Subject Grouping: Technology

In December 1999 the CMS collaboration made the daring decision to change its tracking detector from a design that included gaseous detectors to one constructed entirely from silicon sensors, using both microstrip and pixel technology. On 15 December 2007, teams working in the cavern at Point 5 on the LHC installed the microstrip tracking system into the experiment. The pixel detector will soon follow, completing the CMS Tracker and marking the culmination of eight years of careful work to design, prototype, construct and commission the largest silicon detector ever built.

The collaboration envisaged a tracking system 40 times larger than any existing silicon detector system, with a performance comparable to the vertex detectors used at LEP. The detector would house about 205 m² of silicon sensors (approximately the area of a tennis court) comprising 9.3 million microstrips and 66 million pixels. The aim was to achieve a precision of about 10 μm in spatial and vertex reconstruction resolutions – enough for excellent identification of heavy flavour hadrons – and excellent momentum measurement over a wide momentum range at the LHC. The readout would require 73,000 radiation-hard, low-noise microelectronics chips, almost 40,000 analogue optical links, 1000 power supply units and 500 off-detector readout and control modules. The complete system would be constructed in two halves from nine separate subdetector units: two each of microstrip inner barrels, outer barrels and endcaps, three pixel units in the form of a barrel system and two identical forward units.

In June 2000, the LHC Committee approved the Technical Design Report for the new design and the project formally got underway. A collaboration of more than 500 physicists and engineers from 51 institutions based in Austria, Belgium, Finland, France, Germany, Italy, Switzerland, the UK and the US, as well as from CERN, took joint responsibility for the project. They agreed that the inner barrel would be constructed by an Italian consortium, the outer barrel system by CERN together with Finnish and US groups, and the two endcaps by European teams. Swiss groups would build the central barrel region of the pixel system and a US collaboration would provide the forward pixel units.

The assembly project
The detailed design of each of the subdetector units took several years, including extensive testing of prototype sensors, modules and the readout, cooling and power systems. Production of the microstrip detector modules began in November 2004 using the sensors, hybrids and electronic components developed during the earlier phase – all of which had been thoroughly studied and evaluated to ensure maximum reliability and performance. Production of these modules was complete by March 2006. Then, after further substantial testing and thermal cycling, they were ready for mounting onto low-mass carbon fibre substructures with pre-assembled cooling circuits.

The project also became a massive worldwide logistical activity. The microstrip sensors were manufactured in Japan, with contributions from Italian industry, and shipped to Europe and the US for evaluation. The sensors were then moved to other European and US destinations for construction into modules using customized automated assembly equipment that CMS engineers had devised; and they journeyed further still for assembly into sub-units such as rods for the outer barrel, shells and discs for the inner barrel, and petals for the endcaps. The pixel system involved a similar transporting of parts, starting with commercially manufactured sensors from Norway.

The electronic readout system relied on developments in radiation-hard electronics and innovations in optical links, technologies that evolved rapidly in the 1990s. The CMS system culminated with the APV25 – the first large readout chip for a particle-physics experiment to use 0.25 μm CMOS integrated circuit technology – and novel analogue fibre-optic links. Much of this development was the responsibility of groups in the UK and teams at CERN, who worked closely with other CMS groups to assemble the elements of the readout system. CERN designed a set of control and ancillary chips using 0.25 μm CMOS technology, extensively exploited both in the Tracker and throughout CMS.

The collaboration also subcontracted a great deal of the assembly work to industries in several countries, including Austria, France, Italy, Japan, Switzerland and the UK

The automated assembly pioneered for this enormous system was vital for constructing thousands of modules quickly, so that the 15,200 required could be delivered on time. It also generated a huge interconnection requirement. Each module was assembled from one or two microstrip sensors, which had to be connected to the APV25 readout chip. The module chips also had to be bonded to their low-mass carrier. The intensive use of automatic-wire bonders met this demand and maintained consistent throughput with few delays, despite occasional variations in bond quality and rejection of sub-optimal modules.

The collaboration also subcontracted a great deal of the assembly work to industries in several countries, including Austria, France, Italy, Japan, Switzerland and the UK. In partnership with CMS institutes, the companies manufactured components and produced electronics boards, mounting and aligning semiconductor lasers, optical fibres, photodiodes, and analogue and digital electronics, including field-programmable gate arrays that were then state of the art. All modules were thoroughly tested in industry – often using CMS-constructed test equipment – then re-tested for acceptance in CMS laboratories. It is impossible here to do justice to the efforts of the CMS institutes, all of which took on significant tasks in assembly, evaluation and procurement.

Collaboration members constructed new facilities in many institutions for assembling the subdetectors, as well as expanding and utilizing large laboratories such as at CERN, Fermilab and Pisa. Aachen assembled one of the Tracker endcaps, while Florence, Pisa and Torino jointly integrated the inner barrels and discs. There were intensive reviews at all system levels for each stage of production and integration to ensure that quality and performance were maintained.

On the main CERN site, CMS built a facility to assemble the final detector and to provide an environment where a substantial fraction of it could be fully commissioned before final installation into CMS at Point 5. The Tracker Integration Facility is a 350 m² class 100,000 cleanroom, which was also used to integrate the entire outer-barrel system and the second endcap. Each subdetector underwent testing and thermal cycling before transportation to CERN. Further acceptance tests took place after arrival before final integration into the support structure.

The two halves of the outer barrel were built inside the Tracker support tube, which is a low-mass carbon fibre cylinder 5.4 m long and 2.5 m in diameter. The outer-barrel subdetector was completed in November 2006. The inner-barrel halves arrived at CERN in April and September 2006 for final testing before insertion into the outer barrel. The first half section was placed in position in December 2006 and the second half inner barrel and the endcap followed rapidly, with integration of the second endcap completed on 22 March 2007.

As each subdetector was assembled, the teams re-tested it to ensure that it continued to achieve the required performance. The integration facility included rack-mounted electronics, cooling and air-conditioning, which allowed the Tracker to observe cosmic-ray events before installation underground. This incorporated a quarter of the complete safety, control, power, data-acquisition and computing systems for the Tracker – destined eventually for the CMS caverns – including electrical and optical cables, which were to be re-used to keep down costs.

From March until August 2007, all aspects of the Tracker underwent testing, including safety, control and monitoring systems.Several million cosmic-ray events were recorded at five operating temperatures ranging between –15 °C and +15 °C. The data were reconstructed using the CMS-distributed computing Grid and were analysed throughout the world. All systems operated reliably during this five-month period and the collaboration verified that the assembled detector met the performance specifications.

Analysis of the cosmic-ray data shows that the performance of the microstrip tracker is excellent. The number of inactive strips is below one part in 2000; noisy strips do not exceed 0.5%. The signal-to-noise ratio, which depends on sensor thickness, was about 28 for 300 μm sensors. Measurements showed the track cluster finding efficiency to be better than 99.8%. All of these results meet or exceed expectations, which bodes well for LHC physics.

Final installation
At the CMS experimental area at Point 5, preparation for installing the Tracker began before the solenoid magnet was even lowered into the cavern in February 2007. Installation and testing of the cooling plants, power systems and off-detector readout electronics, as well as control and data-acquisition systems took place throughout 2007.

The final performance of the subdetectors in LHC collisions is crucially dependent on the electrical quality of the underground environment, which will only become fully understood after the experiment is complete. The Tracker’s electronics are exquisitely sensitive to tiny signals and must be protected against unwanted noise. To achieve this, 32 interconnection units (patch panels) serving different sectors of the Tracker were installed at the edge of the CMS solenoid, through which all electrical power and cooling services – as well as optical fibres and monitoring wires – pass. The patch panels filter electronic noise and will permit in situ optimization of the detector’s grounding. They also provide termination for cooling, optical links and electrical cables so that all services could be tested as far as possible in CMS before the Tracker arrived.

By late September 2007, the installation teams had completed the massive task of installing cooling systems for 450 loops, 2300 power and 400 fibre-optic cables. The microstrip tracker was transported overnight to Point 5 on 12 December and installation into CMS was completed over the following two days. Connection of the services from the patch panels to the Tracker, and commissioning the Tracker with the rest of CMS, will be completed this spring.

The pixel system
Although a physically smaller device, the pixel system has about a factor of seven more channels. Being at the centre of the detector, concern about minimum material budget and higher radiation levels necessitates even greater attention. Interconnection technologies – especially fine-pitch bump bonding, which were not yet mature for applications in particle physics – had to be studied and, in some cases, developed in CMS labs to allow construction of the detector. The pixel assembly project followed a similar course to the microstrip tracker, with significant transport of parts around the world. Fermilab was at the centre of US activity, and was where the final assembly of the forward system was completed following plaquette construction at Purdue University. The team at the Paul Scherrer Institute (PSI) assembled the barrel subdetector with the collaboration of Swiss universities. PSI also designed the pixel readout chip, while other chips were developed in PSI and the US; the pixel detectors have also exploited components from the microstrip tracker.

The pixel system is scheduled for insertion into CMS following the installation and bake out of the LHC beam pipe in April. The complete forward subdetector was transported to CERN from Fermi-lab in December 2007 and is now undergoing extensive system tests at the Tracker Integration Facility. The barrel subdetector is also complete and currently being commissioned at PSI. It will be transported to CERN in April.

When the LHC begins to open up a new high-energy frontier, it will achieve the highest concentration of energy in operations with lead ions. The collisions, each involving around 400 nucleons with a total energy of more than 1000 TeV, will create strongly interacting, hot, dense matter – a melting pot of quarks and gluons called quark–gluon plasma. This matter will exist for only an instant, and the main goal of the ALICE experiment is to search for evidence of its existence among the many thousands of particles emerging from each collision. One important piece of evidence will be the detection of dimuons – pairs of muons of opposite sign. For this reason, the muon spectrometer, which incorporates some of the first detectors installed in the ALICE underground cavern at Point 2 on the LHC ring, has a key role.

Dimuons are emitted in the decays of vector mesons containing heavy quarks, such as the J/Ψ, the Ψ’, and members of the Υ family. Dimuons will also reveal the decays of light vector mesons (φ, ρ and ω) and of particles with open charm and beauty. The heavy quarkonia states represent one of the most powerful methods to probe the nature of the medium produced in the early stages of the heavy-ion collisions. Indeed, more than 20 years ago, Tetsuo Matsui and Helmut Satz pointed out that J/Ψ production should be suppressed if a quark–gluon plasma is formed in the collision. This provides a strong motivation for experimental studies of J/Ψ and Ψ’ production, undertaken at the energies of the SPS at CERN and RHIC at Brookhaven National Laboratory (BNL). The LHC will be special, however, because two families of resonances (J/Ψ and Υ) rather than one will be experimentally accessible, thanks to the higher beam energy. In addition, the temperature of the quark–gluon “bath” at the LHC is expected to be high enough to “melt” all or most of the Υ states.

As in many experiments, including ATLAS and CMS at the LHC, the role of the muon spectrometer is to detect muons and measure their momenta from the bending of their tracks in a magnetic field. However, there are some very specific aspects of the spectrometer’s design because the ALICE experiment will specialize in studying heavy-ion collisions. In ATLAS and CMS, the muon spectrometer follows the “barrel and endcaps” construction based on a toroidal or solenoidal magnetic field. ALICE also has a central “barrel” of detectors inside the large-aperture solenoid magnet from the L3 experiment at LEP, but the muon spectrometer – with its own large dipole magnet – is located at one side of the barrel, where it will detect muons emitted at small angles with respect to the beam. Isolating muons in heavy-ion collisions requires a large amount of material (absorber) to reduce the huge numbers of hadrons, but the absorbers also stop low-energy muons. So, the measurement of vector mesons (in particular the J/Ψ and Ψ’) of low transverse momentum (p_t) is feasible only at small angles, where the muons emitted in their decay have rather high energies owing to the Lorentz boost.

Both the special environment of the heavy-ion collisions and the physics involved have led to other important criteria for the design of the spectrometer. For example, the tracking detectors must be able to handle the high multiplicity of charged particles that are produced. Also, the accuracy of the dimuon measurements is limited by statistics (at least for the Υ family), so the geometrical acceptance must be as large as possible.

The main goal will be to resolve the peaks of the Υ, Υ’ and Υ”, which requires resolutions of 100 MeV/c² for masses around 10 GeV/c². This in turn determines the bending strength of the spectrometer magnet as well as the spatial resolution of the muon tracking system. It also imposes the need to minimize multiple scattering in the structure and carefully optimize the absorber. Finally, the spectrometer has to be equipped with a dimuon trigger system that matches the maximum trigger rate handled by the ALICE data acquisition.

Figure 1 shows the main components of the spectrometer. Closest to the interaction region, there is a front absorber, to remove hadrons and photons emerging from the collision. Five pairs of high-granularity detector planes form the tracking system within the field of the large dipole magnet. Beyond the magnet is a passive muon filter wall, followed by two pairs of trigger chamber planes. In addition, there is an inner beam shield to protect the chambers from particles and secondaries produced at small angles.

The absorbers have a crucial role, so the collaboration has taken great care in their design. The front absorber has to remove hadrons coming from the interaction region without creating further particles and without affecting muons that come from vector meson decays. This absorber is located inside the L3 magnet. It has a composite structure of different materials to limit small-angle scattering and energy lost by the muons and to protect other detectors in ALICE from secondary particles produced in the absorber itself.

Building such a complex item was an impressive international effort. The tungsten came from China, the aluminium from Armenia, the steel from Finland, the graphite from India, the borated polyethylene from Italy, the lead from the UK and the concrete from France. Engineers from Russia and CERN designed the absorber, the Chinese assembled it at CERN and the International Science and Technology Centre in Moscow provided part of the funding.

The spectrometer itself is shielded throughout its length by the beam shield. This is a dense absorber tube made of some 100 tonnes of tungsten, lead and stainless steel, which surrounds the beam pipe. The inner vacuum chamber has an open-angle conical geometry to reduce background particle interactions along the length of the spectrometer.

While the front absorber and the beam shield are sufficient to protect the tracking chambers, the trigger chambers need additional protection. This is provided by an iron wall about 1 m thick – the muon filter – located between the last tracking chamber and the first trigger chamber. Together, the front absorber and the muon filter stop muons with momentum of less than 4 GeV⁄c.

The spectrometer design is constructed around a dipole magnet that is among the largest ever built using resistive coils (figure 2). With a gap between poles of about 3.5 m and a yoke about 9 m high, it weighs 850 tonnes. To provide the required resolution on the dimuon mass, it has a field of 0.7 T, with a field integral between the interaction point and the muon filter of 3 Tm.

There are two main requirements that underpin the design of the tracking system: a spatial resolution better than 100 μm and the capability to operate in the high-multiplicity environment. For central lead–lead collisions, even after the absorbers have done their work, a few hundred particles will nevertheless hit the muon chambers, with a maximum hit density of about 5 × 10^–2 cm^–2. Moreover, the system has to cover an area of about 100 m².

These demands all led to the choice of cathode-pad chambers to detect the muons. There are 10 planes of chambers in all, arranged in pairs to form five stations: two pairs before the dipole magnet; one inside it; and two after. Each chamber has two cathode planes to provide 2D hit information. The read-out pads are highly segmented to keep the occupancy down to around 5%. For example, in the region of the first station close to the beam pipe, where the multiplicity will be highest, the pads are as small as 4.2 × 6.3 mm². Then, as the hit density decreases with the distance from the beam, larger pads are used at larger radii. This keeps the total number of electronics channels to about 1 million.

To minimize the multiple scattering of the muons, the chambers are constructed of composite materials such as carbon fibre. This technology allows for extremely thin and rigid detectors, resulting in the chamber thickness as small as 0.03 radiation lengths. The tracking stations vary in size, ranging from a few square metres for station 1 to more than 30 m² for station 5. This led to two different basic designs for the chambers. The chambers in the first two stations have a quadrant structure, with the read-out electronics distributed on their surface (figure 3). For the other stations, the chambers have an overlapping slat structure (figure 4) with the electronics implemented on the side of the slats. The maximum size of the slats is 40 × 240 cm².

The front-end electronics for reading out the signals from the tracking chambers is based on custom-designed VLSI chips, developed within the ALICE collaboration. The system uses the MANAS chip, which was derived from the GASSIPLEX chip used for other detectors in ALICE, and the MARC chip. The gain dispersion between the different channels is about 3% – essential for achieving the desired invariant mass resolution. The electronics are completed by the CROCUS system, which was specifically designed and developed to perform the read out of the tracking chambers.

The alignment of the tracking chambers is crucial for achieving the required invariant mass resolution, so there will be a strict procedure to follow when ALICE is running. There will be dedicated runs without magnetic field for aligning the chambers with straight muon tracks. Then, during standard datataking, a dedicated monitoring system will record any displacement with respect to the initial geometry, which can occur for a variety of reasons, including the switching on of the magnet. The geometry-monitoring system consists of 460 optical sensors installed on the tracking chambers. It projects the image of an object onto a CCD sensor and the analysis of the recorded image then provides a measurement of the displacement. The aim is to monitor the position of all of the tracking chambers with a precision better than 40 μm.

Trigger chambers beyond the muon filter form the final important component of the muon system. The role of the trigger detectors is to select dimuons produced (e.g. by J/Ψ or Υ decays) from the background of low-p_t muons produced by the decays of pions and kaons. The selection is made on the p_t of each individual muon, yielding a dimuon trigger signal when there are at least two tracks above a predefined p_t. This p_t selection needs a position-sensitive trigger detector with a spatial resolution of better than 1 cm – a requirement that is fulfilled by resistive plate chambers (RPCs). These detectors will be operated in streamer mode during heavy-ion runs.

The trigger system consists of four RPC planes, with a total active area of about 150 m², arranged in two stations, 1 m apart, behind the muon filter. The RPC electrodes are made of low-resistivity Bakelite (about 3 × 10⁹ Ωcm) so as to achieve the rate capability in the heavy-ion collisions. They are coated with linseed oil to improve the smoothness of the electrode surface. Extensive tests have shown that the RPCs will be able to tolerate several years of data taking in ALICE with heavy-ion beams.

The front-end electronics for the trigger detectors is based on the ADULT integrated circuit, also developed within the ALICE collaboration. Although designed for optimizing the time resolution when the RPCs operate in streamer mode, the circuit also allows the chambers to operate in “avalanche mode” during the long proton–proton runs that will occur at the LHC. The signals from the trigger detectors pass to the trigger electronics, which performs the p_t selection on each muon. If the muon trigger is fired, a dedicated electronics card called the DARC allows the transfer of the trigger data to the ALICE data acquisition. Thanks to a short decision time of about 700 ns with these electronics, the dimuon trigger forms part of the level-0 trigger for ALICE. A high-level trigger system – based on the analysis by a PC farm of the final two tracking stations – further refines the selection of good events.

The collaboration has developed a detailed simulation to evaluate the performance of the muon spectrometer for the vector meson and heavy-flavour studies, using as input current knowledge about the different processes that contribute to dimuon production at LHC energies. Figure 5 shows an example of such studies, in this case the invariant mass distributions in the regions of the J/Ψ and Υ mass for lead–lead collisions in ALICE. This demonstrates the spectrometer’s capability to detect these resonances against various sources of background.

For the past few years, components built for the spectrometer have arrived at CERN from many different collaborating institutes and suppliers. The two coils for the spectrometer dipole magnet arrived at CERN in September 2003 and the complete magnet was installed in its final position underground in summer 2005. A year later, dimuon trigger and tracking chambers were the first detectors to be installed in ALICE’s underground cavern in July 2006. Since then installation and commissioning have maintained a good pace, and the dimuon spectrometer should be ready for the first global tests of ALICE at the end of the year.

• The design and construction of the ALICE muon spectrometer have been made possible through the joint efforts of many institutions in different countries: CEA/DAPNIA Saclay, IPN Lyon, IPN Orsay, LPC Clermont-Ferrand, LSPC Grenoble and Subatech Nantes (France); CERN Geneva (Switzerland); INFN/University of Cagliari, INFN/University of Torino, University of Piemonte Orientale and INFN Alessandria (Italy); JINR Dubna, PNPI Gatchina and VNIIEF Sarov (Russia); KIP Heidelberg (Germany); Muslim University of Aligarh and SAHA Institute Kolkata (India); University of Cape Town (South Africa); and Yer-Phi Yerevan (Armenia).