4 . The 128 channels APV chip consists of a preamplifier followed by a CRRC shaper with a time constant of r = 50 ns. The shaper output is sampled at 40 MHz and stored in an analog pipeline of 192(APV25) cells. When a trigger arrives, three consecutive samples are processed by a switched capacitor filter which performs a deconvolution algorithm that narrows each pulse down to one single clock cycle in order to identify the exact bunch crossing. The predominant noise source in the detector and readout system is the preamplifier input transistor of the preamplifier. The measured noise in deconvolution mode for APV6 is ENC=1000 + 46 p F - 1 and 400 + 60 p F - 1 for APV25. Other noise contributions such as leakage current, strip resistance or bias resistor together account for about 430 e - which is quadratically added 1. With a capacitive load of 16 pF, which is typical for CMS detectors, we expect ENC = 1426 e~ with the APV25 in deconvolution mode, which is in good agreement with the measured value of 1430e~. One to three readout chips (see Table.l) were mounted on hybrids assembled with a pitch adapter on the detector frame carrying two daisy-chained sensors. Thus the total strip length was 12 cm (16cm for VB25).

3

Experimental setup

All detector modules under study were housed in a cooling box which was operated by two water cooled peltier elements with a total cooling power of ~ 200 W at A T ~ 40° C (Thox = - 2 0 ° C). The box was flushed with dry nitrogen to prevent water condensation. The tests were performed in the PSI 7rMl beam line which provides 350MeV/c pions or protons with a rate of up to 9 kHz/mm 2 and a beam spot of approximately 50 x 50 mm 2 FWHM. In the present study we operated the detectors mostly with pions. Although the 350 MeV/c pions are approximately minimum ionizing particles (MIPS), secondary reactions can produce heavily ionizing particles (HIPS) with up to 1000 times larger ionization losses.

227

4 4-1

Results Front-end electronics

Irradiation may cause two kinds of damages to the readout electronics: permanent or transient. Due to its deep sub-micron fabrication process, the APV25 should be intrinsically radiation tolerant. Eight APV25S1 chips were irradiated to 1.87 • 10147r/cm2 at 300MeV/c momentum 5 . No critical, irreversible damage was observed. The irradiation did not affect the calibration signal SNR within ±5%. We have observed a dependance of the calibration SNR on the temperature of about ~ 25% for AT=30°C. This dependance was expected and is due to variation of the chip settings which have to be optimized for each temperature. The charge released by HIPS can result in the nipping of an APV register cell, called a Single Event Upset (SEU). We measured a total cross section of approximately 2 • 10~ 12 cm 2 for such SEU, corresponding to 2 • 10~ 15cm2 for a single flip-flop cell. Single event upsets also occur in the analog circuitry, but are self-repairing and appear as a negligible increase in noise background. 4-2

Detectors behavior

The Signal to Noise denned as „ " ' ' " ' " • , where Noiseciuster=y *^ , . '-, was analyzed from runs collected in the 350 MeV/c pion beam, at a rate below 100Hz/mm 2 . Figure 1 left presents the most probable (MP) SNRcius values as a function of Vbias- The averaged rms noise at Tb ox = —10° C is independent from the Vuas within 5%. The highest SNR of about 16.5 was achieved with the non-irradiated PD26 detector. As expected, the APV25 readout provides an ~30% increase of the SNR due to lower noise compared to the APV6. The significant difference in the signal obtained for the irradiated detectors BA2 and PD27 can be explained by different production technologies. The BA2 was produced with < 111 > crystal orientation, while the PD27 was manufactured from < 100 > silicon using the oxygenation technique 6 . The irradiated oxygenated sensor (PD27) demonstrated almost the same SNR as the non-irradiated (PD26) at Vbias = 550 V while the non-oxygenated irradiated sensor (BA2) had lost ~ 20% of the signal in comparison with the non-irradiated one (BAl). For the irradiated sensors the maximum SNR was achieved at Vbias ~ 3 times above the expected Vd ep i et j on . Note that all irradiated detector modules have shown stable operation at a bias voltage of 550 V.

228 •

' •

*****

" . : - . • "

m

» ."1 •

A

* A * * T m * 200

BAImirr(atT=I8C) PD26nirr. BA2irr. PD27iir. VB25nirr. PDImirr. HWfrr.

X

•

•

m

1 • A

A

* •

** • m• •

A

A

A

„ A PD26mrr. M BA2irr.

•

PD27UT.

300

Figure 1. Left is a SNR in deconvolution mode versus Vbias at T=—10° C (except BA1). On the right is a cluster size versus Vbias m deconvolution mode.

The cluster size (figure 1 right) is larger for irradiated detectors presumably due to an under-depleted zone near the p + strips and charge trapping. The dependence of the cluster size on Vbias follows the dependence of the interstrip capacitance Cj n t . For irradiated sensors, Ci nt decreases with Vbias, especially for the < 111 > crystals, while for non-irradiated sensors, Ci„t remains constant and for irradiated < 100 > crystals, the dependence on Vbias is weaker. Due to the absence of precise independent tracking and large multiple scattering we could not measure the absolute efficiency. However we can roughly estimate the knee of the efficiency plateau by using the detectors under test as a tracker, see figure 2 left. To study the SNR uniformity the beam spot has been moved along the strips. No variation in the SNR has been found for the largest VB25 detector in a beam spot scan from one end to the opposite end. The measured variation of the SNR across the sensor was below 2.5% for all detectors. Unlike our measurements at PSI, where the beam incidence was usually perpendicular to the detector plane, a wide-spread angular distribution is expected in CMS l. An angular scan has been performed with the 140 ^m pitch VB25 module at room temperature. In figure 2 the maximum probability signals and the cluster sizes are presented versus the incident angle a to

229 APV2S angla Man - signal dustar width Decomokition Mods.Standard MtHng*. 360 MeV/c protoru and pkxil • A 0.9S

Pretoria- duster Width (measured) Pretoria - Fit sqrtfconsP+tan9) F ^ s - C l i i s t e f Width (measured) Piona - Fit aqrt(conet*+tan*)

S .-' _.-•''

. - - * • " * * " %

0.9

-

0.8S

-

/ a

•

20

so

«

so

eo

Taanglean

».»

APV25 angla scan - duatar algnal i. Standard MltkigK 350 MeV/c prolont and plant

0.7S

-

0.7

-

•

BAlirrad.

Pretoria - MP SNR (meaaurad) • Protona - Fit 1/coa Piona - MP SNR (meaaurad) Rons - Fit 1/cos

*,/' t,'"'

•"a^"'"

0.65

1 , . . 1

0

2

4

t

8

10

12 14 SNK

10

20

30

40

50

60

Tlangbnll

Figure 2. Efficiency versus MP SNR for the BA2 detector operated at T=—10° C in deconvolution mode on the left. On the right cluster size (top) and most probable signal (bottom) of the VB25 detector module versus incident angle for 350 MeV/c pion and proton beams in deconvolution

the detector plane for pion and proton beams at a momentum of 350MeV/c. The signal dependence is described with l/cos(a) and the cluster size, with y'c 2 + tan2 [a) functions, where c denotes the cluster width at perpendicular incidence. The ratio of SNRP is about 6.6 and thus in good agreement with the calculation from the restricted Bethe-Bloch theory which predicts 6.0. The detectors have demonstarted a stable operation in the high intensity pion beam at 9 kHz/mm 2 , the leakage current have increased with fluence Ileak ~ $OL, a — 8 1 0 ~ 1 7 A / c m .

5

Summary

Different type of silicon strip detectors were tested in a hadron beam under conditions close to what is expected at the LHC. With a strip length of 12 cm and irradiated oxygenated sensors, we have obtained signal-to-noise value of 15.5 for the APV25 and 10 for the APV6 readout chips in deconvolution mode. The signal-to-noise is uniform along and across the strips within a

230

level of 2.5%. It has been shown that the sensors and the readout chips do survive in the harsh radiation environment of LHC. No critical damage could be observed on the readout chips, and the single event upset rates are sufficiently low so they will cause only negligible corruption of data. In an angle scan, the detector modules behaved as expected from geometrical relations and the measured signals were consistent with the restricted Bethe-Bloch theory for pions and protons. 6

Acknowledgments

We would like to thank K.GabathuIer and D.Renker for their help at PSI and L.Shektman for sharing the beamline with us. References 1. CMS Tracker Technical Design Report, CERN/LHCC 98-6, 1998. 2. A.Ruzin for RD48, NIM A 447 (2000), 116. 3. T.Beckers et al, Proceedings of 9-th Vienna conference on instrumentation 2001(to be published in NIM A). 4. M.French, APV User Manuals, h t t p : //www. t e . r l . a c . uk/med. 5. M.Friedl et al, Proceedings of the Vertex 2001 conference (to be published in NIM A). 6. N.Demaria et al, NIMA 447 (2000) 142-150.

STATUS OF T H E CMS PIXEL D E T E C T O R T. ROHE Paul Schemer Institut, 5232 Villigen, Switzerland e-mail: [email protected] for the CMS Pixel Collaboration The innermost layers of the CMS tracking system will consist of pixel detectors. They will allow pattern recognition in the high track density and will be used as vertex detector. An overview of the system and a status report on the different components will be presented. Emphasis will be given to the latest developments in 2001: The first submission of a full-size radiation-hard readout chip and the latest sensor prototyping.

1

Detector Layout

The tracking unit of the CMS experiment at the Large Hadron Collider (LHC) will contain hybrid silicon pixel detectors for track reconstruction and btagging 1 . It will consist of three barrel layers and two end disks at each side. The barrels will be 53 cm long and placed at radii of 4.4 cm, 7.3 cm, and 10.2 cm (fig. 1). They cover an area of about 0.8 m 2 with roughly 800 modules. The end disks are located at a mean distance to the interaction point of 34.5 cm and 46.5 cm. The area of the 96 turbine blade shaped modules in the disks sums up to about 0.28 m 2 . In the first years when LHC has not reached its final luminosity only the two innermost barrel layers and the first disk on each side will be installed. This system will represent about half of the final system and provide a two hit coverage up to a pseudorapidity" of \rj\ < 2.1. By adding the 3 r d barrel and the 2 n d disk the system will provide three hits over about the same solid angle without extending the region of two hit coverage. In order to achieve the best vertex position measurement the spatial resolution of the sensor should be as good in the ^-direction (parallel to the beam line) as in (r, ip) and therefore a squared pixel shape with a pitch of 150 x 150 um2 was adopted. To improve the spatial resolution analog interpolation between neighbouring channels will be performed. The strong Lorentz deflection in the (r, (^-direction caused by CMS' 4 T magnetic field is utilized to distribute the signal onto several channels. Hence the detectors are not tilted in the barrel layers. The resolution along the z-axis is determined by a

n = — lntan(0/2) where 0 is the track angle relative to the beam axis.

231

232

Figure 1. Perspective view of the CMS pixel system.

the pixel pitch in the region with low pseudorapidity and by charge sharing if the tracks hit the sensors under an angle where the typical cluster size can exceed values of 6 or 7. The best resolution will be reached at the point where the charge is distributed over about two pixels. In the disks where the charge carrier drift is hardly affected by the magnetic field the modules are tilted about 20° resulting in a turbine like geometry visible in fig. 1. 2

Modules and Mechanics

A picture of a barrel module is shown in fig. 2. A 66.3 mm long, 18.45 mm wide, and 300, nm thick sensor is bump bonded to 2 x 8 readout chips. The bump bonding procedure using Indium has been developed at PSI. It is currently used to assemble 64 pixel modules for an experiment at the Swiss synchrotron light source (SLS) at PSI 2 ' 3 . This ensemble is glued with the chips down to a 270 urn thick silicon base plate which is attached to the cooling frame with small screws. Silicon is chosen as material to avoid mechanical stress

233

bumps

readout chips

silicon base plate

Figure 2. View of a barrel module. In the cross section the vertical scale is raised by a factor of 5.

due to thermal expansion. A 50 um thick polyemide hybrid that is thermally matched to silicon is glued on the sensor's back side. The electrical connection to the readout chip is done via wire bonding. The hybrid is equipped with passive components and the so called token bit manager chip managing the readout of the system. Clock, control and data signals are transferred to the barrel periphery via a copper-on-Kapton cable glued and wire bonded to the hybrid. The power is brought in by extra aluminum wires. At both ends of the barrel an end flange is situated where the cables are grouped into sectors and brought to the end of the tracker volume via a 2.2 m long service tube. On this service tube the conversion from electrical to optical signals and vice versa is performed. The end cap modules look slightly different due to their trapezoidal shape. The width of the sensors varies from 2 readout chips per sensor in the innermost region to 5 at the outer end of each blade. The hybrid is placed between the readout chip and the base plate *. The full pixel detector including the service tubes can be preassembled and inserted into CMS as the last component. It will have to be removed at least every second year of LHC running for beam pipe bake out and replacement of the innermost layers which suffer most from radiation damage.

234

Figure 3. Readout scheme of the CMS pixel readout chip. Data of hit pixels will be transmitted to the column periphery via a fast scan and stored there until either rejected or read out via the time stamp and data bus. The control & interface block contains a logic to program the chip and set some reference voltages.

3

Detector Readout

The high bunch crossing rate of 40 MHz requires LHC experiments to readout while data taking continues. This involves a complicated scheme of buffering the data in frontend pipelines up to the time when the trigger decision arrives. The readout architecture of the CMS pixel detector is explained in 4>1. 3.1

Chip Architecture

The readout chip as the most crucial component of the readout chain uses a column drain architecture as described in 5 ' 1 . This architecture assigns as many tasks as possible to the column periphery located at the edge of the chip as shown in fig. 3. This approach keeps the pixel unit cell simple. In its present implementation it contains only about 125 transistors, much less than those of other LHC experiments.

235

When a pixel is hit it notifies the column periphery via a fast OR. A time stamp is created instantly and a token search mechanism is initiated scanning all pixels of a double column as indicated in fig. 3. If a hit cell is identified, its address and analog signal is transmitted to the data buffer in the periphery. In contrast to a fixed association of data buffer cells with each time stamp, a scheme using a pool buffer for all pixel cells of a double column was adopted. It allows to store a variable number of pixel hits per time stamp and ensures that large variations in the hit multiplicity are accepted. This is important for events where the pixels are inside a high pt jet or in case of heavy ion collisions. The readout of several chips in a module is controlled by the token bit manager chip. It sends a token flag to a group of daisy chained readout chips scanning all chips for hit double columns in a similar way as done for the pixel cells in one double column. When a hit double column is found data is sent to the DAQ system via the time stamp & readout bus as indicated in fig. 3. In addition the readout chip will contain a control & interface block for chip programming and setting of reference voltages. 3.2

Latest Results from Prototyping

A chip (PSI 41) with the architecture described above has been produced in the radiation hard DMILL process. It contains 36 x 40 pixel cells which is about half the number of the final chip with 52 x 53 pixels and totals roughly 240000 transistors. The pitch is 150 x 150 um 2 , according to the technical design report 1. The following features are implemented: • Final column drain architecture as explained above. In addition a double hit capability is implemented allowing to record an additional hit in the double column during a hit scan. • The complete double column periphery is included with 8 time stamp buffers and 24 pixel data buffers. • Fully functional readout chain. The readout takes 6 clock cycles per hit. The column and pixel addresses are coded in five analog levels. The pulse height is also read out analog. • Some variations of the analog block are implemented for final optimisation. Still missing in this prototype is the control & interface block which contains 21 DACs to set reference voltages, voltage regulators for the supplies, and a fast I2C-interface for chip programming. It is included in the final chip (PSI 43) submitted in autumn 2001.

236 Rgpr

Rgsh

I—wv-

i—y^r-

Comp

(a)

cnH>-

250fF

1.7fF

Preamplifier

—ycV—i

(b)

A

Shaper

X

Sample and hold

i—yw-

Hh

OT^> Figure 4. Two realisations of the analog block of a pixel unit cell: (a) with a source follower (b) alternative option with lower power consumption.

Tests performed with the PSI 41 chip were successful. Pixels can be enabled or masked, thresholds can be set and tuned individually for each pixel, and calibration pulses can be injected. The column drain mechanism performs as expected. A speed of 2 GHz was reached in the hit scans. Time stamp and data buffers and the event assembly is working well. Problems showed up in the readout and clocking speed which is limited to 15 MHz and 35 MHz respectively instead of the required 40 MHz. However the reasons of the speed limiting problems were identified and fixed for the next chip generation PSI 43. The power consumption of the chip is in the order of 90 uW per pixel equally shared by the digital and the analog part. A significant fraction of the analog power is consumed by the source followers (fig. 4a), which are fed by a separate supply voltage. In order to simplify the system, replacement options for the source followers were investigated. While the first one between preamplifier and shaper can be omitted by slightly reducing the feed back capacitance, the second one has to be replaced as shown in fig. 4b. The power needed by the current mirror is much decreased compared to the source

237

followers and the voltage adjustment is less critical. Therefore this supply voltage can be derived from the analog power line, reducing the number of required supply voltages from four to three. The performance of this circuit is superiour to the original one. The risetime at the end of the shaper which determines the time walk of the comparator is still fast and was measured to about 20 ns with an artificial input load capacity of 106 fF. The output to the sample-and-hold also displays a fast peaking time but a slower return to baseline due to the p-MOS transistors. This provides further robustness as timing of the sampling is less critical. A further advantage of the new circuit is the reduction of the power consumption by 15-20%.

3.3

Future Plans

The full size radiation hard chip PSI 43 with all features necessary for an operation in CMS including the control and interface block is currently in production using the DMILL process. The recticle will also contain the token bit manager chip. Delivery is expected in spring 2002. A translation of the chip into a radiation hard 0.25 um-technology is planned. This technology offers the possibility of a further pitch reduction with a cell size of 100 x 150 um 2 instead of the previous 150 x 150 urn2. In the disks the reduced pixel dimension measures the r-direction with improved accuracy. In the (^-direction the spatial resolution is favoured by charge sharing induced by the 20° tilt of the modules. In the barrel the zresolution in the non-central region is determined by charge sharing due to the tilt of the tracks. A pitch reduction along z would only improve the resolution were the hit multiplicity is well below two, which is the case for -q < 0.5. For the barrel the pitch reduction is realised along the (r,
4

Sensor D e v e l o p m e n t

The most challenging requirement for the sensor part of the hybrid pixel system is the radiation hardness. The innermost barrel layer will be exposed

238

to a fluence6 of about 3 x 1014 n e q /cm 2 per year at full luminosity, the 2 n d and 3 r d layer to about 1.2 x 1014 n e q /cm 2 and 0.6 x 10 14 n e q /cm 2 respectively. Since the readout chip as the most vulnerable part of the system is believed to survive a particle fluence of 6 x 10 1 4 n e q /cm 2 , all components of the pixel detector including the sensor part have to perform well up to at least this fluence. This also implies that the 2 n d layer will have to be replaced once after about 7 years (assuming the first 3 years with reduced luminosity) while the innermost layer must probably be replaced every 1 to 2 years of full luminosity equivalent. An operation of the silicon sensors up to 6 x 1014 n e q /cm 2 requires an operation at less than half the full depletion voltage in the last phase. Therefore an "n on n" aproach has been chosen 1. The first prototypes of the sensor were delivered in 2000. They used p-stops as n-side isolation. In order to prevent accidently unconnected pixel cells from floating to high potentials with respect to their neighbours, the p-stops are not entirely closed. Along a narrow path through the p-stop openings an electron accumulation can form providing a high resistive connection path to the neighbours. Such a connection between all pixels also allows on-wafer testing prior to bump bonding as all pixel cells are connected and can be reached with one probe needle placed at the sensor edge. A large variety of open p-stop geometries has been implemented and the interpixel resistance has exensively been investigated 9 . Before irradiation the resistance depends strongly on the amount of overdepletion. Due to pinch-off mechanisms between the p-stops the resistance rises to values above 10 MQ. Therefore the testability is only given for some of the designs 10 . After irradiation the inter-pixel resistance saturates at values of several G£2 more or less independently of the design. Although unbonded pixels float to high potentials with respect to their neighbours, no harmful effects have been observed. There is no correlation between missing bump bonds and noisy pixels in irradiated samples. However the number of noisy pixels in samples irradiated to levels relevant for LHC turned out to be on an unacceptable level when the bias voltage is risen above 300 V. This is probably due to avalanche breakdown at the p-stop edges, an effect which has also been observed in strip detectors 11>12. The implemented guard ring structure consisting on 10 floating rings with increasing distance towards the edge displays a very good performance up to voltages of l k V 9 . In order to continue the sensor developement and to further optimise the design, a second sensor prototype has been submitted in 2001 10 . It contains several p-stop design options with the aim to reduce the bias dependence of ''fluence is normalised to lMeV neutron equivalent n e q /cm 2 .

239 the interpixel resistance. This will be reached by wider p-stop openings and larger gaps between the p-stops themselves. This implies that only one p-stop atoll ring per pixel will be used. To improve the post radiation breakdown behaviour two aproaches will be followed. One possibility are field plates covering the lateral pn-juction of the p-stops which are held on p-stop potential. This method used in power electronics since the late 1960s 13 showed to be quite successful 12>14. The other approach addresses directly the root of the problem, i.e. the reduction of the p-stop dose from currently about 5 x 10 13 c m - 2 to the minimum value possible. This automatically leads to the p-spray technique 15 which is known for its good high voltage capability in the irradiated state 16>17. A final decision on the sensor design will be made after the inverstigation of the second prototype submission in 2002. 5

Summary

The CMS pixel detector has been described. Latest developement is the submission of a full size radiation hard readout chip in DMILL technology satisfying the CMS specifications. All critical features have been investigated in previous readout chips with a reduced size. A translation of the chip into a radiation hard deep sub-micron CMOS design is planned. This allows a moderate decrease of the pixel size. Sensors have been produced and successfully tested in 2000. A second prototye for a further optimisation has been submitted in 2001. References 1. The CMS Collaboration, CMS Tracker Technical Design Report, CERN/LHCC 98-6. 2. Chr. Bronnimann et al., A pixel read-out chip for the PILATUS project, N I M A465 (2001) 235-239. 3. E. F. Eikenberry et al., PILATUS: A 2-D pixel detector for protein crystallography, Presented at the 10 th International Workshop on Vertex Detectors "Vertex 2001", September 23 rd -28 th , 2001 in Brunnen, Switzerland, to be published in NIM A. 4. D.Kotlinski et al., The CMS pixel detector, N I M A465 (2001) 46-50. 5. R. Baur et al., Readout architecture of the CMS pixel detector, N I M A465 (2001) 159-165. 6. B.Hendrich and R. Kaufmann Lorentz angle in irradiated silicon, Presented at the 5 t h Position Sensitive Detector Conference, London, 1999, accepted for publication in NIM A. Accessible via

240

http://cms.web.psi.ch/cms_conference.reports.html 7. M. Aleppo et al., A measurement of Lorentz angle of radiation-hard pixel sensors, N I M A465 (2001) 108-111. 8. C. Troncon et al., A measurement of Lorentz angle and spatial resolution of radiation-hard pixel sensors, accepted for publication in NIM A 9. R. Kaufmann, Developement of Radiation hard Pixel Sensors for the CMS Experiment, PhD thesis at the faculty of mathematics and science at the university of Zurich, Switzerland, 2001. 10. T. Rohe et al., Sensor development for the CMS pixel detector, Presented at the 5 t h International Conference on Large Scale Applications and Radiation Hardness of Semiconductor Detectors, July 4 t h -6 t h , 2001 in Firenze, Italy, accepted for publication in NIM A. 11. D.Robinson et al., Noise studies of n-strip on n-bulk silicon microstrip detectors using fast binary readout electronics after irradiation to 3 x 1 0 1 4 p c m - 2 , N I M A426 (1999) 28-33. 12. Y. Unno et al., Novel p-stop structure in the n-side of silicon microstrip detector, presented at the Hiroshima symposium on semiconductor devices, held 1997 in Mebourne, Australia. Submitted to the conference proceedings (not published). Accessible via http://jsdhpl.kek.jp/~unno/notes.html 13. B. J.Baliga Modern power devices, Wiley, New York, 1987, pp 116 and references therein. 14. T. Nakayama et al., Radiation damage studies of silicon micro strip sensors, IEEE Trans. Nucl. Sci. Vol. 47, No. 6, December 2000, p 18851891. 15. R. H. Richter et al., Strip detector design for ATLAS and HERA-B using two-dimensional device simulation, N I M A377 (1996) 412-421. 16. M.S. Alam et al. The ATLAS silicon pixel sensors, N I M A456 (2001) 217-232. 17. T. Rohe et al., Design and test of pixel sensors for the ATLAS pixel detector, N I M A (1999) 55-66.

F A B R I C A T I O N OF M I C R O S T R I P D E T E C T O R S A N D INTEGRATED ELECTRONICS ON HIGH RESISTIVITY SILICON

G.-F. DALLA BETTA* M. BOSCARDIN, P. GREGORI, N. ZORZI ITC-irst, Divisione Microsisttmi, 38050 Povo (TN), Italy G. BATIGNANI, S. BETTARINI, M. CARPINELLI, F. FORTI, M. GIORGI, A. LUSIANI, M. RAMA, F. SANDRELLI, G. SIMI INFN-Pisa and Universita di Pisa, 56010 S. Piero a Grado (PI), Italy

INFN-Trieste

L. BOSISIO, S. DITTONGO and Universita di Trieste, 34128 Trieste, Italy

G. U. PIGNATEL Universita di Trento, 38050 Mesiano (TN), Italy P. F. MANFREDI, M. MANGHISONI, L. RATTI, V. SPEZIALI, G. TRAVERSI INFN-Pavia and Universita di Pavia, 27100 Pavia, Italy V. RE Universita di Bergamo, 24044 Dalmine (BG), Italy A fabrication technology has been developed at ITC-irst (Trento, Italy) for the realisation of silicon microstrip detectors with integrated front-end electronics, to be used in high-energy physics and space experiments as well as in medical/industrial imaging applications. The main technological issues are addressed, and experimental results from the electrical characterisation of the first prototype batch are reported, showing that good quality transistors are obtained within the proposed technology while preserving the basic detector parameters.

1

Introduction

The possibility to integrate at least part of the front-end electronics on the same silicon radiation detector substrate can greatly simplify the mechanical assembly of the read-out system. Owing to the minimisation of the stray capacitance associated with the detector-preamplifier connection, this approach can also enhance the noise performance for detectors having a low output capacitance, such as drift chambers and pixel detectors, whereas for microstrip -TEL.+39-0461-314543, FAX. +39-0461-302040, E-MAIL: DALLABESITC.IT

241

242

detectors, which feature a relatively high capacitance ( « l p F / c m ) , the advantages of integrated electronics are less evident. Nonetheless, a fully integrated system is very appealing in applications where compactness, weight, amount of material are crucial. For instance, an integrated preamplifier at the end of a strip detector would allow to move the rest of the readout electronics, which in traditional systems is right next to the silicon detector, further apart, significantly reducing the amount of material and complexity in the active detection area. Besides, a very dense stacking of detectors would become possible, with applications in X-ray detection or active targets. Finally, integrating the entire readout chain at the end of the strip, an extremely compact system with very few connections and external components would be obtained, which is ideal for space applications. To this purpose, we have modified the fabrication technology developed at IRST for PIN detectors with integrated N-JFET's 1>2, in order to realise microstrip detectors and integrated read-out electronics on high resistivity silicon. We report on the main technological issues and on selected results from the electrical characterisation of the first prototype batch. 2

Design of detectors and integrated electronics

As a first step toward the realisation of a fully integrated detection system, we have considered a monolithic structure consisting of a microstrip detector with integrated N-JFET (triode) in the source-follower configuration. Fig. 1 shows the schematic diagram of the monolithic structure, consisting of a strip detector (SD), either (A) DC- or (B) AC-coupled to a source follower, to be connected via the integrated capacitance CA to an external readout circuit (charge sensitive amplifier plus shaper). Theoretical analyses and circuit sim-

Figure 1. Schematic diagram of a microstrip detector with (A) DC-coupled, (B) ACcoupled integrated N-JFET in the source-follower configuration.

243

illations have been carried out, assuming the external electronics to be implemented with the analog section of the AToM chip, developed for the readout of BaBar detectors 3 . Good electrical figures have been predicted, also in terms of Equivalent Noise Charge (ENC), the value of which was found to be in the order of 500 e~ rms in the shaping time range from 100 to 400 ns 4,B . 3

Device fabrication

IRST process for PIN diodes and integrated N-JFET's, detailed in l, has been further developed to include polysilicon (low and high resistivity) and recessed coupling capacitors with a stacked-dielectric insulator (SiC^-TEQS), while maintaning the same basic approach which features: (i) p + and n + implants (shallow and deep) and thermal diffusion for the transistor realisation and (ii) back-side, P-doped poly-Si gettering to ensure low diode leakage current. Moreover, the process thermal budget has not been altered, so as to preserve the most critical characteristic, i.e., the JFET doping profile in the gateregion. The schematic cross section of a monolithic structure, consisting of a microstrip detector and a front-end JFET, is shown in Fig. 2.

I n' SI substrate

I

Figure 2. Schematic cross-section of a monolithic strip-)-JFET structure (not to scale).

4

Experimental results

Measurements on test structures evidenced an adequate process control of the main parameters, such as: (i) diode leakage current (~ 0.5 ± Q.lnA/cm2 at M l depletion); (ii) polysilicon sheet resistance, both for high-resistivity (14.1±0.6 kO/sq.) and low-resistivity (402±8 fi/sq.) resistors; (iii) dielectric thickness of the insulators employed in poly/p + (212±3 mm) and poly/metal (193±7 urn) integrated capacitors. JFET's were tested on-wafer by means of an automatic probe-station. All measurements were carried out at substrate reverse voltage, Vs„&=6GV, higher

244

than the wafer depletion voltage. As an example, Fig. 3 shows the transfer characteristic of a transistor having an aspect ratio W/L=1000/xm/4/zm, namely the same device adopted in the structures of Fig. 1. The pinch-ofF voltage is about —1.15V, while the drain saturation current, Id ss , is about 3mA, resulting in a high transconductance, gm ~7.5mS. The output characteristics of the same device at different Vgs values are shown in the inset of Fig. 3: a good saturating behaviour in the pinch-off region is observed, and high output resistance (rout) values were measured. The main electri-

gate-to-source voltage, V ^ (V)

Figure 3. Id-V 93 characteristic of a J F E T with W/L=1000/mi/4/mi, with output characteristics at different Vgs values in the inset.

frequency (Hz)

Figure 4. Spectral density of the noise voltage as a function of frequency for three J F E T samples with W/L=1000/im/4pm.

cal parameters of JFET's having different width (W) with the same length (L=4/im) are reported in Table 1, evidencing a correct scaling of the electrical figures with the device width. Note that in the Vgs and Vds range of practical interest, the JFET gate current, I 9 , is low, its value being dominated by the leakage current of the p-well/n-substrate junction. On the contrary, the values measured for the input capacitance, C s s s , are quite high, particularly for the device with W=1000/im. Thus, in order for the ENC performance of the strip-)-JFET structure not to be degraded, a smaller transistor width should be preferably used, allowing for a better g m vs. Cgss trade-off. Noise tests have also been performed on JFET's: as an example, Fig. 4 shows the spectral density of the noise voltage as a function of frequency for three devices having W/L=1000)um/4/zm, biased with I
245 Table 1. Summary of the electrical figures of JFET's having different width (W) with the same gate length (L=4 (im). Data refer to the bias point with V^s=0, Vj3=3V, VSU),=60V.

w (100 H 200 400 500 1000

Idas

(mA) 0.39±0.09 0.77±0.19 1.50±0.35 1.85±0.41 3.59±0.69

gm (mS) 0.79±0.08 1.58±0.19 3.06±0.33 3.84±0.44 7.50±0.93

Tout

(kil) 84.5±30.6 44.9±18.7 24.1±9.0 19.2±7.1 9.8±3.0

I9 (PA) 5.61±0.40 8.17±0.70 12.43±0.52 13.86±0.78 27.75±1.27

Ogss

(PF) 1.16±0.11 2.16±0.23 4.30±0.40 5.12±0.50 10.09±0.94

observed, which is typical of fully-implanted devices, whereas the white noise voltage is in the order of 4nV/y/Hz, higher than expected on the basis of the gm values (~1.6nV/VHz). The reason for this difference is likely to be ascribed to a non negligible contribution from the p + -gate series resistance. 5

Conclusions

We have reported on the development of a fabrication technology for silicon microstrip detectors with integrated electronics. Results from the electrical characterisation of the first prototype batch have evidenced that the fabrication process can provide good quality transistors while preserving the basic detector parameters. Functional testing of microstrip detectors with integrated JFET's in the source-follower configuration is under way. Acknowledgments This work has been partially supported by the Italian Ministery for Scientific and Technological Research and by the National Institute for Nuclear Physics of Italy (INFN). References 1. 2. 3. 4.

G.-F Dalla Betta et al, Nucl. Instr. Meth. A417, 325 (1998). G.-F Dalla Betta et al, Nucl. Instr. Meth. A458, 275 (2001). P. F. Manfredi et al, IEEE Trans. Nucl. Sci. 46(6), 1865 (1999). G.-F Dalla Betta et al, "Feasibility studies of microelectrode silicon detectors with integrated electronics", to appear in NIM A. 5. L. Ratti et al, "Integrated front-end electronics in a detector compatible process: source-follower and charge-sensitive preamplifier configurations", to appear in the Proceedings of SPIE, Vol. 4507, 2001.

T H E D I A M O N D P R O J E C T AT GSI - P E R S P E C T I V E S E. BERDERMANN*, K. BLASCHE, H. W. DAUES, P. MORITZ, H. STELZER*, B. VOSS Gesellschaft fur Schwerionenforschung, Planckstrasse 1, 64291 Darmstadt, Germany E-mail: [email protected] * Members of the RD42 Collaboration,

GERN

CVD-diamond detectors operational in various heavy-ion experiments at GSI are described. The results shown demonstrate convincingly the suitability of such detectors for a variety of tasks, where classical well known detectors fail. New applications are introduced, which are being recently developed for the use in measurements of minimum-ionizing particles.

1

Introduction

Heavy-ion experiments at GSI cover nuclear and atomic physics, related basic and applied research in plasma physics and material science as well as biophysics including tumor therapy with carbon ions. After low-energy sections, the heavy-ion synchrotron (SIS) provides pulsed, cooled ion beams up to 2 GeV/amu of beam intensities up to 10 11 ions/spill. All kind of ions from protons with Z = l to 238 U ions with Z=92 are available. Detector and trigger systems have to cope with a signal ratio of 1:9000, a variable beam intensity from a few ions to about 10 11 ions/s and a time structure from 100 ns to 10 s. Whereas the high-energetic ions from the SIS traverse nearly undisturbed the detectors, in the low-energy domain of the accelerator facility the ion range in solid state material is in the order of 1 /xm. Severe radiation damage is observed. The most remarkable properties of CVD-diamond detectors implemented in heavy-ion measurements at GSI are radiation hardness, intrinsic time resolution well below 50 ps and a single-particle count-rate capability ranging from 1 to 109 ions/s x. However, due to the inhomogeneous charge collection inside the polycrystalline diamond bulk, even a rough particle identification based on pulse-height resolution is excluded 2 . 2

CVD-Diamond Heavy-Ion Detectors

Contrary to high-energy physics, where polished highest-quality diamond material and low noise charge-sensitive integrating electronics is needed, the

246

247

highly-ionizing heavy ions allow the use of thin 'as grown' material, which is characterized by a short carrier lifetime. In order to take advantage of the fast collection of charge (~1G0 ps) low-impedance broad-band amplifiers (DBA) are used 3 . The diamond samples are connected via 50 0 micro-strip lines to the amplifiers. This type of amplifier can be driven also, if connected after some meters of impedance-matched transmission lines, without disturbing the signal performance. 2.1

Diamond Detectors in Physics of Dense Matter (W. Koenig et al.)

A beautiful example which demonstrates the variability and power of CVDdiamond detectors in heavy-ion experiments is the HADES spectrometer. 4

Th« Start- V'£to Device Spill Monitors, TO, ToF Detectors

Target ""'*"> sjstt

2 Octogones A » 25 x 15 mm d B = 100 urn

Figure 1. The Start-Veto device (left) of the HADES spectrometer (right).

The High-Acceptance Di-Electron Spectrometer is designed for studies of hadronic properties in nuclear matter. Theoretical models predict a shift in the mass and resonance width of vector mesons like p, w and <j>, if they are produced inside the nucleus. Lepton pairs obtained from the decay of the mesons axe excellent probes for such investigations. A mass resolution better than 1% is required to distinguish p and w. Pig. 1 shows a schematic view of HADES (right) and a zoomed explosion view of the Start-Veto device (left). A pair of identical diamond strip detectors located 75 cm upstream and downstream the target is used to determine the START for the ToF (Time-of-Flight) measurements. The lepton candi-

248

dates are selected by position correlations of hits in the RICH (Ring Imaging Cerencov Hodoscope) and in the segmented plastic-scintillator wall (TOF). The short intrinsic dead time of ~ 1-2 ns 1 is the most important advantage of the diamond detectors in this experiment. The required high collision rates of ~10 8 ions/spill can therefore be accepted. Beam particles not reacting in the target produce a signal in the downstream detector, which is used as a veto for the upstream detector. This provides a START signal with a rate below 107 ions/spill. The veto efficiency of 90% has still to be improved. In Fig. 2 ToF spectra measured in a commissioning run with 52 Cr ions on 27 Al are shown. The distance between the START counter and the segmented plastic-scintillator wall (TOF) is 2.1 m.

t,

10 N

ALL

^ IV -

i j /

o

8

« c 3 o O

6 4 j1

ToF[ns] 0

4

8

12

16

20

ELECTRONS

•J

"0

5

ToF[ns]

^

10

15

Figure 2. Left: ToF spectrum of all particles detected. Right: ToF spectrum of electron candidates, as selected by means of a position correlation between RICH and T O F hits. Data were taken without magnetic field.

The data from all 384 TOF segments are included in both spectra. In the left plot the ToF spectrum of all particles arriving within 20 ns at TOF is shown. By means of an angular correlation between RICH and TOF, leptons are selected, and the ToF spectrum of those candidates is shown on the right plot. The intrinsic time resolution of the Start-Veto device amounts to a— ± 29 ps x. The ToF resolution of a=233 ps is determined by the TOF detector and is affected from the calibration of all TOF segments. 2.2

Focal Plane Detectors of Magnetic Spectrometers (S. Toleikis et al.)

A large area (60 mm x 40 mm) CVD-diamond detector of a thickness of 200 /xm has been implemented as the focal plane detector of a magnetic spectrometer used for beam-foil spectroscopy. Since August 2000 the experimental setup shown in Fig. 3 (left) is in operation.

249 Hydrogenlike ions passing through a target foil produce different excited charge states by electron capture. The 7 radiation of the subsequent decay of

Figure 3. Left: The experimental setup in the atomic physics cave. Right: X-ray spectra of helium-like Au atoms with (lower curve) and without coincidence (upper curve) to the corresponding diamond strips.

the excited states is detected with two Ge(i) detectors, one of them movable. The different ionic charge states are deflected by the magnetic field on different diamond strips. High-resolution spectroscopy and lifetime measurements are performed up to the heaviest ions. Fig. 3 (right) demonstrates the background suppression in X-ray spectra of heliumlike 197 Au ions if the Ge(i) detector is readout in coincidence with the corresponding diamond strips (low spectrum). 2.3

Diamond Detectors in Accelerator Beam-Diagnostics

Beam intensities up to 10 11 ions/spill are particularly challenging for detectors to be used in accelerator beam diagnostics. However, this is the top application of CVD-diamond detectors. It is not just they fulfill the tasks excellent but also, there exists nearly no alternative detector material. A variety of beam-intensity- and beam-profile monitors as well as beam-loss monitors outside the beamlines is installed. Optionally, each of the diamond detectors can be used for high-resolution spill- and bunch-structure investigations. Ion beams from the SIS are extracted either 'slow', distributed over a time interval 0.5 s to 10 s or 'fast' in a bunch of about 100 ns. When monitoring slow extracted beams up to a beam intensity of 109 ions/s the detectors are used in a single-particle mode. An integrated current pulse is obtained from fast extracted bunches. Due to the short intrinsic dead time of the detectors, it was possible for the first time to observe the unexpected internal time structure of the bunches. Fig. 4 shows two 58 Ni bunches of quite different time structure, which were extracted consecutively. Whereas on the left picture the 4 pre-

250 0.08

0.08

•

Spi 12 •a

^

'

1

•

'

'

'

P'l—IT"*",

1

006

CD T3

£

•o c o

0.02

co

0.00

5

'

Spin 3

0.06

%_ 0.04

<

'

= E

0.04

•o c o

0.02

.«

0.00

< ^

0

htfj**^

-0.02

-2e-7 -1e-7

0 1e-7 2e-7 3e-7 t[s]

W

ftystyto

-0.02 -2e-7

100 ns

-1e-7

* V ^

0

^

1e-7 2e-7 3e-7 t[s]

100 ns

Figure 4. Two consecutively extracted nickel bunches.

bunched beam strings are still visible on the right one a bunch compression of about 20 ns seems to be achieved.

2.4

Position-Sensitive

Carbon-Ion Dosimeter

More than hundred patients have been already treated at GSI. 12 C beams of an energy between 100 MeV/amu and 400 MeV/amu and an intensity of 108 ions/spill are scanned over the tumor volume. The results are impressive. In order to check the presently used ionisation chambers and to have an independent determination of the applied dose, single-particle counting is proposed. A prototype large-area position-sensitive carbon dosimeter was developed and first beam tests were performed recently. The 20 x 20 mm 2 CVD-diamond pad detector of a thickness of 100 fj,m contains 16 pads which are glued with conductive silver on a doubble-sided ceramic board. Each pad is connected to a broad-band amplifier. The dosimeter operates as a beamprofile monitor measuring precisely the total fluence F in ions/cm 2 . The dose D applied on each pad can then be calculated taking into account the energy loss AE of a single 12 C ion in matter with density g and thickness d according to D=(AE/ion)*(F/d£>). The similarity of carbon with human tissue is of particular advantage.

251

3

ToF Detectors for M I P s (M. Petrovici, N I N P E Bucharest)

The use of CVD-diamond detectors for fast timing of Minimum-Ionizing Particles (MIPs) has been investigated. First tests using 90 Sr electrons have been performed. Two polished diamond detectors with a collection distance of 210 fim and a thickness of 500 fim were triggered by a plastic scintillator behind the diamonds. DBA amplifiers and low-threshold discriminators were used. The time-difference distribution fitted to a gaussian gives a time resolution of <7=95 ps assuming equal contribution of both detectors. However, due to an equivalent noise charge of 12000 e of the DBA, the detection efficiency obtained was only 3%. New type of amplifiers based on High Electron Mobility Transistors (HEMT) are under construction striving for both, a significantly improved signal-to-noise ratio and a time performance of the present DBA generation. 4

Concluding Remarks

The use of CVD-diamond detectors at GSI has been established. Various CVD-diamond applications in the framework of the tumor therapy are under investigation. We proceed with the study of polished CVD-diamond samples for timing detectors for MIPs. Recently, a joint venture with the Wits University of Johannesburg, Southafrica has been started, aiming to investigate the suitability of HPHT (High Pressure High Temperature) single-crystal diamond for electronic devices and for radiation hard detectors which combine both, the timing properties of CVD diamond and the energy-resolution of silicon detectors. Acknowledgments We would like to thank the GSI target laboratory for the excellent metallisation of the diamond samples, the accelerator crew for support during the beam tests and all colleagues providing experimental data. References 1. 2. 3. 4.

E. Berdermann et al, Diamat. 10, 2001 (1770-1777). E. Berdermann et al, Bormio Proc. 116, 2000 (104). P. Moritz et al, Diamat. 10, 2001 (1765-1769). R. Schicker et al, NIMA. 380, 1996 (586).

R A D I C A L B E A M G E T T E R I N G EPITAXY OF ZNO A N D G A N A. N . G E O R G O B I A N I , V. I. D E M I N A N D M. O. V O R O B I E V P.N.Lebedev

Physical

Institute

of RAS, Leninsky prospect Russia E-mail: [email protected]

53, 117924

Moscow

A.N.GRUZINTSEV, I.I.HODOS Institute

of Microelectroniks

and High Purity Materials of RAS, Russia E-mail: [email protected]

Chernogolovka

M. B . K O T L J A R E V S K Y , V. V . K I D A L O V A N D I. V. R O G O Z I N Pedagogical Institute, Berdyansk Ukraine E-mail: [email protected] P-type ZnO layers with a hole mobility about 23 c m 2 / ( V s), and a hole concentration about 10 1 5 c m - 3 were grown by means of radical-beam gettering epitaxy (the annealing of n-ZnO single crystals in atomic oxygen flux). The effect of native defects on the photoluminescence spectra of the layers was studied. The dominant bands in the spectra peaked at 370.2 and 400 nm. These bands were attributed to the annihilation of exciton localised on neutral V z „ and to electron transitions from the conduction band to singly positively charged V z „ correspondingly. The effect of annealing in atomic nitrogen flux of p-CaN:Mg films on their photoluminescence spectra and on the value of their conductivity were studied. Such annealing leads to appearance of a number of emission bands that peaked at 404.9, 390.8 and 378.9 nm and increases hole concentration from 5 X 10 1 5 to 5 X 10 1 6 c m - 3 , and the hole mobility from 120 to 150 c m 2 / ( V s). The n-ZnO — p-GaN:Mg electroluminescence heterostructures were obtained. Their spectrum contains bands in the excitonic region of GaN at the wavelength 360.2 nm and in the edge region at wavelengths 378.9 and 390.8 nm.

1

Introduction

The synthesis of wide gap ZnO and GaN semiconductors with the necessary intrinsic defects composition is a difficult technology problem. We have found that just the intrinsic defects are responsible for luminescent and electrophysical properties of these materials, including the type of their conductivity. ZnO and GaN obtained by different traditional methods are characterized by the deviation from the stoichiometry to metal component surplus. This causes the presence of such intrinsic defects like Zn or Ga in the interstitial location as well as oxygen and nitrogen vacancies which are donors in ZnO or GaN correspondingly, and their concentration determines the conductivity of n-type.

252

253

The original method - Radical Beam Gettering Epitaxy (RBGE) 1 elaborated by us which is based on the thermal annealing of the crystals in the flux of atoms (radicals) of the not-metal component of the compound is very effective for shifting the intrinsic defects composition in these semiconductors to the side of not-metal component surplus. Such annealing leads to the growth of new layers via sorption of chalcogen atoms from the vapour phase and gettering of metal atoms from the crystal bulk (substrate) 2 . The predominant defects in the materials prepared by this method are metal vacancies. 2

Technology methods

Atomic oxygen and nitrogen were generated by an RF discharge in the flux of oxygen or nitrogen at pressure 0.1 - 10 Pa and with RF power 80 W. By passing molecular gases through the RF discharge, we have obtained a mixture of molecules, atoms, ions and electrons. The charged particles have been removed by a magnetic field, so that only molecules and atoms could reach the treated crystals. The concentration of atomic oxygen was about 1016 — 10 17 c m - 3 . In the case of thermal dissociation at 700°C, this concentration corresponds to a molecular oxygen pressure equal to 105 Mpa 3 , a level which cannot be attained under the ordinary synthesis and heat-treatment conditions. The concentration of atomic nitrogen was an order of magnitude lower. The concentration of atoms was determined by measuring the heat supplied to platinum wire via catalyzed surface recombination of radicals into oxygen molecules 4 . From the measured heat and known recombination energy, we evaluated the flux of atoms and their concentration. 3 3.1

Results and discussion ZnO with hole conductivity

Using RBGE we were the first to obtain single crystal ZnO layers with p-type of conductivity. As-grown ZnO crystals were heat-treated in elemental oxygen at temperatures between 400 and 700°C. The thickness of the layers was up to 50 fim, a specific resistance of 102 fi-cm, Hall mobility ~ 23 cm 2 /(Vs) and hole concentration ~ 1015 c m - 3 . Clearly, the predominant acceptors in the layers were Zn vacancies. The layers were used to study the effect of native defects on the photoluminescence (PL) of ZnO. Figure 1 displays the UV PL spectra of ZnO crystals before and after heat treatment in elemental oxygen. Characteristically, the low-temperature PL spectra of as-grown ZnO crystals (excitation at 337.1 nm) show the bands peaked at 367.7, 369.1, 374.6, 383.5,

254

Ac

— , — - . — i — . — , — i — i — i — i — < — i —

360

370

380

390

400

410

Wavelength, nm

Figure 1. 80-K PI spectra of ZnO crystals in the UV range before (1) and after annealing at 500 C for 1 and 4 h, respectively (2,3).

392.0, and 401.3 nm (curve 1). The 367.7-nm emission is known to be due to the recombination of the free A exciton. The 369.1-nm band arises from the recombination of the exciton bound at a neutral donor such as Zn,-. The other bands are phonon replicas of the free A exciton. In the visible range, the spectrum consists of the only intense band at 505 nm. After heat treatment, the UV PL spectrum is dominated by the band peaked at 370.2 nm, which falls in the region predicted for neutralacceptor-bound excitons (368.7 — 373.6 nm) 5 . Since Zn vacancies are the predominant defects in our layers, the 370.2-nm band arises, most likely, due to recombination of excitons bound at neutral V zn • By chemical profiling, the thickness of the ZnO layer was determined to be 10 fim. In the visible range, the 505-nm emission becomes weaker after annealing at 500°C for 1 hour and vanishes after annealing for 4 hours. This emission is commonly attributed to either interstitial Zn or singly positively charged oxygen vacancies ( F + centre). However, F + centres are usually observed in ZnO irradiated with high-energy particles 6 and are rarely present in unirradiated crystals. It is therefore reasonable to attribute this band to interstitial Zn. The decrease in its intensity can be due to the predominant gettering of Zn from interstitials during the RBGE growth of ZnO layers 3 . With increasing the annealing time, the intensity of the 400-nm band rises, while the excitonic emission vanishes (curve 3). Since Zn vacancies are the dominant defects in the p-type films, it is reasonable to assume that the 400-nm band is due to electron transitions from the conduction band to singly positively charged zinc vacancies. We have also investigated the Electron Paramagnetic Resonance of zinc vacancies in ZnO. Usually, because of a small concentration of these vacancies in the

255

initial samples, EPR-signals connected with these defects are not observed. However, as it was mentioned, after the annealing of the crystals in the atomic oxygen flux the concentration of zinc vacancies increases significantly, that makes it possible to detect the corresponding EPR signal. We have registered a series of EPR lines in the region 2.00217 < g < 2.00790, which correspond to single charged zinc vacancies in a case of measurements in the magnetic field perpendicular to the c axis of the crystal. The lines are anisotropic and were observed without optical excitation. In a case of magnetic field parallel to the c axis all the lines coincided. 3.2

GaN.Mg activated by RBGE

Fabrication of high quality GaN films of n- and p-type conductivity permitted the production of semiconductor lasers for the blue spectral region 7 . We have established 8 the donor-acceptor pair nature of the blue emission with a peak at 2.8 eV, with the nitrogen vacancy as a donor. The improvement of the stoichiometry of GaN films will allow to shift their luminescence to exciton UV region and to increase its efficiency. Consequently, annealing in nitrogen atmosphere is required for this. In order to increase the efficiency of thermal treatment the annealing in atomic nitrogen is preferable. Such annealing allows one to reduce the number of native donor defects, which provide the blue emission, to increase the film conductivity and to improve its crystal structure due to the comparatively low annealing temperatures. We have performed the studies of the effect of annealing in atomic nitrogen flux of p-GaN:Mg films on their PL spectra and on the value of their conductivity. GaN:Mg films obtained by molecular-beam epitaxy on sapphire substrates with (0001) orientation were used as the initial material. Doping of GaN during the growth process with Mg acceptor impurity result only in the formation of high- resistance (5.6 x 104 fi cm) p-type material due to the compensation effect. The film thickness was 0.1 /im. The Hall measurements showed the hole concentration 5 x 10 15 c m - 3 and their mobility 120 cm 2 /(V s). The annealing of the GaN:Mg films for 2 h at the temperature in the range 300 — 700° C in atomic nitrogen flux was carried out in the radical-beam epitaxy setup. The concentration of neutral nitrogen atoms in this method exceeded by 4-5 orders of magnitude their concentration in the inactivated atmosphere under the same pressure and annealing temperature. An nitrogen laser with the wavelength 337.1 nm, with pulse duration 10 ns and the mean radiation power 10 kW was used for the PL excitation. The spectra were analysed with the help of double monochromator, which yield a spectral resolution no worse than 1 meV. The PL spectra of the films prior

256

Energy, eV

Figure 2. 80 K spectra of GaN:Mg films: 1 - initial samples; 2 - the samples annealed in molecular nitrogen, 3 - the samples annealed in nitrogen atoms flux. The annealing temperature was 700°C.

and after annealing in the atmospheres of inactivated and activated nitrogen at various temperatures are shown in Fig. 2. The spectra of initial films (curve 1) contain a broad band peaked at 3.23 eV corresponding to edge emission and a narrow band of bound excitons with a peak at 3.44 eV. One should note the good initial stoichiometry of our films, as evident by practical absence of the blue band peaking at 2.88 eV, which is associated with nitrogen vacancies in GaN 8 . Annealing in a nitrogen atmosphere results in a considerable narrowing of the edge luminescence due to the decrease of in the intensities of its longwavelength components (curves 2,3). In this case, the intensity of the excitonic emission increases, which indicates that the crystal structure improves during the annealing. After the annealing of GaN:Mg films in the atomic nitrogen flux, three peaks at 3.27 (378.9 nm), 3.17 (390.8 nm) and 3.06 (404.9 nm) eV are well pronounced in the edge region of the PL spectrum (curve 3). With the annealing temperature increase the intensities of the first two luminescence bands increase, whereas the intensity of the band peaked at 3.06 eV remains practically constant. The intensity of the exciton peak also remains practically unchanged. Annealing in the nitrogen atoms flux increases the concentration of acceptors. The specific resistance of the films decreases with the growth of the annealing temperature up to 5.8 x 103 fi cm at 700°C. Thus, the hightemperature annealing in atomic nitrogen flux enhances the conductivity of the material almost by an order of magnitude. The concentration of holes in the samples after annealing at 700°C is about 5 x 1016 c m - 3 , and the hole mobility ~ 150 cm 2 /(V s).

257

Ql T " ~ . 2,0

""I . 1 2,5 3,0 Energy, eV

.

1— 3,5

Figure 3. Electroluminescence spectra of ZnO - GaN:Mg diode heterostructures at temperatures: 1-80 K; 2-300 K. In the inset, the schematic representation of a diode structure is given.

3.3

Heterostructures

n-ZnO-p-GaN:Mg

The predominance of the ultraviolet luminescence bands in p-GaN:Mg films allows one to hope that the UV light emitting diode structures based on these films can be produced. The ZnO-GaN:Mg diode heterostructures were obtained on the basis of annealed p-GaN:Mg films. The similarity between the crystal and band structures of these two materials with hexagonal symmetry allows us to expect that this type of structure may contribute much in development of electroluminescence devices. As the layer of n-type conductivity, we used low-resistance layers of zinc oxide 0.5 fim thick obtained by oxidation at 550°C of Zn deposited onto the GaN:Mg film. The oxide layers had perfect crystallinity and had only narrow lines of bound excitons in the emission spectra. Electroluminescence (EL) of the diode structure was observed under forward bias. The electroluminescence spectrum of these structures measured at a voltage of 10 V and at liquidnitrogen temperature is shown in Fig.3. The UV emission bands character to GaN:Mg films are observed; these are exciton bands peaked at 3.44 eV, the edge bands with peaks at 3.17 and 3.27 eV. This indicates that the radiative recombination of carriers occurs mainly in the region of GaN:Mg, although the peak of green emission at 2.30 eV with a rather low intensity can be associated with the luminescence of ZnO layers which is likely to be optically excited by UV radiation.

258 4

Acknowledgements

This work was supported by RFBR grant No 00-02-16421, by the Ministry of Science of Russian Federation as part of the program "Physics of Solid State Nanostructures" (Project 99-1122) and as a part of a program "Physics of quantum and waves processes" - subprogram "Fundamental spectroscopy" (Project 01.08.02.8-4). References 1. Georgobiani A. N. and Kotljarevsky M. B, Radical beam gettering epitaxy of II- VI compounds. Nuclear Phys B 61B (1998) pp. 341-346 2. Georgobiani A. N., Kotljarevsky M. B., Kidalov V. V., et al., Study of ZnO and ZnSe film growth by Radical-Beam Epitaxy on ZnS and ZnSe substrates. Inorg. Mater. (Engl. Transl.) 28 (10) (1993) pp. 1245-1248. 3. Georgobiani A. N., Kotljarevsky M. B., Rogozin I. V. and Kidalov V. V., Characterization of the surface of ZnO layers and the ZnO/ZnSe interface in heterostructures prepared by Radical-Beam Gettering Epitaxy. Inorg. Mater. (Engl. Transl.) 31 (10) (1995) pp. 1249-1254. 4. Kondrat'eva E. N. and Kondrat'eva V. N., Catalytic recombination of electrically active centers as applied to determination of their concentration in the reaction zone, Sov. Zh. Fix. Khim. 20 (1946) pp. 12391247. 5. Kuz'minal. P. and Nikitenko V. A., Zinc Oxide: Preparation and Optical Properties, Moscow: Nauka, 1984, 166 p. 6. Nikitenko V. A., Tarkpea K. E., Nikul'shin S. F. and Kuz'mina I. P., F and F+ Centers in Zinc Oxide. Sov. Zh. Prikl. Spektrosk 47 (5) (1987) pp.834-838. 7. NakamuraS., NagahamaS, Iwasa N, et.ai, Ridge geometry in GaN multiquantum structure laser well diodes. Appl. Phys. Lett 69 (1996) pp. 1477-1481. 8. Kaiser U., Gruzintsev A. N., Khodos I. I. and Richter W., Substrate effects on the structure and optical properties of GaN epitaxial films. Inorganic Materials 36 (6) (2000) pp. 595-598.

GEM D E T E C T O R S FOR COMPASS FRANK SIMON, JAN FRIEDRICH, BORIS GRUBE, IGOR KONOROV, STEPHAN PAUL Physik-Department E18, Technische Universitdt Miinchen D-85748 Garching, Germany E-mail: [email protected] CEM ALTUNBAS, STEFFEN KAPPLER, BERNHARD KETZER, ALFREDO PLACCI, LESZEK ROPELEWSKI, FABIO SAULI CERN CH-1211 Geneve 23, Switzerland For the small-area tracking of the COMPASS experiment, GEM detectors with an active area of 31 x 31 cm2 are employed. These detectors use three cascaded GEM foils with asymmetric voltage sharing and Ar:C02 (70/30) as detector gas. The GEMs have a non-uniformity in gain of less than ±15% and achieve an efficiency of 99.0 ±0.1% and a spatial resolution of 46 ± 3 /xm for minimum-ionizing particles at nominal gain of ~ 8000. The narrow charge correlation (o>at < 0.1) between the orthogonal coordinates of the 2D projective readout improves the reconstruction capability for multiple hits. High rate tolerance and low discharge probability make the GEM detectors well suited for operation in intense muon and hadron beams.

1

Introduction

COMPASS 1 is a two-stage magnetic spectrometer designed to investigate the structure of hadrons using high-intensity muon and hadron beams from the SPS accelerator at CERN. The main component of the small-area tracking are ten GEM stations, each made up out of two detectors, one with horizontal and vertical readoutstrips, the second one rotated by 45° with respect to the first. For the operation in high-intensity beams, the central region of the GEM detectors can be deactivated to reduce occupancy. The GEM stations are mounted to the centers of the COMPASS large area trackers (straws or MWPCs) to provide high spatial resolution close to the beam. The Gas Electron Multiplier2 is a thin (50 /an) Cu-clad kapton foil perforated with holes of 70 /xm diameter at a pitch of 140 /Ltm. Application of a voltage between the two metalized faces of the foil leads to high electric fields in the holes and allows electron amplification. To reach high gains, several such foils can be cascaded.

259

260 2

T h e G E M Detectors

The COMPASS GEM detectors consist of three GEM stages. Since the detectors will be operated in high-intensity hadron beams with a background of heavily-ionizing nuclear fragments, a minimization of the discharge probability is crucial. To this end, the voltage sharing between the three GEM foils is asymmetric, with the highest voltage difference across the topmost foil3. In addition, the foils are subdivided into twelve parallel segments and a circular central sector with 50 mm diameter, which are individually connected to the HV distribution chain. This limits the available energy in case of a discharge. The disc-shaped central sector can be deactivated remotely to permit operation in high-intensity beams. The voltages are applied via a resistor network, so that only one external HV connection is necessary per detector. The two-dimensional orthogonal readout is realized with 768 parallel strips per coordinate at a pitch of 400 /am, the width of the strips having been adjusted to achieve equal charge sharing. The upper strips are 80 pm in width, the lower ones 340 (j,m, down from 350 fan in the first batch of the production to avoid short-circuits. As front-end readout electronics, the APV25-S0 chip 4 developed for the CMS silicon tracker is used. A protection circuit consisting of antiparallel diodes and blocking capacitors shields the chip from large electrical pulses caused by eventual discharges. •.WSLLZ,-,:,

drift honeycomb

jj*

™'*jj

mmEiar

3.0

GEM1 r

2.0

15.8

2.0

JOUt<

m

2.0 33

S0.0

TrLLUOXIlXIJJ.

Figure 1. Cut through a triple-GEM detector, all lengths in millimeters.

Figure 1 shows a cut through a COMPASS GEM. The gas volume (Ar:C0 2 (70/30)) is sandwiched between two honeycomb plates that provide the structural stability. The GEM foils are supported by thin spacer grids. The material budget is 0.72% X 0 averaged over the active area. Circular holes in the honeycomb plates of 35 and 50 mm diameter, respectively, reduce this to 0.42% Xo for the passage of the beam.

261

800

Figure 2. Cu X-ray spectrum from 15 kV X-ray generator. The energy resolution (fwhm of the full-energy K a peak) is A E / E = 20% (80 nm coordinate).

3

"""on,,, "late

1000 1200 pulse height [i

Figure 3. Gain map of a GEM detector (4x4 patches). The patch (1,1) is represented by fig. 2. The average pulse height is 970, the non-uniformity is less than ± 15%.

Laboratory Tests

Before the detectors are equipped with electronics for the installation in the COMPASS spectrometer, they are tested in the laboratory to ensure their functionality, as well as to determine their operational parameters. This is done with Cu X-rays (Ka line at 8.0 keV). An X-ray spectrum taken with a GEM detector is shown in figure 2. The energy resolution of 20% is indicative of the good homogeneity of the foils on a small scale (in the order of millimeters). By taking spectra at several positions over the active area of the detector, maps of gain (proportional to the pulse height of the spectrum) and of energy resolution are created. These maps show the uniformity of the detector over its full active area, see figure 3. The gain variations are less than ± 15%, similar to the variations in energy resolution. The absolute magnitude of the effective gain is determined by measuring the current on the readout and by normalizing the corresponding charge to the primary charge produced in the 3 mm drift region. 4

Performance in Particle Beams

Prior to their installation in the COMPASS experiment the GEM detectors were tested with a mixed secondary beam of 3.6 GeV/c protons and pions from CERN's PS accelerator. With the help of a simple zero-field tracking algorithm the efficiency of the GEM detectors for these minimum-ionizing protons and relativistic pions

262

0.2

0.4

residual [mm]

Figure 4. Hit residuals from the calculated tracks. The width of the distribution is "•/itt = 58 ± 2 ;mi, corresponding to a spatial resolution of 46 ± 3 pm.

100

200

300

400

500

cluster charge (80 urn strips) [a.u.]

Figure 5. Spectrum of the cluster charge for 160 GeV/c muons (Landau spectrum), the solid curve is a fit to a Landau distribution, most probable value 105 a.u.

(Pi = 25) was studied on a sub-millimeter scale for a gain of ~ 8000. The efficiency is reduced in areas covered by the 0.3 mm spacer grid supporting the GEM foils and in areas of segment boundaries (nominal width 200 /mi). In unobstructed regions the efficiency is 99.0±0.1% 5 . From the distribution of hit residuals relative to the track defined by a silicon micro-strip detector and a second GEM, the spatial resolution of the detector under study can be deduced by subtracting small-angle scattering effects and tracking uncertainties. The uncertainty of hit residuals is ahit = 58 ± 2 fiva. (see figure 4), resulting in an average spatial resolution of 46 ± 3 /xm for the two GEM detectors involved5. For the 2001 COMPASS physics running with a 160 GeV/c muon beam from the CERN SPS 14 GEM detectors have been operational. The cluster charge distribution for muons is shown in figure 5. No cut on the cluster amplitude was applied, leading to the noise peak around zero. The clear separation of the Landau distribution from the noise peak indicates high efficiency. Figure 6 shows the ratio of charge collected on both readout coordinates. This ratio has a mean value close to unity and a width of arat < 0.1, demonstrating almost equal charge sharing and excellent charge correlation. With the help of this sharp correlation multiple hits can be resolved by combining hits on projections into space points, thus enhancing the tracking in the COMPASS experiment. The quality of multiple-hit reconstruction using the pulse height correlation has been studied for combinatorial events with arbitrary multiplicity by means of a maximum-likelihood algorithm. Results obtained for the mixed proton and pion beam, which reflects the situation in an experiment with different particles species, are shown in figure 7. During the physics running the GEM detectors performed as expected

263

§ 1600 o o 1400 1200

[

T ' '

' ' I '

,,,,r

''-

l1

: r

600

^ -r

400 200

^'

\

J/

0 1 . . _„|~(-< 0.5

I

I

T H

V

. .

o" g 1-2

5

'

1000 r 800

j

C »o o 3 | o S

1

.... !

^i

0.8

* full event • single hit

"^^^

x^ "~ ""^

0.6

V

0.4

1.S

2

'"" 1

2

3

4

5

6

cluster charge ratio (80 um)/(350 um)

Figure 6. Charge correlation between readout coordinates. The solid curve is a fit to a Gaussian with aTat = 0.076 ± 0.001, data for muons from the 160 GeV/c SPS beam.

':

">.... 02

. N^-J_ . . . . 1

"~ ; ' 1

a 1800

7

8 9 10 hit multiplicity

Figure 7. Association of multiple hits. The single-hit efficiency describes the probability for correct reconstruction of one hit, the fullevent efficiency gives the probability for correct reconstruction of the complete event.

from the excellent test results and showed no problems at the design intensity of 2 x 108 muons/spill (4 s). No discharges were observed during 8 weeks of operation. The quality of the discharge protection for the front-end electronics can only be demonstrated when the detectors will be operated in a highintensity hadron beam during the COMPASS hadron running. Acknowledgements The GEM foils and the detectors are produced in CERN workshops. The project is supported by Bundesministerium fur Bildung und Forschung, Germany and by the Maier-Leibnitz-Laboratorium fur Kern- und Teilchenphysik der Miinchner Universitaten. References 1. The COMPASS Collaboration, "COMPASS: A Proposal for a Common Muon and Proton Apparatus for Structure and Spectroscopy", CERN/SPSLC 96-14, SPSC/P 297 (1996). 2. F. Sauli, Nucl. Instrum. Methods A 386, 531 (1997). 3. S. Bachmann et a]., CERN-EP/2000-151 (2000). 4. L.L. Jones et al, CERN-99-09 (1999). 5. F. Simon, "Commissioning of the GEM Detectors in the COMPASS Experiment", Diploma Thesis, TU Munich (2001), http://www.el8.physik.tu-muenchen.de/~fsimon/.

A R C H I T E C T U R E OF T H E C O M M O N GEM A N D SILICON R E A D O U T FOR T H E COMPASS E X P E R I M E N T BORIS GRUBE, RITA DE MASI, JAN FRIEDRICH, IGOR KONOROV, STEPHAN PAUL, LARS SCHMITT, FRANK SIMON, ROBERT WAGNER, MICHAEL WIESMANN Physik-Department E18, Technische Universitat Munchen D-85748 Garching, Germany BERNHARD KETZER CERN, CH-1211 Geneve 23, Switzerland The readout chain of the GEM and the silicon detectors of the COMPASS experiment at CERN is based on the APV25 frontend chip. The system utilizes optical fibers for data transmission and is designed to stand high event rates. Using the Multi readout mode of the APV 25, giving three samples per event, a very good time resolution of the detectors can be achieved. The high trigger rates require an efficient zero suppression algorithm. The data sparsification that is performed in hardware features an advanced common mode noise correction utilizing a combination of averaging and histogramming.

1

Introduction

The COmmon Muon and Proton Apparatus for Structure and Spectroscopy 1 (COMPASS) is a fixed target experiment at the CERN SPS. The detector is designed to stand high trigger rates of up to 100 kHz at beam intensities of up to 2 • 108 particles per spill. The Silicon and GEM detectors are used for small angle tracking. The double-sided silicon microstrip detector is 5 x 7 cm large, has a pitch of about 50 pm and 1280 + 1024 channels 2 . The GEM detectors have an active area of 30 x 30 cm and a two-dimensional readout with a strip pitch of 400 /mi and 768 + 768 channels 3 . For both detector types the APV 25 4 was chosen as the frontend chip so that they can use the same readout hardware. The APV 25 samples the signal amplitudes at 40 MHz. The samples are stored in an analog pipeline that can buffer the data for up to 4 /is. After receiving a trigger, the chip sends out the data in a sequential format, called 'frame', which contains the 128 analog amplitudes and header information. In the used 'Multi mode' the APV reads for each event three consecutive samples. The output is driven by a 20 MHz clock so that the readout of one event takes 21 ^s. By calculating the ratios of the amplitudes of the different samples, it is possible to determine the position along the assumed pulse shape

264

265

and thus to get precise timing information. Using a first implementation of this method, time resolutions of about 3 ns for the silicon detector and of about 15 ns for the GEM detector were achieved. 2

The readout chain

The readout chain consists of four parts: the frontend chip APV 25, the repeater cards, the ADC card and the GeSiCA readout module (see figure 1). The data processing is based on Field Programmable Gate Arrays (FPGAs). GEM frontend 2x6

FE

chips

APV 25

=n 1 >:iT

Rept.

•

/

•'i

r

APV 25

Rapt.

r

Figure 1. The common GEM and silicon readout system

The analog differential data output of the APVs goes via short flat-cable through a repeater card to the ADC card. In case of the GEM detector one ADC card handles all twelve APV chips of a chamber. In the silicon frontend one ADC card reads ten (for the 1280 channel n-side) respectively eight APVs (for the 1024 channel p-side). The analog signals are digitized by 10 bit differential ADCs. The zero suppression logic processes and sparsifies the digital signals. After a first formatting the data to 32 bit words they go via long optical fibers to the GeSiCA (GEM Silicon Control and Acquisition) board. For this data path the HOTLink protocol 5 is used. Aside from offering a high bandwidth of 40 Mbyte/s for data transfer, the optical fibers isolate

266

the ADC cards from the readout module. This is of particular importance for the silicon frontend, where the two ADC cards that read out one silicon detector lie on different potentials due to the depletion voltage 2 . The GeSiCA is a 9U VME module and is able to process up to four incoming HOTLink data streams, which is equivalent to 48 APVs or 6144 channels. After de-serialization, the data are buffered in a FIFO. Via a common bus the merger unit takes the data from the FIFOs, labels them with the event header it gets from the Trigger Control System (TCS) and writes the data to the S-Link card. Via the optical S-Link connection the data go with 100160 Mbyte/s to the readout computer. GeSiCA also encodes the TCS reference clock, the trigger and the reset signal which are provided by the TCS receiver and distributes these signals through the optical fibers to the ADC cards and the connected APVs. To allow easy configuration and status monitoring of the frontend electronics (ADC cards and APVs), GeSiCA provides an I 2 C interface6 which is accessible via the VME bus. An encoder transforms the I 2 C protocol so that it can be sent together with the trigger and reset signals over the optical fiber. The ADC card decodes the I 2 C signals from the fiber and forwards them to the APVs. In the year 2001 run of the COMPASS experiment 14 GEMs and 1 silicon detector with altogether 24000 channels were read out using 16 ADC cards and 5 GeSiCA readout modules. With zero suppression, trigger rates of up to 20 kHz were reached. 3

The zero suppression with common mode noise correction

The zero suppression is done using a threshold cut on the strip amplitude. Because fluctuations of the baseline of the APV are observed, the data have to be corrected for this 'common mode noise', before a threshold cut can be applied. This correction is performed in the hardware by histogramming the accumulated frequencies of the amplitudes within a window around the assumed baseline of the APV frame. The utilized histogram has 32 bins. The entry in each bin is the number of channels that have an amplitude bigger or equal to an assigned bin 'value', defined as two times the bin number, so that amplitudes in the interval from 0 to 62 are histogrammed. If the strip amplitudes lie in the proper range, the baseline of the frame is approximately the biggest bin value that has an entry bigger or equal to 64. This is because compared to this bin value at least half of the channels in the APV frame have an amplitude which is bigger or equal, so that this value is close to the median amplitude.

267

Because of the limited dynamical range of the histogram the amplitudes have to be shifted in the right region so that the histogramming can work. Therefore the hardware calculates the average pulse height of the pedestal corrected APV frame. If the frame contains no hits the average pulse height lies very close to the median amplitude. Signals obviously create a difference of average and median amplitude, but for sufficiently low occupancy the average pulse height gives an estimate for the median amplitude, close enough for the histogramming method to work. The channel amplitudes are shifted in a way that the average amplitude comes to lie on value 32, in the center of the histogram explained earlier (see figure 2).

[ADCch.]

Figure 2. The hardware zero suppression algorithm: (1) The pedestals are subtracted. (2) The average amplitude is calculated. (3) The data are shifted so that the average amplitude comes to lie at 32. (4) The histogramming that determines the signal baseline.

In the hardware the algorithm is implemented as a pipeline that processes the data of six ADCs in parallel. The pipeline has multiple stages7 that perform the above steps of the zero suppression algorithm.

268

In the high intensity muon beam of the year 2001 run of COMPASS the average occupancy of the GEM detectors was about 17 hits or 13 %. Under the same conditions the silicon detector has an average occupancy of about 12 hits or 9 %. Number of hit strips GEM

Mean = 17.29 RMS = 11.71

Number of hit strips silicon p-eide

Mean 3 10.75 RMS = 7.069

Number of hit stripe silicon n-slde Count F

Meen = 11.91 RMS = 7.49S

Figure 3. The occupancies of the GEM and the silicon detector in the high intensity muon beam

Acknowledgments The project is supported by the Bundesministerium fiir Forschung und Bildung, Germany and by the Maier-Leibnitz-Laboratorium fiir Kern- und Teilchenphysik der Miinchner Universitaten. References 1. The COMPASS Collaboration, "COMPASS: A Proposal for a Common Muon and Proton Apparatus for Structure and Spectroscopy", CERN/SPSLC 96-14, SPSC/P 297 (1996). 2. R. Wagner, Diploma Thesis, TU Munich (2001) a . 3. F. Simon, Diploma Thesis, TU Munich (2001) a . 4. M. Raymond et a.., "The CMS Tracker APV25 0.25 /xm CMOS Readout Chip", Proceedings of 6th Workshop on Electronics for LHC Experiments, Krakow, CERN/LHCC/200-041. 5. CYPRESS, "HOTLink Design Considerations", http://www.cypress.com/hotlink/index.html. 6. Philips, "The I 2 C-Bus Specification Version 2.1", 2000. 7. B. Grube, Diploma Thesis, TU Munich (2001) a . "Available at http://www.el8.physik.tu-muenchen.de/research/compass/

269

Medium and High Energy Physics Experiments Organizers: G. Westfall (Medium Energy Physics) L.

J. Pietraszko H. R. Schmidt P. Moissenz S. Tomassini W. J. Spalding P. Checchia A. Di Ciaccio A. Marin M. Giorgini B. Di Girolamo S. Movchan S. Hagopian S. Tentindo Repond A. Papanestis S. Braccini S. Paoli

Price

(High Energy Physics)

Performance of the Pre-shower System in the HADES Spectrometer The Time Projection Chamber for the CERN-LHC Heavy-ion Experiment ALICE Cathode Strip Chambers Data Analysis A Gas System for a Large Multi-cells Detector Run II Upgrades and Physics Prospects Detectors for a Linear Collider The ATLAS Muon Spectrometer US ATLAS Muon End Cap System Performance of the MACRO Limited Streamer Tubes for Estimates of Muon Energy Exploitation of ATLAS DAQ Prototypes for Test Beam and Lab Activities Cathode Strip Chamber Performance of the CMS ME1/1 Muon Station The Run2 D 0 Muon System at the Fermilab Tevatron The DO Central Tracker Trigger A Proposal for the Alignment of the LHCB RICH Detector Monitored Drift Tube Chamber Production at Laboratori Nazionali di Frascati A Database for Detector Conditions Data of Current and Future HEP Experiments

This page is intentionally left blank

P E R F O R M A N C E OF T H E P R E - S H O W E R S Y S T E M I N T H E HADES SPECTROMETER * J. PIETRASZKO, A. BALANDA, M. JASKULA, L. KIDON, R. KULESSA, E. LUBKIEWICZ, A. MALARZ, J. OTWINOWSKI, W. PROKOPOWICZ, W. PRZYGODA, P. SALABURA, E. WAJDA, W. WALUS, M. PIOSKON, T. WOJCIK M. Smoluchowski Institute of Physics, Jagellonian University 30-059 Krakow, Reymonta 4, Poland tel. (48-12)6324888, fax:(48-12)6342038, e-mail: [email protected] M. KAJETANOWICZ, K. KORCYL, AND A. SKOCZEN Nowoczesna Elektronika, Krakow FOR THE HADES COLLABORATION The Pre-Shower detector system of the HADES spectrometer is applied to electron identification with emphasis on fast hadron rejection at forward angles. The detector is operated in the self-quenching streamer mode (SQS) to simplify on-line recognition of electromagnetic showers. Stable electronics at low noise guarantee robust pattern recognition through the experimental runs. The construction and performance of the detector is presented.

1

Introduction

A High Acceptance Di-Electron Spectrometer (HADES) 1 ' 2 has been proposed at the SIS accelerator of GSI to investigate electron pairs produced in proton, pion and heavy ion induced reactions. The main goal of these studies is to explore in-medium modifications of vector mesons (p, LJ, (j>) properties changes at moderate temperatures and nuclear matter densities predicted by various models. The expected total yield of dielectrons from vector mesons decays is of the order of 10~ 6 per central Au+Au collision at lAGeV. Therefore an efficient electron detection system with large acceptance, high rate capability and highly selective multi-stage trigger scheme allowing pair detection at beam intensities of up to 108 particles/s is required. For HADES a fast electron recognition with two fold identification with on-line image processing is necessary. It is provided by a hadron-blind Ring Imaging Cerenkov Counter (RICH) and a Time Of Flight (TOF) scintillation wall accompanied • P A P E R SUPPORTED BY POLISH-GERMAN COLLABORATION FOUND 528/LN/96 AND PARTLY BY POLISH STATE C O M M I T T E E FOR RESEARCH (KBN GRANT NR 2 P03B 088 11, KBN GRANT NR 5 P03B 140 20).

271

272

by a set of electromagnetic shower detectors (PRE-SHOWER). Due to Mnematical reasons the TOF is able to accomplish sufficient separation between electrons and hadrons in the angular region 45°-88°. However, at polar angles below 45°, the time of flight measurement is not sufficient for the proper separation of electrons from pions due to a high pions momentum. According to calculation, described in ref.3, for timing resolution of at «15Q-200 ps the number of fake electrons is equal to rj/0fee «1 (45° < 0 <88°) and iifake «4-5 (13° < 6 < 45°) respectively. Therefore the electron identification has to be additionally improved by an introduction of the Pre-Shower.

2

Principle and detector construction

The PRE-SHOWER detector is based on the principle that at Sdnetic energy range below a few hundred GeV, electrons and positrons are the only particles which substantially loss their energy by radiation. In the picture (Figure 1) the idea lepton/hadron distinction based on Shower method is presented. A

Figure 1. Schematic representation of the lepton/hadron discrimination by means of the P R E - S H O W E R detector. Lepton/hadron discrimination is performed by comparing the number of particles measured before and after the lead converters. The instrument consist of 2 lead converters inserted between three wire chambers working in the self quenching streamer mode (SQS).

273

stack of three wire chamber layers (referenced later as: plane 1 = pre, plane 2 = postl, plane 3 = post2) separated by a layer of lead in between form the Pre-Shower detector. In order to perform number of particle measurement the wire chambers filled with the argon/isobutane/hepthane mixtures and operate in the limited self-quenching streamer mode (SQS) are used. SQS mode ensure that the produced charge depends only weakly on the particle specific energy loss and the quantity of the charge collected by means of the chamber should be directly proportional to the number of particles which enter the chamber. Counting and comparing the number of particles before and after the converters is appropriate way of electron identification. PRE-SHOWER wire chambers consist of a plane of equally alternately spaced potential and ground wires centered between two cathode planes. The construction principles for all chambers are identical except little difference in dimensions results from adjustment to cover the same solid angle with respect to the target. The details of the main components of the gas chamber (Figure 2) can be summarize as follows: - cathode pad plate, made of fiber glass with copper pads. Pads are organized in a matrix-like structure with rows and columns and are aligned with respect to the straight-line trajectories assumption. 20 mm thick aluminum frame is used to strengthen the pad plate, - two wire planes (ground wires, potential wires) glued and soldered to the rigid frames made from fiber glass equiped with integrated printed board circuits for wire connections, - the 0.7 mm thick stainless steel cathode plate with 20 mm aluminum support frame. The final pad configuration is shown in Figure 2. There are 32 pad rows, the number of columns varies from 32 in the lowermost to 20 in the uppermost row. The height of the pads changes from 3 cm (upper rows) to 4.5 cm (7 lowermost rows) and are aligned with chamber cells (i.e. 4.5 cm pads cover exactly 3 chamber cells). This configuration results in total number of 942 pads for one wire chamber and 16956 for the whole PRE-SHOWER detector. The wire plane is composed from sense and field wires at a distance of 4mm from a cathode planes. The space between ground and potential wires is 7.5mm. The sense and field wires are made from 25 mm gold plated tungsten and 125 mm Cu-Be, respectively. They are spanned on two separate frames made from rigid fiber glass epoxy resins. This scheme ensures very good electric isolation of sense wires at high potential (+ 2800 V) from field wires at zero potential. Typical leakage currents of the chambers are 20-40 nA at 2900 V. There are 89 sense and field wires stretched with a force 40 G and 160 G, respectively.

274

22 cm yS / "

cathode stainless steel

S u D si

/////////////

mi ^?iA plane 160 cm read-out

Figure 2. Schematic drawing of a wire chamber used as a counter in P R E - S H O W E R detector. Pad plane structure with geometrical dimensions in the left-hand side picture is shown. A set of parallel wires mounted (on distinct frames) symmetrically between two cathode planes in the right-hand side picture is depicted.

3

Lepton identification and hadron rejection, simulation and experimental results

The PRE-SHOWER detector compare the number of charged particles before and after the layers of lead to identify electrons. The simplest version of a condition which allows us to classify the particle is: ^ipost\ r\ *%pre

or

^IposVl r\ *°cpre

> F.

(1)

where Qpre, Qposti, QPost2- charges collected from first, second and third chamber respectively. Efficiencies for lepton identification was obtained from experiments with beams of electrons (850 MeV) and protons (2.1 GeV). The results are presented in the picture (Figure 3). In addition, the simulation results are enclosed. For the F=2 efficiency for electron identification is about 90% and only about 10% hadrons we misidentified as electrons. In addition a good agreement between simulation and experimental results was achieved.

275

1 electrons 850 MeV

0.5 a
tut

a>

fakes - protons 2.1 GeV

0.1

*

0.05 simulation data experimental data

0.02 0.01

2.2

2.4 2.6 parameter F

2.8

Figure 3. The measured and simulated electron identification and misidentification (fakes) efficiencies as a function of F parameter.

4

Conclusion

The PRE-SHOWER detector provides an effective way of leptons from hadrons separation. The detector together with dedicated stable electronic system 5 allows to perform online lepton recognition in the second level trigger of HADES. References 1. HADES, the New Electron-Pair Spectrometer at GSI J. Friese, for the

HADES collaboration Nucl. Phys. A 654, (1999) 1017c 2. P. Salabura at al., Acta Phys. Pol. B 27(1,2)(1996)421 3. C. Agodi et al., "The Time of Flight Wall for the HADES Spectrometer", IEEE Transaction on Nuclear Science, vol. 45, no. 3, June 1998 4. J. A. Kadyk, Nucl. Instr. And Meth. A 300 (1991) 436-479 5. Development of fast pad readout system for the HADES shower detector, A. Balanda at al., Nucl. Instr. And Meth. A 417 (1998) 360-370

T H E TIME P R O J E C T I O N C H A M B E R FOR T H E CERN-LHC HEAVY-ION E X P E R I M E N T ALICE H.R. SCHMIDT Gesellschaft fur Schwerionenforschung, Planckstr. 1, D-64S91 Darmstadt, Germany E-mail: [email protected] The ALICE TPC is a conventional TPC based on experience with previous TPCs used in heavy ion beams. However, the unpreceeded high particle multiplicities at LHC Pb+Pb collision has led in detail to many innovations in its design and construction. 1

Introduction

It seems that roughly every 5 years, forced by the augmented energy of heavy ions beams, a major Time Projection Chamber (TPC), setting new standards, comes into operation. This is documented in Table 1, which contains a comparison of design parameters of the NA49 2 TPC at the SPS (y/s = 17 GeV) , the STAR 3 TPC at RHIC ( ^ i = 200 GeV) and the projected ALICE * TPC at LHC (y/s = 5500 GeV) . From this table it can be seen that the ALICE TPC exceeds its predecessors in basically all aspects. This is forced by the expected high charged particles multiplicities of up to 8000 per unit rapidity, which is a factor of 5-10 higher than at RHIC. These unpreceeded high multiplicities are a major challenge both to the construction and the operation of the ALICE TPC. In this contribution we will show how the design of the ALICE TPC readout chambers, being basically conservative and based on the NA49 and STAR T P C s , are optimized to be able to handle the high particle load. 2

T P C Layout

The overall acceptance of the TPC is 0.9 < r\ < 0.9. To cover this acceptance the TPC is of cylindrical design with an inner radius of about 80 cm, an outer radius of about 250 cm, and an overall length in the beam direction of 500 cm schematic layout of the ALICE TPC is shown in Fig. 1 The TPC field cage provides a highly uniform electrostatic field in a cylindrical high-purity gas volume to transport primary charges over long distances (2.5 m) towards the readout end-plates. The field configurations is defined by a high-voltage (up to 100 kV) electrode located at the axial centre of the

276

277 Table 1. Comparison of the NA49, the Star and the planned ALICE TPC

parameter No. of channels gas

gas gain field cage drift voltage minimal pad size Luminosity [cm _ l i s " 1 ]

NA"4¥ SPS fixed target 1995 182k NeCC-2 (90-10) (vertex) ArCH4C02 (90-5-5) (main) 2 x 10* (vertex) 5 x 10 3 (main) W = 39"0cm L = 390cm H = 180cm (main) V = 27m3 „ 13.4/19.5 kV 200/175 V / c m 3.5 X 16 m m 2 = Kfi 0 m m 2

»1025

—ATICE LHC collider

STAR RHIC collider 2000 140k

570k

ArCH 4 (90-10)

NeCOa (90-10)

3.6 x 10 3 (inner) 1.3 x 10 3 (outer)

2 x 10 4

L = 420 cm R = 210cm V = 50m3 31 kV 150 V / c m 3.5 x 11.5 m m 2 = 39..S m m 2

2 X 10 2 5

L = 500cm R = 250cm V = 88m3 lOOkV 400 V / c m 4 x 7.5 m m 2 = 30 0 m") 2 0.5 - 1 X 10 2 7

SL-i-'fc

Figure 1. The ALICE TPC showing the central electrode, the field cage.and the end plates

cylinder. As drift gas a mixture of N e C 0 2 , as is currently used in the NA49 (9Q%/10%) and CERES (80%/20%) experiments at the SPS, is chosen. The readout chambers are basically conventional multiwire proportional

278

chambers with cathode pad readout as used in many TPCs before. In detail, their construction, however, requires to overcome significant technical challenges as discussed below. The overall area to be instrumented is 32.5 m 2 . The azimuthal segmentation of the readout plane follows that of the subsequent ALICE detectors, leading to 18 trapezoidal sectors, each covering 20 degrees in azimuth. The radial decrease of the track density leads to changing the requirements for the readout chamber design as a function of radius. Consequently, there will be two different types of readout chambers, the inner and outer chambers. Each outer chamber is further subdivided into two sections with different pad sizes, leading to a triple radial segmentation of the readout plane, with 557568 readout pads in total.

3

T P C challenges

The most obvious negative consequence of a high track density is the corresponding high occupancy of the readout channels. In the following we show that a simple increase in the readout granularity would be of limited help if not accompanied by a number of other measures. 3.1

Pad Size

A sufficient number of pads per charge cluster in terms of position resolution is 2-3. Thus an increase in the number of pads is sensible only if the induced charge from the (point-like) avalanche spread over no more the 2-3 pads. This can be achieved by reducing the distance between anode wire and pad plane. However, at a certain point the distance HV-GND gets critical. There are also other reasons why it makes little sense to decrease the pad size beyond a certain limit: the width of the charge cloud after 250 cm of drift is of the order of mm depending on the choice of the drift gas and voltage. I.e., any reduction of the pad beyond a certain size result in an oversampling of the track without any gain of information. 3.2

Time direction

The situation in time direction is similar: one could think of increasing the frequency of the time sampling, however, as diffusion occurs also in longitudinal direction this would result as well in an oversampling of the pulse. The choice of a shorter shaping time is limited by the fact that below 150-200 ns shaping time the signal/noise ratio becomes critical.

279 3.3

Optimization

From of the above considerations one is left with the following measures to optimize for best performance in a high density environment: 1) minimization of the diffusion, i.e. choice of a "cold" gas (NeC02, 90-10) and a high drift field (400V/cm). 2) choice of a minimal pad area (A = 30 mm 2 ) which still gives a reasonable signal; this implies a) the proper choice of the anode-pad distance (2 mm) to have the desired pad response function (PRF) and b) a high gain, because the faint signal from the small pad needs high amplification. This can be done both by gas and electronic amplification, however, to optimize the signal/noise ratio (S/N^20) a high gas gain (2 xlO 4 ) and low electronic gain (8mV/fC) is preferable. This choice should lead to a number of equivalent noise electrons below 1000. 3) For a given pad area the proper choice of the aspect ratio (4 x 7.5 mm 2 ) will further decrease the number pads occupied by a cluster. The principal reason for this is that the tracks are oriented in a preferred direction, i.e., radially. 4

Long Term Stability

In principle, the long term behavior of gaseous detectors is not testable, as the only halfways realistic test would require an exposure of the chambers at rates and durations comparable to the experimental conditions. For an expected running time of several years this is not possible for obvious reasons. One resorts therefore to short time tests with high intensity exposure to accumulate at least as much charge per unit length anode wire (where the amplification takes place) as in the experiment. In our case we exposed an anode area of about 1 cm 2 with a strong 55 Fe source for about 2000 hrs. The resulting anode current of 25 nA was monitored and found to be stable for the whole measurement period. The corresponding charge/unit length of the anode wire is calculated to be 60 mC. This has to be compared with an estimated accumulated charge of 1.1 mC per cm wire and ALICE year (1 ALICE year = 10 6 s). 5

Space Charge

There are two distinct sources of space charge in the TPC drift volume:

280

a) positive ions from primary ionization by a charged particle, and b) positive ion leaking back from the amplification zone into the drift space. Owing to the much smaller mobility of the ions as compared to the electrons a quasi-stationary positive charge will distort the drift field significantly. While a) is unavoidable and leads to distortions of the tracks of up to 0.5 mm for NeCC>2 as drift gas, b) could cause much larger distortions if the ion feedback is not sufficiently blocked by the gate. This is particularly dangerous at the high amplification of 2 x 104 in the present ROC's. First tests on prototype chambers showed indeed that ions leak back into the drift space even with gate closed. Two-dimensional calculations of the field configuration revealed the ion-leaks were located at the radial borders of the chamber, i.e., at the discontinuities of the otherwise regular gating grid structure. To circumvent the problem electrostatic "shims" were introduced to optimize the field geometry. The gating inefficiency was assessed by measuring the drift electrode current as a function of the gating offset voltage. At high gating offset voltage the measurement was limited by the sensitivity of the ammeter of « 10 pA. An upper limit of 0.5 x 1 0 - 4 for the gating inefficiency was deduced. This, together with an amplification of 2 x 104 results in less than 20 ions/cm track coming from the amplification and is of the same order of magnitude as the ions from primary ionization. 6

High Rate

So far, previous TPC's have not yet been operated both at high gain and at high track density. It is thus questionable whether under those conditions the chamber can be operated stably at all. A first test was performed at GSI employing a TPC readout chamber formerly used in the NA35 experiment. The chamber was irradiated with secondaries from a 12 C beam hitting a thick target. By varying the target thickness and/or the beam intensity track densities from overlapping events similar LHC P b + P b collisions could be reached. It turned out that the chamber could sustain several tens of fiA anode current without signs of instability. 7

Simulation Results

Even after the optimization steps described above one is left with an occupancy exceeding 50% at the innermost radius for an assumed multiplicity of dN/dy = 8300 plus background ( « 30%). Previous experience from the NA49 experiment demonstrated that the tracking efficiency is reduced dramatically

281

for occupancies above 20%. The situation, however, is different for a fixed target experiment as NA49 and an experiment in collider geometry where the track density decreases quadratically with the radius. The ALICE tracking group has adopted novel tracking algorithms which are based on local methods, i.e., the tracking starts at the outer parts of the TPC and proceeds to smaller radii. No global track model is needed in this case. Employing Kalman filtering leads to an acceptable efficiency, i.e., of 88% of all recognizable track are found with only 2% fake tracks. At present, the momentum resolution is evaluated for the TPC only, i.e., the connection to the other tracking detectors - ITS at small radii and TRD at large radii - is not included in the tracking algorithms. The momentum resolution Ap/p at lGev/c is found to be 2.4%, which is close to the expectation of the Technical Design Report 1. For high momentum tracks with p > 5GeV/c the resolution is at present > 14%, clearly not good enough for high pt physics. However, with the additional information from the other tracking detectors is it expected that the resolution for high momentum track is well below 5% 4 . The simulations yield a dE/dx resolution of 8-9% in the high track environment, while the resolution of a single, isolated track is « 5%, which is close to the optimum. Thus the particle identification properties of the TPC are as good as they can possibly be under the given circumstances. 8

Summary

We have shown that the measures taken to optimize the TPC readout chambers will allow to operate the ALICE TPC even under the highest anticipated particle load. The simulated performance indicates that momentum and dE/dx will be sufficient the reach the physics goals as formulated in 1. References 1. ALICE Collaboration, Technical Design Report, CERN/LHCC 2000-001 2. S. Afanasev et al., The NA49 Large Acceptance Hadron Detector, Nucl.Instrum.Meth. A430:210-244, 1999 3. H. Wieman, Recent Developments in TPC Technology and the STAR TPC at RHIC, this conference 4. ALICE Collaboration, Technical Design Report, CERN/LHCC 2001-021, p. 156

CATHODE STRIP CHAMBERS DATA ANALYSIS I. GOLUTVIN, Y. KIRYSHIN, K. MOISSENZ, P. MOISSENZ, S. MOVCHAN, A. ZARUBIN Joint Institute for Nuclear Research, 6 Joliot Curie, Dubna, Moscow reg., 141980, E-mail: [email protected]

Russia

Main cathode strip chambers (CSC) data analysis tasks are: calibration, reconstruction of the transmission function (transformation of CSC readout information to the coordinate of particle), track finding, optimal track parameters estimation and alignment. Methods for solution of these tasks (excepting calibration) are described. The influence of CSC geometrical parameters, sampling time and a number of signal readout samples, uncorrelated background, overflows and magnetic field to the spatial resolution are analysed. Proposed methods and algorithms are useful for the data analysis as well as for detector optimisation.

1

Introduction

CSC has been chosen for the endcap muon system of the Compact Muon Solenoid (CMS) because it is capable to provide precise spatial and timing resolution in presence of a high magnetic field and high particle background rate.1,2 Each CSC module consists of six layers to provide robust pattern recognition for background rejection and efficient matching of external muon tracks to internal track segments. The layer is a multiwire proportional chamber in which one cathode plane is segmented into strips running across wires. Radial strips and wire groups provide a natural

Reconstruction of the transmission function

An avalanche developed on a wire induces on the cathode plane a distributed charge of a well known shape which is defined by electrostatics.3 Charpak et al. have shown that by interpolating fractions of charge picked up by these strips, one can reconstruct the track position along the wire with precision of 50 urn or better.4 Charge cluster is a sequence of strips (> 2) with collected charge greater then noise and bounded (left and right) by strips without charge. Let Qc is a charge of the central strip in the cluster, Q is a charge of the left neighbour strip, Q,. is a charge of the right neighbour strip, x is the distance between muon and centre of the strip with Qc. There are two methods for x calculation: ratio method and fitting method.5

282

283

2.1

The Ratio Method

Let

a=

'

'

where W is a strip width, then x(ax) =

-w

+ WJp(a)d(a)

0)

(for the uniform distribution of x across the strip!) where p(a) is probability density distribution from a. Spatial resolution is

_ ORATIO-

gV(&-a) 2+ (a-Q) 2 +(a-&) r &'(&-&)-&'(&-a)-a'(&-e,)

(2)

where o is standard deviation of readout channel noise. For x = ±.5-W or x = 0

ORATIO=

vrf^V

(3)

where Q is the cluster charge, Q-q = Q r (q is a fraction of strip charge (from Q) vs distance between strip centre and muon). Sometimes Q,. is not correctly measured, due to the dynamic range of electronics. In this case the following formula can be used6 Qr-Qi

a=

^—=

(A\

and then spatial resolution is

V2(rV(8f + Q, -8 r + , -6,_,)2 +(6f - g ; ) 2 ORATIO

2.2

(Q;-Q'r)(QiM+QlJ

+

(Qr-Ql)(Q'r+l+Q;_1) + 2a'rQl-2Q'lQ,

( )

The Fitting Method

According to the fitting method muon coordinate x and cluster charge Q are calculated from the minimum of F{x,Q) = Jd{^-Qq(x)f

(6)

where n is a number of strips in the cluster, i is strip number in cluster, Qjexp is measured charge in strip i, qi is theoretical value of the measured charge in strip i . Spatial resolution is

284

(7)

One can see that both methods give the similar spatial resolution in the centre of strip and between strips for the narrow clusters. Radial structure of the strips gives the possibility to calculate x, Q, R, and A from the minimum of F(x,e,tf,A) = j r f e M P - 2 [ ( l - 2 ( * , + *2))
Magnetic Field Influence on the Spatial Resolution

Magnetic field causes ionisation clusters to spread along the CSC wires, which deteriorates a chamber spatial resolution. The radial component B r influence can be compensated by rotating CSC at the corresponding Lorentz angle a,. For calculation Or we analysed chamber resolution deterioration vs. track inclinations for differences Bf.7 The longitudinal component B 2 influence can be compensated by rotating the wires at the corresponding Lorentz angle o^. We have made special magnetic field scan of the CSC with rotated anode wires for calculation o^.7 2.4

Switched Capacitor Array (SCA) Influence on the Spatial Resolution (simulation and analytical estimation)

The pulses from the CSC chips are preamplified and shaped. The later is sampled by the SCA at 20 MHz. Dependence of the number of analysed samples on the spatial resolution has been studied using Maximum Likelihood Method for amplifier-shaper "KATOD-1M" (peaking time = 150 ns and pulse recovery time = 500 ns).8 For the single muon more than 5 samples are needed and in comparison with the sample/hold method (1 sample), spatial resolution is better by factor 1.6. Increasing of the background rates leads to distortion of the signal shape of the strip readout

285

electronics and as a result, to degradation of CSC spatial resolution. The effect depends on signal recovery time. Influence of the recovery time value on a probability to detect a background event has been estimated with the following equation R=l-e"tH

(9)

where R is probability of background event with rate of H to take place in t - the time interval.9 For the overlapped signals more than 7 samples are needed in order to separate signals. Spatial resolution degrades dramatically if the distance between particles less than 5 mm and time shift less than 100 ns. 8 3

Track finding

Muon track is the straight line (for the variables 9 and R) lies inside 50 ns time gap. Track finding steps are: • 3D straight line track-candidates searching (by using track road method for peaks of 3D clusters) inside 50 ns time gap. • Transformation searched clusters to coordinates. • 2D straight line track-candidates searching with %2 checking (by using track road method for the coordinates). • Missing coordinates searching. For CMS conditions we have found that uncorrelated background with a rate of about 100 kHz/eh creates insignificaijt influence (around 1%) on the muon track reconstruction efficiency. 4

Optimal track parameteis estimation

Non gaussian distribution of CSC residuals Ax is parameterised by P(Ax) = PrN(0,af) + (l-Pl)N(0,al),

(10)

where N is gaussian distribution, 01(02) is CSC single layer spatial resolution without (with) secondaries, pi is probability to get muon coordinate with precision d\. For the First Muon Station (ME1/1) p! is around .9. Non gaussian distribution requires to use Maximum Likelihood Method to achieve optimal track parameters estimation.5 Finally (for track model x=ao+aiz+a2z2) we have the following system of non-linear equations for calculation a0, ai, a2

286

i=l

M

(=1

i=l

aoZ^o-*()+a,X2,2a-*,)+«2Zz?o-*,)=Xx12ia-*/) i=l

(11)

6 1=1

i:

(»,--a0-fliZji-a2Z|2)2(g2 ""'l2 )

where

&, —

P\<* lw

2 (*, -ap-fliZi -a2zf

! | Q-Pi)gic

2

2

) ( a j -
^

2
^•2 _2
/7,CT 2

The nonlinear system is solved by iterative method. 5

Alignment

Analysis of MEl/1 adjustment error limits shows that shifts and rotation in XY plane have the most significant influence on the spatial resolution.10 Main parameters of the local (polar) coordinate system for layer i in the master coordinate system XYZ (Fig. 1) are: • pi - distance between pole of the local coordinate system and axis Z. • q>j - angle between p! and plane XY. • Yi - angle between polar axis and axis X. Y

*.

A Polar axis

figure 1. Position of the local (polar) coordinate system for layer i (reprezented by trapezium) in the master coordinate system XYZ (axis Z - orthogonal to XY)

Now it's about formulas for transformation the local particle coordinates to the old coordinates in master coordinate system. Let py - polar radius (Fig. 1), qjy -

287

polar angle of a particle, where i - number of layer, j - number of charged particle track then the correspondence py and q>y in the XYZ system are: w | ii+ i 2 P l .cos(a,.+< )_

M

p„=/>r\

or

p^+P/cos(a(+^)

ril

(12)

^i=7i-9i
M

p,.sin(a;+
P?

Now it's about function F for calculation pj, (ft and y,. For example the necessary computation are executed for two overlapping chambers, therefore i=l,2,...,12. Taking into account that muon track is straight line

F=^ [ ( 1 - ^ ? ?

+£l»j + Plcos(al+
+e?J))

+^(p^+er 2 ,+p 1 2 co^ 2 +<+er 2 .))-pf-p,co
p1sin(«1+<+^)i>|

11

+SXrO-M<+«.W. i=\ 1=2 (

.<* , „, , „.

old

Pu + e

(13)

Pi2sin(a12+flf + e ^ ) A P

P.2-+&12;

v A

,»« , „

v,r+/.-

p, s i n («, + < " )

(z — 7 1

where

&. = —

— , zs is Z coordinate of the layer i, N is a number of muon

(Z|2-*l)

tracks intersected the overlapping chambers, & and e* are uncorrected random errors due to measuring polar radius and angle, o p and av are measuring precision of the polar radius and angle. Parameters of the local coordinate systems are calculated from the minimum of F. Analysis of the corresponding system of linear equations shows that for unique solution it is necessary as minimum to fix two values each of p, (p and y along axis Z. Obviously these requirements can be replaced by the additional links. For the example, six reference points are laid in ME1/1, but unfortunately on the back part only. It is very important to orient as minimum two

288

layers one to another with high precision (< 50 um). Two days of CMS work at luminosity 1033 cm'V 1 are necessary for calculation parameters of a layer local coordinate system of single ME 1/1 chamber and 25 days are necessary for the chambers alignment. 6

Acknowledgements

We thank CMS colleagues for the useful discussions and critical remarks. References 1. CMS. The Compact Muon Solenoid. Technical Proposal. CERN/LHC 94-38, LHCC/P1, Geneva, Switzerland, 1994. 2. CMS. The Muon Project. Technical Design Report. CERN/LHCC 97-32, CMS TDR 3, Geneva, Switzerland, 1997. 3. Gatti E. et al., Optimum Geometry for Strip Cathodes or Grids in MWPC for Avalanche Localization Along the Anode Wires, Nucl. Instr. and Meth. 163 (1979) pp. 83-92. 4. Charpak G. et al., High-Accuracy Localization of Minimum Ionizing Particles Using the Cathode-Induced Charge Centre of Gravity Read-Out, Nucl. Instr. and Meth. 167 (1979) pp. 455-464. 5. Zubov K., Karjavin V., Movchan S., Moissenz P., Data Analysis for Cathode Strip Chamber, JINR Communication P10-99-118, JINR, Dubna, Russia (1999). 6. Movchan S., Moissenz K., Moissenz P., Cathode Strip Chamber Transmission Function and Single Layer Spatial Resolution for Clusters with Overflow, JINR Communication P10-2000-108, JINR, Dubna, Russia, 2001. 7. Movchan S., Moissenz P., The Method of Anode Wire Incident Angle Calculation of the First Muon Station (ME1/1) of the Compact Muon Solenoid Set Up (CMS), Particles and Nuclei Letters N4[107]-2001, JINR, Dubna, Russia, 2001. 8. Movchan S., Moissenz P., Khabarov S., The Influence of Readout Number of Samples in Analog Pipeline to Moun Spatial and Timing Resolution of Cathode Strip Chamber of the Compact Muon Solenoid Set Up (CMS), JINR Communication PI0-2000-183, JINR, Dubna, Russia, 2000. 9. Golutvin I. et al.,The Rate Capability of the CSC Readout Electronics, Particles and Nuclei Letters N4[107]-2001, JINR, Dubna, Russia, 2001. 10. Movchan S., Moissenz K., Moissenz P., Alignment of the First Muon Station (ME1/1) of the Compact Muon Solenoid Set-up (CMS), JINR Communication P10-2001-50, JINR, Dubna, Russia, 2001.

A G A S S Y S T E M FOR A LARGE MULTI-CELLS D E T E C T O R L. B E N U S S I , M. B E R T A N I , S. B I A N C O , F.L. F A B B R I , P. G I A N O T T I , M. G I A R D O N I , V. L U C H E R I N I , E. P A C E , L. P A S S A M O N T I , F . P O M P I L I , N . Q U A I S E R , S. T O M A S S I N I ° Laboratori Nazionali di Frascati, Via E. Fermi 40 1-00044 Frascati In complex detectors for high energy and nuclear physics experiments a high number of multi-cell gas detectors (in the range of thousands) are often used. In such cases a parallel, reliable, automatic, continuous monitoring gas system has to be built and operated in order to control and check the operational parameters (gas mixture composition, pressure, flux, gas losses...) that affect the detector performances and safety requirements. FINUDA is an experiment employing a large gas multi-cell detector and the gas system built and operated to flow its straw tubes is described. Straw tubes are used in FINUDA for charged particle tracking. The total number of tubes is 2424 (~ 2.5 m long, 1.5 cm diameter) with a gas volume of about 0.5 1 each. They are mounted on three super layers (each one composed by two staggered layers) one axial and two stereo. The straws are made of thin aluminized mylar and each one is flown individually. The gas system has been designed to offer modularity, parallel layout, immunity from individual straws leakage, safety and remote control. An automatic gas bottle inversion system has also been implemented and will be described.

1

Introduction

In complex experiments in high energy and nuclear physics large systems of gas multi-cell detectors are often employed operating in the non-recycling mode. The complexity of such systems and the need to operate them continuously for long periods of time, their location in areas often not directly accessible during data taking, or, in any case, in closed volumes, has required the development of dedicated systems to operate, control and monitor, from the remote, all the relevant operational and safety parameters related to the gas flux in several (of the order of thousands) independent channels. Such parameters, in fact, as mixture composition, pressure, temperature, flux rate affect the detector performances, while flux losses, especially in case of utilization of flammable gases, can generate, if not monitored, dangerous hazards. Hence, a gas system for a complex multi-cell gas detector has to provide, simultaneously, a continuous monitoring of the actual relevant parameters of the gas flux and, also, to act, or react, to any unwanted variations from the settled values of the operational parameters to readjust them within the admitted tolerances. Moreover, the system has also to signal any flux loss and, in case of exceeding the fixed safety "Corresponding Author. Tel. +39 6 9403 2797, e-mail: [email protected]

289

290

limits, to act consequently in an automatic way, for instance switching off the HV power supply, or, even, shutting down the incoming gas flux. Finally, gas systems for complex multi-cell detectors have to be built taking into account the maximum degree of parallelism, in order than any faults in a single channel, or group of channels, do not affect, or affect in a minimal way, all other channels normally operating. A gas system incorporating all such requirements should (or can) hence have the following general layout. - A storage and formation station, to be located, if possible, in an area always accessible. In such area the gas bottles needed to create the wanted mixture are put, provided with a mechanism of automatic switching from an empty bottle to a filled one for any of the gases of which the mixture must be done, in order to assure an uninterrupted gas supply. In this same station the mixture is formed for the whole detector: the controls on the mixture (composition, pressure, flux and temperature) are also made here. So, only one gas pipe is needed to feed the experimental area, and only another one for the mixture exhaust. - A splitting station, situated in the experimental area, in which the incoming mixture flux is directed to parallel controlled branches, for each of which (active) controls are made for pressure, input and output flux and temperature. Ideally, all single cells should be parallel supplied and controlled in an active way. When the involved cells run in the number of thousands, as in modern experiments, such a provision can not be, however, practically implemented and hence each of the parallel actively controlled branch groups several individual cells, according to a selected modularity. In order, however, to assure the maximum level of parallelism, passive controllers can be used in each single cell, such as calibrated impedances inserted in each channel 1 . A proper impedance selection can, in fact, ensure the detector normal operation in case of failure of individual cells. A gas system having the above summarized characteristics has been built and operated for the FINUDA experiment. The FINUDA experiment 2 at the DA$NE collider in Frascati has the largest, both in number and dimension, array of thin mylar straw tubes ever built and put into operation. Straw tubes are essentially single channel drift cells, where the anode wire is surrounded by a cylindrical cathode that works also as gas mixture container. FINUDA straw tubes detector is the outermost tracking device of the experiment: it consists of an array of straw tubes, arranged cylindrically in six concentric layers of 404 tubes each. The parameters of the tubes are: diameter 15

291

mm, length 2.7 m, mylar wall thickness 30// m, anode goldened-tungsten wire thickness 30/xm. The six layers are arranged in three super-layers of staggered tubes: the first axial and the following stereo, (tilted of ±13 degree respect to the experiment axis). The first layer is at about 1.2 m from the beam axis. 2

The gas system of F I N U D A straw tube array

The gas system of the FINUDA straw tubes foresees two components, but it can be easily extended to use more than two elements. The two gases presently employed are Argon and Ethane, used in a (50%-50% ratio). Argon and Ethane are supplied from high-pressure gas bottles and combined into the desired percentage in a gas room located outside the accelerator building and then the resulting mixture is delivered to gas distribution racks located inside the DA$NE hall, close to the detector. Since the detector has a large number of elements (2424), the gas system needed to feed them must have a parallel structure. For this reason, a gas system with six identical branches has been designed and built, each branch corresponding to each of the six layers of straw tubes. A mass flow controller (FC) is placed on each branch to guarantee the exact flux for each layer and 4848 calibrated impedances are mounted at the inlet and outlet of each tube in order to maintain the same condition independently on each channel. In fact, the use of impedances (passive controllers), decouples the individual cells from each other, and allows to maintain the wanted flux in the tubes even in case a single straw or a small group of straws are subjected to gas leakage. Each branch is splitted into seven parallel sub-sectors. The mixture coming from each sub sector is further split in two parts: hence each branch is (parallel) divided in 14 independent channels. Each of this independent pipes feeds a 32 channel divider to which about thirty straw are connected by means of the calibrated impedances. The gas outlet follows the same scheme of the inlet: thirty straws are connected, by means of the calibrated impedances, to the 32 channel divider, then the resulting outputs of fourteen of them belonging to the same layer are grouped two by two to form the corresponding seven sub-sectors, sent to the exhaust through a FC (one for each layer), that hence measures the output flux of the corresponding layer. In Figure 1 The overall scheme of the FINUDA gas system is shown. The high degree of parallelism turns out also advantageous to search for gas losses or isolate eventual fault straws while minimally affecting the overall system. If a single tube breaks down, the gas loss does not increase so much due the inserted impedances. Moreover, the resulting pressure loss is negligible, and the system keeps on working well. The pressure loss trough the impedance

292 versus gas flux has been studied both analytically and experimentally in order to select the optimal length and diameter for the impedances3. These studies allowed to build a system in which flux stability is maintained at a few % in spite of a total leakage of flux of up to 10% of the straws. The impedances are made of stainless steel capillary 30 mm long, 180 /zm inner diameter obtained by cutting with electric erosion a long pipe. The electric erosion technology has been used in order to preserve the section of the pipe. DAFNE HALL

Figure 1: The overall scheme of FINUDA straw tubes gas system.

Moreover due to the fact that we have the same impedances at the inlet and outlet of each straw we are able to periodically change the direction of flux to prevent micro-dust closing impedances. The control of the mixture composition is assured by the two mass Flow Controller (FC) (each for each gas of the binary mixture) in the gas room. These devices measure the mass flow and are self compensating for temperatures and pressure variations, therefore the mixture composition could only change due to a device calibration drift. For that reason FCs are periodically calibrated by means of a reference instrument also self compensating for temperatures and pressure variations. The calibration of the mass Flow Controllers of each layer is less critical, because does not affect the mixture composition but only the flow of a whole layer. Thus they are periodically calibrated according to mass flow controller in the gas room. The calibration constant are obtained by measuring the flow read by each FC for a given settled gas flux. The absolute errors of the calibration constants so obtained is better

293 than 1%. The calibration factors obtained during the calibrating procedure are then inserted in the remote control software. Each FC is connected to a digital readout and control unit (LabBox) and all the LabBoxes are connected to a PC. Two programs written in Labview provide all the operations. They allow to set the flow in each layer, to read pressure and temperature, to display leakages and to change the mixture composition. The flow meters at the outlet of each layer are also calibrated according to mass flow controller in the gas room. Since the flow at inlet and outlet of each layer is measured, the monitor software is able to continuously control the gas leakage. The monitor software also reads pressure and temperature and makes a backup copy for the offline analysis. The written files, readable also via web, allow to check off-line the system stability over long term periods. 3 3.1

Controls and quality checks Control logic

The gas flow is sent to each layer by means of Tylan Mass Flow Controllers. To make each mass flow controller work well, the pressure at the input of the controllers must be kept beyond a critical limit. Moreover it is necessary to provide a fairly constant pressure. If the mass flow controllers regulating the flow of each layer consume more gas than the mixer supplies, the pressure in the supply line will decrease and vice versa. To provide a constant pressure of about 80 mBar (the pressure value typically used) into the mixer, a feedback control has been developed. In the Labview control software there is a subroutine that reads the pressure in the mixer, and if the value is less than 80 mBar, it increases the set point of the two mass flow controllers in the formation station by a factor Af=k AP, where k is a constant; if on the contrary the pressure increases, it decreases the set point by the same factor. The scheme of the control logic is reported in Fig 2. 3.2

Mixture quality

Due to their particular structure, straw tubes cannot have a hermetic closure 4 . Therefore during their operation gas leakage can be observed with passive contamination of the operating mixture with atmospheric components (H2O, O2, N 2 ). It has been experimentally determined that in order to preserve the mixture composition a flow rate not less than 10 1/h per channel (layer) must be set. Moreover, gases coming from bottles (Argon and Ethane) pass through purification cartridges just after the primary decompression. For the Ethane line we use only oxisorb and hydrosorb cartridges while for the Argon we use

294 Exh;1US L

((

FC1

To layer 1 | )

From Bottles

Argon FC ^•'••"'""

• - - . .

../•-•*

MIXER

From bottles

Ethan FC

\\

y-

K

\ \ \\

' • • . . / ' • • • /

/

" " »

>

((

FM1

^

1 FM2

-• «• •

To layer 2 ) | From layer 2

1 FC3

,/' *p

|

FC2

. F'»»>Weri

1 FC4

>

^

1 FM3

^

1 FM4

U

To layer 3 \ \ From layer 3

+ ({

—> • *

To layer 4 M from layer 4

|

PC Control L

FC3

^ <<

^

1 FM5

To lajrcr 5 I | From layer 5

1 FC6

„ {(

*

1 FM6

To layer 6 1 I From layer 6

1

1

Figure 2: Control logic scheme. The simultaneous measurement of the inlet and outlet flux for each layer allows the on line remote monitoring of flow losses with a precision better than 1%

also an activated carbon cartridge. Pollution level is kept under 2% in volume and straw tube overpressure also does not allow the air backnow. Moreover mixture composition is periodically controlled by a mass chromatograph on the exhaust line and by calibrating the two mass flow controller in order to have always the same gas gain and preserve the detector performances. The performances of the used mass flow controllers are: accuracy ± 1.0% of full scale (FS), repeatability ± 0.2% (FS), linearity ± 0.5% (FS) and step response time (0-100%) 500 ms. 3.3

Safety and system protection

The safety of the whole system is based on hardware and software protections. Six excess flow valves, one for each layer, open when the pressure is higher then 250 mBar and need to be manually reset. Many sensor points provide alarm signals to the gas monitor and can close the gas supply if setting point limits are overcome. The primary purpose of this system is to allow a safe handling of the flammable gas and to measure the mixture losses. A second but equally important function is to protect the detector hardware. Finally, this system

295

serves to help insure the integrity of the data collected by the detector alerting the experimenters on shift if a condition arises which might affect the detector performance. Conditions monitored are 1) flammable mixture leak detection, 2) low main supply pressure, 3) high delivery pressure, 4) mixture composition. When a fault condition occurs, for example the mixture composition change beyond the settled limits, gas flow into the detector is turned off, closing the valves. An audible alarm sounds in both the gas room and the counting room, and one or more red lights on the interlock panels in both locations indicate the specific fault detected. Finally the high voltages of straws are turned off. In order that the unavoidable gas losses to not cumulate in high flammable concentrations in closed volumes surrounding the straw tubes location, a continuous cleaning of such volumes is provided with a inert gas (N2) flux that corresponds to a complete volume change every six hours. 4

Control software

An important part of the FINUDA gas system is the software developed for operating it. All the programs are written using the National Instruments Lab VIEW and are running on a PC with Windows as operating system. In the following, a short description of the written programs is given. 4.1

The SET program

The operator can easily change the percentage of the gas mixture and set the gas flow of each layer. The program will sum all the flow values that have been set and calculates the total gas flow that have to be set at mixture level. The actual gas flows are read from the Flow Meters on the mixer station and on the input and output of each layer and displayed on the PC screen. The set values are written on a log file. 4.2

The MONITOR program

The MONITOR program reads continuously some crucial parameters of the gas systems: Ethan and Argon flows at the mixer and gas flows at the input and the output of each layer, and temperature and overpressure values at the same points. These data are recorded each minute on a log file and some of them are elaborated and continuously displayed on the computer screen (see Fig. 3). Values plotted on the screen are: input and output flow and pressure values for each layer, flow loss of each layer, pressure at the mixer, Argon and Ethane flows, temperatures at each layer.

296

The MONITOR program also controls special and dangerous conditions that are depicted in the following. For each of these conditions, an alarm is generated that is registered on a log file. If one gas of the gas mixture has a null flow, also the low of the other gas is set to zero, as well the flows on the distributors, and an alarm is generated. When the gas flow is zero, a message is sent to the program that controls the high voltage of the straw tubes that are switched off for safety reasons. When the internal pressure that is read in the mixer or in each layer becomes higher than the settled limit, a warning, and then an alarm, are generated: when the alarm is generated, the gas flow is closed.

Figure 3: Plot of the information shown on line on the PC screen. The settled gas flux, and the actual temperature, pressure, inlet and outlet fluxes and flux losses of each layer are continuously displayed

4-3 The off-line analysis The values recorded on file are very important because they allow to monitor the system behavior over long periods. A set of easy tools exists for plotting temperature, overpressure and gas low values read at various points (on the

297

mixture and on the input and output of each distributor). Furthermore, gas leakage can be plotted for each distributor, (see Fig. 4 for an example of quantities stored in the files.)

0.6 .

0,5

r

;

•:

' - ' : '

:

:

'

•

'

-

•

.

0.2 0.1 0

!

:.

t

-

0.4 0.3

:

:

•'

•

•^jajy i|ii I;!:•: ; i :j; !i-:Ji ; .ii.

i-

I plpl -;I

1

,

1

1

.

I

.;

[''','• . '

.

f • ,,- i i

-i

I

i ; i

'.

1

.

i :.i

I

1

iayer2..(inputFlow-outputF!ow)finputFtew..vs..time

tey©rS..(inputFlow-ou^!UtFlow)/inputF]ow..¥S..tim®

Figure 4: The fractional iow losses of layer 2 and 5 recorded for the whole 2000 year. The different regions correspond to different settled values and to maintenance, test or data taking periods. As seen, in this way it is possible to study the long term behavior of the system, checking for any systematic variation or drift of working parameters

5

Gas Cylinder Inversion System

The architecture of the inversion system is shown in Fig. 5. The system allows the ramp inversion from one bottle to the new one when either the weight or the pressure of the used bottle drops below a fixed value. A cross

298 M«ss fx~v*

f5^<~.

V«>*!g

J

S

WAV* WAV:?

,_~,

1 #<#?»»*. } J t ,

kr.euua' ir pressure Solenoid eas Cylinder. Weight transducer

Figure 5: Scheme of the gas cylinder inversion system.

check of the cylinder weight and pressure is done. Cylinder pressure for both right and left ramps is measured by a 4-20 mA pressure transducers and it is reduced in a single stage de-compressor to 10 bars. A pneumatic valve for each ramp controls the gas passage to the outlet single stage pressure decompressor where gas pressure is further reduced to 5 bars. Finally a mass flow controller is used to set the out going flux. This allows controlling the gas mixture ratio. The inlet pressure transducer and the mass flow controller are interfaced to the control unit which also operates the solenoid valves to close or open the pneumatic valves. The controller is built around a single board embedded computer (TERN type processor) with a 40 MHz CPU. An on board 12 bit multi-channel ADC equipped with signal conditioning instrumentation amplifiers is used to read the 4-20 mA transducers and 0-5 V signal from the mass flow controllers. The CPU uses a 12-bit DAC integrated circuit to set the value on mass flow controller. It also employs semiconductor relays in order to operate on solenoids for closing and opening the pneumatic valves. The bottle weight is measured by two extensimetric cells and a digitalized signal is sent

299

to the embedded controller by means of a weight transducer. User interface is provided through a touch sensitive LCD and the gas cylinder status is displayed on the control PC. 6

Conclusions

The gas system developed for a complex multi-cell detector, the FINUDA straw tubes detector, has been described. It incorporates a high degree of modularity and parallel operation for the gas flux of the 2424 straws tubes, employing both active and passive controllers. The straw gas system is automatically operated and PC controlled. It allows to safely flux the wanted (flammable) gas mixture for long periods without direct operator intervention, handling also emergency situations. The gas flow system software allows also the automatic ramp inversion and can be monitored and even controlled remotely through the Web. The operation of the system for long periods of operation (weeks, months) is also saved on files storing all relevant information, to allow stability checks to be performed off line. The system, in operation almost continuously in the past two years, proved to be stable and reliable. ACKNOWLEDGEMENT We acknowledge S. Sarwar for his activity in the early phase of this project, A. Mecozzi for the involvement in the construction of the mechanical frames, D. Pierluigi and A. Russo for the continuous assistance in running the gas system. References 1. G. Anzilino et al., The LVD tracking system chambers Nucl. Instrum. Methods A 329 (1993) 521. 2. M. Agnello et al., The FINUDA Technical Report LNF-95 95/024/R. 3. S. Bianco, A. Mecozzi, Gas impedance measurements for straw tubes: test setup and status report, FINUDA note 25/LNF/PUB/95 4. L. Benussi et al, Design and Realization of a Simple Gas Leakage Test Setup for the Finuda Straw Tube Detector at Dafne, FINUDA note 39/TO/DC/1996

R U N II U P G R A D E S A N D P H Y S I C S P R O S P E C T S

W. J. SPALDING Fermi National Accelerator Laboratory for the CDF and DO Collaborations This article describes the present status and physics prospects for Run 2 at the Fermilab Tevatron accelerator. The accelerator complex and both the collider detectors, CDF and DO, have completed extensive upgrades resulting in a significant increase in luminosity and physics capability.

1

The Accelerator Complex for R u n 2

The Tevatron accelerator at Fermi National Accelerator Laboratory is the highest energy accelerator in the world, colliding protons and antiprotons at a center of mass energy of almost 2 TeV. Since the completion of Run 1 in 1996, a new 150 GeV accelerator, the Main Injector, has been built to inject the proton and anti-proton beams into the Tevatron. The number of bunches has been increased from six in Run 1 (with a bunch spacing of 3.5 /isec) to 36 for the start of Run 2 (bunch spacing 396 nsec), with a further upgrade to 100 (132 nsec) later in the run. In addition the energy has been increased from 1.8 TeV to 1.96 TeV, which although apparently quite modest, will nevertheless result in a 30-40% increase in the number of ti events produced. Commissioning with the Main Injector started with a short "engineering run" in October 2000, and continued with the beginning of Run 2 in March 2001. In 2001 the peak luminosity was typically about 7 x 1030 cm~ 2 s _ 1 . During 2002 the lumninosity is expected to increase towards the initial design goal of 8 x 10 31 c m ~ 2 s - 1 . By that time a second new accelerator, the Recycler, will be commissioned. The Recycler is an 8 GeV ring of permanent magnets housed in the same new tunnel as the Main Injector. Its role is to collect and re-use the anti-protons remaining at the end of each Tevatron store, providing an additional boost to the luminosity. Both CDF and DO predict that the silicon trackers will suffer significant radiation damage after about 4 f b - 1 , so Run 2 is divided into two sections, 2a and 2b, with a shutdown at the end of 2004 to allow the replacement of the silicon detectors. By that time the accummulated luminosity will be about 21b-1.

Figure 1 shows the instantaneous and cumulative luminosity projected for Run 2. The goal for Run 2b is to operate at a peak luminosity of 5 x 1032 c m " 2 s - 1 , accumulating more than 4 f b - 1 per year. The Run 2

300

301 6

16

Figure 1. Projected luminosity for Run 2: peak luminosity in E32 cm integrated luminosity in f b _ 1 (right)

2

s * (left) and

total is 15 fb *, sufficient luminosity to allow a thorough search for the Higgs boson. 2

The U p g r a d e s t o C D F a n d DO

Both CDF and DO have completed major upgrades to the detectors since Run 1. In particular DO has installed a 2 Tesla superconducting solenoid and a new tracking system inside the existing liquid argon calorimeter. CDF has also completely replaced the tracking system for Run 2, and both CDF and DO have upgraded the trigger and DAQ systems to accommodate the higher luminosity and decreased bunch spacing. Other upgrades include new plug calorimeters, extended muon coverage, and a time-of-fiight system for CDF, and an improved muon system and new pre-shower detectors for DO. There are talks on several of these upgrades at this conference1. 2.1

Tracking

While the physics goals for the experiments are similar, they have chosen quite different solutions for the tracking detectors. Both employ silicon detectors at the inner radii. The DO silicon tracker includes six short four-layer barrel sections with disk detectors between each barrel to provide forward coverage. Additional disc detectors at each end of the whole assembly provide tracking to |-^71 < 2.5. The CDF silicon tracker is arranged entirely in a barrel geometry, with a total of seven layers in the central region (|?j| < 1) where the outer tracking chamber provides coverage, and eight layers out to |r/| < 2 for standalone silicon tracking. The innermost layer for CDF, LOO, is supported directly

302

Figure 2. The CDF detector with the silicon tracker being installed, and the end-plug calorimeter ready to close, and the DO detector after roll-in to the collision hall, showing the extensive anion system.

Figure 3. Assembly of one of the three CDF silicon detector barrels, and the DO silicon detector. For DO, the six barrel-disk assemblies are installed as combined units.

on the beampipe at a radius of only 1.5 cm. This layer employs radiation-hard single-sided detectors connected to the readout chips via very thin Kapton cables. The primary role of LOO is to improve vertex resolution for softer B decay tracks which are degraded by multiple scattering. Both experiments use double-sided silicon with a mix of small angle and 90-degree stereo, with 722,000 channels for CDF and 790,000 for DO. The

4s > >

v

«;

m%

uf'

8*0 * f^4 ".<&&*» 5 * f^f

) *57y5 %s

Figure 4. J/f and lf« signals seen in DO using silicon stand-alone tracking.

Figure 5. The COT assembly and an on-line display of a 3-jet event.

readout electronics is similar, but while DO uses the SVX2 amplifier+ADC chip, CDF uses the SVX3 chip which allows simultaneous digitization and readout of a previous event while acquiring the silicon signals for a new event. The silicon trackers are working well. Figure 4 shows reconstructed J/ip and Ks signals seen in DO using silicon stand-alone tracking. Outer tracking for CDF is performed by an open cell drift chamber, the COT. This chamber has 30,240 wires arranged in 96 planes (in eight super layers), with an outer radius of 132 cm. The track resolution is 180 fan. Because DO added a solenoid and tracking system inside the original calorimeter, the tracking system is necessarily more compact than in CDF. The outer tracking in DO is provided by a scintillating fiber tracker, the CFT, with an outer radius of 51 cm.

304

Figure 6. The Z -*• e+e~ mass distribution aed W -* ev transverse mass from CDF, and a Z —• /*+it~ candidate event in DO showing full efficiency of the central and forward muan tracking.

The CFT is read out via Visible Light Photon Counters, VLPC, which are used for both the tracker and the preshower counters. The VLPCs have a remarkably high quantum efficiency of around 60% and provide excellent resolution of individual photoelectron peaks. The downside is that they must operate at 9-deg Kelvin, and thus require a cryogenics system. At this time the CFT is only partially instrumented with readout boards. The electronics will be complete by the end of 2001. 2.2 Calorimetry, Muon Systems and Particle ID DO continues to use the liquid-argon calorimeters from Run 1, upgraded with new electronics for the new bunch-spacing. CDF has added new scintillating tile-fiber end plug calorimeters to improve coverage and resolution in the forward direction. Both experiments use scintillator and drift-tube layers for muon identification, and have in particular upgraded the coverage in the forward direction. dE/dx information from the tracking layers is used to tag kaon and proton tracks - particularly important in B-physics. CDF has installed a new scintillator time-of-flight system between the COT and the solenoid to provide v/K separation up to about 1.6 GeV, complementary with the dE/dx range.

305

2.3

Trigger and DAQ Systems

Both CDF and DO have implemented new three-level trigger systems which are very similar in design philosophy. The initial two levels are implemented in hardware+firmware and the third level in a Linux pc farm. Primitive objects such as "track", "muon", "jet" or "missing-energy" are identified at Level 1, with extrapolation and matching between detectors at Level 2. For CDF, Level 2 adds the silicon tracking and impact parameter information using the SVT processor. The transverse impact parameter in SVT has an r.m.s. width of 50 (im - a combination of the size of the beam spot and the silicon tracking resolution. Typically a trigger for hadronic B decays will cut at an impact parameter of about 120 jzm. DO is developing a similar trigger scheme for implementation next year. The Level 3 farms provide essentially full event reconstruction and a tape logging rate of several tens of hertz. One difference between the two experiments is in the rate capability at Level 1. DO will operate with a Level 1 rate of 5-10 kHz, whereas in CDF a fully pipelined DAQ at Levels 1 and 2 allows a Level 1 rate of 40-50 kHz. During summer 2001 the Tevatron delivered about 12 p b - 1 , and the experiments collected "engineering" signals for calibration of the detectors and the reconstruction programs. Ks, J/i>, and A signals provide samples for tracking studies and efficiency measurements, and are precursors to physics signals. Figure 7 shows signals from CDF for 4 p b - 1 of data, including the first B signal from Run 2. 3

Physics Prospects in R u n 2 - with 400 p b - 1 (2002), 2 fb" 1 (2004) and 15 f t r 1 (2007)

The Tevatron collider provides a wealth of physics data over a very broad range of topics - both sharpening the precision of measurements within the framework of the standard model and searching for new phenomena. A series of workshops was held in the last year to focus attention on the physics of Run 2 2 . Already, by the end of 2002, the delivered luminosity of 400 p b - 1 will be several times that delivered in Run 1. QCD, B-physics, top-physics, Higgs, and SUSY will be energetically pursued, and new physics topics, will be accessible with the upgraded detectors. The Bs mixing parameter xs will be measured by CDF by triggering with SVT on hadronic B decays, with the vertex resolution enhanced by LOO and particle ID from the new TOF system. With only 400 p b - 1 CDF will be able to cover the range xs < 35. The

308 .,. v ... 5 ..„ r „.,.... 1 .„. ! „„j r

"*!*Svi(ft*ws,s3*rf*Tj**M^&s^

-<

J

i>A

5* ottcuwatt <ws sr- « ragmen: a n as invariant Mass Ktev'i

1

»£.#•

f

> ?

"e^s^iiJ

<™_4_j™J_

&.«-:'

8

a

'

i

*»

11i

IJP 4vHi I I, WW * r x t ft t fe i » • f

! i _*,^

:T

S.33

$.4

S45

SJS

*§ §.es s.* 5 » e.a SUM s a t

7. A (top left), Ks (bottom left), J/ip (top right) and the first B signal from Run 2, J/V tf+, from CDF.

full program of B physics at CDF and DO will pin down many of the CKM parameters in the standard model. As the luminosity increases through Run 2a and into Run 2b, Higgs and SUSY searches will become the major focus. Increased precision of the top and W masses will improve the limits on the mass range allowed for a standardmodel Higgs and with the full luminosity of Run 2b direct searches will cover the range up to a Higgs mass of 190 GeV. JFor Higgs masses below about 140 GeV the dominant decay mode is H ->• 66. Unfortunately this mode suffers from significant QCD background, so the plan is to search for Higgs produced in association with a W or Z, which can then provide a clean trigger and background rejection. Above 140 GeV, the dominant decay is if -4 W+W~. Combining the projected data sets for CDF and DO, a Higgs at 115 GeV can already be excluded at 95 percent confidence level with 2 fb _ 1 per experiment. Discovery, at the ha level, requires the higher luminosity of Run 2b.

307

Figiire 8. Limits on the mass of a standard model Higgs from measurements of the W and top masses (for 2 f b - 1 and 10 fb - 1 ), and prospects for direct discovery of the Higgs versus mass and luminosity.

4

Conclusion

Eun 2 has just begun and the detectors are starting to take their first physics data. This is an enormously challenging effort, but the prospect for newdiscoveries are very exciting. The luminosity will increase dramatically over the next few years and we anticipate significant results, maybe discoveries, and very likely some surprises before the LHC takes the lead at the high energy frontier. Acknowledgments Thanks to many people from CDF and DO who contributed to this talk. CDF and DO rely on the hard work of the technical staff at Fermilab and the participating institutions, and the support of their funding agencies. References 1. Related talks in these proceedings E. Kajfasz: Production and Operation of the DO Silicon Microstrip Tracker A. Bean: Design of an Upgraded DO Silicon Micmstrip Tracker

308

A. Patwa: DO SciFi Based Tracker and Pre-shower Detectors S. Tentindo Repond: The New Online Central Track Trigger of DO S. Hagopian: Run II Muon System of the DO Detector S. D'Auria: Commissioning and Operation of the CDF SVX Detector E. Palmonari: CDF II Silicon Tracking System M. Bishai: CDF Silicon DAQ System and Front-End Electronics I. Fiori: CDF Online Silicon Vertex Tracker S. Cabrera: Run 2b CDF Upgrade of the Silicon Vertex and Layer 00 Detectors 2. Physics at Run 2 Workshops (http://fnth37.fnal.gov/run2.htm) B Physics at the Tevatron: Run II and Beyond Fermilab-Pub-01/197 Report of the Tevatron Higgs Working Group, hep-ph/0010338 Report of the SUGRA Working Group, hep/ph/0003154

D E T E C T O R S FOR A LINEAR COLLIDER P. CHECCHIA I.N.F.N, sezione di Padova, via Marzolo 8 35131 Padova, Italy An overview of the detectors foreseen for an e+e~ Linear Collider is presented. The Physics program of such a type of machines requires high precision detectors. The solutions proposed for the European project (TESLA) are presented in detail with short recalls to the alternative solutions proposed in the other regions.

1

Introduction

At present, three e+e~

Linear Collider projects with center-of mass energy

up to about 500 GeV in a first step and to about 1 TeV in a successive stage are under study in Asia 1, Europe 2 , and North-America 3 . In this note the main LC Physics items are given and the different machine conditions are summarised giving more details for the European project TESLA 2 . The Detector for TESLA is then illustrated and examples of the different approaches followed in the other regions are given. 2

Physics at a LC

In this section the most relevant Physics items of an e+e~ Linear Collider are briefly illustrated in order to emphasise the requirements to the Detector. 2.1

Higgs profile

If the Higgs exists, the clean environment of a e+e~ LC allows an accurate study of its production and decay properties in order to obtain experimental evidence of the Higgs mechanism role in the electroweak symmetry breaking. A precision measurement of Mu is possible both in the case of a light (< 140 GeV) or of a heavy Higgs. The precision expected after combining all the possible measurements at y/s = 350 GeV for an integrated luminosity of 500 fb" 1 goes from 40 MeV to 80 MeV in the MH range from 120 to 180 GeV. An excellent momentum resolution is mandatory for such a precision. In the Standard Model the couplings to the fermions and to the gauge bosons depend only on their mass. As a consequence at a LC the measurements of the Higgs branching fractions (Fig. 1) and decay widths together with the precision measurements of the production processes (Higgs-strahlung and WW fusion, see Fig. 1) give an unique way to measure the Higgs cou-

309

310

Figure 1. Left): The expected precision on the experimental measurements of SM Higgs branching ratios as a function of the Higgs mass. Right): Missing mass distribution in bivP events for 500 f b - 1 at y/l = 350 GeV.

plings and, from the comparison with the SM expectations, to exploit the predictions of the SM supersymmetric extensions. In order to make these measurements with the required high precision, an excellent vertex detector and an excellent jet-energy reconstruction are necessary.

2.2

Supersymmetry

If LHC will discover SUperSYmrnetry, the LC is the best machine to study SUSY characteristics and to determine its fundamental parameters. In this context, a fundamental role is played by the beam polarisation. Here only a few very significant measurements are recalled. The sleptons mass and width can be obtained by the end point in the energy spectrum of the decay lepton and by the measure of the pair production cross-section at threshold. The lepton spectrum can be used for the sneutrino mass determination and leptons or jet distribution together with threshold scans can provide precise measurements of the chargino and neutralino masses. In a similar way the jet energy reconstruction can be used in the chargino studies. Several channels are characterised by the presence of missing Energy and therefore the detector hermeticity is a fundamental requirement for all the SUSY measurements.

311

y/8 (GeV) Gradient (MV/m) C (10Mcm-2s-1) RF-freq. (GHz) Bunch spacing (ns) Bunch per train Beamsstrahl. (%)

TESLA 500 (800) 22 (35) 3.4 (5.8) 1.3 337 (176) 2820 (4886) 3.2 (4.4)

NLC/JLC-X 500 (1000) 50 2.2 (3.4) 11.4 1.4 190 4.6 (8.8)

JLC-C 500 34 0.43 5.7 2.8 72

Table 1. Parameters of the European (TESLA), American (NLC) and Asiatic (JLC) projects.

2.3

Electroweak precision measurements

A LC can provide a very precise determination of the top mass (5mt ~ 120 MeV ) by measuring the it production cross-section near threshold. Since the dominant t —> b + W channel produces events with 6 jets having high track multiplicity, the efficiency for track reconstruction must be extremely high. High luminosity runs at low energy (Z° pole, W-pair production region) with beam polarisation can ensure very high accuracy on the SM parameters. In particular the left-right asymmetry at the Z° could be obtained by combinations of the cross-sections of different polarisation states and the W mass can be precisely determined with accurate measures of the W-pair production cross-section near threshold. Precision in cross-sections measurements requires a good luminosity determination. 3

Machine C o n d i t i o n s

Projects for e+e" Linear Collider are studied in America (NLC), Asia (JLC) and Europe (TESLA). The two first projects are based on warm highfrequency high-field cavities, while the latter relies on superconductive cavities. The most relevant characteristics of the three proposed solutions are shown in table 1. In the following the TESLA conditions are given. 3.1

Data Acquisition

System

The European project has a long bunch spacing with obvious advantages from the point of view of detector design. The total length in time for a 2820-bunches train is 950 (J,S and the time between two trains is 199 ms. This implies, taking into account the Physics and background rates and the

312

detector structure as described below, an expected Data volume per train of 220 MBytes. With these conditions a Data Acquisition System with the following elements can be proposed: 1) hardware trigger unnecessary; 2) data in pipeline for 1 ms, no dead time; 3) 200 ms for pipeline ready; 4) software event selection. The overall throughput for event building is moderate if compared with the LHC requirements and it should not be a problem. 3.2

Beamstrahlung

The interaction of one beam with the high electromagnetic field produced by the other implies a significant increase of the luminosity but also an intense emission of hard photons (beamstrahlung) with a mean energy loss in the e + e~ center-of-mass of 3.2% (TESLA) and a long low energy tail. As a consequence, to determine the energy dependence of cross-sections at threshold the Luminosity spectrum must be measured 4 . 3.3

Background

The secondary interactions of beamstrahlung photons produce mainly e + e~ pairs with typical energies of a few GeV. The pairs in presence of the detector magnetic field move along the beam pipe but they can be deflected when they reach the quadrupoles. Then they can hit the beam pipe or the quadrupoles creating a large number of secondaries. To shield the detector from these backgrounds a tungsten mask system has to be designed as shown in Pig. 2.

figure 2. Left):The TESLA detector mask system. A portion of the Inner Mask and of the Tungstea shield axe instrumented (LCAL: Luminosity Calorimeter, LAT: Low Angle Tagger). Right) Hits per bunch crossing (BX) produced by pair background on the vertex detector layers for different energies and detector magnetic fields.

A portion of the masks will be instrumented (LCAL and LAT) for Luminosity measurement and low angle beam scattering detection.

313

There is an important interdependence among pair background, magnetic field intensity and minimum radius of the first vertex detector layer ( and consequently on the accuracy on the impact parameter determination) as shown in Fig. 2. With a 4T magnetic field and a radius of 1.55 cm the hit density (<0.05 hits/mm 2 ) is below critical levels being still acceptable with a 3T field. Other background sources (hadronic interactions of the beamstrahlung photons, radiative bhabha's, neutrons, muon induced background) are expected not to create problems to the detector and to the feasibility of background subtraction as described in detail in 5 . 4

Detector General Concepts

As a consequence of the Physics program and of the machine conditions a Detector at an e+e~ LC must be designed taking into account the following points: a) an excellent vertex resolution to identify heavy flavours (b,c,r) as required to measure the Higgs couplings gHff'i b) a tracking system with good momentum resolution and high efficiency for multi-jet events as required to obtain good mass resolution (i.e. HZ -»• Hl+l~); c) good Energy flow reconstruction for W,Z-> qq(t) measurements and SUSY particle reconstruction; d) hermeticity and good forward energy determination in order to efficiently select SUSY signatures; e) good Luminosity evaluation including the luminosity spectrum due to beamstrahlung as required for precision measurements of cross-sections at threshold; f) an excellent lepton identification. The detector proposed for TESLA 8 is shown in Fig.3. It is based on a large volume coil for a 4T magnetic field containing the tracking system with a TPC as central detector and the Electromagnetic and Hadron Calorimeters. The performance goals for the most important detectors are listed in table 2. Two options for the Detector are taken into account in the American studies 6 : the first is similar to the European one with a large coil detector, while the second is based on a more compact geometry with an intense field (5 T) and several layers of Si^u-strips as central tracking. The Asiatic project foresees a detector based on a drift chamber as central tracker and two options (2T and 3T) for the magnetic field 7 . 5

Tracking system

The detector for TESLA has a tracking system based on a vertex detector (VTX), an intermediate tracker (SIT) and a TPC for the barrel region. In

314 7450

YOK 44S0 3850

con —

2977

HCA 1908 1680

ECA

TPC

>_ .

32C 160

0

vrx srr

b

|i|1150|||

42S0

MOO

7400

Figure 3. TESLA Detector.

Subdetector VTX TPC ECAL HCAL COIL Low Angle 4- Lumi CAL Tracking EFLOW

Performance 5± < 2 • 10- 4 10

Pt

*§• < °- ~m^) ® °-01' gr8mularity f

< 0.50 ]f} - © 0.04, granularity 4 T, uniformity< 10~3 coverage down to ~ 5 mrad (veto) " 5 • lQ" 6(GeVjcj =1 1J <

^f ~0.3 y/B(GaV)

Table 2. Detector performance required for an experiment at TESLA

addition planes of detectors orthogonal to the beam direction are foreseen in the intermediate region (FTD) and close to the TPC end plates (FCH). In the following only a few aspects of VTX and TPC are described, in particular those which are relevant for future R&D programs.

315

5,1

VTX

In order to optimise the impact parameter measurement (table 2) as required by Physics, the radius of the layer closest to the interaction point should be as smallest as possible. As seen in the previous section a Radius of 1.5 cm is compatible with background level with a 4T magnetic field, In addition the multiple scattering and consequently the detector thickness should be minimised. Furthermore the detector should cover the maximal solid angle. Three technology options are proposed to realize the VTX: CCD 9 ,CMOS 10 and Hybrid Pixels " . The advantages of the two first options are the very good point resolution and the possibility, still under study, of realizing an unsupported detector which minimises the material in the tracking volume. In particular the CMOS option is a new, very attractive solution based on charge collection by diffusion in undepleted epitaxial layer (Fig. 4) which could optimise the performance (test results gives a ~ 2fj,m with 20/im pitch) at low cost.

Figure 4. Charge collection in a CMOS detector.

On the other hand, the pixel technology has already been implemented in HEP experiments and will be adopted at LHC. It is then a well established technique which can presumably reach a point resolution o-<7/im by means of capacitive coupling read-out and an implant pitch of < 25/trn. 5.2

TPC

TPC's are detectors used in many HEP and Heavy Ions experiments. At a LC the performance should be pushed to the highest level. Few points of a TESLA TPC design (Fig. 5) are recalled: 1) the internal radius is determined by the size of the mask system; 2) the external radius is constrained by the fact that the calorimeter is inside the coil and by the desired momentum resolution of the TPC aloae (table 2); 3) events from nijyiy BX. are superimposed and hence the timing information should be precise enough to disentangle the events; 4) the TPC should be continuously operative during one train; 5) read-out

316

technologies ( GEMs 12, Micromegas13 ) are considered in alternative to the usual wire chambers. This is because, close to the wires where the avalanche multiplication of drift electrons occurs, the electric and the magnetic fields are no longer parallel and this produces a broadening of the electron cloud with consequent resolution deterioration. The required performance can be achieved with 200 points per track obtained with pads of 2 x 6 mm2 corresponding to a total of 1.2 M channels.

Figure 5. Layout of a TPC quadrant.

6

Calorimeters

Key point for the design of calorimeters for experiments at a LC is the capability of the whole detector to reconstruct precisely the jet energy. This fact is as important as the usual requirements for Electromagnetic (energy resolution, position reconstruction, e/ir separation etc.) and Hadron CALorimeters. Two opposite strategies can befollowed:1) measure the Energy as the sum of different contributions: E = Scfcp'* + £ 7 .E eco ' + Y,nentEhcal, proiting of the extremely good precision on the momentum measurement of charged tracks. In this case one of the most important issues of the ECAL ( and of the HCAL as well) is the high granularity which is necessary to disentangle the different contributions (European approach); 2) measure the energy following a calorimetric approach which implies good resolution and compensation in the HCAL (Asiatic approach).

317

6.1

ECAL

Two solutions are considered in the TESLA TDR: 1) SiW with extremely high granularity 14 ; 2) Shashlik with very high granularity 1 5 . The former is a calorimeter with Tungsten absorber and Silicon pads of 1 cm2 size as active detector. With 40 SiW layers the shower separation is maximised but the total Silicon area (about 3000 m 2 ) and the total number of channels (more than 30 M) are a serious challenge. The Shashlik solution is based on a R&Dproject 16 which demonstrated the feasibility of a 2-3-fold longitudinal segmentation. That was obtained by means of thin vacuum photodiodes placed in the first part of the detector or using scintillators with different decay times (Fig. 6).

Figure 6. Left) Signal time distribution for fast and slow scintillators. Right) Energy in the slow scintillator versus total energy for electrons and pions.

Other solutions with Pb (W) and scintillator with transversal WLS fibre read-out ( ff-tail) are considered in the AGFA project. In a European R&D project 1 7 this technique is proposed together with the insertion of 3 layers of 1 cm 2 Si pads in order to keep part of the advantages of the SiW calorimeter with much less silicon area and channels. 6.2

HCAL

Two technical solutions are foreseen for the hadron calorimeter: 1) an absorber-scintillator calorimeter with cr-tail read-out; 2) a digital ( binary) calorimeter with iron absorber and active layers with 1 cm 2 pad units. The former solution ( Europe and Asia) requires different absorber materials (iron or lead), different ratios passive-material/scintillator and different granularity in order to optimise the shower separation or the hadronic compensation. The

318

2 n d solution can be realized by means of RPC or limited Geiger wire chambers or wire chambers which simply record the passage or not of particles. This implies several millions of channels ( but they can be single bits). It is clearly proposed in order to exploit to the maximum level the granularity and to push to the extreme the energy flow based on particle separation. 7

Expected Performance

Several studies concerning the performance expected from the proposed detectors were done. In particular full simulation based on GEANT 3 package has been performed for the TESLA detector. Good results in agreement with the requirements (table 2) have been obtained. As an example the purity vs efficiency for heavy flavour tag is shown in Fig 7. 1

H**>r»»

0.9 0.8 0.7

•"•>"•%.,, SLD-b

•

•9 T

i / -

,

c(bbksr)

SLD-c

• r ^

t

0.6 0.5

"«

u

: m -

"

1

:

m

v

r

0.4 0J 0.2 ^i . , . i

.

.

»-

1 .

,

I

,

.

.

1

0.8

.

.

,

1

•

1 efficiency

Figure 7. Purity vs efficiency for heavy flavour tag in Z decays.

8

Conclusions

The Physics program of an e+e~ Linear Collider requires a high precision Detector. The machine conditions allow to design a Detector which utilises quite standard technical solutions pushed in order to obtain the best possible performance. The R&D programs when necessary are mainly oriented to exploit the high precision limits rather than to solve principle problems. References 1. ACFA LC Working Group KEK report 2001-11 hep-ph/0109166. 2. TESLA Technical Design Report DESY 2001-011 ECFA 2001-209.

319

3. American LC Working Group BNL-52627, CLNS 01/1729, FERMILABPub-01/058-E, LBLN-47813, SLAC-R-570, UCRL-ID-143810-DR. 4. M.N.Frary, D.J.Miller DESY-92-123A (1992) 379. 5. TESLA TDR 2 Part IV pg 127-138. 6. ALCWG 3 pg 379-412. 7. ACFA report 1 pg 209-375. 8. TESLA TDR 2 Part IV pg 1-7. 9. C.J.S. Damerell et al. LC-DET-2001-023 http://www.desy.de/~lcnotes. 10. G. Deputch et al. LC-DET-2001-017 http://www.desy.de/~lcnotes. 11. M. Battaglia et al. LC-DET-2001-042 http://www.desy.de/~lcnotes. 12. F.Sauli Nucl. Instrum. Methods A 386, 531 (1997). 13. Y. Giomataris et al. Nucl. Instrum. Methods A 376, 29 (1996). 14. TESLA TDR Part IV pg 65-73. 15. TESLA TDR Part IV pg 73-78. 16. A.C. Benvenuti et al. LC-DET-2001-027 http://www.desy.de/~lcnotes. 17. S. Bertolucci et al., DESY PRC R&D 00/02.

T H E ATLAS M U O N S P E C T R O M E T E R A. DI C I A C C I O Universita'

di Roma

" Tor Vergata", Dipartimento di Fisica, scientifica, 00139 Roma, Italy E-mail: anna, diciaccio @roma2. infn. it

via della

ricerca

ATLAS is a general purpose proton-proton experiment at the Large Hadron Collider (LHC) at CERN. The primary goal of the experiment is to operate at high luminosity (10 3 4 cm~2 s _ 1 ) with a detector that provides as many event signatures as possible. The quality of the muon measurement is insured by a system of three large superconducting air-core toroid magnets, precision tracking detectors with 80 f/,m intrinsic resolution and a dedicated trigger system. T h e ATLAS muon system is now in the construction phase. The main features of the muon spectrometer are presented in this paper together with the status of the chamber mass production.

1

Introduction

The Large Hadron Collider with its 14 TeV center of mass energy and design luminosity of 10 34 cm~2 s _ 1 opens a new frontier in particle physics. Beam crossing are 25 ns apart and at this luminosity there are 23 interactions per crossing. The potential of the ATLAS experiment for physics discoveries, such as Higgs bosons and supersymmetric particles, is insured by a high quality measurement of electron, gamma, jet, muon, missing transverse energy and bquark tagging *. The muon signature is particularly important in the difficult high-rate environment of the LHC collisions. Therefore the muon spectrometer 2 has been designed to provide stand-alone capability and to maintain high performances in terms of detection efficiency, momentum resolution and fake tracks rejection at the highest luminosities. Figure 1 shows the momentum resolution of the muon spectrometer as a function of the muon transverse momentum 3 . At low pr the momentum resolution is dominated by energy loss fluctuations, at intermediate energy (30 GeV < pr < 100 GeV) by multiple scattering and at higher momentum by the intrinsic detector resolution. Good momentum resolution is important for the detection of decays such as H —> ZZ* —> 4/i above large background. Figure 2 shows the mass resolution, essential for particle searches, for the SM Higgs decay H° —> ZZ* —* Afi as a function of the invariant mass.

320

321

10

p T (GeV) Figure 1. px dependence of the muon momentum resolution averaged over the azimuthal angle <j> and for \q\ < 1.5.

£3.4

(b)

5'3.2

+

i_iO

-

+

-

"> 3

5 <

+

2.8 2.6 2.4 2.2

•

2 1.8

•••••

+

•

*

300

400

500

•

1.6 100

200

600

700

800

900

M 4p (GeV)

Figure 2. Mass resolution as a function of t h e invariant mass, using the muon spectrometer only; no Z mass constraint was applied.

322

Figure 3. Layout of the ATLAS muon spectrometer.

2

T h e muon system

The layout of the muon spectrometer is presented in figure 3. It is based on the magnetic deflection of muon tracks in the large superconducting air-core toroid magnets, instrumented with separate trigger and high precision tracking chambers. Over the pseudorapidity range |TJ| < 1.0, magnetic bending is provided by the large barrel toroid (BT) consisting of independent coils arranged with an eight fold symmetry outside the calorimeter. For the range 1.4 < \t]\ < 2.7, muon tracks are bent by two smaller end-cap toroids (ECT) inserted into the BT at each end. This magnet configuration provides a field that is mostly orthogonal to the muon trajectories, while minimizing the degradation of the resolution due to the multiple scattering. The performance in terms of the bending power is characterized by the field integral / Bdl, where B is the azirnuthal held component and the integral is taken on a straight line trajectory between the inner and the outer radius of the toroid. The BT provides 2 to 6 Trn and the ECT contributes 4 to 8 Tm in the 0.0-1.3 and 1.6-2.7 pseudorapidity ranges respectively. The bending power is lower in the transition regions where the two magnets overlap (1.3 < \rj\ < 1.6). The overall dimensions of the magnetic system are 26 m in length and 20 m in diameter.

323

3

The muon chambers

In the barrel region, tracks are measured in chambers arranged in three cylindrical layers (stations) around the beam axis; in the transition and end-cap regions the chambers are installed vertically, also in three stations. Wherever possible, the chambers measure the sagitta of the curved tracks in three positions: at the inner field boundary, close to the field maximum and at the outer field boundary. In the end-cap regions (|?;| < 1.4) the magnet cryostat does not allow the positioning of the chambers inside the field volume, so the deflection of the tracks that have traversed the ECT are measured taking advantage of a large lever arm between the two outer measurement stations. Over most of the pseudorapidity range, a precise measurement of the track coordinates in the main bending direction of the magnetic field is provided by Monitored Drift Tubes (MDTs) over \T)\ < 2. At large pseudorapidities and close to the interaction point, Cathode Strip Chambers (CSCs) with higher granularities are used in the innermost plane over 2 < |r;| < 2.7 to stand the high rate and background conditions. The trigger system covers the pseudorapidity range |^| < 2.4. Resistive Plate Chambers (RPCs) are used in the barrel at |?y| < 1.05 and Thin Gap Chambers (TGCs) are used in the end-caps. The trigger chambers serve several purposes: • bunch crossing identification, requiring a time resolution better than the LHC bunch spacing of 25 ns; • a trigger with well defined muon pr discrimination; • second coordinate measurement in the non-bending projection with a typical resolution of 5-10 mm. The trigger system is visible in figure 4. For the RPCs a low pr coincidence (with a resolution smaller than the 25 ns bunch-crossing interval) is defined by those tracks that have hits in at least three of the four middle station trigger planes. A high pr coincidence is the logical AND of a low pr coincidence and at least one hit in the two planes of the outer stations. A similar algorithm is used in the case of the TGCs. The low pr thresholds (6-10 GeV) allow to trigger on muons coming from beauty decays, the high pr (20-35 GeV) thresholds for heavy particle search. These thresholds have been chosen to ensure good trigger efficiency for beauty and Higgs physics and a reasonable background rejection. Table 1 summarizes the number of chambers, the covered area and the number of read-out channels for the four chamber technologies. The chamber mass production should be finished by the end of 2004. Assembly and installation of the muon stations is foreseen for the years 2004-2005. Each

324 TGC 2 - H

Figure 4. ATLAS Level-1 muon trigger scheme.

Table 1. Overview of the muon chamber system. T h e covered area refers to the chamber modules which normally contain several detector layers.

Numbers of chambers Numbers of readout channels Covered area (ra 2 )

Precision Chambers

Trigger Chambers

CSCs MDTs

RPCs TGCs

32 1.163 31.000 370.000 27 5.500

1136 1584 385.000 322.000 3.650 2.900

step of the production is monitored by a very severe quality control program. The chamber performances are measured with cosmic ray test stations. 3.1

The Monitored Drift Tubes

The basic detection elements of the MDT chambers are aluminum tubes of 30 mm diameter and 400 fj,m wall thickness, with a 50 urn diameter central WRe wire. The tubes are operated with a non flammable Ar - CO2 (93%/7%) mixture at 3 bar absolute pressure. The envisaged working point provides a gain of 2 x 104 (to minimize ageing) and a maximum drift time of 700 ns. The single wire resolution, measured with various prototypes, is ~ 80 firn except for tracks passing very close to the wire. To improve the resolution of a chamber beyond the single wire limit and to achieve adequate redundancy for

325

pattern recognition, the MDT chambers are constructed from 2 x 4 monolayers of drift tubes for the inner and 2 x 3 monolayers for the middle and outer stations. The tubes are arranged in multilayers of three or four monolayers, respectively on either side of a rigid support structure (spacer frames) that provides accurate positioning of the drift tubes with respect to each other and support to the components of the alignment system. Mechanical deformations are monitored by an in-plane optical system with an accuracy of ~ 10 /xm. A system of light rays and sensors implemented in RASNIK technology allows for the measurements of a chamber displacement. The MDT chambers are presently under construction in thirteen production sites and ~ 20% of the bare chambers have been already produced with the required performances. 3.2

The Cathode Strip Chambers

The CSCs are multiwire proportional chambers with cathode strip read-out and with a symmetric cell in which the cathode-anode spacing is equal to the anode wire pitch. The precision coordinate is obtained by measuring the charge induced on the segmented cathode by an avalanche formed on the anode wire. Good spatial resolution is obtained by segmentation of the readout cathode and by charge interpolation between neighboring strips. The cathode strips for the precision measurements are oriented orthogonal to the anode wires. The anode wire pitch is 2.54 mm and the cathode readout pitch is 5.08 mm; r.m.s. resolution of better than 60 fim has been measured in several prototypes. Other important characteristics are smaller electron drift time (< 30ns), good time resolution (7 ns) and low neutron sensitivity. A measurement of the transverse coordinate is obtained using the orthogonal strips, i.e. oriented parallel the anode wires, which form the second cathode of the chamber. The CSCs are arranged in 2 x4 layers. The gas mixture is non flammable and composed of 30% Ar, 50% C02 and 20% CF4. 3.3

The Resistive Plate Chambers

The RPCS is a gaseous detector, working in avalanche mode, with a typical space-time resolution of 1 cm x 1 ns with digital read-out. The basic unit consists of a 2 mm gas gap formed by 2 mm parallel resistive bakelite plates, separated by insulators. The gas mixture is 96.7% C2H2F4, 3% iso — C4H10 and 0.3% SFs- The signal is read out via a capacitive coupling by metal strips on both sides of the detector. A trigger chamber is made of two detector layers, each one read-out by two orthogonal series of pick strips with a variable width from 2.5 to 3.8 cm. The first 3% of the chamber production has been achieved. Cosmic ray tests show that the produced chambers have the foreseen

326 performances in terms of efficiency, cluster size and time resolution. After an ageing test of 0.3 C/cm2, corresponding to 10 ATLAS years, performed at CERN GIF irradiation facility, the RPC shows a rate capability of 200-300 Hz/cm2 adequate to the LHC needs 4 . 3.4

The Thin Gap Chambers

The TGCs are proportional chambers operating with a thin gap; both signals from the proportional wires (for trigger and bunch-crossing identification) and cathode strips (for trigger and second coordinate measurement) are used for the digital read-out. The main dimensional characteristics of the chambers are cathode-cathode distance (gas gap) of 2.8 mm, a wire pitch of 1.8 mm and a wire diameter of 50 \im. The operating high voltage is 3.1 KV. The electric field configuration and the small wire distance provide for a short drift time and a time resolution good to identify the bunch-crossing. The TGCs are assembled in doublets and triplets. They will be operated with a gas mixture of 55% CO2 and 45% n - C5H12. The TGCs costruction is well advanced in two sites and ~ 25% of the chambers have been made with a high quality control of the production 5 . 4

Summary

The ATLAS muon spectrometer is now in the construction phase with a tight schedule to be ready for the first collisions in 2006. The performances of the detector are fully adequate to explore the rich physics program of the LHC. References 1. ATLAS Collaboration, Technical Proposal, CERN/LHCC/94-93, LHCC/P2, 15 December 1998. 2. ATLAS Collaboration, Muon Spectrometer Technical Design Report, CERN/LHCC/97-22, 31 May 1997. 3. ATLAS Collaboration, Detector and Physics Performance Technical Design Report, CERN/LHCC/99-14, 23 May 1999. 4. G.Aielli et al., Further advances on RPC ageing studies, Proceeding of the Conference: "Ageing phenomena in gaseous detectors", Hamburg (DESY), 2-5 October 2001. 5. K.Ishii et al., ATLAS muon spectrometer, Proceeding of the International Conference on High Energy Physics, Budapest, Hungary, 12-18 July 2001.

US ATLAS M U O N E N D C A P S Y S T E M ALEX MARIN Mailing Address *' HEPL/Harvard Univ, 42 Oxford st, Cambridge, MA 02138, USA E-mail: [email protected] We present the status and the Progress Report of the U.S. Muon End Cap system.

1 1.1

INTRODUCTION The ATLAS Muon End Cap System Tasks

The US ATLAS End Cap Muon system is covering the rapidity region between 1 and 2.7 The main tasks of the US ATLAS team are summarized as follows: 1. Design and construction of the Monitored Drift Tubes ( M D T ) : rapidity between 1 and 2; 2. Design and construction of the Cathode Strip Chambers (CSC): rapidity between 2 and 2.7; 3. Design and construction of the Alignment System for the entire muon end cap system; 4. Design and construction of the Associate Electronics for US and other ATLAS groups. In Figure 1 we present the end cap muon system layout. Various type and shapes of chambers are used. The talk discuss mainly the MDT chambers as they are built at the present time. 2

US M D T C h a m b e r s

The US ATLAS MDT collaborations designed and built the tooling for the End Cap MDT chmabers. The production sites are Harvard University (BMC), Seattle (UW) and Ann Arbor (UM). 240 MDT Chambers EI and EM chambers will be built in the present approved scope. *For the U.S. Muon End Cap Collaboration: Boston Muon Consortium (BMC : Boston Univ., Brandeis Univ., Harvard Univ., MIT, Tufts Univ.), Brookhaven National Lab. (BNL), University of California Irvine, University of Michigan (UM), Northern Illinois University, SUNY Stony Brook, University of Washington (UW) Presented at: 7th Int. Conf. on Tech. and Particle Phys., Como, Italy, 15-19 Oct, 2001

327

328

W-044C

'"'" • • • ' • • • •

<=?

yMO-l*

% j $ $ & ; : ^

:

Figure 1. ATLAS End Cap Muon Chamber Layout.

The US ATLAS group is also involved in building parts of the read out electronics for the MDT chambers as follows: - Signal HedgeHog cards for U.S. chambers - Mezzanine readout C8JT ds for entire ATLAS - Design of Chamber Service Module University of Washington design and build the kinematic mounts and the associate hardware for all US MDT chambers.

2.1

Tube Production

Each site is producing the tube according with the chamber in production. Typical production rate is of the order of 50 tubes per day. Figure 2 presents the UM tube production factory.

329

Figure 2. ATLAS Tube Factory (UM) and UW QA and QC data.

2.2

Chamber Production

Figure 3 show a base chamber produced at UW. All three sites produced about 80 chambers at this date, with good results. The CERN tomo surveys indicated that the wire in the US MDT chambers are within the specs.

2.3

Services and Cosmic Ray tests

Presently, all three sites are instating the Faraday Cages, gas systems, as well as the rest of the sensors required. Figure 4 show a detailed picture of one chamber with Faraday cage and gas system installed. The BMC group built a cosmic ray stand, which presently is used for various tests using the series prototype of EILl chamber. The chamber is filled and monitored in simmilar conditions with the actual ATLAS specs. One important results is presented, showing that the chamber filled with the ATLAS gas performes at the design resolution.

330

Figure 3. ATLAS Base Muon Chambers (UM).

3

CSC chambers for the Muon End Caps

The CSC chambers are design and built primarily by BNL. Figure 5 shows the schematics of the CSCs which axe now in production at BNL. 4

Alignment System

Brarideis University plays a major role in the Muon end cap alignment system for the entire muon end cap chambers. They are also designing and manufacturing all the components for the alignment bars, inplane alignment system as well as various components of the RASNIK/readout systems used by most of our collaborators.

331

Figure 4. BMC Cosmic Ray Stand Layout and results.

'I-.-.' J " * * * * * «>*•«•»*%*'•«&£* fc"«*

Figure 5. ATLAS End Cap CSC Layout (BNL).

P E R F O R M A N C E OF T H E M A C R O LIMITED S T R E A M E R T U B E S FOR ESTIMATES OF M U O N E N E R G Y M. G I O R G I N I for the MACRO Dept.

of Physics

Collaboration*

and INFN, V.le C. Berti Pichat 6/2, 1-40127 Bologna, E-mail: [email protected]

Italy

The MACRO limited streamer tubes can be operated in drift mode by using the TDCs included in the Q T P system. In this way a considerable improvement in the space resolution is obtained, allowing the analysis of muon tracks in terms of multiple scattering effects and the energy estimates of muons crossing the detector. We present the results of two dedicated tests, performed at CERN PS-T9 and SPSX7 beams, to provide a full check of the electronics and to exploit the feasibility of the analysis. Using a neural network, we are able to estimate the muon energies up to Ey, ~ 40 GeV. The test beam data provide then an absolute energy calibration, which allows to apply the method to the MACRO data.

1

Introduction

The MACRO experiment can study atmospheric neutrinos via the detection of neutrino-induced muons. Recent results concerning upthroughgoing muons, produced in the neutrino interactions in the rock below MACRO and crossing the detector, showed a flux deficit and a distortion of the zenith angle distribution with respect to the MonteCarlo (MC) expectations 1 . The data can be interpreted in terms of neutrino oscillations. As the oscillation probability of neutrinos depends on oscillation parameters and on the ratio L/Ev (where L is the distance between neutrino production and interaction points and Ev is the neutrino energy), it is important to estimate the residual muon energy. Since MACRO is not equipped with a magnet and the mass is not large enough to stop the most part of upgoing muons, the only way to estimate the muon energy is by a Multiple Coulomb Scattering (MCS) analysis. The r.m.s. of the lateral displacement of a relativistic muon crossing a layer of material with depth X and radiation length X° is °MCS * ^ ^ ^ / ^ ( l

+ 0.0038 MX/X°))

(1)

where pM (GeV) is the muon momentum. The measurement of the track * For the full list of the Collaboration see the second paper of ref.

332

1

.

333

deflection has reasonable precision when

OMCS

is larger than the detector

space resolution. In MACRO, on the vertical, (JMCS — 10(cm)/.EM(GeV). The space resolution of the tracking system (streamer tubes used in digital mode) is ~ 1 cm, providing a muon energy estimate through MCS up to ~ 10 GeV, not enough to fully exploit the physics potential of neutrino-induced muons. To improve the space resolution, we used the MACRO limited streamer tubes in drift mode, by using the TDCs implemented in the QTP electronic system 2 , designed for magnetic monopole searches. A check of this electronics with two dedicated tests, performed at the CERN PS-T9 and SPS-X7 beams, and the estimates of muon energies in terms of MCS effects are described. 2

The M A C R O limited streamer tubes in drift mode

The MACRO streamer tube system 3 consists of about 5600 chambers; each chamber is made of 8 streamer tubes, each with cross section (3x3) cm2 and 1200 cm length, for a total of about 50000 wires. The QTP system 2 consists of an ADC/TDC system and acts as a 640 ^isec memory, during which the charge, the arrival time and the width of the streamer pulse of the particle crossing the cell are recorded. The TDC bin width is AT ~ 150 ns, quite coarse for drift time measurements, because the maximum drift time for our streamer tubes, operated with a He(73%)/npentane(27%) mixture, is ~ 600 ns. The ultimate resolution we can obtain with such device is a ~ VdHft • A T / \ / l 2 ~ 2 mm, enough to estimate upgoing muon energies up to 30 -j- 40 GeV. Such electronics can be used to operate the streamer tubes in drift mode, retrieving the TDCs informations corresponding to each fired wire. Selecting only planes with a single fired tube, the association with the fired QTP channel is uniquely determined. To avoid systematic effects and to fully understand the capability of the QTP system in such context, we decided to check the electronics in a test beam at CERN PS-T9. 3

The C E R N PS-T9 test b e a m

The goals of the test beam were: (i) the study of the QTP-TDCs linearity; (ii) the study of the drift velocity in He/n-pentane; (Hi) the test of the software used for muon tracking. Moreover, the comparison of the muon energy reconstructed with MCS analysis with respect to the beam nominal energy

334 101

bass**

Figure 1. Test beam layout.

offered the possibility of performing an absolute energy calibration. We reproduced a slice of MACRO which was esposed t o unions with energies ranging from 2 to 12 GeV. The test beam layout is described in Fig. 1: the trigger was provided by the fast coincidence of the scintillators S1,S2,S3. The last scintillator, following a 60 cm iron slab, suppressed the w±, K*1 contamination at high energy. The analog output of a chamber was sent to a QTP channel, while the digital output, OR of the chamber signals, was sent to a Lecroy 2228A TDC (AT = 250 ps). Such double measurement of the drift time allowed us to make a comparison between QTP-TDCs and Lecroy TDCs event by event. In 80 runs we collected about 108 muons. For the check of QTP electronics, we used runs with absorbers out. We evaluated the QTP-TDCs linearity by comparing them with the Lecroy TDCs. Fig. 2a shows the profile plot of such variables: for a given value of the QTP-TDC system (75 ns, 225 ns, 375 ns, 525 ns), we performed a gaussian fit of the Lecroy TDCs time distribution. Horizontal errors on the figure represent the bin width of the QTP-TDCs, while errors on vertical represent the sigma of such gaussians: the QTP-TDCs response is linear within 8%. We then studied the drift velocity in He/n-pentane mixture. Since in the test beam configuration the muons hit the detector perpendicularly: dN

dz dN

dN

'df-'di^-

Vdrift

-'d7

=

K Vd ift

- '-

(2)

335

10

Test beam data GARFIELD

100

200

300

400

500

600

0.2

0.4

0.6

r(cm)

QTP t i m e ( n s )

(a)

(b)

Figure 2. (a) Lecroy TDCs drift times as a function of QTP-TDCs times, (b) Drift velocity measured at test beam (black points) compared with the GARFIELD MC expectation. The dotted lines represent the effect of a 15% gas mixture variation.

The evaluation of Vdnjt can be obtained fitting the Lecroy TDCs spectrum distribution. Fig. 2b shows the experimental results obtained (black points), superimposed to the GARFIELD 4 MC. The experimental data are in good agreement with the simulation. The test beam data can be also used to measure the space resolution. Fig. 3a shows the residuals distribution for streamer tubes in drift mode using the Lecroy TDCs and the QTP-TDC system. In the first case a resolution of ~ 500^m is found, while for the second one we obtained a resolution of a ~ 2 mm, very close to the QTP-TDCs resolution a = Vdrift x 150ns/\/l2 ~ 4cm///s x 150ns/\/l2 ~ 1.9 mm. Summarizing, the test beam data analysis showed that: (i) no "hot-effects" are present in the QTP electronics, which can then be successfully used to operate streamer tubes in drift mode; (ii) the QTP-TDCs have a reasonable linearity (within 8%); (m) the drift velocity in He/n-pentane mixture is in good agreement with the GARFIELD simulation; (iv) the software tools are adequate to perform a reasonable muon track fit; (v) the space resolution of the streamer tubes in drift mode measured at the test beam is <JQTP ^ 2 mm. 4

Study of the M A C R O space resolution

In order to estimate the performance of the streamer tubes in drift mode in MACRO, we analysed a sample of downgoing muons ((SM) ^ 300 GeV).

336 3

3

< jgO.08 c

<„ 0.14

L

•

?0.07

Leeroy TDCs

1

QTP-TDCs

E0.06 2 0.05

1 z3

iI J

0.03

/

n

0.02 0.01

-1

-0.8

.^J? -0.6

-0.4

-0.2

(a)

•

*?!

^0.12 O E

0.04

Tj.

S 4 '

MACRO data (OTP) MontoCario(QTP) MACRO data (no QTP

4

0.1

*

0.08

0.06

\

4 0.04

\

0.02

Si^-.

0.2

0.4

0.6

0.8 1 Ax (cm)

yf\.

.,-••-


-^ ^

/

V\

.. rvs^_ j ^ ^ . Ax (cm)

(b)

Figure 3. (a) Test beam data: residuals distribution from Leeroy TDCs (continuous line) and QTP-TDCs (dotted line), (b) MACRO data: residuals distribution from the Q T P T D C s (a ~ 3 mm) compared with the MC expectation. T h e dotted line represents the same distribution obtained using the streamer tubes in digital mode (c ~ 1 cm).

Among the muon tracks reconstructed with the standard MACRO tracking (streamer tubes in digital mode), we selected the hits made of a single fired tube. For each hit we looked at the corresponding QTP-TDC value in a time window of 2 fis, to avoid background hits. After converting the TDC values in drift radii using the drift velocity measured at the test beam, a global fit of the track tangent to the circles is performed. Fig. 3b shows the distribution of the track residuals for the MACRO streamer tubes in drift mode (black circles) and the simulation (continuous line). The residuals of the downgoing muons have a a ~ 3 mm, in good agreement with the simulation; the continuous line shows the residuals distribution obtained by the streamer tubes in digital mode [a ~ 1 cm): an improvement of a factor ~ 3.5 has been obtained. As far as MACRO data is concerned, the resolution is spoiled with respect to that measured at the PS-T9 test beam (~ 2 mm). This difference comes from 5-rays and radiated photons produced in the rock absorbers which may produce early streamers with respect to those due to the muon, resulting in smaller drift radii, from residual downgoing muon MCS and finally from streamer tube gas mixture variations which may affect the drift velocity. This hypothesis has been demonstrated during a second test beam, per-

337

^

0. I-

032

• PS-T9 Test beam data » SPS-X7 Test beam data ^fiASRS* - a Monte Carlo simulation ...

t!

M "-" £0.225

E ^ GeV

S

•S...S

/

Eo.075

,.,

^

•

a"-*

D

•

S>

0) .Q

I-*1

1

0.1

{ r

Eo.075 1

1

1

1

.,...,,

3 °.os 0.025

P

j^

•s5?7..., i

Log,0E„(GeV)

(a)

^ , t.

''

Log1()(E/"(GeV))

© 0.2 »0.176

0.2

Q0.175 "o °" W0.12S

D

J-\L k T>% J ^L

£o.2» E M =40 GeV

EM=12 GeV

£o.22S D D -

0.02S

Log^E^GeV))

D

T

"oi-0.12$ °-1s "

r-S**!

f=

E„=4 GeV

0-2

O0.17S

n

J*D 1

"5 o.is r feO.IK

kr

\ . . ,Lm

J]

Eo.075 3 0.05 0.025

l~*

u/

A>r

fc

l

, I , . . i n- . _LJtAs^tvT, Log (E'«(GeV))

Log^E^GeV))

10

(b)

Figure 4. (a) Average neural network output as a function of l o g 1 0 E M : PS-T9 test beam data (full circles), SPS-X7 test beam data (full triangles) and MC prediction (empty squares), (b) Reconstructed muon energy for 2, 4, 12, 40 GeV muons: the full symbols have the same meaning as in (a), the continuous line represents the MC simulation.

formed at CERN SPS-X7, where the same layout of Fig. 1 was exposed to muons with energies up to 100 GeV. The sigma of the residuals obtained with E^ — 100 GeV and rock absorbers inserted was a = (0.294 ± 0.003) cm, in good agreement with that obtained using the MACRO downgoing muon data. 5

Muon energy estimate

For each muon event we computed the following multiple scattering sensitive variables: the highest residual between the 14 available, the average of the residuals, the difference of the residuals of the three farthest hits along the track, the slope and the intercept of the "progressive-fit". We define "progressive fit" the linear fit of the absolute value of the residuals as a function of the streamer tube plane (i=l,14). It gives an almost small slope for high energy muons because their energy is almost the same in different planes, while such slope is larger for low energy muons because they loose a high fraction of their energy crossing the detector. We followed in this analysis a neural network approach (NN), choosing JETNET 3.0, a standard package with a multilayer perceptron architecture and with back-propagation updating 5 . The NN was configured with the 7

338

£ „ (GeV) Eruec (GeV)

2. (23. 6 )

3.

(3+ft)

5. (5+i8)

10. (10+2")

40. (40±^)

Table 1. Reconstructed muon energy with ±1
input variables quoted above and 1 hidden layer, selecting the Manhattan upgrading function. The average neural network parameter output increases as a function of the muon energy up to E^ ~ 40 GeV, showing saturation at higher energies. The use of the streamer tube system in drift mode combined with a neural network approach allows therefore to estimate the muon energy up to ~ 40 GeV. Fig. 4a shows the comparison between the PS-T9 data (black circles), the SPS-X7 data (full triangles) and the MC prediction (empty squares). The NN output obtained with the test beam data is properly reproduced by the simulation. The muon energy can be reconstructed by inverting the curve shown Fig. 4a. Fig. 4b shows the reconstructed muon energy E™c for En = 2,4,12 and 40 GeV: the data collected at the PS-T9 test beam (full squares) and at the SPS-X7 test beam (full triangles) are compared with the MC expectation (continuous line), showing a reasonable agreement. 6

Conclusions

The CERN PS-T9 test beam excluded the presence of undesired effects in using the MACRO QTP electronics to operate the streamer tubes in drift mode. This method, allowing to improve the space resolution of a factor ~ 3.5 (from a ~ 1 cm to a ~ 3 mm), was followed by a neural network approach which estimates the energy of muons crossing the detector up to ~ 40 GeV. This analysis can be applied to neutrino-induced upgoing muons in MACRO to test the hypothesis of neutrino oscillations. References 1. S. Ahlen et al, MACRO Coll., Phys. Lett. B 357, 481 (1995); M. Ambrosio et al, MACRO Coll., Phys. Lett. B 434, 451 (1998). 2. M. Ambrosio et al, Nucl. Instrum. Methods A 321, 609 (1992). 3. S. Ahlen et al, MACRO Coll., Nucl. Instrum. Methods A 324, 337 (1993). 4. R. Veenhof, Nucl. Instrum. Methods A 419, 726 (1998). 5. C. Peterson et al, Comput. Phys. Commun. 8 1 , 185 (1994).

EXPLOITATION OF ATLAS DAQ PROTOTYPES FOR TEST BEAM AND LAB ACTIVITIES BENIAMINO DI GIROLAMO CERN, Geneva, Switzerland E-mail: Beniamino.Di. GirolamoQcern. ch A functional Implementation of a vertical slice of the ATLAS Data Acquisition has been exploited in the past two years as a DAQ system for testbeam and lab test with satisfactory results. Here the setup descriptions and figures of merit are reported.

1

Introduction

The ATLAS DAQ prototype for test beam and laboratory tests is a functional implementation of a vertical slice of the ATLAS Data Acquisition. The prototype, named DAQ/EF -1, includes all the elements from the detector output to the mass storage (figure 1). DETECTOR

EF

Event Filter Farm

Figure 1. Schematic view of the ATLAS DAQ/EF-1 prototype. The elements in the shaded area are simulated by hardware and software components.

After a successful development and implementation completed in 1999, it

339

340

has been chosen as the baseline for the ATLAS test beam DAQ. It has been re-engineered for the exploitation with detectors and integrated for the first time with the ATLAS Tilecal hadronic calorimeter during the spring-summer 2000. A modified version has been integrated during summer 2001 with the ATLAS Muon Drift Tubes detector. A simplified version of the DAQ prototype has been also exploited for laboratory tests of the prototypes of the ATLAS subdetector Read Out Buffers (RODs) that are currently being built and certified. In this case the system has been modified to be even simpler and more cost effective. The various implementations and their performance are described here. Finally a plan for the future evolutions is also drawn. 2

Testbeam Data Acquisition

The ATLAS experiment will run at the Large Hadron Collider (LHC) at CERN starting from the beginning of 2006 1 . Most of the ATLAS subdtectors are now being produced and pre-assembled with their final electronics. 2.1

The ATLAS Tilecal testbeam DAQ

The ATLAS Tile Hadron Calorimeter (Tilecal) 2 is one of the most advanced subdetectors that started to take data in summer 2000 with its final frontend electronics at the CERN SPS H8 beam line. At the beginning of 2000 the subdetector community decided to adopt as a testbeam DAQ system the ATLAS DAQ/EF-1 prototype. The various DAQ components have been successfully integrated with the detector software and hardware. The Tilecal testbeam setup includes two barrel production modules (about six meters long, 90 readout channels on two output optical links each) and two extended barrel modules (about three meters long, 28 readout channels on one output optical link each). The maximum number of optical links for this setup is six. The ATLAS readout architecture is such that the optical links coming from the front-end electronics have to be connected to the detector specific Read Out Drivers (RODs) that collect the front-end information and output the Level 1 accepted events to the so-called Read Out Buffers (ROBs) via optical links, the Read Out Links (ROLs). The events are bufferized in ROBs until a Level 2 decision is made and then sent to the Event Building (EB). A detailed description of this architecture can be found in the ATLAS HLT,DAQ and DCS technical Proposal 3 . The six optical links coming from the Tilecal modules have been collected

341

in two ROD emulators. Since the ATLAS Tilecal ROD is still in development, a VME single board computer based emulator has been used as a substitute. The ROD emulators have been built using CES RI02 boards 4 on which it is possible to connect up to two PMC cards; to have the capability to control with one CPU more input and output cards, a CES PMC extender PEB6084 4 has been plugged to the RI02 boards to allow the connection of a maximum of four mezzanine cards. Therefore each ROD emulator can accept up to three optical links and can output the collected events over a fourth optical link to a ROB. The ROD emulators are both housed in a VME crate, the ROD crate (figure 2).

Figure 2. The ATLAS Tilecal DAQ setup.

An additional crate is used by the detector to collect the data coming from the trigger counters, the beam position chambers and from other ancillary detectors used for triggering purpose and for detecting the non contained fraction of the hadronic showers. The collected information from the modules of this additional element, the Beam crate, is sent to a third ROB (figure 2). The VME crate containing the ROBs is the Read Out Crate (ROC). The first implementation of the ROC functionalities is based on VME and it includes also three additional components: ® The LDAQ module, that is dedicated to the crate control and is connected, via Ethernet, to the general DAQ services via the Back End software that is another component of the DAQ/EF-1 prototype providing all the necessary services (connection with a configuration database, mes-

342

sage reporting system and information services, graphical user interface, just to mention the main elements). • The TRG module, that is the interface module to the detector trigger that is replacing the missing Level 2 selection. • The EBIF module, that collects and assemble the fragments from the ROBs and send them to the Event Building infrastructure via a dedicated link (a Fast Ethernet link in this implementation). The Event Building has been implemented on a Linux based PC, here referred as the SFC PC. The SFC PC has two Fast Ethernet network card, one used for a dedicated connection with a central data recording system and one connected point-to-point with the EBIF module of the ROC crate. The VME implementation of the ROC has been adopted in Tilecal. This implementation is the best suited when external links have to be handled concurrently and independently, that is the situation that the DAQ will face during the data taking at LHC. It allows to have one process per processor and can potentially run a large number of ROBs that are then multiplexed into one Event Building link. However, when the Level 2 rejection is absent, it suffers from the VMEbus limitations: limited bandwidth and lack of broadcast. Another limiting factor is the high cost. The performance of this implementation have been evaluated. In figure 3 a plot of the throughput versus the fragment size is shown. A throughput of about 5 MB/s is obtained for fragments of 2 kB. This performance is acceptable for a typical testbeam data taking, but it has to be remarked that the Fast Ethernet link used for the measurements is exploited at a maximum of 35 % of its bandwidth, due to the poor T C P / I P implementation of LynxOS. On the practical side the new DAQ system is a huge improvement for the Tilecal data taking performance. The improvement achieved can be expressed as the time needed to take a certain amount of data. In 1999 and before, with the previously used DAQ system, it was possible to take 20000 events in 30 minutes, while from 2000 it has been possible to collect 75000 events in six minutes, that is an improvement of a factor 20. 2.2

The ATLAS MDT testbeam DAQ

The Muon Drift Tubes (MDT) is one of the four muon subdetectors of ATLAS. The subdetector has started in 2001 a large activity of beam tests of production chambers. The testbeam data taking has been dedicated to the study of three different barrel chambers 5 and to the setup of the new DAQ system, based on a further evolution of the DAQ/EF-1 prototype.

343

Figure 3. Throughput vs fragment size for the VME based ROC with a EBIF module point-to-point connected to the Event Building PC via a Fast Ethernet link.

Figure 4. The ATLAS MDT DAQ setup.

The DAQ setup (figure 4) is similar to the one adopted by Tilecal, but the VME ROC crate has been replaced by a PC based ROC that runs all the data acquisition tasks in a single application. The optical links connecting the detector crates (ROD and Beam crates) are of the same type used in Tilecal and in this implementation the receiving cards are mounted in the PC ROC via PCI interface cards. The Linux OS has been extended for specific hardware access by mean of dedicated DAQ patches, but it is mainly based on the kernel 2.2.x. The higher processing power available is an advantage of this implementation as well as the lower cost when compared to the VME based implementation. However the limited number of PCI slots available in off-the-shelf PC is a limitation of this system. Another limitation is given by the sequential handling of the

344

links in the present implementation of the readout software. The performance of the system (figure 5) have been evaluated in two situation: a Fast Ethernet point-to-point link is established between the PC based ROC and the PC based SFC (circles); a Gigabit Ethernet point-to-point is established instead (triangles).

600

1000

1500

2000

ROD fragment size (B)

Figure 5. Throughput vs fragment size for the PC based ROC with a EBIF module pointto-point connected to the Event Building P C via a Fast Ethernet link (circles) and via a Gigabit Ethernet link (triangles).

The plot shows how the Fast Ethernet link is exploited to its maximum bandwidth already for small fragments (about 250 byte), while a maximum throughput of 30 MB/s is reached for the Gigabit Ethernet link in the range of the fragment sizes explored. 3

Laboratory test DAQ system

The experience with the testbeam setups has proved the high modularity of the DAQ system in different implementations. Therefore the need of the subdetectors to test the prototypes of their final electronics, especially the ROD modules, lead to a further implementation of the DAQ/EF-1 prototype for laboratory tests. In the minimal system configuration (figure 6) the main three functions of the DAQ, the collection of the data from the detector electronics, the data recording and the overall online control, are dispatched to three dedicated PC. Starting from this situation a reduced system has been buil, where all these functions are implemented on a single PC. This setup has been implemented as a baseline system for ATLAS subdetector ROD test data acquisition and the performance have been evaluated by using a fast SCSI drive as a storage medium. The performance achieved are very promising since the system is able to process 30 MB/s. The con-

345

Figure 8. Evolution of a minimal DAQ system to a reduced DAQ prototype for laboratory tests,

current running of the Dataflow and the Online software does not result in performance penalties (figure 7), although that might not be true if an heavy monitoring task is also required and that in the measurements here discussed was kept to the minimum. 35

30 25 20

15 4 10 •

V

8

e 500

1000 1503 2000

2800

Figure 7. Performance achieved by the minimal DAQ system for laboratory tests.

4

Conclusions a n d future

The preparation of ATLAS subdetectors for their final installation and commissioning are going in parallel with the exploitation of the ATLAS DAQ prototypes in beam and laboratory tests to get the detector people acquainted with the modern data acquisition system that will be used during the testing phases and the final data taking with a clear benefit. On the other ends the

346

integration of the DAQ prototypes with the various subdetectors helps the data acquisition experts to verify and modify the user requirements for a fully functional data acquisition for LHC. In future the current prototypes will be exploited for the coming testbeams in the period 2002-2004, as well for the various electronics tests. Starting from next year, there will be also the possibility to send the detector outputs, before and in parallel to the storage stage, to a minimal scale event filter farm. This last element will become part of the testbeam DAQ infrastructure once completely certified. Acknowledgements I would like to thank Livio Mapelli and the CERN EP ATD group for their work and support for the efforts done for the DAQ integration with the ATLAS detectors, as well as all the subdetectors experts involved in the process. References 1. ATLAS Collaboration, Technical Proposal for a General-Purpose pp Experiment at the Large Hadron Collider at CERN, CERN/LHCC/94-43, 1994. 2. ATLAS Collaboration, ATLAS Tile Calorimeter Technical Design Report, CERN/LHCC/96-42, 15 December 1996; L. Price, ATLAS Tile Calorimeter, Proceedings of this conference, to be published. 3. ATLAS collaboration, ATLAS High-Level Triggers, DAQ and DCS Technical Proposal, CERN/LHCC/2000-17, 31 March 2000. 4. Creative Electronics Systems, product catalogue, Geneva, Switzerland; http://www.ces.ch. 5. ATLAS Collaboration, ATLAS Muon Spectrometer, CERN/LHCC/9722, 31 May 1997.

CATHODE STRIP CHAMBER PERFORMANCE OF THE CMS ME1/1 MUON STATION YU. ERCHOV, I. GOLUTVIN, N. GORBUNOV, I. GRAMENITSKY, V. KARJAVIN, S. KHABAROV, V. KHABAROV, YU. KIRYUSHIN, V. LUSIAKOV, G. MECHTCHERIAKOV, I. MELNICHENKO, P. MOISSENZ, S. MOVCHAN, V. PALICHIK, V. PERELYGIN, D. SMOLIN, A. ZARUBIN, AND E. ZUBAREV Joint Institute for Nuclear Research, 6 Joliot Curie, Dubna, Moscow reg., 141980, E-mail: [email protected], [email protected]

Russia

O. DVORNIKOV, N. SHUMEIKO, A. SOLIN, AND V. TCHEKHOVSKI National Center of Particle and High Energy Physics, Minsk,

Belarus

Cathode strip chamber (CSC) was chosen as the baseline detector of the CMS Endcap Muon System. First forward muon station ME1/1, located in the solenoid magnetic field, is the key station because it should provide precise matching between Muon Sytem and Inner Tracker. The ME1/1 CSC spatial resolution must be about 75 um per station to achieve required muon momentum resolution and timing resolution of few ns for bunch crossing identification. Uncorrelated background rate is 1 kHz/cm2 or 100kHz per readout channel. The solenoid magnetic field is up to 4 T. CSC design is described. R&D results of the CSC performance study are presented: spatial resolution, timing resolution, rate capability, track reconstruction efficiency, influence magnetic field on CSC parameters. Test of many CSC prototypes demonstrated that ME1/1 CSC performance meets CMS requirements.

1.

Introduction

The innermost first forward muon station ME1/1 is placed in a slot between the Endcap hadron calorimeter HE and the return yoke disk YNl. 1 ME 1/1 station has to provide spatial resolution a x < 75 (im, timing resolution a, < 4ns and give precise matching with inner tracker. It should operate in strong axial CMS magnetic field up to 4 T at highest background rate in CMS Muon system up to 103 particles/cm2-sec (100kHz per readout channel). There are 36 ME 1/1 6-layers cathode strip chambers, CSCs, per each Endcap. CSCs are arranged to form a disk with the angular acceptance 1.6 (azimuth) and R (radius) coordinates for muon registration. The overall sensitive area covered by the chambers is more than 34 m2. CSC represents a 10° sector of the CMS ME1/1 Endcap muon station. It is a unit of six trapezoidal proportional chambers, layers, with cathode strips and anode wires readout } ' 2

347

348 2.

CSC design

Each layer of CSC is formed by two cathode electrodes with the gap of 7.0 mm and anode wires electrode in the middle. The panel consists of a "honeycomb-like" structure sandwiched between two electrodes, one with continuous copper surface and the other with milled strip pattern. The "honeycomb-like" filler of rectangular shape grid is made out of 0.5 mm FR4 strips. The size of a cell is 20x40 mm2. Electrodes are made out of 0.8 mm single side copper clad FR4 sheets. The thickness of the copper lamination is 18 urn. The unflatness of the panel each side is in a range of +50 urn. Center of gravity of charges induced on strips gives the precise measurement of coordinate along the anode wire. This radial strip structure covers the tp-angle range of ±5.42° to provide the overlapping with the neighboring CSCs. For strip electrode production a milling machine has been designed and assembled. The radial shape of strips made by a 0.35 mm thick rotating diamond disk. A crosscut of the strips provides a radial split of strips into two groups in order to minimize background rate per cathode channel. Strips length is equal 1060mm in the CSC top part (1.6
349 pads to the strip-electrode. The rubber seal is used to isolate the CSC gas volume from the atmosphere. The CSC assembly procedure is the following. Each panel has two reference bushings glued with the accuracy of ±25 ujn respect to strips. The two alignment pins are inserted into the panel reference bushings of the first panel. The second panel is put on the first one. Pins are inserted into the reference bushings of the second panel and so on. Finally CSC is screwed by 35 M6 bolts. 3.

CSC spatialresolutionand muon trackreconstructionefficiency in the presence of uncorrelated and correlated background 3.1 Uncorrelated background

P3 prototype was placed into the superconducting magnet Ml at H2 beam line (CERN) and turned to 10° in respect to beam line in order to provide a ratio of radial (BR) and axial (Bz) components of magnetic field of BR/BZ=0.1. In this experiment a deterioration of CSC spatial resolution vs. pion beam intensity have been studied (uncorrelated background). ' The trigger counters have monitored the beam intensity. The trigger count and beam profile have been taken for calculation of the strip counting rate. To study the background contribution to the registered events one P3 layer have been taken as testing while the five others - for pion track reconstruction. Layer spatial resolution as a function of strip counting rate was measurement. At counting rate of 100 kHz/strip the spatial resolution of single CSC layer is about a = 70 u.m. The track is regarded as efficient if strip clusters are found at least in 4 out of 6 CSC layers (criterion 4/6) with measured coordinates within the ± 3 a corridor. Track reconstruction efficiency is about 97.5% at counting rate of 100 kHz/strip. 3.2 Correlated background The Integrated Test experiment was carried out to test CMS Endcap prototypes. The prototypes of the preshower, the electromagnetic calorimeter (ECAL), the hadron calorimeter (HCAL) and the muon detector (ME1/1) have been installed into the superconducting magnet Ml at the H2 beam line (CERN). The axial magnetic field was oriented along the beam. The value of the magnetic field in the ME1/1 was equal to 2.5-3.0 T. Data taking was performed with a negative charge muon beam having momenta of 100, 150, 225 and 300 GeV/c. In this experiment a deterioration of CSC layer spatial resolution and muon track reconstruction efficiency in the presence of the secondaries have been studied (correlated background).1'5

350

The experimental data have shown that the high energy muons (100-300GeV) produce 20-25% of events with electromagnetic secondaries. CSC layer spatial resolution can be presented by two Gaussians: "good'Vsingle muon tracks with RMS=75 urn and muon with secondaries - RMS=lmm. Total muon track reconstruction efficiency for "good" muon tracks and muons with secondaries is 99%. Only 1% of muon tracks is lost. Such high efficiency is obtained due to splitting wide strip clusters into subclusters and restoring the charge on strips with overflow.6"8 4.

CSCtimingresolution

P3 prototype has been installed at Gamma Irradiation Facility (GIF) at X5 beam line (CERN).1 It has been turned to 10° in order to correspond to ME1/1 acceptance. The 137Cs radioactive source (740GBq) provides 662keV gamma background. A combination of filters is used to change the absorption factor from 1 (about 2x10 y/cm2sec on P3 surface) to 104. Trigger counters separate the muon beam. The prototype has been instrumented with anode readout electronics based on groups have been used for readout of 430 cm 2 sensitive area. The full width of the anode spectrum (99% of events) is equal to 25ns. The ratio strips/anode time spectrum width is about 1.3.

Figl Anode time spectrum and time distributions of timing signals. The full width of time distribution (99% of events) of the first output signal from 6 layers (1/6) is 10.5ns. The same value for 2nd signal - 12ns, 3"* signal - 14ns, 4th signal 16.5ns, 5th signal - 22.5ns and 6th signal - 26.5ns. The efficiency of timing signal (2 nd ) and charged particle track identification signal (4 th ) is about 98% within 25ns gate. Bunch-crossing (BX) at LHC will occur every 25ns. This means that for unambiguous BX identification 2nd, 3 ^ , 4th and even 5th signals can be used (2nd signal is baseline).

351

The study of background influence on BX identification have been made with Cs radioactive source. Insignificant influence of background on 1/6, 2/6, and 4/6 timing signals and muon identification efficiency was found for GIF flux attenuation factor 35 (rate - 100kHz per anode channel). At high background rates the probability of random coincidences from y-rays grows up. The generation efficiency of the false anode BX signals and "muon" tracks is negligible (less than 0.1%) for the rate about lOOkHz/ch. 137

5.

Conclusion

Test results of ME 1/1 prototypes equipped with Minsk front-end electronics confirmed that ME1/1 CSC performance meets CMS requirements. We thank CMS colleagues for the help, useful discussions and critical remarks. References 1. CMS. The Muon Project. Technical Design Report. CERN/LHCC 97-32, CMS TDR 3, Geneva, Switzerland, 1997. 2. Yu.Erchov et al." P4 - the pre-production prototype of the ME1/1 CSC", JINR Communication E13-2000-26, JINR, Dubna, Russia, 2000. 3. Movchan S., Moissenz P., The Method of Anode Wire Incident Angle Calculation of the First Muon Station (ME 1/1) of the Compact Muon Solenoid Set Up (CMS), Particles and Nuclei Letters No4[107]-2001, JINR, Dubna, Russia, 2001. 4. Golutvin I. et al., The Rate Capability of the CSC Readout Electronics, Particles and Nuclei Letters No4[107]-2001, JINR, Dubna, Russia, 2001. 5. Golutvin I. et al., Increasing of muon-track reconstruction efficiency in ME1/1 Dubna prototype for the CMS/LHC, JINR Rapid Communications Nol[93]-99, JINR, Dubna, Russia, 1999. 6. Golutvin I. et al., Robust estimates of track parameters and spatial resolution for CMS muon chambers, Computer Physics Communications 126 (2000) pp.7276. 7. Zubov K., Karjavin V., Movchan S., Moissenz P., Data Analysis for Cathode Strip Chamber, JINR Communication P10-99-118, JINR, Dubna, Russia (1999). 8. Movchan S., Moissenz K., Moissenz P., Cathode Strip Chamber Transmission Function and Single Layer Spatial Resolution for Clusters with Overflow, JINR Communication P10-2000-108, JINR, Dubna, Russia, 2001.

T H E R U N 2 D 0 M U O N S Y S T E M AT T H E F E R M I L A B TEVATRON S. H A G O P I A N Dept.

of Physics,

Florida State University, Tallahassee, E-mail: [email protected]. edu for the D0

FL, 32312,

U.S.

A.

Collaboration

The Run 2 D 0 muon detector at the Fermilab Tevatron has three subsystems: Proportional Drift Tubes (PDTs), Mini-Drift Tubes (MDTs) and trigger scintillation counters. The P D T s were used in the 1992-1996 data taking run and provide tracking coverage for pseudorapidity |?j| < 1.0. The forward muon tracking system, new for Run II, uses planes of mini-drift tubes and extends muon detection to \i)\ = 2.0. Scintillation counters are used for triggering and for cosmic muon and accelerator backgrounds rejection. Toroidal magnets and special shielding complete the muon system. All subsystems interact with 3 levels of triggers. Level 1 generates trigger information synchronously with the beam crossing. Level 2 operates asynchronously with a maximum decision time of 0.1msec. All three muon detector subsystems use a common readout system based on a 16-bit fixed point digital signal processor, which buffers the data from the front-end, re-formats the data if accepted by Level 2 and sends it to the Level 3 trigger system, which is a farm of Linux workstations running software trigger filters. Muon triggers accepted by Level 3 are written to tape for offline reconstruction.

1

Introduction

The recent upgrade of the Tevatron Proton-Antiproton Collider at Fermilab near Chicago for increased luminosity up to 2x 10 3 2 cm~ 2 s - 1 and smaller beam bunch spacing of 396 ns requires a corresponding upgrade of the D 0 detector x2 ' . The Run 2 D 0 muon system will enable D 0 to trigger, identify and measure muons in the new high rate environment 3 . The central muon system has been supplemented with additional scintillator layers for triggering, cosmic ray rejection, and low momentum muon measurements. The Run I forward muon system has been completely replaced with scintillator pixels and minidrift tube chambers. New shielding has been added to decrease background rates. The muon trigger has been redone to accommodate the high trigger rate and increased number of interactions per beam crossing. The upgraded central tracking system consisting of the Central Fiber Tracker and the Silicon Microstrip Tracker improves the momentum measurement of muons as well as other charged particles. The detection of leptons (muons and electrons) is very important for

352

353

physics at the energy frontier. The study of intermediate vector bosons decaying into leptons will give precise measurement of the mass of the W, results on forward backward asymmetry and the measurement of anomalous gauge boson couplings. Multileptons are a signature of supersymmetric particles in many models. The search for leptoquarks and heavy vector bosons uses lepton final states. Massive stable particles can appear as slowly moving muon-like objects. Muons are also used to tag b-jets for B physics, top physics and higgs searches. 2

Central Muon Detectors

The central muon tracking system, with pseudorapidity coverage \r]\ < 1.0, consists of 94 proportional drift tube chambers built for Run I 4 . The A layer is between the calorimeter cryostat and the 2 Tesla muon toroid magnet. The A layer chambers on the top and sides have 4 decks to help in rejecting backgrounds, while those on the bottom only have 3 decks due to space constraints. The B and C layers outside the toroid have three decks each. See figure 1 for the layout. The chambers are rectangular aluminum tubes with 5.7 cm by 10 cm cells. The drift distance resolution is about 1mm. The momentum resolution from the PTDs is ~ 30% for muons with pr = 100 GeV/c, where PT is the momentum transverse to the beam direction. But when the muon track is matched with tracks from the DO central tracking system, the resolution is improved for all central muons. For muons with PT=100 GeV/c, the resolution using central tracking is ~ 15% . Layers of scintillator, called the Cosmic Cap, on the top and upper sides of the central muon detector were used in Run 1 to help reject cosmic rays. Coverage was completed for Run 2 with the addition of the Cosmic Bottom counters. A new layer of scintillators, called the A counters, was added between the A layer and the calorimeter 5 . These counters have segmentation of 4.5 degrees. The A counters are used for muon triggering, rejection of out-of-time scattered particles and identifying low pr muons. 3

Forward Muon System

The Forward Angle Muon Detection System, which consists of mini-drift tubes (MDTs) and pixel scintillators, is entirely new for Run 2. The Run 1 forward toroids are used, and new shielding has been added. The MDT system covers the region 1.0 < \rj\ < 2.0 6 . The mini-drift tubes have 8 cells of 1 cm x 1 cm cross-section, and are made of aluminum extruded combs and plastic sleeves. The A-layer chambers are in front of the forward toroid magnet and the B

354 V!!3K

ff

^

1

4X " ^ f ^

tV

i*7*<"*

V ^ M

*,,"'"• ' }

C«) o

.1

1 --10

i

i

i

J

1 -S

i

!

i

;

J

j

{.'

i

i

i

i

i

,i

5

Figure 1. The Rue 2 D 0 Muon Detector

and C layers are behind it (see figure l.).The layers are divided into octants with the length of the tube depending on its position in the octant. As in the central region, the MDT A-Layer has four decks of drift tubes and the B and C Layers have three decks each. The coordinate resolution is 0.7 mm/deck. The momentum resolution is 20% for low momentum tracks. The Muon Forward Scintillator Pixel system covers the same eta region 7 . The segmentation of 4.5 degrees matches the segmentation of the Central Fiber Tracker. The r\ segmentation is 0.1. The typical size is 20 cm x 30 cm. The counters are made out of Bicron 404A scintillator. Kumaiin W t S bars are used for light collection into PMTs. The scintillators are used for triggering and track reconstruction.

355

Large backgrounds in the forward direction in Run I were, in general, due to the interaction of beam jets with the forward elements of the DO detector and the accelerator hardware. For Run 2 shielding was built in several large moveable sections. These extend from the endcap calorimeters and contain the low beta quadrupole magnet inside a case of 20 inches of iron, six inches of polyethylene and two inches of lead.

4

Triggers and Electronics Upgrades

The DO Run 2 Trigger System consists of 3 levels 8 . Level 1 is a pipelined hardware stage. It processes information from individual subdetectors in Field Programmable Gate Arrays (FPGAs) in a decision time of 4.2 fis. The trigger accept rate, output from Level 1, input to Level 2, is lOKHz. Level 2 is a second hardware stage which uses Dec Alphas. It refines Level 1 information and adds more information if available with preprocessors for each subdetector. A global processor combines information from the subdetectors. Level 2 has a maximum decision time of 100 microseconds. The accept rate out of Level 2 is 1 KHz. Level 3 has two stages: a custom-built data acquisition system and a Linux farm of processors which makes the final trigger decisions. The farm does partial online event reconstruction and uses filters to accept or reject events. The decision time depends on the number of farm nodes, and is about 50 msec for the beginning of the run. The sustained trigger rate out of Level 3 is 20 Hz, with an output event size of 250 Kilobytes. In order to handle the high input data rate, the front end electronics of all the muon subsystems was upgraded. Digital signal processors (DSPs) are used to buffer and reformat the data 9 . The DSPs make muon stubs from hits and buffer the Level 1 accepted data from the front-end readout, while a Level 2 decision is pending. If the trigger is accepted by Level 2, the DSPs reformat the data and send it to the Level 3 trigger system. The muon trigger has 3 levels plus an extra trigger level between Level 1 and Level 2 called SLICs (Second Level Input Computers). 10 . Level 1 triggers uses wire positions, scintillator hits in the A, B and C layers and central, north and south octants to define and/or trigger terms. The SLICs use 80 DSPs to find muon stubs in from nearby hits in a single layer. Level 2 combines muons, calorimeter and central tracks. Timing, pr, eta, phi and quality values are calculated for all muon candidates. Level 3 uses muon hits, makes muon segments and combines them into muon tracks which are matched with central tracks and calorimeter information. Events passing the triggers requirements are written to tape.

356

5

Conclusions

Run 2 of the Tevatron has started. The upgraded D 0 Muon System is wellmatched to the upgraded Tevatron. It should do an excellent job of triggering, measuring and identifying muons. Very exciting physics with muons is just around the corner. Acknowledgments The success of the Run 2 Muon Detector and associated triggering and software is due to many years of hard work by the Run 2 Muon Detector, Muon Algorithm and Muon ID groups. My special thanks to Dmitri Denisov, Tom Diehl, Neeti Parashar, Steve Doulas, Dave Hedin and Christos Leonidopoulos for their input, advice and help in preparing this talk and paper. References 1. DO Collaboration, S. Abachi et al, "The D 0 Upgrade - the Detector and its Physics", FERMILAB-PUB-96-357-E (Oct. 1997) 2. T. LeCompte and H.T. Diehl, Annu. Rev. Part. Sci., 50, 71 (2000). 3. DO Collaboration, "The D 0 Upgrade - Forward Preshower, Muon System and Level 2 Trigger", D 0 Note 2894, FERMILAB-FN-641. 4. DO Collaboration, S. Abachi et al, Nucl. Instrum. Methods A 338, 185 (1994). 5. DO Collaboration, B. Baldin et al, "Technical Design of the Central Muon System", D 0 Note 3365. 6. DO Collaboration, G. Alexeev et al, "Technical Design of the Forward Muon Tracking Detector Based On Mini-drift Tubes", D 0 Note 3366. 7. DO Collaboration, V. Abramov et al, "Technical Design of the D 0 Forward Trigger Scintillation Counters", D 0 Note 3237. 8. G. C. Blazey (D0 Collaboration), "The D 0 Run II Trigger", Xth IEEE Real Time Conference, Beaune, France, 22-26 September 1997, Editor C.E. Vandoni, pg 83-87. 9. N. Parashar et al, "Real-Time Data Processing in the Muon System of the D 0 Detector", IEEE Trans. Nucl. Sci, 47, 276, (2000), FERMILABConf-01-083-E (Jul. 2001) 10. C. Leonidopoulos, "The Muon Trigger at D0-,- to be published in the Proceedings of CHEP2001, Beijing, China, Sept. 3-7, 2001.

THE D 0 CENTRAL TRACKER TRIGGER SILVIA TENTINDO REPOND Department of Physics, Florida State University, Tallahassee, FL 32306, USA E-mail: [email protected] The goals of high energy physicists for the next decade require new designs for the online systems of collider experiments. We describe the new D0 Central Tracker Trigger (CTT) System, which makes heavy use of field programmable gate arrays (FPGA) and digital signal processors (DSP) to allow the system to cope with the greatly increased data rate anticipated at the Fermilab Tevatron. We describe briefly how the CTT system meets the physics goals of the collaboration.

1

Introduction

The upgraded Fermilab Tevatron pp Collider, which operates at a center of mass energy of 1.96 TeV, will reach an instantaneous peak luminosity of 1032 c m - 2 s _ 1 and eventually of 5 x 1032 c m - 2 s _ 1 . The bunch crossing time is 396 ns, and later will be reduced to 132 ns. The quest for new physics has required an upgrade of the D 0 detector: the upgrade includes a better Muon Trigger, new electronics for the Calorimeter and data acquisition system, and a new Central Tracker (Fig. 2), comprised of a Silicon Microstrip Tracker (SMT), a Central and Forward Fiber Tracker (CFT), a Central and Forward Preshower (CPS and FPS) and a Central Solenoid. The upgraded detector allows the measurement of charged particles momenta, improved tracking resolution, identification of secondary vertexes and enhanced identification of electrons and photons. The main characteristics of the upgraded D 0 are summarized in Table 1. The improved momentum resolution and the tagging of long lived particles (through secondary vertex and impact parameter measurements) will allow for an enriched sample of events with b quarks in the final state. In order to record such events with high efficiency a powerful trigger system is required. The larger background to signal ratios need a trigger with higher processing and rejection power. The reduced beam crossing time means that processing must happen faster (algorithms must be run in parallel processors) and with reduced dead times (pipelining must be used). The large scale and systematic use of Field Programmable Gate Arrays (FPGAs), Digital Signal Processors (DSPs) and Alpha Processors has allowed the D 0 collaboration to fulfill all these requirements.

357

358 Table 1. Characteristics and performance of the D0 detector

Characteristics momentum resolution vertex reconstruction primary secondary minimum Pt for muons minimum Pt for electrons

2

Performance APt/Ptz = 0.002 a — 15 — 30/zm a = 35jum (r, ) cr = 80fJ,m (r, z) |r/| < 2 (P t > 1.5Gev/c) \r]\ < 2.5 (Pt > IGev/c)

The D 0 Trigger

The D 0 trigger system 1 , as shown in the block diagram of Fig. 1, is a multilayer hardware structure composed of three levels called Level 1, Level 2 and Level 3. The input rate to LI from the detector front ends is 8 MHz. Since the events can be written to tape at a rate of 20-50 Hz, this implies that a rejection factor of 4 x 105 must be obtained by D 0 in Run II. The trigger systems of D 0 share the same architecture. The Level 1 Trigger (LI) in Fig. 1 includes information from the Calorimeter and Muon System, the Central Fiber Tracker (CFT) and the Central and Forward PreShower (CPS and FPS). The last three systems provide information to the LI Central Tracker Trigger (L1CTT). In the L1CTT special boards provide the interface to the Level 2 Silicon Track Trigger preprocessor (L2STT). The D 0 Level 2 Trigger (L2) shown in Fig. 1 is comprised of two stages: an array of preprocessors that format and sort LI data, and a Global preprocessor that correlates information across the whole detector. The input rate from LI to L2 is 10 KHz, the output rate to L3 is 1 KHz. L2 uses Alpha processors 2 : one Alpha controls all data paths (LI, L2, L3) to a trigger control computer, and the others process data. Among the preprocessors, the L2PS, L2CFT and L2STT constitute the L2 Central Tracker Trigger. The L2CFT receives tracks from LI CFT, sorts them by Pt and truncates their number before sending them to L2 Global. The L2STT processes the Silicon Microstrip Tracker data and improves tracking resolution by merging the information from the CFT and the SMT. Tracks are sorted by impact parameter, and optionally vertex information is provided by the L2STT processor.

359

Figure 1. The DO Trigger System.

3

The Fiber Tracker, PreShower and LI Central Tracker JL3, J. Is, Kg *5 J.

The LI Central Tracker Trigger 3 uses the information from two detectors: the CFT and the PS (see Fig. 2). The Central Fiber Tracker detector (CFT) contains 16,000 channels of electronics, connected to scintillation fibers, 2 degrees pitch, mounted on eight cylinders (20 < r < 50 cm) of alternating axial and stereo doublets. The CFT provides tracking in the central region. Its position resolution is about 400/am. The PreShower detectors, Central and Forward (CPS and FPS), contain 16,000 channels, where signals are generated in triangular scintillator strips. The position resolution is about SQOfim. Hits from the CFT are used in L1CTT to form tracks or preshower clusters; the number of tracks/clusters is counted and several categories are made before the tracks/clusters are pipelined for output to L2. Heavy use is made of FPGAs to implement the algorithms, as well as to perform interface and framework tasks. FPGAs (XVC600 and XCV400) are used to find tracks in 4 Pt bins. In the lowest bin there are about 8,000 track equations, stored as

360

Solenoid Figure 2. The Central Tracker System of D 0 :

Look Up Tables. Another FPGA (XCV300) does track/cluster matching 2 . 4

The Silicon Tracker and the L2 Central Tracker Trigger

One essential component of the Central Tracker Trigger of D 0 is the Silicon Tracker Trigger (L2STT) 4 , that receives the information from the Silicon Microstrip Tracker detector (SMT) (Fig. 2). The SMT consists of 800,000 channels, half of which are used by the trigger. The SMT is composed of six cylindrical sections where rectangular detectors are arranged in four concentric layers. All the axial strips have a pitch of 50^m , and the stereo strips, at 2 or 90 degree angle wrt to the axial direction (direction parallel to the beam), have 60 — 150/x m pitch. The hit resolution is 15p m, and the impact parameter resolution obtained from offline reconstruction is about 18/x

361

m. The secondary vertex resolution is 35/x m (in r — cf>), and 80/x m (in r — z). In contrast to the other L2 preprocessors, the L2STT hardware is based on custom designed VME boards. The L2STT system contains six VME crates. In each crate there is one Fiber Road Card (FRC), nine Silicon Trigger Cards (STC) , and two Track Fitting Cards (TFC). Each STT crate receives the information from one SMT sector, 60 degrees in azimuth. Each FRC contains three Altera FPGAs (FLEX10K) where the tracks are received from the L1CFT, formatted, buffered and transmitted to L2 and L3. Each STC card receives SMT hits via a VBD (VME Buffer Driver) Transition Module, and LlCFT tracks via the FRC. A set of FPGAs performs cluster calculations from the SMT hits; the bad hits are rejected after gain correction and calibration. Downloaded Look Up Tables allow for rapid calculations. Another set of FPGAs in each STC receives the LI tracks and associates clusters to a CFT track. The STC data are forwarded to TFC track cards, each of which uses eight Digital Signal Processors (DSPs) TI-TMS20C6203 for fitting, and three Altera FLEX10K100 FPGAs for the control logic on the board 5 . 5

Physics Potential of the Central Tracker Trigger

The expected performance for the CTT is based on detailed simulation studies 4 . Typically, for a 10 GeV track, the Pt resolution is 18% (if one CFT layer is used in the tracking), 9% if eight CFT layers are used in the tracking, and 6% if the SMT is used for tracking also. The tracking efficiency in L2STT reaches about 85%. Pt measurements, as well as impact parameter measurements, will be used at reconstruction level for analysis in many physics channels. At the trigger level (L2STT), cuts on Pt and use of trigger objects that include the impact parameter will increase the ratio signal/background making the rate of interesting events compatible with the band width in L2, without a significant reduction in efficiency. The inclusion of the STT in the L2CTT improves the Pt resolution by a factor 2-3. The improved Pt resolution can be used in an Et/Pt cut for electrons at L2. For electrons of Pt — 5 GeV the better resolution allows a background rejection improvement of a factor 4, and at Pt = 20 GeV, of a factor 2. The physics channels that will benefit the most from this improvement include: Top production. Ratio signal/background (s/b) improves by a factor 2 to 7. Higgs associated production. Ratio s/b improves by a factor 7. Z boson electroweak production. Ratio s/b improves by a factor 10.

362

B physics . Though at least one lepton is still required at LI, an increased s/b of a factor 3 is obtained at L2. References 1. J. Linnemann, The D0 Trigger System, Proceedings of this Conference. 2. D. Edmunds et a/., Technical Design Report for the Level2 Global Processor, D 0 Note 3402 (February 98). 3. M. Martin et al, CTT Technical Design Report, D 0 Note 3551 (August 99). 4. The D 0 Collaboration, A Silicon Trigger for the D0 Experiment in RunlI, D 0 Note 3516 (September 98). 5. W.Taylor for the D 0 Collaboration, An Impact Parameter Trigger for the D0 Experiment, IEEE Trans. Nucl. Sci., vol.48, pp. 557-561, June 2001.

A PROPOSAL FOR THE ALIGNMENT OF THE LHCB RICH DETECTOR ANTONIS PAPANESTIS Rutherford

Appleton Laboratory, Chilton, Didcot, E-mail: [email protected]

0X11

OQX,

UK

The strict requirements in Cherenkov angle resolution combined with the size and complexity of the LHCb RICH2 detector will pose a challenge in the quality and alignment of the mirrors. There are 28 spherical and 20 planar mirror segments in each half of RICH2, and the required Cherenkov angle resolution is below 0.6 mrad. We have studied extensively the mirror misalignment effect on Cherenkov angle resolution and particle identification performance, and to what extent alignment with data can compensate for mirror misalignment. Our results show that the initial mirror alignment must be better than 1 mrad to achieve a negligible effect of the alignment on the total Cherenkov angle resolution.

1 1.1

Introduction The LHCb RICH Detector

Particle identification is a fundamental requirement of the LHCb experiment. The RICH detector has been designed to provide particle identification, and especially pion-kaon separation, in the momentum range 2-150 GeV. It consists of two separate RICH detectors and three different radiators. RICH1 is positioned just after the vertex detector and covers the full LHCb acceptance. It has two radiators, aerogel and C4F10, four segmented spherical mirrors and provides particle ID up to 70 GeV. RICH2 is using a lighter radiator, CF4, to separate pions from kaons up to 150 GeV. The two spherical mirrors cover 120 mrad in the horizontal plane and 100 mrad in the vertical plane. In order to keep the photo-detectors outside the LHCb acceptance, a second set of flat mirrors is used. The spherical mirrors are made from 28 hexagonal segments, and the flat mirrors from 20 rectangular segments. 1 The information of both detectors is combined in the particle identification algorithm. 1.2

Alignment

The accurate reconstruction of the Cherenkov angle for each photon from a given track assumes accurate knowledge of the geometry and orientation of the mirrors, in both RICH1 and RICH2. Ideally, the segments that form the spherical and plane mirrors should be aligned to form one single mirror. However, it is still possible to reconstruct the Cherenkov angle without error,

363

364

Figure 1. A Cherenkov ring resulting from mirror misalignment. Point C is the real centre of the circle, but point C , derived from the track, is used to calculated the Cherenkov angle.

if the orientation of every mirror segment is known, and if the segments that each photon was reflected on can be identified. As a result, the contribution of the alignment to the total error in the reconstructed Cherenkov angle will depend on the misalignment of the mirror segments, the accuracy with which their orientation can be determined and the number of photons that cannot be assigned to specific mirror segments. 2

Determination of mirror orientation parameters using data

Cherenkov photons emitted in the radiator from each particle track are distributed on a circle on the photo-detector surface, with the radius representing the Cherenkov angle and the centre given by the track. Mirror misalignment has the result of moving the circle, but not the centre, as in Figure 1 where all distances represent angles. The real centre of the circle is C, but the point given by the track is C . So the measured Cherenkov angle is i9c/,, while the real Cherenkov angle is i90-2 fich - t?o = dcos{cj)ch + 4>o) or $ch - i?o = acos(<j>ch) + bsin{(j)ch)

(1)

A plot of "&ch - i?o against <j>ch can be made for each mirror segment, and the parameters a and b can be extracted from the fit. As a result of the uncertainty of the photon emission point, there is an ambiguity on the mirror segments upon which each photon was reflected. To avoid this, the Cherenkov angle solution was calculated for two photons, one emitted at the beginning of the radiator and one at the end. If both photons were reflected on the

385

Background subtracted

Figure 2. A 2D histogram of A0 against 0cfc. The background was subtracted by applying a cut at 30% of the maximum on each column separately.

same mirrors, the photon was declared unambiguous and was used to fill the histogram for the particular mirror. Photons coming from other tracks act as background in the histograms, which can be easily subtracted (Figure 2). 2.1

RICH1

BICH1 is a one mirror system, so the parameters a and b give directly the absolute horizontal and vertical tilts of each mirror segment. This can then be used in the reconstruction program with very good results. 2.2

RICH2

RICH2 is a two mirror system, so a different histogram must be filled for every spherical/flat mirror combination. Furthermore, the parameters a and b give the total tilts of the two mirrors in each combination. The only way to extract the parameters for the individual segments is to minimise a function of the form:

£ [(aij - xj - Xj)2 + (bij - yt - y/)2]

(2)

where i counts the spherical segments, j the flat mirror segments, a and b are the measured total tilts and x and y are the tilts of individual mirror segments.

366

All photons Perfect —— Aligned —— Not oligned -

Jl

Perfect Unambiguous' Ambiguous photons

RMS 0.58 RMS 0.67

"gUarer (Inn-ad mtoaMgnnwtt) '

RMS 0.58 RMS 0.59 RMS 0.90

anal* arrar (1mr«) tninManmant)

Figure 3. The effect of misalignment in the resolution of the Cherenkov angle in RICH2.

With this procedure it is possible to align the mirrors relative to one mirror, but not absolutely as any tilt of the spherical mirrors can be compensated by assuming an equivalent tilt in the opposite direction of the flat mirrors. In practice, the error on the measured Cherenkov angle depends on the geometry and the path followed by the photon. Since every mirror combination has a limited acceptance of photon angles, it is possible to measure a "calibration" constant for each mirror, using the LHCb Monte-Carlo simulation. Equation 2 can be changed to: (dij-pfjXf

J 2

"


+

(bij-plvt-qfjyjr

(3)

^n

to include the calibration constants p and q (which depend on the mirror combination) and the errors in the measurement of a and b. This expression can then be minimised using Minuit 3 , or a similar program, and extract all the mirror tilts relative to a chosen mirror segment. 3

The error in the Cherenkov angle and particle identification performance

As mentioned in the introduction, the initial misalignment of the mirror segment will affect the performance of the detector as there is no plan to move the mirrors during data taking. The percentage of ambiguous photons in RICH2 is 20, and, by looking at the error in the Cherenkov angle reconstruction, about 30% (6% of the total) of them are assigned to the wrong mirrors. We performed a case study by misaligning all the RICH mirror segments by 1 mrad in a random direction. During the study we found that it is beneficial to iterate the procedure more than once on the same data.

367 Table 1. Comparison of performance (%) between a perfect and an aligned detector Particle type Perfect Aligned

3.1

e 36 31

Efficiency IT K 74 29 69 28

P 93 94

e 26 23

Purity K 74 50 73 47 7T

P 79 78

RICH1

After the alignment procedure the sigma of the reconstruction of the Cherenkov angle was 1.4 mrad, the same as a perfectly aligned detector. 3.2

RICH2

The effect of misalignment in the resolution of the Cherenkov angle in RICH2 can be seen in Figure 3. On the left we see the error distribution before and after alignment compared with a perfectly aligned detector. Even after alignment the error is increased mainly due to ambiguous photons as can be seen on the right. The distribution of the unambiguous photons is almost identical to the perfect detector, but the error of the ambiguous photons is increased significantly. Table 1 shows the particle identification performance of the RICH2 detector (particles above 70 GeV) for this case study. It is obvious that after alignment the performance has been affected. 4

Conclusions

We have conducted a case study by misaligning all the segments of the LHCb RICH detector by 1 mrad and have shown that it is possible to align the detector using data to 0.1 mrad. However, ambiguous photons with wrongly assigned mirrors in RICH2 have a non-negligible effect on the detector performance and therefore we conclude that the initial alignment of the mirrors must be better than 1 mrad. References 1. S. Eisenhardt The LHCb Ring Imaging Cherenkov Detectorsin these proceedings. 2. A. Gorisek et al, Nucl. Instrum. Methods A 433, 408 (1999). 3. http://wwwinfo.cern.ch/asdoc/minuit

MONITORED DRIFT TUBE CHAMBER PRODUCTION AT LABORATORI NAZIONALI DI FRASCATI S. BRACCINI (ON BEHALF OF THE ATLAS FRASCATI GROUP) Istituto Nazionale di Fisica Nucleare, Laboratori Nazionali di Frascati, Via E. Fermi, 40 - 1-00044 Frascati (Rome), Italy E-mail: [email protected] The muon Spectrometer of the ATLAS experiment is instrumented with precision tracking chambers made of layers of drift tubes assembled together with very high mechanical positioning accuracy. The chambers are arranged in three stations (inner, middle and outer). The middle station in between the coils of the toroid magnet in the barrel is built at Laboratori Nazionali di Frascati of INFN, where methods for the construction of the drift tubes and the precision assembly of the chambers have been developed. These methods aim at both achieving the required performance of the detector as well as a high level of automation in the serial production of the chambers. The construction methods and the first results of the production are presented.

1

Introduction

High momentum final state muons are the most promising and clean signature of physics at high energy hadron colliders. In order to fully exploit the discovery potential of the LHC machine, a very accurate measurement of the momentum of the muon tracks is mandatory. To reach this goal, the muon spectrometer of the ATLAS experiment is designed to measure the momentum of the muons with the very high precision of Ap/p ~ 10~ 4 x p [GeV] in the stand alone mode. The pseudorapidity range \TJ\ < 1.0 is covered by the barrel spectrometer which is based on three stations of precision drift chambers located inside a large air-core toroid super-conducting magnet x , as shown in Figure 1. These chambers, denominated Monitored Drift Tube (MDT) chambers 2 , are composed of successive layers of drift tubes glued together. All the chambers are equipped with an optical in-plane alignment system to monitor possible deformations. To obtain the required accuracy on the measurement of the momentum, the very high mechanical precision of 20 fiia r.m.s. on the positioning of the wires must be attained. The middle station in between the coils of the toroid magnet is completely built at Laboratori Nazionali di Frascati (LNF) and is composed of

368

3i9

Figure 1. A union track detected by the AT- Figure 2. Schematic view of a MDT chamber LAS muon spectrometer. of the middle station. 94 MDT precision chambers, denominated BML (Barrel Middle Large) chambers, equipped with about 30000 drift tubes. These chambers have two multilayers with three layers of drift tubes each separated by a light aluminum support structure called spacer made of two longitudinal beams and three cross plates, as shown in Figure 2. The length of the chambers is 360 cm while the width ranges from 70 cm up to 170 cm. The construction of such a large number of precision chambers is based on specific techniques able to match the requirements of precision and automation for serial production. 2

Tube production

The basic detection element of the precision BML chambers is a 30 mm diameter, 380 cm length aluminum drift tube with a W-Re 50 (i,m diameter central wire. The tube operates with a gas mixture of Ar CO2 at 3 bar absolute pressure. The assembly of the drift tubes is performed by the "Tub-o-matic" wiring machine shown in Figure 3, completely designed and constructed at LNF. All the operations for the assembly of a drift tube are performed automatically and the intervention of an operator is required only in case of problems. The main operations for the assembly of a drift tube axe summarized here. The wire is first passed through the aluminum tube by a pneumatic system and then is passed through the two high precision end-plugs which assure the

370

Figure 3. The "Tub-o-matie" automatic Figure 4. The gas leak test station based on wiring machine. a helium spectrometer.

correct positioning of the wire inside the tube. The end-plugs are crimped within the aluminum tube by a high pressure pneumatic system and jaws crimp the wire at one end. The wire is pre-tensioned at 450 g for about one minute before the nominal tension of 350 g is applied and the wire is crimped at the other end. With this machine a production rate of 50 tubes per day is easily achieved, consistent with the needs for the assembly of the chambers. As soon as a tube is assembled, the tension of the wire is measured by exciting the wire with a variable frequency square wave voltage which allows to determine the mechanical resonant frequency. The wire tension is required to be within 350±15 g. For each tube the dark current is required to be less than 7 nA and the gas leaks less than 1 0 - 8 bar liter/s in Ar. The gas leak test station, shown in Figure 4, designed and constructed at LNF, is based on a Helium spectrometer. The tension of the wire is measured again about 30 days after the wiring in order to detect creeping effects. About 98% of the tubes are found to fulfill all these specifications. 3

C h a m b e r assembly

The assembly of the MDT chambers is based on the gluing of successive layers of very accurately positioned drift tubes. In order to position the tubes of one layer with the required accuracy of 20 fim r.m.s., a very precise jig on a granite table has been conceived and constructed at LNF. As shown in Figure 5, the jig is composed of precision aluminum combs spaced every 50 cm. The combs localize the tubes and keep them in position by vacuum suction. The correct positioning of the wires with respect to the jig is assured

371

Figure 5. The high precision jig on which the Figure 6. The assembly of a MDT chamber tubes are positioned. at LNF.

by the high mechanical precision of the end-plugs of the tubes. The planarity of each layer is checked by a system based on a micrometer in order to avoid interference effects between consecutive layers. Once a layer of tubes is positioned on the jig, it is glued to the rest of the chamber which is lowered on it by a crane, as presented in Figure 6. In order to obtain a positioning within the required accuracy, a system based on six precision spheres attached temporarily to both the ends of the cross plates is adopted. These spheres are positioned into precision cones which are located upon accurately constructed interfaces. There are three sets of six precision interfaces with equal height within 10 yaa. Each set is used to assemble two layers of tubes symmetrically with respect to the spacer. To remove overconstraints, only one sphere is completely fixed and only one sphere on the same reference side is free to move only in the direction along the tubes to allow for temperature dilatation. All the other four spheres are free to move in a plane parallel to the granite table. The positioning of the chamber and its displacements during all the steps of the assembly procedure are monitored by an optical system. By this system, a reproducibility of the positioning within 5 fim is achieved. The deformations of the chamber are also monitored by the in-plane optical alignment system shown in Figure 2. In order to reduce the effect of the sag of the cross plates, a compensation force corresponding to the 80% of the weight is applied on the longitudinal beams (Figures 2 and 6). In this way the sag of the cross plates is constantly kept under 5 pm. The gluing rate is of one layer per day, due to the drying time of the glue. Taking into account the time required to build and glue the spacer, the minimum time to assemble one chamber is 7 working days.

372

In order to check the mechanical precision on the positioning of the wires, about 10% of the MDT chambers are measured by X-ray tomography 1 at CERN. Four of the chambers built at LNF have been measured and an average accuracy of 14.7/xm r.m.s. is achieved, showing good stability and reproducibility of these high precision serial construction methods. 4

Conclusions

In the year 2001 the serial production of the MDT chambers has started at LNF. More than 6000 drift tubes have been produced and 16 BML chambers have been assembled which represent 17% of the total. After a period in which the assembly procedure has been optimized for the serial production, the maximum production rate of 7 working days per chamber has been achieved, stable over more than four months. The results from the monitoring systems show good reproducibility of the positioning of the chamber during all the steps of the assembly. This is confirmed by the very good accuracy on the positioning of the wires measured by the X-ray tomography at CERN. Acknowledgments I am indebted to many colleagues of the ATLAS group at LNF for the very constructive discussions and suggestions. I would like to acknowledge here the important contribution of the technicians of the ATLAS group in Frascati E. Capitolo, G. Pileggi, B. Ponzio and V. Russo, in mastering the challenging technical aspects of the project and in operating smoothly and efficiently all the production facilities. The mechanical design of all the facilities was developed by C. Capoccia, G. Catitti and S. Cerioni of the LNF Research Division design department and realized by the LNF mechanical workshop. References 1. ATLAS Detector and Physics Performance Technical Design Report, CERN/LHCC/99-14 and CERN/LHCC/99-15; ATLAS Muon Spectrometer Technical Design Report, CERN/LHCC/9722. 2. M. Curatolo, The Monitored Drift Tube (MDT) chambers for the muon precision tracking in the ATLAS spectrometer, Nucl. Phys. B (Proc. Suppl.) 78 (1999) 422.

A DATABASE FOR D E T E C T O R C O N D I T I O N S DATA OF CURRENT A N D FUTURE HEP EXPERIMENTS S.PAOLI CERN,

CHI 211, Geneva

23,

Switzerland

This paper presents the Conditions Database, a C + + library (currently based on Objectivity/DB) developed by the Database group of the CERN IT Division, for use in the current and future (LHC) experiments at CERN. The package can be used to manage conditions d a t a of a detector into a database; these are the conditions that vary with time and are necessary for the reconstruction and analysis of raw data. The library provides an abstract interface, independent of the underlying DBMS, which means that moving from an implementation based on a certain DBMS to another one would have very little impact on the user code.

1

Introduction

The Database group 1 in the Information Technology (IT) Division at CERN was created in January 2000 to regroup the support of production services on the two database management systems currently deployed at CERN (Objectivity/DB 2 and Oracle 3 ). Several CERN experiments showed in 1999 interest for an early version of a Conditions Database package developed at SLAC for the BABAR experiment. This package was adapted to CERN specific environment and then supported in its adoption by the NA45 experiment. In this way, important knowledge on user and software requirements for such a package had been acquired. Subsequently, in spring 2000, several experiments at CERN requested significant enhancements on the product; as the design of the BABAR product was not easy to evolve, the change requests led to a complete new design and implementation of the package. For similar reasons also the BABAR product was redesigned at SLAC 4 . In this paper we present the main features of the Conditions Database package, the current status of development and the productions experience gained to date. 2

The purpose of the Conditions Database

The Conditions Database is a software package, based on a database system, to store, retrieve and manage conditions data. For the HEP experiments, conditions data are any detector conditions, as well as any other experimental

373

374

area condition data, that can be used to characterize the status of the detector. Examples are data from calibration, alignment, geometry, environmental parameters (temperature, pressure, etc.), any slow control parameter. Conditions data are needed by reconstruction and analysis programs in order to reconstruct the events from the raw event data, and to further analyse this data. With respect to the Conditions Database, conditions data are generic information data that characterize the status of a detector element at a given time; this data evolve with time, and has therefore an interval validity range. A detector is characterized by a great number of conditions, that have to be managed in a similar way. The picture in fig. 1 illustrates these concepts, showing the evolution of four conditions; a box indicates the validity interval of that condition value.

Figure 1. Evolution in time and version of four detector conditions.

The condition types for different measurements can be very different, varying from simple numeric types (e.g. temperature) to complex types (e.g. alignment). In any given condition, many different values can be inserted for the same time, as a result of different calculations, representing refined evaluations of the experimental data (see "VDET alignment" in the picture). Reconstruction and analysis programs will then look up in the Conditions Database for all necessary conditions at a given event time.

375

3

The architecture and features of the Conditions Database

The Conditions Database is a C + + library; the user will have to integrate the library within the experiment specific code. The library API has been designed in a way not to expose details of the specific DBMS solution; specifically, no C + + include files from the DBMS are exposed to the user code. Moreover the API is abstract; this means that the user code is not bound to any specific implementation; for instance, moving from the implementation based on a DBMS to a different implementation on a different DBMS should not require changes to the user code. The Conditions Database offers support for generic types of conditions data. The user defines conditions types ( C + + classes) by inheriting from a given abstract class provided by the package; two methods have then to be implemented by the users, to convert the object into a byte stream and vice versa. Although time is a natural parameter for the conditions validity, the Conditions Database does not require this; the parameter is a generic 64 bit integer, called CondDBkey. A user can use any type convertible to this 64 bit integer, and having the "greater than" operator. For example, by using an external library Anaphe/HepUtilities 5 , a time class can be used, and transparently converted to the 64 bit integer in the API calls. Features. The Conditions Database allows to organise detector conditions in a tree structure, similar to the file system. A user can structure a detector in sub components and elements (tree nodes), called CondFolderSet, and for each of them define the condition types (tree leaves) associated, called CondFolder. Once a CondFolder is created, conditions data values can be stored inside. The look up of conditions data is then done specifying the CondFolder and the value of CondDBKey to look at; this will return the most recent version of conditions data with validity range containing the given CondDBKey. An important feature, called Tag, has been added to facilitate the look up of chosen condition values, independently of future changes in the CondFolder. At a given time the status of a CondFolder can be associated to a string called Tag; this Tag identifies a complete sequence of condition values at any time. In this way, even if additional condition values are inserted in the CondFolder for past values of CondDBKey, the previously identified values can be easily retrieved by specifying the values of CondDBKey and the Tag. The picture fig. 2 illustrates the concept; by specifying CondDBKey=10 and Tag=" production 2000" the right condition value will be retrieved. Tags can be associated to single CondFolder, or collectively to the whole tree.

376 f version 2000" time Figure 2. Definition of a tag within a CondFolder.

Additional features are provided to browse the stored condition value, either vertically (i.e. at a given CondDBKey value) or horizontally (within a Tag). A prototype GUI browser has also been implemented in Java, accessing the C + + API via JNI, with the objective of investigating this technology, and to analyse possible user needs. The Objectivity/DB implementation. A first complete implementation has been done based on Objectivity/DB. This product was chosen because the first experiments that would have adopted the package were already using Objectivity/DB for their event data. To have good performance with the kind of data look up (by time) performed, the use of a B-tree index is necessary. Objectivity/DB offers scalable persistent collections based on B-tree; one of these, the ooTreeSet, has therefore been extensively used for all the collections of conditions data. Other persistent collections provided by Objectivity/DB have been used as well. This has resulted in a simple and performant design and implementation. Using the DBMS specific hint, in the Folder creation API, the user may pass the name of existing database files to be used by the Conditions Database; this flexibility is important for manageability. 4

Status and outlook

The project started in spring 2000 with requirements gathering and analysis. People from the participating experiments have actively contributed to the definition phase. The implementation and unit testing has then been performed by the IT-DB group, while the experiment members have performed the system testing, validation, and integration in their experiment specific software. This process has ensured user involvement and an important resource sharing. The first production release has been made in May 2001. The product is now in production in the experiments Harp and Compass at CERN. The

377

experiments LHCb and ATLAS have integrated it in their software framework GAUDI, and they are increasingly adopting it for test beams and preproduction activities. The Harp experiment is currently storing in the Conditions Database all non event data from the experimental area, and plans to store all other conditions data produced by the offline software in the same way soon. The total size of packed data is of the order of lOMb/day. Although this does not represent a high data rate, the value added by the Conditions Database is important for reasons of manageability and easy of use. It allows a fast and efficient data quality control, which results in maximum saving of precious beam time. The focus of the Conditions database support by IT-DB is now on production activities; this will bring important experience, and may trigger enhancements and new requirements. Certain experiments, using Oracle instead of Objectivity/DB as their DBMS solution, have also asked for an implementation of the Conditions Database on Oracle. Many other additional features identified for the medium term may also be implemented in the near future. Given the rather generic application requirements it is likely that other areas of applications of the Conditions Database, beyond HEP detector conditions, could be possible; this should be analysed, as it would certainly bring benefits to the product. 5

Conclusions

The development of the Conditions Database has been a successful example of collaboration between the CERN experiments and the IT division for the creation of a common solution. The product is currently in production in several experiments, and the user base is expected to widen, as soon as more experiments approach the production phase, where a comprehensive database solution for the conditions data is needed. References 1. 2. 3. 4.

CERN Database Home, h t t p : / / c e r n . c h / d b . Objectivity, http://www.objectivity.com. Oracle, h t t p : //www. o r a c l e . com. I.Gaponenko et. al. (2000) "An overview of the BABAR Conditions Database", CHEP 2000. 5. Anaphe, h t t p : / / a n a p h e . w e b . c e r n . c h / a n a p h e / h e p u t i l i t i e s . h t m l . 6. Conditions DB, h t t p : / / c e r n . c h / d b / o b j e c t i v i t y / d o c s / c o n d i t i o n s d b .

This page is intentionally left blank

379

Calorimetry Organizer: C. Leroy C. Leroy H. F. Heath S. Simion A. Ziegler F. Djama L. E. Price P. Sempere Roldan F. Steinbuegl V. Hagopian R. Orr J. Pinfold

Convener's Report Overview of the CMS Electromagnetic Calorimeter The Readout of the ATLAS Liquid Argon Calorimeter A New W/Scintillator Electromagnetic Calorimeter for ZEUS Performance of the ATLAS Liquid Argon Electromagnetic Calorimeter Modules under Test Beam Status of ATLAS Tile Calorimeter and Study of Muon Interactions Construction of the First CMS-ECAL Fully Operational Module (400 Lead Tungstate Crystals) A New Concept for an Active Element for the Large Cosmic Ray Calorimeter ANI What's New with the CMS Hadron Calorimeter Overview of the ATLAS Liquid Argon Calorimeter System The ATLAS Hadronic Endcap Calorimeter

This page is intentionally left blank

CALORIMETRY (Convener's report) CLAUDE LEROY Groupe de Physique des Particules, Departement de Physique, Universite de Montreal, C.P. 6128, Succ. "Centre-Ville", Montreal (Quebec) H3C 3J7 Canada; E-mail: [email protected] The calorimetry sessions put an emphasis on ATLAS and CMS experiments which are the two major general purpose experiments exploring new physics frontiers through the study of head-on collisions of protons accelerated each to 7 TeV at an expected peak luminosity of l.OxlO 34 c m - 2 s _ 1 at the Large Hadron Collider (LHC). Major progresses have been reported for both experiments. A large part of ATLAS calorimetry is using liquid argon as active medium. Robert Orr gave an overview of the ATLAS liquid argon system. These liquid argon calorimeters have the capability to achieve the measurement, with sufficient resolution, of the energy and direction of jets, electrons and photons, and Ej? tss over a wide pseudorapidity interval. Particle identification is provided , particularly 7/77° and e/7r separation. Robert Orr discussed the structure of the various liquid argon calorimeters regarding energy and angular resolutions to be achieved. The accordion structure has been adopted for the barrel electromagnetic calorimeter. The " Spanish fan" configuration has been selected for the electromagnetic endcap calorimeter to maintain a spatially constant sampling fraction avoiding the accordion structure difficult to realize in the region covered by this detector. Large irradiation levels will be encountered at LHC. In particular, the hadronic endcap calorimeter (HEC) and the forward calorimeter (FCAL) will be exposed to high fluences of particules and their design takes this constraint into account. The HEC is based on a conventional plate design, compact and radiation hard. The FCAL, covering the pseudorapidity range 3< \r)\ < 5, is the most exposed to radiation (1016 neutrons c m - 2 year - 1 ). It consists of three sections: FCAL1, FCAL2, FCAL3. The structure of FCAL2 and FCAL3 consists of a paraxial arrangment of tungsten rods inside copper tubes. The liquid argon gap between the rod and tube provides the ionization region. FCAL1 is very similar with the exception that the matrix and electrode rods are made of copper. FCAL construction is well on schedule and will be ready for the start of the ATLAS data taking campaigns. The FCAL having received a particular attention in Robert Orr's presen-

381

382

tation, the other parts of the ATLAS liquid argon calorimetry system were also covered in details by several speakers. We heard a presentation of Fares Djama reporting on test results of ATLAS barrel and end-cap prototype modules tested with electron beams at CERN. The energy resolution measured for the two prototypes are fulfilling the physics requirements. Module stacking and cabling are well underway with 25% of the modules completed by the time of the conference. The module production will continue towards commissioning in 2004. The HEC was presented by Jim Pinfold who gave a detailed report on its construction. The talk he gave presented the plans for assembly of the HEC and its insertion into the endcap cryostat. The results of test beam studies were reported. The response of the module to electrons, muons and pions were measured with electron and pion energy resolutions following the design and physics requirements. Jim Pinfold also described further tests to be performed and involving ATLAS electromagnetic and forward calorimeter modules. Finally, the schedule of assembly of the HEC was exposed. Stefan Simion provided a status report of the readout of the ATLAS liquid argon calorimeter. The readout system has been redesigned to improve its radiation hardness and therefore to comply with radiation tolerance requirements. In particular, six radiation hard ASICs have been developed and are being operated, after undergoing successful functional and radiation tests, on a new prototype of front-end electronic board. He noted also that DSM technology has been retained for the switched capacitor array (SCA) controller, the gain selector and for clock distribution circuitry. The front-end board serial configuration protocol and the optical output link both use DMILL technology. The only commercial ICs still used on the front-end board are the ADCs, the differential drivers for the SCA address bus and the G-link serializer for the optical output link. To conclude the series of presentations on ATLAS calorimetry, Larry Price presented the hadron ATLAS tile calorimeter which is a sampling device made out of steel as absorber and scintillating tiles as active material. It realizes a simple and very well proven idea of calorimetry. Wavelength shifting fibers collect the scintillation light from the tile at both of their open - azimuthaledges and bring it to photomultipliers at the periphery of the calorimeter. The full tile calorimeter consists of a barrel cylinder at the collision point which is 564 cm long and extended barrel cylinder 292 cm long at each end of the barrel and separated by a gap for detector services. Photomultiplier tubes and front-end electronics are packaged in drawers which are inserted into the girders at the outer radius of each module. Several talks presented the current developments and status of CMS. He-

383

len Heath presented the status of the CMS precision electromagnetic calorimeter (ECAL) made of lead tungstate (PbW0 4 ) crystals. The use of P b W 0 4 material allows calorimeter compactness due to its short radiation length and small Moliere radius. It has good radiation hardness properties and its fast scintillator and peak emission frequency is well adapted to the short LHC interbunch crossing time (25 ns). After a successful R&D phase, the ECAL project is now moving to a construction phase. Pablo Sempere Roldan presented a review of the construction of the first large size ECAL barrel unit. This unit was comprising 400 PbWC-4 crystals. Various aspects of the construction sequence were discussed such as uniformization of light collection, gluing the crystals to photodetectors, assembly process, module thermal regulation. The status of the CMS hadron calorimeter was reported by Vasken Hagopian. This calorimeter includes a central barrel and two endcaps, made of brass and scintillators. Two forward calorimeters extend the pseudorapidity coverage up to large pseudorapidity (\T]\ < 5). Energy leakage from the central barrel are measured with scintillators located outside the magnet coil, within the muon system. After several design changes to simplify the calorimeter and reduce the cost, the construction of the calorimeter is about 50% complete. Then, it is expected to have a working CMS hadron calorimeter at the start of the data taking. We had also a report from calorimetry activities in DESY/HERA. Andy Ziegler gave a talk on the new ZEUS tungsten/scintillating fiber spaghetti electromagnetic calorimeter for the luminosity monitoring system after the upgrade of the HERA luminosity. The detector was successfully tested and achieved an energy resolution of 17.0%/-/B+15.7%/E and a position reconstruction resolution of 0.93x0.51 mm 2 . Finally, Franz Steinbuegl presented a new concept of an active element for the large cosmic ray calorimeter ANI located on the Mount Aragatz, Armenia. A new concept for active medium has been translated into a prototype which has undergone a first set of tests. The basic element is a long tube filled with water (10-40 m long and 30x30 cm2 cross section) and read out by two photomultipliers at both ends. The significant fraction (about 50%) of the Cerenkov light produced by the passage of a charged particle through water is absorbed by a dissolved wavelength shifter dye and re-emitted around 420550 nm. A new type of reflector material has been used to optimize light transportation over long distances.

OVERVIEW OF THE CMS ELECTROMAGNETIC CALORIMETER H. F. HEATH H.H.Wills Physics Department, Bristol University, Tyndall Ave, Bristol, BS8 1TL, England E-mail: Helen.Heath @bristol.ac.uk The Compact Muon Solenoid (CMS) is one of two ominpurpose experiments to be constructed at the Large Hadron Collider (LHC) at CERN. CMS incorporates a precision Electromagnetic Calorimeter which will be the largest crystal calorimeter ever constructed. The harsh environment at the LHC places stringent demands on the detector components. Following an extensive development period production of parts for the CMS ECAL is under way. An overview of the current project status is presented including results from recent prototypes and quality control tests on production components.

1

Introduction

The CMS experiment1 for the LHC has made high precision electromagnetic calorimetry a priority and is constructing the largest crystal calorimeter ever built. It will contain 80,000 lead tungstate crystals. The calorimeter will be located inside a superconducting solenoid which will produce a field of 4T. 2

Physics Constraints

If the Higgs has a mass of less than 150GeV/c2 then the most likely discovery channel is the decay to two photons. The natural width of a Higgs of this mass is very small and so the measured width depends only on the detector resolution. The mass resolution is given by the expression: aMIM = (fj£] IEX @
GEIE =

alJE@b®clE

The "stochastic" term a arises from photoelectron statistics and shower fluctuations. The "constant term" b has contributions from non-uniformities and from shower leakage. The "noise" term c is due to electronics noise and pile-up. The design goals2 for the CMS ECAL barrel and endcap are a = 2.7% and 5.7%

384

385

respectively and b <0.55% for both sections of the detector. Expressing the noise as transverse energy the goals are, at low luminosity, c = 155MeV and 205 MeV and at high luminosity 210MeV and 245MeV for the barrel and endcap respectively. Results from beam tests with prototype crystals3,4 have demonstrated that these goals are attainable. 3

Lead Tungstate

To reach its design goals CMS has chosen a homogeneous crystal calorimeter. Lead Tungstate (PbWCty has been chosen as the detection medium. It the advantages of a short radiation length (0.89cm) and small Moliere radius(2.19cm) which means that the ECAL can be relatively compact. A compact ECAL reduces the overall size of the experiment including the solenoidal magnet. Lead Tungstate is intrinsically radiation hard, which is essential for the LHC environment. It is a fast scintillator and the peak emission frequency matches well with the requirements of photodetectors. An extensive research and development program5 has produced crystals of the required quality. Recent developments have succeeded in increasing the size of the boules, from which the crystal are cut, enabling two crystals (and possibly four) to be produced from the same boule. This results in a dramatic increase in the potential production rate. The current status is that the order of the barrel crystals has been placed and the order for the endcap crystals is to follow shortly. All the procedures for monitoring the crystal quality are well defined and have been exercised thoroughly with preproduction crystals. 4

Mechanics

Mechanically the CMS ECAL is divided into three sections, a central barrel section and two endcaps. The crystals are arranged such that intercrystal gaps are offpointing by 3°. To achieve this the barrel has a total of 34 different crystal shapes 17 pairs of mirror image crystals - approximately 20x20x230mm3 in size. The barrel is constructed of 36 supermodules containing 1700 crystals. Each supermodule contains four modules each comprising 400 or 500 crystals. The first fully operational CMS ECAL barrel module has been constructed and a description is given elsewhere in these proceedings6 The Endcap design uses a single crystal shape approximately 30x30x220mm3. The crystals are slightly tapered and twenty-five of them are inserted into a carbon fibre "alveolar" unit, each with the same orientation. The resulting tapered supercrystal is the basic unit from which the Endcap is constructed. The supercrystals are mounted onto D-shaped back plates with angled mounting blocks. All supercrystals in one quadrant have the same orientation so that the supercrystal

386

angle to the horizontal increases with distance from the beam pipe. This maintains the desired off-pointing. A full-sized supercrystal has been tested at in electron beams at CERN3 and in mechanical tests at the Rutherford Appleton Lab a fully loaded supercrystal has been mounted on a mock up of the Dee back plate(Fig 1).

Figure 1. A fully loaded endcap supercrystal suspended from a mock up of the endcap Dee backplate.

5

Photodetectors

Two different photodetector technologies are used in the CMS ECAL to meet the constraints of the environment in the barrel and endcaps. The barrel uses solid state Avalanche Photodiodes(APDs). These devices have the advantages that they can operate in a transverse magnetic field, have a high quantum efficiency(~70%) and the thickness of the layer in which charge from shower leakage can be amplified (the nuclear counter effect) is small (6-8|xm). Their disadvantages are that they are small in area and the bulk leakage currents increase after radiation. Development work4 has reduced the excess noise factor (the fluctuations in the amplification process) and improved the stability of the gain with respect to voltage and temperature variations. To address the problem of the small active area a pair of APDs, mounted in a capsule, is used on each crystal. By October 2001 120,000 APDs had been ordered and over 11,000 delivered. The quality control and testing procedures are well defined. Assembly of capsules has been running since Autumn 2000. The ECAL endcap is instrumented with Vacuum Phototriodes (VPTs) which are essentially single stage photomultiplier tubes. The use of APDs in this region is ruled out because the increase in electronic noise which would be induced by radiation damage is unacceptable. VPTs will operate in magnetic field provided the

387

tube axis is at less than 45° to the magnetic field axis. Development work for CMS7 has addressed the issues of improving the performance in magnetic fields, increasing the effective photocathode area and producing radiation hard glasses for the face plates. VPTs have a rather low Quantum efficiency (-18%) and gain (~8) and so large area tubes are essential in order to collect as much light as possible. The current status is that "1-inch" VPTs from a preproduction batch of 500 have been tested in the UK in magnetic fields of up to 4.7 T - fully exercising the quality control and acceptance procedures. The first order for 7000 production tubes has just been placed.

6

Monitoring

During the lifetime of CMS there will be an inevitable loss in response from the crystals. The basic scintillation mechanism is unaffected by radiation damage but the formation of colour centres leads to a loss in optical transparency. In order to compensate for this change in response it is necessary to track the changes throughout the lifetime of the experiment. For this purpose CMS is using a laser monitoring system in which light of wavelength 440nm is injected into each crystal. Tests have shown that the response to laser light is linearly related to the response to charged particles8. Energy = ISO. GtV. angle =
600 500

CombiiiDd ajtiil pkji prnbowcnapuaae

400 KeqMnn vitboit

300 200 100

16016Z5165167.5170172.3175177J180182,5.-,- 185 Eiurgy (GeV)

Figure 2 The energy resolution seen in a prototype endcap supercrystal for 180GeV electrons with and without the preshower correction.

388

7

The Preshower

The reducible contribution to the background to signals with a photon in the final state (including the Higgs decay) is from a jet which fragments into a leading TC°. In the barrel region a 7t° with a E T of 60GeV will fragment to two photons with a separation of 0.8cm. Here it is possible to reduce the n° background using the information from the crystals alone. In the endcap the separation of the photons from the same n° would be a few millimetres and the crystals are larger. In order to reduce this background a detector is needed which has a finer granularity than the crystals. CMS has chosen a preshower with lead as the absorber and two orthogonal layers of silicon strip detectors with 1.9mm strip pitch. The 7t° rejection which can be achieved in the endcaps varies from 60-70% at ET = 60GeV, depending on r\. The main purpose of the preshower is to improve n° /y separation but it also improves isolation and reduces the longitudinal leakage from the ECAL. The presence of the lead absorber leads to a degradation in the resolution of the ECAL but this can be largely compensated for by applying a correction to the measured energy using the information from the preshower (Fig.2). 8

Summary

The CMS ECAL project is now moving from a successful research and development phase into the construction phase. Orders have been placed for many of the critical components. The quality control procedures for acceptance of these components are in place and have been exercised with preproduction components. References 1. CMS Collaboration (G.L.Bayatian et al.), CERN/LHCC 94-38 2. CMS Collaboration (G.L.Bayatian et al.), CERN/LHCC 97-33 3. M.Apollonio et.al. Test Results from a Prototype Lead Tungstate Crystal Calorimeter with Vacuum Phototriode Readout for the CMS Experiment. Accepted for Publication by Nucl. Instr. and Meth A 4. E. Auffray et.al. Nucl. Instr. and Meth A 412 (1998) 223 5. I. Dafieni Lead Tungstate Crystals for the CMS Electromagnetic Calorimeter at the LHC. 6. P. Sempere Roldan. Construction of first CMS-ECAL fully operational module (400 lead tungstate crystals) 7. K.W.Bell at.al. Nucl. Instr. and Meth A 469 (2001) 29, N.A.Bajanov et. al. Nucl. Instr. and Meth A 442 (2000) 146 8. G. Davies et. al. A Study of the Monitoring of Radiation Damage to CMS ECAL Crystals, Performed at X5-GIF. CMS NOTE-2000/020

T H E R E A D O U T OF T H E ATLAS LIQUID A R G O N CALORIMETER STEFAN SIMION Columbia University, Nevis Labs, 136 S Broadway, Irvington NY 10533, USA The ATLAS Collaboration The ATLAS liquid argon front-end system will instrument 190000 calorimeter channels, providing low-noise, highly-accurate digital signals to the read-out drivers (RODs) and to the second-level trigger, as well as fast analog outputs to the firstlevel (LVL1) trigger. The bipolar-shaped ionization signals are sampled at the 40 MHz LHC frequency and temporarily stored into analog memory, during the fixed 2.5 /is LVL1 latency. For each LVL1 trigger, a configurable number of samples (typically 5) are digitized and sent to the ROD via optical links. Most of the digital components have been redesigned to comply with radiation tolerance requirements. The logic has been implemented in DMILL and/or 0.25 nm ASICs, and feature error-correcting circuitry for protection against single-event upsets. Radiation tolerance studies of the ASICs have been performed using a high-intensity proton beam. Radiation results are presented.

1

Introduction

The architecture of the ATLAS Liquid Argon (LAr) calorimeter readout has been described extensively elsewhere1. Ionization signals induced on Kapton electrodes are routed to front-end electronics boards (FEBs) mounted in close proximity (on top) of the LAr cyrostats. Each FEB instruments 128 calorimeter channels, and feeds one optical link with digitized information for the second-level trigger. Only a brief overview of the analog chain is given here, followed by a more detailed description of the digital components, with emphasis on radiation tolerance requirements. The front-end system is designed to withstand a total ionizing dose of 50 krad; a NIEL equivalent of 4 10 13 n/cm2; and a fluence of hadrons with kinetic energy greater than 20 MeV, capable to produce single-event effects (SEE), of 10 13 h/cm2. On the FEB, the ionization signals are amplified using fast current preamplifiers, which are four-channel plug-in subassemblies, realized with discrete components in bipolar technology. There are a few flavours of preamplifiers which differ only by their transimpedance and input impedance, so as to match the dynamic range of the calorimeter section being instrumented. The preamplifiers have been shown to tolerate a NIEL fluence of 10 14 n/cm2 with no loss of performance. The amplified signals are fed into a three-gain bipolar shaper 2 whose

389

390

time response has been chosen to minimize the total noise (electronics and pile-up noise) at the design luminosity of the LHC. Three gains are required to adapt the calorimeter (and preamplifier) dynamic range of about 17 bits, to that of the analog memory (SCA, see below) and of the ADC. The shaper ASIC (in bipolar AMS technology) instruments four calorimeter channels. In addition to the 12 outputs to the SCA, each shaper ASIC features an additional trigger sum output, which is the first stage of the analog chain providing the calorimeter input to the LVL1 trigger. The high-precision shaper outputs are sampled at the LHC frequency as they are stored into a 144-cell deep analog memory, or Switched Capacitor Array. The SCA can be written to, and read from, simultaneously, provided that different cells are accessed. This allows reading out the calorimeter at up to 100 kHz LlA rate, with essentially no dead time. The analog memory provides sufficient storage to accommodate the 2.5 fis LI latency, plus the time needed to digitize other LI events queued for read out. Reading out an event is a slow operation, as the analog signals from 8 channels are multiplexed (by the SCA circuitry) to drive the inputs of the ADC operating at 5 MHz. The SCA memory ASIC has been migrated 3 from the 1.8 /xm HP technology, to the DMILL radiation-hard technology. The DMILL SCA features a noise level of 290 fj,V per storage cell, a pedestal (or baseline) dispersion of 250 /iV per channel (i.e. across an entire row of 144 cells), a sampling jitter smaller than 70 ps, and a dynamic range of 13 bits. All of the chips will be tested individually on a dedicated test setup, before they are being shipped for assembly. One key parameter which will eventually determine the production yield, is the leakage current of the 1 pF storage cells. Measurements performed on first batches have shown a yield of 65% if only chips with all cells below 5 pA leakage are accepted. The DMILL SCA memory has been tested to 300 krad and 2 10 13 n/cm2 with no degradation of performance. In addition, a Module 0 FEB" was equipped with 16 DMILL SCA chips (instead of the rad-soft HP SCA) and operated succesfully. 2 2.1

The SCA Controller Principle of operation

The SCA Controller provides the SCA addresses for both the analog write (sampling) and analog read out operations. A simplified diagram of the SCA "radiation-soft prototype realized in 1998, used to read out the calorimeter modules during beam tests at CERN

391

Controller is shown in Figure 1. The SCA Controller continuously drives, at the 40 MHz LHC frequency, the 8-bit wide SCA write address bus. The write addresses are being read from a free cells list (FIFO), presented on the SCA write bus, and written to a separate latency FIFO whose depth can be programmed to match the L1A latency. At the read end of the latency FIFO, the addresses are either put back into the free FIFO (in case no LlA trigger has been received) or on the contrary queued for digitization by going into a read FIFO. Addresses are taken out of the read FIFO and transmitted serially to the SCA, thus specifying which analog storage cells will Figure 1. Simplified diagram of the drive the ADC input. After digitization 6 SCA Controller address flow. of the specified sample, the corresponding address ends up again in the free cells list. The actual behaviour is more complicated than has been described above. For instance, to keep digital noise as low as possible, the write addresses are gray-encoded so that only one address bit changes every 25 ns, provided the addresses are consecutive. Therefore, the free cells list is actually implemented as two separate FIFOs, followed by a sequencing logic which maintains the write address ordering, as much as possible. 2.2

SCA Controller radiation-tolerant

ASICs

The SCA Controller was initially implemented in a Xilinx FPGA for the (rad-soft) Module 0 FEB. To comply with radiation-hardness requirements, two ASIC options have been developed and tested. One is a 80 mm2 DMILL chip, with all storage (FIFOs) implemented in registers; due to the large chip area, no room has been left for error detection and/or correction logic, to avoid SEU-related corruption of the address generation. Nevertheless, this DMILL SCA Controller has been successfully operated at up to 50 MHz, and the preliminary yield was 70% (a total of 40 devices were tested). In parallel, a SCA Controller ASIC has been developed4 using a 0.25 fim (deep sub-micron or DSM) technology. This 16 mm2 chip uses SRAM cells developed at CERN 5 for storage of SCA addresses, and features error detecfc

for each of the 8 channels being multiplexed to one ADC

392 tion and correction circuitry to cope with SEUs. The circuit is fully functional up to 66 MHz.

2.3

SCA Controller radiation setup and test results

Four DSM SCA Controller devices were tested during irradiation at the Harvard Cyclotron Facility0. The cyclotron delivers 158 MeV protons with high intensity, up to 4.4 10 10 p/cm2/s. The devices were placed in close proximity to the main magnet, inside the cyclotron vault. The test setup includes an Altera FPGA which generates the L1A trigger to the SCA Controller, and monitors the write addresses and the serial read commands generated in response to the trigger. This FPGA includes an exact model of the SCA Controller behaviour, which serves as a reference for comparing the SCA Controller outputs. Finally, the FPGA connects to the read and write serial configuration ports of the SCA Controller, used for downloading parameters such as the number of samples to be digitized and the LlA latency. The FPGA is placed in the beam area, close to the chip being irradiated, and is protected against scattered radiation using lead bricks. This test jig is connected using twisted-pair differential cables to a dataaquisition computer located outside the cyclotron restricted area. Whenever a SEU is detected by the Altera, and event dump (a snapshot of the SCA Controller outputs) is generated and transferred to the computer. The computer is also used to continuously monitor the supply current of the device being tested (and thus to record the possible occurence of latch-up) as well as the instantaneous beam intensity. The four samples have been irradiated with up to 14.5 10 13 p/cm2 (corresponding to 9.9 MRad of TID at 158 MeV). No latch-up and no significant damage has been observed. Figure 2 shows the supply current of the SCA Controller, during operation at 40 MHz and during irradiation, as a function of the integrated proton fluence. Most of the SEEs observed are SRAM single-bit errors and do not affect the circuit behaviour due to the self-correcting logic. However, whenever such an error is corrected by the SCA Controller, a bit is set in the status output, for monitoring purposes. The SRAM-related SEU cross-section has been measured: 1.7 10 - 1 1 cm 2 at 158 MeV. Figure 3 shows the SEU crosssection per SRAM cell, as a function of the proton kinetic energy.

c

44 Oxford Street, Cambridge, Massachusetts 02138

393 I

-

•

1

X

.

4

' 10*13

2e»13

3e+13

4e+13

<s* 5e»13

6e*13

: 7©»13

lluenco |pfcm 2 )

Figure 2. SCA Controller supply current during irradiation, as a function of the integrated proton fluence.

. _

(

._

SamptoA

6

„S © Sample D

i

1

|

Sample C

i

UJ

-

™ 2 StmpltE

i

Figure 3. SRAM cell SEU cross-section as determined during the SCA Controller proton test. T h e difference in cross-section between samples A and C should be considered as a measure of the systematic error.

3 3.1

The Gain Selector Principle of Operation

This circuit is responsible for deciding, event by event and for each channel, which of the three gains stored separately in the SCA analog memory are to be digitized before being sent out (via optical links) to the read-out driver. The Gain Selector controls which gains (or rows) of the SCA are allowed to

394 drive the ADC input, sequentially for each of the 8 channels being multiplexed to one ADC. For each LI trigger, every channel is first digitized in a pre-determined gain, which can be programmed in the Gain Selector. The ADC data value is then compared to two 12-bit thresholds, which are also programmed for each channel and reside in the Gain Selector internal storage. When the ADC value is larger than the upper threshold, the lower of the remaining two gains will be chosen for digitization; and when the ADC value is smaller than the lower threshold, the higher of the remaining two gains will be chosen. Otherwise, the gain is left unchanged. When the automatic gain selection is enabled, the ADC performs 6 conversions for every channel (assuming that 5 samples are read out). The Gain Selector can also be programmed for fixed-gain operation6*. 3.2

The Gain Selector ASICs

The actual circuit 6 is a dual Gain Selector which serves 16 channels (two ADCs and four SCAs). For the Module 0 FEB, this circuit was implemented in an Altera 6k series FPGA. For operation in ATLAS, two rad-tolerant ASICs were developed, one in DMILL and the second in a 0.25 fim technology. Both ASICs feature error detection and correction circuitry to protect the internally-stored gain selection parameters (thresholds etc.) from being corrupted by SEUs. The DMILL and DSM ASICs are based upon the same Verilog code. Proton kinetic Energy (MeV) ~50 100 158

Fluence 10 13 p/cm2 2A 4.0 20.8

Single bit errors Uncorrected errors Count a (cm2) Count a (cm2) 0 n~M. 0 n.a. 14 3.5 10~ 13 4 1.0 10~ 13 212 10.2 10~ 13 38 1.8 10~ 13

Table 1. Summary of DMILL Gain Selector SEE measurements

Five DMILL Gain Selectors have been tested 6 during irradiation with protons of 50, 100, and 158 MeV and fluences up to 20 10 13 p/cm2. The SEU measurements are summarized in Table 1. When extrapolated to the entire calorimeter, one single-bit error is expected to occur every 30 minutes, while unrecoverable errors (which require explicit software reconfiguration of the circuit) are expected every 168 minutes. Four DSM Gain Selectors have been tested under irradiation with up to 22 10 13 p/cm2 (158 MeV protons). No degradation and no latch-up has been d e

ln fixed-gain mode, any one, two, or all three gains can be read out for one event T h e test setup is similar to that used for the SCA Controller.

395 observed. The SEU cross-section for this circuit is 10 magnitude lower than for the DMILL version. 4

13

cm2, one order of

Conclusion

The readout system of the ATLAS electromagnetic calorimeter has been redesigned to comply with radiation requirements. In particular, six radiationtolerant digital ASICs (two of which are described in this report) have been developed, have undergone functional and radiation tests, and are now being operated on a new prototype FEB. The DSM technology has been retained for the SCA Controller, the Gain Selector, and for the clock distribution circuitry. The FEB serial configuration protocol (SPAC) and the optical output link both use DMILL technology. The only commercial ICs still used on the FEB are the ADCs; the differential drivers for the SCA address bus; and the G-link serializer for the optical output link. Acknowledgements We thank E.W. Cascio and the cyclotron personnel for their excellent operation of the Harvard Cyclotron Facility used for the irradiation measurements. References 1. CERN/LHCC/96-11, ATLAS Liquid Argon Calorimeter Technical Design Report. 2. C. de la Taille at al, The LAr Tri-Gain Shaper, ATLAS note LARG-92 and ATL-AL-ES-0006 (EDMS). 3. P. Borgeaud et al, The HAM AC rad-hard Switched Capacitor Array, a high dynamic range analog memory dedicated to ATLAS calorimeters, ATL-AL-EN-0021 (EDMS). 4. S. Botttcher, J. Parsons, W. Sippach, D. Gingrich, The SCA Controller for the ATLAS LAr Calorimeter1. 5. F. Faccio, K. Kloukinas, G. Magazzu, A. Marchioro, SEU effects in registers and in a Dual-Port Static RAM designed in a 0.25 p,m CMOS technology for applications in the LHC, CERN-OPEN-2000-123. 6. S. Bottcher, J. Parsons, W. Sippach, The Gain Selector ASIC for the ATLAS LAr Calorimeter.

f available at http://www.nevis.columbia.edu/~atlas/electronics/asics

A NEW W/SCINTILLATOR ELECTROMAGNETIC C A L O R I M E T E R FOR ZEUS A. ZIEGLER, U. HOLM, N. KRUMNACK, K. WICK, AR. ZIEGLER Institut fiir Experimentalphysik, Universitat Hamburg, Lumper Chaussee 149, 22761 Hamburg, Germany J. CRITTENDEN, R. WICHMANN Deutsches Elektronen-Synchrotron (DESY), Notkestr. 85, 22607 Hamburg, Germany A new tungsten/scintillating fiber spaghetti calorimeter was constructed for the ZEUS detector at the electron-proton collider HERA. The calorimeter will be located 5.3 m from the interaction point, inside one of the HERA quadrupole magnets. It will detect electrons coming directly from the interaction point under very small angles. Test beam measurements with the fully assembled detector show an energy resolution of 17.0%,/ \/~E ($1.5.7%/ E and a position reconstruction resolution of 0.93 x 0.51 mm 2 .

1

Introduction

The upgrade of the HERA luminosity requires a modification of the ZEUS luminosity monitoring system. At ZEUS the luminosity is determined using the bremsstrahlung process ep —> epj. Until now the photon flux of the bremsstrahlung process has been measured about 100 m away from the interaction point. The new HERA conditions produce a much higher synchrotron background and more than one bremsstrahlung photon per bunch-crossing is expected. These circumstances require modifications of the existing detector. ZEUS decided to retain and rebuild the photon calorimeter of the luminosity monitoring system and to add a second complementary system which independently measures the luminosity by counting bremsstrahlung photon conversions in the beam pipe exit window. These conversions are counted by electron-positron coincidences in a pair of two small electromagnetic calorimeters 1 . The measurement of the bremsstrahlung photon in coincidence with the bremsstrahlung electron" allows a cross check of the absolute energy calibration of the photon detectors and a direct measurement of their acceptance. "In the following electron will be used meaning either electron or positron, according to the HERA running mode.

396

397

lightguide PMTs hv system *H

more channels

VME link

Figure 1. A sketch of the test beam setup showing the trigger system, the detector and the readout. The triggers are a scintillator plate covering the whole beam, S2, a veto trigger with a 3 mm hole, SI, and two crossed finger counters behind the hole that covered a plane of 10 x 10 m m 2 , T l and T2.

For the detection of the bremsstrahlung electron an exit window was designed for the electron beam pipe 5.3 m from the interaction point. This window exists inside one of the HERA quadrupole magnets leaving a space of 100 x 100 x 30 mm 3 (WxLxH) where a small electron detector will be placed which performance is reported. The new detector was designed to work under extreme conditions. The W/scintillating fiber spaghetti design was chosen due to the described strong space constraints. In addition a dose rate of 10 to 20 kGy/year is expected as well as a temperature of up to 40° C, where ageing effects of the fibers become important. At the foreseen position the detector measures electrons with an energy from 4 to 9 GeV with an angle of incidence between 3° and 7° with respect to the beam axis. In addition kinematic peak events with an electron energy up to the beam energy of 27.5 GeV can be detected. 2

Construction of the calorimeter

The calorimeter is made out of 85 1 mm thick Densimet D 180 K plates (18.0 ± 0 . 1 g/cm 3 and 95% W) of dimensions 24 x 100 mm 2 . On both faces semicylindrical grooves (0.56 mm diameter) are machined. The stacked plates are held together by three 1.5 cm wide steel straps which are screwed to a support bar on the side of the detector. This support bar has a thickness of 1 cm and is placed opposite to the beam pipe. At the back side of the detector a 2 mm thick PVC plate with holes is glued to a strap. The fibers are put through the PVC plate and the detector and are glued to the PVC mask. If it is necessary to replace the fibers they can easily be extracted and

398

the tungsten can be reused. The detector contains 1890 SCSF-38M fibers with a diameter of 0.5 mm made by Kuraray. The fibers have a length of 1.5 m and are machined with a diamond cutter on both sides. A detailed investigation of these fibers can be found elsewhere2. The front face of the calorimeter is painted white to improve the light collection efficiency and the uniformity. Always 27 fibers are grouped to one readout cell. In total there are 70 cells with an effective size of 6 x 4.68 mm 2 , resulting in a grid of 14 horizontal times 5 vertical cells. Outside the magnet the 70 cells are readout with PMT 5600 (Select 3) from Hamamatsu, placed in a /u-metal box fixed below the magnet. An LED system distributes amber light through clear fibers to all 70 PMTs for PMT monitoring. 3

Test beam results

The calorimeter was tested in the DESY test beam with electrons of energies up to 6 GeV. The setup for the test beam measurement consisted of a trigger system, the detector and the readout (Figure 1). The triggers were a scintillator plate at the beam exit window that covered the whole beam, S2, a veto trigger with a 3 mm hole, SI, and two crossed finger counters behind the hole, T l and T2, that covered a plane of 10 x 10 mm 2 . The detector was placed in such a way that the angle between the electron beam and the fiber direction was 6°. The calibration was performed with 5 GeV electrons triggered with S2. The aim was to get a calibration constant of about 8 pC/GeV/cell so that a 27.5 GeV electron can be detected with only one cell. The trigger selected events which were equally distributed over the whole detector. The events that centrally hit the cells can be obtained in two steps. In the first step of the calibration of the cells the contamination of low energy depositions was avoided. Therefore a Gaussian was fitted from the mean value of the distribution up to the higher energetic tail. The mean of the fit for each cell determined the factor that had to be applied to get the calibration constant. In the second step only hits for a cell were accepted in which this cell contained at least 60% of the sum over a 3 x 3 cluster around this cell. This procedure was iterated twice and the energy distribution became Gaussian. The energy reconstruction is linear within 1%. The energy resolution of the calorimeter was determined with electron beams of 1 to 6 GeV and the same trigger configuration as for the calibration. A minimum signal of 10 pC in the detector was required to suppress background events except for the 1 and 2 GeV runs. The calorimeter response

399 1

noise = 0.25 o a o

3x3 cells

Stat

0.2-

-5x5 cells

noise stat

0.15

\

=0.124210.001917 =0.1839 ±0.0003547

H |> o

3x3 cells

5x5 cells

|

=0.1574 ±0.002552 =0.1696 ±0.0006312

0.1

0.05

5 6 beam energy / GeV Figure 2. The observed resolution
signal was calculated by summing up the energy signals of 3 x 3 and 5 x 5 cells, respectively, around the cell with the highest deposited energy. A volume of 3 x 3 cells offers 36 central cells that can be used, for the volume of 5 x 5 cells the reconstruction only works for the 10 central cells but permits a better energy reconstruction. The energy resolutions can be parameterized by (Figure 2) a/E = 18.4%/y/E © 12 A%/E a/E = 17.0%/y/E © 15.7%/E

for 3 x 3 cells for 5 x 5 cells

Monte Carlo simulations of the calorimeter performance with the EGS4 code 3 were done 4 . The simulation takes care of the final geometry of the calorimeter, the leakage losses at all 6 sides and the cell structure. Noise and photostatistics are not included. The Monte Carlo predicts an energy resolution of a/E = 15.8%/VB® 1.2% a/E= 14.1%/>/S©1.3%

for 3 x 3 cells for 5 x 5 cells

400

The high statistical term in the fit to the test beam data is due to the length of the gate (188 ns) in the test beam. The dependence of the gate length on the energy resolution was studied. The energy resolution will be better with the gate length used in the ZEUS readout (69 ns) and much more similar to the Monte Carlo prediction. The reconstruction of the position of the incidence electron was measured with 5 GeV electrons, requiring that no energy was deposited in the counter Si with the 3 mm hole but in T l , T2 and S2. The detector was moved in 2.5 mm steps horizontally and vertically, respectively. The deviation from uniformity of the position reconstruction was 2.5% for 5 x 5 cells and 3.0% for 3 x 3 cells. The reconstructed position was linear within 0.2 mm but depends on the electron energy and the angle of incidence of the electron. The obtained resolution is 0.93 mm in horizontal and 0.51 mm in vertical direction. 4

Summary

A new small electromagnetic calorimeter was assembled and tested for ZEUS. The W/scintillating fiber spaghetti design had to be chosen due to the limited space. The detector will be located close to the beam pipe behind an exit window 5.3 m from the interaction point inside a HERA magnet. The detector will measure bremsstrahlung electrons in the energy range from 4 to 9 GeV and scattered electrons up to the beam energy. It is a vital component for the ZEUS luminosity monitoring system and will measure the electron in coincidence with the photon detectors to determine the acceptance of these calorimeters. The detector was tested at the DESY test beam. The position of the electron is linear within 0.2 mm with a resolution of 0.93 x 0.51 mm 2 for a 6 x 4.68 mm 2 cell. The energy reconstruction is linear within 1% and the achieved energy resolution is 17.0%/y/E © 15.7%/E. The energy resolution and the agreement with the Monte Carlo will improve when the shorter gate length of the ZEUS readout is used. References 1. S. Paganis, A Luminosity Spectrometer for ZEUS experiment at HERA, Frascati Physics Series Vol. XXI, 523 (2000). 2. Ar. Ziegler, In situ measurement of radiation damage in scintillating fibers, these Proc. 3. Walter R. Nelson et al., The EGS4 code system, SLAC-Report 265 (1985). 4. N. Krumnack, diploma thesis, University of Hamburg (2001)

P E R F O R M A N C E OF T H E ATLAS LIQUID A R G O N ELECTROMAGNETIC CALORIMETER MODULES UNDER TEST BEAM

F. Djama * CPPM,

CNRS/IN2PS - Univ. Mediterranee, Marseille E-mail: [email protected]

- France

The ATLAS collaboration will start to collect d a t a at the LHC proton—proton collider in 2006. It has chosen a liquid argon-lead sampling electromagnetic calorimeter using the accordion geometry. In 1999 and 2000, barrel and end-cap prototype modules have been tested under electron beams at CERN. Detailled results from these tests are shown.

1

Introduction

The ATLAS experiment will record proton-proton collisions data at the future LHC collider, starting from summer 2006, together with CMS and LHCb experiments. Prototype modules (barrel and end-cap) of the ATLAS electromagnetic calorimeter have been put under test beam in 1999 and 2000. This talk was devoted to the results and the outcome of these tests. 2

Physics Requirements

The calorimeter performances needed to achieve the ATLAS physics goals have been set-up by simulating a few benchmark channels 1 . The calorimeter must be radiation tolerant, and have fast electronics, to insure safe operation in the LHC environment (radiation and pile-up). Hermeticity, large acceptance, identification capabilities are required by a large set of physics analysis. Search for the Higgs boson decay into two photons (H —* 77) sets the main requirements on energy resolution, energy scale and polar angle resolution. A 10 %/st/E(GeV) sampling term and a constant term below 1 % are necessary to achieve a 77 mass resolution of 1 %. To get rid of the longitudinal spread of LHC vertices, an intrinsic polar angle resolution for the electromagnetic calorimeter of 50 mrad/\/E{GeV) is required. *On behalf of the ATLAS liquid argon group.

401

402

3 3.1

D e t e c t o r Description Mechanics

T h e calorimeter consists of lead plate absorbers alternated with polyimidecopper electrodes. T h e liquid argon gap between absorbers and electrodes is denned by honeycomb nets. T h e barrel and the two end-caps will be housed in three separate cryostats. Pseudorapidity coverage extends u p to 3.2, with a separation between the barrel and the end-caps at 1.4. Absorbers and electrodes are bent to accordion shape. T h e bends are parallel to the beam axis in the barrel, and are radial in the end-caps. T h e barrel is divided into 2 x 16 modules. Each module has 64 doublegaps (figure 1). The end-cap consists of 2 concentric wheels, separated at 7) = 2.5 and is divided in azimuth into 8 wedge-shaped modules. T h e end-cap modules (figure 2) have 96 double-gaps in the outer wheel and 32 in the inner one. T h e whole detector has a n external diameter of 4000 m m , a n d a n active depth varying from 22 to 37 radiation lengths. Up to 7) = 1.8, a presampler is put in front of the calorimeter, to recover the energy lost in dead material. A detained description can be found elsewhere 2 .

Figure 2. View of an end-cap module. Figure 1. View of a barrel module. first and last absorbers are drawn.

3.2

Only

Read-Out

Large three-layer copper-polyimide electrodes are used for read-out. High voltage is applied on the 2 external copper layers (absorbers being grounded), while signal is read-out on the internal layer. Granularity along r) and longitudinal segmentation are achieved by etching the three layers of the electrodes. There are 3 longitudinal samplings in the

403 Table 1. Detector Granularity: ATJ X A(/>.

rj range 0 - 1.350 1 . 3 5 0 - 1.400 1 . 4 0 0 - 1.475 1 . 4 7 5 - 1.52

presampler 0.025 X 0.1 0.025 X 0.1 0.025 X 0.1 0,025 X 0.1

1.375 - 1.425 1 . 4 2 5 - 1.5 1 . 5 - 1.8 1 . 8 - 2.0 2.0 - 2.4 2 . 4 - 2.5 2.5 - 3.2

0.025 X 0.1

-

Barrel front 0.003 X 0.1 0.003 X 0.1 0.025 X 0.1

middle 0.025 X 0.025 0.025 X 0.025 0.075 X 0.025

End-Cap 0.050 0.025 0.003 0.004 0.006 0.025

X 0.1 X 0.1 X 0.1 X 0.1 X 0.1 X 0.1

-

back 0.05 x 0.025

-

0.050 0.025 0.025 0.025 0.025 0.025 0.1

X 0.025 X 0.025 X 0.025 X 0.025 X 0.025 X 0.025 X 0.1

0.05 0.05 0.05 0.05 0.1

X 0.025 X 0.025 X 0.025 X 0.025 X 0.1

whole acceptance, except in the end-cap inner wheel, where there are only 2 samplings. In azimuth, neighbouring electrodes are summed to define physical cells by the mean of dedicated summing boards. Details of granularity are shown in Table 1. Mother boards, which house signal connectors for read-out cables are plugged on the top of summing boards. 3.3

Electronics

Signals are routed outside cryostats via feedthroughs. Then they cross the electronic chain: OT-preamplifier, tri-gain shaper, analog pipe-line, ADC and buffering space. These components are parts of the Front End Board. Each FEB reads 128 channels. The FEBs are plugged on the Front End Crate, which is fixed on the warm side of the feedthrough. The FEC houses also the calibration boards. Signals are sampled each 25 ns and the samples are stored in the 144 capacitor deep pipeline 2 ' 3 . 4

Prototype Modules

Barrel and end-cap prototype modules were built during spring 1999 at LAPPAnnecy and CPPM-Marseille. They were fully equipped with absorber plates and half equipped with read-out electrodes. Presampler sectors were mounted in front of both modules. Modules were put under test beam during 1999 and 2000. Electron beams,

404 with a muon component were used (H6 and H8 CERN facilities). Beam position was monitored by a set of 4 b e a m chambers. Pions and muons events were tagged by dedicated scintillators. 5 5.1

Data Analysis Optimal

Filtering

Five samples are recorded for each physical cell. T h e optimal filtering method allows the computation of the signal amplitude (which is proportional to energy) as a linear combination of the recorded samples: E = £ ] a j S j 4 - T h e dj coefficients are function of the signal shape, its derivative, and the samples noise correlation matrix. T h e shape of the response to ionisation signal has been built by using the asynchronous nature of the test b e a m . Several methods are under study to derive signal shape from calibration shape. One of these have been used for the barrel module analysis 5 , while an ad-hoc sigmoid function 6 has been used for the end-cap 7 . 5.2

Clustering

and Geometrical

Effects

Cluster size has been optimized to minimize electronic noise, while keeping leakage at an acceptable level. In the middle sampling, it is 3x3 or 5x5 cells, depending on pseudorapidity. T h e back sampling is not used for energies below 50 GeV. Cluster energy has to be corrected for geometrical effects. Lateral leakage in rj is corrected by a parabola, while
Results and Performances Cross-Talk

P a t t e r n s of cells were pulsed with the calibration input, and response of neighbouring cells were studied, b o t h in amplitude (cross-talk value) and in shape (origin of the cross-talk) 8 ' 9 . Obtained values of cross-talk in the barrel module in the 1999 test period are shown in Table 2. All cross-talk values are within expectation, except the middle-middle, middle-back and back-back cross-talk, which are 2-3 times higher. These are of inductive origin, and their large values have been traced back to insufficient grounding on the mother boards. New mother boards have

405 been designed and produced. Beam test data in 2000 for the barrel and lab measurements for the end-cap 9 showed that the problem has been solved. Table 2. Barrel Gross-Talk Values in %. (Lines are pulsed a n d we look at columns). front middle back

6.2

front 4.3 0.10 0.02

middle 0.25 0.90 0.50

back 0.08 0.70 0.80

Energy Resolution

After cluster building and geometrical corrections, presampler weights should be determined. They depend on the amount of dead material in front of the calorimeter. In ATLAS, the dead material (trackers, cryostats walls, services,...) will be typically in the range of 2-3 radiation lengths, with a maximum at the barrel-endcap transition region (5-8 XQ). Lead plates of different thicknesses have been put in front of cryostats as dead material. Figure 3 shows the end-cap presampler performance 11 .

Figure 3. Presampler weight determination (a) and improvement of the resolution with presampler (b).

Cluster energy is computed as the sum of energy in a l the samplings. Energy distribution is built and fitted with a gaussian. Electronic noise is

406

evaluated with random events and is quadratically subtracted. Sampling and local constant terms a and b can then be determined by fitting row. A 0.6% uniformity is obtained. Similar results were obtained in dimensions, both in barrel and end-cap 12 . This is an encouraging result, althought we can not claim that this is the final detector uniformity, nor the global constant term, since this concerns limited area of the detector. Dead zones, narrow beam and limited time prevent us from the complete scan of the modules. The global constant term will be determined from the ATLAS modules. Energy resolution \

Ebeam-!f2S0,Gey Ebeam = 20 GeV

71=0.3375

C ° full resolution T ° noise contribution . \ * subtracted resolution 9.

:

— dE/E = a/VE®b a = 9.24 ± 0.10%

:

b = 0.23 ± 0.04% '• •

8

=1.04

:

.t^.>^^^ty^,

^ ^ \

• •

D

°

°

D

.•?.."

D.

*>=10

:

5

10

15

20

25

30

35

40

45

Middle cell T\ number

Figure 4. Energy resolution in the barrel module.

6.3

Figure 5. Uniformity along 77 in the barrel module.

Polar Angle Resolution

Pseudorapidity energy barycentres are computed in the front and the middle samplings, using respectively a 3 strips and a 3x3 cells cluster. If we combine these 2 measurements, and knowing the longitudinal barycentres of the electromagnetic showers in both samplings, the polar angle of the shower can be determined using only the calorimeter information. An example in the end-cap module is shown in figure 6. The initial requirement of 50 mra is satisfied.

407 Polar angle resolution at r\ = 1.9

= ( 4 8 . ± 1.) mrad GeV"'" = ( 8 3 . ± 5.) mrad GeV"' = ( 0 . ± 1.) mrad

0.01

0.008

0.006

0.004

0.002

0

20

40

60

BO

100

120

140

160

180

200

Beam energy (GeV)

Figure 6. Polar angle resolution in the end-cap module.

7

Conclusions and Schedule

T h e major physics requirements on energy resolution were fulfilled by the two ATLAS liquid argon electromagnetic prototype modules. T h e observed problems (dead zones, inductive cross-talk) were all traced back to hardware a n d solved. Solutions are being applied t o the ATLAS modules. T h e ATLAS modules stacking began in summer 2001. In November 2001, about 25 % of the modules were stacked and cabled. T h e module production will end by April 2003, a n d the detector commissioning will start in October 2003, J a n u a r y 2004 and September 2004, respectively for the first end-cap, the barrel, and the second end-cap. T h e LHC pilot run is scheduled in summer 2006. A cknowledgment s I wish to thank S. Tisserant, Ch. Benchouk and L. Serin for their help. Barrel performance plots were kindly provided by L. Di Ciaccio and J. Vossebeld. References 1. ATLAS Calorimeter Performance, C E R N / L H C C / 9 6 - 4 0 , December 1996.

408

2. ATLAS Liquid Argon Calorimeter Technical Design Report, CERN/LHCC/96-41, December 1996. 3. See contribution from S. Simion in the same session. 4. W. E. Cleland and E. G. Stern, Nucl. Instrum. Methods A 338, 467 (1994). 5. L. Neukermans, P. Perrodo and R. Zitoun, ATLAS Note ATL-LARG2001-008. 6. B. Mansoulie and J. Schwindling, Using Multi Layer Perceptions in PAW, http://schwind.home.cern.ch/schwind/MLPfit.html 7. P. Barrillon, F. Djama, L. Hinz and P. Pralavorio, ATLAS Note ATLCOM-LARG-2001-018. 8. F. Hubaut, B. Laforge, D. Lacour and F. Orsini, ATLAS Note ATLLARG-2000-007. 9. P. Pralavorio and D. Sauvage, ATLAS Note ATL-LARG-2001-006. 10. F. Hubaut, ATLAS Note ATL-LARG-2000-009. 11. V. M. Aulchenko et al, ATLAS Note ATL-LARG-2001-016. 12. L. Di Ciaccio, L. Neukermans, P. Perrodo and R. Zitoun, ATLAS Note ATL-LARG-2001-009, P. Barrillon, F. Djama, L. Hinz and P. Pralavorio, ATLAS Note ATLLARG-2001-012, P. Barrillon, F. Djama, L. Hinz and P. Pralavorio, ATLAS Note ATLLARG-2001-014.

STATUS OF ATLAS TILE CALORIMETER AND STUDY OF M U O N INTERACTIONS

L. E. PRICE Bldg 362, Argonne National Laboratory, Argonne, IL 60439, USA E-mail: [email protected] (For the ATLAS Tile Calorimeter

Collaboration)

In this paper, we provide a description and status report on the barrel hadronic calorimeter for the ATLAS detector at the CERN Large Hadron Collider. We describe measurements taken with prototype and initial modules of the calorimeter, in particular those involving muon energy loss.

1

Brief Description of the ATLAS Tile Calorimeter

The ATLAS 1 Tile Calorimeter2 is a sampling device made out of steel and scintillating tiles, as absorber and active material respectively. It realizes a simple and very well proven idea of calorimetry, particularly suited for the LHC environment. The absorber structure is a laminate of steel plates of various dimensions, connected to a massive structural element referred to as a girder. The highly periodic structure of the system allows the construction of a large detector by assembling smaller sub-modules together. Since the mechanical assembly is completely independent from the optical instrumentation, the design becomes simple and cost effective. Simplicity has been the guideline for the light collection scheme used as well: fibers are coupled radially to the tiles along the outside faces of each module. The laminated structure of the absorber allows for channels in which the fibers run. The use of fiber readout allows the definition of a three-dimensional cell read-out, creating a pseudo-projective geometry for triggering and energy reconstruction. A compact electronics read-out is housed in the girder of each module. Finally, the read-out of the two sides of each of the scintillating tiles into two separate photon detectors (in our case photomultipliers, PMTs) guarantees a sufficient light yield and provides a redundancy which might be needed during the long expected period of operation of the ATLAS experiment. A conceptual design of this calorimeter geometry is shown in Figure 1. The absorber structure is a laminate of steel plates of various dimensions stacked along Z. The basic geometrical element of the stack is termed a period. It consists of a set of two master plates (large trapezoidal steel plates, 5 mm thick, spanning along the entire X dimension) and one set of spacer plates (small trapezoidal steel plates, 4 mm thick, 10 cm wide along X). During construction, half-period elements are pre-

409

410

assembled starting from an individual master plate and the corresponding 9 spacer plates. The relative position of the spacer plates in the two half periods is staggered in the X direction, to provide pockets in the structure for the subsequent insertion of the scintillating tiles. Each complete stack, called a module, spans 2?t/64 in the azimuthal angle

Figure 1. Conceptual layout of ATLAS Tile Calorimeter

(Y dimension), 100 cm in the Z direction and 180 cm in the X direction (about 9 interaction lengths, \h or about 80 effective radiation lengths, X 0 ). Each module has 57 repeated periods. The module front face, exposed to the beam particles, covers 100x20 cm . The scintillating tiles are made out of polystyrene material of thickness 3 mm, doped with scintillator. The iron to scintillator ratio is 4.67 : 1 by volume. The calorimeter thickness along the beam direction at the incidence angle of 0 = 10 deg. (the angle between the incident particle direction and the normal to the calorimeter front face) corresponds to 1.49 m of iron equivalent length9. Wavelength shifting fibers collect the scintillation light from the tiles at both of their open (azimuthal) edges and bring it to photo-multipliers (PMTs) at the periphery of the calorimeter (Fig. 1). Each PMT views a specific group of tiles through the corresponding bundle of fibers. The modules are divided into five segments along Z and they are also longitudinally segmented (along X) into four

411 depth segments. The readout cells have a lateral dimension of 200 mm along Z, and longitudinal dimensions of 300, 400, 500, 600 mm for depth segments 1 - 4 , corresponding to 1.5, 2, 2.5 and 3 Xt at © = 0 deg. respectively. Along Y, the cell sizes vary between about 200 and 370 mm depending on the X coordinate.

2 2.1

Calorimeter Fabrication and Status Mechanics

The full Tile Calorimeter consists of a barrel cylinder centered at Z=0 which is 564 cm long and an extended barrel cylinder 292 cm long at each end of the barrel and

Figure 2. A submodule during stacking separated by a gap for detector services. As of October, 2001, 65% of the barrel modules had been completed at the JINR Laboratory in Dubna, Russia. Extended barrel modules are being constructed in the U.S., where 50% of the modules were completed and in Spain where 73% were completed. As a practical matter, the steel stacks are assembled first in units called sub modules which are then combined into modules (see Fig. 2). A barrel module is made of 19 submodules and an extended barrel module contains 9 submodules. In October, 2001, 90% of the 2405 submodules had been fabricated at sites in Russia, U.S., Spain, and Italy. Extensive measurements are done on each submodule and then on completed modules to ensure that mechanical standards are maintained, especially that the completed module is contained inside its specified envelope and will fit properly with other modules into a barrel structure.

412 2.2

Instrumentation with Tiles and Fibers

After a module is assembled, the plastic scintillator tiles are inserted into the gaps left in the steel structure. Wavelength shifting fibers are held against the edges of the tiles and carry the light to plastic "cookies" which are viewed by photomultiplier tubes inside the girders. All of the roughly 400,000 scintillator tiles have been

Figure 3. Fiber routing stage at end of instrumentation work on a module produced in Russia and inserted into the Tyvek sleeves which protect them and reflect light. 65% of the fibers had been inserted into plastic "profiles" by automatic machinery in Portugal. As of October, 2001, 50% of the modules had been fully instrumented. Fig. 3 shows the fiber routing stage of instrumentation where fibers are sorted and gathered into their plastic cookies. 2.3

Electronics and Readout

Photomultiplier tubes and front-end electronics are packaged in drawers which are inserted into the girders at the outer radius of each module. A drawer is shown in Fig. 4. In October, 2001, 45% of the 10010 PMTs and 3% spares had been delivered and tested in six different collaborating institutes.

413

Figure 4. A readout drawer containing photomultipliers and front-end electronics

2.4

Calibration and Testing

Each module will be tested and calibrated with a Cs137 source which moves through T L B * 12 N e R R u n # 2 1 2 3 8 S a l e V * *IVIOAII

13.23

Figure 5. Typical response data from the Cs source

414

the module and illuminates each scintillator tile. Kg. 5 shows an example of the data obtained on a pass of the source, with a peak as the source passes each scintillator tile. 3

Calorimeter Beam Tests

Extensive measurements have been made with modules in a test beam, using the arrangement shown in Kg. 6. Exposures have been made with muons, pions, and electrons. Fig. 7 shows the response of the calorimeter at 180 GeV.

Figure 6. Arrangement of modules for test beam 3.1

Calibration and light yield

The light yield has been estimated from the statistical variation of differences in signal between the fibers and photomultiplier tubes. We find typically 50 to 60 photoelectrons per GeV energy deposited. Gains on the photomultiplier tubes axe adjusted to yield 1.2 pC per GeV.

415

2001/0a/2S 13.48 '

I

1

1

'

4000

J 35 OC 30 00

-

2300

i

2000

'•

1500 tooo

-

500

90

°_

0

XI

too

150

, \

200 2S0 CellAlO

Figure 7. Response to 180 GeV muons, pions, electrons 3.2

Muon Energy Loss in Iron and Nuclear Form Factor Effects

Energy losses of muons at very high energies, up to 10 TeV, have been measured in cosmic-ray experiments3"5. In these experiments muon energies were measured with a magnetic spectrometer, and reasonable agreement between data and calculations was found, but not in the region of very small energy losses . Energy losses of muons up to 300 GeV were measured in various accelerator experiments . A reasonable agreement with theory was reported in6"8. Preliminary results of 300 GeV muon energy loss measurements in iron (lead) indicated9 about 7% (10%) higher probability compared to Monte Carlo predictions. A measurement was performed in 1998 with 180 GeV positive muons incident on a preseries module of the ATLAS Tile Calorimeter (Module 0). In this setup muons traversed 5.6 m of finely segmented iron and scintillators, thereby providing high statistics and high granularity data. Contamination from hadrons and muon decays in flight are eliminated using the first 1.5 m of the muon track in the calorimeter. The results are compared with theoretical predictions in Fig. 8. Particular attention is given to muon bremsstrahlung which is the dominant process leading to large energy losses. In this region we clearly observed for the first time the suppression of bremsstrahlung due to the nuclear elastic form factor.

416

00 0.2 1

' (U

0.5 1

l

i

i

l

l

1

2

1

•

5

HE

Scv&tned Hucleas

O

II

10

100

, 1

"•I

U:o.scr«5H**l n i t c k u s

L

"4

T L ITJUI _ £ f L i f l T f J ^ J ^ 5t - ^ ^*TT T T • n -1-ItJL

4

£o

i Ju i 1 ^ y**f ~i

0)

frrfTW

\ a

^-0.2 - •

1«08«Vmiioo«(«e*ilci)

4

1»S«Vmu«w(SS
-0A

1

.

^

t rv. 4^1 t-M ( \. >1 Jt- T \ 1\

h

i , / - 1.63 ± 0.17^ ± 0.23^ ± 2 S . „ CPubHshtX! eak--:ilatif«)s give- 0.5 to i .5;

l

._ ,_ , Figure 8. Response to 180 GeV muons, pions, electrons References 1.

ATLAS Collaboration, ATLAS Technical Proposal for a General-Purpose pp Experiment at the Large Hadron Collider, CERN/LHCC/94-93, CERN, Geneva, Switzerland, 1994. 2. ATLAS Collaboration, ATLAS Tile Calorimeter Technical Design Report, ATLAS TDR 3, CERN/LHCC/96-42, CERN, Geneva, Switzerland, 1996. 3. W. Stamm et al., Nuovo Cim. 51 A, (1979) 242 4. K. Mitsui et al., Nuovo Cim. 73A, (1983) 235 5. W.K. Sakumoto et al., Phys.Rev. D45, (1992) 3042 6. J.J. Aubert et al., Z. Phys. C10, (1981) 7. R. Kopp et al., Z. Phys. C28, (1985) 171 8. R. Baumgart et al., Nucl. Instrum. Methods A258, (1987) 51 9. M. Antonelli, G. Battistoni, A. Ferrari, P.R. Sala, Proceedings of the 6th International Conference on Calorimetry in High-energy physics, 1996, Frascati, Italy, p. 561. see also ATLAS Collaboration, Calorimeter Performance Technical Design Report, CERN/LHCC 96-40, CERN, 1997, p. 150-152 10. E. Berger et al., Z. Phys. C73, (1997) 455^163; CERNPPE/96-115, CERN 1996 (1981) 635

C O N S T R U C T I O N OF T H E FIRST CMS-ECAL FULLY O P E R A T I O N A L M O D U L E (400 L E A D T U N G S T A T E CRYSTALS)

E T I E N N E T T E A U F F R A Y , F R A N C E S C A CAVALLARI, JULIEN C O G A N , PAUL LECOQ, LAURA P E R E Z P R A D O AND PABLO S E M P E R E ROLDAN* CERN, Geneva, * Universidad de Santiago

Switzerland de Compostela,

Spain

MARC SCHNEEGANS LAPP,

Annecy,

France

The CMS Electromagnetic Calorimeter (ECAL) will be made of almost 80000 lead tungstate crystals. T h e barrel part will be composed of 36 supermodules with 1700 crystals each, divided in 4 modules. This paper reports on the construction of the first Module, called Mod # 0 ' , that will be tested under beam conditions. The optimum quality of the crystals that form Mod # 0 ' together with the light collection uniformization techniques applied and the results obtained will be described. In addition, different gluing techniques and results for the 400 Mod # 0 ' crystals will be also reported in this paper. Other aspects, such as the mounting sequence that follows the gluing of the crystals to the photodetectors or the thermal regulation of the module will be also discussed.

1

Introduction

The Large Hadron Collider (LHC) is an instrument that will explore new Physics at the higher energies ever achieved, aiming to find the Higgs boson. The LHC is being built at CERN and by 2005 it will be ready to produce head-on collisions of protons at a centre-of-mass energy of 14 TeV. Several spectrometers will be installed along the 27 km LHC ring. One of them, the so-called Compact Muon Solenoid (CMS), will consist of a central inner tracker and a calorimeter system (electromagnetic and hadronic) both located inside a solenoidal coil producing a 4T magnetic field. In addition, a muon detection system outside the coil completes the CMS detector. CMS will have its electromagnetic calorimeter (ECAL) composed of ~ 80000 Lead Tungstate (PbW04) crystals divided into a barrel region and two endcaps. In this article, the construction of the first large size ECAL barrel unit is reviewed. Such a unit (a Module) is called Mod # 0 ' and comprises 400 pre-production PWO crystals.

417

418

2

Improvement in P W O crystals quality

The choice of the Lead Tungstate (PbW0 4 ) for the ECAL of the CMS collaboration was performed in 1994. Since that moment, an intensive R & D program started in collaboration with the producers in order to improve the optical properties and radiation hardness of the crystals. By 1998, the evident improvements achieved1 allowed to enter into the pre-production phase. The main objectives of this phase for the producers were to increase the production rates and to improve even more the crystals quality and the homogeneity of their properties. For the ECAL community, the goals were to set up all the activities performed at the Regional Centres (RC) and the installation of the measuring devices called Automatic Crystal COntrol Systems (ACCOS). During this phase, 6000 crystals were delivered by BTCP° producers to CERN RC were they were characterized using ACCOS machines 2 . The excellent results achieved3 for such a large amount of crystals in terms of the stringent specifications defined4 lead to the end of the pre-production phase in year 2001 and the beginning of the production phase. The crystals used for the Mod # 0 ' were chosen among the 6000 pre-production ones. 3

Uniformization of Light Collection

The CMS ECAL will play an essential role in the potential discovery of a light Higgs (approximate mass range: 100 GeV < m # < 150 GeV), the main decay of which is given by the H —+ 77 channel. In order to detect the gamma rays from such a decay, the energy resolution should be well below 1%, the stochastic and constant term being the most important contributions to the energy resolution at this gamma energies. The former is assumed to be low for an homogeneous calorimeter such as the CMS ECAL (if enough light is collected), whereas the latter requires a special effort to be minimized. The light collection non-uniformity around the shower maximum (8X0) constitutes a major contribution to the constant term 5 . In order to quantify this non-uniformity, we define the Front Non-Uniformity or Fnuf as the percentual change in the light collected per radiation length in the Front region (4 to 13 XQ), and similarly we define the Rear Non-Uniformity or Rnuf in the same terms but in the Rear region (13 to 22 Xo). In order to limit the contribution to the constant term of these non-uniformities to less than 0.3%, we require the Fnuf and Rnuf parameters to be within these limits: -0.35 < Fnuf < +0.35 a

,

-1.80 < Rnuf < +0.25

Bogoroditsk Techno-Chemical Plant, Bogoroditsk, RUSSIA.

(1)

419

where Fnuf, Rnuf and the limits are expressed in units: %/XQ. The light collection profile of a crystal is defined by the balance of two effects: the focusing, induced by the pyramidal tapered shape of CMS barrel crystals and the absorption. The former trends to enhance the light emitted far from the photodetector (located in the Rear face), whereas the latter shows an opposite behavior. As explained in Sec. 2, PWO crystals present nowadays optimum optical properties, thus the absorption is very weak. This causes the focusing effect to be dominant, and so, when crystals are completely polished, the light collection profile yields Fnuf values ~ 1.3%/Xo (far from our acceptable limits, expressed in Eq. (1). This means that we must find a method to uaiformize the light collection. This method 5 , defined at CERN RC in 1998, is based on the depolishing of a lateral face of the PWO crystals. By doing this, we destroy the total internal reflection on that face, and thus, we strongly reduce the focusing. This technique is nowadays applied by the producers, yielding excellent results, as is shown in Fig.l, where the final Fnuf and Rnuf values for the 400 crystals chosen for Mod # 0 ' are depicted.

Frani NOP (%/XaJ

R ™ N ^ . iStfXe)

Figure 1. Pnuf (Left) and Rnuf (Right) distribution for the 400 Mod # 0 ' crystals.

4

Gluing of t h e C r y s t a l s t o t h e P h o t o d e t e c t o r s

The glue used to join the ECAL barrel PWO crystals to their photodetectors must fulfil certain requirements. First of all, it must present an index of refraction at least as high as the one from APDs protective epoxy window (to maximize the angular acceptance of photons), in addition, it must present a large attenuation length in order to minimize the absorption losses. Some other important requisites are: to be chemically inert with any of the parts participating in the gluing, to remain unchanged during the whole CMS life-

420

time and of course to be resistant to the high radiation levels present in CMS.

Figure 2. Some related features t o t h e gluing: Left: Photograph of t h e gluing bench, with several stored multiboxes containing recently glued crystals. Right Photograph of a gluing acquired with the bubble-viewer through a 23 cm length P W O barrel crystal.

The gluing of the 400 Mod # 0 ' crystals was performed using two different glues: the RTV 31456 and the UV-curing NOA 61 c . For this gluing task we have used several tools. Among them, we can mention, a gluing bench, that allows to glue and store during curing time an amount of 150 crystals placed inside the so-called multiboxes d , as shown in Fig.2-Left. In addition, a bubble viewer allows the operator to check the presence of air bubbles simply by looking through the crystal (this is possible since the crystals are very transparent, as Fig.2-Right shows). The results obtained after the Mod # 0 ' gluing task reveals that the glue RTV 3145 gave the best results in terms of bubble-free gluing and acceptable adhesion after a curing time of 1-2 days. 5

T h e Assembly P r o c e s s

The assembly process begins with the gluing of the crystals to the APDs to which they have been paired. After it, both the crystal and the photodetector become a new ECAL unit named sub-unit. Then, ten sub-units are inserted into the alveolar structure that will hold them inside the final detector becoming a new ECAL unit called sub-module. Several submodules are assembled to 6

Dow Corning Silicon Adhesive RTV 3145. Norland Optics Adhesive 61. Norland Products, Inc. d Box containing 5 barrel crystals specially conceived to be adaptable to ACCOS machines.

c

421

create another unit called module. There are four existing module types: Type 1 comprises 50 submodules, and Types 2, 3, and 4 comprise 40 submodules. The assembly of these four module types constitutes a super-module6. Six different steps of the real Mod # 0 ' assembly sequence are shown in Fig.3. It

Figure 3. Sequence of the mounting of Mod #0'.

is important to mention that during the whole assembly process, the objectoriented database named C.R.I.S.T.A.L. 7 provides continuous assistance to the operator. In order to provide such assistance, all the manipulations that constitute the assembly have been arranged into flow-charts or work-iows (sequences of smaller activities that must be successfully finished to complete the whole manipulation). In this manner, C.R.I.S.T.A.L. always indicates to the operator which is the next operation to perform, and hence avoids possible human mistakes. In addition, it records all the operators comments and observations, allowing to keep a track of the activities. 6

Module Thermal Regulation

The crystals and the APDs are very sensitive to temperature. In fact, if the temperature is increased by one degree, the former ones produce a 2% lower light output, whereas the latter ones give a 2.3% lower gain. In consequence, we must provide a way to guarantee the thermal stability of the ECAL. This fi

The sketch of all the ECAL barrel elements can be found elsewhere6

422

is achieved with two independent circuits. The so-called regulating circuit is conceived to keep the temperature of crystals and APDs at 18°C with a maximum tolerable spread of 0.1°C. This circuit is located inside the grid, which is drilled with this purpose, allowing to cool it down by internal water circulation (see Fig.4~Left). Besides, there is a power circuit designed to evacuate the heat dissipated by the electronics, as not more than 10% of this heat is allowed to leak to the grid (see Fig.4-Right). The cooling system described in

Cnstais

'W~~ a>rcnto < c - 1 W / K >

Figure 4. Schema of the cooling circuits. Left: Regulating Circuit. Right: Power Circuit.

Fig.4 corresponds to Mod # 0 proposal. However, during its construction, it was observed that the leakage through the electronic boards connectors was excessive, thus, for the construction of Mod # 0 ' several modifications were performed in order to improve the thermal stability of this ECAL unit: reshaping of some screening pieces and substitution of the copper braids shown in Fig.4 by flexible inox tubes attached to the electronic boards copper plates. 7

Conclusions

The construction of Mod # 0 ' was performed using optimum optical quality crystals, thus, we expect from it the best performances. On the other hand, the construction of this module allowed the CERN Regional Centre to validate the operations carried out, as well as the tooling specifically designed for this mounting. In addition, the assembly of Mod # 0 ' was a great chance for the operators to get a good training in order to achieve the neccessary skills for such complex activities as the ones involved in the assembly process are. The excellent capacities of C.R.I.S.T.A.L. as data management system were also confirmed.

423

Acknowledgments I would like to thanks the CERN Regional Center operators Bruno Buisson, Herve Cornet, Eric Gitton, Alain Machard and Rene Morino for the essential role played in the uniformisation and characterization task and also in the assembly sequence. In addition, I would like to thank Michel Lebeau for his contribution to the development of the uniformization method. References 1. P. Lecoq, Large scale production of Lead Tungstate crystals in Russia, Proceedings of the International Workshop on Tungstate Crystals (Rome, 1988). 2. E. Auffray et al, Performance of ACCOS, an Automatic Crystal quality Control system for the PWO crystal of the CMS calorimeter, Nucl. Instr. Meth. A 456 (2001) 325-341. 3. E. Auffray et al, Status on PWO crystals from Bogoroditsk for CMSECAL, Proceedings of International Conference SCINT 2001, Chamonix (France), September 2001. 4. E. Auffray et al., Specifications for lead tungstate crystals preproduction, CMS Note 98/038. 5. P. Sempere et al, Uniformization of light collection in lead tungstate crystals in view of a high resolution electromagnetic calorimeter, Proceedings of the IX Int. Conference on Calorimetry in High Energy Physics, Annecy (France), October 2000. 6. E. Auffray, CMS/EC AL Construction and quality control, Proceedings of the VIII Int. Conference on Calorimetry in High Energy Physics, Lisbon (Portugal), June 1999. 7. J.M Le Goff et al., C.R.I.S.T.A.L. / Concurrent Repository & Information System for Tracking Assembly and production Lifecycles. A data capture and production management tool for the assembly and construction of the CMS ECAL detector, CMS Note 96/003.

A N E W CONCEPT FOR A N ACTIVE ELEMENT FOR THE LARGE COSMIC RAY CALORIMETER ANI

F. STEINBUEGL, J. GEBAUER, E. LORENZ AND R. MIRZOYAN Max-Planck-Institute for Physics, Foehringer Ring 6, 80805 Munich, Germany E-mail: [email protected] A. CHILINGARIAN Yerevan Physics Institute, Yerevan,

Armenia

D. FERENC UC Davis, Davis, CA, USA B. JOKELE Technical University of Munich, Munich,

Germany

For the half completed ANI sampling calorimeter (1600 m2 detection area, 6 concrete absorber layers of 1 m thickness each) at Mount Aragats, Armenia, a cheap and efficient active detector element is needed. A new concept for such a detector element and first results from a reduced size prototype are presented.

1

Introduction

T h e energy spectrum of cosmic rays comprises more t h a n 12 orders of magnitude and extends to t h e enormous energies u p t o 10 2 1 eV. T h e Cosmic ray " b e a m s " bring information about most energetic processes in t h e Universe and could be used for investigation of strong interaction parameters at energies well above achievable on modern particle colliders. Extremely low fluxes of t h e Cosmic Rays (CR) of t h e highest energies require a large detector area, respectively volume. Cost is a limiting factor and there is a need t o find cheap active elements of large volumes. A typical example is t h e need for t h e active element for the large calorimeter project ANI 1 . ANI is located on t h e Mount Aragats and is half completed after the collapse of t h e former Soviet Union. Besides an operational scintillator array t h e ANI-detector comprises a 40x40 m 2 concrete absorber of 6 layers of 1 m thickness each, interspersed by 40 cm gaps. T h e high altitude of 3200 m asl makes this detector particularly interesting for studies of charged CRs above 10 1 4 eV where besides t h e operational Kascade array at Karlsruhe 2 no other large experiment is planned.

424

425

2

T h e basic concept and some prestudies

Here we want to present both the concept and the first results from a reduced size prototype for the active element for such a calorimeter or similar applications 3 . The basic element is a long tube filled with water and read out by 2 photomultipliers (PMT) at both ends. A typical configuration would consists of 20(40) m long tubes of, say, 30x30 cm 2 crossection, arranged alternatively in x and y direction in consecutive gaps. The inner surface of the tubes are lined with a new highly reflective (specular reflector) foil ensuring good light piping over long distances, respectively of detectors with large aspect ratios (1/d). Fast charged particles passing through the water produce Cherenkov light. A significant fraction ( « 50%) of the light is absorbed by a dissolved wavelength shifter dye (WLS) and and re-emitted around 420500nm. The re-emitted light is isotropic and well matched to the spectral range of the high reflectivity-liner and the sensitivity of standard PMTs. Figure 1 shows the conceptual design of such an element.

Figure 1. Conceptual design of a detector element with readout.

The main problem in past approaches of similar configuration was inefficient light transport over long distances in thin (d/1 98 % reflectivity between 400 and 700 nm (We are actually

426

negotiating with 3M about extending the high reflectivity down to 300 nm.). The foil does not contain any metal and is completely inert against many

40 60 80 100 120 140 160 D i s t a n c e (light e m i t t e r - r e c e i v e r ) ( c m j

180

Ploxifllasmodow

p,-^

Tcstod Male,irtal

"V

fr-

Figure 2. Signal reduction along a tube of aspect ratio 71 for isotropically emitted light and different lining.

liquids and solutions (for example salt dissolved in water or alcohol-water

mixtures). As readout elements PMTs (9390B from ET) with a diameter of typically half the diameter of the tubes are used. Using short Winston type light concentrators of the same 3M foil material one can in principle economise on the size of the PMTs and also improve on the time dispersion by rejecting large angle photon paths at the expense of small signal loss. As WLS many fluorescent dyes can be used, for example dyes used for optical brighteners in washing powder, white paper etc. These dyes are nonpoisonous, mass produced, dissolve easily in water and have a quantum efficiency (QE) close to 100%. Fig. 3 shows the absorption and emission curve of a tested WLS, the absorption cut-off of water, the reflectivity of the 3M foil VM 2000 and the QE of a standard PMT with bialkali photocathode. Calibrations and monitoring of the tube performance will be realized by i) LED pulsers, ii) by charge pulses injected at the inputs of the charge sensitive preamps and iii) by cosmic muons (2000/sec/20 m tube at 3200 m asl). LEDs are either blue ones (< A > = 428 nm or < A > = 470 nm) or Nichia UV-LEDs (< A > = 370 nm) exciting the nearby WLS. Some problems might occur during subzero temperatures in winter, requiring either external heating or replacing the water by a water alcohol mixture, respectively the admixture of

427

•••-

•:•••••;

iu,;,;,;,

£ «...

o

•

250

275

300

326

350

Absorption of WLS • - • • Transmittance(1m, pure water)

•

375 400 425 Wavoiongth [nmj

450

Emissions of WLS Reflectivity of VM2000-FOII

475

500

525

550

575

» Quantum Efficiency of PMT

Figure 3. Some optical parameters of the active element.

salt to the water. 3

Test of the basic element

In the spectral range between 250 and 400 nm a total of 280 Cherenkovphotons/cm pathlength are generated by a traversing charged particle of 0 « 1. The diameter of our prototype tube is 20cm (length = 6m). For the tested WLS (Radiant D282) of fts 95 % QE about 2600 photons are emitted isotropically, which are emitted basically half and half in the two directions. Due to water absorption and reflectivity losses only some fraction of the photons will reach the PMTs. Figure 4 B shows the decrease of photons as a function of distance from the PMT. A muon at the far end produces ~ 7 photoelectrons (QE of the used PMTs is about 17% in the range or interest). Figure 4 A shows typical PMT-signals. The different amplitudes , different onsets and shapes reflect the passage of a muon close ( « 1 m) to PMTl. From both the amplitude and timing measurements we concluded that a position resolution of <30 cm can be reached for a single particle. 4

Summary and outlook

Test results show that the presented concept for an active detector element is working as expected. The length for a 50% loss of photons is li/2 « 2.5m — 3.5m. This makes the extension of the tube length to 20 or 40m possible (if a single muon calibration over the entire length is not required)

428 DIslalK* to PMT1:1 ,00m

• PMT1 • PMT2 • PMTHPMT2

DlslancotoPMT2:5.00i

S 'M

75 100 125 150 175 200

rm« [n»i

Distance to PMTI/PMT2[OTI]

Figure 4. A: Typical PMT-signal (seen by an oscilloscope before the shaping amplifier) caused by a crossing particle. B: Decrease of photons as a function of distance t o traversing muons (No Winston cone, no optical grease between the P M T and the window).

and consequently the detector element has the potential for the application in large area sampling calorimeters. A first cost estimate shows that a tube element can be built for $ 1500-2000 (dominated by the PMT price) for both the 20 and 40 m long units. A total of 800 (40 m version) respectively 1600 tubes of 20 m length would be needed for the ANI calorimeter. The test shows a significant improvement potential: i) Search for more suitable wavelength shifter dyes, ii) Influence of WLS-concentration to the efficiency of the detector element, iii) Influence of antifreezing compounds to the absorption and emission spectra of the WLS-dyes. Also long-term aging tests need to be performed. Acknowledgments We want to thank our colleagues from the HEGRA collaboration, S. Borngrebe and the workshop of the MPI for the production of the prototype. We also thank 3M for providing the new VM 2000 foil sample. References 1. 2. 3. 4.

Danilova T.V. et al., Nucl. Instrum. Methods A 323, 104-107 (1992). Doll P. et al., KASCADE Collab., Karlsruhe, Report KfK 4648 (1990) Weller M.F. et al., Sience Vol., 287, 2451, (2000). Steinbuegl F., Diploma thesis, TU-Munich, (2002), unpublished.

WHAT'S N E W W I T H T H E CMS H A D R O N CALORIMETER VASKEN HAGOPIAN For the CMS HCAL Collaboration Department of Physics, Florida State University, Tallahassee, FL. 32306, USA E-mail: [email protected] The CMS Hadron Calorimeter is designed to measure hadron jets, single hadrons and single JJ'S. The Central Barrel and the two End Caps, made of brass and scintillators cover the | r\ | range of 0.0 to 3.0. The two Forward Calorimeters made of iron and quartz fibers extend the | n | range to 5.0. Scintillators are also placed outside of the magnet coil, within the muon system to measure the energy leakage from the Central Barrel. The construction of the calorimeter is about 50% complete. Several design changes were made to simplify the calorimeter and reduce the cost. The longitudinal segmentation of the central barrel and end caps was reduced by one unit. The quartz fiber diameter was doubled from 300 to 600 microns. Improvements were made to the Hybrid Photodetectors (HPD) and various other components. The special purpose ADC (QIE) and other electronics are in prototype stage.

1

INTRODUCTION

The CMS is one of the major general purpose facilities for the CERN Large Hadron Collider (LHC). The Hadron Calorimeter x (HCAL) Central Barrel (HB) is 9 meters long, one meter thick and 6 meters in outer diameter. The HB is made of 36 wedges of brass and scintillator. The two End Caps (HE) are also made of brass and scintillator, with a diameter of 0.8 to 6.0 meters and a thickness of 1.8 meters. Both HB and HE are inside the 4-tesla solenoid coil. The r\ — <j> segmentation of HB and HE is 0.087 x 0.087, except near | T) | 3.0 where the size of the segmentation is doubled. The longitudinal segmentation for HB is one unit while for HE varies from one to three. The forward calorimeters (HF) are designed to measure forward jets. Since the HB is only 6.5 interaction lengths thick, the Outer Calorimeter (HO) (scintillators inside the muon barrel system, outside of the solenoid coil) measure the HB energy leakage. Monte Carlo studies and test beam results show that HO improves the energy resolution. To compensate for the radiation damage at | T] | above 2.0, HE has extra longitudinal segments to allow correction for signal loss. Various sections of the calorimeter have been tested using pion, proton, electron and muon beams 2 . For the test beam at CERN a large motion table has been constructed on which a 40° segment of the calorimeter can be placed. Besides the active and passive materials, the calorimeter has optical readout systems using 19 pixel Hybrid Photodiodes (HPD), 12 kV high

429

430

Figure 1. Hadron half-barrel calorimeter in CMS assembly hall

voltage system, low voltage system, special design ADC (QIE), calibration systems using moving radioactive source, laser and LED and special purpose electronics. Details can be found in prior conference proceedings 3 . 2

P R O G R E S S T O DATE

The Felguera factory in Spain constructed the barrel calorimeter wedges. The wedges were assembled in two barrels in the factory to make sure all the components fitted together and verify that the distortions were within acceptable limits. The wedges were then disassembled and shipped to CERN. In CERN the scintillator megatiles were inserted and the first half has been shipped to the CMS assembly hall and reassembled.

431

Figure 2. Hadron half-barrel calorimeter in MZOR

Figure 1 shows the first half-barrel comprising of 18 wedges. HB will be supported inside the magnet cryogenic housing. The first End Cap brass portion of calorimeter is completed and shipped to CERN, where the scintillator megatiles will be inserted. Figure 2 shows a HE End Cap on the MZOR factory floor in Russia. Prototypes of HO and HF have also been constructed and tested in the test beams. Various radiation tests are continuing to better understand the damage. Now that the major construction projects are well under way, the emphasis has shifted to the other parts of the calorimeter. This is where several changes were made recently for a variety of reasons, including simplification and lowering of the cost. HO is being constructed in India. HF construction is a joint project of many countries, including US, Russia, Hungary and Turkey. The major changes and improvements are discussed below.

2.1

HPD

The original design of the HPD's suffered a large amount of cross talk, where about half the signal in one pixel leaked into other pixels. Three sources of cross talk were identified. 1. Electrical cross talk between pixels. This cross talk was eliminated by creating a low impedance diode bias voltage electrode. 2. Optical cross talk due to internal reflection of the light signal. The optical cross talk has been almost eliminated by antireflection coating that produced destructive interference of the reflected light. 3. Back scattered electrons. The magnetic field of the CMS will spiral these electrons in tight circles. This

432

ftemmijl QfES Response Me>tt-lrtvssrti«g Input £ SI. • IMOtgT

$000

1Q0GO

1S900

20000

26000

-\

30»0

Input €h#O0C*C;)

Figure 3. QIE, multirange, nonlinear prototype

solution requires that the HPD axis be well aligned with the magnetic field. 2.2 QIE Figure 3 shows the four-range QIE prototype that almost meets all the specifications for the calorimeter. The QIE will measure charge from 1 fC to 10 pC. The next step is producing several production units that should be the final design and also be used in the test beam at CERN. 2.3 Longitudinal Segmentation Initially HPD's under consideration were the 19 and 73 pixel units, each 5 cm in diameter. The 73 pixel HPD required 66 QIE channels with 11 QIE cards. The width of the 73-pixel readout module became too large to be serviced in the very tight space available in CMS. The original proposed longitudinal segmentation for each tower was to read one layer immediately after the electromagnetic calorimeter and the remaining layers connected to single pixel.

433

This arrangement required the 73 pixel HPD. Using test beam results and Monte Carlo studies, it was found out that the active first layer did not add substantially to the resolution. The passive addition of the first layer to the other layers, with a weight maximizing the resolution, gave almost the same results. Since physical space was tight and to reduce costs, the choice was made to reduce the longitudinal segmentation to one unit. For HE, the longitudinal segmentation was also reduced. These changes reduced the number of HPD's and QIE's by about 15% and 40% respectively. 2.4

Forward Calorimeter - HF

Two major changes were made in the design of the HF calorimeter. The HF is in a very high radiation area and the active material chosen for this subsystem is quartz fibers. The initial choice was 1% quartz fiber of 300-micron diameter. Monte Carlo studies were performed to study the resolution as a function of the quartz fiber diameter keeping the fraction of active material the same. Figure 4 shows a plot of the resolution as a function of the energy for four quartz fiber diameter-space of 2.5 mm (300 mic), 5.0 mm (600 mic), 7.5 mm (900 mic) and 10.0 mm (1,200 mic). The resolution deteriorated very little from 300 to 600 microns. Doubling the diameter reduces the number of fibers to be installed by a factor of four, thereby reducing the manpower. The second design change is replacing the quartz fiber with quartz clad with quartz fiber with plastic clad. This change reduces the cost of the fibers substantially. Test beam studies and radiation damage studies of quartzquartz and quartz-plastic fiber have shown small difference in degradation due to radiation damage. 3

CONCLUSION

The CMS Hadron Calorimeter is about 50% constructed. Some critical improvements have been done that reduced the cost. Focus has now switched to electronics, assembly, installation and calibration. Front-end electronics is also progressing well. We expect to have a working hadron calorimeter on the first day of data taking. ACKNOWLEDGEMENT The CMS HCAL collaboration is composed of about 45 institutions from about a dozen countries. It is truly an international collaboration, with various nationalities taking lead responsibility of various sub-systems. This tremen-

434 Ef G$¥ IOD0O T—r—n

1—T--T"

"•^•^•xt^^

. ...

~ ^ f , •"•"***•„

0,05 1—

!

J

%X05

i

A..

0.04

olio

0.02

0.01 E'^.

n GsV,-***

Figure 4. Fractional energy resolution vs. (energy)1/2 for 300, 600, 900 and 1,200 micron quartz fibers

dous progress is due to the hard work of many members of these institutions, and I gratefully acknowledge the many scientists whose help was crucial for this report. References 1. The CMS - Hadron Calorimeter - Technical Design Report. CERN/LHC 97-31 (1997). 2. V. Abramov, et al, Nucl. Instrum. Methods A 457, 75 (2001). 3. V. Hagopian, Nucl. Phys. B 78, 182 (1999); V. Hagopian in Calorimetry in High Energy Physics, ed. G. Barreira, B. Tome, A. Gomez, A. Maio, M. J. Varanda (World Scientific, Singapore, 2000).

OVERVIEW OF T H E ATLAS LIQUID A R G O N CALORIMETER S Y S T E M ROBERT S. ORR Department of Physics, University of Toronto, 60 Saint George Street, Canada M5S 1A7 (On behalf of the ATLAS Liquid Argon Group) E-mail: [email protected]

Toronto,

An overview is given of the ATLAS liquid argon calorimeter system. The liquid argon system uses several different design technologies in different pseudorapidity regions of the detector. We discuss how the design goals have led to each of these technology choices. For each of the systems we describe the mechanical structure, the measured performance, and the status of construction.

1 1.1

Introduction LHC Environment

The Large Hadron Collider (LHC) at CERN is designed to have an energy and luminosity sufficient to experimentally elucidate the mechanism of electroweak symmetry breaking, and to generally investigate the centre-of-mass energy regime up to 1 TeV constituent collision energy. Given the available machine technology the LHC has been designed to achieve a proton-proton centre-ofmass energy of 14 TeV at a luminosity of 1 0 3 4 c m - 2 s _ 1 . While these machine parameters bring the opportunity for new and exciting discovery physics, they make stringent requirements on the design of the detector elements. 1.2

Calorimeter Design Goals

The ATLAS experiment is designed to exploit the full capability of the LHC at both the design luminosity, and the much lower luminosity expected at the turn-on of the machine. For this reason the complete detector must satisfy the somewhat conflicting needs of both discovery physics such as the Higgs and Supersymmetry, and also "Standard Model" physics such as precision investigations of the top and bottom quarks. In physics terms, the calorimeters must measure, with sufficient resolution, the energy and direction of jets, electrons and photons, and E™lss over a wide rapidity interval; the ability to identify the bunch crossing, and some capability to identify muons are also important requirements. In addition to energy measurement, particle identification is important, crucial areas being 7/7r° and e/n separation.

435

436

EM Endcap Calorimeter and presampler

Figure 1. A cross section through one hemisphere of the liquid argon cryostats, showing the various calorimeter systems.

In terms of system design considerations there are many requirements which have to be optimized. The high LHC bunch crossing frequency requires a fast readout scheme with high segmentation to reduce occupancy, the high luminosity requires that the calorimeters be generally radiation resistant, and the high center-of-mass energy requires a high dynamic range from 1 mip to 5 TeV and a pseudorapidity coverage out to \r]\ w 5. General operating considerations also require uniformity of response, long term stability, and ease of calibration. Finally all of these must be combined with cost considerations, modular construction, and a feasible mechanical scheme for installation in the ATLAS detector. The detector requirements vary somewhat depending upon the position and role of the detector. Clearly physics such as H —> 77 requires a very high resolution on electromagnetic energy combined with excellent position resolution, while the energy resolution for hadronic final states can be considerably relaxed. Similarly, the forward calorimeter (FCAL) must be much more radiation resistant than the barrel electromag-

437

netic calorimeter. Considerations such as these have led us to choose different technologies for different elements of the liquid argon calorimeter system 1. A general arrangement of the calorimeter elements is shown in figure 1. 2 2.1

Calorimeter Technologies Barrel Electromagnetic

Calorimeter

Both the barrel and the endcap electromagnetic calorimeters are lead-liquid argon devices. The accordion structure has been chosen as it meets the requirements on energy and angular resolution, high segmentation, speed of read-out with low pileup, and hermeticity. The design requirements are, for energy resolution:

*" ^ 10%

n*» °-

27

—< . = e o . 7 % ® _,„, TA E ~ ^EjGeV) E (GeV)

,

(i)

and for angular and spatial resolution,

o-e <

40mrad ,

" ^(GeV)

, a? <

8mm .

.

,„. (2)

V£(GeV)

The device is designed to have an energy scale uncertainty of 1%, and an energy response linearity of 0.5%. These resolution requirements must be met over a pseudorapidity range of 0 < \r]\ < 3.2 for the calorimeters, and 0 < \r]\ < 1.8 for the pre-sampler. Each module of the barrel (as shown in figure 2(a)) is 3100mm long, and two such modules cover the pseudorapidity range —1.4
438

(c)

(d)

Figure 2. Details of the construction of the electromagnetic barrel calorimeter: (a) schematic of one module, showing the general structure with the pre-sampler at the bottom, (b) general arrangement of the barrel modules and the pre-sampler in the barrel cryostat, (c) a barrel module during construction, showing t h e accordion structure of lead, electrodes, and honeycomb spacer, (d) a barrel module during construction.

2.2

Endcap Electromagnetic

Calorimeter

The endcap electromagnetic calorimeter is similar to the barrel in conception. However the realization of the accordion structure in this region is mechanically difficult. In order to maintain a spatially constant sampling fraction ATLAS has chosen the "Spanish Fan" configuration; this is in essence a variable fold angle with radius. This geometry can be seen in figure 3(a). Mechanical constraints limit the radius over which this geometrical configuration can be used, and the endcap calorimeter is divided into a small and large radius wheel; again this is clear in figure 3(a).

439

(a)

(c)

(b)

(d)

Figure 3, Details of the construction of the endcap electromagnetic calorimeter: (a) schematic of a sector module showing how t h e accordion structure is assembled, (b) general arrangement of the sector modules into t h e wheel, (c) an end view detailing the accordion structure, (d) a general view of t h e module zero before cabling.

The arrangement of modules into the endcap wheel is shown in figure 3(b), and an impression of the structure can be gained from figures 3(c) and (d). The endcap wheel covers the pseudorapidity range 1.4 < |»7| < 3.2. The modules consist of layers of lead absorber and copper-kapton readout electrodes arranged as an electrostatic transformer with an accordion structure, figure 3(c), similar in concept to that already described for the barrel . There is one wheel at each end of the detector, with 8 modules per wheel. There are 96 gaps in the outer wheel which vary in width from 2.8mm to 0.9mm in order

440

to realize the constant sampling fraction; the inner wheel has 32 gaps varying in width from 3.1mm to 1.8mm. Each wheel assembly of 8 modules has a diameter of 4000mm, corresponding to a depth varying between 22 X0 and 37 X0 as a function of pseudorapidity. There are three longitudinal samples and the lateral granularity is A77 x A