Preprint typeset in JINST style - HYPER VERSION
The ATLAS experiment at the CERN Large HadronCollider
The ATLAS Collaboration
ABSTRACT: This paper describes the ATLAS experiment as installed in its experimental cavern atpoint 1 at CERN.
KEYWORDS: ATLAS, LHC, CERN.5
Contents
1. Overview of the ATLAS detector 91.1 Introduction 9
1.1.1 Physics motivation 910
1.1.2 Overall concept of ATLAS 101.1.3 Tracking 131.1.4 Calorimeters 14
1.1.4.1 The LAr electromagnetic calorimeter. 151.1.4.2 The hadronic calorimeters. 1715
1.1.5 Muon system 181.1.5.1 The toroid magnets. 191.1.5.2 Muon chamber types. 191.1.5.3 Muon chamber alignment and B-field monitoring. 20
1.1.6 Readout and control systems, trigger and data acquisition system 2120
1.1.6.1 Readout. 211.1.6.2 The ATLAS trigger and data acquisition system. 221.1.6.3 DCS. 23
1.1.7 Experimental conditions and interface to the LHC machine 231.1.7.1 Radiation levels. 2425
1.1.7.2 Shielding elements. 241.1.7.3 LHC Machine interface. 241.1.7.4 Beampipe. 241.1.7.5 Forward detectors. 25
1.1.8 Outline of the paper 2530
2. Magnet system and magnetic field 262.1 Magnet system 26
2.1.1 Solenoid 272.1.2 Barrel toroid 282.1.3 End-cap toroids 2935
2.1.4 Magnet services 302.1.4.1 Cryogenics. 302.1.4.2 Vacuum system 322.1.4.3 Electrical circuit 322.1.4.4 Magnet controls. 3240
2.2 Magnetic field determination 332.2.1 Performance specifications and measurement concepts 342.2.2 B-field modelling 352.2.3 Magnetic field instrumentation and reconstruction 37
2.2.3.1 B-field sensors 3745
– 1 –
2.2.3.2 B-field reconstruction 372.2.4 Solenoid-mapping measurements 38
2.2.4.1 Mapping campaign 382.2.4.2 Mapper geometry, survey and auto-calibration 382.2.4.3 Map fitting 3950
2.2.5 Experimental validation of the field map in the muon spectrometer 402.2.6 Towards an overall field map for ATLAS data-taking 41
3. Background radiation and shielding 423.1 Introduction 423.2 Description of the shielding 4355
3.3 Calculation of particle fluences and absorbed doses 453.3.1 The inner detector and calorimeter regions 463.3.2 The muon spectrometer region 48
3.4 Background monitors 503.4.1 Monitors in the inner detector 5060
3.4.2 Monitors in the muon spectrometer 523.5 Activation 53
4. Inner Detector 554.1 Introduction 554.2 Inner detector sensors 5865
4.2.1 The Pixel and SCT detector sensors 584.2.2 TRT straw tubes 59
4.3 Inner detector modules 614.3.1 Pixel modules and staves 614.3.2 SCT modules 6370
4.3.3 TRT modules 664.4 Readout of the inner detector 69
4.4.1 Front-end electronics 694.4.1.1 Pixel front-end electronics 694.4.1.2 SCT front-end electronics 7175
4.4.1.3 TRT front-end electronics 734.4.2 Data transmission and power-supply services and routing 744.4.3 Data transmission to the readout drivers 74
4.4.3.1 Pixel and SCT readout 744.4.3.2 TRT readout 7780
4.4.3.3 Pixel, SCT and TRT readout drivers (ROD’s) 774.5 Electronics and detector power supplies and services 784.6 Structure and mechanical integration of the inner detector 79
4.6.1 Pixel structure and integration 804.6.2 SCT structure and integration 8485
4.6.3 TRT structure and integration 87
– 2 –
4.6.4 Integration of the SCT and TRT 894.7 Inner detector environment and general services 89
4.7.1 Inner detector DCS, safety and interlocks 904.8 Material distribution of the inner detector 9090
4.9 Inner detector performance 914.9.1 Electrical performance of the integrated detector 914.9.2 Cooling performance 964.9.3 Cosmic-ray track reconstruction 974.9.4 Inner detector installation and performance in situ 9895
5. Calorimetry 1005.1 General description, cryostats and feed-throughs 100
5.1.1 General Description 1005.1.2 Cryostats and associated feed-throughs 101
5.1.2.1 Cryostat description 101100
5.1.2.2 Signal feed-throughs 1025.1.2.3 High-voltage feed-throughs 104
5.2 Electromagnetic calorimetry 1045.2.1 Accordion geometry 1045.2.2 Barrel geometry 104105
5.2.3 End-cap geometry 1075.2.4 Calorimeter alignment 1085.2.5 High-voltage distribution 1095.2.6 On-detector cold electronics 1105.2.7 Quality control tests 111110
5.3 Hadronic calorimeters 1115.3.1 Tile calorimeter 111
5.3.1.1 Overview 1115.3.1.2 Mechanical structure 1125.3.1.3 Optical components 114115
5.3.1.4 Instrumentation with scintillator and fibres 1155.3.2 Hadronic end-cap calorimeters (HEC) 116
5.3.2.1 Overview 1165.3.2.2 Wheel assembly and installation 119
5.3.3 Forward calorimeters 119120
5.3.3.1 Overview 1195.3.3.2 Module description 1215.3.3.3 Integration into the cryostat 1225.3.3.4 Alignment survey results 1225.3.3.5 Tests during installation/integration QC/QA results 123125
5.4 Gap Instrumentation 1235.5 Calorimeter readout electronics, calibration and services 124
5.5.1 Electronics readout 124
– 3 –
5.5.1.1 Overview 1245.5.1.2 LAr calorimeter front-end electronics 124130
5.5.1.3 Tile calorimeter front-end electronics 1265.5.1.4 Back-end (BE) electronics 128
5.5.2 Calorimeter calibration systems 1315.5.2.1 LAr calibration 1315.5.2.2 Tile calibration 132135
5.5.3 Calorimeter power supplies and services 1335.5.3.1 Low-voltage power supplies 1335.5.3.2 High-voltage power supplies 1335.5.3.3 Other services 1345.5.3.4 Detector control systems 135140
5.6 Beam test results 1365.6.1 Electromagnetic module performance 1365.6.2 Hadronic endcap performance 1385.6.3 FCal stand alone beam test performance 1415.6.4 Tile calorimeter performance 141145
6. Muon system 1446.1 Overview 1446.2 Precision chambers 1456.3 Monitored drift tube chambers 147
6.3.1 Structure and function of the drift tube 147150
6.3.2 Mechanical structure of MDT chambers 1486.3.3 Signal path and readout electronics 1516.3.4 The MDT alignment system 1536.3.5 Performance of the MDT chambers 155
6.4 The Cathode strip chambers 156155
6.4.1 Layout of the CSC system 1566.4.2 Spatial and time resolution 1576.4.3 Mechanical design 1596.4.4 Readout electronics 1616.4.5 Performance of the CSC 162160
6.5 Trigger chambers 1636.6 Resistive Plate Chambers 165
6.6.1 Principle of operation 1666.6.2 Mechanical structure 1676.6.3 Signal path and readout electronics 168165
6.7 Thin gap chambers 1706.7.1 Introduction 1706.7.2 Principle of operation 1716.7.3 Mechanical structure 1726.7.4 Signal path and readout electronics 173170
– 4 –
6.8 Commonalities in the muon system 1746.8.1 The gas supplies in the muon system 1746.8.2 Electronics services and power consumption 175
7. Trigger, Data Acquisition, and Controls 1777.1 Introduction to event selection and data acquisition 177175
7.2 The L1 trigger 1787.2.1 Calorimeter trigger 180
7.2.1.1 The analogue front-end. 1817.2.1.2 The pre-processor 1827.2.1.3 The cluster and jet/energy-sum processors 184180
7.2.1.4 The cluster processor module 1847.2.1.5 The Jet/Energy module (JEM) 1857.2.1.6 The common merger module (CMM) 1877.2.1.7 The processor back-plane 1877.2.1.8 The Read Out Driver (ROD) 188185
7.2.2 Muon trigger 1887.2.2.1 Muon barrel trigger 1897.2.2.2 Muon end-cap trigger 1917.2.2.3 Muon to Central Trigger Processor Interface (MUCTPI) 194
7.2.3 Central trigger processor 195190
7.2.3.1 Functional overview 1957.2.3.2 System implementation and latency 196
7.3 HLT/DAQ 1987.3.1 Overview 1987.3.2 HLT event selection 199195
7.3.3 Control 2007.3.4 Configuration 2017.3.5 Monitoring and information distribution 2027.3.6 Readout system 2027.3.7 L2 trigger 204200
7.3.8 Event-building 2057.3.9 Event filter 2067.3.10 Event output 206
7.4 Implementation and capabilities of the DAQ/HLT 2077.5 Detector control system 210205
8. Integration and installation 2138.1 Introduction 2138.2 Organisational issues 213
8.2.1 Organisational processes 2148.2.2 Organisational computing tools 214210
8.3 Mechanical integration 215
– 5 –
8.3.1 Envelopes (individual, global, dynamic) 2158.3.2 Survey and placement strategy 216
8.3.2.1 Survey reference grid in the cavern 2168.3.2.2 Stability measurements of the floor and the bed-plates 216215
8.3.2.3 Placement of ATLAS sub-systems 2188.3.2.4 Monitoring of the placement of the experiment 221
8.4 Infrastructure and detector services 2218.4.1 Civil engineering 2228.4.2 Electrical power distribution 223220
8.4.3 Air-conditioning and cooling systems 2248.4.4 Gas distribution 2258.4.5 Cryogenic systems 2268.4.6 Racks and cables 2288.4.7 Drag-chains and mobile services 229225
8.5 Built-in safety features 2308.6 Support and access structures 231
8.6.1 Feet and rail system 2318.6.2 Trucks 2338.6.3 Surrounding structures (HS and HO) 233230
8.6.4 Muon barrel access structures 2348.6.5 Big wheels 234
8.7 Detector Installation 2358.7.1 Phase 1: infrastructure in the main cavern, feet and rails 2358.7.2 Phase 2: barrel calorimetry and barrel toroid 236235
8.7.3 Phase 3: end-cap calorimeters and muon barrel chambers 2368.7.4 Phase 4: muon big wheels, inner detector and completion of muon barrel 2378.7.5 Phase 5: end-cap toroid magnets and muon small wheels 2388.7.6 Phase 6: beam-pipe and forward shielding 238
8.8 Access and detector opening 239240
8.8.1 Access scenarios 2398.8.2 Movement system 240
8.9 Beam-pipe 2408.10 Interface to the LHC machine 242
9. Expected performance of the ATLAS detector - 55 pages 245245
9.1 Introduction, 1 page 2459.2 Tracking (17 pages) 247
9.2.1 Charged particles in the inner detector (11 pages) 2479.2.1.1 Tracking performance for single particles and jets 2479.2.1.2 Alignment of the inner detector 247250
9.2.1.3 Vertexing performance 2499.2.1.4 Particle identification, reconstruction of electrons and photon
conversions 249
– 6 –
9.2.2 Muon measurements, 6 pages 2539.2.2.1 Introduction 253255
9.2.2.2 Calibration and alignment 2539.2.2.3 Muon performance 2539.2.2.4 Muon performance in situ 254
9.3 Electron and photon identification and measurements (8 pages) 2579.3.1 Calibration of the electromagnetic calorimeter 257260
9.3.2 Electron and photon reconstruction and identification 2589.3.3 Strategies for validation of performance in situ 258
9.4 Jet reconstruction, 8 pages 2599.4.1 Jet clustering and reconstruction 2599.4.2 Particle jets 259265
9.4.3 Calorimeter jets 2599.4.4 Jet calibration 259
9.4.4.1 Calibration at the level of particle jets 2609.4.4.2 Calibration at the level of parton jets 260
9.4.5 Jet reconstruction performance 260270
9.4.5.1 Jet reconstruction efficiency 2609.4.5.2 Low-pT jet vetoing 2629.4.5.3 Forward jet tagging 262
9.5 Missing transverse energy, 4 pages 2669.5.1 Reconstruction and calibration ofEmiss
T 266275
9.5.1.1 Contribution from calorimetry and noise suppression 2669.5.1.2 Contribution from cryostats 2669.5.1.3 Contributions from muons 2669.5.1.4 Refined calibration ofEmiss
T 2669.5.2 Evaluation ofEmiss
T performance 266280
9.5.3 Measurement ofEmissT direction 266
9.5.4 Use ofEmissT for mass reconstruction 266
9.5.5 FakeEmissT 267
9.5.5.1 Sources of fakeEmissT 267
9.5.5.2 FakeEmissT from muons 267285
9.5.5.3 FakeEmissT from Calorimeter 267
9.6 Hadronicτ -decays (5 pages) 2709.6.1 Reconstruction and identification of hadronicτ -decays 270
9.6.1.1 Efficiency and quality of track reconstruction 2709.6.2 π0 subclusters in single-prong decays 270290
9.6.3 Identification of hadronicτ -decays and rejection of jets 2719.6.4 Initial physics withτ leptons 271
9.7 Flavour tagging (4 pages) 2749.7.1 b-tagging using tracking and vertexing 274
9.7.1.1 Impact parameter of tracks 274295
9.7.1.2 Secondary vertices 274
– 7 –
9.7.1.3 Likelihood-ratio taggers 2759.7.1.4 Labelling and purification 2759.7.1.5 Expected performance 276
9.7.2 Soft lepton tagging 276300
9.7.3 Impact of alignment 2769.7.4 Estimation of performances from data 277
9.7.4.1 Measuring b-tagging efficiency 2779.7.4.2 Estimating mistag rates 278
9.7.5 Calibration 278305
9.7.5.1 JetProb calibration 2789.7.5.2 Calibration of likelihood ratio-based taggers 278
9.8 Trigger performance (8 pages) 2799.8.1 Overview 2799.8.2 Selection strategy 279310
9.8.3 Trigger menus 2799.8.4 Examples of trigger performance 2799.8.5 Trigger commissioning 2819.8.6 Trigger efficiency from data 281
9.9 Overall ATLAS energy scale and alignment (4 pages) 283315
10. Outlook and conclusions 28410.1 From end 2007 until start-up (1 page) 28410.2 From start-up till operation of complete detector at design luminosity(1/2 page) 28510.3 Forward detector systems (2-3 pages) 285
10.3.1 The LUCID detector 285320
10.3.2 The ZDC detector 28610.3.3 The ALFA detector 28710.3.4 B-layer replacement (1 page) 28910.3.5 Super-LHC upgrade (1-1.5 page) 28910.3.6 Concluding remarks 290325
11. Bibliography: main references and complete list of back-up papers 291
– 8 –
1. Overview of the ATLAS detector
1.1 Introduction
Inside the LHC, bunches of 1011 protons will collide 40 million times per second to provide 14 TeV330
proton collisions at the design luminosity of 1034 cm−2s−1. With an inelastic proton proton cross-section of 100 mb, this gives approximately 25 events per bunch-crossing, or a total rate of 109
inelastic events/sec. This means around 1000 particles will emerge from the collision points every25 ns within|η | less than 2.5.
Two general purpose detectors have been constructed for theLHC: ATLAS and CMS. This335
paper describes the ATLAS detector system. The ATLAS detector is now installed and is preparingfor first collisions at the LHC.
1.1.1 Physics motivation
The formidable LHC luminosity and resulting interaction rate are needed because of the small crosssections expected for many of the LHC benchmark processes such as Higgs-boson production and340
decay. Some of the processes include new physics scenarios such as supersymmetry and extradimensions. A serious experimental difficulty at LHC is therefore that every candidate event fornew physics will on average be accompanied by 25 inelastic events occurring simultaneously in thedetector.
The nature of proton proton collisions imposes another difficulty. QCD jet production cross-345
sections are very large compared to the rare processes mentioned above and therefore one oftenhas to study final states involving leptons, photons, or characteristic experimental signatures suchas missing transverse energy (Emiss
T ) or secondary vertices to strongly reduce QCD backgroundprocesses. Looking for such final states among already rarely produced particles imposes furtherdemands on the luminosity needed, and on the detectors particle identification capabilities.350
More specific requirements for the LHC detector systems [1] have been defined using a set ofbenchmark processes covering most new phenomena one can hope to observe at the TeV scale:
Higgs: The search for the Standard Model Higgs Boson has been used asa benchmark to es-tablish the performance of important subsystems of the LHC detectors. It is a particularly importantbenchmark since there is a wide range of decay modes depending on the mass of the Higgs boson.355
At low masses (MH < 2MZ), the natural width is only a few MeV and the observed width will bedefined by the instrumental resolution. The predominant decay mode into hadrons is difficult dueto QCD backgrounds, and the two photon decay channel will be important. Other channels, asfor example associated production of H as ttH, WH, and ZH, with H→bb, are also important, us-ing a lepton from the decay of the accompanying particle for triggering and background rejection.360
Above 130 GeV, Higgs-decay into two Z bosons, where both Z’s decay to charged leptons (oneZ is virtual when theMH is below the ZZ threshold) will be the most interesting channel. Aboveapproximately 600 GeV, WW and ZZ decays into jets or involving neutrinos are needed to extracta signal. Tagging of forward jets in the region 2< |η | < 5 from the WW or ZZ fusion produc-tion mechanism are important forMH masses close to 1 TeV. Beyond the Standard Model Higgs365
searches, such particles as the A, H± of the Minimal Supersymmetric extension of the StandardModel (MSSM) require sensitivity to processes involving taus and good b-tagging performance.
– 9 –
Supersymmetry: The decays of supersymmetric particles, such as squarks andgluinos, in-volve cascades that, if R-parity is conserved, always contain a lightest stable SUSY particle (LSP).As the LSP interacts very weakly with the detector, the experiments will measure a significantEmiss
T370
in the final state. The rest of the cascade results in a number of leptons and jets. In GMSB schemeswith the LSP decaying into a photon and gravitino, an increased number of hard isolated photonsis expected.
Extra dimensions: Several new models propose the existence of extra dimensions leading toa characteristic energy scale of quantum gravity in the TeV region [2]. In terms of experimental375
signatures this could lead to the emission of gravitons thatescape into extra dimensions and there-fore generateEmiss
T , or Regge-like excitations that manifest themselves as Z-like resonances with∼ TeV separations in mass. Other experimental signatures could be anomalous high mass dijet pro-duction and mini black hole production with spectacular decays involving democratic productionof fundamental particles such as jets, leptons, photons, neutrinos, W’s and Z’s, etc...380
Standard Model tests and beyond the SM objects and effects:The LHC will also enablestudies of QCD, electroweak, and flavour physics. New, heavy, gauge bosons W’ andZ′ could beaccessible to the LHC for masses up to 5-6 TeV. To study their leptonic decays, high-resolutionlepton measurements and charge identification are therefore needed in thepT range of a few TeV.Another class of signatures of new physics can be provided byvery highpT jet measurements. As385
a benchmark process, quark compositeness has been used where the signature would be observabledeviations in the jet cross-sections from the QCD expectations. Top quarks will be produced at theLHC with a rate measured in Hz, providing the opportunity to test the couplings and spin of the topquark under the condition that good b-jet identification is possible. Searches for flavour changingneutral currents, lepton flavour violation throughτ → 3µ or τ → µγ, measurements ofB0
s → µµ ,390
measurements of triple- and quartic gauge couplings. can also open a window onto new physics.
1.1.2 Overall concept of ATLAS
These physics goals can be turned into a set of requirements for the LHC detectors.
• First of all, due to the experimental conditions at the LHC, the detectors need fast, radiationhard electronics and sensor elements. In addition, a very high granularity is needed to be395
able to handle the particle fluxes and to reduce the influence of overlapping events.
• Large acceptance in pseudorapidity (η ) with almost full azimuthal (φ) angle coverage isrequired. Here,η = − ln tan(θ/2), whereθ is the polar angle, measured from the beamdirection.
• Good charged particle momentum resolution and reconstruction efficiency in the inner tracker400
are essential. For efficient triggering and offline tagging of τ ’s and b-jets, pixel detectorsclose to the interaction region are required to observe secondary vertices.
• Very good electromagnetic (EM) calorimetry for electron and photon identification and mea-surements, complemented by full-coverage hadronic calorimetry for accurate jet and miss-ing transverse energy (Emiss
T ) measurements are considered important requirements, as these405
measurements form the basis of many of the studies mentionedabove.
– 10 –
• Good stand alone muon identification and momentum resolution over a wide range of mo-menta and the ability to determine unambiguously the chargeof high pT muons are funda-mental requirements.
• Triggering on low transverse momenta objects is important to maintain high kinematic ef-410
ficiency with sufficient background rejection to realise an acceptable trigger rate for mostphysics processes of interest at the LHC.
The overall detector layout is shown in Fig. 1 and the main performance goals are listed inTable 1. The magnet configuration is based on an inner, thin superconducting solenoid surroundingthe inner detector cavity and large superconducting air-core toroids consisting of independent coils415
arranged with an eight-fold symmetry outside the calorimeters.
Figure 1. The ATLAS detector. The diameter and barrel toroid lengths are both 26 m and the overall lengthis 46 m. The overall weight of the ATLAS detector is approximately 7000 tons.
The Inner Detector (ID) is contained within a cylinder of length 5.5 m and a radius of 1.15 m,in a solenoidal magnetic field of 2 T. Pattern recognition, momentum and vertex measurements, andelectron identification are achieved with a combination of discrete high-resolution semiconductorpixel and strip detectors in the inner part of the tracking volume, and continuous straw-tube tracking420
detectors with transition radiation capability in its outer part.High granularity liquid-argon (LAr) electromagnetic sampling calorimetry, with excellent per-
formance in terms of energy and position resolution, cover the pseudorapidity range|η | < 3.2.In the endcaps, the LAr technology is also used for the hadronic calorimeters, which share thecryostats with the electromagnetic endcap calorimeters. The same cryostats also house the spe-425
cial LAr forward calorimeters which extend the pseudorapidity coverage to|η | = 4.9. The LAr
– 11 –
calorimetry is contained in a cylinder with an outer radius of 2.25 m and extends longitudinally to± 6.65 m along the beam axis. The bulk of the hadronic calorimetryis provided by a scintillator-tilecalorimeter, which is separated into a large barrel and two smaller extended barrel cylinders, oneon each side of the barrel. The outer radius of the scintillator-tile sampling calorimeter is 4.25 m430
and its half length is 6.10 m.The calorimeter is surrounded by the muon spectrometer. Theair-core toroid system, with a
long barrel and two inserted end-cap magnets, generates a large magnetic field volume with strongbending power within a light and open structure. Multiple-scattering effects are thereby minimised,and excellent muon momentum resolution is achieved with three stations of high precision tracking435
chambers. The muon instrumentation includes, as a key component, trigger chambers with veryfast time response. The muon spectrometer defines the overall dimensions of the ATLAS detector.The outer chambers of the barrel are at a radius of about 11 m. The half-length of the barrel toroidcoils is 12.5 m, and the third layer of the forward muon chambers, mounted on the cavern wall, islocated about 22 m from the interaction point.440
The trigger systems of the ATLAS experiment have three distinct levels. The first level, basedon custom-built processors, uses a subset of the total detector information to make a decision in2.5 µs on whether to continue the processing of an event or not, reducing the data rate to around100 kHz. During this time the event data are buffered in the front-end electronics circuits onthe detector. Higher levels, using a network of several thousand processors, fast switches, and445
networks, access gradually more information and run algorithms that resemble offline data analysisto achieve the final reduction.
Table 1. General detector performance goals
Detector component Minimally required resolution, η coveragecharacteristics Measurement Trigger
Inner detector 5 x 10−4 pT ⊕1% ±2.5Enhanced electron identification ±2.5τ - and b-tagging ±2.5Secondary vertex detection at
initial luminosities ±2.5
EM calorimetry 10%/√
E⊕0.7% ±3.2 ±2.5
Presampler detector Enhancedγ–π0 andγ–jet ±2.4separation, direction measure-ments
Jet andEmissT
Calorimetrybarrel and end-cap 50%/
√E⊕3% ±3.2 ±3.2
forward 100%/√
E⊕10% 3.1 < |η | < 5.2 3.1 < |η | < 4.9
Muon detection 10% atpT = 1 TeV ±3 ±2.2in stand-alone modeat highest luminosity
– 12 –
1.1.3 Tracking
The layout of the Inner Detector (ID) is shown in Fig. 2. It combines high-resolution detectorsat the inner radii with continuous tracking elements at the outer radii, all contained in the CentralSolenoid (CS) which provides a nominal magnetic field of 2 T. The CS extends over a length of450
5.3 m and has a bore of 2.5 m. The position of the CS in front of the EM calorimeter demandsa careful minimisation of the material in order to achieve the desired calorimeter performance.As a consequence, the CS and the LAr calorimeter share one common vacuum vessel, therebyeliminating two vacuum walls.
Figure 2. The ATLAS inner detector.
The momentum and vertex resolution requirements from physics call for high-precision mea-455
surements to be made with fine-granularity detectors, giventhe very large track density expected atthe LHC. Semiconductor tracking detectors, using pixel andsilicon microstrip (SCT) technologiesoffer these features. The highest granularity is achieved around the vertex region using silicon pixeldetectors, followed by silicon strip detectors. The total number of precision layers must be limitedbecause of material introduced and cost. Typically, three pixel layers and eight strip layers (four460
space points) are crossed by each track. A large number of tracking points (typically 36 per track)is provided by the straw tube tracker (TRT), which provides continuous track-following with muchless material per point and at lower cost. The combination ofthe two techniques gives very robustpattern recognition and high precision in bothφ and z coordinates. The straw hits at the outer radiuscontribute significantly to the momentum measurement, since the lower precision per point com-465
pared to the silicon is compensated by the large number of measurements and the higher averageradius. The electron identification capabilities are enhanced by the detection of transition-radiationphotons in the xenon-based gas mixture of the straw tubes.
– 13 –
The ID cavity sits within a volume of radius 115 cm and total length 5.5 m. Mechanically,the ID consists of four units, the first three consisting of integrated SCT/TRT detectors at radii470
between 23 and 108 cm and two identical endcaps extending between 85< |z| <273 cm, and thelast comprising the complete pixel detector and its proximity services at radii between 5 and 23 cm.The precision tracking detectors (pixels and SCT) thus cover fiducially |η |< 2.5, whereas the TRTonly covers|η | < 2.0, because its outermost end-cap wheels have been staged.
In the barrel region, the high-precision detector layers are arranged on concentric cylinders475
around the beam axis, while the end-cap detectors are mounted on disks perpendicular to thebeam axis. The pixel layers are segmented in Rφ and z, while the SCT detector uses small angle(40 mrad) stereo strips to measure both coordinates, with one set of strips in each layer measuringφ. The barrel TRT straws are parallel to the beam direction. All end-cap tracking elements arelocated in planes perpendicular to the beam axis. The strip detectors have one set of strips running480
radially and a set of stereo strips at an angle of 40 mrad. The continuous tracking consists of radialstraws arranged into wheels.
The basic layout parameters and cell sizes are summarised inTable 2. The layout providesfull tracking coverage over|η | ≤ 2.5, including impact parameter measurements and vertexing forheavy-flavour andτ tagging. The secondary vertex measurement performance is enhanced by the485
innermost layer of pixels, at a radius of about 5 cm, as close as practical to the beam pipe. Thelifetime of such a detector will be limited by radiation damage and may need replacement after afew years, the exact time depending on the luminosity profile.
Item Radial extension (mm) Length (mm)
Overall ID envelope 36< R < 1150 0 < |z| < 2727 (or 3400)
Pixel Beam pipe R=36Overall envelope 36< R < 230 0 < |z| < 3300
3 cylindrical layers Sensitive barrel 50.5< R < 122.5 0 < |z| < 4012 x 3 disks Sensitive end-cap 89< R < 150 495< |z| < 650
SCT Overall envelope 230< R < 559 (barrel) 0 < |z| < 746230< R < 635 (end-cap) 847.5< |z| < 2727
4 cylindrical layers Sensitive barrel 299< R < 514 0 < |z| < 7462 x 9 disks Sensitive end-cap 270< R < 560 847.5< |z| < 2727
TRT Overall envelope 559< R < 1150 (barrel) 0 < |z| < 715559< R < 1150 (end-cap) 847.5< |z| < 2727
73 straw planes Sensitive barrel 559< R < 1080 0 < |z| < 715160 straw planes Sensitive end-cap635< R < 999 847.5< |z| < 2727
Table 2. Key geometric envelopes of the inner detector and its sub-detectors.
1.1.4 Calorimeters
A view of the ATLAS calorimeters is presented in Fig. 3. The calorimetry consists of an electro-490
magnetic calorimeter covering the pseudorapidity region|η | < 3.2, a hadronic barrel calorimetercovering|η |< 1.7, hadronic end-cap calorimeters covering 1.5< |η |< 3.2, and forward calorime-
– 14 –
ters covering 3.1 < |η | < 4.9. Over the pseudorapidity range|η | < 1.8, the EM calorimeter ispreceded by a presampler detector.
Figure 3. The ATLAS calorimeter system.
The EM calorimeter is a lead/liquid-argon detector with accordion geometry. The barrel495
EM calorimeter is contained in a barrel cryostat, which surrounds the inner detector cavity. Thehadronic barrel calorimeter is a cylinder divided into three sections: the central barrel and twoidentical extended barrels. It is based on a sampling technique with plastic scintillator plates (tiles)embedded in an iron absorber. Two end-cap cryostats house the LAr end-cap EM and hadroniccalorimeters, as well as the integrated forward calorimeter. The hadronic end-cap calorimeter is500
a copper LAr detector with parallel-plate geometry, and theforward calorimeter, a dense LArcalorimeter with rod-shaped electrodes in a copper and tungsten matrix. The barrel and extendedbarrel tile calorimeters support the LAr cryostats and alsoact as the main solenoid flux return.
The pseudorapidity coverage, granularity, and longitudinal segmentation of the calorimetersare summarised in Table 3.505
1.1.4.1 The LAr electromagnetic calorimeter. The EM calorimeter is divided into a barrel part(|η |< 1.475) and two endcaps (1.375< |η |< 3.2). The barrel calorimeter consists of two identicalhalf-barrels, separated by a small gap (6 mm) at z=0. Each end-cap calorimeter is mechanicallydivided into two coaxial wheels: an outer wheel covering theregion 1.375< |η | < 2.5, and aninner wheel covering the region 2.5 < |η | < 3.2. The EM calorimeter is a lead LAr detector with510
accordion-shaped kapton electrodes and lead absorber plates over its full coverage. The accordiongeometry provides completeφ symmetry without azimuthal cracks. The lead thickness in theabsorber plates has been optimised as a function ofη in terms of EM calorimeter performance inenergy resolution. The total thickness of the EM calorimeter is > 24 radiation lengths (X0) in thebarrel and> 26 X0 in the endcaps. Over the region devoted to precision physics(|η | < 2.5), the515
EM calorimeter is segmented into three longitudinal sections. For|η | > 2.5, i.e. for the end-cap
– 15 –
Barrel End-cap
EM calorimeterNumber of layers and|η | coverage
Presampler 1 |η | < 1.52 1 1.5 < |η | < 1.8Calorimeter 3 |η | < 1.35 2 1.375< |η | < 1.5
2 1.35< |η | < 1.475 3 1.5 < |η | < 2.52 2.5 < |η | < 3.2
Granularity∆η ×∆φ versus|η |Presampler 0.025×0.1 |η | < 1.52 0.025×0.1 1.5 < |η | < 1.8
Calorimeter 1st layer 0.025/8×0.1 |η | < 1.4 0.050×0.1 1.375< |η | < 1.4250.025×0.025 1.4 < |η | < 1.475 0.025×0.1 1.425< |η | < 1.5
0.025/8×0.1 1.5 < |η | < 1.80.025/6×0.1 1.8 < |η | < 2.00.025/4×0.1 2.0 < |η | < 2.40.025×0.1 2.4 < |η | < 2.50.1×0.1 2.5 < |η | < 3.2
Calorimeter 2nd layer 0.025×0.025 |η | < 1.4 0.050×0.025 1.375< |η | < 1.4250.075×0.025 1.4 < |η | < 1.475 0.025×0.025 1.425< |η | < 2.5
0.1×0.1 2.5 < |η | < 3.2Calorimeter 3rd layer 0.050×0.025 |η | < 1.35 0.050×0.025 1.5 < |η | < 2.5
Number of readout channelsPresampler 7808 1536 (both sides)Calorimeter 101760 62208 (both sides)
LAr hadronic endcap (HEC)|η | coverage 1.5 < |η | < 3.2
Number of layers 4Granularity∆η ×∆φ 0.1×0.1 1.5 < |η | < 2.5
0.2×0.2 2.5 < |η | < 3.2Readout channels 5632 (both sides)
LAr forward calorimeter (FCal)|η | coverage 3.1 < |η | < 4.9
Number of layers 3Granularity∆x×∆y(cm) FCal1 3.0x2.6
FCal2 3.3x4.2FCal3 5.4x4.7
Readout channels 3524 (both sides)
Scintillator tile calorimeterBarrel Extended barrel
|η | coverage |η | < 1.0 0.8 < |η | < 1.7Number of layers 3 3
Granularity∆η ×∆φ 0.1x0.1 0.1x0.1Last layer 0.2x0.1 0.2x0.1
Readout channels 5760 4092 (both sides)
Table 3. Parameters of the ATLAS calorimeter system.
inner wheel, the calorimeter is segmented in two longitudinal sections and has a coarser lateralgranularity than for the rest of the acceptance. This is sufficient to satisfy the physics requirements(reconstruction of jets and measurement ofEmiss
T ).
– 16 –
The total material seen by an incident particle before the calorimeter front face is approxi-520
mately 2.3 X0 at η = 0, and increases with pseudorapidity in the barrel because of the particleangle. In region|η | < 1.8, a presampler is used to correct for the energy lost by electrons andphotons upstream of the calorimeter. The presampler consists of an active LAr layer of thickness1.1 cm (0.5 cm) in the barrel (end-cap) region.
At the transition between the barrel and the end-cap calorimeters, i.e. at the boundary between525
the two cryostats, the amount of material in front of the calorimeter reaches a localised maximumof about 7 X0. In this region, the presampler is complemented by a scintillator slab inserted in thecrack between the barrel and end-cap cryostats and coveringthe region 1.0< |η |< 1.6. The region1.37< |η |< 1.52 is not used for precision physics measurements involvingphotons because of thelarge amount of material situated in front of the EM calorimeter.530
1.1.4.2 The hadronic calorimeters. The ATLAS hadronic calorimeters cover the range|η | <
4.9 using different techniques suited for the widely varying requirements and radiation environmentover the largeη -range.
An important parameter in the design of the hadronic calorimeter is its thickness: it has toprovide good containment for hadronic showers and reduce toa minimum punch-through into the535
muon system. Close to 10λ of active calorimeter are adequate to provide good resolution for highenergy jets. The total thickness, including 1.5λ from the outer support, is 11 interaction lengths(λ ) at η = 0 and has been shown both by measurements and simulation to besufficient to reducethe punch-through well below the irreducible level of prompt or decay muons. Together with thelargeη -coverage, this will also guarantee a goodEmiss
T measurement, which is important for many540
physics signatures and in particular for SUSY particle searches.
The Tile calorimeters: The large hadronic barrel calorimeter is a sampling calorimeter usingiron as the absorber and scintillating tiles as the active material. The tiles are placed radiallyand staggered in depth. The structure is periodic along z. The tiles are 3 mm thick and the totalthickness of the iron plates in one period is 14 mm. Two sides of the scintillating tiles are read out by545
wavelength shifting (WLS) fibres into two separate photomultipliers (PMTs). The tile calorimeteris composed of one barrel and two extended barrels. Radially, the tile calorimeter extends froman inner radius of 2.28 m to an outer radius of 4.25 m. It is longitudinally segmented in threelayers, approximately 1.4, 4.0 and 1.8 interaction lengthsthick atη = 0. Azimuthally, the barreland extended barrels are divided into 64 modules. Inη , the readout cells, built by grouping fibres550
into PMTs, are pseudo-projective towards the interaction region. The total number of channels isapproximately 10,000. The calorimeter is placed behind theEM calorimeter. The total thicknessat the outer edge of the tile-instrumented region is 9.2λ at η = 0. The barrel cylinder covers theregionη < 1.0. A vertical gap of 68 cm provides space for cables from the ID, feed-throughs, andservice pipes for the EM calorimeter and the CS; it also houses front-end electronics for the EM555
calorimeter. The extended barrel covers the region 0.8 < |η | < 1.7. The energy lost in the inactivematerials in the gap between the tile barrel and extended barrel calorimeters is sampled by the Inter-TileCal (ITC) scintillators, which has the same segmentation as the rest of the tile calorimeter. It iscomposed of two radial sections attached on the face of the extended barrel. The ITC is extendedfurther inwards by a scintillator sheet, covering the innerpart of the extended barrel and extending560
to the region between the LAr barrel and end-cap cryostats over 1.0 < |η | < 1.6. This scintillator
– 17 –
samples the energy lost in the cryostat walls and other dead materials.
LAr hadronic end-cap calorimeters (HEC): Each HEC consists of two independent wheelsof outer radius 2.03 m. The upstream wheel is built out of 25 mmcopper plates, while the secondone further from the interaction point uses 50 mm plates. In both wheels, the 8.5 mm gap between565
consecutive copper plates is equipped with three parallel electrodes, splitting the gap into four driftspaces of about 1.8 mm. The readout electrode is the central one.
Each wheel is built out of 32 identical modules, assembled with fixtures at the periphery andat the central bore. Each wheel is divided into two longitudinal segments. The weight of the first(second) wheel is 67 (90) tons.570
To minimise the dip in the material density at the transitionbetween the end-cap and the for-ward calorimeter (around|η | = 3.1), the HEC calorimeter reaches|η | = 3.2, thereby overlappingthe forward calorimeter.
LAr forward calorimeter: The forward calorimeter (FCal) has a front face at about 4.5 mfrom the interaction point and is integrated into the endcapcryostat as this provides clear benefits575
in terms of uniformity of the calorimetric coverage as well as reduced radiation background levelsin the muon spectrometer. In order to minimise the amount of neutron albedo in the ID cavity,the front face of the FCal is recessed by about 1.2 m with respect to the EM calorimeter frontface. This severely limits longitudinal space for installing about 9.5 active interaction lengths, andtherefore calls for a high-density design, which also avoids energy leakage from the FCal to its580
neighbours. The FCal consists of three sections: the first one is made of copper, while the othertwo are made of tungsten. In each section the calorimeter consists of a metal matrix with regularlyspaced longitudinal channels filled with concentric rods and tubes. The rods are at positive highvoltage while the tubes and matrix are grounded. The LAr in the gap between is the sensitivemedium. This geometry allows for an excellent control of thegaps which are as small as 0.25 mm585
in the first section.
In terms of electronics and readout, four rods are ganged on the detector, and the signal iscarried out by polyimide insulated coaxial cables.
1.1.5 Muon system
The conceptual layout of the muon spectrometer is shown in Fig. 4. It is based on the magnetic590
deflection of muon tracks in the large superconducting air-core toroid magnets, instrumented withseparate trigger and high-precision tracking chambers. Over the range|η |< 1.4, magnetic bendingis provided by the large barrel toroid. For 1.6 < |η | < 2.7, muon tracks are bent by two smallerend-cap magnets inserted into both ends of the barrel toroid. Over 1.4< |η |< 1.6, usually referredto as the transition region, magnetic deflection is providedby a combination of barrel and end-cap595
fields. This magnet configuration provides a field that is mostly orthogonal to the muon trajecto-ries, while minimising the degradation of resolution due tomultiple scattering. The anticipatedhigh level of particle fluxes has had a major impact on the choice and design of the spectrome-ter instrumentation, affecting performance parameters such as rate capability, granularity, ageingproperties, and radiation hardness.600
In the barrel region, tracks are measured in chambers arranged in three cylindrical layers (sta-tions) around the beam axis; in the transition and end-cap regions, the chambers are installed ver-
– 18 –
Figure 4. The ATLAS muon system- this figure is a placeholder for a more appropriate one.
tically, also in three stations.
1.1.5.1 The toroid magnets. A system of three large air-core toroids generates the magneticfield for the muon spectrometer. The overall dimensions of the magnet system are 26 m in length605
and 20 m in diameter. The two end-cap toroids (ECT) are inserted in the barrel toroid (BT) ateach end and line up with the CS. They have a length of 5 m, an outer diameter of 10.7 m andan inner bore of 1.65 m. The performance in terms of bending power is characterised by the fieldintegral
∫
Bdl, where B is the azimuthal field component and the integralis taken on a straight linetrajectory between the inner and outer radius of the toroids. The BT provides 2 to 6 Tm and the610
ECT contributes with approximately 1 to 8 Tm in the 0.0-1.4 and 1.6-2.7 pseudorapidity rangesrespectively. The bending power is lower in the transition regions where the two magnets overlap(1.4 < |η | < 1.6).
Each of the three toroids consists of eight coils assembled radially and symmetrically aroundthe beam axis. The ECT coil system is rotated by 22.5 with respect to the BT coil system in615
order to provide radial overlap and to optimise the bending power in the interface regions of bothcoil systems. The BT coils are housed in individual cryostats taking up the forces between thecoils. The toroidal structure consists of eight cryostats and the linking elements between them,called voussoirs and struts, that provide mechanical stability. Services are brought to the coilsthrough a cryogenic ring linking the eight cryostats to a separate service cryostat, which provides620
connections to the power supply, helium refrigerator, vacuum systems, and control system. EachECT also consists of eight racetrack, double-pancake coilsin an aluminium alloy housing. Theyare cold-linked and assembled as a single cold mass, housed in one large cryostat. Therefore theinternal forces in the toroids are taken by the cold supporting structure between the coils, a differentdesign solution than in the BT.625
1.1.5.2 Muon chamber types. Over most of theη -range, a precision measurement of the trackcoordinates in the principal bending direction of the magnetic field is provided by Monitored Drift
– 19 –
Tubes (MDTs). At large pseudorapidities, Cathode Strip Chambers (CSCs) with higher granularityare used in the innermost plane over 2< |η |< 2.7, to withstand the demanding rate and backgroundconditions. Optical alignment systems have been designed to meet the stringent requirements on630
the mechanical accuracy and the survey of the precision chambers. The precision measurementof the muon tracks is made in the R-z projection, in a direction parallel to the bending directionof the magnetic field; the axial coordinate (z) is measured inthe barrel and the radial coordinate(R) in the transition and end-cap regions. The MDTs provide asingle-tube resolution of 80µmwhen operated at high gas pressure (3 bars) together with robust and reliable operation thanks to635
the mechanical isolation of each sense wire from its neighbours.The trigger system covers the pseudorapidity range|η | < 2.4. Resistive Plate Chambers
(RPCs) are used in the barrel and Thin Gap Chambers (TGCs) in the end-cap regions. The triggerchambers for the ATLAS muon spectrometer serve a threefold purpose. The trigger chambers pro-vide bunch-crossing identification, well definedpT thresholds, and the measurement of the second640
coordinate in a direction orthogonal to that provided by theprecision chambers.The main parameters of the muon chambers are given in Table 4.
Table 4. Main parameters of the ATLAS muon chambers.
Drift Tubes MDTs- Coverage |η | < 2.0- Number of chambers 1172- Number of channels 354 000- Function Precision measurement
Cathode Strip Chambers- Coverage 2.0 < |η | < 2.7- Number of chambers 32- Number of channels 31 000- Function Precision measurement
Resistive Plate Chambers- Coverage |η | < 1.05- Number of chambers 556- Number of channels 374 000- Function Triggering, second coordinate
Thin Gap Chambers- Coverage 1.05< |η | < 2.4- Number of chambers 3588- Number of channels 318 000- Function Triggering, second coordinate
1.1.5.3 Muon chamber alignment and B-field monitoring. While the intrinsic resolution ofthe precision chambers is in the 80µm range, the overall performance over the large areas in-volved, particularly at the highest momenta, will depend onthe alignment of the muon chambers645
– 20 –
with respect to each other and with respect to the overall detector. The high accuracy of the AT-LAS stand-alone muon measurement necessitates a high precision of 30µm on the alignment forthe positioning of chambers within a projective tower. The accuracy required for the relative posi-tioning of different towers to obtain adequate mass resolutions for multi-muon final states is in themillimetre range. This accuracy is easily achieved by the initial positioning and survey of cham-650
bers installation. The relative alignment of muon spectrometer, calorimeters, and inner detectorwill rely on high-momentum muon trajectories. In ATLAS, approximately 5000 optical alignmentsensors mounted with high precision, and around 1800 magnetic field sensors, are used to track themovements of the chambers, and to map and track the magnetic field. All alignment systems arebased on optical monitoring of deviations from straight lines. Owing to geometrical constraints,655
different schemes are used to monitor chamber positions in the barrel, in the end-caps, and thedeformations of large chambers (in-plane alignment). For magnetic field monitoring, the goal is todetermine the field throughout the whole volume to an accuracy of approximately 20 G, requiringthat all magnetic materials are also accurately mapped and described.
1.1.6 Readout and control systems, trigger and data acquisition system660
The ATLAS TDAQ readout system, the Timing, Trigger, and Control system, and the DetectorControl System are partitioned into subsystems, typicallyassociated with sub-detectors, whichhave the same logical components and building blocks.
It is not possible to process and select events within the 25 ns available between successivebunch crossings. Furthermore, the size of the detectors andthe underground caverns imposes a665
minimum transit time between the detector electronics and trigger electronics, and the first leveltrigger calculations themselves need to be sophisticated enough to identify clear signatures of newphysics among an overwhelming rate of less interesting events, reducing the rate by a factor almost100. During the transit and processing time - around 2.5µs, the detector data must be time stampedand held in the buffers of the radiation hard front end circuits.670
The proton-proton interaction rate at the design luminosity of 1034 cm−2s−1 is approximately1 GHz, while the event data recording, based on technology and resource limitations, is limited to200 Hz (for an event size of order 1.5 Mbyte). This requires anoverall rejection factor of 5×106
against minimum-bias processes while maintaining maximumefficiency for the new physics soughtby ATLAS. The ATLAS trigger systems has three distinct levels. Each trigger level refines the675
decisions made at the previous level and, where necessary, applies additional selection criteria. Thefirst level being based on custom built processors using a very limited amount of the total detectorinformation takes a decision in around 2.5µs reducing the rate to around 75 kHz. The higherlevels use a network of several thousand commercial processors and fast switches and networks toachieve the final reduction, using gradually more information and algorithms resembling more and680
more the offline data analysis. The total amount of data recorded will roughly be 1.5 MB/event ata final rate of up to 200 Hz.
1.1.6.1 Readout. The front-end electronics and the Readout Drivers (ROD) aredetector specificelements; however, they are formed of standard blocks and are subject to common requirements.
The front-end electronics subsystem includes different functional components:685
• The front-end analogue or analogue-digital processing;
– 21 –
• The L1 buffer in which information (analogue or digital) is stored and is retained for a timelong enough to accommodate the L1 trigger latency. This element had to be designed toaccommodate a L1 trigger latency of at least 2.5µs;
• The derandomising buffer in which the data corresponding toa L1 trigger accept (L1A) are690
stored before being sent to the following level. This element is necessary to accommodatethe maximum instantaneous L1 rate without introducing deadtime. It must contain enoughevents so that a maximum dead-time of 1% is introduced;
• Dedicated links or buses are used to transmit the front-end data stream to the next stage.Their speed, in conjunction with the derandomiser size, hasto be high enough in order to695
accommodate the 1% maximum dead-time requirement.
After accepting the event by the L1 trigger system, the data from the pipelines are transferredoff the detector to the readout driver. The ROD is the functional element of the front-end systemwhere one can reach a higher level of data concentration and multiplexing by gathering informa-tion from several front-end data streams. Elementary digitised signals are formatted as raw data700
prior to be transferred to the DAQ system. The ROD is a detector specific device but has to followsome general ATLAS rules, including the definition of the data format of the event, the error detec-tion/recovery mechanisms to be implemented, and the physical interface for the data transmissionto the DAQ system as well as constraints on the monitoring functionalities.
1.1.6.2 The ATLAS trigger and data acquisition system. The required event reduction factor705
is achieved by a three-level decision process. The first level (L1) is implemented in high-speedpipelined custom electronics operating at the LHC bunch-crossing frequency. The subsequent twolevels of reduction are provided by the High-Level Trigger (HLT), which is comprised of twodistinct parts: the Level-2 Trigger (L2) and the Event Filter (EF).
The L1 trigger searches for high transverse-momentum muons, electrons and photons, jets,710
and taus decaying into hadrons, as well as large missing and total transverse energy. Its selection isbased on reduced-granularity information from a subset of detectors. High transverse-momentummuons are identified using dedicated trigger chambers in thebarrel (RPCs) and endcap regions(TGCs) of the detector. The calorimeter selections are based on reduced-granularity informationfrom all the calorimeters. Results from the L1 muon and calorimeter triggers are processed by the715
Central Trigger Processor, which implements a trigger ’menu’ made up of combinations of triggerelements. Prescaling of the trigger menu items is also available, allowing optimal use of the ratebandwidth as luminosity and background conditions change.
The maximum rate at which the front-end systems can accept L1triggers is limited to 100 kHz.During the L1 decision time of 2.5 µs, the detector signals are stored in detector-specific pipelined720
buffers. Events fulfilling the L1 trigger selection are transferred to the next stages of the detector-specific electronics and subsequently to the data acquisition via point-to-point links.
The L1 trigger also provides regions-of-interest (RoI), i.e. geographical coordinates inη andφ, of regions within the detector where its selection processhas identified interesting features inthe event. In addition to the coordinates of the objects, theRoIs also contain information on the725
type of feature identified and the criteria passed, e.g. threshold. This information is subsequentlyused to seed the selection performed by the HLT.
– 22 –
The L2 selection is seeded by the RoI information provided bythe L1 trigger over a dedicateddata path. L2 selection uses, at full granularity and precision, all available detector data withinthe RoI, approximately 2% of the total event data. L2 is designed to reduce the trigger rate to730
approximately 3.5 kHz, i.e. provide an event reduction factor of approximately thirty, for an averageevent treatment time of approximately 10 ms. The final stage of the event selection is provided bythe event filter, which reduces the event rate to about 200 Hz.Its selection is implemented usingoffline analysis procedures and has an average event processing time of order 1 s.
The data acquisition system receives and buffers the event data, approximately 1.5 Mbyte, at735
the L1 trigger accept rate from the detector-specific readout electronics over 1600 point-to-pointreadout links. On request, it subsequently moves the data within the RoIs to the L2 trigger. Forthose events fulfilling the L2 trigger’s selection criteria, it performs event building at the L2 triggeraccept rate, i.e. 3.5 kHz. The assembled events are then moved by the data acquisition to theevent filter, and the events selected by the event filter are moved to permanent event storage. In740
addition to the movement of data, the data acquisition also provides for the configuration, controland monitoring of all ATLAS components for and during data-taking. It does not provide for thesupervision of the detector hardware, e.g. gas systems, as this functionality is provided the DetectorControl System.
1.1.6.3 DCS. In order to enable coherent and safe operation of the ATLAS detector, a Detector745
Control System (DCS) has been defined and implemented. It hasthe task to set up the detectorhardware in a selected state and to continuously monitor itsoperation. The DCS consists of twoparts: the Front-end systems (FE) and the Back-end (BE) control stations.
The FE connects to the detector hardware and the equipment tosupervise ranges from simplesensors to complex devices like software controlled power supplies. A small set of commercial750
devices has been selected as a standard, such as crates and power supplies. A general purpose I/Oconcentrator has been developed, called Embedded Local Monitor Board (ELMB). It comprises amultiplexed ADC (64 channels with 16 bit resolution), 24 digital I/O lines and a serial bus to driveexternal devices. The ADC part of the ELMB board can be configured for various types of sensors.
The BE is organised in 3 layers: the Global Control Stations (GCS) with human interfaces755
in the ATLAS control room for overall operations, the Subdetector Control Stations (SCS) forstand-alone operation of a subdetector, and the Local Control Stations (LCS) for process control ofsubsystems.
Each subdetector has one SCS and an additional one, called Common Infrastructure Control(CIC) has been set up to supervise the common environment andservices. Apart from data read by760
its LCS, the CIC uses also data collected via the network by the Information Server (DCS IS in theGCS layer) from several systems. Data include summary information and status of the electricitydistribution for the experiment, the operational parameters of all subdetector gas systems and ofthe LAr cryogenics system. All status information of the ATLAS magnets is read in this way.
1.1.7 Experimental conditions and interface to the LHC machine765
The experimental conditions at the LHC are difficult. Critical parameters are in particular theradiation levels, leading to activated elements and detector background. To reduce the radiationslevels as much as possible ATLAS has optimised its shieldinglayout and, together with the LHC
– 23 –
machine, the beam-pipe design. Nevertheless, a good communication between the LHC machinecontrols and ATLAS will be needed during injection and normal data-taking, also to prevent beam-770
related accidents that can damage critical detector elements. A set of forward detectors are used toprovide key measurements related to measurements of the luminosity and also for forward physicsstudies. These detectors are carefully integrated along the beampipe in the forward direction, closeto various accelerator or shielding structures as described below.
1.1.7.1 Radiation levels. At the LHC the primary source of background radiation at fulllumi-775
nosity is collisions at the interaction point. In the inner detector, charged hadron secondaries frominelastic proton-proton interactions dominate the radiation backgrounds at small radii while furtherout the effects of other backgrounds, such as neutrons, become more important.
In ATLAS, most of the energy from primaries is dumped into tworegions, the TAS collimatorand the forward calorimeter. These two sources are however somewhat self shielding. On the other780
hand, the beam vacuum system spans the length of the experiment and is in the forward regiona major source of radiation backgrounds. Primary particlesfrom the interaction point strike thebeam pipe at very shallow angles, so the projected material depth is large. Studies have shown thatthe beamline material contributes more than half of the radiation backgrounds in the muon system.The deleterious effects of background radiation fall into anumber of general categories, increased785
background and occupancies, radiation damage and ageing ofdetect components and electronics,single event upsets and single events damages, and finally creation of radionuclides that impactsaccess and maintenance scenarios.
Replace the following paragraph with a table:The ATLAS tracker has been designed to with-stand high radiation doses (500-1,000 kGy for the innermostpixel layers, and up to 100 kGy for790
the systems further away from the interaction point, after 10 years of operation). These radiationlevels impose severe limitations on almost all aspects of the design and implementation of the innerdetector sensors and electronics. The radiation doses for the calorimeters reach 200 kGy for theelectromagnetic part and 1,000 kGy for the hadronic part at the highestη . However, in the case ofthe calorimeters, the electronics are placed off the detector reducing the requirements on radiation795
tolerance.In the muon area ... missing...
1.1.7.2 Shielding elements. In order to minimise the radiation effects, almost 3000 tonsofshielding has been added to the experiment. The inner layer is made of materials which providea large number of interaction lengths (such as iron or copper) and is designed to stop high energyhadrons and secondaries. A second layer, consisting of doped polyethylene that is rich in hydrogen,800
is used to moderate the neutron radiation escaping from the first layer and the low energy neutronsare then captured by a boron dopant. Photon radiation is created in the neutron capture processand these photons are stopped in the third shielding layer which consists of steel or lead.Theseelements are placed ... to be written.
1.1.7.3 LHC Machine interface. A shorter version of Section 8.10 is still to be made.805
1.1.7.4 Beampipe. The 38 m beam-pipe section in the ATLAS experimental area consists of 7parts, bolted together with flanges to form a fully in-situ bakeable Ultra-High Vacuum system.The central chamber, called Vacuum Inner detector (VI), is centred around the Interaction Point
– 24 –
(IP) and is integrated and installed with the pixel detector. It has a 58 mm inner diameter and isconstructed from beryllium metal, 0.8 mm thick. The remaining 6 chambers, made of stainless810
steel, are installed symmetrically on both sides of the IP. They are supported by the end-cap LArcryostats, the end-cap toroids and the forward shielding, respectively. The principal pump fordesorbed gasses in the system is a Non-Evaporable Getter (NEG) film sputtered onto the whole ofthe inner surface. The whole length of the vacuum system is permanently equipped with a mass-minimised system of heaters, thermocouples and insulationwhich allow the NEG to be re-activated815
annually by heating the beam-pipe to 200C. Chemically inert gasses not pumped by the NEG areremoved by two minimised sputter-ion pumps.
1.1.7.5 Forward detectors. Three smaller detector systems are built to cover the forward regionof ATLAS. These are closely connected to the luminosity determination in ATLAS, but are in addi-tion foreseen to study forward physics. In order of their distance from the ATLAS interaction point,820
the first system is a Cerenkov detector called LUCID (LUminosity measurement using CerenkovIntegrating Detector). LUCID is the main luminosity monitor in ATLAS, detecting inelastic ppscattering in the forward direction, and is located 17 m awayfrom the interaction point. The sec-ond system is the so-called Zero Degree Calorimeter (ZDC) which is located at a distance of 140 mfrom the IP. This corresponds to the location where the LHC beam-pipe is divided into two and825
the ZDC is located between the beam pipes just after the splitinside the so-called TAN absorber.The ZDC will measure neutral particle at 0 polar angle. The most remote system is the AbsoluteLuminosity For ATLAS (ALFA) system. It consists of scintillating fiber trackers located insideroman pots at a distance of 240 m from the ATLAS IP, and determines the absolute luminosity inATLAS by measuring elastic scattering in the Coulomb interference region.830
1.1.8 Outline of the paper
This paper has been structured in the following way: section1 presents an overview of the ATLASexperiment as installed in terms of its main characteristics and section 2 summarises the key fea-tures of the solenoid and toroid magnet systems and of the overall ATLAS magnetic field as mappedand measured in recent campaigns. The shielding and radiation levels expected are presented in835
section 3. The next three sections, sections 4, 5, and 6, describe respectively in more detail the innertracking system, the calorimetry, and the muon spectrometer. Section 7 discusses in broad termsthe hardware aspects of the trigger and data acquisition system. Section 8 then presents the mainfeatures of the infrastructure in the ATLAS cavern, including the overall integration aspects, theplacement strategy and the survey results, the services (including the cables and pipes, the gas and840
cooling systems, the cryogenics, the back-up power units, etc...), the beam-pipe, and the accessand maintenance scenarios. Section 9 presents an overview of the global performance expectedfrom the ATLAS detector, as obtained from combined test-beam measurements and from the latestanalysis results based on the large-scale simulations donefor the commissioning of the computingsystem. Finally, section 10 briefly summarises the current status of installation and commissioning845
and the expectations for the ultimate completion of the detector and its operation in 2007 and 2008.
– 25 –
2. Magnet system and magnetic field
This chapter first briefly describes the as-built ATLAS magnet systems (section 2.1), which com-prise one solenoid and three toroid systems (one barrel and two end-caps). Section 2.2 then pro-ceeds with a description of the current understanding of themagnetic field across the whole exper-850
iment. This includes both measurements performed to-date (mapping of the solenoid field and firstmeasurements of the barrel toroid field with the as-installed Hall-probe system) and calculations,which have been ongoing for many years and which have as main goal the determination of a de-tailed field map with the required accuracy and performance specifications to be used extensivelyin all ATLAS simulation and reconstruction applications.855
2.1 Magnet system
ATLAS uniquely features a hybrid system of four large superconducting magnets. This magneticsystem is 22 m in diameter and 26 m in length, with a stored energy of 1.6 GJ. After approximately15 years of design, construction in industry, and system integration at CERN, the system has beeninstalled in the underground cavern between 2005 and 2007. The properties of the as-built magnets860
and the services required for their safe and efficient operation are presented briefly in this section.More details can be found in [3] for the solenoid and in [4] forthe three toroid systems.
Figure 1 shows the general layout, the four main layers of detectors and the four supercon-ducting magnets providing the magnetic field over a volume ofapproximately 12,000 m3, definedas the volume in which the magnetic field exceeds 50 mT:865
• the solenoid, which is aligned on the beam axis, provides a 2 Taxial magnetic field for theinner detector (see section 2.1.1);
• the barrel toroid (see section 2.1.2) and the two end-cap toroids (see section 2.1.3) provide atoroidal magnetic field of approximately 0.5 T and 1 T for the muon detectors in the centraland end-cap regions, respectively.870
The main parameters of the magnets are listed in Table 5.
Figure 5. Bare central solenoid in the factory aftercompletion of the coil winding.
During the design and manufacturingphases, regular project overviews and statusreports of design and production have beenpublished [5, 6]. Approximately 15 years ago,875
the first pre-design was sketched and in 1997the technical design reports of the magnetsystems were published [7, 8, 9, 10]. Thecold-mass and cryostat integration work be-gan in 2001. The solenoid was lowered in880
the ATLAS cavern in 2004 (together with theLAr barrel calorimeter). This was followedby the eight independent barrel toroid coils(during 2005 and 2006) and finally by theend-cap toroids in 2007.885
– 26 –
General Specific Unit Solenoid Barrel toroid End-cap toroidsproperty featureSize Inner diameter m 2.46 9.4 1.65
Outer diameter m 2.63 20.1 10.7Axial length m 5.29 (cold) 25.3 5.0Number of coils 1 8 2× 8
Mass Conductor Tonnes 3.8 118 2× 20.5Cold mass Tonnes 5.4 370 2× 140Total assembly Tonnes 5.7 830 2 x 239
Coils Turns per coil 1154 120 116Nominal current kA 7.73 20.5 20.5Magnet stored energy GJ 0.04 1.08 2 x 0.25Peak field T 2.6 3.9 4.1Field range T 0.9–2.0 0.2–2.5 0.2–3.5
Conductor Overall size mm2 30 x 4.25 57 x 12 41 x 12Ratio Al:Cu:NbTi 15.6:0.9:1 28:1.3:1 19:1.3:1Number of strands (NbTi) 12 38 40Strand diameter (NbTi) mm 1.22 1.3 1.3Critical current (at 5 T and 4.2 K) kA 20.4 58 60Critical-current margin at 4.5 K % 20 30 30Residual resistivity ratio for Al > 400 > 800 > 800Temperature margin K 2.7 1.9 1.9Number of units× length m 1× 9100 8 × 4× 1730 2× 8× 2× 800Total length (produced) km 10 56 2 x 13
Heat load At 4.5 K W 130 990 330At 60–80 K kW 0.5 7.4 1.7Liquid helium mass flow g/s 7 410 280
Table 5. Main parameters of the three different ATLAS magnet systems.
2.1.1 Solenoid
The solenoid [3], shown bare in Fig. 5 after completion of themanufacturing process, is designedto provide a 2 T axial field (1.998 T at the nominal 7.73 kA operational current) with a minimalthickness of∼ 0.6 X0 [11] and its main parameters are listed in Table 5. The single-layer coil iswound with an Al-stabilised NbTi conductor inside a 12 mm thick Al 5083 support cylinder. The890
inner and outer diameters of the solenoid are 2.46 m and 2.63 mand its axial cold length is 5.29 m.The coil has a mass of 5.7 tons, with a stored energy of 39 MJ. When installed in ATLAS, theflux is returned by the iron in the hadronic calorimeter and its girder structure, which also providea 4.1% enhancement of the magnetic field at the geometrical centre of the coil. The coil wasmanufactured and pre-tested in the factory [12], came to CERN for integration in the LAr cryostat,895
underwent the on-surface acceptance test in its semi-final configuration [13], and was installed inits final central position in ATLAS in October 2005. The one week cool-down and commissioningtest up to nominal field was successfully completed in the summer of 2006 [14] and the solenoid isnow ready for detector operation.
– 27 –
Figure 6. The barrel toroid installed in the underground cavern. Notethe symmetry of the supportingstructure. Some temporary platforms and scaffolding can berecognised, but these have been removed af-ter completion of the installation. Also note the scale withrespect to the person standing in between thetwo bottom coils in front. In the background, the solenoid and the LAr barrel calorimeter are waiting formovement on the rails towards the centre of the experiment and their final position.
2.1.2 Barrel toroid900
The cylindrical volume around the calorimeters and both end-cap toroids (see Fig. 1) is filled withthe magnetic field from the barrel toroid, which is composed of eight coils in a racetrack-shapedstainless-steel vacuum vessel with a tube diameter of 1.1 m,a length of 25.3 m and a radial widthof ∼ 5.4 m. The eight coils are supported by eight inner and eight outer diameter rings of struts.The position of the eight coils in the toroid can be clearly recognised in Fig. 6. The overall size of905
the barrel toroid system as installed is 25.3 m in length, with inner and outer diameters of 9.4 mand 20.1 m, respectively.
The design features of the barrel toroid are explained elsewhere [15]. The conductor and thecoil windings are essentially the same in all coils and are based on winding a pure Al-stabilisedNb/Ti/Cu conductor [16] into pancake-shaped coils followed by vacuum impregnation.910
The cold-mass integration [17] and the cryostat integration were performed at CERN [18] dur-ing a period of approximately three years and completed in summer 2005. In parallel, all coilssuccessfully underwent on-surface acceptance testing [19]. The details of the coil testing are pub-lished elsewhere, in [20] for the first coil, in [21] for an overall summary, and in [22] and [23] forquench behaviour and quench losses, respectively.915
The net Lorentz forces of approximately 1400 tons per coil directed inwards and the self-weight of the toroids are counteracted by the warm structureof Al-alloy struts mounted in betweenthe eight coils. However, the barrel toroid structure does deflect significantly (a few cm) duringthe load transfer. After release of the temporary support structure and systematic loading of thetoroid with its own weight of 830 tonnes and the additional 400 tonnes of weight of the muon920
– 28 –
chambers, the final shape of the toroid bore was designed to benearly cylindrical. The toroid coilswere installed in calculated positions on an oval, longer by30 mm in the vertical direction, to allowfor structure deflection during load transfer from the temporary support structure. Since the releaseand removal of the installation supports, the toroid has moved downwards by about 26 mm, whichdemonstrates that the design values had been well established and that the installation was precise925
to within a few millimetres.
The installation of the barrel toroid in the ATLAS cavern commenced in October 2004 and ittook about 11 months to install the complete toroid, as depicted in Fig. 6 and discussed in moredetail in section 8.7 within the context of the overall ATLASinstallation, for which this phasewas one of the most demanding ones. The overall structure design and installation experience are930
reported in [24].
2.1.3 End-cap toroids
The end-cap toroids generate the magnetic field required formaximum bending power in the end-cap regions of the muon spectrometer system. The two end-captoroids are positioned on thecentral rails, which facilitates the opening of the ATLAS detector for access and maintenance.935
Each end-cap toroid has a single cold mass built up from eightflat and square coil units and eightkeystone wedges, bolted and glued together into a rigid structure to withstand the Lorentz forces(see Fig. 7). Design details are given elsewhere [25], and the production in industry of the coilmodules and vacuum vessels is described in [26].
Figure 7. End-cap toroid cold mass inserted into thecryostat.
The cold masses were assembled and in-940
serted into their cryostats at CERN. Figure 7shows the first end-cap toroid interior justprior to the closing of the vacuum vessel. Acrucial step in the integration process is theadjustment of the cold mass supports [27].945
The weights of cold mass and vacuum ves-sel are 140 and 80 tonnes respectively. Withthe exception of the windings, of the coil sup-ports, and of the bore tube, the entire structureis made out of Al alloy. The fully integrated950
weight is 240 tonnes and the end-cap toroidswere the heaviest single objects to be loweredinto the ATLAS cavern.
The end-cap-toroid cold masses will eachbe subject to a Lorentz force of 280 tons,955
pushing them into the barrel toroid. Achiev-ing the correct sharing of the forces in the ax-ial tie-rods has therefore been a critical designgoal. Prior to their installation in the cavern in summer 2007, both end-cap toroids passed on-surface tests at 80 K to check the magnet mechanics and electrical insulation after thermal shrink-960
age. Once the end-cap toroids are powered in series with the barrel toroid, the peak stress in the
– 29 –
barrel-toroid windings, in the areas where the magnetic fields overlap, will increase by about 30%.
2.1.4 Magnet services
2.1.4.1 Cryogenics. As described more generally for all cryogenics systems in ATLAS in sec-tion 8.4, the overall magnet cryogenics system is divided into external, proximity, and internal965
cryogenics sub-systems, which are linked to each other via transfer lines. These transfer lines arefixed for the solenoid and barrel toroids, whereas they are mobile for the end-cap toroids, whichhave to be recessed for access and maintenance (see section 8.8).
The layout of the underground part of the various cryogenic systems is shown in Fig. 8. Theexternal cryogenics consist of two refrigerators (the mainrefrigerator and the shield refrigerator),970
a distribution transfer line, and a distribution valve box.The main refrigerator cold box has arefrigeration capacity of 6 kW at 4.5 K equivalent, while theshield refrigerator cold box has arefrigeration capacity of 20 kW at 40–80 K.
The gas buffers are located on the surface with the refrigerator compressors, while the refrig-erator cold boxes are installed in the USA15 side cavern. Thecommon distribution transfer line975
makes the link to the distribution valve box in the main cavern. All proximity cryogenics equip-ment, including the storage dewar, cold pumps, cryostat phase separator, and distribution valvebox (except for the valve unit of the solenoid) are positioned near the wall of the main cavern, asschematically shown in Fig. 9 (left).
Figure 8. Layout of the magnet cryogenics system in and around the surface hall (compressors and gasstorage) and services cavern (shield refrigerator and helium liquefier). They deliver cold gas and liquid tothe distribution valve box in the experimental cavern, fromwhich the solenoid and the toroid proximitycryogenics are fed.
The distribution valve box distributes the fluids to two independent proximity cryogenic sys-980
tems, one for the toroid magnets (barrel cryo-ring and two end-caps) and one for the solenoid. Forthe toroid magnets, there is a storage dewar with a capacity of 15,000 litres of liquid helium. Therealso exists a phase separator dewar with two centrifugal pumps and a storage capacity of 600 litres
– 30 –
Figure 9. Left: Layout of underground services connections to the barrel toroid. The two big helium dewarscan be seen on the side of the main cavern. Also shown are the fixed cryogenic lines supplying the solenoidand the cryo-ring for the barrel toroid coils at the top. The cryogenics lines in the flexible chains supplythe two end-cap toroids and follow them whenever they have tomove for access and maintenance of thedetector. Right: schematic of the liquid-helium supply in the barrel toroid. The cryo-ring sectors consistof six standard ones, the bottom sector with a valve box, where the input flow per coil is measured andcontrolled, and the top sector (with a connection to the current lead cryostat) connect the coils where alllines pass.
of liquid helium. The solenoid has a control dewar with a storage capacity of 250 litres of liquidhelium.985
The proximity cryogenic equipment supplies coolant to the magnet internal cryogenics, whichconsist mainly of cooling pipes attached to the cold mass andthe thermal shield. The aluminiumcooling tubes are either welded to the outer surface of the Al-alloy support cylinder (solenoid) orembedded in the Al-alloy coil casings enclosing the pancakecoils (toroids).
The toroid magnets are cooled with a forced flow of boiling helium through cooling lines990
attached to the cold mass that encases the conductor cable ofthe magnet. Helium enters the magnetsfrom the top. In the case of the barrel toroid (Fig. 9, right),helium is supplied from the currentlead cryostat positioned on the top sector, runs down to the distribution valve box at floor level witheight control valves regulating the flow in the eight coils, then goes up and enters the eight coilsseparately, while the return line returns to the top. A totalof 1200 g/s of slightly sub-cooled liquid995
helium is circulated by means of centrifugal pumps, which take the liquid from the phase separatordewar. The system is equipped with two such pumps for redundancy reasons and the spare pump iscalled into operation when the first one fails, which minimises the risk of detector down-time. The
– 31 –
liquid helium in the storage dewar will be used in the event ofa failure with the main refrigeratorto provide the required cooling capacity to safely ramp downthe toroid magnets over a two-hour1000
period.
The solenoid, with a cold mass of approximately five tonnes, is cooled with a direct Joule-Thompson flow from the main refrigerator and is slightly sub-cooled via a heat exchanger in the250 litre helium control dewar.
The flow in the solenoid and ten toroid cold masses is controlled individually to cope with1005
variations in flow resistance and to guarantee helium quality in all coils. Given that the end-captoroids and solenoid each have a single cold mass, there is a single flow control and the branchesof cooling pipes (two for the solenoid and sixteen for each end-cap toroid) are put in parallel.
2.1.4.2 Vacuum system The insulating vacuum is achieved with diffusion pumps directly at-tached to the barrel and the end-cap toroids, two per coil, each with a capacity of 3 kl/s. In addition,1010
two roughing and three backing pumps are used in the low stray-field area at the cavern wall. Un-der normal conditions with a leak rate less than 10−4 mbar· l/s, a single pump would be sufficient;however, in order to minimise detector down-time, extra pumping capacity and redundant unitswere installed. Since the solenoid is installed inside the cryostat of the LAr barrel calorimeter, theinsulation vacuum is controlled by the LAr cryogenic system(see section 8.4.5) and not by the1015
magnet control system.
2.1.4.3 Electrical circuit The three toroids are electrically connected in series to the 20 kA/16 Vpower supply system shown schematically in Fig. 10 (left). They are however individually voltage-protected by the two diode/resistor ramp-down units. The power supply, switches, and diode/resistorunits are located in the side cavern and approximately 200 m of aluminium bus-bars provide the1020
connections to the magnets in the experimental cavern. Ramping up is accomplished with a rateof 3 A/s, leading to two hours maximum ramping-up time. In thecase of a slow dump, the magnetsare de-energised across the diode/resistor units in about 2.5 hours. Quench detection is by clas-sical bridge connections across the entire barrel toroid, across the end-cap toroids and across thesolenoid, as well as across individual coils, using differential voltage measurements with inductive1025
voltage compensation.
There is a six-fold redundancy in the toroid quench detection grouped in two physically-separated units and cable routings. Quench protection is arranged by firing heaters in all toroidcoils so that a uniform distribution of the cold-mass heating is achieved. Given the normal-zonepropagation of 10–15 m/s, a toroid coil is switched to back the normal state within 1–2 seconds.1030
The electrical circuit of the central solenoid is similar and shown in Fig. 10 (right).
As for the quench detection, the quench-protection heater circuits including power supply,cabling, and heaters show a two-fold redundancy. During a quench, the cold-mass temperaturerises to about 70–80 K in the toroids and to about 100 K in the solenoid, both of which are verysafe values. An overview of the magnet services can be found in [28].1035
2.1.4.4 Magnet controls. The magnet control system provides the process controls needed toexecute automatically the various running modes of the magnet system. The implementation isrealised within the context of the overall ATLAS detector control system, as described in Sec. 7.5.
– 32 –
Figure 10. Left: electrical circuit showing the toroids connected in series, fed by a 20 kA power converterand protected by a voltage-limiting diode/resistor ramp-down unit. Right: the electrical circuit of the centralsolenoid is fed by a 8 kA power converter.
The hardware designs rely on a three-layer model, using distributed input/output connected viafield-networks or directly by wiring to a process-control layer, the last layer being the supervision.1040
The main control functions are:
• perform automatic operational sequences on a given magnet (sub-system tests);
• communication interface with the power converter;
• regulation of the helium flow in the magnet current leads as a function of the magnet current;
• information exchange between the control system and other sub-systems such as vacuum or1045
cryogenics;
• monitoring of all critical parameters in the coil (temperatures, strain and displacement gauges);
• calculation of non-linear sensor corrections (temperature sensors, vacuum gauges).
The supervision system displays the main process parameters on a synoptic view, communi-cates with the power supply, collects both continuous and transient data, allows visualisation of any1050
collected data on trend charts and archives collected data.For long-term storage and for correlationof data between different systems, a central data-logging system will regularly receive a pre-definednumber of data items from each magnet system. A subset of the main control parameters is sent tothe ATLAS detector safety system and beyond that to the LHC machine (see section 8.10).
2.2 Magnetic field determination1055
The specifications on the determination of the magnetic field(see section 2.2.1) are rather differ-ent in the inner detector (ID) and the muon spectrometer. In the ID cavity, the driving consid-eration is the absolute accuracy of the momentum scale. In the muon spectrometer, the field ishighly non-uniform and uncontrolled bending-power uncertainties would translate primarily intodegraded muon momentum resolution. Detailed magnetic modelling (see section 2.2.2) and novel1060
– 33 –
instrumentation (see section 2.2.3) have allowed a high-precision mapping of the solenoid field (seesection 2.2.4) as well as a preliminary experimental validation of the field measurement and recon-struction strategy in the muon spectrometer (section 2.2.5). Studies are in progress to combinemagnetic models with field measurements into an overall fieldmap [29] for ATLAS data-taking(see section 2.2.6).1065
2.2.1 Performance specifications and measurement concepts
In the inner detector, the systematic error affecting the momentum measurement of charged tracks isdominated by the relative alignment of detector componentsand by bending-power uncertainties,the former being the more demanding. A high-precision measurement of theW-boson mass isclearly the most challenging goal for such measurements: a lepton fromW-boson decay carries1070
typically a transverse momentum of 40 GeV, resulting in a sagitta of approximately 1 mm as thelepton traverses the ID cavity. The systematic alignment uncertainties in the ID are unlikely toimprove beyond the 1µm level or 0.1% of the sagitta. This suggests setting a targetof ∼ 5 × 10−4
for the fractional bending power uncertainty, so that it remains negligible when determining theATLAS absolute momentum scale. Such stringent requirements can only be achieved reliably1075
by in-situ mapping, using dedicated instrumentation inside the ID cavity, with all the relevantmagnetic materials in place and just before the final installation of the ID itself. Eventual long-term drifts of the absolute scale will be detected to a much higher accuracy, using permanentlyinstalled NMR probes.
In the muon spectrometer, the expected sagitta is approximately 0.5 mm for a muon with a1080
momentum of 1 TeV. The extraction of the momentum from the chamber measurements (MDT’s)requires a precise knowledge of the field integral between consecutive chambers along the muontrajectory. Because the field gradient can reach 1 mT/mm, local bending-power uncertainties trans-late into fluctuations of the momentum scale from one region in space to another, adding in quadra-ture to the overall momentum resolution. In addition, the interpretation, in terms of spatial coordi-1085
nates, of the drift-time measured in the MDT’s is sensitive to the local electric and magnetic fieldsexperienced by the ionisation electrons in each tube. The corresponding functional requirementsare extensively discussed in Ref. [30] and summarised in Table 6.
Bending power accuracy MDT drift propertiesPerformance criterion ∆σpT /σpT < 5% overall Single-wire resolution degraded by< 5%
Field-measurement accuracy ∆Bφ/Bφ < 2 – 5 ×10−3 ∆Bx,y,z < 4 mT (relative over chamber)Reconstructed position of toroid ∆R∼ 1 – 12 mm,∆(Rφ) ∼ 1 – 6 mm, -conductors with respect to MDT tower ∆z∼ 2 – 30 mmµ chamber 2nd-coordinate resolution 1.7–5.5 mm 6 to > 20 mm
Table 6. Summary of magnetic-field-related performance specifications in the muon spectrometer. Thespread reflects theη −φ variations in field gradient and/or strength.
For a given muon trajectory, three sources of uncertainty affect the measured curvature: fieldmeasurement errors; the accuracy on the relative position of muon chambers and magnet coils; and1090
trajectory measurement errors, in particular along the direction of MDT wires. For the purposeof setting specifications, it has been required (somewhat arbitrarily) that the combined effect ofthese sources degrade the momentum resolution by no more than 5% in relative terms; each source
– 34 –
Figure 11. Longitudinal and radial dependence ofthe magnetic field in the inner detector cavity.Thesymbols are the measured axial and radial field com-ponents and the lines are the result of the fit de-scribed in section 2.2.4.
η0 0.5 1 1.5 2 2.5
m)
⋅B
dl
(T
∫
-2
0
2
4
6
8
barrel regionregionend-cap
tran
sitio
n re
gion
=0φ
/8π=φ
Figure 12. Field integral as a function of|η | fromthe innermost to the outermost MDT layer in onetoroid octant, for infinite-momentum muons. Thecurves correspond to equally-spaced azimuthal an-gles.
should then contribute no more than∼3% of fractional resolution degradation, anywhere in thespectrometer volume.1095
In-situ mapping of the spectrometer by conventional techniques would have been impracticalbecause of the rapidly-varying field and very large volume. Instead, the muon system is equippedwith an array of B-field sensors; their readings are comparedwith magnetic simulations and usedfor reconstructing the field in space. This strategy was shown [30] to meet the field-map specifica-tions above, provided the B-sensor readings, after correcting for perturbations induced by magnetic1100
materials, are accurate to∼ 1 mT (absolute) and the field direction is measured to within± 3 mrad.
2.2.2 B-field modelling
The total field in the ID cavity, the calorimeters, and the muon spectrometer is computed as thesuperposition of the Biot-Savart contributions of all magnet windings with those of the magne-tised calorimeter and with the localised perturbations induced by other ferromagnetic structures.1105
In order to reach the required accuracy, the calculation combines numerical integration of the con-tributions of the solenoid, barrel-toroid and end-cap-toroid windings with finite-element modellingof magnetic structures.
The solenoid conductor model is described in section 2.2.4.The magnetised iron (tile calorime-ter and solenoid flux-return girder), which surrounds the IDcavity, is predicted to modify the field1110
by 4.1% at the geometrical centre of the coil. At nominal current, the total field is 1.998 T at thegeometrical centre and drops steeply from∼ 1.8 T at z = 1.7 m to∼ 0.9 T at the end of theID cavity (Fig. 11).
The toroid windings are, at this stage, described using their nominal geometry. The densityof the integration mesh is tailored to the local field gradient to ensure an accurate representation1115
of field variations (as also done for the solenoid). Depending on the radiusR and azimuthφ,the field varies from 0.15 T to 2.5 T, with an average value of 0.5 T, in the barrel region, andfrom 0.2 to 3.5 T in the end-cap region [30]. The analysing performance of the toroid system canbe roughly quantified by the field integral experienced by particles originating from the interaction
– 35 –
Figure 13. Sources of magnetic perturbations in-duced by metallic structures in or near the muonspectrometer.
Figure 14. Schematic representation of themagnetic-sensor layout and coil deformation model,used to reconstruct the magnetic field inside a barreloctant. The MDT nomenclature is defined in Ta-ble 31 (see Section 6.3).
point and propagating in a straight line (the ultimate criterion is the momentum resolution: a zero1120
field integral does not necessarily imply infinite resolution). This available bending power is shownin Fig. 12 as a function of|η |. It shows good magnetic field coverage up to|η | ∼ 2.5−2.7. Theregions with low field integral, around|η | = 1.4 and|η | = 1.6, correspond to trajectories in theplane of an end-cap coil or of a barrel coil, where the fringe field of one magnet largely cancels thebending power of the other.1125
A number of large magnetisable components, shown schematically in Fig. 13, distort the Biot-Savart field at different levels. Although amenable to experimental spot-checks (Sec. 2.2.5), suchperturbations can only be determined using field simulations.
The highly anisotropic structure of the tile calorimeter cannot be satisfactorily modelled usingonly a scalar permeability and an effective iron-packing factor: a formalism incorporating a mag-1130
netic permeability tensor, as well as a more sophisticated treatment of magnetic discontinuities atmaterial boundaries, is called for. The problem is compounded by the superposition of the solenoidand toroid fields in the partially-saturated flux-return girder and in the tile calorimeter itself. Anovel approach to magnetic-field modelling in such structures has therefore been developed [31],and embodied in the B-field simulation package ATLM [32]. This package, which incorporates a1135
careful description of the toroid and solenoid conductors as well as a detailed mathematical modelof the tile calorimeter, is used both to compute the Biot-Savart field by numerical integration (asdescribed above), and to predict, by a finite-element method, the field distortions caused by thetile calorimeter, the flux-return girder and the shielding disk in both the ID cavity and the muon
– 36 –
spectrometer. Altogether, these distortions affect the field integral in the muon spectrometer by up1140
to 4%, depending on|η | andφ; in addition, they induce, at the level of the inner MDT layers, localfield distortions of up to|∆B| ∼ 0.2 T.
A few discrete magnetic structures, either inside the muon spectrometer or close to its outerlayers, induce additional, localised magnetic perturbations. Their impact has been evaluated usingthe 3D finite-element magnetostatics package TOSCA [33]. The largest are caused by the air pads,1145
jacks and traction cylinders that allow the calorimeters and the end-cap toroids to slide along therails. These affect primarily the field distribution acrossthe BIS and BIL chambers in sectors 12to 14, and in addition impact the field integral at the level ofup to 10% over small islands inη −φspace.
The field perturbations caused in the outside MDT layers by the massive steel frame and1150
platforms (HS structure described in section 8.6), which surround the detector, range from|∆B| ∼2 mT up to∼ 50 mT and rapidly decrease as one moves inwards from the outerto the middlechamber layer. While their impact on B-sensor readings and MDT drift properties does need to betaken into account, they barely affect the bending power, except possibly in a few narrow regions.
The other components in Fig. 13 have much less of an impact because either they lie in a low-1155
field region, they intercept a very small fraction of the end-cap muons, or they are made of stainlesssteel with a high-field relative permeability very close to 1.
2.2.3 Magnetic field instrumentation and reconstruction
2.2.3.1 B-field sensors The inner detector is equipped with four NMR probes fixed to the wallof the inner warm vessel nearz ∼ 0 and equally-spaced in azimuth. These probes measure the1160
field strength with an accuracy of around 0.01 mT and will remain in place to monitor the ID fieldstrength throughout the lifetime of ATLAS.
Because NMR probes only measure|B| and because they cease functioning in a gradient ofa few Gauss/cm, the solenoid mapper, described in section 2.2.4, and the muon chambers areequipped instead with 3D Hall cards [30, 34]. These consist of a rigid printed-circuit board carrying1165
a small glass cube, with a Hall probe on each of three orthogonal faces to measure each fieldcomponent. Every card includes its own readout electronics, as well as a thermistor for localtemperature compensation.
All the Hall cards were calibrated in a highly uniform field monitored by a NMR probe. Theachieved absolute Hall-card accuracy on|B| is 0.2 mT up to|B| = 1.4 T and 1 mT up to 2.5 T,1170
whereas the angular accuracy achieved on the measured field direction is 2 mrad.
2.2.3.2 B-field reconstruction In an air-core magnet, the magnetic field can in principle be cal-culated by direct application of the Biot-Savart law, once the geometry of all conductors is knownand assuming material-induced magnetic perturbations arenegligible. In practice however, theconductor position and shape are known only approximately,owing to fabrication tolerances and1175
to deformations of the magnet structure under gravitational and magnetic loads. The exact locationof each magnet coil, as well as the relative positions of the end-cap and barrel toroids, will be re-producible, after a power cycle or an access period, to a finite precision only. Therefore, the fieldmust be measured under running conditions, with all detector components in place and under themutual influence of the different magnets and magnetic structures.1180
– 37 –
To this effect, the muon spectrometer is equipped with an array of approximately 1700 Hallcards, which remain mounted permanently and precisely on the MDT chambers and continuouslymeasure all three field components. Two NMR probes, installed at low-gradient locations in thebarrel toroid, will complement the system, with the aim of detecting eventual long-term drifts inthe response of the Hall cards. The 3-D sensor readings are compared with field calculations that1185
include both the contributions of the magnet windings and those of nearby magnetised structures,and are used for reconstructing the position and the shape ofthe toroid conductors with respect tothe muon chambers (see Fig. 14). Once the geometry of the coils is known, the field can be calcu-lated anywhere in the muon spectrometer. Simulation studies using a simplified coil deformationmodel have shown that the magnetic field can be reconstructedto a relative accuracy of 0.2% [30].1190
2.2.4 Solenoid-mapping measurements
2.2.4.1 Mapping campaign The field was mapped [35] in August 2006 by a machine, whichscanned a Hall-card array over a volume slightly larger thanthat now occupied by the inner detector.During this mapping campaign, the barrel and end-cap calorimeters were all in their final positions.Although the shielding disks were not yet installed, their differential contribution is small enough1195
(< 0.2 mT in the ID tracking volume) that it can be reliably accounted for later. The same is trueof corrections for the absence of the toroid excitation during mapping.
Mapping data were recorded with solenoid currents of 7730, 7850, 7000 and 5000 A, with afinal set of data back at the nominal operating current of 7730A. Each data set contains at least20,000 points, and is sufficient by itself to fit the field with negligible statistical uncertainty. Each1200
map took about four hours, during which the solenoid currentremained stable to within 0.1 A, asconfirmed by the NMR probes.
2.2.4.2 Mapper geometry, survey and auto-calibration The mapping machine had four armsmounted on a common axle in a windmill configuration, with twelve Hall cards on each arm, atradii ranging from 0.118 to 1.058 m, that directly measured the field components Bz, BR and Bφ.1205
The machine could be rotated around its axle and translated in z along the ID rails by means ofpneumatic motors. Optical encoders allowed control of the mapper movements and readout of itsstop positions with an accuracy of 0.1 mm. A number of surveyswere necessary to determinethe positions of each individual Hall sensor for all possible longitudinal mapper positions andazimuthal settings of the windmill arms. After combining all the information, the estimated overall1210
accuracy on the position of a map point in the cryostat coordinate system is approximately 0.2 mm.
The redundancy and internal consistency of the mapping measurements makes it possible toextract individual probe misalignments from the data themselves to an accuracy of±0.1 mrad.The strong constraints from Maxwell’s equations on physically realisable fields in the absenceof any current sources or magnetic materials, combined withthe fact that the field at the origin1215
can be almost completely determined from the measurements of a single Hall probe, allow us todetermine all three probe alignment angles and to normalisethe Bz component to a common scalefor all probes.
The NMR probes, which were operational throughout the field-mapping campaigns, are usedto set the overall scale of the Hall sensors with an accuracy of about 0.3 mT, the limitation coming1220
from the extrapolation uncertainty from the mapper arms outto the position of the NMR probes.
– 38 –
The NMR data also show that there is negligible hysteresis inthe solenoid system: the field at7730 A remained constant within±0.01 mT from the first excitation cycle onwards, provided thatthis current was approached from below. A small saturation effect is visible in the NMR data, withthe field at 5000 A being 0.34 mT higher than would be expected by simply scaling down from1225
7730 A.
2.2.4.3 Map fitting Using the measured magnet current and a detailed model of thesolenoid ge-ometry, the Biot-Savart law is integrated to produce a field model that should account for most ofthe measured field. The conductor model is based on engineering drawings, with as many param-eters as possible taken from surveys of the as-built solenoid. The coil cross-section is assumed to1230
be perfectly circular. The winding was mechanically assembled from four separate sections, eachwith a slightly different average pitch, and joined together by welds that are represented electricallyby turns having just under twice the average pitch. Also modelled are the welds at the coil endsand the return conductor that runs axially along the outsideof the support cylinder. The expecteddistortion of the solenoid, relative to the room-temperature survey and caused by thermal shrinkage1235
and magnetic pressure, is taken into account as well.
The fit to the mapping data has 11 free parameters. Two overallscale factors allow fine tuningof the conductor model: one common to all longitudinal dimensions, and an independent one forthe radial dimension. Five more free parameters quantify the three offsets and two rotations of theconductor relative to the mapper coordinate system. The calorimeter-iron contribution is modelled1240
by a Fourier-Bessel series with four terms. These parameters are determined by minimising aχ2
function that includes the longitudinal and radial field components at all mapped points.
Scale factors in conductor model (Rscale,z scale) = 0.9993, 1.0012Fitted offsets from solenoid centre to centre of cryostat (∆x,∆y,∆z) = 0.30, -2.34, 0.51 (mm)Fitted rotations of solenoid around cryostatx andy axes (θx,θy) = 0.13, 0.09 (mrad)RMS fit residuals σ(∆Bz,∆BR, ∆Bφ) = 4.3, 3.6, 3.5 (Gauss)
Table 7. Typical fit results of solenoid-mapping measurement at 7730A.
This field model is fitted to several representative data setsunder varying assumptions, withand without implementing various corrections (Hall-card alignment,z-dependent carriage tilt, resid-ual perturbations induced by slightly magnetic mapper components, number of Fourier-Bessel1245
terms...). The longitudinal scale factors emerge as very close to unity, as shown in Table 7, sug-gesting that the coil survey data are well understood. The fitted offsets and rotations with respectto the centre of the reference coordinate system (barrel LArcryostat) are stable at the 0.2 mmand 0.1 mrad level respectively, confirming the vertical -2 mm offset of the solenoid axis indi-cated by the survey results before and after installation inthe experimental cavern (see Table 471250
in section 8.3.2.3).
The on-axis fractional iron contribution, as estimated from the Fourier-Bessel series, is consis-tent with the magnetic field model to better than 2 mT, although thez-dependence remains slightlydifferent. The overall fit is excellent, as illustrated in Fig. 11 and as confirmed by the resultingRMS residuals of∼ 0.4 mT for all three field components (see Table 7). The fit quality is best1255
measured in terms of the fractional sagitta residual,δ s/s, which is shown in Fig. 15), as evalu-
– 39 –
Figure 15. Residual fractional sagittavs. η atequally-spaced azimuthal angles, as extracted froma fit to the solenoid field-mapping data at nominalmagnet current.
(mT)φ B∆-25 -20 -15 -10 -5 0 5 10 15 20 250
1
2
3
4
5
6BMS: mean=-0.3, RMS=1.2
BOS: mean=2.2, RMS=2.6 BIS: mean=-7.8, RMS=7.9
Figure 16. Field reconstruction residual∆Bφ forone middle (green, solid), outer (blue, dashed) andinner (red, dot-dashed) MDT layer.
ated along an infinite-momentum trajectory from the interaction point to the point where the trackcrosses the outer radial or longitudinal boundary of the inner detector. To determine the overallsystematic uncertainty on the magnetic field scale, the achieved mapping accuracy,(δ s/s)RMS
of ∼ 2 × 10−4, must be combined with the absolute scale uncertainty of∼ 1.5 × 10−4, asso-1260
ciated with the NMR normalisation, and with the survey-transfer uncertainty from the coordinatesystem of the mapper to that of the inner detector as installed.
2.2.5 Experimental validation of the field map in the muon spectrometer
The tests carried out in Fall 2006 for the barrel toroid provided the first full-scale test of the B-sensor system, and an initial validation of the magnetic models and field-reconstruction strategy1265
in the muon spectrometer [36]. The end-cap toroids were not yet installed at the time and thesolenoid was turned off. Since the muon-chamber installation was still in progress, only 400 MDTHall cards were available for readout, thus providing sensitivity for field reconstruction in aboutone third of the barrel region.
The sensor signals were extremely clean (∼ 0.01 mT of RMS noise at full field), and repro-1270
ducible to∼ 0.05 mT between magnet cycles separated by up to one week. Non-linear effectsremain very small (< 4 mT in the BIS layer, close to the calorimeter iron, over the full currentrange). The absolute field scale, as determined by an NMR probe located in the azimuthal mid-plane of coil 3, at a point where iron-induced perturbationsare negligible and the field gradientbelow 0.2 mT/cm, agrees with the Biot-Savart prediction to better than 0.2%.1275
The field reconstruction algorithm outlined in section 2.2.3 and detailed in [30] has been ap-plied to B-sensor data collected at nominal field in the barrel toroid. Because the muon alignmentsystem was still being commissioned and the MDT survey not yet completed, it is necessary, atthis stage, to assume that all muon chambers and B-sensors are in their nominal position. Forthe three coils bracketed by the available sensors, the reconstructed conductor shape is qualita-1280
tively consistent with that measured at room temperature before insertion of the windings intotheir respective cryostats. Figure 16 displays the difference, at each active sensor in Sector 2 ofthe muon spectrometer, between the azimuthal component of the measured field (corrected for
– 40 –
perturbations from magnetic materials) and that of the Biot-Savart contribution predicted by thefield-reconstruction fit. A perfect description of the conductor geometry and of magnetic pertur-1285
bations should yield∆ Bφ = 0. The agreement is best in the middle chambers (BM), where thegradients are smallest: the distribution is well centred and exhibits a spread∆BRMS
φ ∼ 1.2 mT. Inthe outer chamber layer (BOS), the distribution of∆ Bφ shows a moderate bias of 2.2 mT and aspread of 2.6 mT. In view of the larger field gradient in these chambers, such a spread is consistentwith the current± 5 mm uncertainty on the as-installed MDT chamber positions.The situation1290
is similar but somewhat worse in the inner chambers (BIS). These preliminary results reflect thecumulative effect of errors in the assumed sensor and chamber geometry, of residual imperfectionsin the magnetic model of the calorimeter iron, and of the performance of the reconstruction fit.
Validation of the TOSCA simulations, which describe the distortions induced by other supportand service structures was carried out using dedicated Hall-card arrays installed at critical locations1295
in the bottom muon sector and between the outer muon chambersand the HS structure (see Fig. 13).The agreement between measured and predicted perturbations typically ranges from 2 to 5 mT atthe location of the Hall cards and should be better within thespectrometer volume. It is satisfac-tory at most locations, although discrepancies as large as 50 mT are observed very close to a fewlocalised and well-identified steel supports. A more extensive magnetic characterisation campaign1300
is planned during the full magnet-system test scheduled forthe end of 2007.
2.2.6 Towards an overall field map for ATLAS data-taking
The default field map in the ID tracking volume will mirror thevery accurate fit obtained for thesolenoid mapping data and illustrated in Fig. 11. This approach automatically takes into accountthe magnetised iron surrounding the ID cavity without having to rely on any field calculations.1305
The fit function is required to satisfy Maxwell’s equations and will include empirical corrections tomatch the measured map as closely as possible, as well as small (< 0.2 mT) additional correctionsfor the shielding disks (which were absent at the time of mapping) and barrel-toroid contributions.
In the calorimeters, the map will be based on the ATLM simulation, with the magnetic param-eters describing the calorimeter iron adjusted to fit the solenoid-only and toroid-only field measure-1310
ments performed in 2006. This simulated map will be smoothlyconnected to the fitted solenoidmap in the future: the potential discontinuity remains to becharacterised, but is estimated not toexceed 2 mT over a very narrow interface region.
In the muon spectrometer, the map will reflect the superposition of the winding contributionswith the predicted distortions associated with the calorimeter iron and other significant magnetic1315
structures inside or near the spectrometer volume. So far, the Biot-Savart calculation presentedabove has been performed only in a 1/16th slice, which spans 45 in azimuth and is longitudinallysymmetric with respect to the interaction point: this is theminimum angular size required to handlecorrectly the symmetries of the full toroid system. Extending it to the case of an arbitrary geom-etry (without any symmetry assumptions) is currently in progress and the final implementation1320
will depend on the extent to which the actual coil geometry, as eventually revealed by the field-reconstruction procedure, deviates from the ideal configuration. Similarly, studies are in progressto assess the magnetic impact of shape or position imperfections in the tile-calorimeter geometry:their outcome will indicate to which extent such deviationsfrom the ideal configuration must betaken into account when describing the field inside the calorimeter and/or muon spectrometer.1325
– 41 –
3. Background radiation and shielding
3.1 Introduction
In contrast to previous and existing colliders, the dominant primary source of background radiationat the LHC, when operating at design luminosity, arises fromcollisions at the interaction point.The rates expected from beam-halo particles and beam-gas interactions are negligible in compar-1330
ison. In the inner detector, charged hadrons from inelasticproton-proton interactions dominatethe radiation backgrounds at small radii, while the effectsof other backgrounds, such as neutrons,become more important further out (see [37] for detailed studies of the various radiation sources,radiation levels, neutron fluences and activation levels expected in ATLAS throughout the lifetimeof the experiment).1335
In ATLAS, most of the energy from the primary particles is dumped into two regions, thecopper collimators (TAS) and the forward calorimeters (FCal) depicted in Fig. 17 and Fig. 18,which are therefore among the strongest sources of secondary radiation. These two sources aresomewhat self-shielding, and since they are compact, they have been further shielded with layers ofdense material and cladding. The beam-vacuum system spans,on the other hand, the whole length1340
of the experiment and is, in the forward regions, another major source of radiation backgrounds,since the primary particles produced at large values of pseudorapidity η strike the beam-pipe atvery shallow angles. Through this mechanism, the beam-pipebecomes an extended line sourceilluminating the interior of the forward cavity. Detailed studies have shown that the beam-linematerial is responsible for more than half of the fluences expected in the muon system [37].1345
A thorough understanding of the impact of background radiation has been a critical element inthe design phase of most of the components of the experiment and a number of deleterious effectshave been considered:
1. Increased detector occupancy, which, in tracking detectors, can lead to inefficiencies, de-graded resolutions, and increased rates of fake tracks. In calorimeters, the pile-up fluctua-1350
tions at high luminosity degrade the energy resolution.
2. Occupancies dominated by hits generated by slow neutronsin the muon spectrometer system,an effect which had not been of any concern in previous colliders.
3. If the backgrounds consist of penetrating tracks, the rates of spurious triggers can increase.Also, increased occupancies can increase the rates of random triggers.1355
4. Radiation damage of silicon detectors and readout electronics.
5. Interactions leading to anomalous deposits of local radiation can change the logical statusof electronic signals (single-event upset) or permanentlydestroy components (single-eventdamage).
6. Wire detectors can experience "ageing" (reduced gain andtherefore efficiency) due to poly-1360
merised deposits on the wires caused by radiation interacting with certain components of thedetector gas.
– 42 –
7. Nuclear interactions in dense materials lead to the creation of residual radio-nuclides. The re-sulting dose rates from radio-activation of certain materials will lead to radiological hazards,which impact access and maintenance scenarios.1365
8. The large fluences expected at the LHC design luminosity may lead to significant radiationhazard from the prompt component of the radiation.
The largest impact from background radiation is of course tobe expected close to the beam-pipe, in particular in the region of the inner detector and ofthe forward calorimeters. Given thelack of available space and the large contribution from primaries, only a limited amount of mod-1370
erator shielding could be installed to minimise the impact of background radiation, as describedin section 3.2.
Very large reductions in the expected background rates in the muon spectrometer have beenachieved by designing a large amount of shielding around theTAS. A total shielding weightof 2825 tonnes (1887 tonnes of metals, 920 tonnes of concrete, and 18 tonnes of plastics) has1375
thus been added to the experiment. Since different types of radiation require different types ofshielding materials, a multi-layered shielding approach has been used. The inner layer’s purpose isto stop high-energy hadrons and their secondaries. This layer is made of materials such as iron orcopper, which provide a large number of interaction lengths. In the case of iron, studies have shownthat a minimum carbon content of a couple of percent is advantageous since it efficiently moderates1380
the neutron energies down to lower values. A second layer, consisting of doped polyethylene (richin hydrogen), is used to moderate the neutron radiation escaping from the first layer and the low-energy neutrons are then captured by the boron-based dopant. Photon radiation is created in theneutron-capture process and these photons are stopped in the third shielding layer which consistsof steel or lead. Lead is more effective in stopping photons,but induces more neutron radiation1385
than steel.
3.2 Description of the shielding
Figure 17 shows the locations of the different shielding components in ATLAS.
The moderator shielding on the front face of each of the end-cap and forward LAr calorimetersreduces the neutron fluences in the volume of the inner detector and protects the inner detector from1390
back-splash of neutrons from the calorimeter. It is made of polyethylene, doped (by weight) with5% boron in the form ofB4C. Reactor tests have shown that the choice of this particulardopantresults in a plastic, which is more radiation-hard than if other boron dopants are used. This isimportant since the shielding in front of the forward calorimeters is exposed to a very large ionisingdose over the lifetime of the ATLAS experiment.1395
There are three brass shielding elements inside each of the end-cap/FCal calorimeter cryostatsat the back. The largest one is attached to the rear end-plateof the cryostats and has a diameterof 387 cm. Closer to the beam-line are two other shielding plugs. One of these is a cylindrically-shaped extension of the forward calorimeters. The main purpose of these shielding elements is toprotect the end-cap inner muon stations from the backgroundradiation.1400
The next protection element is the shielding disk, which serves in fact a threefold purpose:it supports the muon chambers in the first end-cap muon station, it shields these chambers from
– 43 –
Figure 17. Schematic view of major ATLAS detector systems and of the main shielding components (seetext).
background radiation emerging from the calorimeters, and it provides a well-defined path for themagnetic field flux return from the solenoid magnet. The bulk of this shielding disk consists of avertical steel disk with a diameter of 872 cm. This disk supports end-cap muon trigger chambers1405
(see section 6.7). At the centre of the disk and surrounding the beam-pipe is a stainless steel tubecontaining a set of cylindrical shielding pieces made of leaded red brass. This tube also supportscathode strip chambers (CSC) and monitored drift tubes (MDT). Brass shielding has been added tothe disk in order to protect the CSC chambers. There is a polyethylene layer on the outside of thisbrass shielding, which is doped withB2O3, to moderate the neutrons, while photons created in the1410
neutron absorption process are stopped in a third layer madeof lead.The next protection element is the end-cap toroid shielding, which consists of two parts, one
located outside the toroid and enclosing the beam-pipe and one inside the cryostat:
• the first one is a cylindrical structure made of ductile cast iron, which surrounds the beam-pipe on the inside of the two end-cap toroid cryostats. The front piece has a large hole in the1415
centre, into which the stainless steel tube of the shieldingdisk fits. On the outside of the castiron is a polyethylene layer doped withB2O3 (5% by weight). The photons created in thepolyethylene layer are stopped by the stainless-steel boretube, which supports the shieldingin the end-cap toroid;
• the second part of the toroid shielding consists of various polyethylene structures, which1420
are located in the vacuum of the end-cap toroid cryostats. The polyethylene is doped withB4C, which causes less out-gassing problems than other dopants. Photons created when theneutrons are absorbed by the boron are stopped by the aluminium of the cryostat itself.
– 44 –
The purpose of the two forward shielding assemblies is to protect the middle and outer end-capmuon stations from background particles created in secondary interactions in the beam-pipe, the1425
calorimeters and the TAS collimators. These shielding elements, which are removable and will bestored in the surface building during maintenance of ATLAS,consist of two parts: a cylindricalcore and a set of octagonal pieces in the rear. All pieces are made of cast ductile iron, surroundedby a layer of polyethylene doped with boron in the form ofH3BO3 and followed by a 3 cm thicksteel layer. The core pieces are enclosed in a 5 cm thick polyethylene layer, while a 8 cm thick1430
layer surrounds the octagonal pieces. These polyethylene layers are made of 10,000 bricks of threedifferent shapes.
The final shielding element, or nose shielding as depicted inFig. 17, supports the TAS col-limator and protects ATLAS from the radiation created in this collimator, which is designed toprevent the first LHC quadrupole from quenching due to the energy deposited by the particles1435
emerging from the interactions in ATLAS. The nose shieldingis permanently installed in ATLASand, unlike the forward shielding assemblies, cannot be removed during shutdowns. The maincomponent of this shielding is the cylindrical 117-tonne heavy "monobloc", which has an outer di-ameter of 295 cm. It is made of cast iron and supported by a tube, which is anchored in a 460 tonneconcrete structure. The 200 tonne heavy "washers", which are located around the support tube,1440
increase the radial thickness of the iron shielding by 112 cmin a region where the monobloc isthin.
3.3 Calculation of particle fluences and absorbed doses
A vast and systematic effort has been made in the design phaseto optimise the shielding in ATLASby using different simulation programs [37] and simulatinghundreds of different geometrical op-1445
tions. These studies have required significant computing resources, since the secondary particlesin the hadronic showers had to be followed down to very low energies. Different event generatorsand transport codes have been used in an attempt to assess thesystematic uncertainties in the cal-culations. When optimising the shielding configuration andmaterials in the limited space availablein ATLAS, it was very often necessary to make trade-offs between different background types, e.g.1450
neutrons versus photons. It has therefore been quite important to also understand the detector re-sponse to different types of background radiation, typically particles in the MeV range, in order toconverge to the optimal solution [38], [39].
The expected particle fluences (integrated over energy) agree to typically better than 20%, aswas shown by comparing two of the most commonly used minimum-bias event generators, namely1455
PHOJET1.12 [40, 41, 42] and PYTHIA6.2 [43]. Larger differences of up to 50% were observed forpions, kaons, and muons with energies above several GeV. However, these particles provide onlya small contribution to the total fluence. The program most used for the shielding optimisationin ATLAS has been the GCALOR package [44], which contains theCALOR code [45] with aninterface to GEANT3 [46]. FLUKA2001 [47] is another transport code, which has been and is1460
widely used for studies of hadronic and electromagnetic cascades induced by high-energy particles,and which has been extensively used in simulations of background radiation in ATLAS. In orderto investigate transport-code differences, GCALOR was compared not only to FLUKA but alsoto MARS14(2002) [48]. Different comparisons for simplifiedgeometries as well as for the mostdetailed descriptions of the ATLAS experiment have been carried out.1465
– 45 –
The results of these studies are extensively reported in [37]: the overall conclusion is that thepredictions of FLUKA, MARS and GCALOR are in good agreement for energy-integrated neutron,charged hadron, photon ande+e− fluences. For most regions in the inner detector, the differencebetween the FLUKA and GCALOR values is below 40%. In the pixelvertexing layer differencesas large as 80% are however observed for charged hadrons. An excellent agreement, typically1470
to within 20%, between the respective photon and neutron fluences in the muon spectrometer isobserved when comparing the FLUKA and the GCALOR results. The charged hadron and leptonfluences in the muon spectrometer show much larger discrepancies, but the differences are alwayswithin a factor of 2.5 (an overall safety factor of 5 has been used in the design of the ATLAS muonspectrometer [49]).1475
The absorbed dose is the mean energy deposited per unit mass,taking into account all energy-loss mechanisms (but corrected for rest-mass effects). Thedominant energy-loss mechanism is usu-ally ionisation, but non-ionising energy loss is also important for understanding detector and elec-tronic damage effects. The ionising dose is defined in the following as the integrated dE/dx energyloss in the detector material from charged particles, excluding ionisation energy loss from nuclear1480
recoils. It is given in units of Gy/year, where one year corresponds to 8× 1015 inelastic proton-proton collisions (assuming an inelastic cross section of 80 mb, a luminosity of 1034 cm−2 s−1 anda data-taking period of 107 s). Comparisons of the calculated ionising dose in the innerdetectorbetween FLUKA and GCALOR show differences of up to a factor oftwo.
3.3.1 The inner detector and calorimeter regions1485
Figure 18 shows a GCALOR calculation of the ionising dose in the region closest to the interac-tion point. The forward calorimeters (FCal) will be exposedto up to 160 kGr/year, whereas thecorresponding number for the end-cap calorimeters is 20 kGr/year. This will lead to very largeintegrated doses over the full lifetime of the experiment and is one of the main reasons why onlythe LAr technology with its intrinsically high resistance to radiation is used in the end-cap and for-1490
ward regions. The main concern in the design phase has been for the electrode materials, primarilypolymers such as polyimide, which had to be chosen with care and thoroughly tested for radiationhardness [50, 51].
The tile calorimeter, with its scintillator samplings readout by wavelength-shifting fibres, isprotected by the LAr electromagnetic calorimeter and is exposed to less than 15 Gr/year, i.e. 10,0001495
times less than the forward calorimeters. The scintillators and fibres were nevertheless also thor-oughly studied under irradiation [52, 53, 54, 55] in order todetermine their degradation during thelifetime of ATLAS.
In the inner detector, a very large effort had to be devoted over many years to the understandingof the impact of irradiation on silicon sensors, on front-end electronics circuits and on ageing1500
phenomena in the ionising gas used for the straw tubes.Two main mechanisms lead to the degradation of the performance of silicon devices under
irradiation. First, there is the effect of damage to the devices due to ionising energy loss. This canlead to the creation of trapped charges, in particular in theoxide layer of the sensor, which alters itselectric properties. The second effect is bulk damage, or displacement damage, which is caused by1505
the displacement of silicon atoms in the lattice. In the study of bulk damage to silicon devices, it isuseful to introduce a quantity called the 1 MeV neutron equivalent fluence (NEF). This fluence is
– 46 –
Figure 18. The total ionising dose per year calculated by GCALOR (see text) in one quarter of the centralpart of the experiment. The locations of the inner detector sub-systems, of the different calorimeters andof the inner end-cap muon stations are indicated. The scale on the left gives the integrated dose per yearcorresponding to the various iso-lines.
obtained by convoluting the various particle energy spectra and fluences with silicon displacement-damage functions, normalised to the value of the damage function for 1 MeV neutrons [56].
Table 8 lists the particle rates, NEF values and ionising doses predicted by FLUKA in the inner1510
detector regions shown in Fig. 18. In the pixel detector, theparticle fluence is dominated by chargedpions and photons. The latter are produced mostly in neutroncapture processes but also directlyfrom the primary collisions and from interactions in the beam-pipe and its related equipment. Thepredicted ionising dose in the innermost layer of the barrelpixel detector is 160 kGr/year, i.e., thesame as for the forward calorimeter, while the NEF is expected to be 3× 1014 cm−2/year. In1515
the SCT detector, the charged hadron and neutron fluences arecomparable and the NEF and theionising dose are reduced by about a factor of 20 with respectto the first pixel layer.
While most of the charged hadrons originate from the interaction point, most of the neutronsin the inner detector are the result of albedo (backsplash from the calorimeters). The purpose ofthe moderator shielding described in section 3.2 is to moderate the neutrons from the end-cap and1520
forward calorimeters to lower their energies to values for which their contribution to the total NEFis minimised. The polyethylene in the moderator shielding is doped with boron, which has a large
– 47 –
cross-section for the capture of thermal neutrons. Nevertheless, the inner detector cavity will befilled during LHC operation by an almost uniform "gas" of thermal neutrons with a fluence of 1–2 MHz/cm2 and the sensitive detectors will be exposed to fluences of 2–10 MHz/cm2 of low-energy1525
photons originating from the interactions themselves and from neutron capture. The dominantlong-term impact of these particle fluences is not only radiation damage but also activation of thedetector components (see section 3.5).
Table 8. Particle fluences and doses in key locations of the inner detector sub-systems (see Fig. 27 forthe definitions and positions of the inner detector layers).The FLUKA program has been used for thiscalculation and the statistical uncertainties are typically less than 10%.
Particle rates (kHz/cm2) NEF Ion. dose
Region R (cm) γ Protons Neutrons π± µ± e− (10−12 cm−2/y) (Gy/y)
> 30 keV > 10 MeV > 100 keV > 10 MeV > 10 MeV > 0.5 MeV
Pixel layer 1 5.05 45800 2030 4140 34100 300 8140 270 158000
Pixel layer 3 12.25 9150 280 1240 4120 190 1730 46 25400
SCT barrel layer 1 29.9 4400 80 690 990 130 690 16 7590
SCT barrel layer 4 51.4 3910 36 490 370 67 320 9 2960
SCT end-cap disk 9 43.9 7580 73 840 550 110 470 14 4510
TRT outer radius 108.0 2430 10 380 61 7 53 5 680
3.3.2 The muon spectrometer region
The effects of the absorbed ionising dose in the most critical muon spectrometer regions have been1530
studied [57]. The CSCs in the inner end-cap stations will be exposed to the highest dose. Figure 18shows that in this region the ionising dose will vary between3 and 20 Gr/year. The chambersclosest to the beam-line in the middle end-cap stations are expected to see at most 10 Gr/year.Most of the muon spectrometer will, however, be exposed to less than 1 Gr/year.
Although care had also to be applied to the choices of materials, to the design of the front-end1535
electronics circuits and to the choice of the ionising gasesfor all the muon chamber technologies,radiation damage due to the ionising dose has not been the only concern in the muon spectrometerregion. Signals in the detectors from background particleshave in reality been the main issue,because these background signals may significantly reduce the muon track-finding efficiency and,more importantly, introduce large rates of fake triggers. The expected particle fluences based on1540
a simulation of the background radiation in the ATLAS muon spectrometer using the GCALORprogram are shown in Fig. 19. The highest fluences are expected in the innermost end-cap muonstations, in particular in the CSCs, which will therefore have to cope with a large backgroundcounting rate, subject in addition to significant systematic uncertainties in the calculation.
Background hit rates caused by neutrons and photons in the relevant energy range have been1545
measured at the gamma-irradiation facility at CERN and havebeen calculated with GEANT3 forall muon-chamber technologies. The detailed geometry description of the muon chamber setup inthe cavern has been used. Energy-dependent efficiency curves have been estimated for neutrons,
– 48 –
Figure 19. Particle fluences in the various muon spectrometer stationsat high luminosity (1034 cm−2 s−1) aspredicted by GCALOR. The neutron and photon fluences are in units of kHz/cm2 and the muon and protonfluences in Hz/cm2.
photons, and electrons [37]. Calculations have been done for the various chamber types takinginto account the angular distributions of the particles at the chamber locations. Fluxes tend to1550
be isotropic in the barrel, while in the end-cap a substantial fraction of the particles originatesfrom the interaction region and from the beam-pipe in the region of the end-cap toroids, whichis the main local source of secondary radiation. The fake L1 trigger rate in the presence of thesebackground hits was studied in simulation including large contingency factors to account for thevarious uncertainties in the predictions.1555
The single-plane efficiency curves have been compared between existing experimental dataand simulation and found to be in good agreement. Average single-plane chamber efficiencieshave been obtained by folding efficiency curves with the energy spectra predicted at each chamberlocation. Uncertainties due to the shape of the energy spectrum, the angular distribution, or thesurrounding material have been studied and amount to a factor of 1.5. Predicted counting rates in1560
the barrel stations are of the order of 10–12 Hz/cm2 for both the MDTs and the RPCs. These ratesare dominated by the photon contribution (80%), followed byneutrons and protons (10% each).In the inner barrel stations, the contribution from muons rises to about 15% and that from punch-through pions to a few percent. In the end-cap regions, photons contribute less to the counting rate.In the CSCs for example, photons account for about half of therate, while muons account for 30%1565
and protons for 10%. The predicted single counting rates in the muon spectrometer are summarisedin Fig. 20.
– 49 –
Figure 20.Average expected single-plane counting rates in Hz/cm2 at 1034 cm−2 s−1 and for various scoringregions in the muon spectrometer.
3.4 Background monitors
Measurements of particle fluences in ATLAS will provide a precise bench-marking of the particletransport codes used in the calculations and will also directly monitor the absorbed doses in the1570
various detectors. Large-scale projects such as ATLAS haverelied heavily on these transport codesand future high-energy physics projects such as the LHC machine upgrade or the internationallinear collider (ILC) will also rely on them. In addition, possible beam losses near the experimentshould be monitored with specific detectors designed to provide fast feedback to the acceleratoroperations team. The motivation for equipping ATLAS with a reliable set of background monitors1575
in various regions of the detector is therefore obvious.
3.4.1 Monitors in the inner detector
The inner detector region of ATLAS contains a set of small detectors, which are sensitive to dose,to the 1 MeV neutron equivalent fluence (NEF) and to thermal neutrons. These detectors consistof:1580
1. Field-effect transistors (RADFETs), which measure the total ionising dose;
2. PIN-diodes, which measure NEF;
3. Radiation-hardened transistors, which measure thermalneutron fluences.
– 50 –
These detectors will measure the integrated doses and fluences in the inner detector and will alsoto some extent provide bench-marking estimates of the different contributions (charged particles,1585
neutrons and photons).
One of the worst-case scenarios during LHC operation arisesif several proton bunches hit thecollimators in front of the detectors. While the accumulated radiation dose from such unlikely ac-cidents corresponds to that acquired during a few days of normal operation, and as such providesno major contribution to the integrated dose, the enormous instantaneous rate might cause detector1590
damage. The ATLAS Beam Conditions Monitor (BCM) [58] systemconsists of a set of detectorsdesigned to detect such incidents and trigger an abort in time to prevent serious damage to the de-tector (see also section 8.10). Stray protons and beam-gas backgrounds frequently initiate chargedparticle showers, originating well upstream (or downstream) from the ATLAS interaction point.Thanks to their very fast response time and intrinsically very high resistance to radiation, the BCM1595
detectors will be used throughout the lifetime of the experiment to distinguish stray beam particlesfrom those originating from proton-proton interactions.
The BCM system consists of two stations, each with four modules (see Fig. 21 for a close-up view of one station installed around the beryllium beam-pipe), designed to tolerate doses ofup to 500 kGr and in excess of 1015 charged particles per cm2 over the lifetime of the experi-1600
ment. The stations are located symmetrically around the interaction point atz = ± 184 cm andR = 55 mm, which corresponds to a pseudorapidity of 4.2. Each module, as depicted in Fig. 22,includes two radiation-hard diamond sensors [59, 60] read out in parallel by radiation-tolerantelectronics with a 1 ns rise-time [61]. The difference in time-of-flight between the two stations,∆t,distinguishes particles from collisions (∆ t = 0, 25, 50 ns, etc.) from those arising from lost1605
protons (∆ t = 12.5, 37.5 ns, etc.). The in-time and out-of-time multi-module coincidences aredetermined by an FPGA-based back-end, which digitises the signals, monitors the detector perfor-mance and generates beam-abort signals if warranted [62]. Preliminary analysis of data on one ofthe modules in a high-energy pion test-beam shows a signal-to-noise of 14± 2 in an operationalgeometry, where minimum ionising particles are incident onthe BCM sensors at a 45 angle. A1610
full description of the design, construction and test-beamcharacterisation of the BCM system canbe found in [58].
Figure 21. Close-up view of one BCM station in-stalled at 1.8 m from the centre of the pixel detector,which can be seen at the far end of the picture. Eachone of the four modules can be seen in position at aradius of 5.5 cm very close to the beam-pipe.
Figure 22. Top view of a BCM module, showingthe diamond sensors (left), the HV supply and signal-transmission lines, the two amplification stages andthe signal connector (right).
– 51 –
Figure 23. Picture of one set of background monitors, to be installed inthe TGC layer of the middle end-capmuon station. The eight different types of detectors are described in the text.
3.4.2 Monitors in the muon spectrometer
Several sets of detectors have been installed in the end-capmuon stations to monitor the back-ground fluences and thus to constrain further the particle transport codes used in the calculations1615
described in section 3.3. These detectors are installed in the inner, middle, and outer end-cap sta-tions. Figure 23 shows one set of the detectors which have been installed. They were chosen toprovide a reliable response to neutrons or photons in various energy ranges:
1. Boron-lined proportional tubes operating with Argon/CO2 gas are used to measure thermaland slow neutrons (energies below 10−5 MeV). Each interactionn + 10B → Li + α sends1620
a slowLi or α -particle into the tube. The cross-section for this reaction is inversely propor-tional to the energy of the neutron. The large ionisation pulse associated with theLi or α -particle is used for pulse-height discrimination against Compton electrons and minimum-ionising particles. These detectors are therefore relatively insensitive to photons and chargedparticles.1625
2. Boron-loaded plastic scintillator (BC-454) is sensitive to the neutron interactions describedabove and is also used to study thermal and slow neutrons.
3. Detectors with a plastic disk loaded withLiF and coated with a thin layer ofZnS(Ag) scin-tillator are sensitive to the tritium andα -particles produced in the neutron capture process inlithium.1630
4. AnotherZnS(Ag) scintillator embedded in plastic is used to study fast neutrons (with ener-gies of a few MeV). The plastic is rich in hydrogen, from whichincoming neutrons scat-ter to produce recoil protons. These protons produce large ionisation pulses compared tominimum-ionising particles or low-energy electrons. Pulse-height discrimination schemesshould therefore provide good rejection against these backgrounds.1635
– 52 –
5. A liquid scintillator, with pulse-shape discriminationelectronics, is used in combination withplastics to measure fast neutrons.
6. Scintillator detectors with NaI and LSO crystals are usedto measure the low-energy photonspectrum (0 to 10 MeV). The spectrum is dominated by photons,but also contains a neutroncomponent, which can be separated out using fitting techniques and detailed simulations.1640
7. The total ionising dose is measured by small ionisation chambers.
3.5 Activation
Induced radioactivity will be a major problem at the LHC, andATLAS is the experiment with thehighest levels of induced radiation. This is due to the smallradius of the ATLAS beam-pipe, thesmall bore of the forward calorimeters, and to the shieldingelements close to the beam-pipe. A1645
comprehensive study has been made of the expected activation in different regions and for differentdata-taking and cooling-off scenarios. The methods and assumptions used in the calculation of theinduced activity are given in [37]. The main conclusion of theses studies is that the beam-pipe willbe the major source of radiation in ATLAS.
QUADTAS
ForwardToroid
Toroid
ForwardToroid
LAr Cal.
HAD
FCAL
Muons
100-2000
50-100
10-50
Sv/hSv/h
Rails
Sv/h
Sv/h
HAD CAL.
EM CAL.
Inner Detector
Shielding
Barrel Toroid Coil
NoseShield.
OuterMuons
MiddleMuons
a) Standard Access
Figure 24. The inner region of the ATLAS experiment during one of the main long-access scenarios. Thepredicted dose rates have been calculated for 10 years of operation at 1034 cm−2 s−1 and for five days ofcooling off. The standard access scenario (a) has the beam-pipe in place.
Two different access scenarios are foreseen for ATLAS during shutdowns, as described in1650
more detail in section 8.8. They are depicted in Fig 24 and Fig25 and described below:
(a) In the standard access scenario, the beam-pipe remains in place, but then acts as a linearsource of photon radiation as can be seen in Fig. 24. Because of the high level of radiation,the area around the beam-pipe, out to a radius of about one metre, has to be fenced off afterhigh-luminosity running. This will ensure that people working in ATLAS during standard1655
access will not be exposed to dose rates larger than 0.1 mSv/h. The only detector, which istruly inside the cage is the inner detector. During standardaccess, maintenance of the innerdetector will therefore be severely limited. With the beam-pipe in place, it will in any case
– 53 –
LAr Cal.
HADHAD CAL.
EM CAL.
FCALQUADTAS
Outer
Rails HF truck
100-500 Sv/h
10-50 Sv/h
50-100 Sv/h
Inner Detector NoseShield.
MuonsMiddleMuonsBarrel Toroid Coil
b) Inner Detector Access
Figure 25. The inner region of the ATLAS experiment during one of the main long-access scenarios. Thepredicted dose rates have been calculated for 10 years of operation at 1034 cm−2 s−1 and for five days ofcooling off. The inner detector access scenario (b) has onlythe inner detector section of the beam-pipe inplace. The expected dose rates are greatly reduced in the inner detector scenario.
be difficult to do extensive maintenance to the inner detector, so its very limited availabilityis not considered to be a problem.1660
(b) In a second access scenario, called the inner detector access scenario, all the beam-pipesections except the one inside the inner detector volume areremoved as well as the smallmuon wheel (or inner end-cap muon stations) and the end-cap toroids. Two hot spots canclearly be seen in the final configuration, as shown in Fig. 25.One is the end-piece of theinner detector beam-pipe, which is made of aluminium, whereas the rest of the inner detector1665
beam-pipe is made of beryllium. The expected dose rate can reach 0.2 mSv/h at this location.The other hot spot is in front of the forward calorimeters, where the dose rate is predicted toreach very high values of up to 0.5 mSv/h. These relatively small-size regions will thereforebe temporarily shielded with lead blocks during maintenance of the inner detector.
While the beam-pipe section inside the inner detector is mostly made of beryllium, the rest1670
of the beam-pipe is made of stainless steel and has to be removed in the case of the inner detectoraccess scenario, since it will become very radioactive witha contact dose rate of 3–5 mSv/h. Thiscould in certain cases inflict several mSv of integrated doseto personnel performing the interven-tion. One way of reducing the dose to personnel would be to make the beam-pipe out of aluminiuminstead of stainless steel. This is expected to give a factor10–50 reduction of the dose levels. If1675
the beam-pipe material were instead to be changed to beryllium over the whole length of the ex-periment, the dose rate would decrease by a factor of 100–1000 and would no longer be a problem.This is, however, very costly and will only be discussed further in the context of the LHC upgradeprogramme.
– 54 –
4. Inner Detector1680
4.1 Introduction
The ATLAS Inner Detector (ID) is designed to provide hermetic and robust pattern recognition, ex-cellent momentum resolution and primary and secondary vertex measurements [63, 64] for chargedtracks above a givenpT threshold (nominally 0.5 GeV, but as low as 0.1 GeV in some ongoingstudies of initial measurements with minimum-bias events)and within the pseudorapidity range1685
|η | < 2.5. It also provides electron identification over|η | < 2.0 and a wide range of energies (be-tween 0.5 GeV and 150 GeV). This performance, which is required even at the highest luminositiesexpected from LHC collisions, is consequently at the limit of existing technology.
The ID layout, as shown in Fig. 26, reflects the performance requirements. The ID is containedwithin a cylindrical envelope of length±3.4 m and of radius 1.15 m, within a solenoidal magnetic1690
field of 2 T (see section 2.2.4). Figure 27 shows the sensors and structural elements traversed by10 GeV tracks in respectively the barrel and end-cap regions. The ID consists of three indepen-dent but complementary sub-detectors. The envelopes of each sub-detector are listed in Table 2(see section 1.1.3). At inner radii, high-resolution pattern recognition capabilities are available us-ing discrete space-points from silicon pixel layers and stereo pairs of microstrip (SCT) layers. At1695
larger radii, the transition radiation tracker (TRT), withmany layers of gaseous straw tube elementsinterleaved with transition radiation material, providescontinuous tracking, with typically 35 hitsper track, to enhance the pattern recognition and improve the momentum resolution. The TRTalso provides electron identification using signals from transition radiation photons absorbed in theXe gas over|η | < 2.0.1700
The high-radiation environment (see section 3.3.1) where the ID is located imposes stringentconditions on its sensors, on-detector electronics, mechanical structure and services over its 10-year design lifetime1. In this period, the layer 1 and layer 2 pixels must withstandan integratedfluence2 of ∼8×1014 neq/cm2. Inner parts of the SCT must withstand 1–2×1014 neq/cm2. Tomaintain an adequate noise performance after radiation damage, the silicon sensors must be kept at1705
low temperature (∼ −5 to−10C) implying coolant temperatures∼−25C . In contrast, the TRTis designed to operate at room temperature.
The above operating specifications imply mechanical requirements (Table 9) that serve asstrong upper limits on the silicon module build precision, the TRT straw tube position, and themeasured module placement accuracy and stability. This leads to:1710
a) a good build accuracy with radiation-tolerant materialshaving adequate detector stability andwell understood reproducibility following repeated temperature cycling between –20C and+20C, with a temperature uniformity on the structure and modulemechanics that minimisesthermal distortion;
b) an ability to monitor the position of the detector elements using charged tracks and, for the1715
SCT, laser interferometric monitoring;
1The pixel Layer−0 will be replaced after 3 years of operation at design luminosity.2Normalised using non-ionising energy loss (NIEL) cross-sections to the expected damage for 1 MeV neutrons.
– 55 –
Figure 26. Plan view of a quarter section of the ATLAS inner detector showing each of the major detectorelements with its dimensions.
Figure 27. Drawings showing the sensors and structural elements traversed by charged tracks of 10 GeVpT
in (a) the barrel (η=0.3) and (b) the end-cap (η=1.4 and 2.2). The barrel track traverses successively theberyllium beam pipe, 3 silicon pixel layers with individualsensor elements of 50×400µm2, 4 stereo layersof silicon microstrip sensors (SCT) of pitch 80µm and the 4 mm diameter straws of the transition radiationtracker with its support structure.
– 56 –
Item Intrinsic Alignment tolerances
accuracy (µm) (µm)
Radial Axial z Azimuth Rφ
Pixel
Layer 0 10 (Rφ) 115(z) 10 20 7
Layers 1 and 210 (Rφ) 115(z) 20 20 7
Disks 10 (Rφ) 115(R) 20 100 7
SCT
Barrel 17 (Rφ) 580(z)1 100 50 12
Disks 17 (Rφ) 580(R)1 50 200 12
TRT 130 (drift time) 302
1. Arises from±20 mrad stereo angle for back-to-back sensors on SCT modules
with axial (barrel) or radial (end-cap) alignment of one side on the structure.
Thez-resolution results from two axial-stereo points with opposite stereo directions.
The result is pitch dependent for end-cap SCT modules.
2. This alignment accuracy applies is related to the TRT drift-time accuracy.
Table 9. Intrinsic measurement accuracies and mechanical alignment tolerances for the ID sub-detectors asdefined by the performance requirements of the ATLAS experiment.
c) a trade-off between the low material budget needed for optimal performance and the sig-nificant material budget resulting from a stable mechanicalstructure with the services of ahighly granular detector.
The performance requirements imply a short-term stability, which adds negligibly (<20% when1720
added in quadrature) to the alignment precision.
This chapter describes the construction and early performance of the as-built inner detector.In section 4.2, the basic detector sensor elements are described. Section 4.3 describes the de-tector modules, section 4.4 details the readout electronics of each sub-detector and section 4.5describes the detector power and control. Section 4.6 discusses the mechanical structure for each1725
sub-detector, as well as the integration of the detectors and their cooling and electrical services. Theoverall ID environmental conditions and general services are briefly summarised in section 4.7.Section 4.8 catalogues the material budget of the ID, which is significantly larger than that ofprevious tracking detectors. Finally, section 4.9 indicates some initial results on the operationalperformance.1730
– 57 –
4.2 Inner detector sensors
This section describes the detector sensors of the pixel, SCT and TRT sub-detectors - silicon pixeland micro-strip sensors, and Xe-CO2 filled straw tubes. As discussed in section 3.3, the detectorsensors are subject to large integrated radiation doses. They have therefore been developed andcontrolled to withstand the expected irradiation, with a safety factor of approximately two.1735
4.2.1 The Pixel and SCT detector sensors
The pixel and SCT sensors [65, 66] are required to maintain adequate signal performance over thedetector lifetime at design luminosity. The integrated radiation dose has important consequencesfor the sensors of both detectors. In particular the required operating voltage, determined by theeffective doping constant, depends on both the irradiationand the subsequent temperature sen-1740
sitive annealing. The sensor leakage current also increases linearly with the integrated radiationdose. The n-type bulk material effectively becomes p-type after ∼ 2×1013 neq/cm2 radiation. Theeffective doping constant then grows with time in a temperature-dependent way. To contain thisannealing and to reduce the leakage current, the sensors will be operated in the range -5C to -10C. The sensors must further meet major geometrical constraints on the thickness, granularity1745
and charge collection efficiency.The pixel sensors required the most leading-edge and novel technology to meet the very strin-
gent specifications on radiation hardness, resolution and occupancy in the innermost layers. Thesensors are 250µm-thick detectors, using oxygenated n-type wafers with readout pixels on the n+-implanted side of the detector. Despite its larger cost and complexity, this novel design involving1750
double-sided processing was used because:
a) the n+ implants allow the detector to operate with good charge collection efficiency aftertype inversion, even when operated below the depletion voltage, because the depletion zonegrows from the pixel side.
b) highly oxygenated material has been shown to give increased radiation tolerance to charged1755
hadrons, with improved charge collection after type inversion and lower depletion voltage;
All of the 1744 pixel sensors are identical. The pixels of each pixel sensor (external dimen-sion 1.9×6.3 cm2) have a minimum pixel size of 50×400µm2, dictated by the readout pitch of thefront-end electronics. There are 47232 pixels on each sensor, but for reasons of space there are fourganged pixels in each column of the front-end chip, thus leading to a total of 46080 readout chan-1760
nels. A common bias grid ensures DC-biasing to all pixels. Toguarantee optimal post-irradiationperformance, a p-spray insulation technology has been used[67]. Each pixel of a sensor is bump-bonded through a hole in the passivation layer to an element of the front-end readout integratedcircuit as part of the module.
For reasons of cost and reliability, the 15912 sensors of theSCT use a classic single-sided p-1765
in-n technology with AC-coupled readout strips. The sensors will initially operate at∼150 V biasvoltage, but operating voltages of between 250 and 450 V willbe required for good charge collec-tion efficiency after 10 years of operation, depending on thesensor radius, the integrated luminosityand the length of warm-up periods. The sensor thickness of 285±15 µm is a compromise betweenthe required operating voltage, the primary signal ionisation and the simplicity of fabrication. The1770
– 58 –
Sensor Cut length Outer Inner Strip Interstrip
type (mm) width (mm) width (mm) pitch (µm) angle (µrad)
Barrel 63.960 63.560 63.560 80.0 0
EC W12 61.060 55.488 45.735 56.9–69.2 207.0
EC W21 65.085 66.130 55.734 69.9–83.0 207.0
EC W22 54.435 74.847 66.152 83.4–94.2 207.0
EC W31 65.540 64.635 56.475 70.9–81.1 161.5
EC W32 57.515 71.814 64.653 81.5–90.4 161.5
Table 10.External cut dimensions of the SCT barrel and end-cap (EC) sensors. The tolerance on all externaldimensions is± 25 µm; the mask accuracy is at the level of± 1 µm. The inter-strip angle is that betweenadjacent strips of the sensor. The sensors are fabricated from 4-inch wafers.
strip pitch was determined by the required digitising precision, granularity, particle occupancy andnoise performance. A strip pitch of 80µm with two 6 cm sensors daisy-chained was chosen forthe rectangular barrel sensors and radial strips of constant azimuth with mean pitch∼80 µm werechosen for the trapezoidal end-cap sensors. There are a total of 768 active strips per sensor, plustwo strips at bias potential to define the sensor edge. The detector dimensions are summarised in1775
Table 10 (see also Table 12). Using the binary readout electronics described in section 4.4, a noiseoccupancy per channel of<5×10−4 for a threshold of 1 fC is specified, corresponding to a 1800 e−
rms noise for fully irradiated modules. The detectors were specified to have<1% of bad readoutstrips at 350 V bias before and after irradiation to 3×1014 24 GeV protons/cm2, and to operatestably at 500 V bias.1780
All sensors were carefully tested, and a sub-sample were subjected to extended pixel or strip-by-strip studies [68]. A sub-sample was also used for extensive post-irradiation performance stud-ies [69, 70]. Apart from precautions related to the humiditysensitivity noted above, the rejectedstrip sensors were at the level of 1%. The pixel sensor rejection rate was somewhat higher becauseof a requirement to carefully control the profile and doping concentration of the p-spray insulation.1785
4.2.2 TRT straw tubes
Polyimide drift (straw) tubes of 4 mm diameter are the basic TRT detector elements [71]. Thestraw tube wall, especially developed to have good electrical and mechanical properties with min-imal wall thickness, is made of two 35µm thick multilayer films bonded back-to-back. The barematerial, a 25µm thick polyimide film, is coated on one side with a 0.2µm Al layer that is pro-1790
tected by a 5–6µm thick graphite-polyamide layer. The other side of the film is coated by a 5µmpolyurethane layer used to heat-seal the two films back-to-back. Mechanically, the straws are sta-bilised using carbon fibres. After fabrication, the tubes were cut to length (144 cm for barrel tubesand 37 cm for the end-caps) and leak-tested at 1 bar over-pressure. The straw (cathode) resistance
– 59 –
was required to be<300Ω/m.1795
The straw anodes are 31µm diameter tungsten (99.95%) wires plated with 0.5-0.7µm gold,supported at the straw end by an end plug. They are directly connected to the front-end electronics(see section 4.4) and kept at ground potential. The anode resistance is approximately 60Ω/m andthe assembled straw capacitance is<10 pF. The signal attenuation length is∼4 m and the signalpropagation time∼4 ns/m. The cathodes are operated at typically -1530V to givea gain of 2.5×104
1800
for the chosen gas mixture of 70% Xe, 27% CO2 and 3% O2 with 5-10 mbar overpressure3. Withthese operating conditions, the maximum electron collection time is∼48 ns and the operationaldrift time accuracy is∼130µm. Low energy transition radiation (TR) photons are absorbed in theXe-based gas mixture, and yield much larger signal amplitudes than charged particles (MIPs). Thedistinction between TR and tracking signals is obtained on astraw-by-straw basis using separate1805
low- and high- thresholds in the front-end electronics.
For the barrel straws, the anode wires are read out from each end. In the centre, the wiresare split electrically by a fused glass capillary of 6 mm length and 0.254 mm diameter to reducethe occupancy. In the inner nine layers of type-1 barrel modules (see section 4.3.3), the wires aresub-divided into three segments keeping only the 31.2 cm-long end-segments active.1810
To guarantee stable operation, the wire offset with respectto each straw centre is required tobe<300µm. This is essentially a requirement on straw straightness since the wire sag is<15 µm.Wires with larger offsets have been disconnected in the finalbarrel and end-cap acceptance tests.To maintain straw straightness in the barrel, alignment planes with a matrix of 4.3 mm diameterholes are positioned each 25 cm along the module.1815
The stable operation of TRT straws with the Xe-based gas mixture requires a re-circulating gassystem with continuous monitoring of the gas quality. To avoid pollution from permeation throughthe straw walls or through leaks, the straws are operated in an envelope of CO2.
At LHC rates, significant heat is generated in the straws as positive ions created by the ionisingparticles drift to the cathode. The heat dissipation is proportional to the single straw counting rate1820
and is estimated to be 10mW to 20mW per straw at the LHC design luminosity. For gas gainuniformity the gradient along each straw is required to be<10C. The heat is evacuated differentlyfor the barrel modules and end-cap wheels (see section 4.3.3).
At the nominal LHC luminosity, the straw counting rate will reach 20 MHz in the most criticaldetector regions, and the ionisation current density will reach 0.15µA per cm of anode wire.1825
The total accumulated charge after 10 years of operation will reach∼10 C/cm in some straws.Many studies, including direct ageing tests lasting hundreds of hours indicate satisfactory strawtube operation during their operational lifetime. Minute levels of pollution cannot be excluded andorgano-silicone impurities, for which the relative concentrations must be kept below 10−11, areparticularly harmful. A gas filter is installed for this reason. The filter is also effective in removing1830
the ozone produced during gas amplification. In case ageing affects the detector performance, theuse of a Ar/CO2/CF4 gas mixture during two days of normal LHC operation has been shown toclean Si-based deposits on the anode wire.
3A mixture 70% Ar, 30% CO2 has been used during quality control and cosmic ray studies.
– 60 –
4.3 Inner detector modules
4.3.1 Pixel modules and staves1835
There are 1744 modules in the pixel detector. A schematic view and photograph of a moduleare shown in Fig. 28. A pixel module consists of a stack, from the bottom up, of the followingcomponents:
a) 16 front-end electronics chips, each with 2880 electronics channels;
b) bump bonds (In or PbSn), which connect the electronics channels to pixel sensor elements;1840
c) the sensor tile 63.4 x 24.4 mm2, approximately 250µm thick;
d) a flexible polyimide circuit board (flex-hybrid) with a module-control chip glued to the flex-hybrid;
e) a Polyimide pig-tail with Cu lines and a connector (barrelmodules) or a wire micro-cable(end-cap modules) bonded to the flex-hybrid.1845
Figure 28. (a) Schematic view of a barrel pixel module and (b) photograph of a barrel pixel module.
The sensors and electronics chips are connected by bump bonding technology to form baremodules. Both solder (PbSn) and indium bump bonding technologies have been used to make pixel
– 61 –
Indium PbSn solder Total
Number % Number % Number %
Module starts 1468 1157 2625
Modules accepted after bump-bonding1296 1122 2418
Modules accepted after flex-hybrid glued1190 100 1122 100 2312
Modules accepted for staves 1025 86.1 1075 95.8 2100 90.8
Module sub-set acceptable for layer 0 281 23.6 445 39.7 726 31.4
Table 11. Yield of pixel modules after bump-bonding of the electronics channels to the pixels. A sub-setof the highest quality modules were selected for the layer 0.Two materials for bump-bonding were used:indium and lead-tin solder balls.This table will be simplified or left to the back-up paper.
modules. Including reworked modules, the production statistics of bare modules are summarisedin Table 11. Flex hybrids with attached and tested module-control chips are glued to accepted baremodules. In total, 2312 modules were available for final electrical and mechanical characterisation.1850
After construction, the pixel modules were tested electrically at room temperature and at−10C,the approximate operating temperature. Thermal cycling was performed on each module prior tocompletion of electrical testing. A ranking was made to separate the modules into those acceptablefor the Layer-0 radius (highest quality) and for the outer barrel and disk regions. Modules withthe best ranking were loaded on barrel staves and disk sectors, leaving modules with the poorest1855
ranking as spares. A brief summary of the production statistics and ranking after testing is given inTable 11.
In the barrel region, 13 pixel modules are mounted on each stave using robotic tools and thenglued (Fig. 29a). The staves are themselves mounted on carbon-fibre structures (see section 4.6.1).A bare stave consists of machined carbon-carbon (C-C) plates, an aluminium cooling tube and1860
a carbon-fibre composite piece that captures the aluminium tube and is glued to the C-C pieces.The C-C pieces are precisely machined in a step pattern and are one-half the length of a stave.They are joined in the middle during the stave fabrication. Acustom extrusion was used to makean aluminium tube with a flat surface at the interface with theC-C material. A custom fitting iswelded to the end of the aluminium pipe. A thermal compound isused to conduct heat between the1865
C-C and the tube. The tube is held in place by a carbon-fibre piece glued to the C-C. The electricaland thermal performance of each stave was measured after assembly. Custom low-mass cables areconnected to each stave via a connector on each module and attached to the back of the stave toreach the ends of the stave. The staves are joined to form bi-staves (Fig. 29a), which form thecooling unit in the barrel region. A custom-welded aluminium U-link is attached to each bi-stave.1870
The end-cap equivalent of the stave is a sector. The two pixelend-caps each have three identi-cal disks. Each disk is composed of eight sectors. Six pixel modules are directly mounted on eachsector (Fig. 29b). The sectors are composed of thin, C-C faceplates with a rectangular aluminiumcooling tube and vitreous carbon foam between the faceplates. The cooling tube is bent into a W-
– 62 –
Figure 29. (a) Close up of a bi-stave loaded with modules. The insert shows the U-link cooling connectionbetween staves. (b) A pixel disk sector during attachment ofmodules. There are also three modules on theback of the sector.
like shape to fit within the sector and makes contact with the faceplates with a compliant, thermally1875
conducting adhesive. Each cooling circuit in the disk region serves two sectors.
Tested end-cap modules are positioned and glued on each sector with a precision of 1-2µm inthe plane of the module and about 10µm perpendicular to the module plane. The module locationson each sector were optically surveyed using fiducial marks in the corners of the pixel sensors andother information. The survey precision with respect to themounting bushings is estimated to be1880
better than 5µm in the plane of the module and about 15µm in the direction perpendicular to thisplane.
Add table describing pixel modules, power per module, yields of working modules, similar toTable 12.
4.3.2 SCT modules1885
The 2112 barrel SCT modules [72] use 80µm pitch micro-strip sensors [66] connected to binarysignal readout chips [73] described in section 4.2. The barrel module is shown, with its components,in Fig. 30. The module parameters are shown in Table 12. The 4 sensors, 2 each on the top andbottom side, are rotated with their hybrids by±20 mrad around the geometrical centre of thesensors. They are glued on a 380µm-thick thermal pyrolitic graphite (TPG) base-board, which1890
provides the thermal and mechanical structure. This extends side-ways to include beryllia facingregions. A polyimide hybrid [74, 75] with a carbon-fibre substrate bridges the sensors on each side.The two 770-strip (768 active) sensors on each side form a 128mm long unit (126 mm active witha 2 mm dead space). High voltage is applied to the sensors via the conducting base-board.
Precision alignment criteria were applied during assembly. The important in-plane tolerance1895
for positioning sensors within the back-to-back stereo pair was< 8 µm and the achieved variancewas 2µm. In the module plane, no distortions were measured after thermal cycling. Out-of-plane,the individual components and the assembly jigging and gluing determine the module thickness andthe intrinsic bow of the sensors determines the out-of-plane shape. A common distortion profile
– 63 –
Parameter Description
Strips 2 × 768 active strips,± 20 mrad stereo rotation
Nominal resolution17 µm in-plane lateral (Rφ)
580µm in-plane longitudinal (z or R))
Module dimension
Barrel Active length 126.09 mm + 2.09 mm dead space between sensors
Outer end-cap Active length 119.4 mm + dead space, radius 438.77−560.00 mm
Middle end-cap Active length 115.61 mm + dead space, radius 337.60−455.30 mm
Inner end-cap Active length 63.58 mm + dead space, radius 275.00−334.10 mm
Specified build Barrel back-to-back in plane:<8 µm (lateral)<20 µm (longitudinal)
tolerance End-cap back-to-back in plane:<5 µm (lateral)<10 µm (longitudinal)
Out of plane (module thickness and sensor bowing)<70 µm.
Module fixation points with respect to module centre<40 µm in plane.
Hybrid power 5.5−7.5 W
consumption
Sensor power Up to 460 V bias,<1W at -7C
consumption
Table 12.SCT barrel and end-cap module specifications.It is planned to add an achieved build accuracy ifconsistent numbers become available. A similar table is required for pixels..
has been established for the sensors at the level of a fewµm and a module thickness variation of1900
33 µm was maintained during fabrication. Following thermal cycling, the out-of plane distortionschanged by a fewµm RMS. When cooled from room to operating temperature, profile deviationsdid not exceed 20µm, even at the sensor corners not supported by the base-board.
Figure 31 shows the construction of an end-cap module [76]. There are three module types(Table 12). Each of the 1976 modules has two sets of sensors glued back-to-back around a central1905
TPG spine with a relative rotation of±20 mrad to give the required space-point resolution inR−φandR. The module thickness is defined by the individual components and variations are compen-sated by the glue thickness (nominally 90µm). The TPG spine conducts heat from the sensors tocooling and mounting points at the module ends and serves as the bias contact to the sensors. Glassfan-ins attach one end of the spine to a carbon base-plate with the polyimide flex-hybrid glued to1910
it. The modules are arranged in tiled outer, middle and innerrings.
The precision alignment criteria applied to the end-cap modules were similar to those of barrelmodules. The in-plane tolerance for positioning sensors within the back-to-back stereo pair was<
– 64 –
Figure 30. Photograph (a) and drawing (b) of a barrel module, showing its components.
Figure 31. (a) Photograph of the 3 end-cap module types (outer, middle and inner) with (b) an explodedview showing different components for a middle module.
5 µm transverse and 10µm longitudinally. The rms spread of the module survey measurementsafter construction was 0.6µm in the back-to-back position of the stereo pair measured transverse1915
to the strips and 2.8µm in the position of the mounting hole and slot measured transverse to thestrips. The mean offsets measured after construction were slightly larger. In the module plane, nodistortions were measured after thermal cycling. Out of theplane, the modules are less rigid, andare affected by variations of the spine thickness and bowingof the sensors. A common distortionprofile has been established for the sensors at the level of a few µm and a mean module thickness1920
variation of 27µm was maintained during fabrication. Following thermal cycling, the out-of planedistortions changed by only a fewµm rms.
The barrel and end-cap sensors are specified to operate at−7C, with a maximum variationwithin and between modules of 5C, to reduce the bulk leakage current after radiation damage. Thehybrid power will be 5.5− 7.5 W per module, and the sensor load will reach∼1 W per module1925
after 10 years of operation. In addition, a convective load of ∼0.8 W per module is expected withan additional convective load of∼0.8 W per module at the top of the barrels and outer disks. Theheat is extracted by evaporating C3F8 at ∼ −25C, circulating in cooling pipes attached to eachmodule.
– 65 –
For the barrel, the sensor and hybrid heat leaves via the base-board (respectively hybrid1930
substrate) to the large beryllia facing on the base-board, interfaced to an aluminium block witha ∼100 µm layer of thermal grease and a copper-polyimide capacitiveshunt shield. The blockis itself soldered to a 3.6 mm diameter Cu/Ni cooling pipe. Each cooling loop serves 48 barrelmodules.
For the end-cap, the sensor heat leaves via the spine, while the hybrid heat is transferred via the1935
carbon fibre hybrid substrate to a cooling block that is splitto avoid heat transfer between the sensorand hybrid. At full load, the ASIC and sensor temperatures are expected to be respectively∼30Cand∼10−15C above the coolant temperature. A layer of thermal grease isapplied between themodules and the cooling block. The blocks are soldered to a Cu/Ni cooling pipe that serves up to33 modules.1940
All modules were tested electrically at room temperature and at the operating temperature.Thermal cycling was performed on each module prior to completion of electrical testing. Morethan 99.9% of the strips operate satisfactorily.
4.3.3 TRT modules
The barrel TRT is divided into 3 rings of 32 modules each, supported at each end by a space frame1945
of the barrel support structure (see section 4.6). Each module consists of a carbon-fibre laminateshell and an internal array of straws embedded in a matrix of 19 µm-diameter polypropylene fibresserving as the transition radiation material. The straws, described in section 4.2, form a uniformaxial array with a mean spacing of∼7 mm. The modules are designed to minimise the dead regionfor high pT tracks. The main barrel parameters are shown in Table 13. Figure 32a shows a quadrant1950
of the TRT barrel during assembly. Aφ-slice showing single outer, middle and inner modules ishighlighted.
The dimensional specifications are set by the intrinsic straw R-φ resolution of 130µm, imply-ing that each straw position should be constrained to within40 µm and that the ends be specifiedto be within±50 µm. The module shell, 400µm thick high thermal conductivity carbon fibre flat1955
to within 250µm, is measured to satisfy maximum distortions of<40 µm under full load.
The module shells also serve as a gas manifold for CO2 that circulates outside the straws toprevent the accumulation of Xe due to possible gas leaks thatwould capture the transition radiationphotons. The heat dissipated by the barrel straws is transferred to the module shell by conductionthrough the CO2 gas envelope. Each module shell is cooled by two cooling tubes located in the1960
acute corners. These tubes also serve as return pipes for theC6F14 cooling circuits of the front-endelectronics.
The module end with its different components is shown in Fig.33a. The central element isthe HV plate that has stringent requirements on flatness and cleanliness to prevent discharges andon the straw feed-through accuracy to ensure mechanical precision of the straw location. Given1965
the machining tolerances and other accumulated effects, the variance of the wire centre position isexpected to be∼90 µm. The HV plates were individually surveyed after machining. The tensionplate of Fig. 33a mounted on the HV plate is a printed circuit board holding the wire ends (andensuring the wire tension) and providing electrical connections. It also closes the active gas volumeand serves as a Faraday cage for the active module elements.1970
– 66 –
|z|min |z|max Rmin Rmax Number Number Straws per Weight
(mm) (mm) (mm) (mm) modules layers module (kg)
Barrel 0 780 554 1082 96 73 52544 700
Module Type 1 inner400 712.1 563 624 32 9 329 1.2
Module Type 1 outer7.5 712.1 625 694 10
Module Type 2 7.5 712.1 697 860 32 24 520 2.2
Module Type 3 7.5 712.1 863 1066 32 30 793 3.1
End-cap 827 2744 615 1106 20 160 122880 1117
Type-A wheels 848 1705 644 1004 12 8 6144 24
Type-B wheels 1740 2710 644 1004 8 8 6144 36
Table 13. Parameters of the TRT barrel and of one TRT end-cap. The quantities shown in bold are globalparameters including services and electronics. All other quantities are for individual modules and the activeregion. Type-1 modules include two straw types as describedin the text.
Mechanical and electrical tests on the modules and individual straws were made at succes-sive assembly stages and following delivery of the modules to CERN. The module lengths wererequired to be in the range 1461.5-1462.9 mm and the twistingof individual modules was requiredto be< 1 mm. Stringent criteria were applied to the gas tightness, wire tension, straw straightness,and high voltage stability. The modules installed in the barrel had 98.5% operational channels.1975
The TRT end-caps each consist of two sets of independent wheels (Fig. 32b and Table 13).The set closer to the interaction point contains 12 wheels, each with eight successive layers spaced8 mm apart. The outer set of wheels contains eight wheels, also with eight straw layers but spaced15 mm apart. Each layer contains 768 radially oriented straws of 37cm length with uniform az-imuthal spacing. The space between successive straw layersis filled with layers of 15µm-thick1980
polypropylene radiator foils separated by a polypropylenenet.
Each eight-layer wheel consists of two basic four-layer assembly units. To assemble a four-plane wheel, straws were inserted and glued into precisely drilled holes in grounded inner and outercarbon fibre rings. The rings and the straws constitute the main mechanical structure of the wheels.The successive straw layers, interleaved by the radiators,are rotated from one layer to the next by1985
3/8 of the azimuthal straw spacing in a given layer. For high pT tracks from the interaction point,this ensures optimal uniformity in the number of crossed straws, which varies radially from∼6to ∼4 across an 8-layer wheel.
Two-layer flex-rigid printed-circuit boards provide high-voltage and signal connections to theend-cap wheels (Fig. 33b) . Each of the flexible layers has conducting paths on one side connecting1990
to the rigid part of the board. To provide a reliable electricconnection, petals in the high-voltagelayer are forced into contact with the inner straw wall through the insertion of a plastic plug. Apress-fit between petals in the signal layer and a metallic crimp tube position and fix the anode
– 67 –
Figure 32. (a) Photograph of one quarter of the barrel TRT during assembly. The shapes of single outer,middle and inner TRT modules are highlighted (in red). The barrel support structure space-frame is alsoshown. (b) Photograph of a TRT 4-plane end-cap wheel.
Figure 33. a) A detail of the end of the TRT barrel modules. b) A detail of the straw connection schemeused for end-cap TRT straws.HV plate to be explicitly highlighted in (a). Figure (b) to beimproved.
wires. There are 32 such boards per 4-layer wheel, each serving aφ-sector of 96 straws. Eachsector is further segmented into three groups of 32 readout channels and 12 high-voltage groups1995
of eight straws sharing a common fuse and blocking capacitor. The carbon-fibre ring holding thestraws and flex-rigid boards, together with a third carbon fibre ring and a simpler glass fibre boardprovide a rigid structure around the outer wheel perimeter that also serves as a gas manifold. Theinner gas manifold is made from reinforced polyimide material.
The heat dissipated by the end-cap straws is evacuated through the CO2 gas envelope that2000
is forced to flow along the straws from the inner to outer radius. Each group of wheels has itsown CO2 cooling circuit, passing the gas sequentially through all the wheels of the group. Heatexchangers cooled with C6F14 extract heat from the gas between adjacent wheels. The high flow
– 68 –
rates required - 50 (30) m3/hr for the inner (outer) wheels - necessitates a closed-loop systemcapable of maintaining a stable (±10 mbar) gas pressure inside the detector.2005
As for the barrel modules, all end-cap wheels passed qualitycontrol procedures during con-struction and after delivery to CERN, resulting in>99% fully operational channels.
4.4 Readout of the inner detector
The readout architecture of the ID is optimised separately for each of the three sub-detectors, butis characterised by the following common elements:2010
a) the reception of a 40.08 MHz clock signal synchronous withthe LHC bunch-crossings usedto time-stamp the signal generated in low noise front-end electronics;
b) signal generation in the front-end electronics and storage in binary or digital buffers forapproximately∼3.2µsec, compatible with the L1 trigger latency of 2.5µsec;
c) following a L1 trigger, the subsequent transfer of the buffer content associated with the2015
bunch-crossing or possibly several bunch-crossings to a readuut driver (ROD) off the de-tector.
The readout of the pipelines is subject to the reception of a L1 trigger signal from the calorime-ters or the muon detectors via the central trigger processor, as discussed in section 7.2.3. The IDis not part of the L1 trigger. External power supplies provide regulated voltages to the front-end2020
electronics and to the sensors (see section 4.5).
4.4.1 Front-end electronics
4.4.1.1 Pixel front-end electronics Each front-end readout ASIC [65, 77, 78] of the pixel de-tector contains 2880 readout cells of 50µm×400µm size arranged in a 18×160 matrix. SixteenASICs are bump-bonded to each sensor (module). The ASICs arefabricated using commercial2025
radiation tolerant 0.25µm CMOS technology. The schematic and basic functionality of the pixelcircuit is shown in Fig. 34.
Each readout cell contains an analogue block where the sensor charge signal is amplified andcompared to a programmable discriminator threshold. The digital readout then transfers the hitpixel address, a hit time stamp and a digitised amplitude information (the time over threshold2030
(ToT)) to buffers at the chip periphery. These hit buffers monitor each stored hit by inspecting theassociated time stamp.
The charge sensitive amplifier uses a single-ended folded-cascode topology optimised for anominal capacitive load of 400 fF and designed for the negative signal expected from DC-coupledn+-on-n sensors. Attention has been paid to the pre-amplifier design because, following the irra-2035
diation expected at the LHC, the sensor leakage current (50 nA) is two orders of magnitude largerthan the signal, that is itself reduced due to carrier trapping inside the silicon. The pre-amplifier hasan approximate 5 fF DC feedback capacitance with a 15 ns risetime. The total analogue front-end(pre-amplifier, second stage amplifier and discriminator) has a bias current of only 24µA per pixelfor the default DAC settings. To ensure the separation of contiguous bunch crossings, a front-end2040
time walk of<25 ns is required. To fulfill the requirements of sensor leakage current, a compen-sation circuit is implemented that drains the leakage current and prevents any influence on the bias
– 69 –
Chip-levelreadoutcontroller
Hit buffer
Column-levelreadoutcontroller
Data output
Clock
L1 Power supplies
8-bit time stamp
Address & time stamps
1. readout cell 2. readout cell
Analogue part
Configuration bits
Digital part
Time
Delayed BC clock L1
?Hit address
Hit amplitudeControl
Hit buffer cell
D/A convertor
Bump-bondpad
Slow control
Chip address
Sync
stamp
Figure 34. Layout and schematic description of the front-end readout ASIC for the pixel detector.
current of the fast feedback circuit used to discharge the feedback capacitor. Each pixel has severalparameters that are tuned and stored in a 14-bit control register, for example the feedback trimmingand threshold levels. The digital circuitry in the readout cells generates the required hit information2045
to measure the charge and associate the hit to the bunch crossing.
The readout is made using the column based Readout Controller. The first task of the controlleris the generation of the readout sequence to transfer the hitinformation. The second task is thedigital processing of hit data in the front-end chip buffers. Once a L1 trigger signal arrives, hitswith a corresponding time-stamp are flagged for readout. Flagged hits are transmitted to a serialiser2050
and sent out of the chip. Hits older than the trigger latency are cleared from the front-end chipbuffers.
The module control chip [65, 79] is a digital chip running with the same 40 MHz clock. Ithas three main system tasks: the loading of parameter and configuration data in the front-end chipsand in the module control chip itself, the distribution of timing signals such as bunch-crossing,2055
L1 trigger and resets (TTC functions), and the front-end chip readout and event building. Thedesign of the module control chip reflects the required pixelperformance during LHC operation:the association of signals to a bunch-crossing, the expected bandwidths at the highest luminosity,the maximum L1 trigger rate of 100 kHz and the number of front-end chips, which are controlledin a module. Because of the high radiation environment, particularly in layer-0 modules, special2060
attention has been given to ensure a single-event upset (SEU) tolerant design.
Extensive tests have been made on ASIC chips, single front-end chip assemblies and full mod-ules, before and after irradiation. Some production modules were irradiated to the end-of-life doseexpected at LHC (500 kGy). The noise and threshold dispersion after full irradiation and after
– 70 –
Figure 35. Distribution of threshold (left) and noise (right) of an irradiated module (500 kGy). The noisedistribution is shown for "normal pixels", for 600µm-long pixels between front-end chip edges and forganged pixels where two pixels are connected to the same pixel pre-amplifier in the region between thetwo rows of eight front-end chips.Show rather plots of pixel performance in terms of noise, time-walk andintrinsic accuracy before and after irradiation.
threshold tuning are shown in Fig. 35. Both quantities only modestly increase during irradiation2065
and remain within the operating specifications; the noise increase is∼10-15% while the differencein threshold dispersion after retuning is negligible.
4.4.1.2 SCT front-end electronics The readout hybrid of each SCT module (see section 4.3)houses 12 identical 128-channel ASICs [73] to read a total of1536 sensor strips per module. The2070
ASIC is fabricated in radiation tolerant bi-CMOS DMILL technology. The successive blocks ofthe ASIC are shown in the circuit schematic of Fig. 36. A pre-amplifier and shaper and tunablediscriminator exists for each channel. A 132-length binarypipeline stores the hit information foreach channel associated to the beam crossing for a period of∼3.2 µsec. Following a L1 trigger,the chip compresses the data pertinent to that beam crossingand serialises it for output. An 8-deep2075
de-randomising buffer after the pipeline ensures that the dead-time is negligible for the expecteddata rates.
Two critical module performance specifications are the detection efficiency (>99%) and noiseoccupancy (<5×10−4), for signals from the 12 cm long silicon strips with capacitive load∼20 pF.These have led to the choice of an ASIC discriminator threshold of 1 fC and a requirement on2080
the maximum Equivalent Noise Charge (ENC) of 1500 e RMS for anun-irradiated module and1800 e after the expected integrated radiation dose. Extensive studies have been made using ASICchips, single front-end chip assemblies and full modules, before and after irradiation. A sample ofproduction modules was irradiated in a 24 GeV proton beam to adose of∼3×1014 protons per cm2,
– 71 –
Figure 36. Schematic of the readout ASIC for the SCT detector, showing the successive signal processingsteps.
Figure 37. (a) The efficiency (circles) and noise occupancy (triangles) for SCT end-cap modules beforeirradiation. (b) The efficiency and noise occupancy measurements after exposure to a dose of∼3×1014
protons per cm2 in a 24 GeV proton test beam. The nominal operating thresholdis 1 fC.
equivalent to the damage of 1-2×1014/cm2 1-MeV-neutrons. The efficiency and noise occupancy2085
are shown as a function of the discriminator threshold in Fig. 37. At the nominal operating thresholdof 1 fC, the efficiency and noise-occupancy specifications are easily met before irradiation andalmost met after irradiation.
The chips are daisy-chained so that all the data of one moduleare read out over two serial
– 72 –
Figure 38. Front End readout of the TRT detector.
links. Several design features provide fault tolerance. For example, a faulty chip can be bypassed2090
in the serial data path and if one link should fail, it is possible to send the data using the remainingactive link. Likewise, if the primary clock and command lines to the ASICs on a module fail, it ispossible to instead use the clock and command signals from anadjacent module.
4.4.1.3 TRT front-end electronics The analogue signal processing and threshold discrimination2095
to detect signals from both minimum ionising particles and transition radiation, as well as thesubsequent time digitisation and data-pipelining are implemented in two on-detector ASICs. Thesignal is shown in Fig. 38 at each stage of the TRT signal readout chain.
a) The 8-channel ASDBLR ASIC [80], fabricated in bi-CMOS radiation tolerant DMILL tech-nology, performs the amplification, shaping and base line restoration. It includes two dis-2100
criminators, one at low threshold for minimum ionising signal detection and one at highthreshold for transition radiation detection;
b) A subsequent 16-channel ASIC fabricated in commercial radiation tolerant 0.25µm CMOStechnology [81] performs the drift time measurement (∼3 ns binning). It includes a digitalpipeline for holding the data during the L1 trigger latency,a derandomising buffer and a2105
40 Mbits/s serial interface. It also includes the necessaryinterface to the timing, trigger andcontrol as well as DACs to set the ASDBLR discriminator thresholds and test pulse circuitryfor mimicking analogue inputs to the ASDBLRs.
– 73 –
P P 1
T y p e I I c a b l e s1 3 6 L V c a b l e s3 0 4 D C S c a b l e s
1 0 - 1 3 mC r y o s t a t i n n e r b o r e
U S 1 5U S A 1 5
T y p e I V c a b l e s4 0 2 D C S & S e r v i c e s
c a b l e s5 0 - 5 5 m
T y p e I I I c a b l e s1 3 6 L V c a b l e s
4 0 - 7 5 m3 0 4 D C S & I n t e r l o c k
c a b l e s2 5 - 4 0 m
P P 0 T y p e I c a b l e s2 8 0 H V a n d L V c a b l e s
3 0 4 D C S c a b l e s2 9 4 8 - w a y r i b b o n f i b r e s
3 m
P P 3
P a t c h P a n e l I Ii n s i d e t h e m u o n s p e c t r o m e t e r
S l o w c o n t r o l e l e c t r o n i c sV o l t a g e r e g u l a t o r s
T y p e I I I c a b l e s1 4 4 H V c a b l e s
4 2 6 4 - f i b r e o p t i c a l c a b l e s6 0 - 9 0 m
P P 2
Figure 39. The routing of data links and power supply cables from each side of the pixel detector to theoff-detector electronics (respectively power supplies) in the service cavern, together with the number, typeand utilisation of the cables and optical links.
These ASICs are housed on front-end boards attached to the detector. There are 12 differentboards for the barrel and three different boards for the end-cap. The electronics are cooled by a2110
liquid mono-phase fluorinert (C6F14) cooling system.
At the TRT operating low threshold used for tracking (equivalent to∼15-20% of the averagesignal expected from minimum-ionising particles), the mean straw noise occupancy is∼2%, but asmall fraction of 1% have a noise occupancy exceeding 10%, which however remains small com-pared with the expected maximum straw occupancy of up to 60%.The full front-end electronics2115
chain was exposed to a neutron dose of∼4×1014/cm2 and to aγ-ray dose of 80 kGy. Changes ofup to 25% were observed in the ASDBLR gain, but with no change in the effective thresholds andnoise performance after a standard voltage compensation procedure.
4.4.2 Data transmission and power-supply services and routing
The transmission of data from the ID modules to the off-detector electronics in the service cavern,2120
as well as the digital transmission of the clock and control commands to the modules differs forthe 3 sub-detectors. Figures 39, 40 and 41 summarise the layout and technology of the readout andcontrol services for each of the pixel, SCT and TRT sub-detectors. The locations of key patch-panelconnection boards are shown: PP0 close to the ends of the pixel detector, PP1 at the edges of the IDvolume, PP2 in specifically designed parts of the muon spectrometer system and PP3 outside the2125
ATLAS active detector volume. The numbers and lengths of lines for each module (pixel, SCT) orfront-end board (TRT) are tabulated for both the barrel and end-caps. Similarly, the cables used forthe readout electronic bias as well as the silicon sensor andTRT straw high-voltage lines are alsoshown.
4.4.3 Data transmission to the readout drivers2130
4.4.3.1 Pixel and SCT readout The digital transmission of the clock and control commandsto the pixel and SCT module and data from the module to off-detector electronics in the service
– 74 –
Figure 40. The routing of data links and power supply cables from each side of the SCT to the off-detectorelectronics (respectively power supplies) in the service cavern, together with the number, type and utilisationof the cables and optical links.
P P B 1
P P F 1
T y p e I I c a b l e s1 2 9 2 m i n i - c o a x i a l H V l i n e s 3 5 2 0 l o w v o l t a g e c a b l e s
1 5 9 3 6 m i n i a t u r e t w i s t e d p a i r s f o r d a t a r e a d - o u t , T T C a n d D C S
1 2 - 1 5 m
C r y o s t a t i n n e r b o r e
P P 2
M A R A T O ND C - D C
B a l c o n i e s i n U X 1 5
U S A 1 5
P a t c h P a n e l I Ii n s i d e t h e m u o n s p e c t r o m e t e r
E l e c t r i c a l t o o p t i c a l c o n v e r t o r s L V D S r e p e a t e r s
S l o w c o n t r o l e l e c t r o n i c sV o l t a g e r e g u l a t o r s
T y p e I I I c a b l e s3 4 H V m u l t i c o n d u c t o r s c a b l e s
3 2 C A N B u s c a b l e s2 8 8 M u t i - p a i r c a b l e s
9 6 4 - f i b r e o p t i c a l c a b l e s6 0 - 1 0 0 m
T y p e I V c a b l e s1 2 M u l t i - p a i r c a b l e s1 2 3 8 0 V c a b l e s
3 5 - 6 0 m
T y p e I I I c a b l e s5 4 4 L V c a b l e s2 5 - 4 0 m
Figure 41. The routing of data links and power supply cables from each side of the TRT to the off-detectorelectronics (respectively power supplies) in the service caverns, together with the number, type and utilisa-tion of the cables and optical links.
cavern is made via optical links [73, 82, 83, 84]. A detailed block diagram illustrating the pixellink structure is shown in Fig. 42.
As summarised in Table 14 each pixel module uses one or two up-link fibres according to the2135
required bandwidth; the bit rate is either 40 or 80 Mb/s per fibre link. One down-link fibre permodule is used to transmit trigger, timing, clock signals and configuration data. Opto-boards onthe detector side convert the electrical signals from the module control chip to optical signals. Thedown-link uses a bi-phase mark encoded format to transmit both the 40 MHz bunch-crossing clockand the data. The bi-phase mark light signal, detected by a PIN diode, is decoded by a digital2140
opto-receiver integrated circuit. In the up-link the module control chip output is converted to lightby a VCSEL driver chip coupled to a VCSEL. This latter chip andthe corresponding SCT chipshave four channels each. They have been made in the same 0.25µm CMOS technology as the
– 75 –
RX plugin
TX plugin
DORIC
Opto-Board
Opto-Board
Up
Down
LVDSPP0 PP1
rackROD
16 or 8
8
84
4
(8)
8
8
8
272
272
272
272
Back of Crate Card
VDC
6 or 71 or 2
2 or 4
272
2
PiN
array
VCSEL
array
Fibres Fibres
PIN
array
VCSEL
array
DRX
BPM-12
Module
MCC
Figure 42. Detailed schematic of the pixel optical link architecture.For each pixel module, one (layer 2) or2 (layers 0, 1 and the disks) optical fibres transfer data to the RODS, and 1 fibres transfer control and clocksignals to the module. The SCT optical links have a similar design.
Subdetector Link Speed Format Links per Total number
Mbit/s module of links
Pixels Data 40/80 NRZ 1 (layer 2)
40/80 NRZ 2 (layers 0,1 and disks)2812
TTC 40 BPM 1 1750
SCT Data 40 NRZ 2 8176
TTC 40 BPM 1 4088
Table 14. Summary of the optical links used for the pixel detectors (NRZ: Non Return to Zero, BPM:Biphase Mark).
module control chip and have been produced on the same silicon wafers. Opto-boards can serviceeither six (half-staves or disk-sectors) or seven (half-staves) modules depending on which part of2145
the detector they are connected to. The opto-boards have been placed on service quarter-panelsnear the internal surface of the pixel support tube (see section 4.6), to reduce the radiation doseseen inflicted over 10 years of LHC operation.
In the case of SCT modules, the service harness that providespower to 5-6 modules alsoincludes two data fibres and one trigger/control fibre for each module, as well as VCSEL and pin-2150
diode opto packages for electrical to optical conversion together with their control ASICs. EachSCT barrel module is serviced by a polyimide-aluminium low-mass tape that provides redundancyand electrically isolates the modules. For the end-caps, polyimide-copper tapes are used for thecontrol lines and individual copper-clad aluminium wires for the power lines. The module is con-nected via the low-mass tape to the PP1 patch panel outside the SCT and from there by conventional2155
cables to the power supplies. The services include 2 data fibres and 1 trigger/control fibre for eachmodule plus VCSEL and PIN diode opto packages for electricalto optical conversion together withtheir control ASICs.
– 76 –
On the other side of the link, back-of-crate cards interfacethe opto-signals with the electricalsignals in the readout driver (ROD). These cards are the samefor the pixel and SCT detectors and2160
they contain two kinds of electro/optical converter plug-ins: a RX plug-in with an array of PINdiodes and a data receiver, and a TX plug-in with an array of VCSELs driven by a bi-phase markencoding ASIC. The SCT uses a radiation-hard step index multi-mode fibre for the full length fromthe back-of-crate cards to the detector. The pixel detectoruses a long length of radiation-tolerantgraded-index fibre, spliced to 7m lengths of the radiation-hard fibre inside the detector.2165
4.4.3.2 TRT readout As described in section 4.3.3, the TRT readout is segmented in 32 φ-sectors in order to simplify the data transmission for L2 triggers. The initial 40 Mbits/s LVDS datareadout uses small custom designed twisted pair lines from the ASIC to boards at patch panel PP2,located after the first muon chambers. The data are then serialised in a Gbit serialiser [85] and an2170
electrical-to-optical conversion is made. The complete readout of the TRT requires 768 1.6 Gbit/soptical links. Timing and control signals to and from the TRT-TTC module are electrically trans-mitted.
4.4.3.3 Pixel, SCT and TRT readout drivers (ROD’s) Each pixel and SCT ROD crate is a 9U2175
VME crate with up to 16 ROD cards per crate. Each ROD card is paired with a back-of-crate card,which is plugged into the crate back-plane and provides the optical to electrical interfaces [83, 86].The back-of-crate card can accept 40 Mbps optical links fromSCT modules or 80 Mbps from pixelmodules. Each ROD card services up to 48 SCT modules (96 data links) or between 6 and 32 pixelmodules, depending on the layer. The lower limit of pixel modules is due to the larger quantity2180
of data from each pixel module, which demands more servicingon the ROD. The pixel and SCTROD’s are identical units except for the data treatment algorithms. The ROD crates receive clock,trigger and fast commands from the ATLAS TTC system through aTTC interface module locatedin each crate which transmits these signals across the crateback-plane to the ROD’s and back-of-crate cards. Clock and command signals are then transmitted to the detector modules using a2185
simpler protocol over the optical fibres mentioned above.
Following a L1 trigger, each ROD card receives data from all of its modules. It performserror checking and combines the data for all its modules intoa single event packet associatedto the L1 trigger. This event packet is sent over a high speed optical link (the ATLAS definedS-Link), again via an optical interface on the back-of-crate card and in a standard data format,2190
to the readout system, which is common to all ATLAS sub-detectors. The ROD also performsdata monitoring functions including the accumulation of monitoring histograms, for example thechannel hit frequency. The ROD can also send special calibration and diagnostic commands to thedetector modules [83, 86, 87].
The TRT off-detector electronics also uses two custom-designed TRT-ROD and TRT-TTC2195
modules [88]. Each ROD module receives serialised data from1/16 of one barrel side or 1/32of one end-cap side, using 8 optical links operating at 1.6 Gbits/s. After the optical-to-electricalconversion, the ROD’s perform de-serialisation, error checking, data compression and local eventbuilding tasks. The data compression scheme does not introduce data losses. As for the pixel
– 77 –
Pixel SCT TRT
Voltage (max) 700 V 500 V 2000 V
Bias voltageVoltage (nominal)150-500 V 150-350 V 1600 V
supply Current (max) 4 mA 5 mA 3 mA
Segmentation 1 module 1 module ∼200 straws
Electronics Voltages Vdd, VddA Vdd, Vcc ±3 V analogue
supply VddC Vpin, Vvcsel +2.5 V digital
Current 3.7 kA 6 kA 6.5 kA
Segmentation 6-7 modules1 module 1/32 end-cap
1/32 barrel
Table 15. Summary of the ID sensor bias voltage (HV) and front-end electronics (LV) requirements andgranularity. The different electronic LV voltages are described in the text. This table may become toocomplex to be kept as is. It needs to be corrected for the fact that through the LV regulators each pixelmodule is "independently" biased and that the HV is suppliedindividually to each module. It should alsosummarise power consumption on detector and in the cables ifpossible. The LV drops should probably besummarised here rather than just for the SCT in the text.
and SCT, the event packets are then transferred over the ATLAS S-link to the readout system in a2200
standard data format.
The TRT-TTC module provides an interface between the standard ATLAS TTC system andthe TRT front-end electronics. It also feeds the ROD with allnecessary L1 trigger information(event ID, bunch crossing ID, trigger type). Each TRT-TTC module interfaces to the front-endusing 40 links and to two ROD’s via a dedicated back-plane using the VME 9U P3 connector.2205
4.5 Electronics and detector power supplies and services
Both the sensors and the front-end electronics require power. Table 15, together with Figs. 39 to 41summarise the high voltage (HV) sensor bias and low voltage (LV) electronics power requirementsfor each of the pixel, SCT and TRT sub-detectors, as well as the segmentation used for the powerdistribution. The power distribution lines are connected at PP1 on the cryostat wall, PP2 external2210
to the tracker after the first muon chambers, and PP3 at the outside of the detector.
The pixel power-supply system has four main components: theLV and HV power supplies,the regulator station, and the supply and control for the opto-links. Two commercial LV suppliesprovide the analogue and digital parts of the front-end readout electronics. To protect the front-end electronics against transients, remotely-programmable regulator stations are installed at PP2,2215
outside the calorimeter. Separate HV supplies are able to deplete the sensors up to 700 V (forleakage current values below XXµA). The LV and HV lines are connected to respectively thelow voltage and high voltage patch-panels that distribute the power and monitor the currents of
– 78 –
individual lines. The SC-OL, a complex link consisting of three voltage sources and a controlsignal, delivers the adequate levels for the operation of the on-detector part of the optical link.2220
The SCT [89] has chosen to maintain independent electrical services to each of the 4088modules. Each module receives multi-voltage LV channels providing power and control signals tothe readout chips, the optical electronics, timing and control electronics, as well as HV (0-500V)to the silicon sensors. The SCT LV power module provides the analogue (Vcc) and digital (Vdd)voltages for the SCT front-end ASICS and optical link components. Each LV power module also2225
includes the hybrid temperature readout as well as the digital RESET and SELECT control lines.The HV bias provides a stable and controlled voltage of 0-500V (for leakage currents belowYY µA). The LV and HV modules are based on DC-DC converters and since a single powerline serves each module, twelve identical 4-channel LV and 6identical 8-channel HV boards arehoused in each crate. Each card in the crate is connected through a System Interlock Card (SIC) to2230
the Detector Control System (DCS). The four crates of each rack are powered by a single 48V DCsupply. The power distribution lines are in 4 parts: low masspolyimide tapes for the most innerregion of the detector connected to PP1 on the cryostat wall,thin cables along the cryostat wall thatare spliced to medium-size cables in outer parts of the detector and, after PP3, thick cables goingto PS crates. Because of a high voltage drop in the cables (4-4.5V), the supplied voltage exceeds2235
the maximum allowed voltage of the ASICSs that could be problematic, for example in case ofa drop in the power consumption of the front end system. To avoid such an accident a voltagelimiter circuit is introduced at PP3. To dampen common-modenoise, inductors on all power andlow-current control and monitoring lines are also housed atPP3.
The TRT front-end electronics requires three low voltage power supplies (+2.5 V digital,±32240
V analogue). Commercial supplies deliver power to boards housed at PP2. The PP2 boards houseradiation tolerant voltage regulators delivering power toeach front-end board. The size of the ca-bles feeding the PP2 power boards and the front end boards is acompromise between the availablespace and the power dissipation in the cable trays allowed bythe cooling system. The TRT strawHV is nominally 1.6 kV. Sets of about 200 straws are powered bya single commercial HV source2245
able to deliver up to 3 mA at 2 kV. About 2000 HV channels are needed for the whole detector.Standard multiwire HV cables are used up to the level of PP2, then custom miniature HV cablesare used to reach the detector. The cable layout and inventory is shown in Fig. 41.
Additional power supplies exist for items such as the thermal enclosure heaters and the heatersof the evaporative cooling system.These will have to be briefly listed and described here.2250
Add a paragraph on grounding and shielding for each of the pixel, SCT and TRT here or im-mediately below as a sub-section, depending on what is in theoverview. For the SCT there is onecommon SCT ground inside ATLAS and the ground and shield of all PS cables are referenced tothis. There is no contact with local grounds such as at the PS racks.2255
4.6 Structure and mechanical integration of the inner detector
Prior to integration as part of the full ATLAS detector, the barrel and end-caps for each of the pixel,SCT and TRT sub-detectors were separately assembled and fully tested on the surface. This sectiondescribes the overall mechanical structure of each of the sub-detectors. The subsequent integration2260
– 79 –
Figure 43. The TRT barrel support structure, which also serves as the support for the complete barrel ID.The SCT detector is supported inside the carbon-fibre TRT inner cylinder. The outer carbon fibre cylinder isnot shown. The space frame geometry is designed to support individual TRT barrel modules.
of the ID in ATLAS comprises 4 steps that are also described inthis section: the integration of theSCT and TRT barrel, the integration of the SCT and TRT end-caps, the integration of the barreland end-cap pixel detectors with the beam pipe, and finally the insertion of the pixel package.
The mechanical support for the SCT+TRT barrel is the barrel support structure, shown inFig. 43. It is designed for high stiffness and long-term stability, with <10 µm displacements under2265
the expected temperature and humidity variations. It consists of two 20 mm thick carbon fibrespace frames, joined by inner and outer carbon fibre cylinders. The SCT+TRT end-caps are eachsupported from a pair of girders, sliding on rails inside thebarrel calorimeter cryostat. The separate6.6 m-long pixel and beam-pipe package includes the pixel support tube, which slides inside theSCT and is attached to the SCT end-caps.2270
Tables 16 for pixels and 17 for SCT should probably go back to the module section. TheLorentz angle effects, the dependence of Lorentz angle on radiation, and the tilt angles chosenmust all be discussed in one place, for pixel and SCT.
4.6.1 Pixel structure and integration
The pixel detector and the pixel support tube (PST) within the ID are shown in Fig. 26. The detector2275
with its associated services and the beryllium beam pipe within the ID are precisely located insidethe PST. The pixel services (cooling, power and monitoring)are routed to the ends of the PST.
The active region of the pixel detector is shown in Fig. 44. The pixel detector consists of threebarrel layers and two end-caps each with three disk layers. The basic detector parameters are listedin Table 16. The total active area of silicon is approximately 1.7 m2. The 112 barrel staves and 482280
end-cap sectors (8 sectors per disc) form the barrel and disclayers.
– 80 –
Number of Number of Number of
Barrel Radius(mm) staves modules pixels
Layer 0 50.5 22 286 13.2×106
Layer 1 88.5 38 494 22.8×106
Layer 2 122.5 52 676 31.2×106
End-cap (one side) z(mm) Sectors Modules Pixels
Disk 1 495 8 48 2.21×106
Disk 2 580 8 48 2.21×106
Disk 3 650 8 48 2.21×106
Barrel+both End-caps 1744 80.4×106
Table 16.Parameters of the pixel detector. The pixel envelope is shown in Table 2. The active barrel lengthis 801 mm and the inner (outer) active radii of each disk are respectively∼89 (∼150) mm. The quoted barrelradii are average values since the barrel staves are tilted at 20 with respect to a tangent vector at the givenradius. The diskz-positions are also nominal.Tables 16 for pixels and 17 for SCT should probably go backto the module section. The Lorentz angle effects, the dependence of Lorentz angle on radiation, and the tiltangles chosen must all be discussed in one place, for pixel and SCT.
Figure 44. A perspective cut-away view of the pixel detector. The view shows individual barrel and end-capmodules, supported with their associated services on staves and disks within an octagonal support frame.
In the barrel region, the bi-staves are mounted in half-shells (Fig. 45a). Each half-shell is a thincarbon-fibre shell formed with facets to match the number of staves, with cut-outs to reduce themass and with mounting rings that position the staves at five locations. The location of the stavesin the half-shells was measured using a coordinate measuring machine to reference the position of2285
– 81 –
modules to known features on the shells. The disks of the pixel end-cap detector are bolted withprecision bushings to a carbon-composite support ring. Thedisks are then held with four mountswithin a section of the octagonal support frame to form an end-cap. An end-cap during the finalstages of assembly is shown in Fig. 45b, after connection of the cooling circuits. Two fully-loadedhalf-shells form a barrel layer. Layer 2 after this step is shown in Fig. 46. Each barrel layer is2290
inserted into the supporting octagonal frame and connectedto end-cone structures with fingers tomate precisely with mounting brackets on the barrel. Capillaries and outlet cooling tube extensionsare added.
Figure 45. (a) A pixel barrel half-shell, with its cutouts, being loaded with barrel bi-staves and services. (b)A pixel end-cap at the last stage of assembly, after connection of its cooling circuits.
Following the integration of the beam pipe, the barrel octagon and the 2 end-cap octagons,the barrel cooling pipes and cables are passed over the outside of the end-cap frame. All end-cap2295
services are on the inside of the frame.
The PST itself consists of three sections. Each section is a cylinder with external stiffeningrings and precision rails. The barrel section is made from carbon-fibre composites. Each end-section is composed of carbon fibre and fibreglass composites. The rails are carbon-fibre compos-ites and are accurately positioned within each cylinder. The cylinders are joined (at each end of the2300
barrel) by bolted carbon-fibre flanges. Heater panels (copper-on-polyimide printed circuit boards)are glued to the cylinders. These are activated if there is a failure in the dry environment aroundthe pixel detector.
The barrel PST section is precisely located with respect to the barrel SCT structure. Mountingpoints on the barrel PST receive mounts on the pixel detectorsupport frame described below and2305
locate the pixel detector to approximately 100µm with respect to the barrel SCT. An octagonalcarbon-composite frame supports and positions the barrel and the two end-caps. The two end-capsections are joined with composite bolts to the barrel frame. Mounts that position the pixel detectorwithin the PST are located on endplates on the end-caps.
Each pixel module receives LV power directly from the LV regulators located at PP2 (see sec-2310
tion 4.5). Furthermore, there are 88 cooling circuits for the pixel detector and monitoring functions
– 82 –
Figure 46. Barrel pixel layer 2, loaded with bi-staves, viewed along the axis after the joining of the half-shells.
require connections within the pixel detector volume. The electrical and cooling connections passthrough the end regions of the PST. Since the pixel detector is inserted from one end, all services,including connectors, must fit within the PST inner envelope, a radius of 230 mm. The electricalservices and cooling pipes are contained within service quarter-panels, which deliver one-quarter2315
of the required services at each end of the pixel detector. Power and monitoring wiring is routedby individual twisted pairs and soldered to printed circuitboards near the active detector (PP0)that contain miniature connectors. The low-mass cables from each module are plugged into theseconnectors. At the other end (PP1) the wires are soldered to printed circuit boards to penetratethe sealing plate at the end. On the outside of the PP1 region,other twisted pairs are soldered to2320
the printed circuit boards and terminated in commercial connectors. Optical transceivers are alsolocated in the PP0 region and convert electrical signals from the detector to light transmitted byfibres to connectors at PP1. Control signals from outside thedetector are also transmitted by fibresto the transceivers for conversion to electrical signals (see section 4.4).
Each cooling circuit includes a custom heat exchanger that consists of an inlet and an out-2325
let tube glued together along the length of the service quarter-panel. These tubes penetrate theplate at PP1 and dry-gas integrity is maintained by a bellowsseal that also allows for the 2-3 mmcontraction of these aluminium pipes when the detector is operating.
The service quarter-panels and the beam-pipe are supportedby a composite beam-pipe supportstructure. The beam-pipe supports are adjustable from the ends of the PST to position the beam-2330
pipe. The overall detector integration is illustrated in Fig. 47.
The pixel detector is sensitive to the high instantaneous rates that might occur during accidentalbeam losses. For this reason, a set of small, fast and radiation-hard diamond detectors, called theATLAS beam-conditions monitor (BCM) has been built and integrated into the pixel package tomonitor the beam conditions and to distinguish lost beam particles from proton-proton interactions.2335
The BCM is described in more detail in section 3.4.1 and shownas installed near the beam-pipe
– 83 –
Figure 47. The pixel detector during integration of the barrel, end-caps and their services: (a) the end-capregion; (b) the barrel detector region; (c) Patch Panel 1 (PP1) region; (d) Patch Panel 0 (PP0) region and (e)region of the optical transceivers on the service quarter-panels. See text for details.
in Fig.21.
4.6.2 SCT structure and integration
As indicated in Fig. 26, the SCT consists of 4 coaxial cylindrical layers in the barrel region and twoend-caps each having 9 disk layers [63]. The layers are tiledwith 63 m2 of silicon modules [72, 76]2340
that provide almost hermetic coverage with at least four precision space-point measurements overthe fiducial coverage of the inner detector (see Tables 17 and18).
The low mass barrel cylinders are designed to be extremely stable to both temperature andhumidity variations, and to long term creep [90]. They are made from 3–layer (0,+60,-60)carbon fibre skins of 200µm thickness over a carbon fibre/ cyanate ester honeycomb coreto form2345
a 6 mm sandwich. The cylinder ends are closed with flanges, incorporating holes that are machinedto high precision. Pads for the precision mounting of modulebrackets are attached to the surfaceand machined. The insert position accuracy is specified to be±20 µm. A similar precision isspecified for the inserts of an alignment system mounted on each barrel, and for the machinedholes on the end flanges. Because of poor adhesion for a few pads, all the pads were subsequently2350
attached via plastic screws, with a slight precision loss. The external radius of all cylinders hasbeen maintained to within 1 mm.
The barrel modules [72] are mounted in rows of 12, on individual carbon fibre brackets(Fig. 48). The module is rotated by±20 mrad, alternating from barrel to barrel, to align thestrips of one side along the cylinder axis and attached to thesupport structure at 3 points, two on2355
the beryllia facing (cooling side) and one on the far side [91]. When mounted on the barrel, thevariance of the mounting precision in z as measured by the mounting robot is 60µm, but there hasbeen no systematic survey of the variance in R–φ. To avoid HV breakdown, a distance of>1 mm
– 84 –
Barrel Radius Full length Module tilt Number of
cylinder (mm) (mm) angle (deg)modules
3 284 (299)1530(1498)11.00 384
4 355 (371) 11.00 480
5 427 (443) 11.25 576
6 498 (514) 11.25 672
Total 2112
Table 17. SCT barrel cylinder parameters and the number of modules. There are 12 modules per row. Thequoted radius and length are those of the outer barrel surface and the mean active sensor radius or length(brackets). The tilt angle is with respect to the tangent to the support cylinder.Tables 16 for pixels and 17for SCT should probably go back to the module section. The Lorentz angle effects, the dependence of Lorentzangle on radiation, and the tilt angles chosen must all be discussed in one place, for pixel and SCT.
Disk 1 2 3 4 5 6 7 8 9
|z| mm 835.8934.01091.51299.91399.71711.42115.22505.02720.2
Outer 52
Middle 40 none
Inner none 40 none
Table 18. Thez-position for each SCT end-cap disk and the number of moduleson each disk (total 1976).For acceptance reasons, disk 9 has only outer modules, whiledisks 1 and 7 have no inner modules. Themiddle modules of disk 8 have only one sensor, again for acceptance reasons.
is maintained between the sensor edges and any ground potential. The centres of adjacent modulesin each row are radially separated by 2.8 mm using the tiling arrangement.2360
As for the barrels, the end-cap disks support end-cap modules (Fig. 49) with tight stabilityand accuracy requirements, as well as electrical, mechanical and alignment services. The 8.7 mmthick disks consist of carbon-fibre face skins (200µm thick) with an aramid/phenolic honeycombcore. The arrangement of the plies and the choice of materials minimises the effects of thermal andhumidity changes. Individual modules are attached to cooling blocks held by inserts glued to the2365
disk. A large cooling block at the hybrid end (230 mm2) defines the position of the module whilea slot at the end of the module defines the± 20 mrad rotation of the module when attached to thesmaller block (78 mm2), to within ±1 mrad. The rms spread of the surveyed module placementpositions on each disk was 10µm and an estimated 50−100 µm in the placement of disk withinthe support cylinder. .2370
The disks of an end-cap are supported at 12 points around their diameter by springs in the
– 85 –
Figure 48. a) A detail of the barrel modules mounted on the support cylinder, together with the moduleservices (signal and power polyimide cables, and cooling tubes). b) Drawing of the mounting brackets thatare attached to the barrel, and the attachment of the module and cooling pipes to the bracket.
Figure 49.Modules mounted on end-cap SCT disk with (a) outer and inner modules and (b) middle modules
support cylinder. The springs are soft in the radial direction but otherwise stiff, allowing for theradial expansion of the disks and cylinder. The cylinder composition is similar to that of the disks.Services leave the disks through apertures in the cylindersand run along the cylinders before exitingat the far end of the end-cap thermal enclosures. Each support cylinder is in turn supported by two2375
flat composite panels (of similar construction to the cylinder). These panels rest on the same railsthat support the TRT.
Prior to the mounting of SCT modules on the support structures, each barrel or disk wasequipped with electrical services and cooling loops. In thebarrel, each cooling loop services 4rows of 12 modules. The loops are mounted on the module mounting fixtures and connected to2380
the modules using thermal grease. For the end-caps, each cooling loop is attached to the disk andtraverses the module mounting blocks.
– 86 –
Each service assembly and each cooling loop were tested and all modules of a given barrel ordisk were powered and read to verify the full module functionality [92]. Less than 0.5% of modulesneeded intervention. For a final test of completed barrels and disks, test systems were constructed2385
to operate and read up to one million channels simultaneously [93]. After assembly, individualbarrels were transported to CERN for final integration.
The integration of individual barrels and disks differ due to their different support structures.For the end-cap, an assembled disk is inserted into the overall carbon-fibre support cylinder andpolyimide power tapes and cooling pipes are connected at theouter disk circumference, and run2390
along the outside of the cylinder to a patch panel at the end. Each completed end-cap assemblywas fully tested and characterised. After assembly, each ofthe end-cap cylinders was transportedto CERN for final integration and test.
Both the barrel and end-cap are surrounded by low-mass outerand inner thermal enclosures.Their role is to: prevent condensation during operation by maintaining a low temperature and low2395
humidity N2 environment; prevent the out-flow of N2 gas surrounding the SCT modules that woulddamage the TRT performance and the in-flow of the ID environmental gas (CO2); and provide aFaraday shield to protect the SCT from external electrical noise. The outer thermal enclosures andthe end-surfaces are covered with resistive-pad heaters toensure thermal neutrality.
To complete the SCT barrel from four individual barrels, theouter thermal enclosure was2400
mounted into a support cradle and the 4 SCT barrels sequentially inserted. All services were sealedinto slots of the outer thermal enclosure feed-through on the outer circumference of the SCT barrel.Inlet cooling capillaries, outlet exhaust cooling pipes and ground-reference connections were addedbetween individual barrels and thermal-enclosure bulkheads. Finally the outer and inner thermalenclosures were sealed and the air-tightness of the barrel enclosure checked.2405
4.6.3 TRT structure and integration
As indicated in Fig. 26, the TRT occupies the outer radial regions of the ID. There are 3 modulelayers with axially oriented drift tube straws in the barrelregion [94] and 20 wheels with radialstraws in each of the end-cap regions [95]. The active regions of each detector are shown in Table 13(see section 4.3.3). The TRT contains up to 73 layers of straws interleaved with fibres (barrel) and2410
160 straw planes interleaved with foils (end-cap), which provide transition radiation for electronidentification. All charged tracks withpT > 0.5 GeV and|η | < 2.0 will traverse at least 35 straws,except in the barrel-end-cap transition region (0.8< |η | < 1.0), where this number decreases to aminimum of 22 crosses straws. Typically, 7 to 10 high-threshold hits from transition radiation areexpected for electrons with energies above 2 GeV.2415
Each of the 96 barrel modules is supported at each end by the barrel support structure, whichalso provides the required overall module stability. It is a20 mm thick carbon fibre disk, machinedto a triangular strut array and attached to a support cylinder (see Fig. 43 and Fig. 50a).
As described in section 4.3.3, each of the TRT end-caps consists of twelve type-A wheels witha straw layer spacing of 8 mm and of eight type-B wheels with a spacing of 15 mm. The geometry2420
of the lower density type B wheels maintains the required number of straws crossed by a particlefrom the interaction point as well keeping the material of the active detector approximately constantas a function ofη .
– 87 –
Figure 50. (a) Face view of the TRT barrel structure, during the final attachment of cooling and electricalservices. (b) A completed TRT end-cap during the final service integration, showing from the left 12 type Awheels, 8 type B wheels and the services.
Each 8-layer end-cap wheel consists out of the basic block of4-layer sectors mounted in aninner and outer electrically grounded carbon fibre ring. Therings and the straws serve as the2425
mechanical support structure of the wheels. Each eight-plane wheel is covered with a thin metal-clad polyimide membrane on each front side connected at the inner radius and providing a signal-return path with defined electronics ground. Openings at theouter radius of the electronics shieldallow a path for the CO2 cooling gas. The weight of a complete end-cap including all services,pipes and cables has been measured as 1117 kg± 10 kg.2430
The overall discussion of the weight of the ID deserves a separate paragraph in the materialsection. By now, all components are known to a high accuracy.In fact, the end-cap numbers shouldbe known to 1-2 kg.
As shown in Fig. 50b, the type A and B wheels are assembled in two independent groups andtheir wheels are held together using tie-rods between two solid carbon-fibre membranes of 10mm2435
thickness. The end-membranes of a stack are supported on rails fixed to the ATLAS barrel cryostat.Each group is sealed at the inner radius by a glass-fibre cylinder covered with a thin copper-cladpolyimide foil. This cylinder serves as both, electrical Faraday shield and CO2 gas manifold.
The end-cap CO2 gas circulating between and within sectors is cooled by heatexchangerspositioned at the outer ends of the straws between each second eight-plane of type A and each2440
eight-plane of type B before flowing to the next wheel of the end-cap. That circulating in barrelmodules is cooled by conduction at the acute edge of the barrel modules. The leak tightness ofthe barrel and end-caps for the CO2 has been tested. To prevent pressure changes of the coolinggas that might cause mechanical stress and deformation (discharge) of the straws, a set of passivesafety valves are installed.2445
Supplies for high-voltage, active gas, electronics cooling and the collection of the detectorsignals to the back-end system located outside the ATLAS experimental hall are organised sector-wise. Along a sector all pipes and cables are brought in cabletrays from outside the ATLAS
– 88 –
Figure 51. Insertion of SCT barrel into the TRT barrel. The 3 module types of the TRT barrel are clearlyidentified. The SCT Outer Thermal Enclosure is visible, together with the barrel services extending onsupport frames from each end.
cryostat to the detector and distributed to the individual wheels.
4.6.4 Integration of the SCT and TRT2450
The completed SCT barrel and end-caps were finally inserted into the corresponding TRT sub-detectors. The barrel procedure is described here: the end-cap integration followed a similar proce-dure. The SCT is supported on a cantilever frame (foregroundof Fig. 51). The completed TRT wastransferred into the final support and lifting frame (background of Fig. 51). The TRT was guidedon rails over the SCT. During the movement the mechanical alignment and electrical isolation of2455
the sub-detectors were verified. After insertion, the SCT was positioned on rails inside the TRTwith a precision of∼250µm. The mechanical survey of the SCT and TRT barrel shows a horizon-tal displacement of∆x=-0.3 mm and a rotation around the y-axis of 0.221 mrad of theSCT withrespect to the TRT barrel. These values are in good agreementwith the global alignment found ina later cosmic run (see section 4.9).2460
The final mechanical survey has to be confirmed with Andrea.
4.7 Inner detector environment and general services
This should address the overall mechanical ID integration (ID end-plate and interface to beam-pipe with the bake-out scenario), the evaporative cooling,the environmental gas and the general2465
DCS, safety and interlocks.
– 89 –
4.7.1 Inner detector DCS, safety and interlocks
The ID-specific DCS operates within the general context of the overall ATLAS DCS, as describedin section 7.5. It controls the detector powering and monitors and controls the environmentalparameters, in particular the pixel and SCT cooling, and theTRT gas gain. The sub-system-specific2470
environmental monitoring is similar for all three sub-systems and is described briefly below.
In the context of general inner detector DCS, the evaporative and mono-phase cooling systems,which must remove∼ 75 kW of heat from the ID and maintain a temperature stabilityof ±2C, arecontrolled and monitored. Critical temperature, pressureand flow parameters are monitored andcontrolled, including the pressure regulators and the heater temperatures and current. Additional2475
sensors monitor and control the thermal enclosure heaters,the electronics rack temperatures andvoltages, and the cable and patch-panel temperatures.
For the SCT [96], the environmental monitoring measures approximately 735 temperature andhumidity sensors across the detector. Three types of temperatures are measured: sensors are locatedat the exhaust of each of the 116 cooling pipes; sensors attached to the mechanical structure of the2480
detector to monitor the possible deformation due to temperature changes; and sensors located nextto the laser interferometric survey monitoring system to measure the gas temperature inside thedetector volume. Radiation-hard humidity probes are installed in various locations and are used tocalculate the dew point and hence to avoid condensation on the modules.
The monitored cooling-pipe temperatures also trigger the interlock, protecting the pixel and2485
SCT modules if the cooling stops. It is implemented in hardware without the use of DCS micro-processors nor software. The custom-built hardware compares the temperature sensor values to apreset threshold and signals the appropriate power supply cards to turn off in< 1 second if thethreshold is reached.
Within the TRT system, about 200 parameters of the gas distribution system are monitored.2490
In addition, the gas composition is monitored using a55Fe source, and the DCS system is used toadjust the straw HV to maintain constant gas gain. The activegas and mono-phase liquid coolingtemperatures are also monitored.
4.8 Material distribution of the inner detector
The performance requirements of the ATLAS inner detector are more stringent than any tracking2495
detector built so far for operation at a hadron collider. Theharsh environment and the pile-up frommultiple interactions per bunch crossing make a high detector granularity mandatory, with elec-tronics, readout services and cooling within a detector volume, which must have good mechanicalstability. The overall weight (∼4.5 tonnes) and material budget of the ID (in terms of radiation andinteraction length) are therefore much larger than those ofprevious tracking detectors. The conse-2500
quences of this are quite serious and are currently the focusof many studies (see section 9.2.1):
a) many electrons lose most of their energy through bremmstrahlung before reaching the elec-tromagnetic calorimeter;
b) approximately 40% of photons convert into an electron-positron pair before reaching the LArcryotat and the electromagnetic calorimeter;2505
– 90 –
c) even in the case of low-energy charged pions, a significantfraction will undergo an inelastichadronic interaction inside the inner detector volume.
A detailed modelling of the ID has been made in the Geant4 framework. Figure 52 shows amap of photon conversions in the ID volume, integrated over azimuth, for photons coming fromthe primary vertex. Need to provide a few numbers in addition here or to remove figure: for2510
example, what is integrated fraction of photons which converted and for which energy spectrum?How much running time to collect an equivalent amount of data(probably less than an hour at200 Hz data-taking rate).
Figures 53 and 54 show the integrated X0 (a) andλ0 (b) traversed by a straight track as afunction of | η | at the exit of the ID envelope. The most striking feature is the onset of non-active2515
service and structural material at the interface of the barrel and end-cap regions. This includescooling connections at the end of the SCT and TRT barrels, TRTelectrical connections, SCT andTRT barrel services extending radially to the cryostat to the PP1 connection box and then alongthe cryostat wall. Another service contribution is from thepixel services (| η |≥ 2.7) that leave thedetector along the beam pipe; their extended range in| η | is evident. A large fraction of the service2520
and structural material is external to the active ID envelope, therefore deteriorating the calorimeterresolution but not the tracking performance. It should be noted that end-plates of both the barreland end-cap supports are ribbed, and therefore azimuthallydiscreet.
Table 19 lists the contribution to X0 as a function of radius from different elements of the IDfor straight tracks at| η |=0 and| η |=1.8.2525
The material breakdown is particularly important at small radius. The pixel barrel radiationlength for perpendicular incidence is approximately 10.7%for the three pixel layers. The break-down is approximately: electronics+bump bonds (1.4%), sensor(1.1%), hybrid (1%), local supportstructure with cooling(5.4%), cables (0.3%) and global support (1.5%). That of SCT barrel layersis 11.8% when averaged over the active area (2.96% per layer including 1.33% for modules, 1.15%2530
for services and 0.48% for the support structure). The module budget includes sensors (0.61%),hybrids (0.22%), pigtails and base-boards. The equivalentmaterial per disk is 3.75% (modules :1.81%, services: 1.47% and support structure: 0.47%). Services include the electrical services,cooling blocks and pipes. The support structure includes the outer support cylinder. Off-barrel oroff-disk services, especially at the interface between thebarrels and end-caps and in the forward2535
direction from pixel services, and items such as the thermalenclosures, are not included.
4.9 Inner detector performance
This section must be re-discussed following full ID integration into the pit. It should include shortnoise summaries of each detector, short summaries of bad channels for each detector, a summary ofDCS monitoring results (cooling for pixel and SCT, power supplies for all) and if possible cosmic2540
rays in pit. It will be limited to 5 pages.
4.9.1 Electrical performance of the integrated detector
As noted in section 4.5, the electrical performance of individual pixel, SCT and TRT moduleswas monitored throughout assembly, and some modules were irradiated to the total dose expectedafter 10 years of LHC operation. Following delivery to the CERN SR1 building, the SCT barrels2545
– 91 –
z(mm)-3000 -2000 -1000 0 1000 2000 3000
radi
us(m
m)
-1000
-500
0
500
1000
Figure 52. Mapping of photon conversions as a function ofη and radius, integrated overφ, for the ID. Themapping assumes 150000 photons originating from the primary vertex.
Figure 53. Material distribution (X0, λ0) at the exit of the ID envelope, including the services and thermalenclosure. The distribution is shown as a function of| η |, averaged overφ. The breakdown indicates thecontribution of external services, and individual sub-detectors, including services in their active volume.
and end-caps were fully connected to the SCT DAQ. The equivalent noise charge (ENC), noiseoccupancy at the nominal 1 fC operating threshold and the number of defective channels werethen measured and compared with previous data. After the SCTbarrels and end-caps were fullyintegrated into the TRT, connectivity checks were repeated.
Prior to installation in the ATLAS pit, two opposite sectorsof the SCT and TRT Barrel were2550
tested. The connected sectors comprise 1/8 of the TRT and 468modules of the SCT barrels asshown in Fig. 55. A partial test was also made of end-cap C. Onequadrant (247 modules) of theSCT and 1/16 of the TRT wheels were connected. The tests emulated the final pit configuration, inparticular the service routing and detector grounding. Onepixel end-cap was also operated under
– 92 –
|η|0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Thi
ckne
ss in
Rad
iatio
n Le
ngth
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
|η|0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Thi
ckne
ss in
Rad
iatio
n Le
ngth
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2 OutsideSupports/OtherCablesCooling
ElectronicsActiveBeam Pipe
|η|0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Thi
ckne
ss in
Nuc
l Int
er L
engt
hs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
|η|0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Thi
ckne
ss in
Nuc
l Int
er L
engt
hs
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7 OutsideSupports/OtherCablesCooling
ElectronicsActiveBeam Pipe
Figure 54. Material distribution (X0, λ0) at the exit of the ID envelope. The distribution is shown as afunction of | η |, averaged overφ. The breakdown shows the contribution of different ID components,independent of the sub-detector.
| η |=0 | η |=1.8
Sampling Sampling
radius (mm)X0 radius (mm)X0
Exit beam pipe 36 0.004536 0.014
Exit pixel layer 057 0.037 57 0.105
Exit pixel layer 2172 0.108 172 0.442
SCT entry 253 0.119 253 0.561
TRT entry 552 0.205 621 0.910
Exit TRT 1081 0.458 907 1.074
Table 19. Integrated radiation length (X0) estimated as a function of radius R for| η |=0.0 and| η |=1.8using Geant4. The quoted radii are the radii at which the sampling is made. The data are averaged overφ.
realistic conditions and cosmic ray data studied.2555
The goals of the SCT-TRT combined tests included the commissioning of DCS issues such asthe cooling system, measurements of the noise performance of the combined detectors under a widerange of operating conditions, to ensure the absence of cross-talk, to study timing and synchronisa-tion issues and to obtain experience in the combined operation as preparation for commissioning.Cosmic ray tracks through the SCT/TRT barrels and end-caps were also used to study their per-2560
formance. Following the SCT-TRT integration, a total of 0.3% SCT channels (Table 20) and<2%TRT channels (Table 21) were found to be defective.
The number of defective channels in the end-cap and barrel pixel detector elements is<0.2%as summarised in Table 22.
The pixel detector was not included as part of the combined SCT-TRT barrel cosmic test since2565
– 93 –
Figure 55. Photograph of the ID Barrel setup for cosmic ray studies (left) and the configuration of modulegroups chosen for this test (right).
Total Not Dead Not Part Noisy Other Total
channelsbondedchannelsreachedbonded defectsdefects
Barrel 3244032 793 3885 1536 401 3130 117 0.304%
End-cap3035136 811 6464 - 352 230 5 0.259%
Table 20.Number of defective channels in the SCT after integration.These numbers will later be updated.
Total Dead Noisy Total
channelschannelschannelsdefects
Barrel side A 52544 926 <1% 1.8%
side C 52544 1050 <1% 2.0%
End-capside A 122880 521 <1% 0.4%
side C 122880 921 <1% 0.7%
Table 21.Number of defective channels in the TRT after integration.These numbers will later be updated.
it is installed as a complete unit into the ID. However, one pixel end-cap was oriented verticallysuch that a reasonable cosmic ray rate (about 1 Hz) could be obtained using a simple scintillatortrigger. A prototype of the complete internal pixel services was used to bring power and cooling to
– 94 –
Defective channels (%)
End-cap A Disk 1A Disk 2A Disk 3A Average defects (%)
0.14 0.23 0.43 0.22
End-cap C Disk 1C Disk 2C Disk 3C
0.14 0.23 0.43 0.22
Barrel Layer 0 Layer 1 Layer 2
0.07 0.18 0.26 0.20
Pixel average 0.19
Table 22. The percentage of defective pixel channels prior to the finalintegration of the detector.Thesenumbers will later be updated.
the end-cap, and to allow for optical readout. The external cables, power supplies and read out wereclose to those used in the final detector. The complete end-cap (144 modules) was also operated.2570
The noise of the pixel, SCT and TRT sub-detectors was measured in many different configu-rations from data collected using random triggers.
The modules in the pixel end-cap were tuned for a threshold of∼ 4000e. The average noise foreach active module was∼170eand the distribution is shown in Figure 56a. The average noise seenin the cosmic ray test was about 10% lower than that measured during individual module testing2575
under similar, but not identical, circumstances. The observed pixel occupancy per BCID (beamcrossing ID) included fixed pattern noise from a small fraction of the channels. Most of these hot(noisy) pixels were previously identified during individual module tests using an241Am source.The pixel occupancy for the active modules in one of the end-cap disks is shown in Figure 56bafter masking of the hot pixels. About 5×10−5 of the active pixels were masked for the modules2580
shown. BCID 5 corresponds to the peak of the cosmic ray timingdistribution with small tailsbefore (BCID 4) and after (BCID 6). The pixel occupancy per BCID for other BCIDs was foundto be 10−9−10−10 and is indicative of the random noise occupancy for these operating conditions.
For the SCT detector, the ENC noise performance and noise occupancy has been compiled forthe individual barrels and end-caps. For example, the distribution of measured noise occupancies of2585
all active module sides of the barrel and end-cap cosmic ray runs are shown in Figure 57a and hasa mean of< 5×10−5. The corresponding ENC values extracted from threshold scans are shownin Figure 57b for all active barrel and end-cap C modules. Allnumbers are within specifications.
In addition, the SCT noise was measured in extreme conditions: varying the trigger rate from5 Hz to 50 kHz, with and without TRT operation, with the cooling heaters on, off and switching2590
between the two states, with several grounding schemes and while the TRT was being read out.No increase of noise was observed in any of the tested configurations. Similar tests were alsoperformed for the TRT barrel: the noise was verified before and after the insertion of the SCT, ina configuration in which the TRT analogue ground was connected to the SCT power return, fordifferent SCT thresholds and while the SCT was being readout. The straw noise occupancy was in2595
– 95 –
Figure 56. (a) The average noise in electrons for each active module in the pixel end-cap cosmic raytest.(b) Pixel occupancy for active modules for one end-capdisk during the pixel end-cap cosmic ray test asdescribed in the text. BCID=5 corresponds to the peak of the cosmic ray timing distribution and BCID=4,6corresponds to adjacent time bins. The occupancy for other BCID values represents a measurement of therandom pixel noise occupancy.
ENC)-Noise (e600 800 1000 1200 1400 1600 1800 2000
Num
ber
of c
hips
1
10
210
310 ENC-Barrel : 1586 e
ENC-EC Outer : 1575 e
ENC-EC Middle : 1551 e
ENC-EC Inner : 1105 e
Figure 57. (a) Distribution of mean noise occupancy for all active module sides of the SCT barrel and end-cap (outer or middle) modules obtained at a 1 fC threshold. (b) Distribution of the ENC values of all activemodules in the SCT barrel and end-cap C.
all cases around 2%.
4.9.2 Cooling performance
A satisfactory cooling performance of the pixel and SCT detectors is crucial to their operation, withup to 75 kW dissipated inside the ID detector (17 kW for the pixels, 35 kW for the SCT and 23 kWfor the TRT) and a required temperature stability within theSCT and pixels of∼2C.2600
The performance of the pixel end-cap sectors was measured during the cosmic ray tests de-scribed above. The evaporative cooling was operated with a base temperature (no power) of approx-imately−25C. Temperatures on each module were monitored. The average increase in tempera-ture on a module for nominal power (23W for the six modules on asector) was 9-10C (operationat -16 to -15C) for the eight sectors on a disk and projected to be 13-14C for end-of-lifetime2605
– 96 –
Module position along the stave5 10 15 20 25
C)
oA
vera
ge T
(
25
25.5
26
26.5
27
27.5
28
28.5
29
29.5
30
Figure 58.The mean and rms (indicated by error bars) of all hybrid temperatures for a given module position(from the cooling inlet to the cooling exit), averaged over all cooling loops in the combined SCT-TRT run.The exhaust cooling-pipe temperature was∼10C.
power (about 38W). The barrel pixel system has not been measured in such a realistic way. How-ever, extensive measurements have been made on single and bi-staves. The expected differencein temperature between the coolant (nominally−25C) and a module on a bi-stave is expected tobe∼15C (due to some corrosion problems with the aluminium pipes, some staves in layer 2 hadto be reworked and the temperature gradient may be as high as∼23C in these cases).2610
The various phases of integration have allowed a preliminary study of the SCT cooling per-formance. For the combined barrel test, a cooling pipe exhaust temperature of approximately 10Cwas used. The average temperature on the barrel module hybrids (read by a thermistor on eachmodule side) was∼10C above the coolant temperature (unpowered) and∼17-19C (powered).When averaged over the operating cooling loops, the mean andspread of the hybrid temperature for2615
each module along a barrel cooling loop is shown in Figure 58.For the end-cap, this temperaturedifference was slightly higher,∼15C, with respect to the selected exhaust temperature of 1C.However, the sensor temperature is slightly lower than thatof the barrel because of the existenceof 2 cooling contacts for each module. The total power dissipated on the SCT barrel is 11.7 kWcorresponding to∼265 W per cooling loop. In each end-cap, the power dissipation is 11 kW.2620
4.9.3 Cosmic-ray track reconstruction
The software infrastructure developed in ATLAS to simulateand reconstruct the LHC data has beenadapted for the cosmic rays data. A detector description wascreated for both the SR1 pixel andcombined SCT-TRT barrel setups with no magnetic field [97, 98] and a simulation of cosmic raystraversing the detector setup was implemented to test the full reconstruction chain and to compare2625
the simulated and reconstructed data. Specific pattern recognition algorithms [99] developed withinthe tracking framework [100] and two of the track fitters designed to work at the LHC (one basedon aχ2 minimisation method and the other on a Kalman Filter fit) wereused.
Several alignment methods [101, 102, 103] are being developed to align the ID using LHCdata and they have been applied to the SCT/TRT cosmic ray setup. They are the so-called silicon2630
globalχ2, local χ2 and robust algorithms. A combined algorithm [104] also based on a global andlocal χ2 minimisation was also developed to deal with the SCT and TRT alignment simultaneously
– 97 –
and was tested for the first time with cosmic data. Due to the lack of magnetic field, however,the track momentum cannot be properly estimated and therefore the ultimate detector resolutioncannot be extracted (the standard deviation of the residualdistributions are a convolution of the2635
detector resolution and track parameter uncertainties). The track reconstruction efficiency and theassociated hit efficiencies have been evaluated for each SCTbarrel module and layer side, andexceed 99%. Studies of the TRT alignment are in progress.
Global SCT-TRT barrel misalignments have been studied by applying an alignment algorithmthat extracts single relative translation and rotations from track segments that have been fitted sep-2640
arately in SCT and TRT. The SCT segments are propagated to thenearest TRT hit to compute aresidual in that TRT straw layer surface and theχ2 minimised. Since the TRT only measures trackparameters in thexy plane, misalignments along thez-axis are not measured. Furthermore, thecosmic trajectories were nearly parallel in the setup and therefore both∆x and∆y, nor∆rot−x and∆rot− y, are not reliably extracted simultaneously. Since tracksare more vertical than horizontal2645
∆x, ∆rot−y and∆rot−z were obtained. Their values are shown in Table 23 and are in reasonableagreement with the survey measurements (see section 4.6).
Run ∆x [mm] ∆rot−y (mrad) ∆rot−z(mrad)
3007 -0.290 0.277 0.254
3099 -0.289 0.293 0.226
Table 23.Mean re-alignments for the global alignment of TRT and SCT.
The cosmic ray test of the pixel end-cap was used to determinepreliminary alignment con-stants for each working module in the end-cap. This measurement took advantage of the overlapregion (about 20%) between modules on the front and back of the disks. Alignment constants2650
were obtained from the cosmic ray data and applied in both thelocal X coordinate (the precisioncoordinate with pitch of 50µm) and in local Y (pixel length 400µm). The pixel residuals beforeand after applying the alignment constants are shown in Figure 59 and compared with Monte Carlopredictions (using a simple cosmic ray Monte Carlo) in Table24. The values shown in Table 24 arederived from fits to residuals in Fig. 59. The improvement in accuracy after alignment is significant2655
and the agreement with the expectations from the simple Monte Carlo is good.
4.9.4 Inner detector installation and performance in situ
To be filled in if possible for final draft.
– 98 –
Before [µm] After (µm) MonteCarlo (µm)
σ(dX) 21 18 16
σ(dY) 119 117 117
Table 24. Pixel residuals from a cosmic ray test of a pixel end-cap before and after alignment corrections,and a comparison with Monte Carlo predictions.
dLocX (mm)-0.25 -0.2 -0.15 -0.1 -0.05 -0 0.05 0.1 0.15 0.2 0.250
100
200
300
400
500
dLocY (mm)-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1
0
50
100
150
200
250
300
350
Cosmic Run 1129
Nu
mb
er o
f E
ven
tsN
um
ber
of
Ev
ents
Mean = -0.003437
Sigma = 0.1697
Mean = -0.00837
Sigma = 0.02999
(a)
Corrected dLocX (mm)-0.2 -0.1 0 0.1 0.2
Nu
mb
er o
f E
ven
ts
0
100
200
300
400
500
Corrected dLocY (mm)-1 -0.5 0 0.5 1
Nu
mb
er o
f E
ven
ts
0
50
100
150
200
250
300
350
Cosmic Run 1129 After Correction
Mean = 0.0001231
Sigma = 0.02522
Mean = 0.0009362
Sigma = 0.166
(b)
Figure 59. (a) Distribution of pixel residuals before alignment corrections. (b) Distribution of residuals afteralignment corrections. See text for the definition of local coordinates LocX and LocY.
– 99 –
5. Calorimetry
5.1 General description, cryostats and feed-throughs2660
in italic points to be developed or remarksin roman characters the text itself as.
5.1.1 General Description
The ATLAS calorimetry [105] system, as shown in Fig. 3 (see section 1.1.4), is composed of sev-eral sampling detectors, disposed on an inner and an outer cylinder along the beam axis. The2665
inner calorimeters are shown in Fig.60 and comprise three cryostats, one barrel and two end-caps. The barrel cryostat contains the electromagnetic barrel calorimeter, whereas the two end-cap cryostats each contain an electromagnetic end-cap calorimeter (EMEC), a hadronic end-capcalorimeter (HEC), located behind the EMEC, and a forward calorimeter (FCal) at small angle.They use liquid argon as the detector medium, which has been chosen for its intrinsic linear be-2670
haviour, its stability of response over time and its radiation tolerance.
Figure 60. Sketch of the ATLAS liquid argon calorimeters.The electromagnetic barrel, the electromagneticend-cap (EMEC), the hadronic end-cap (HEC) and the forward calorimeter (FCal) are indicated with thenumbers 1 to 4. One can also see the three cryostats housing these calorimeters.
The precision electromagnetic calorimeters are lead-liquid argon detectors with accordion-shape absorbers and electrodes. This geometry allows to have several active layers in depth, three inthe precision-measurement region (0<|η | < 2.5) and two in the higher-η region (2.5< |η | < 3.2)and in the overlap region between the barrel and the EMEC. In the precision-measurement region,2675
the electromagnetic calorimeters are completed by presamplers, an argon layer, which provides ameasurement of the energy lost in front of the electromagnetic calorimeters. An accurate position
– 100 –
measurement is obtained by finely segmenting the first compartment in η . The η -direction ofphotons is determined by the position of the photon cluster in the first and the second compartment.
The calorimetry system also has electromagnetic coverage at high η (2.5 < |η | < 4.9) pro-2680
vided by the Forward Calorimeter (FCal). The resolution is not as good as at low|η | and thereis no tracking to allow electron/photon separation. However this coverage will be important inphysics channels such as the measurement of electro-weak couplings via the forward-backwarddecay asymmetry of a heavy Z’.
For the outer hadronic calorimeter, the hadronic barrel calorimeter, the sampling medium is2685
scintillator and the absorber medium is iron. The barrel calorimeter is composed of three parts,a central and two extended barrels. The choice of tile calorimeter technology provides maximumradial depth for the least cost for ATLAS. The hadronic central barrel covers the 0. < |η | < 1.7range. The hadronic calorimetry is completed at large pseudorapidity by the HEC, a copper-liquidargon detector, and the FCal, a copper/tungsten-liquid argon detector. Theη coverage for hadronic2690
calorimetry extends to|η | = 4.9.The numbers of radiation and interaction lengths in front ofand in the electromagnetic calorime-
ters are shown in Fig. 61 and Fig.62.
Table 3 shows theη coverage for the electromagnetic and hadronic calorimetry. There is a fullφ coverage for all calorimeters.2695
Section 5.1.2 describes the LAr cryostats and feed-throughs. Sections 5.2 and 5.3 are devotedto the description of the electromagnetic and hadronic calorimetry respectively. The front-endreadout electronics, back-end electronics and services are described in section 5.5. Finally, test-beam performances of production modules of the different calorimeters are shown in section 5.6.
5.1.2 Cryostats and associated feed-throughs2700
As mentioned above and illustrated in Fig. 60, the liquid argon calorimeters are housed in threedifferent cryostats. The barrel cryostat houses the electromagnetic barrel calorimeter and the su-perconducting solenoid coil, whereas the two end-cap cryostats each house one electromagneticend-cap calorimeter (EMEC), one hadronic end-cap calorimeter (HEC) and one forward calorime-ter (FCal).2705
5.1.2.1 Cryostat description The barrel cryostat and the two end-cap cryostats (Fig. 63) and thebarrel cryostat are each composed of two concentric aluminium vessels, a cold and a warm vessel.Each vessel forms a cylindrical torus centred on the beam axis. The space in between the vesselsis evacuated. Furthermore the central solenoid is housed inthe insulating vacuum of the barrel2710
cryostat and supported by the inner cylinder of the warm barrel vessel.The different vessels are closed at both ends by bulkheads. All bulkheads are circular shaped
plates with passage holes to allow the insertion of the innertracking detector into the inner warmbarrel cylinder bore and to allow the passage of the vacuum beam pipe. All four bulkheads andthe inner cold vessel of the barrel cryostat have tapered walls to minimise the material between the2715
interaction region and the electromagnetic calorimeters.To also minimise the amount of material,the cold and warm front end-cap bulkheads are flat and insulating spacers allow the warm bulkheadto sit on the cold one.
– 101 –
Pseudorapidity0 0.5 1 1.5 2 2.5 3
0X
0
2
4
6
8
10
Pseudorapidity0 0.5 1 1.5 2 2.5 3
0X
0
2
4
6
8
10Before accordion
Before presampler
Pseudorapidity1.4 1.45 1.5 1.55 1.6 1.65 1.7
0X
0
5
10
15
20
25
30
35
40
Pseudorapidity1.4 1.45 1.5 1.55 1.6 1.65 1.7
0X
0
5
10
15
20
25
30
35
40 End-cap accordionBefore end-cap accordionBefore end-cap presamplerBefore scintillatorBarrel accordionBefore barrel accordionBefore barrel presampler
Pseudorapidity0 0.2 0.4 0.6 0.8 1 1.2 1.4
0X
0
5
10
15
20
25
30
35
40
Pseudorapidity0 0.2 0.4 0.6 0.8 1 1.2 1.4
0X
0
5
10
15
20
25
30
35
40 Layer 3Layer 2Layer 1
Before accordion
Pseudorapidity1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2
0X
0
5
10
15
20
25
30
35
40
45
Pseudorapidity1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2
0X
0
5
10
15
20
25
30
35
40
45 Layer 3Layer 2Layer 1
Before accordion
Figure 61. Amount of material, in units of radiation lengthX0 and as a function of|η |, in front of theelectromagnetic calorimeters. The top left-hand plot shows separately the total amount of material in frontof the presampler layer and in front of the accordion itself over the fullη -coverage. The top right-hand plotshows the details of the crack region between the barrel and end-cap cryostats, both in terms of material infront of the active layers (including the crack scintillator) and of the total thickness of the active calorimeter.The two bottom figures show, in contrast, separately for the barrel (left) and end-cap (right), the thicknessesof each accordion layer as well as the amount of material in front of the accordion.
The cold vessels rest on four feet made out of fibre-glass epoxy composite, which provideselectrical and thermal insulation to the warm vessels. The outer warm cylinders are equipped at2720
the lower half of their two ends with four support interfaces. The barrel cryostat load is transferreddirectly to the ATLAS main rail system through the tile-calorimeter support saddles. This is notpossible in the case of the end-cap cryostats, so the load is transferred through the inner radiusof the extended tile calorimeter through an adjustable support located on the side of the end-capcryostat pointing towards the interaction point.2725
5.1.2.2 Signal feed-throughs The signal feed-throughs, 64 for the barrel and 2× 254 for the
4For each end-cap, 20 feed-throughs are used by the EMEC, fourby the HEC and one by the FCal (the EMEC usesalso part of the four HEC feed-throughs.
– 102 –
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50
2
4
6
8
10
12
14
16
18
20
Pseudorapidity0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Inte
ract
ion
leng
ths
0
2
4
6
8
10
12
14
16
18
20
EM caloTile1
Tile2
Tile3
HEC0
HEC1
HEC2
HEC3
FCal1
FCal2
FCal3
Figure 62. Amount of material, in units of interaction length, as a function of |η |, in front of the electro-magnetic calorimeters, in the electromagnetic calorimeters themselves, in each hadronic compartment, andthe total amount at the end of the active calorimetry. Also shown for completeness is the total amount ofmaterial in front of the first active layer of the muon spectrometer (up to|η | < 3.0).
Front−end Crates
Forward CalorimetersHadronic End−CapCalorimeters
ElectromagneticEnd−Cap Calorimeter
Feedthroughs and
Figure 63. An artist’s view of an end-cap cryostat with a cut-away showing the positions of the three end-cap calorimeters. The outer radius of the cylindrical cryostat vessel is 2.25 m and the length of the cryostatis 3.17 m.
end-caps bring 122 880 and 2×48 000 signal, monitoring, calibration and spare lines through aninsulating vacuum, from the the liquid argon into the warm environment of the barrel and end-capcryostats.
A feed-through [106], consists primarily of a warm flange anda cold flange, with a flexible2730
bellows welded between them. The volume between the two flanges is pumped. Each flange
– 103 –
houses 4 gold-plated pin carriers providing a total of 1920 signal connections per feed-through.Both flanges are interconnected with flat, polyimide strip-line flexible 33Ω5 cables. To minimisethe lengths of the connection lines, the signal feed-throughs are distributed radially around eachend of the barrel cryostat and at the high|z| end of end-cap cryostat.2735
5.1.2.3 High-voltage feed-throughs High-voltage (HV) feed-through ports are mounted on eachcryostat6, close to the highest point. A port consists of a single warm bulkhead erected on the warmvessel from which the wire bundle leads to the calorimeter via a tube connected to the cold vessel bystainless steel bellows. The liquid rises up to the bellows followed by a 70 cm high column of argongas at 1.25 bar7. The HV wire consists of a solid 0.41 mm diameter constantan conductor with2740
a 0.30 mm thick PEEK (PolyEther-Ether-Ketone) cladding. Each port holds about 840 HV wires.This number of channels allows to serve the standard HV linesand the remainder are spares, whichare used to supply single electrodes or split sectors.
5.2 Electromagnetic calorimetry
5.2.1 Accordion geometry2745
The accordion geometry has been chosen for the absorbers andthe electrodes of the electromag-netic calorimeters. Such a geometry allows a full coverage in φ and a fast extraction of the signalat the rear or at the front of the electrodes (Fig. 64). In the barrel the accordion waves are parallelto thezdirection and run inφ, the folding angles of the waves are varying withR to keep the argongap constant. In the endcap the waves are parallel to ther direction and run inz, and as the liquid2750
argon gap increases with the radius, the wave height and the folding angle vary with the radius.The absorbers are made of lead plates (thickness 1.53 mm forη < 0.8, 1.13 mm forη > 0.8
for the barrel and 1.7 mm forη < 2.5 and 2.2 mm forη > 2.5 for the endcap), glued betweentwo 0.2 mm thick stainless steel sheets by resin-impregnated glass fiber fabric. The stainless steelsheets provide mechanical strength to the absorber.2755
The readout electrodes [107] consist of three conductive copper layers separated by insulatingpolyimide sheets. The two outer layers are at the high voltage potential and the inner one is usedfor reading out the signal via capacitive coupling. The segmentation of the calorimeter inη andin depth is obtained by etched patterns on the different layers (Fig. 64). Theφ segmenation isobtained by ganging the appropriate number of electrodes. Each barrel gap between two absorbers2760
is equipped with two electrodes, one for|η | < 0.8 and one for|η | > 0.8 whereas each endcap gapbetween two absorbers|η |< 2.5 is equipped with one electrode and each endcap gap between twoabsorbers for|η | > 2.5 is equipped with another electrode.
5.2.2 Barrel geometry
The ATLAS EMB calorimeter is made of two half barrels, centered around the z-axis (ATLAS2765
beam axis). One half barrel covers thez > 0 region (pseudorapidityη > 0) and the other one theregionz < 0 (pseudorapidityη < 0), from |η | = 0 to |η | = 1.475; the length of each half barrel is
5This impedance value is in between the 25Ω and 50Ω signal cables used to connect the calorimeter to the front endboards.
6They are located at each end of the barrel cryostat and at the high-|z| end of each of the end-cap cryostats.7RF gaskets between the ports and the warm vessel ensure gas-tightness.
– 104 –
Figure 64. The figure shows the etched pattern, before folding, of the signal layer of the two electrodes ofan half barrel.
3.2 m, the inner and outer diameters are 2.8 m and 4 m respectively. To take into account the effectof the material in front of the calorimeter, the barrel calorimeter is complemented with a liquidargon presampler detector placed in front of its inner surface, over the fullη range (see below).2770
A half barrel is made of 1024 accordion shaped absorbers, interleaved with readout electrodes.The electrodes are kept in the middle of the gap by honeycomb spacers. The drift gap on each sideof the electrode is 2.1 mm, which corresponds to a total drifttime of about 450 ns for an operatingvoltage of 2000 V. Once assembled, there is no discontinuityalong the azimuthal angleφ; but forease of construction, a half barrel is divided into 16 modules, each covering a∆φ angle of 22.5.2775
The total thickness of a module is at least 22 radiation lengths (X0), increasing from 22 X0 to 30 X0
between|η | = 0 and|η | = 0.8, and from 24 X0 to 33 X0 between|η | = 0.8 and|η | = 1.3.
At the inner and outer edges, each absorber is encased into the groove of a precision machinedG10 fiberglass-epoxy composite bar. The purpose of these bars is to position each absorber withrespect to its neighbours, and also to provide space for the connectors of the electrodes. The piling2780
of these bars define the cylindrical geometry.
Seven stainless steel outer rings support and give rigidityto a half barrel. Each ring is made of16 ring-pieces corresponding to the 16 modules. They are allidentical and have a cross-sectionalshape of an I-beam except for the two ring-pieces at the levelof the cryostat rail. Similarly, 8composite inner rings help to define the inner geometry of a half barrel. Each inner ring is also2785
made of 16 ring-pieces. The absorber bars are screwed onto the ring-pieces.
A module (Fig. 66 and Fig. 65) has three compartments in depth(front, middle, back as viewedfrom the interaction point). The readout granularity of thedifferent compartments is shown in Ta-ble 3. In total there are 3424 readout cells per module, including the presampler cells. The amountof dead material in front of the presampler and between the presampler and the first calorimeter2790
compartment as well as the thickness of each calorimeter compartment, in units of X0, is shown inFig. 61.
The presampler [108] is needed to correct for the energy lostupstream of the calorimeter,especially at low energy. For this purpose a separate liquidargon presampler, providing shower
– 105 –
Figure 65. The right figure shows a sketch of a barrel module where the different compartments are clearlyvisible and also the ganging of electrodes inφ.
sampling in a thin liquid argon layer (11 mm in depth), is placed in front of the electromagnetic2795
calorimeter inside the common barrel cryostat. It is made of64 identical azimuthal sectors (i.e. 32for a half barrel). Each sector is 3.1 m long and 0.28 m wide thus covering the half barrel length andgiving a coverage inη andφ of 1.52 and 0.2 respectively. It is composed of eight different sizedmodules with a length increasing withη to obtain the sameη coverage of 0.2 for each moduleexcept the one at the end of the barrel whoseη coverage is reduced to 0.12.2800
The modules are made of interleaved cathode and anode electrodes glued between fiberglassepoxy composite plates. The electrode spacing slightly varies with presampler module type from1.9 to 2.0 mm. The cathodes are double-sided printed circuitboards while the anodes have 3conductive layers separated by fiberglass epoxy composite layers. The required segmentation(∆η = 0.025,∆φ = 2π/64) for each module is obtained by putting together the appropriate num-2805
ber of anodes in theη (or z) direction and by subdividing (i.e. etching) each anode into two halvesin theφ-direction. A +2 kV high voltage potential is applied to the outer layers of the anodes andthe signal is readout through capacitive coupling to the central layer at ground potential. Each mod-ule is equipped with a mother board which is a five layers printed circuit. The mother boards collectsignals from the readout cells. In addition the mother boards are equipped with a set of precision2810
surface-mounted resistors (0.1 % precision and 25 ppm/C) for the injection of calibration pulses.
– 106 –
5.2.3 End-cap geometry
The ATLAS EMEC calorimeters [109] consist of two wheels, oneon each side of the electromag-netic barrel. Each wheel, 63 cm thick with external and internal radii of∼ 200 cm and 30 cm,2815
weighs 27 tons. It covers the pseudorapidity range 1.375< |η | < 3.2 . In the transition regionbetween the barrel and the endcap calorimeters, the material in front of the calorimeter amounts toa fewX0. In order to improve the energy measurement in this region, aliquid argon presampler isimplemented in front of the endcap calorimeter, covering the pseudorapidity range 1.5< |η | < 1.8.
Each endcap calorimeter consists of two coaxial wheels. Theboundary between the inner and2820
the outer wheel, 3 mm wide, is located at|η | = 2.5 and is mainly filled with low density material.It is projective and it matches the boundary of the rapidity range covered by the ATLAS chargedparticle tracking system|η | < 2.5. Each endcap wheel is divided into eight wedge-shaped moduleswithout introducing any discontinuity along the azimuthalangle (thanks to the accordion concept).A sketch of a module is shown in Fig. 66. Each endcap is made of 768 absorbers interleaved with2825
readout electrodes in the outer wheel and 256 in the inner wheel. The total active thickness of anendcap calorimeter is larger than 24X0, except for|η | < 1.475. It increases from 24 to 38X0 when|η | runs from 1.475 to 2.5, and from 26 to 36X0 for 2.5< |η | < 3.2.
Figure 66. The left figure shows a partly stacked barrel module. The 6 outer rings on which are screwed theabsorbers can be seen. Also the backbone (yellow part) and the assembling bench (white part) are visible.The right figure shows the side view of an EMEC module (the beamaxis is vertical on this photo). The firstacccordion absorber of each wheel is clearly visible, as well as the summing, motherboards and cables.
In the outer wheel, signals from the different pads are readout at two sides of the electrode, i.e.as above. In the inner wheel, because of the higher radiationlevel, signals are all read out at the2830
back side only. The electrodes are kept centered in the middle of the gaps by honeycomb spacers.
The calorimeter support frame is composed of six support rings: three on the front and threeon the back. The total weight of the detector is supported mainly by the two large external rings.
As for the barrel electromagnetic calorimeter, the precision region in the endcap calorime-ters (1.5 < |η | < 2.5) is divided in depth in three longitudinal compartments. The front compart-2835
ment, about 4.4 X0 thick (about 6 X0 when including the dead material in front of the calorime-ter), is segmented with narrow strips along theη direction. The projective tower transverse size
– 107 –
in the middle compartment is the same as defined in the barrel electromagnetic calorimeter :∆η × ∆φ = 0.025 × 0.025. The back compartment has a twice coarser granularity inη .The low pseudorapidity (|η | < 1.5) region of the inner wheel and highest pseudorapidity region2840
of the outer wheel, are segmented in only two longitudinal compartments and have a coarser trans-verse granularity. Table 3 summarizes the longitudinal andtransverse readout granularity in theelectromagnetic endcap calorimeter as a function of theη range. The∆η granularity in the frontsampling varies withη in order to keep the copper strip width larger than a few mm. The ∆φgranularity is obtained by gathering the signals from adjacent electrodes. In the outer wheel, 122845
adjacent electrodes are summed for a front cell and 3 for a middle or back cell. In the inner wheel,4 adjacent electrodes are summed for any cell. Each endcap calorimeter counts in total 31872readout channels (including the 768 from the presampler).
Each endcap presampler consists of 32 identical azimuthal sectors (modules). It is placed ina 5 mm deep cavity in the back side of cryostat cold wall. The granularity of the presampler is2850
∆η ×∆φ = 0.025× (2π/64). One endcap presampler module consists of two, 2 mm thick activeliquid argon layers, formed by three electrodes parallel tothe front face of the EMEC calorimeter.The electrodes, made from double sided printed circuit boards, are separated by honeycomb spac-ers and glued together at the inner and outer radius with 2 mm thick G10 bars. A negative highvoltage is applied to the external electrodes and the signals are read out from the central electrode2855
segmented into pads. The pads are connected by pins to the 50Ω strip readout lines printed on thetop external electrode. The strip lines are connected at theouter radius of the module to the 50Ωcoaxial cables which lead the signals to the feedthroughs. The same signal, calibration and HVcables as for the endcap calorimeter are used. Two HV cables feed separately the left and right sideof a presampler module. Three calibration lines (one per 8 channels) are connected to the signal2860
pads through calibration resistors of 3.48 kΩ (0.1% tolerance).
5.2.4 Calorimeter alignment
After insertion in the cryostat, the circularity of the EMB calorimeter was measured. The deforma-tion due to the weight can reach up to 3 mm at the top and bottom of EMB and is in a fair agreement2865
with a finite element calculation [110]. At cold the deformation is reduced due to the Archimede’sthrust caused by the liquid argon (density 1.4). The centersof the fitted inner and outer circlescoincide to better than 0.5 mm. Finally the sagging of the absorbers was measured as a functionof the azimuthal angle, with a maximal value of 2.5 mm near thehorizontal plane. The same setof measurements have been performed for both endcaps, once they were in vertical position. The2870
measured deformation is up to 2 mm at the top.
Due to difficulties in the positioning of the half barrels inside the cryostat, both half barrelswill be at cold 4 mm too low and 2 mm laterally displaced with respect to the cryostat’s central axis[110]. Part of the vertical misalignment will be compensated by positioning the cryostat inside AT-LAS about 2 mm higher than originally planned. The remainingmisalignments will be corrected2875
by software.
– 108 –
5.2.5 High-voltage distribution
To have redundancy, the two HV sides of an electrode are fed independently. The nominal highvoltage for the EM barrel is 2000 V (see table 25). If one side of an electrode is not powered, only2880
half of the signal will be collected. The variation of the signal with HV, mainly due to the variationof the drift velocity with HV, is shown in Fig.67. When reducing the high voltage by a factor 2, themeasured signal is only reduced by 33%.
The EMB segmentation is∆φ× ∆η = 0.2× 0.2, meaning that 32 electrodes sectors arepowered simultaneously. The EMB preshower segmentation isthe same as above.2885
0
25
50
75
100
125
150
175
200
225
250
0 250 500 750 1000 1250 1500 1750 2000
Figure 67. The 2 figures show the cluster energy as a function of the applied HV. Left figure: Barrel module(245 GeV electrons shown with open circles, 100 GeV shown with diamonds, 100 GeV results scaled to thelast point of the 245 GeV results shown with star symbols). Two fits are superimposed: the full HV range fitis shown with a dotted line, the fit to all points with HV> 400V is shown with a dashed line. Right figure:endcap module (193 GeV electrons)
Contrary to the barrel part, the drift gap on each side of the electrodes is not constant for theEMEC, but is a function of the radius (R). It varies from 2.8 mmto 0.9 mm in the outer wheel andfrom 3.1 mm to 1.8 mm in the inner wheel. Thus to get anη -independent detector response onewould need a high voltage continuously varying withη (open circles in Fig. 68). In practice, it isapproximated with a variation by steps (closed triangles inFig. 68). Table 25 explicitly gives these2890
nominal high voltage values as function of theη range, defining the 7 (2) HV sectors of the outer(inner) wheel. Therefore the HV segmentation is 0.2 inφ as in the barrel and around 0.2 inη asindicated in table 25.
Different solutions have been implemented in case of HV problems: one electrode sector ispowered individually if the problem was seen at room temperature, one HV sector is divided in2895
two (i.e. ∆φ×∆η = 0.1× 0.2 in the barrel) and the two halves are powered separately, iftheproblem was seen only in the cryostat at liquid argon temperature. The effect of not supplying highvoltage to one side of the electrode was studied in the test beam. A slight deviation to the expectedfactor 2 in the signal height observed at the transition region with a good sector was understood asa geometrical effect and is well reproduced by simulation and is therefore easy to correct.2900
– 109 –
Electromagnetic barrel (EMB) |η | range0./1.475 HV 2000
Barrel preshower |η | range 0./1.52 HV 2000
Electromagnetic endcap (EMEC)
|η | range 1.375/1.5 1.5/1.6 1.6/1.8 1.8/2.0 2.0/2.1
HV 2500 2300 2100 1700 1500
|η | range 2.1/2.3 2.3/2.5 2.5/2.8 2.8/3.2
HV 1250 1000 2300 1800
Endcap preshower |η | range 1.5/1.8 HV -2000
Hadronic endcap (HEC) |η | range 1.5/3.2 HV 1800
Forward calorimeter |η | range 3.1/4.9 HV 250/375/500
FCal1, FCal2, FCal3
Table 25.Nominal high voltage values, in Volts, for the liquid argon calorimeters as function ofη .
5.2.6 On-detector cold electronics
0.81
1.21.41.61.8
22.22.42.62.8
1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2
HT (kV)
ETA
Figure 68. EMEC high voltage distribution. A uni-form calorimeter response would need an high volt-age continuously variable alongη (open circles). Ac-tually it is approximated by discrete values (solid tri-angles).
The front section of each calorimeter is readout at the inner core, whereas the middle andback sections are read out at the outer core.2905
The readout is accomplished by several in-terconnected boards. These include summingboards that generate analog sums from adja-cent calorimeter gaps and mother boards foranalog sum readout and for calibration pulse2910
distribution.
The summing boards are connected tothe electrode connectors grouping the sig-nals inφ. Each EMB (outer EMEC wheel)summing board services 16 (12) electrodes2915
(∆φ = 0.1). In the front section, all electrodes are summed into one readout cell, while in themiddle and back section they are summed in 4 adjacent readoutcells (∆φ = 0.025). The innerEMEC wheel summing boards group 8 electrodes (∆φ = 0.2). For both first and second samplings,4 adjacent electrodes are summed to form readout cells (∆φ = 0.1).
A mother board provides readout for a∆η × ∆φ zone of 0.2 × 0.2 . The receptacles at the2920
output of the summing board and the input of the mother board are connected via pins mounted on2 thin PC-boards. Each channel is then connected twice. For the outer radius, one out of every twoPC-boards also houses a low capacitance transient voltage suppressor (TVS) [110] with a turn-onat 6 V to protect against accidental discharges that may damage the calibration resistors mounted
– 110 –
on the mother board.2925
The mother boards route the outputs to the readout cables through a “low profile” connec-tor. The mother boards also include 0.1 % precision tantalumnitride calibration resistors (non-hygroscopic with a temperature coefficient of 70± 5ppm/C) to distribute calibration pulses to allreadout channels. The value of the calibration resistor in various parts of the detector are chosen tomatch the expected currents from particle showers [110, 109]. The calibration pulse is injected at2930
the analog input to the mother board, the closest point to theorigin of the analog signals from theelectrodes.
5.2.7 Quality control tests
At each stage of the construction and of the assembly of the calorimeters, several quality control2935
tests have been performed. The most important tests were those performed after insertion of thecalorimeters inside their cryostats and after the first cool-down and first liquid argon filling. All thesignal and calibrations channels have been tested. Table 26shows that the number of missing orbad channels is well below our predefined acceptable number of faults. Similarly, a high voltagetest at the nominal value was performed. After this first coldtest, the calorimeters were heated at2940
room temperature before transportation and installation inside the ATLAS pit. High voltage testshave also been performed on the EMB and one EMEC in their final position in ATLAS. About 2%of the HV channels with shorts have been found. (waiting for final numbers). In the case of theEMB preshower and of the EMEC, the shorts can be eliminated byapplying a HV discharge [111].Otherwise the sectors with shorts will be powered at a reduced value as explained above.2945
5.3 Hadronic calorimeters
The ATLAS hadronic calorimeters are the tile calorimeter for the central region, the liquid argonhadronic end cap calorimeter (HEC) and the liquid argon forward calorimeter (FCal) in the forwardregions . As FCal has no electromagnetic calorimeter in front, its first compartment is used as anelectgromagnetic calorimeter. They are described in the following three subsections.2950
5.3.1 Tile calorimeter
5.3.1.1 Overview The TileCal calorimeter [112] is a sampling calorimeter using steel as theabsorber and scintillator as the active medium. It is located in the region|η | < 1.7, behind theLiquid Argon calorimeter and is subdivided into a central barrel, 5.8 m in length, and two “extendedbarrels”, 2.6 m in length with each having an inner radius of 2.28 m and and outer radius of 4.25 m.2955
The radial depth of the calorimeter is approximately 7.2λ . Each barrel consists of 64 modules,or wedges, of size∆φ = 0.1, made of iron plates and scintillating tiles. The 4 and 5 mm thickiron plates are assembled to form a frame with pockets in which 3 mm thick scintillating tilesare inserted. The assembled module forms an almost-periodic iron-scintillator structure with aratio by volume of about 4.7:1. The geometry is sketched in figure 69. The orientation of the2960
scintillator tiles radially and normal to the beam line, in combination with wavelength-shiftingfiber readout on the tile edges provides almost complete azimuthal calorimeter coverage. The gapregion between the barrel and the extended barrel is instrumented with special modules made of
– 111 –
layer Total numberMaximal Maximal Nb. found Nb. found
of channels rate number at room in liquid
acceptedacceptedtemperatureargon
Electromagnetic barrel calorimeter
presampler7808 0.05% 4 0 0
1st layer 57216 2/mod. 64 9 11
2nd layer 28672 0.05% 14 2 5
3rd layer 13824 0.05% 6 0 4
barrel-end 2048 0.05% 1 3 3
calibration 8192 0.05% 4 1 1
Electromagnetic endcap calorimeter
presampler1536 0.05% 1 0 1
1st layer 28544 2/mod. 32 3 4
2nd layer 23424 0.05% 12 2 7
3rd layer 10240 0.05% 5 0 2
calibration 5952 0.05% 3 2 2
Table 26. Table showing the number of signal channels in each calorimeter layer and of calibration chan-nels, the predefined acceptable maximum rate and number of faults, the faults found after insertion at roomtemperature and afterwards in liquid argon.
iron-scintillator sandwiches, having the same sampling fraction as the rest of the calorimeter, andwith thin scintillator counters where the free space is limited. These devices allow to partially2965
recover the energy lost in the crack regions of the detector.
The electronics and readout are highly integrated with its mechanical structure. The low volt-age power supplies which power the readout are contained in asmall external section called a“finger”, which has the cross-section of the support girder.The photomultiplier tubes and all front-end electronics are mounted in 1.4m long aluminum units, called “drawers”, which are inserted2970
inside the support girder. The frontend electronics provide a analog sum of subsets of the channels,forming trigger towers, for the ATLAS first level trigger. Finally, the calorimeter is equipped withthree calibration systems: charge injection, laser and a radioactive source. These systems test theoptical and digitized signals at various stages, and are used to set the PMT gains to a uniformity of±3%.2975
5.3.1.2 Mechanical structure The mechanical structure of the tile calorimeter is designed as aself-supporting, segmented structure comprising 64 modules subtending 5.625 degrees in azimuth
– 112 –
for each of the three sections of the calorimeter [113]. The module sub-assembly is shown inFigure 70. Each module contains a precision-machined strong-back girder whose edges are usedto establish a module-to-module gap of 1.5mm and whose key isused to register the elements of2980
the absorber structure. To maximise the use of radial space,the girder provides both the volumein which the tile calorimeter readout electronics is contained and the flux return for the centralsolenoidal field. The readout fibers, suitably bundled, penetrate the edges of the girders throughmachined holes into which plastic rings have been preciselymounted. These rings precisely matchthe position of photomultipliers.2985
Wavelength Shifting Fiber
Scintillator Steel
Source
Tubes
PMT
Figure 69. Schematic showing integration of the optical readoutand mechanical assembly of the tile calorimeter. The variouscomponents of the optical readout - tiles, fibers and the photo-multipliers (PMT) - are shown.
The fundamental element ofthe absorber structure is termed ahalf-period consisting of a 5mmthick master plate, on which 4mmthick spacer plates are glued to2990
form the pockets in which thescintillator tiles are located. Themaster plate was fabricated byhigh-precision die stamping to re-alise the dimensional tolerancing2995
needed to meet the ATLAS speci-fication for the module-to-modulegap. [114]. At the module edges,the spacer plates align into recessedslots in which the readout fibers3000
run. Holes in the master and spacerplates allow the insertion of stain-less steel tubes for the radioactivesource calibration system.
Figure 70. Module to module azimuthal geometry and general layout (left) and internal geometry of amodule showing a readout fiber in contact with a scintillatortile, and three more fibers contained in a plasticchannel that optically isolates them (right)
Each module is constructed3005
by gluing half-periods into sub-
– 113 –
modules on a custom stacking fixture. These are bolted onto the girder to form modules, withcare being taken to ensure that the azimuthal alignment meets specification. The calorimeter isassembled by mounting and bolting modules to eachother in sequence. Shims are inserted at theinner and outer radius load-bearing surfaces to control theoverall geometry and yield a nominal3010
module-to-module azimuthal gap of 1.5mm and a radial envelope which is generally within 5mmof nominal [115, 116]
5.3.1.3 Optical components Eleven sets of scintillating tiles of thickness 3mm and withradiallengths from 97 mm to 187 mm and azimuthal lengths from 200 mm to 400 mm form the activemedium of the calorimeter. Ionizing particles crossing thetiles induce the production of ultraviolet3015
scintillation light in the base material (polystyrene) which is subsequently converted to visiblelight by wavelength-shifting fluors (the scintillator is doped with 1.5% PTP as the primary fluorand 0.04% POPOP as the secondary fluor.) Over 460,000 scintillating tiles were produced forthe Tilecal by injection molding of individual tiles, as this eliminated the need for machining toform the trapezoidal shapes and drilling to cut the holes through which the calibration tubes must3020
pass. The tolerance for all dimensions was held to±0.10mm. 5% of the tile production was testedwith a radioactive source and the results were used to characterize the light output of each small(20) group of tiles in terms of maximum intensity and attenuation length [117]. Two sources ofraw polystyrene were used for tile fabrication, and during instrumentation the groups of tiles weresorted so that tiles with similar response were inserted in contiguous areas of the detector.3025
Irradiation tests of tile-fiber assemblies indicated that in the first longitudinal sampling for anintegrated dose corresponding to 10 years a light loss of less than 10% is expected. Smaller losseswill occur in the other samplings, where the radiation dose is less.
Prior to insertion into the calorimeter, the tiles are inserted into a TyvekTM(1055B) sleeve. Thissleeve both protects the tile, and improves the scintillation light yield due to its high reflectivity3030
(95%). A mask pattern is printed on the sleeve to improve the optical uniformity. The resultingnon-uniformity over the surface of a tile is generally no worse than 5%.
Wavelength-shifting fibers placed in contact with the tile edges collect the scintillation lightproduced in the scintillators and convert it to a longer wavelength. Each fibre collects light fromone or two tiles and transmits it to the PMTs located inside the girder. The fibres used are 1 mm3035
diameter double clad Kuraray Y11TM(200)MSJ, with an emission peak at 476 nm and decay timeof 6ns. As part of the quality control, light output and attenuation were measured for a sample offibres from each of the 65 preforms used for fiber production [118]. The fibers had a spread inattenuation length of 4% and in relative light output of 3%. To improve light output, the fibres arealuminized at the end opposite to the PMT. The aluminium mirrors were deposited using magnetron3040
sputtering on bundles of 1261 fibres. The reflectivity was required to be 75% and its rms variationwas measured to be 7%, based on measurements from a sample of fibers in each bundle [117].
To facilitate instrumentation, a novel approach was developed to manage the some 540,000fibers required to read out the scintillator tiles and form calorimeter cells. The fibers (4 for thecentral barrel and 3 for the extended barrel calorimeters) are inserted into a plastic channel (much3045
like a straw). The channels are opaque to avoid cross talk, and painted with a high reflectivity inkon the side facing the tiles to maximize the light collection. Slots in the channels allow fibers topass from a location in which they face a tile to a location in which they do not in order to decouple
– 114 –
the fibers at different depths in the calorimeter. The fibres were inserted in these channels using arobot [119], and are glued to the channel at a spot near the aluminized end of the fiber.3050
Figure 71. Average cell response unformity, measured us-ing the cesium calibration system.
5.3.1.4 Instrumentation with scintilla-tor and fibres The light produced inthe scintillating material is collected atthe edges of each tile using two wave-length shifting fibers as indicated in3055
Fig. 69. The fibers are grouped togetherand coupled to photomultipliers (Hama-matsu R7877) (PMT’s) housed at theouter edge of each module. The fibergrouping is used to define a three di-3060
mensional cell structure in such a wayas to form three radial sampling depths(1.4, 4.0 and 1.8 λ thick) composed ofcells having dimensions equal to∆η ×∆φ = 0.1×0.1 (0.2×0.1 in the last lon-3065
gitudinal layer). The depth and rapid-ity segmentation of the barrel (LB) andextended barrel (EB) modules is shownin Fig. 72. The fibers coupled to eachedge of the scintillating tiles are read out3070
by two different PMTs to provide re-dundancy and also allow for the partialequalization of signals produced by par-ticles entering the scintillating tiles at different impact positions.
0.1 0.2 0.3 0.4 0.5 0.6
0.7
0.8
0.9
1.0
D
BC
A
0.8 1.0 1.1 1.20.9
1.3
1.4
1.5
1.7
1.6
Tile Calorimeter
Cells and Tile Rows
Half-Barrel Extended Barrel
0 500 1500 mm
D0 D1 D2 D3
BC1 BC2 BC3 BC4 BC5 BC6 BC7 BC8 B9
A1 A2 A3 A4 A5 A6 A7 A8 A10A9
D5D4 D6
B11 B12 B13 B14 B15
A A13 A14 A15 A16
C
1
1000
0
12
Figure 72. Left: Barrel (LB) module segmentation; Right Extended (EB)module segmentation. The bottomof the picture corresponds to the inner radius of the calorimeter. The detector is symmetrical about theinteraction point at z=0.
– 115 –
Figure 73. glued fiber bundle in girder insertion tube (left) and fiber routing (right)
Module instrumentation [115] comprised many steps, which ranged from cleaning the slots3075
in the absorber structure to a final module quality measurement based on the individual tile fiberresponse. Following insertion of tiles and channel/fiber assemblies into a module, the fibers aregrouped together to form cells and glued as a bundle into the fiber insertion tube as shown inFigure 73. These tubes are then fixed into the girder plastic rings mentioned above, to obtain aprecise match to the position of the photomultipliers. The tubes and fibers are then cut and polished3080
inside the girder to give the optical interface to the PMT. This interface requires that these fibers bephysically present at the time of module instrumentation. However, the gap and crack scintillatorsare only mounted following calorimeter assembly in the cavern. Therefore an optical connector isused to couple the light from their readout fibers to the already glued and polished optical fiberswhich penetrate the girder.3085
Quality control checks are made at several points during instrumentation: the fiber bundlingand routing; fiber gluing, cutting and polishing; tile-fiberoptical coupling when the tile is excited byeither a blue LED or cesium radioactive source. Tile-fiber pairs whose response was more than 25%below the average response of the tile row for that cell were repaired in most cases (typically by re-insertion of the plastic channel or replacement of the fiber). The average cell response uniformity3090
for production modules,measured using the cesium source, is shown in Figure 71, and meets thedesign specification of 10%.
5.3.2 Hadronic end-cap calorimeters (HEC)
5.3.2.1 Overview The hadronic end-cap calorimeter (HEC) [120, 121, 105] is a copper liquid-argon sampling calorimeter in a flat plate design. It covers the pseudo-rapidity range 1.5 < |η | <3095
3.2. The HEC shares each of the two liquid argon endcap cryostats with the electromagnetic endcap(EMEC) and forward (FCal) calorimeters (see Fig. 63). The HEC calorimeter consists of twowheels in each endcap cryostat: a front wheel (HEC1) and a rear wheel (HEC2). The wheelsare cylindrical with an outer radius of 2030 mm. Each of the four HEC wheels is constructedof 32 identical wedge-shaped modules (see Fig. 74). A stainless steel connecting bar system at3100
the at the outer wheel perimeter guaranteed the mechanical integrity of the wheel structure. Alsotwo sliding rails, which supported the wheel inside the cryostat, where part of the wheel definingmechanical structure. At the inner radius small copper connecting bars linked the plates of eachset of neighbouring modules. The geometrical precision of the wheel was given by 32 datum pins
– 116 –
Figure 74. Artist’s impression of a HEC module, with cutaway showing the read-out structure and ‘activepad’ electronics.
on the assembly table. The final vertical deformation of the wheel structure was measured for the3105
four wheels to be 0.3 mm on average. The wheels remained perpendicular to their axes within±1.0 mm.
The modules of the front wheels are made of 24 copper plates, each 25 mm thick, with a 12.5mm front plate, while those of the rear wheels are made of 16 copper plates, each 50 mm thick,with a 25 mm front plate. The gaps in between the plates are 8.5mm in all cases. The resulting3110
sampling fractions for HEC1 and HEC2 are 4.4 % and 2.2 % respectively. The wheels have aninner radius of 372 mm for the first 9 plates of HEC1 and 475 mm for the remaining 16 platesof HEC1 and for all 17 plates of HEC2. The structural strengthof the modules is obtained byseven stainless steel tierods with 12 mm (16 mm) diameter forthe front (rear) modules. Annularhigh precision spacers on the tie rods maintain the 8.5 mm gaps between the copper plates. Three3115
electrodes divide the 8.5 mm gap into four separate LAr driftzones of 1.8 mm width each. TheFig. 75 shows schematically the arrangement of the read-outstructure. The space between theelectrodes is maintained using a honeycomb hexcel sheet, which is also captured by the seven tierods. Each drift zone is individually supplied with high voltage (HV) (see Table 25). The middleelectrode carries a pad structure covered by a high resistivity layer, serving as the read-out electrode3120
and defining the lateral segmentation of the calorimeter. The two other electrodes carry surfaces ofhigh resistivity to which HV is applied. These high voltage planes form an electrostatic transformer(EST). Electrically this structure corresponds to a doublegap of 3.6 mm each. The EST structurehas the advantage of a lower HV for operation, of a double HV safety and of a reduced ion densityproblem at largeη . Figure 76 shows a schematic view of the design of a HEC module. Given the3125
nominal HV of 1 800 V, the typical drift time for electrons in the drift zone is 430 ns.
– 117 –
Figure 75. Schematic of the arrangement of the read-out structure in the 8.5 mm inter-plate gap. Alldimensions are in mm.
Figure 76. Schematicr −φ (left) and r − z (right) view of the hadronic endcap calorimeter. Note thesemi-pointing layout of the read-out electrodes, as indicated by the dashed lines. Units are in mm.
The read-out cells are defined by pads etched on the central foil in each gap. The arrangement
– 118 –
of these pads provides a semi-pointing geometry (see Fig. 76). The size of the read-out cells is∆η ×∆φ = 0.1×2π/64 in the region|η |< 2.5 and 0.2×2π/32 for larger values of pseudorapidity.
The signals from the pads of this electrode structure are amplified and summed employing the3130
concept of active pads [122]: the signals from two consecutive pads are fed into a separate amplifier(based on GaAs electronics). The use of cryogenic GaAs preamplifiers provides the optimumsignal-to-noise ratio for the HEC. An important aspect of the HEC is its ability to detect muons,and to measure any radiative energy loss. The density of the electronics on the HEC wheels withtheir rather modest number of read-out channels (5632 in total) and the low power consumption of3135
the GaAs IC’s (15 mW for one preamplifier channel,∼ 200 mW for the entire chip) is such thatthe heating effect of the electronics on the liquid argon does not produce bubbling. The output ofthe amplifiers is summed on the same GaAs IC to produce one signal from each tower. The signalsent to the feedthrough for each tower is thus comprised of the amplified and summed signals ofthe 8 or 16 cells with the sameη andφ within a read-out depth.3140
All 134 HEC modules (including 3 front and 3 rear spare modules) passed the full cold testprocedure according to the required specifications.
5.3.2.2 Wheel assembly and installation During the wheel assembly, each module had to passa series of final tests: HV hardness, capacitance control, electronic cabling and full signal recon-3145
struction. These tests were repeated after the wheel assembly, after the wheel rotation, after thewheel insertion, after the full cabling of the HEC1 and HEC2 wheels inside the cryostat and finallythrough the feed throughs of the cryostat. Fig. 77 shows a HECwheel fully assembled on the as-sembly table. After closing the endcap cryostat, each endcap has been cooled down, filled with LArand the final cold tests prior to the movement to the ATLAS pit have been performed. For the HEC3150
these tests included TDR measurements and cabling checks ofall signal and calibration lines, fullcalibration and delay scans of all signal channels, full pulse shape analysis of all signal channels,detailed noise measurements of all channels including coherent noise and a long term HV test ofall HV lines. Typically 3 signal channels per endcap were found non operational, correspondingto 0.1 %. The level of HV lines with shorts was∼ 2.5 % in LAr and∼ 0.6 % after emptying the3155
cryostat. Most importantly: with four individual HV lines per individual HEC gap, all HEC regionsremain active.
5.3.3 Forward calorimeters
5.3.3.1 Overview The ATLAS Forward Calorimeters (FCals) are located in the same cryostatsas the EndCap calorimeters and provide coverage of 3.1 < η < 4.9. The close coupling between3160
systems results in hermetic design that minimizes energy loss across the calorimeter systems andlimits the backgrounds that reach the muon system. As the FCal modules are located at highη , ata distance of approximately 4.5 m from the Interaction Point(IP), they are exposed to high particlefluences. This has mandated the need for small liquid argon gaps which have been realized in thedetector by using an electrode structure of small diameter rods centred in tubes that are oriented3165
parallel to the beam direction. The liquid argon gaps chosenare smaller than the usual 2mm gap ofthe electromagnetic barrel calorimeter to avoid ion buildup problems. These smaller gaps also leadto a much faster signal. In FCal1 the triangular current pulse at the electrode has a full drift time of
– 119 –
Figure 77. A HEC wheel fully assembled on the assembly table showing the‘active pad’ electronics.
450 550500400350 600 650
60
50
40
30
20
10
0
R (
cm)
Z (cm)
EMEC
FCal 1 FCal 2 FCal 3
shiel
ding
(EM) (Had) (Had)
(front)HEC HEC
(back)
Poly Shield
Pump
Coppe
r allo
y
Figure 78. The left hand diagram shows the three FCal modules located inthe cryostat. The material infront of the FCal and the shielding plug behind it are shown. The black shaded regions are structural partsof the cryostat. This diagram has a larger vertical scale forclarity. The right hand diagram shows a view ofthe hadronic module absorber matrix including a set of tungsten rods and copper tubes surrounded by 1 cmlong tungsten slugs.
60 ns as opposed to 450 nsec for a more conventional gap size. After 25ns, 65% of the signal hasalready been accumulated on the electrodes. For FCal2 and FCal3 the full drift time scales with the3170
gap size as the field in the gaps is similar for all three modules.
The detector is split into three 45 cm deep modules; one electro-magnetic module (FCal1) andtwo hadronic modules (FCal2 and FCal3), see Fig. 78. To optimize resolution and heat removalFCal1 was made of copper, while mainly tungsten was used in FCal2 and FCal3 to provide con-tainment and minimize the lateral spread of hadronic showers. A shielding plug made of a copper3175
alloy is mounted behind FCal3 to reduce backgrounds in the muon system. A full description ofthe FCal detector can be found in Reference [123].
– 120 –
FCal1 FCal2 FCal3
Function EM HadronicHadronic
Mass of module (kg) 2119 3826 3695
Main absorber material CopperTungstenTungsten
LAr gap width mm 0.269 0.376 0.508
Radiation lengthX0 27.6 91.3 89.2
Absorption lengthλ 2.66 3.68 3.60
Number of electrodes 12260 10200 8224
Number of read out channels 1008 500 254
Table 27. Table of “as built” parameters for the three FCal modules
5.3.3.2 Module description The basic parameters of the FCal modules are listed in Table 27.The FCal1 is made of 18 copper plates stacked one behind the other. The plates have 12,260 holes3180
drilled in them through which the electrode structures are inserted. An electrode consists of acoaxial copper rod and copper tube separated by a precision,radiation resistant PEEK fibre woundaround the rod. The arrangement of electrodes and the effective Molière radius for the modulescan be seen in Fig. 79. Mechanical integrity is achieved by a set of four tie rods that are boltedthrough the structure. The electrode tubes are swaged at thesignal end to provide a good electrical3185
contact. The hadronic modules FCal2 and FCal3 are optimizedfor a high absorption length. Thisis achieved by maximizing the amount of tungsten in the modules. These modules consist oftwo copper endplates, each 2.5 cm thick, that are spanned by electrode structures, similar to theones used in FCal1, except for the use of tungsten rods instead of copper rods. Swaging of thecopper tubes to the endplates is used to provide rigidity forthe overall structure and good electrical3190
contact. The space between the endplates and the tubes is filled with small tungsten slugs as shownin Fig. 78. The inner and outer radii of the absorber structure formed by the rods, tubes and slugsare enclosed with copper shells.
Signals are read out from the interaction point side of FCal1and the non interaction point sideof FCal2 and FCal3. This arrangement keeps the cables and connectors away from the region of3195
maximum radiation damage near the back of FCal1. Readout electrodes are hardwired togetherwith small boards on the faces of the modules in groups of four, six and nine for FCal1, FCal2 andFCal3 respectively. The signals are then routed using miniature kapton co-axial cables along theperiphery of the FCal modules to summing boards that are mounted on the back of the HadronicEndCap (HEC) detector. The summing boards are equipped withtransmission line transformers3200
that sum four inputs. High voltage (see Table 25) is also distributed on the summing boards via aset of current limiting resistors as shown in Fig. 79. The signal summing at the inner and outer radiiof the modules are in general different due to geometric constraints and higher count rates at theinner radius. These differences are discussed in detail in Reference [123]. Calibration test pulsesfor the FCal system are applied at the baseplane of the Front End Crates (FEC) where the signals3205
– 121 –
split into two. One of the split pulses goes directly into thefront end boards, the other goes thoughthe cold electronics chain, reflects off the electrodes and is observed as a delayed pulse. The directpulses are used to calibrate the electronics in the FECs and the delayed pulses are used to examineproblems with the FCal and the cold electronics.
3210
Figure 79. Left hand plot shows the electrode structure of FCal1. The darkened disc has radiusRM, theMolière radius. The right hand plot shows a schematic of the module cabling from the electrodes to thecryogenic feedthrough.
5.3.3.3 Integration into the cryostat There are two sets of FCal modules, one on each side ofthe IP. Each set of modules, along with a shielding plug, are mounted inside a cylindrical tube,known as the FCal support tube which is shown as the black horizontal line between the FCals andthe HEC in Fig. 78. This support tube, along with an upstream conical “nose” and a downstreamflange are the structural components at the inner radius of the EndCap cryostat. To load the modules3215
into the support tube a mandrel was used as a temporary support for the modules. This stage ofthe assembly can be seen in the left hand picture of Fig. 80. This arrangment allowed the readoutcables to be dressed in the cable troughs along the sides of the modules. The support tube was thenslid over the modules, the module’s weights were transferred to the support tube and the cone andflange were attached to the support tube, as can be seen on the right of Fig. 80. The complete unit3220
was then installed into the endcap cryostat at which point the cables were connected to the FCalsumming boards.
5.3.3.4 Alignment survey results There is only a small clearance of the FCal modules withinthe FCal support tube such that the x-y position is fixed relative to the tube. The central axis of thedetectors is about 1 mm below the central axis of the support tube. The distance between modules3225
is constrained by bumpers attached to the endplates of the modules. In this manner the modulescan move in the z direction within the tubes, but will be within about 1 mm of the design location.The survey of the FCal is then a matter of a survey of the FCal support tube. This was carried outafter the installation of the FCal into the cryostat. The central axis of the support tube is within
– 122 –
Figure 80. Assembly of FCal modules. At left can be seen from left to right the three modules plus thecopper alloy plug on the mandrel with most of the cabling in place. At the right the completed endcapassembly can be seen with the bulkhead and cone attached justbefore insertion into the EndCap cryostat.
about 2 mm of the central axis of the cryostat outer warm vessel coordinate system and is surveyed3230
to a precision of about 0.3 mm.
5.3.3.5 Tests during installation/integration QC/QA results A series of quality control checkswere carried out on the modules that included HV testing, capacitance measurements and reflectiontesting. The results of the capacitance measurements provided a measure of the uniformity of theelectrodes. HV checks were carried out at each stage of assembly and integration of the detectors3235
both at warm and at cold. It was found that the number of shortsdid vary during this process. Thereflection tests were used to look for other detector faults such as disconnected channels and brokenground braids. A summary of the faults observed in the FCal modules is shown in table XX. (Tablecomes soon)
5.4 Gap Instrumentation3240
Figure 81. Schematic of the region in the ATLAS calorimeter whereadditional scintillator elements are mounted to provide corrections forenergy lost in dead material such as the liquid argon cryostats and sup-ports.
The energy measurement insome regions of the AT-LAS calorimeter system isadversely affected by thedead material of the crys-3245
tat walls, supports and detec-tor services. Two pairs ofadditional scintillators, readout separately, are mountedto the EB calorimeters, as3250
shown in figure 81, to al-low correction for the en-ergy loss in this dead ma-terial.The first of these (thegap scintillators) covers the3255
– 123 –
rapidity 1.0< |η | <1.2 andprovides correction for theloss of energy in hadronicshowers passing through thecryostat supports and the in-3260
ner detector cables and services. The second of these (the crack scintillators) covers the rapidity1.2< |η | <1.6 contributes to the measurement of primarily electromagnetic showers. In earlyrunnning, 16 of the crack scintillators covering the rapidity region from 1.2< |η | <1.4 are notavailable as their readout is used for the scintillators installed to trigger on minimum bias events.
5.5 Calorimeter readout electronics, calibration and services3265
5.5.1 Electronics readout
5.5.1.1 Overview This section describes the on-detector (front-end) and off-detector (back-end)electronics for the ATLAS calorimetry. The major tasks required of the readout electronics are:
1. to measure, for triggered beam crossings, the energy deposit in each calorimeter cell to betterthan 0.25% at high energy. The dynamic energy range covers a maximum of 3 TeV down to3270
a lower limit of 10 MeV set by thermal noise for the electromagnetic calorimeters. Coherentnoise over the many cells used to measure the jet energy should be kept below 5% of theincoherent noise. The readout should proceed without any dead time up to a trigger rate of75 kHz.
2. to provide the trigger system with the energy deposited intrigger towers of size∆η ×∆φ =3275
0.1×0.1.
The necessity for a large signal dynamic range and for very low level coherent noise combinedwith the high radiation levels expected throughout the cavern (see section 3) favours remote elec-tronics combined with preamplification very near the detector (just above the signal feedthroughsfor the LAr calorimetry and in drawers incorporated into thegirders for the tile calorimeter). As3280
described for the experiment as a whole in section 3, all electronics components situated in thecavern have to be radiation-tolerant to specifications including large safety factors. In particular,the radiation levels in the vicinity of the LAr cryostats andtile drawers (integrated fluences of 1012
neutrons per cm2 per year and ionising doses of 20 Gy per year), have led to the use of specificallydesigned radiation-hard electronics (mainly implementedin DMILL technology) for some of the3285
key components of the different boards of the calorimeters.
As described in more detail below, the back-end electronicssystems for all LAr calorimetersand for the tile calorimeter have been designed with the samearchitecture although with importantdifferences in the functional implementation. More specifically, a common overall electronicsreadout system [124, 125] has been chosen for all LAr calorimeters.3290
As an example, the readout architecture of the LAr systems issketched in Fig. 82.
5.5.1.2 LAr calorimeter front-end electronics The sensitive analogue electronics [124] arehoused on the detector. Inside the cryostat, the calorimeter electrodes are grouped to form readout
– 124 –
Figure 82.Block diagram of the LAr readout electronics. The cold electrical circuit is depicted at the bottomof the figure, followed above by the on-detector Front End electronics crate (FEC), while at the top left areschematized a ROD crate boards and a TTC crate.
cells and small coaxial cables bring the cell signals to the cold-to-warm feedthroughs (see subsec-tion 5.1.2). Front End electronic custom crates are mountedat the feedthroughs. A Front End Crate3295
contains several types of boards:
1. The calibration board [126] injects precisely known current pulses through high precisionresistors located in the cold, to simulate energy deposits in the calorimeters.
2. Front End Boards [127] (FEB) which:• amplify and shape the analog signals8;3300
• sum the calorimeter cells by trigger tower within each layerdepth, and prepare the inputsignals for the tower builder board;• store the signals in an analog memory waiting for the decision by the Level-1 trigger;• digitize the selected pulses and transmit the multiplexed digital results on 70 m long opticalfibres to the Read Out Drivers (RODs) situated in the Level 1 cavern.3305
8The HEC also uses cold preamplifiers.
– 125 –
3. Tower Builder Boards (TBBs), for the EM barrel and endcap calorimeters, perform the finallevel of analog summation to form trigger tower signals and transmit the analog signals to theLevel-1 trigger processor where they are digitized and processed. As well, the board whichbuilds the HEC trigger towers is called the Tower Driver Board (TDB) since no summationis implied here.3310
4. Front End Crate controller boards (FECcont) receive and distribute the 40 MHz clock, theLevel-1 trigger accept signal (L1A), as well as other fast synchronous signals and informa-tion, to configure and control the FEBs and calibration boards.
5. Various boards which monitor the information from the mechanical sensors and stresses ofthe detector, the information from the temperature gauges situated in the liquid argon and the3315
information from the purity monitors.
Figure 83. The figure shows the triangularpulse of the current in a barrel LAr cell andthe FEB output signal after bi-polar shap-ing. Also indicated are the sampling pointsevery 25 ns.
The FEBs perform the first signal processing step.Each FEB receives the signals of 128 LAr detector cells(a total of 1524 FEBs are used to read all LAr calorime-ters). The triangular-shaped pulse of the ionization cur-3320
rent of each detector cell is first amplified. In the caseof the HEC only, the preamplifiers are installed directlyon the detector and not on the FEBs. The preamplifieroutput is sent to shaper chips which amplify the pulse,split it into three gain scales in the ratio 1/9.3/93 and3325
apply a bi-polar shaping function to each scale. Thetime constant, tp = 15 ns, of the FEB shapers is cho-sen to minimize the overall noise level. The two maincontributions are the electronic noise and pile-up noise,which respectively decrease and increase with the peak3330
shaping time. The signal is then sampled at 40 MHz inSwitched Capacitor Array (SCA) chips which store thesamples in analog form during the Level-1 trigger la-tency (< 2.5 µs). Figure 83 shows the triangular pulsecurrent in the LAr cell as well as the FEB output signal3335
after bi-polar shaping and sampling.
When the Level-1 trigger decision arrives, the optimal gainscale is selected on an event-by-event basis. The SCA samples are then digitized at a 5 MHz ratein a 12-bit Analog-To-DigitalConverter which, together with the gain selection procedure, fulfills the required 16-bit dynamicrange over the whole energy interval. The FEB data are finallysent via the 1.6 Gbit/s G-link optical3340
links to the RODs.
5.5.1.3 Tile calorimeter front-end electronics The Front End electronics of the Tile Calorime-ter are housed inside the so-called Tilecal drawers locatedinside the Girders which constitute theexternal support frame of the Tilecal Modules [112]. on the plastic rings into which the bundles3345
– 126 –
of optical fibers are glued during module instrumentation. The drawers hold PMT Blocks (whichcontain photomultipliers and their associated electronics), their readout and trigger electronics;electronics to regulate the high voltage for each channel plus the input/output links and communi-cation with the external world. A schematic of the tile calorimeter front end electronics and readoutcomponents inside drawer is shown in Figure 84.3350
Figure 84. Tile Calorimeter Front End Electronics Block Diagram
A key element in the readout is the PMT block [128]. It is a mechanical structure comprisinga steel cylinder and mu-metal shield for magnetic shielding, which contains a light mixer, photo-multiplier tube, voltage divider and the so-called 3-in-1 card. The light mixer is an optical plasticinsert which mixes the light from the readout fibers to insureuniform illumination of the photo-cathode. Hamamatsu R7877 PMT’s, with a compact 8-dynode structure, are used to measure the3355
scintillation light [129]. The PMTs were burned in and tested for linearity, stability, dark currentand operating voltage for nominal gain of 105 [115]. The average operating voltage for nominalgain is 680V. The assembled PMT Blocks are inserted inside precision slots inside the aluminiumstructure of the drawers, which ensure accurate placement of the light mixer relative to the fiberbundle for each readout cell.3360
Most analog functions of the front-end electronics are contained on the 3-in-1 card [130]. Thisprovides signal shaping to yield a pulsewidth of 50ns and twolinear outputs with a relative gain of64 to achieve the overall 16-bit dynamic range with full scale on the low gain channel being 800 pC.Charge-averaging amplifiers on the 3-in-1 card are used for cell calibration and monitoring, and acharge injection system is incorporated to calibrate the readout over its full dynamic range. The3365
3-in-1 card also produces the analog signal for the Level 1 Trigger.
The digitiser system [131] samples the incoming data from the 3-in-1 cards every 25ns using2 dedicated 10 bit ADCs (one each for high and low gain channels) and stores the information in a
– 127 –
fixed length pipeline waiting for a level 1 trigger accept. The pipeline latency is 2.5µs as requiredby the ATLAS specification. Each triggered event is recordedover an extended time frame, with3370
a programmable length up to 16 samples. In normal data-acquisition mode, following a first levelaccept, 7 samples are kept: 1 close to the peak, 4 before (withthe two first samples giving accessto the signal pedestal) and 2 after the peak.
There is one Interface card [132] per pair of drawers (Super-Drawer.) It receives and distributesthe Timing Trigger and Control (TTC) signal in the drawer, collects data from the digitiser cards,3375
aligns it into event frames and transmits the digitised datavia optical link. The interface board alsoperforms cyclic redundancy checks on the input and ouput datastreams.
The high voltage for each PMT is regulated with an accuracy better than 0.5V by micro-controllers inside the drawer. A single micro-controller services 24 PMT channels and has a mem-ory into which the nominal and set values of the voltage for each PMT are loaded. Each HV3380
Distributor system drives 7 temperature probes located at key locations inside each drawer. SlowControl of this system is via a CANbusTM network configured in a daisy chain of 16 Super-Drawersconnected at the level of the micro-controller.
A system of four linked Mother Boards form the base of the readout system. They distributepower to the 3-in-1 cards, ADC Integrator, Mezzanine card, and the Adders. Each Mother Board3385
also contains circuitry that generates a precise referencevoltage used by the 3-in-1 cards for chargeinjection and integrator calibration. The Mezzanine card is mounted on the first Mother Board inthe drawer. It decodes control commands send via the LHC TTC B-channel and sends them ondifferential serial lines to the 3-in-1 cards. These commands control the charge injection, integratorgain setting and calibration, trigger output gating, and integrator bus access for each 3-in-1 card.3390
The Mezzanine card can also receive commands via the ADC Integrator card, which is mountedon the first motherboard. It sends and receives messages via adedicated CANbus daisy chain. TheADC card microprocessor cues commands to the Mezzanine card. It sets the integrator gain on a3-in-1 card, multiplexes the 3-in-1 card onto the integrator readout bus and reads out the chargeinduced by the Cesium source, minimum bias events, or the on board calibration DAC. The ADC3395
integrator can also read back 3-in-1 card status and controlinformation.
Adder cards [133] are distributed along the drawer mounted on the Mother Boards. Each adderreceives the analog trigger outputs from up to six 3-in-1 cards. The inputs are connected to formtrigger towers. The adders perform an analog sum of the inputsignals and send two output signalsvia long cables to the Level 1 trigger system. One signal (tower signal) comprises the sum of all3400
samples, the second (muon output) signal contains only the last Tilecal sampling. In the case of thegap and crack scintillators, the adder board provides additional amplification to match better matchthe small signals from these scintillators to the requirements of the L1 trigger boards. This adderboard provides the sum of the signal in all four of the gap and crack scintillators. The muon outputfrom this board provides the signal from the scintillator covering the region 1.2<η<1.4. In early3405
running, for 16 modules, this latter channel is used for the trigger from the minimum bias triggerscintillators.
5.5.1.4 Back-end (BE) electronics
Overview The BE system [125] for all calorimeters is located in a cavern at 70 m away
– 128 –
from the detector. It contains three different sub-systems: the RODs which constitute the core of3410
the BE, the TTC and the Level-1 receiver. The BE system reads the FE electronics, receives theTTC signals, processes the data and sends it to the acquisition system at a Level-1 rate of up to75 kHz for the physics mode and at a 10 kHz rate for the calibration mode. The BE system alsodistributes the timing clock and trigger to the FE electronics and the RODs, and also configures andcontrols the FECs. All of these tasks must be performed fast enough to cope with the above trigger3415
rates.
Four VME crates are used to implement the timing trigger and control (TTC) system, whereup to two partitions are implemented in a single crate. Each TTC partition contains essentially acontroller and a local trigger processor.
The trigger latency, which is the delay between the bunch crossing time and the time when3420
the L1A signal arrives to the FE electronics, has been minimized to a value below 2.5µs. TheTTC rack location has been chosen to minimize the length of the TTC fibres to the FECs and theassociated contribution to the trigger latency. In addition, the programmable delay lines of thecalibration boards are preset to reproduce the timing of signals generated by particles originatingfrom the interaction point.3425
The receiver system is part of the trigger sum chain and interfaces the TBBs/TDBs to thecalorimeter Level-1 processor. One function of the triggersum chain is to convert the signal fromenergy to transverse energy. The final gain adjustment is left to the receiver. Each receiver cratecontains 16 receiver modules, two monitoring modules, and one controller module. The LAr sys-tem consists of 6 receiver crates: 2 for the EMB, 2 for the EMECs and 2 for the FCal and HEC.3430
Two additional crates are for the Tile calorimeters.
LAr back-end electronics The optimal filtering (OF) method [125] is at the core of the BEprocessing role. The OF algorithm, implemented in the ROD Digital Signal Processors (DSPs [125])calculates the energy for each cell while minimizing the noise contribution. For cells with an en-ergy above a certain threshold (approximately 5% to 10% of all cells), the precise timing of the3435
signal as well as aχ2 like data quality factor, allowing to flag cells with large pile-up contribu-tion, are determined and transmitted as well. For an even lower fraction, the samplessi are eventransmitted. Complementary to this functionality, the RODcreates and fills online histograms tobe readout through VME and used by the online monitoring of ATLAS.
The signal pulse is sampled every 25 ns. For physics, 5 samples are typically used, whereasfor calibration, up to 32 samples can be taken. The energyE , timeτ and simplifiedχ2 of the signalpulse are calculated in a weighted sum of the sample amplitudes,si :
E =n
∑i=1
ai(si − ped) τ =1E
n
∑i=1
bi(si − ped) χ2 =n
∑i=1
(
si − ped−E(gi − τg′i))2
,
wheren is the number of samples andped is the pedestal value of the corresponding readout3440
channel. The optimal filtering weights,ai andbi , are evaluated while minimizing the dispersion inE arising from electronics and pile-up noise, taking into account the noise autocorrelation matrix.gi
andg′i are respectively the normalized pulse shape and its derivative. These are estimated from themeasured calibration pulse and from the electrical model parameters of the readout circuit [125].
– 129 –
For each trigger, data from the TTC stream and from the FE electronics are pushed into the3445
ROD modules where they are checked, processed, formatted and sent on a Read Out Link (ROL)to the ATLAS DAQ.
To increase modularity and allow for concurrent running of the various parts, the LAr calorime-try is split into 6 partitions: one for each half-barrel (EMB-A and EMB-C), one for each endcap(EMEC-A and EMEC-C), one for the A side hadronic endcap and forward calorimeter (HEC-A3450
+ FCal-A) and one for the C side (HEC-C + FCal-C). Each partition is associated to a PartitionMaster computer, which controls and monitors the system andto a TTC sub-system. A typicalbarrel partition (see [125] for the other partitions composition) is composed of 4 ROD crates andthe associated FE electronics. Each ROD crate contains 14 ROD boards and drives and controls 4FECs.3455
Tile calorimeter back-end electronics The main elements of the TileCal back-end electron-ics are the readout driver (ROD) system [125, 134] and the optical multiplexer board (OMB) [135].The OMB has the responsibility of checking data coming from the FE electronics. There are twofibers per FE drawer carrying the same data and the OMB has to select the error free one in realtime by checking CRC information of the data received. It canalso act as data injector for the ROD3460
system.
The Tile ROD system is divided in four independent partitions. Each partition is composedby a 6U VME trigger crate and a 9U VME ROD crate and reads up to 64front-end drawers. ARODmb is a custom 9U VME board totally controlled and configured through VME. There areeight optical inputs in each RODmb, each input receives one fiber coming from one front-end3465
drawer of the TileCal detector. Eight G-Links, HDMP1024, deserialize the data and send it to fourstaging FPGAs. These FPGAs drive the data to four processingunits (PUs) that perform onlinealgorithms to reconstruct the energy deposited in the calorimeter in order to send this informationto the second level trigger. Four output controller FPGAs receive the processed data from the PUsand send it to the transition module through the standard VMEP2 and custom P3 backplane.3470
The ROD PUs are designed to perform quality checks of the data, processing tasks and onlinemonitoring. Each PU contains two input FPGAs that verify thequality of the data with a CyclicRedundancy Check. The PUs are also equipped with two digitalsignal processors (DSPs), whichsynchronize the data comparing the trigger information coming from the front-end electronics withthe same information distributed by the TBM through the P3 backplane. At the same time, the3475
DSP performs online algorithms to reconstruct energy deposited in the calorimeter, signal timingand quality factor of the reconstruction.
A digital filter, termed an Optimal Filter (OF) [136], is the choice for reconstructing the energydeposited in the calorimeter and the timing of the signal. OFreconstructs the amplitude and thetiming of the digitized shaped signal by means of a weighted sum of its samples, expressed as:3480
A =n
∑i=1
aisi , τ =1A
n
∑i=1
bisi , (5.1)
wheren is the number of samples andsi is the sample taken at timeti . The amplitude,A, is thedistance between the maximum and the signal baseline. The phase,τ , is the time between thecentral sample and the peak of the signal. Thea andb coefficients are calculated using the pulse
– 130 –
Figure 85. Comparison of the use of the Optimal Filter charge reconstruction with that obtained by a flatfilter in charge injection calibration events.
Figure 86. Injection scheme of the calibration signal in the electromagnetic LAr calorimeters.
shape (at the input of the digitizer) and the noise autocorrelation matrix. A Lagrange multipliermethod is used to calculate the weights minimizing the noisecontribution to the variance of the3485
amplitude and phase.
Figure 85 shows the performance of the Optimal Filter in the reconstruction of charge in chargeinjection events. A clear improvement is observed for low signals by comparison to a Flat Filterwhich weights all samples identically. An even greater improvement is expected in pp collisions inthe presence of pileup.3490
5.5.2 Calorimeter calibration systems
The LAr and tile calorimeters have very precise charge-injection systems and calibration boardsto calibrate the response of the front-end electronics boards to a very high accuracy. The tilecalorimeter is equipped with two additional calibration systems to monitor the behaviour with timeof the photomultipliers and optical components, as described below.3495
5.5.2.1 LAr calibration The aim of the electronics calibration is to determine the conversionfactor, measured signal in ADC counts to signal inµA, for each individual channel and to monitorthis factor with time. The conversion of the signal inµA to a signal in GeV can be derived from firstprinciples [137]. The measurement of the conversion factorhas a direct impact on the performancesof the uniformity response of the calorimeter, therefore onits global constant term. Furthermore3500
several calibration procedures are used to determine the OF’s coefficients (see subsection 5.5.1.4).
– 131 –
The use of fast shaping for the ATLAS calorimeter readout implied to distribute the calibrationsignal on injection resistors directly placed at the input of the detector cell and not at the input of thepreamplifier located outside the cryostat at warm as shown onFig. 86 for EMB and EMEC, wherethe best accuracy is needed. The calibration pulse is injected in front of the HEC cold preamplifiers.3505
However, calibration test pulses for the FCal system are applied at the baseplane of the FrontEnd Crates (FEC) where the signals split into two. One of the split pulses goes directly into thefront end boards, the other goes though the cold electronicschain, reflects off the electrodes and isobserved as a delayed pulse. The direct pulses are used to calibrate the electronics in the FECs andthe delayed pulses are used to examine problems with the FCaland the cold electronics.3510
The calibration boards [126] are built to deliver a uniform,stable, linear and precise voltagedriven signal whose amplitude and shape are as close as possible to the calorimeter ionisationcurrent signal over a dynamic range from 0 to 200 mA. The triangular physics pulse is in factapproximated by an exponential shape. The chosen injectionscheme, typically one channel everyfour signal channels, allows also the measurement of the crosstalk. Calibration boards contain3515
programmable delays which are used to adjust the timing between the different front end boards.Apart from the pedestals and gain measurements, the pulse shape versus time is determined thanksto the help of the precise delays of TTCRx chips.
In total 132 boards have been produced. They fulfill the expected performances of an integrallinearity better than 0.1%, an uniformity response better than 0.2%, and a stability with time better3520
than 0.1%, performances necessary to achieve the needed accuracy for physics.
5.5.2.2 Tile calibration Each TileCal cell can be divided into three sections for calibration pur-poses; the optical part consisting of scintillator and fibers, the phototubes, and the front-end elec-tronics which shape and digitize the light signals. A calibration and monitoring system is used tocertify each of these parts independently [138].3525
The charge injection system is designed to calibrate the front end shaping and digitizing cir-cuits to an accuracy of 1%. The charge injection system utilizes a precision DAC on each of the3-in-1 cards in conjunction with solid state switches and precision capacitors. Dedicated chargeinjection runs are taken scanning the full range of both gains. Fits to the pulse convert ADC countsto pC, and yield a typical non-linearity of about 2 ADC countsover the full dynamic range, with a3530
stability of better than 1% over many months of monitoring.
The laser system is used to calibrate and monitor the phototube response with an accuracyof about±0.5%. A Nd:YVO4 laser synchronised to the TTC clock generates 10nsec pulses at awavelength of 532 nm. The laser light is split in USA15 and sent via clear plastic fibers to eachcalorimeter finger where it is again split to each photocathode. The pulse-to-pulse variation is3535
monitored at the laser by several pin diodes and, by means of afilter wheel, the laser light intensitycan span a dynamic range of 10,000. Global phototube non-linearity is found to be<0.5% above80pC. IN addition, the system can be used to set the global calorimeter timing and to investigatesaturation recovery techniques. To calibrate and monitor the TileCal scintillator and optical systema∼ 10 mCi 137Csγ-source is moved hydraulically inside the calorimeter bodyin dedicated runs,3540
with the source traversing each of the 463,000 tiles in the detector [139]. The phototube currentis integrated by the 3-in-1 cards as the source passes withinthe respective cell. From the digitizedcurrent the response of each scintillator tile can be extracted. The accuracy of a single tile response
– 132 –
is better than±2% with the mean response accuracy for a calorimeter cell is±0.3%. Based onmonitoring a single module over a period of 4 months, the long-term stability of the system was3545
determined to be 1%. A procedure to set the gains of each phototube based on the integrated cellsource current allows intercalibration of cells and modules to better than±3% [138].
5.5.3 Calorimeter power supplies and services
This section describes briefly the LAr and tile calorimeter power supplies, cooling systems andcontrols.3550
5.5.3.1 Low-voltage power supplies The low-voltage power supplies for the LAr and tile calorime-ters are located in the cavern as close as feasible to the on-detector electronics boards. To this end,they had to be custom built and validated to operate in a high radiation environment and in a sig-nificant residual magnetic field.
LAr low-voltage power supplies Each front-end crate (FEC) is powered by a low voltage3555
power supply (LVPS) located close to it, in a so-called fingergap and delivering the seven DCvoltages necessary to the FEC. This location has imposed several constraints : limited volume,limited access, high radiation environment and significantresidual magnetic field. There is a totalof 58 LVPS’s, 2×16 for the EMB and 2×13 for the end-cap calorimeters. Each LVPS is poweredby a 280V DC power supply located in the underground countingroom. The LVPS’s and associated3560
280V power supplies are controlled and monitored through CANbusTM using a custom interfaceboard (ELMB).
Tile low-voltage power supplies Low voltage power to the tile drawer electronics is sup-plied by a custom power supply located in an extension to the module (finger) [140]. These arelocated on either side of the central barrel and on the outer surface of each extended barrel at the3565
support girder and must operate in stray magnetic fields up toabout 1000 Gauss and be radiationtolerant. Each power supply consists of 8 isolated switch DC-to-DC convertors (bricks) which arepowered by a custom external 200V bulk power supply. The bricks themselves are controlled andmonitored at two levels through CANbusTM using a custom interface board (ELMB): inside thedrawer using current loops and voltage feedback loops, and remotely in a low area area from which3570
a reset can be sent in the case of brick failure due to Single Event Upset for example.
5.5.3.2 High-voltage power supplies High voltage is delivered to the LAr and tile calorimetersthrough industrial or custom-built external power supplies located in the underground countingroom.
LAr high-voltage power supplies Several different versions of the same industrial power3575
supply are used to power the different LAr detectors, according to the voltage and intensity needed.The maximal voltage and the maximal intensity are given in Table 28.
Furthermore, the EMB and EMEC channels showing a problem, like a resistive short, arepowered with special power supplies (2000 V and 3 mA). The power supplies modules contain twoboards with 8 or 16 channels. They are remotely controlled via a CANbusTM and a set of 6 PCs. In3580
particular, the current value at which the HV is automatically switched off is remotely controlled.In total there are 20 crates of 8 modules for the LAr system.
– 133 –
Electromagnetic barrel (EMB) 2500 75µA
Barrel preshower 2500V 75µA
Electromagnetic endcap (EMEC)1500-2500200µA
900-2000 200µA
Endcap preshower -2500 200µA
Hadronic endcap (HEC) 2500 75µA
Forward calorimeter (FCal) 600 6 mA
Purity monitors ±2500 75µA
Table 28.Maximal voltage (or range) in Volts and maximal intensity for the LAr calorimeters HV supplies.
Tile high-voltage power supplies High voltage is delivered to the tile drawers by a customand external bulk power supply. The supply delivers two voltages to the input of the drawer (-830Vand -950V) to match the PMT operating requirements and provides a maximum current of 20mA.3585
A single bulk supply provides high voltage to 16 drawers.
5.5.3.3 Other services The cryogenics services for the LAr calorimeters are described in somedetail in section 8.4.5 and shown in Figure 145.
The cooling systems for the LAr and tile calorimeters are both based on demineralised wa-ter supplied to a leakless system as described below for the on-detector implementation and in3590
section 8.4.3 for the off-detector systems.
LAr calorimeter cooling system Each FEB channel has a consumption of∼ 0.7 W. Thetotal consumption of LAr front-end readout electronics exceeds 200 kW. Therefore each electronicboard in a front end crate is conductively cooled using two aluminum plates, put on each side of theboard. These aluminum plates are implemented with a leakless cooling system. The temperature3595
of a FEB is maintained constant at±1C. Under these conditions no temperature dependence isobserved on the pedestals, the gains and the delays of the readout electronics. The LVPS’s are alsocooled.
Tile calorimeter cooling system A heat dissipation of 300 W is expected from 1 super-drawer [141] and a leakless cooling system [142] has been designed to provide cooling inside the3600
drawers. The cooling system supplies demineralized water at 18C to the modules and works withsub-atmospheric pressure to prevent the cooling liquid from leaking out of the system should abreak or holes occur in a cooling pipe.
The performance requirements of this system were determined with a prototype system [141]which was used during the calibration of production modulesin a particle beam. By varying theinput temperature and flow rate it was possible to change the temperature inside the drawer whilemeasuring the response of the calorimeter to a high energy pion beam. The energy variation as a
– 134 –
function of the temperature of the PMT block was found to be:
∆EE
= 0.2%/CPMT (5.2)
and that the variation in response of the PMT gain as a function of the change in temperature ofthe cooling water is 0.15%/Ccooling. To ensure a PMT gain stability of±0.5%, the temperature3605
variations of the PMT block must be smaller than±2.5C. Long terms tests carried out during cal-ibration of production production modules showed that the system could maintain the temperaturein a drawer to within±0.3C.
5.5.3.4 Detector control systems Within the context of the overall ATLAS detector safety (DSS)and control (DCS) systems described in section 7.5, this section describes the main operating pa-3610
rameters monitored by the specific DCS implementations of the different calorimeter systems.
LAr DCS The LAr DCS controls and monitors two important issues: liquid argon tempera-ture and purity.
• Temperature measurementsThe measured sensitivity [137] of the signal to the temperature is−2% per degree. There-3615
fore numerous calibrated (∆T = 10mK) temperature probes (PT100 platinum resistors) havebeen installed in each calorimeter to measure precisely theliquid argon temperature. A totalof 192 temperature probes are glued on the absorbers of the EMB at two radii (inner andouter) regularly distributed inz andφ (6×32), whereas 32 (18) temperature probes are lo-cated at the front and back side of the external (internal) EMEC wheel close to their exterior3620
radius. They are also uniformly distributed inφ. A total of 192 HEC temperature probesfixed on copper blocks at three radii (inner, middle and outer) located at the HEC wheels rearsurface and regularly distributed inφ. Finally the three FCal modules are equipped with atotal of 14 temperature probes, 4 at the outer radius and 10 atthe inner radius where heatingfrom beam interactions and heat transfer through the cryostat wall is highest. These sets of3625
temperature probes are completed by other sets, used to control and monitor the temperatureduring the cooling or heating of the cryostats.
• Purity monitorsTo control the argon purity, 10 purity monitors have been installed on the outer radius of the3630
barrel calorimeter between the support rings. They are located in the median azimuthal planeand at the top and the bottom of the half barrels. Also 10 purity monitors are installed in eachendcap cryostat, 2 in the EMEC region, 6 in the HEC region and 2in the FCal region.
The impurity measurement is based on the energy deposition by radioactive sources inthe liquid argon. The ionisation charge is collected by an electric field and measured by3635
a cold pre-amplifier. Each device consists of two radioactive sources: an241Am sourceemits 5.5 MeVα -particles and a207Bi source 1 MeV conversion electrons. The ratio of themeasured chargesQBi/QAm is used to extract the absolute oxygen content in the liquid ar-gon [110].
– 135 –
The other quantities controlled and monitored by the LAr DCSconcerns mainly the readout elec-3640
tronics and are summarized below:
1. on the detectors, DCS controls also the LVDT-type position monitors, the voltage, the inten-sity and the temperature of HEC preamp PS;
2. on the HV system, DCS controls the voltage, the intensity and the preset value for switchingoff of each HV channel. The temperature of each HV module is also monitored.3645
3. on the front-end electronics, DCS controls the voltage, the intensity and the temperatre ofthe 280V PS powering FEC LVPS, of the LVPS FEC crates (the temperature of each FEB ismonitored by DAQ);
4. on the back end electronics, the voltage, the intensity and the temperature of each crate ismonitored.3650
Tile DCS The Tilecal main Distributed Control Systems (DCS) controland monitor the on-detector Low Voltage (LV), High Voltage (HV) and electronics cooling as well as the off-detectorhigh voltage power supply and electronics racks. The systemoperates within a Finite State Machineframework for control, monitoring and error handling implemented in a hierarchical structure.
The communication with the bulk power supply crates is done using RS422. CANbusTM is3655
used for the communication between the micro-controller cards located inside the drawers and thePC equipped with Kvaser PCIcan cards. A client-server system is being developed for the controland monitoring of the high voltage. Control and monitoring of the on-detector low voltage powersupplies is implemented via a custom interface board, whichmultiplexes currents and temperaturesfrom sensors inside the drawers. The 200V DC power supplies are controlled and monitored using3660
Modbus.A total of more than 25000 parameters are monitored in this system. Configuration data such
as system structure (lists and hierarchies of devices), device properties (configuration of archiving,smoothing, etc.) and settings (output values, alert limits) are stored in the configuration database.The data produced by PVSS is stored in the DCS ORACLETM archive. Due to the large amount of3665
data monitored by DCS, smoothing has to be applied to reduce the amount of data stored. A subsetof this data (for example PMT high voltages) is also available to offline reconstruction.
5.6 Beam test results
5.6.1 Electromagnetic module performance
Four barrel (out of 32) and three endcap (out of 16) series modules have been subjected to test-3670
beam, with a setup including a dedicated cryostat and movingtable allowing detailed positionscans. These tests have been carried out on the CERN’s SPS H8 and H6 beam lines, using electronand positron beams with various energies. A comprehensive analysis of module performance hasbeen finalised [143, 144, 137, 145].
To reconstruct the energy of the electron, the following steps are made. For example, for theEMB, an electron cluster is first constructed from the secondaccordion compartment, where cellswithin a square window of 3× 3 cells around the cell with the highest energy are merged. For
– 136 –
the other accordion compartments all cells intersecting the geometrical projection of this squarewindow are included. Thus a cluster includes 17× 2 cells in the first, 3× 3 in the second and3× 2 in the third compartment. In the EMB preshower 3× 2 cells are used. The energy in eachcompartment, E0 for the preshower, E1, E2, E3 for the front, middle and back compartments ofEMB is the sum of the energy of the selected cells in the cluster. Then the energy of the electron isgiven by the algorithm :
E = offset + W0×E0 + W01×√
E0E1 + λ (E1 +E2+E3) +W3×E3
where the coefficients, offset and W0, corrects for the energy lost upstream of the preshower, W013675
corrects for the energy deposited between the presampler and the first calorimeter compartment,λ corrects for the lateral leakage outside of the electron cluster and W3 corrects for the energydeposited downstream of the calorimeter. These coefficients depend on the electron energy and theη value and have been determined by simulating the test-beam setup using GEANT4. A similarscheme is used to reconstruct the electron energy in EMEC.3680
For the electromagnetic barrel, the linearity of the response versus the energy and the energyresolution have been studied in the range from 10 to 245 GeV, at η = 0.687. The experimentalmeasurements, after noise subtraction, have been fitted with the expression [144]:
σE
E=
a√
E[GeV]⊕b
wherea is the sampling term, andb the constant term reflecting local non-uniformities in theresponse of the calorimeter. A stochastic term of 10% and a constant term of 0.17% have beenobtained. In the energy range 15≤ E ≤ 180 GeV, the reconstructed energy response is linearwithin ±0.1% (Fig. 87). These results are in agreement with dedicated Monte-Carlo simulations ofthe test-beam setup and meet the ATLAS design. Similar results have been obtained for the EMEC.3685
The response uniformity as a function ofη has been measured using an electron beam of 245GeV for the barrel and of 119 GeV for the endcap. The ATLAS goalis to achieve a constant term of0.7% or smaller over the full calorimeter acceptance. Non uniformities of the response on the testedmodules9 [137] do not exceed 0.7% and can reach 0.5% in the case of barrel modules (Fig 88).3690
The overall constant term in the energy resolution, using the above formula, range between 0.5%and 0.7% and meet well the initial calorimeter design performance.
At low energy, the calorimeter response to minimum ionizingparticles has allowed us a de-tailed exploration of the material structure of the detector. Since muons are non interacting parti-cles, their energy deposit is much more localized than the electron energy deposits. Typically, they3695
only cross one middle cell inη and 1 or 2 inφ. They can be used to study the fine structure of thecalorimeter, without having to deconvolute the effects of the showering process. The signal overnoise ratio, evaluated as the ratio of the most probable energy deposit divided by the r.m.s of thepedestal, goes from 7 to 12 for middle barrel cells [145] and from 6 to 7 for middle endcap cells.
The performance of the electromagnetic calorimeter with respect to its finely segmented first3700
sampling has been studied by using electron, photon and pionbeams. The position resolution
9one barrel module was excluded from this analysis due to the presence of unexpected material in the beam line.
– 137 –
[GeV]beamE0 50 100 150 200
beam
/Ere
cE
0.99
0.992
0.994
0.996
0.998
1
1.002
1.004
1.006
1.0‰±
Databeam energy uncertainty uncorrelatedbeam energy uncertainty correlated
[GeV]beamE0 100 200
/E
Eσ
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05Sampl. Const. Term
] [%](GeV)[% * 0.04±0.1 0.17±10.1
DataData noise subtractedNoise
Figure 87. The left figure shows the ratio of the reconstructed energy tothe beam energy as function of thebeam energy. All points are normalised to the value measuredat E = 100 GeV. The inner band illustratesthe uncorrelated uncertainty of the beam energy measurement; in the outer band the correlated uncertainty isadded in quadrature to the inner band. The right figure shows the fractional energy resolution as a functionof the beam energy. Shown are the data before (closed circles) and after (open circles) the gain dependentnoise subtraction. Overlayed as a line is the result of the fitof the resolution. The open squares indicate thesubtracted noise contribution.
along η was measured to be about 1.5× 10−4 and 3.3× 10−4 (in units of pseudorapidity) forfront and middle compartments, allowing to achieve a polar angle resolution in the range 50-60(mrad)/
√
E(GeV). Theπ0 rejection was found to be 3.54±0.12stat at pT = 50GeV/c. Theseresults meet the ATLAS requirements.3705
Finally in 2004, a full slice of the central ATLAS detectors,from the inner detector to themuon chambers, has been exposed to beams of electrons, photons, pions and muons and protonsin the energy range 1 GeV≤ E ≤ 350 GeV at the CERN H8 beam line. A spare EMB module,identical to the series modules, was part of the full slice. The amount of material in front of theEMB module was very close to the material in ATLAS and therefore it has been checked that the3710
linearity, energy resolution and uniformity performances[146] are not degraded as can be seen inFig. 88 and Fig. 89. A study [147] of the response of the EMB module to low energy electronsbetween 1 and 9 GeV has been done.(More to come before end july.
5.6.2 Hadronic endcap performance3715
About 25 % of the series production modules were exposed to beams of electrons, pions and muonswith energies up to 200 GeV [148]. Two “partial HEC wheels” consisting of three HEC1 and threeHEC2 modules were used in a standard setup. The goal was not only to prove the uniformity of theproduction modules as defined by the hardware tolerances, but also to study the performance andcalibration as obtained from pions, electrons and muons. The analysis of the data [148] taken with
– 138 –
0.96
0.98
1
1.02
1.04
0.5 1 1.5 20
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
η
Nor
m. A
v. E
nerg
ies
+2 %
+1 %
-1 %
-2 %All Barrel Modules 0.43% All EMEC 0.62%
Overall Barrel and EMEC 0.54%(882 Cells) (2455 Cells)
OV
ER
LAP
η 0 0.1 0.2 0.3 0.4 0.5 0.6
<E
> (
Ge
V)
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
(23 Cells)
M0 module 0.46%
2 4 6 8 10 12 14
Figure 88. The left figure shows a two dimensional histogram of the average energies measured in allcells of all tested modules normalised to the mean energy of the modules. In the barrel the energie was∼ 245 GeV while it was∼ 120 GeV in the EMEC. The distributions are normalised to the number of middlecells scanned inφ for eachη value.The right figure shows the uniformity obtained in the combined test runat an energy of∼ 180 GeV
.
Ebeam (GeV)0 50 100 150 200 250
<E
-Eb
ea
m/E
be
am
>
-0.02
-0.015
-0.01
-0.005
0
0.005
0.01
0.015
0.02
Ebeam (GeV)0 50 100 150 200 250 300
/<E
> (%
)Eσ
0.01
0.02
0.03
0.04
0.05 / ndf 2χ 78.48 / 3p0 0.0014± 0.1059 p1 0.000448± 0.004371
MC
Data
Figure 89. The figures shows respectively the linearity and the resolution obtained in the combined testbeam.
electrons gives an energy resolution:
σ(E)
E=
a√
E[GeV]⊕b (5.3)
with a sampling constanta = (21.4±0.1) %√
GeV and a constant termb compatible with zero,in very good agreement with Monte Carlo (MC) simulations. Horizontal and vertical scans withbeams across the surface of the calorimeter showed a homogeneity of the electron signal of±1%without corrections. The irradiation with pions is of particular importance for the prediction ofthe final performance of the calorimeter for jets. Using the parameterization mentioned above, a3720
sampling constant of(70.6± 1.5) %√
GeV and a constant term of(5.8± 0.2)% were obtained.The data have been compared in detail to MC simulations and the results are published in [149].
In 2002 a new phase of combined beam tests was started: the calibration of full φ wedges ofthe three ATLAS liquid argon endcap calorimeters as specified already in [121]. The main purpose
– 139 –
(GeV)beamE0 50 100 150 200
(GeV)beamE0 50 100 150 200
, noi
se s
ubtr
acte
dre
co)
/ Ere
co(Eσ
10-1
-πTB+πTB
Geant3LHEPQGSP
Figure 90. Energy dependence of the energy resolution forπ− andπ+ data taken during the first combinedbeam test in comparison to different MC predictions. The analysis employs the cluster weighting approach.The lines represent the result of the related fits.
is to define calibration procedures and constants for initial operation at ATLAS. The first combined3725
beam test carried out in 2002 was devoted to the region 1.6 < |η | < 1.8 [150, 151]. First steps ofthe ATLAS hadronic calibration strategy have been tested. A3D clustering algorithm and signalweighting approach, as used already in previous experiments, have been tested and the first resultsyield a very good pion resolution. Fig. 90 shows the energy dependence of the energy resolution ascompared to various MC predictions. Fits to the data with formula (5.3) yields sampling constants3730
of (84.6± 0.3)%√
GeV and(81.7± 0.4) %√
GeV for π− and π+ respectively, and constantterms of zero within errors. The vanishing of the constant terms (after correcting for leakage)shows the effectiveness of the energy weighting approach inachieving a good compensation. TheGEANT 4 simulations with different hadronic physics lists are closer to the data than the GEANT3 predictions, but neither yield an optimal description of the data.3735
The second combined beam test was carried out in 2004 in the region around|η |= 3.2. This isa complex region of overlap of the three endcap calorimeters. The EMEC, HEC and FCal moduleswere positioned as in ATLAS, including all details of cryostat walls and supports (dead material).One quarter of the full HEC1 and HEC2 wheels were assembled with dedicated small moduleshaving reduced|η | coverage and encompassing only the forward region. Similarly, one EMEC3740
inner wheel module (1/8 of the full EMEC wheel) and 1/4 of the full FCal1 and FCal2 detectorswere assembled. The analysis of the data is in progress[152], in particular in the EMEC/HEC andthe FCAL regions. The response and energy resolution for electrons have been compared with MCsimulations and show good agreement. Also the response of pions has been studied in detail, usingthe em scale first and moving on to separate calibration constants per calorimeter. The comparison3745
with MC is also under study.(more to come)
In conclusion, the beam tests demonstrate that the HEC performance matches well the require-
– 140 –
[GeV]BeamE0 20 40 60 80 100 120 140 160 180 200
(N
oise
Sub
trac
ted)
[%]
reco
/Ere
coσ
0
2
4
6
8
10
12
14
16
18
20
0.00733±0.133 & Constant=3.73±Sampling=27.5
[GeV]BeamE0 20 40 60 80 100 120 140 160 180 200
(N
oise
Sub
trac
ted)
[%]
reco
/Ere
coσ
0
5
10
15
20
25
30
ATLAS Forward Calorimeter Pion Energy Resolution
0.161±0.699 & Constant=2.62±Radial weights with no clustering: Sampling=78.20.113±1.04 & Constant=7.75±Flat weights using a 16 cm cylinder cluster: Sampling=96
0.119±0.988 & Constant=7.09±EM Scale using a 16 cm cylinder cluster: Sampling=99.8
ATLAS Forward Calorimeter Pion Energy Resolution
Figure 91. Left hand plot: Electron resolution showing a stochastic term of 28% and a constant term of3.7%. Right hand plot shows the FCal pion resolution based onthe response of all three modules. A fit tothe data gives a stochastic term of 78% and a constant term of 2.6%.(Final figures soon)
ments as given in [121, 105].
5.6.3 FCal stand alone beam test performance
There were two FCal stand-alone beam tests, one in 1998 and one in 2003. The 1998 test used3750
engineering modules which were full depth segments of FCal1and FCal2. This test allowed toextract the electron response of FCal1 and an estimate of theEM scale for FCal2 [153]. The 2003beam test was carried out with the final production modules and used electronics that were verysimilar to final ATLAS electronics. Data were taken with electron and pion beams with momentafrom 10 to 200 GeV/c. The data were analysed using the standard LAr technique of Optimal3755
Filtering (OF) in the ATHENA framework. Electronics noise was calculated from the data andwas subtracted in quadrature before energy resolution plots such as those shown in Fig. 91 wereproduced. The points in these plots include data from all three modules and have been fitted withthe standard parametrization yielding stochastic and constant terms of 28% and 3.7% for electronsand 78% and 2.6% for pions. It should be noted that the very lowconstant term for pions is a feature3760
of the analysis technique and a better figure of merit is the measured 6% calorimeter resolution for200 GeV/c pions. The pion data analysis uses a radial weighting technique which compensatesfor the distribution of EM and hadronic energy within a shower. Using this technique realizes thefull potential of the fine transverse segmentation of the FCal modules. The resolutions obtainedin the beam tests exceed the design specification of the ATLASTechnical Proposal which were3765
stated as stochastic and constant terms of 100% and 10% respectively for hadrons. Further detailsof the results of the 2003 test beam can be found in Reference [154], including a discussion of theabsence of significant tails in the FCal energy resolutions which are important for optimizing theoverall missing energy resolution.
5.6.4 Tile calorimeter performance3770
Muons of 180 GeV/c with trajectories at 90 to the module symmetry plane were used to studythe photoelectron (p.e.) yield for all 11 tile sizes. Two independent techniques were used, based:a) on the spread of the difference of the signals from the two PMTs reading the same cell; b) onthe spread of a laser-generated signal reproducing a muon signal in the same cell. The results of
– 141 –
the two methods are in excellent agreement; the light yield is independent of the size of the tiles3775
(within 1-2% errors). These results also reproduced the batch-to-batch light output variation seenduring tile production, and also the systematic differenceassociated with the two sources of rawpolystyrene (the p.e. yield is about 80/GeV for one, and about 100/GeV for the other). In theExtended Barrel modules the yield is 10% to 20% lower, reflecting the choice that was made to usebatches with higher light yield for the barrel modules. The same 90 muon data allowed studying3780
the local and global features of the calorimeter response. The RMS spread of the mean signalsfrom the 11 tile rows (sizes) is 2.0% (2.8%) for Barrel (Extended Barrel) modules and the spreadwithin each tile size is about 3%. The average response within a module varies between moduleswith an RMS of 1.1% (1.2%) in the Barrel (Extended Barrel), while the variation within a moduleis typically 4-5%.3785
E (pC)
Eve
nts/
0.11
pC
0
1000
2000
3000
4000
5000
6000
0 2 4 6 8 10
Figure 92. Signal detected, summed in depth, for muons of 180 GeV/ctraversing the tile calorimeter. The spike near zero is the signal pedestal.
In the ATLAS experi-ment, muon trajectories areprojective and may clip theedges of cells. Summingover a tower (three radial3790
segments) signal/noise ratiosare typically 44, whereasS/N 18 in the last radial seg-ment. The response to 180GeV muons over angles (-3795
1.49< η <1.35) has an RMSvariation of 1.9%, exclud-ing angles where coverage isnot complete. The typicalmuon signal, summing over3800
a tower, is shown in figure92.
The modules were ex-posed to electron beams of10 to 180 GeV/c to set the energy conversion scale (i.e. the PMT HVs) on a significant frac-3805
tion of the entire calorimeter. This basic calibration willbe transferred using the cesium system tothose modules which were not calibrated using high energy beams. To avoid a dependency on theperiodicity inherent in the scintillator/iron structure,the electron response is extracted by averagingsignals measured over one scintillator/iron period, as seen at the chosen angle of incidence: atη =0.35 (about 20 to the symmetry plane of the calorimeter). Electron showersare mostly contained3810
in the first radial compartment, therefore precise responsevalues can only be extracted from thisfirst layer of cells. The RMS spread of these responses over the entire exposed module sample is2.6%. The electron response is linear, within deviations ofabout 1%. From a linear regression tofive beam energies the EM signal conversion factor of 1.04 pC/GeV is determined. The electronresponse is observed to vary with the angle of incidence, as expected because of the variation in3815
the effective calorimeter granularity with angle. Betweenextremes, corresponding to angles to thetile/iron plate planes from 0 to 90, the response varies from 0.995 pC/GeV to 1.075 pC/GeV.
– 142 –
Systematic measurements were made with hadron beams of 20 GeV/c ≤ p≤ 180 GeV/c, witha few additional runs at higher (350 GeV/c) and lower momenta(3 GeV/c to 10 GeV/c). Beamstypically contained a mixture of pions, kaons and protons. The beam line included a Cerenkov3820
counter, set up to separate pions and protons from 50 GeV/c upwards, which was used to obtainthe results reported next.
The response to pions, expressed as the ratio of the detectedenergy obtained using the elec-tron conversion factor, to the beam energy is less than one. However this ratio grows only from0.82±0.02 to 0.85±0.01 over the range 20 GeV/c≤ p ≤ 350 GeV/c. Hadronic shower leakage3825
corrections are of order 2-3% and do not significantly affectthe energy dependence of this ratio.The fractional energy resolutionσ /E, parametrized as usual as the quadrature sum of a statis-
tical A/√
E and a constant term B, displays a significant dependence onη because the granularityof the calorimeter gets finer, and leakage decreases, with increasingη . This leads to a decrease ofboth A and B terms with increasingη : for η= 0.25, A = 57.5%± 0.5% and B = 5.9%± 0.1% ,3830
while atη= 0.55, A = 53.1%± 0.4%, B = 5.2%± 0.1%. These resolutions are in good agreementwith GEANT4 simulations, using the QGSP 2.7 physics list.(Results on pion signal in 2004 combined run expected soon.)
– 143 –
6. Muon system
6.1 Overview3835
ChambersChambers
Chambers
Chambers
Cathode StripResistive Plate
Thin Gap
Monitored Drift Tube
Figure 93. The Muon Spectrometer with its fourchamber subsystems(Correct EO wheel)
The muon system forms the outer part of theATLAS detector and is designed to recordtracks penetrating the calorimeters in the cen-tral and forward region up to|η | < 2.7.Chambers in the central region (”barrel”) are3840
interleaved with the eight coils of the SCbarrel toroidal magnets, while the forwardparts are in front and behind the two end-cap toroidal magnets. Theφ symmetry ofthe toroidal magnets is reflected in the sym-3845
metric structure of the muon chamber sys-tem, consisting of eight octants. Each oc-tant is subdivided in two sectors, containingtwo chambers with slightly different lateralextension, leading to a small region of over-3850
lap in φ. This overlap of the chamber bound-aries allows for cross calibration of adjacentchambers, as tracks in the overlap region arerecorded in both chambers.
The chambers in the barrel are arranged in three cylindricalshells around the z-axis of the3855
detector at radii of about10 5, 7.5, and 10 m. In the two end-cap regions muon chambers formlargecircular disks at distances of 7.4, 10, 13, and 21 m from the central plane. Fig. 93 shows the overalllayout of the muon system while Figs. 94 and 95 give a cross section in transverse and longitudinaldirection.
The aim of the layout was to achieve a momentum resolution of about 10% for tracks at 1 TeV,3860
which translates into a sagitta of about 500µm to be measured with a resolution of≤ 50 µm.
Basic element of the precision momentum measurement is the Monitored Drift Tube Chamber(MDT), combining high measuring accuracy with mechanical rigidity and simplicity of construc-tion. The tubes are operated at an absolute pressure of 3 atm,improving the measuring accuracy byabout a factor of
√3 compared to operation at normal pressure, leading to a resolution of 80 µm3865
per tube.
In the very forward region(2 < |η | < 2.7) Cathode Strip Chambers (CSC) are used becauseof higher rate capability and time resolution. The CSC’s aremultiwire proportional chambers withcathodes segmented into strips, the direction in both cathodes being orthogonal to each other. Thisallows to measure both coordinates from the distribution ofthe induced charge. The resolution3870
of the CSC’s in the bending plane will be 60µm and about 5 mm in the transverse plane, thedifference being due to different spacing of the readout channels.
10the exact values are listed in table 31
– 144 –
resolution in meas’mts/track Number of
Type Function η φ time barrel end-cap chambers channels
MDT precision 35 µm – 750 ns 20 20 1172 (1108) 354k (339k)
CSC precision 40 µm 5 mm 4 ns – 8 (4) 64 (32) 61.4k (30.7k)
RPC trigger 3 cm 3 cm 20 ns 6 – 622 (560) 373k (359k)
TGC trigger 0.7-3.6cm2-3 cm 20 ns – 8 3588 318k
Table 29. Parameters in the four subsystems of the muon detector. Numbers in bracket are for the initialphase of the experiment.
In order to monitor the position of the MDT chambers with the required accuracy an alignmentsystem based on infrared light rays was chosen. Chambers arealigned internally to monitor defor-mations, at close distance to monitor the relative positionamong neighbours and along projective3875
lines, corresponding to infinite momentum tracks, to monitor displacements in a row of sequen-tially transversed chambers. In total, 5800 and 6400 alignment rays are used for barrel and end-capalignment, respectively, see Section 6.3.4.
An important design criterion of the muon system was the capability to trigger on muon tracks.The precision chambers had therefore to be complemented by asystem of fast chambers, capable3880
to deliver track information a few tens of nanoseconds afterthe passage of the particle. In the barrelregion Resistive Plate Chambers (RPC) were selected for this purpose, while in the end-cap ThinGap Chambers (TGC) are used. Both chamber types deliver signals inside 15-20 ns, being able totag the beam crossing.
The trigger chambers readbothcoordinates of the track, one in the bending (η ) and one in the3885
non-bending (φ) plane, while the MDT’s are only measuring theη coordinate, the position of thetrackalongthe tube being unknown. Combining the two track coordinatesfrom a trigger chamberswith the one from the MDT allows to assign the second coordinate to the MDT’s.
The muon system had to be designed to tolerate the radiation levels in the experimental hall,which may induce ageing effects in the detectors and create damage in the electronics. All com-3890
ponents were therefore tested to withstand at least five times the radiation levels predicted by thesimulation studies.
Table 29 gives an overview of the main parameters of the four subsystems of the muon detector.
6.2 Precision chambers
The concept for the precision chambers is to determine the coordinate of the track in the bending3895
plane with the MDT’s, deriving the coordinatealong the tube, (i.e. in the non-bending plane) froma combination of MDT chamber and trigger chamber information. In the trigger chambers bothcoordinates are measured, though with less precision. After matching of the MDT and triggerchamber hits in the bending plane, the trigger chamber’s coordinate in the non-bending plane isadopted as the "second coordinate" of the MDT measurement, i.e. the location of the track along3900
the tube.
– 145 –
Figure 94. The muon barrel consists of threecylindrical chamber layers, segmented in 16 sec-tors of Large and Small chambers.(x-scale to be
added)
Figure 95. In the longitudinal direction muon cham-bers are arranged in projective towers seen from theinteraction point (schematic).(CSC’s to be added)
This method assumes that in any MDT/trigger chamber pair a maximum of one track per eventbe present, as for two or more tracks theη andφ hits cannot be combined in an unambiguous way.In fact, simulations have shown that the probability of a track with pT > 6 GeV is about 6×10−3
per beam crossing, corresponding to about 5×10−6 per chamber. Assuming uncorrelated tracks,3905
this leads to a negligible probability to find> 1 track in any MDT/trigger chamber pair.11
The dimensions of the MDT in inner, middle, and outer layer are projective, i.e. their respec-tive size increases in relation to their distance from the origin.
While the segmentation in the azimuthalr −φ plane follows a hexadecimal symmetry, ina sequence of small and large sectors (Fig. 94), the structure in the longitudinalr − z-plane is3910
segmented in projective "towers" with three towers in barrel and end-cap each. Hence, infinitemomentum tracks will travel inside the same sector and tower, while finite momentum tracks maycross tower boundaries due to their curvature. Fig. 95 showsthe tower structure in ther − z-direction.
The MDT chambers cover the region|η |< 2. There are 1172 MDT chambers of 18 main types3915
in the detector with a total area of 5500 m2. Besides the ”main” type chambers a considerablenumber of chambers with ”special” shapes has been built to minimize acceptance losses in theregions around the magnet coils and support structures.
In the very forward region(2 < |η | < 2.7) a higher rate of correlated tracks may be encoun-tered. This problem is particularly severe in the first chamber layer, where tracks have not been3920
separated by the magnetic field. The CSC chambers are designed to measure both coordinates ofa track and in addition resolve ambiguities if more than one track is present, using the measuredcharge deposition in both directions. In the final installation each track will be measured in eightsubsequent CSC layers as is the case for the adjacent MDT’s. 32 CSC’s on each side will then beinstalled, covering an area of about 65 m2 in total.3925
11Correlated close-by muon pairs may be caused by 2-body-decays of low-mass particles. The corresponding ambi-guities may be resolved by matching muon track candidates with tracks from the inner detector.
– 146 –
6.3 Monitored drift tube chambers
6.3.1 Structure and function of the drift tube
The basic element of the Monitored Drift Tube chambers is thepressurized drift tube with a di-ameter of 30 mm, operating with Ar/CO2 gas (93:7) at 3 bar. Ionization is collected at the centraltungsten-rhenium wire with a diameter of 50µm, at a potential of about 3080 V. The wire is held3930
in position at the tube ends by a cylindrical element (”endplug”) which guarantees the centricity ofthe wire with respect to the tube with an accuracy ofσ < 10 µm. The central conductor holdingthe wire also serves for the gas transfer in and out of the tube. Signal transmission to the electronicsand connection to the HV supply system are at opposite ends. Fig. 96 shows a cross section throughthe endplug.3935
Figure 96. Longitudinal cut through a MDT tube.Figure 97. Cross section of a MDTtube.
The MDT’s chambers, built of these tubes, combine good measuring accuracy with high me-chanical precision and strength. A high level of reliability for the chamber can be expected, thefailure of a single tube not affecting the operation of the others.
Another advantage of the tubes for our application lies in the radial shape of the electric field.In this geometry the point measuring accuracy does not depend on the angle of incidence of the3940
track onto the chamber plane, as the coordinate of the track is determined by the radius of the circlearound the wire to which the track is tangential (see Fig. 97). This is important as the angle ofincidence of infinite momentum tracks onto the chamber planeextends up to 45o. In chamberswith rectangular drift geometry the time of arrival would bedegraded by the smearing-out of theionisation along the drift field.3945
A disadvantage of the radial drift geometry, on the other hand, is the long pulse train causedby the track segments far from the tangential point, which may produce several threshold crossings(”hits”) per track. In our tubes the maximum drift time from the wall to the wire is about 750 nsfor Ar/CO2 (93:7). A track passing close to the wire thus generates a pulse train with a durationof this order, while only the arrival time of the central track part, closest to the wire, is relevant3950
for the track coordinate (rmin in Fig. 97). To prevent an inflation of the data volume by multipletrack hits, an adjustable dead time has been implemented into the front-end of the readout chain,see section 6.3.3.
Naming of chambers is based on their location in barrel or end-cap (B,E), their assignmentto Inner, Middle or Outer chamber layer (I, M, O) and their belonging to a Large or a Small3955
sector (L,S). The sector number (1–16) and the sequence number in a chamber row are added tocompletely specify a MDT chamber. A BOS chamber, for example, is located in a Small sector
– 147 –
of the Barrel, Outer layer, while an EML lies in the Large sector of the Middle layer of the end-capBecause of mechanical interference with the end-cap toroidal magnet the Inner layer of the end-cap MDT’s is separated in two concentric rings of chambers with an offset in z of about 2.50 m.3960
Here chambers in the inner ring are called EIS and EIL, while the ones in the outer ring, moreremote from the interaction point, are named EEI and EEL (derived from ”Extra” chambers).
6.3.2 Mechanical structure of MDT chambers
MDT chambers had to be built in many different shapes to optimize solid angle coverage whilerespecting envelopes dictated by magnet coils, support structures and access ducts. MDT’s are3965
rectangular in the barrel and trapezoidal in the end-cap their outer dimensions being defined by thesegment/tower structure (Fig. 94 and 95). All regular MDT chambers consist of two packages of
Parameter Design value
Tube material AlMn1
Outer tube diameter 30 mm
Tube wall thickness 0.4 mm
Wire material W-Re (97:3), gold-plated
Wire diameter 50 µm
Gas mixture Ar-CO2 (93:7)
Gas pressure 3 bar (absolute)
Gas gain 2 x 104
Wire potential 3080 V
Maximum drift time ∼ 750 ns
Avg. resolution per tube ∼ 80 µm
Table 30.Main MDT chamber parameters.
tube layers (multilayers) separated by a support structure, where each multilayer consists of threetube layers in the outer and four tube layers in the inner layers of the muon detector. Fig. 98 showsthe structure of a barrel chamber with 2× 3 tube layers. The height of the support beam between3970
the multilayers (”Spacer”) depends on the chamber type, varying between 6.7 and 317 mm, seeTable 31 and 32. Detailed information about chamber dimensions and other parameters is availablein [155].
During chamber construction high precision of tube placement and a high level of mechanicalstrength had to be achieved in order to maintain the inherentresolution of the drift tubes. The3975
two multilayers of a MDT chamber are mounted on a support frame of solid aluminium beams,providing mechanical strength and rigidity to the structure. The lateral support beams (”CrossPlates”), designated by RO for readout and HV for the high voltage supply side, are interconnected
– 148 –
by two longitudinal beams (”Long Beams”). Three kinematic mounting points, not shown in thefigure, are attached to the frame for installation into the rail system of ATLAS.3980
The frames also carry most of the interfaces to gas supplies,electrical, monitoring and align-ment services. The 3 (4) tube layers of a multilayer are jointtogether with epoxy glue, layer bylayer, after having been carefully aligned on a granite table. Reference surface for the tubes is theprecisely milled circumference of the endplug, which also serves as reference for wire positioning.This way, a high precision of relative wire positioning is assured during construction.3985
The straightness of the tubes is required to be better than 100 µm. The relative positioningof wires, reached during production, has been verified to be better than 20µm. The gap betweenadjacent tubes filled by glue is 60µm.
Figure 98. Mechanical structure of a MDT chamber. Three spacer bars connected by longitudinal beamsform an aluminum space frame, carrying two so-called multilayers of three or four drift tube layers. Fouroptical alignment rays, two parallel and two diagonal, allow for monitoring of the internal geometry ofthe chamber. RO and HV designate the location of the ReadOut electronics and High Voltage supplies,respectively.
Due to gravitational forces, chambers are not perfectly straight but suffer a certain elasticdeformation. The BOS chambers for example, with a total length of 3.92 m, have a gravitational3990
droop of about 800µm when supported at the two ends in the horizontal position. The wires inthe tubes have only 200µm droop at their nominal tension of 350 g. In order to re-establish thecentricity of the wires, the droop of the multilayers can be corrected by the so-called sag adjustmentsystem, which applies an adjustable force to the central cross plate. Using the in-plane alignmentsystem as reference, deformations can be corrected with a precision of about 10µm. Thus, for3995
each angle of installation in the ATLAS detector the droop ofdrift tubes and wires can be matched,leading to wire centricity and circular symmetry of the drift field.
The precision achieved in construction has been verified in cosmic ray tests and in the X-rayfacility operated at CERN (”X-ray tomograph”). The result was an average deviation of the wirepositions of≤ 20 µm from the nominal over the full area of the chamber.4000
Inspite of the solid construction of the MDT chambers, deformations are expected to occurin the various mounting positions in ATLAS and may change in time when thermal gradients are
– 149 –
present. Therefore, a chamber-internal alignment system was implemented, which continuouslymonitors potential deformations of the frame. The alignment system consists of a set of four opticalalignment rays, two running parallel to the tube direction and two in the diagonal direction between4005
the corners of the chamber. The lenses for the light rays are housed in the middle, while LEDs andCCD sensors are located in the outer spacers. This system canrecord deformations of a fewµmand is meant to operate during production, installation, and experimental running in ATLAS.
Name Layer stand.spec. tube locat. tubes/tubes width chamberspacer
chambers layers in r layer total along z length height
BIS inner 90 38 2×4 4550 30 240 916 1820 6.5
BIL inner 72 48 2×4 4949 36 28 1096 2820 170
BMS middle 48 32 2×3 8095 48 288 1497 3220 170
BML middle 35 59 2×3 7139 56 336 1697 3700 317
BEE middle 32 1×4 4415 48 192 1457 1060 –
BOS outer 62 44 2×3 10569 72 432 2177 3920 317
BOL outer 48 48 2×3 9500 72 432 2177 5110 317
387 273
Table 31. MDT chambers in the barrel. Dimensions (mm) in column 9 referto the standard chambers.”Special” chambers have reduced dimensions resp. cut-outsto fit between magnet coils, support structureetc.
The exact definition of the wire positions in a MDT, due to tight tolerances in production andcontinued monitoring of global deformations, results in the accuracy of the track segment in the 64010
(8) tube layers being only limited by the error of the single tube measurements of about12 80 µm.Thus the resolution of the central point of a track segment ina multilayer of 3 (4) tubes is 50(40) µm, while the one in a chamber of two multilayers is 35 (30)µm. For a track crossing threeMDT chambers in a projective tower a sagitta resolution of 45µm is expected, corresponding to amomentum resolution ofδp/p= ∆s× p(TeV)/500µm. The exact position of the chambers inside4015
a tower is monitored by the projective alignment system witha design accuracy of about 30µm,see section 6.3.4.
To monitor thermal deformation each chamber contains 6-18 temperature sensors, while thelocal magnetic field is monitored by a pair of B-field sensors per chamber. Both environmentalparameters are read out via the Detector Control System (DCS), in which each chamber is repre-4020
senting a node.An important design criterion in chamber construction was electrical integrity, i.e. low-impedance
connections had to be provided between all metallic components. The bonds between most cham-ber components being epoxy-glue, beams, gas bars, and support structures were therefore con-
12This value deteriorates for significant levels of background radiation, see section 6.3.5
– 150 –
Name Layer # of tube locat. rad. tubes/tubesradial outer chamb.spacer
ch’s layers in z ch’s layer total width length height
EIS inner 32 2×4 7261 2 36 288 1096 1745 121
EIL inner 64 2×4 7675 4 54 432 1637 3295 121
EES middle 32 2×3 10276 2 40 240 1216 2951 121
EEL middle 32 2×3 11322 2 40 240 1216 4703 121
EMS middle 80 2×3 13878 5 64 384 1937 3860 170
EML middle 80 2×3 14294 5 64 384 1937 6008 170
EOS outer 96 2×3 21424 6 48 288 1457 4202 170
EOL outer 96 2×3 21840 6 48 288 1457 6503 170
512
Table 32.MDT chambers in the end-cap. Numbers and dimensions in columns 7–10 apply to the outermostof the trapezoidal chambers in each radial row.
nected with solid wire straps, while tubes were screwed to a continuous metallic sheet (”ground4025
plane”) which in turn was connected to the support structure. The protective cover of the electron-ics (”faraday cage”), made from thin aluminum sheets, was chromatized to assure a conductive,unoxidizable surface.
To avoid ground loops among chambers the kinematic mounts, sliding on the rails, have a ce-ramic inner surface, while non-conductive transition pieces are used to connect the gas bars to the4030
external supplies. Care was also taken to isolate the readout of the in-plane alignment (”RASNIK”)and the DCS from the chamber ground. After installation the resistance between individual cham-bers and the ATLAS structure was> 1 GΩ. To assure a low-impedance connection to the safetyground, rows of chambers were finally connected toonecommon ground line, which, in turn, wasconnected to the ATLAS structure at a single point.4035
6.3.3 Signal path and readout electronics
The architecture of the MDT readout electronics chain is shown in Fig. 99. In the first stagethe raw signals from the tubes are amplified, shaped and discriminated, eight tubes being servedby one monolythic ASD chip (Amplifier/Shaper/Discriminator). The outputs of the ASD, binarydifferential signals, are routed to the Time-to-Digital Converter (TDC), where the arrival times of4040
leading and trailing edges are stored in a large buffer memory together with an identifer of thecorresponding tube and a flag for the leading respectively trailing edge. The time is measured inunits of the TTC system clock (Timing, Trigger and Control) of 40 MHz, 12 bits being used for thebeam crossing identification (BXID) and 5 bits for the subdivision of the 25 ns-period (fine time).The fine time period of 0.78 ns therefore leads to a RMS timing error of 0.23 ns corresponding to4045
an average position error of about 5µm.
– 151 –
Figure 99. Schematic diagram of the MDT readout electronics.
An additional feature of the ASD is the measurement of the pulse height of the signal, whichallows monitoring of the gas gain as well as pulse height dependent corrections ("slewing"). Thepulse height is measured by a Wilkinson ADC and encoded as delay between leading and trailingedge, a delay of 150 ns corresponding to the maximum pulse height. 13 The ASD contains a4050
programmable dead time to prevent more than one leading edgeinside a given time. It can beselected in the range 200-750 ns, the latter corresponding to the maximum drift value.
Programming of ASD and TDC is done, following the JTAG protocol, via a shift register,allowing for a 128 bit string for ASD and TDC each. This way many parameters and functions canbe controlled, like the setting of discriminator thresholdand deadtime as well as triggering of test4055
pulsing for calibration or deactivation of noisy channels.The TDC with three ASDs, serving 24tubes, is implemented on a 9 cm x 11 cm PCB ("Mezzanine Card").The Mezzanine Card is thebasic readout element of the MDT’s and is conntected to the tubes via a signal distribution card("Hedgehog Board"). There are about 15000 mezzanine cards to serve the 355000 MDT tubes inthe system.4060
MDT chambers contain up to 18 mezzanine cards, which are controlled by a local processor,the Chamber Service Module (CSM). The CSM broadcasts the TTCsignals to the TDCs and col-lects the hits sent back from them for each LVL1 trigger. Subsequently, the data are formatted,stored in a large derandomizing buffer and sent to the MDT Readout Driver (ROD) in USA15 viaan optical link. The ROD is a VME module, serving up to six CSMs. Its main task is formatting4065
of the front-end data for rapid transfer to the Read Out Buffer (ROB), where data are stored untilthe event has been retained or discarded by the Level 2 Trigger logic. Table 33 summarizes themodularity of the readout system.
An important feature of the ROD is its capacity to monitor theincoming data, due to its largeprocessing power. The data stream received from the CSMs canbe monitored for tube and chamber4070
occupancies and deviations from nominal values, possibly pointing to malfunction. As the RODssee the full L1 event rate they can accumulate significant statistics in a short time, allowing for anearly recognition of errors. Depending on average event size and L1 rate, a variable fraction of theevents will be monitored in order to not slow down data transfer to the ROBs, which has priority.A detailed description of the MDT readout is given in [156] and [157].4075
– 152 –
tubes ASDs mezz cdsCSMschambersRODs power
Barrel 19190424162 8054 624 656 110 14.7 kW
Two end-caps16224020280 6760 516 516 94 12.2 kW
Total 35414444442 14814 1140 1172 204 26.9 kW
Table 33.Modularity of the MDT readout electronics. In some cases in the barrel two chambers are servedby one CSM, leading to a excess of 32 chambers with respect to the number of CSMs. The allocation ofRODs to barrel or end-cap is approximate, some RODs serving CSMs from both.
Figure 100.Principle of the alignment of the ATLAS muon spectrometer.
6.3.4 The MDT alignment system
In the ATLAS spectrometer the MDT chambers will be installedwith about 5 mm and 2 mradaccuracy with respect to their nominal positions. In order to achieve the required momentumresolution, however, the actual chamber positions must be known with an uncertainty≤ 30 µm,which is more than 2 orders of magnitude below the installation accuracy. To reach this level4080
of precision, an alignment system was developed which, based on optical imaging, referred thepositions of each chamber to its neighbours in the layer and along r-z in the tower, taking theposition of the toroidal coils as the master reference.
Fig. 100 shows the schematics of the alignment system for onequarter of the detector. In thebarrel, the chambers inside a MDT row are referred to each other by chamber-to-chamber alignment4085
rays (”praxial” and ”axial” rays), while the projective system connects the three layers of a tower,simulating the trajectories of an infinite momentum track originating from the interaction point.In the end-caps, due to the presence of the toroidal magnets,the projective alignment proceedsin two steps using precise rulers (”alignment bars”), as intermediate reference. The projectiverays, describing infinite momentum tracks, connect the alignment bars, while the chambers of the4090
13Thus, unlike the leading edge the trailing edge does not correspond to the discriminator output.
– 153 –
four wheels (EI, EE, EM and EO) are referenced to these bars. With the internal accuracy of theMDT’s being guaranteed by construction, and their planarity being monitored by optical in-plane-alignment (see section 6.3.2), the alignment system thus forms a tight, stable grid for surveying theposition of each MDT wire in the muon system. A detailed presentation of the alignment strategyand technical implementation is given in [158] and [159].4095
Figure 101.The RASNIK alignment system.
As an example, Fig. 101 shows theschematics of the RASNIK system, the mostfrequently used element of the muon align-ment, where an image sensor looks througha lens towards a lighted target carrying a4100
chequerboard-like pattern. The sensor imageof the target is analysed online, retaining fourparameters to characterize the deviation fromthe nominal geometry, namely: two translations perpendicular to the optical axis, the rotationaround and the longitudinal position along the axis, the latter being derived from the optical mag-4105
nification. For a symmetric RASNIK, with the lens halfway between image sensor and mask, atranslation resolution of 1µm and a magnification resolution of 2× 10−5 has been obtained. Adescription of the RASNIK alignment can be found in [160] and[161].
A more detailed layout of the alignment rays for three adjacent barrel sectors is shown inFig. 102. The praxial systems (two short, crossed RASNIKs connecting adjacent chamber corners)4110
and the axial systems (interleaved long RASNIK systems along the chamber) control the geometryof a each row of six chambers. The relative position of the chamber layers perpendicular to theprojective direction is controlled by the projective rays.
Figure 102. Layout of the alignment lines for threeadjacent barrel sectors.
Figure 103.Layout of the end-cap alignment system.
The projective system works with RASNIKs, where BIL holds the mask, BML the lens andBOL the image sensor. Small chambers are not equipped with projective lines and will be aligned4115
with respect to the large chambers by common tracks in the overlap region.The layout of the alignment rays in the end-cap is shown in Fig. 103. Due to the presence of
the toroid, there are only two projective rays, plus a polar ray passing through the horizontal tube(”stay tube”) connecting the two outer flanges of the cryostat of the toroidal magnet. To cope with
– 154 –
this limitation, a grid of optically connected alignment bars using a quasi-projective alignment has4120
been used. The chambers are referenced to these bars like to precision rulers. Each wheel carrieseight bars, with lengths up to 9 m, placed at the border between large and small sectors. The barstraightness and elongation is controlled with a set of internal RASNIKs, interleaved among eachother, and with a number of temperature sensors.
Barrel Alignment Endcap Alignment
Name Type # Function Name Type # Function
Inplane R 2110 MDT planarity Inplane R 2048 MDT planarity
Praxial R 2006 Ch-to-ch distanceBar monitor. R 384 Bar deformation
Axial R 1036 Ch-to-ch planarityRadial B 96 Bar deformation
Projective R 117 Tower alignment Polar B 208 Bar-to-bar link
Reference S 256 Link to toroid Azimuthal B 608 Bar-to-bar link
CCC S 260 Sm-to-Large link Proximity R 2560 Ch-to-bar link
BIR-BIM R 32 BIR/BIM link Saloon Door B 512 Ch-to-bar link
5817 6416
Table 34. The alignment components in barrel and end-cap. Three different types of imaging systems areused, adapted to the special application: RASNIK (R), BCAM (B) and SaCam (S).
6.3.5 Performance of the MDT chambers4125
MDT chambers have been tested under various conditions trying to match the environment of thefinal experiment as closely as possible.
There were tests of a set of 12 MDT chambers in a high energy muon beam in 2003 and 2004in the H8 area at CERN. The 12 chambers were arranged similarly to a large sector in the barreland end-cap. Main purpose was to verify functionality and performance of the readout electronics4130
and the projective alignment system. The resolution dependence on thresholds and gas gain wastested as well as efficiency and noise rates. The alignment system was used to monitor the chamberpositions relative to each other recording, for example, day-night displacements of the chambersdue to temperature. The alignment data were compared to the momenta measured in the MDT’s ina similar way as will be done in the experiment and a perfect correlation was found.4135
In 2004 tests were done together with the muon trigger chambers and with other subsys-tems of ATLAS like pixel, SCT, TRT, and calorimeter. The mainresult was a confirmation of theperformance with respect to the resolution parameters, readout functionality and stability of theelectronics as well as, once again, the correct functioningof the alignment system.
Additional test were done in 2003/2004 at the Gamma Irradiation Facility (GIF) at CERN in4140
order to test the behaviour of the chambers in a high background environment as is expected in theATLAS hall. In the GIF test area a high energy muon beam was available as well as an adjustable,
– 155 –
intenseγ-source. The muon tracks were defined by a silicon telescope and the resolution of theMDT tubes could be studied accurately as a function of the track’s distance from the central wireand of theγ-intensity. Expectation was that electrons from Compton conversions in the tube walls4145
would lead to high occupancies, while positive ions, slowlydrifting towards the wall, would locallydistort the drift field, leading to a radius-dependent degradation of the resolution.
Fig. 104 shows the results. The degradation of resolution with gamma intensity is clearlyvisible being highest for the longest drift path close to thewall (r = 15 mm). The highest irradiationlevel expected for MDT chambers in ATLAS is≤ 150 hits/(s×cm2). In the centre of the tube(r =4150
7.5 mm) this would degrade the resolution from 60µm to about 80µm which seems acceptable.
0 2 4 6 8 10 12 14
990 / (cm² s)670 / (cm² s)342 / (cm² s)
0187 / (cm² s)
Threshold: (16±2) p.e.300
225
275
175
200
150
125
100
50
Irradiation rates:
75
250
spatialre
solu
tion
σ(µ
m)
r (mm)
Figure 104. Resolution as a function of the impact parameter of the trackwith respect to the tube wire atvarious levels ofγ-irradiation. The maximum rate expected for the MDT’s is≤ 150 hits per cm2 and sec.
6.4 The Cathode strip chambers
6.4.1 Layout of the CSC system
The limit for safe operation of the MDT’s is at counting ratesof about 150 Hz/cm2 which will beexceeded in the region|η | > 2. In the first layer of the end-cap the MDT’s are therefore replaced4155
by Cathode Strip Chambers, a MWPC type which combines high spatial, time and double trackresolution with high-rate capability and low neutron sensitivity. Operation is considered safe up tocounting rates of about 1000 Hz/cm2, which is sufficient up to the forward boundary of the muonsystem at|η | = 2.7.
To achieve a certain homogeniety of the architecture the layout was made similar to the one4160
of the adjacent MDT’s, where two multilayers of four tube measurements are combined, resultingin eight measurements along the track. In the CSC layout two groups of four CSC detector planesare used along the track, giving likewise eight independentmeasurements. There is a sequence ofLarge and Small CSC chambers inφ, comparable to the MDT layout. The whole CSC system will
– 156 –
consist of two disks with 32 chambers, 16 small and 16 large ones as shown in Fig. 105.14 The4165
CSC’s are multiwire proportional chambers with both cathodes segmented into strips perpendicularrespectively parallel to the wires. The position of the track is obtained by interpolation between thecharges induced on neighbouring cathode strips. The CSC wire signals are not read out.
Figure 105. Layout of a CSC end-cap with eightsmall and large chambers, made of two identicalmodules each.
The resolution achieved with this proce-dure depends mainly on the segmentation of4170
the cathode and the readout pitch, the lat-ter being the main cost driving factor for thereadout electronics. With a readout pitch of5.08 mm in the bending direction the ATLASCSC reaches a resolution of 60µm per CSC4175
plain, to be compared with the 80µm resolu-tion of a MDT tube layer. In the non-bendingdirection the cathode segmentation is coarserleading to a resolution of a few mm.
Apart from the precision and relative4180
simplicity of the coordinate determination,there are a number of other characteristicswhich make the CSC’s suitable for regions ofhigh particle densities:
a) Good two-track resolution and reliable4185
pairing of the projective measurements via thepulse height.
b) Electron drift times less than 30 ns re-sulting in a timing resolution of about 7 ns perplane.4190
c) Low neutron sensitivity because of thesmall gas volume and the absence of hydro-gene in the chamber gas (Ar/CO2/CF4). Theneutron sensitivity is< 104 while the photonsensitivity is of the order of 1%. A detailed description of the CSC-system is given in [162]. The4195
operating parameters of the CSC are shown in Table 35.
6.4.2 Spatial and time resolution
Fig. 106 shows the symmetric cell structure of the CSC, the anode-cathode spacing being equalto the anode wire pitch. The distribution of the charge induced onto the cathode strips by theavalanche on the wire is given in Fig. 107.4200
The segmentation of the cathode aims to sample the induced charge distribution as precisely aspossible while limiting the number of electronic readout channels. The following scheme has beenchosen: between two neighbouring readout strips (connected to amplifiers) there are two interme-
14In the first period of data taking only one module per chamber will be installed, so each track will be measured byfour instead of eight chamber planes. In this article all numbers refer to the final system.
– 157 –
diate (”floating”) strips capacitatively coupling the induced signal to the readout strips. The widthof readout and intermediate strips was selected to minimizethe differential non-linearity of the po-4205
sition measurement to≤1%. Fig. 108 shows the segmentation of the CSC cathode, wherereadoutand intermediate strips are 1.07 mm resp. 1.63 mm wide with a gap between strips of 0.25 mm.The interstrip capacitance is about 10 times the strip capacitance to ground. The intermediate stripsare connected to ground via a high resistance path to define the DC potential.
Anode wires
Cathode strips
d
d
WS
Figure 106. Structure of the CSC-cell. Right: cath-ode segmentation in the bending plane.
As the precision coordinate in a CSC4210
is obtained by therelative measurement ofcharges induced on adjacent strips, the per-formance is immune to the variation of fac-tors encountered by the whole chamber, likegas gain, temperature variations or pressure.4215
The primary limiting factor for the spatial res-olution of the CSC’s is electronic noise ofthe preamplifiers, and therefore only a smallnumber of strips around the centre is used inthe algorithm. In our geometry the best results4220
are obtained with 3 to 5 strips around the peak of the distribution.
0
0.2
0.4
0.6
0.8
-20 -10 0 10 20x (mm)
Cha
rge
frac
tion
Figure 107. Charge distributionon the cathode induced by theavalanche on the wire.
Figure 108. The readout and in-termediate strips in the bending-plane.
Figure 109. The inclination ofthe CSC’s towards the interactionpoint.
Parameter Value
Operating voltage 2600
Anode wire diameter 30 µm
Gas gain 4×104
Gas mixture Ar/CO2/CF4 (30/50/20)
Total ionization (normal track) 90 ion pairs
Table 35.Operating parameters of the CSC
– 158 –
The second most significant contribution to the spatial resolution of the CSC is the effect ofinclined tracks and the Lorentz angle. The charge interpolation is optimal when the avalanche isformed on a single point along the wire, while a significant extent of the anode charge results ina degradation. The spreading of the charge deposition can becaused by delta electrons, inclined4225
tracks and a Lorentz force along the wire, as the magnetic field is not collinear with the electricfield around the wire. While the effect of the Lorentz force cannot be corrected in a CSC, it doesnot lead to a systematic shift of the measured coordinate, but only slightly reduces the resolution.The effect of inclined tracks is minimized by tilting the chamber towards the interaction point sothat tracks are normal at the centre of the chamber (Fig. 109).4230
The strips in the second cathode plane are parallel to the wires and measure the position of theavalanche in the non-bendingΦ-plane. The spacing of the readout strips is larger than the one inthe bendingη -plane (15.7 mm) leading to a resolution of about 5 mm.
If more than one track is present in a given event, the pulse height is used to match the mea-surements in the two projections. Because of the broad pulseheight distribution in a gas chamber4235
and the high resolution of the ADCs, both tracks will have, inmost of the cases, a different pulseheight reading, allowing for an unambiguousη/φ assignment of the tracks from the separate mea-surements inη and Φ. In the rare cases of near equality, the information from theother threechamber layers will have to be used. In using the informationof all four chamber layers, theη/φassignment for two or more tracks seems to be sufficiently robust.4240
The time resolution of the CSC is an important parameter for the correct tagging of the beamcrossing. The maximum drift path of a primary electron to theclosest wire being only 1.25 mm, onewould expect a maximum drift time of about 20 ns, given the average drift velocity of 60µm/ns.However, the vanishing drift field at the boundary between two cells creates regions of very lowdrift velocity leading to long tails in the distribution of the arrival times, which prevents reliable4245
tagging of the beam crossing. By OR-ing the signals of corresponding cathode strips in the fourchamber planes, however, the tails disappear, as virtuallyno track can pass close to a cell boundaryin all four layers. The time of arrival distribution thus obtained is symmetric with an r.m.s. valueof 3.6 ns, resulting in a reliable tagging of the beam crossing.
6.4.3 Mechanical design4250
The CSC’s are located at about 7 m distance from the interaction point. They are mounted togetherwith the MDT’s and TGC’s on the shielding disk (”Small Wheel”) and occupy the radial spacebetween 881 and 2081 mm, corresponding to the pseudorapidity range|η | = 2−2.7.
The CSC’s come in two types of modules as shown in Fig. 110 and are arranged in two rings ofeight chambers each, as shown in Fig. 105. Two identical modules form a chamber, similar to the4255
two multilayers of a MDT. Each chamber module consists of four wire planes leading to a similarconfiguration as in the MDT system, but with much finer granularity. The modularity of the CSCsystem is summarized in Table 36.
The CSC design utilizes low-mass materials to minimize multiple scattering and detectorweight. A four layer module is formed by five flat rigid panels,each made of an 18.75 mm thick4260
sheet of nomex honeycomb and two 0.5 mm thick copper-clad FR4laminates, where the 17µmthick copper cladding forms the cathodes. The panel frames are made from machined rohacell, ahigh stiffness lightweight foam. Precision FR4 strips glued on the panels provide the 2.54 mm step
– 159 –
Figure 110.Shapes of the CSC modules.
Wires
Strips
Rohacell
Cathode read-outSpacer bar
Sealing rubber EpoxyWire fixation bar
Conductive epoxyHV capacitor
Anode read-out
Gas inlet/ outlet
0.5 mm G10 laminates
Nomex honeycomb
Figure 111.Structure of the CSC
for the anode wire plane. The anode wires have a diameter of 30µmand are made from gold-platedtungsten with 3% rhenium.4265
Figure 112.Model of a CSC chamber with four planes showing the location of the readout electronics.
There are 280 and 432 wires per chamber plane along the bending direction in the Small andLarge modules, respectively (Table 36). The segmentation of the cathode readout strips in thebending plane has a spacing of 5.08 mm, with 192 readout strips per plane, while in the transversedirection the spacing is 15.7 mm, with 48 readout strips. Thecathode strips are lithographicallyetched for highest precision. The five panels are precisely positioned with respect to each other4270
with the aid of locating pins. The whole assembly is rigid enough in order not to require a localalignment system. The outer copper-clad laminates of each module form an electromagnetic shieldfor the detector. A cutout view of one gap formed by two panelsis shown in Fig. 111. Signals fromthe cathode strips are transferred via ribbon cables to the ASM-I boards located on the chamberedges (Fig. 112).4275
The 16 chambers of an end-cap are mounted onto a precisely manufactured, solid supportstructure, defining the relative positions of the chambers with respect to each other.
– 160 –
Wires Strips R/O ch’s Strips R/O ch’s total R/O FE chipsASM-I
along wires across wires channels
Plane small 280 48 48 214 192 240 10
large 432 72 48 204 192 240 10
Module small 1120 192 192 856 768 960 40 10
large 1728 288 192 816 768 960 40 10
Octant sm+large 2848 480 384 1672 1536 1920 80 20
1 side 16 oct’s 45568 7680 6144 26752 24576 30720 1280 320
System 2 sides 91136 15360 12288 53504 49152 61440 2560 640
Table 36.Modularity of the CSC chamber system and readout modularity.
6.4.4 Readout electronics
The CSC chambers have to operate at the innermost part of the Small Wheel, a region which ischaracterized by a high density of detector components, difficult accessibility and high radiation4280
levels. As standard components like CMOS-processors, FPGAs or commercial ADCs are notexpected to operate reliably in this environment, the readout electronics had to consist of customdesigned building blocks, fabricated in radiation hard technologies, combining performance withsimplicity and reliability. To limit number and complexityof custom components, the front-endfunctionality had to be reduced to a minimum. The design therefore aimed at shifting the data4285
as early as possible – i.e. after the LVL1 trigger latency – tothe ReadOut Drivers (RODs) in theshielded area in USA15.
Figure 113.Schematics of the CSC front-end electronics.
The readout architecture of the CSC’s is shown in Fig. 113. Atthe first stage the chambersignals are amplified and shaped (ASM-I). Secondly the pulsetrain is stored in a Storage CapacitorArray (SCA) while awaiting the L1 trigger decision. On a LVL1confirmation the stored data are4290
digitized, multiplexed and transferred to the ROD.
A particularly demanding aspect of the centroid-finding system is to achieve a intercalibration
– 161 –
of < 0.5% between nearby channels. This requires high stability ofthe front-end amplifiers andsmall leakage in the analog storage system, as well as an appropriate calibration system.
Figure 114.The CSC readout architecture.
Signal processing in the ASD, crucial4295
for the system performance, is as follows:subsequent to the charge-sensing preamplifierthe pulse-shaping stage has a peaking timeof 70 ns with a semi-Gaussian shaping cir-cuit to reject low-frequency noise providing4300
a prompt return of the signal to the baseline.The choice of these parameters was based onan assumed maximum rate of about 600 kHzper strip at full LHC luminosity, includingthe usual safty factor of 5. Preamplifier and4305
Shaper are realized as monolythics with 24channels per chip. The 0.5µm CMOS tech-nology (Agilent) was used because of knownreliability and high radiation tolerance.
Each CSC module has 768 readout strips4310
in the precision (bending) and 192 in thetransverse direction, so 32 and 8 chips (24channels) are needed for the readout of bothdirections, respectively. A group of four chipsis mounted on a PC-board (ASM I), such that the 40 chips of a chamber are served by 10 ASM4315
I boards, which are mounted on the lateral frame of the CSC, see Fig. 112. Two ASM I boardsin turn are served by one ASM II board which sends the signals to the DAQ via optical fibers. Abreakdown of components used in the CSC readout is given in Table 36.
An additional constraint for the construction of the readout system came from the fact thatnatural air convection is practically absent around the CSC’s, and liquid cooling had to be provided4320
to evacuate the dissipated heat.
6.4.5 Performance of the CSC
The CSC has been tested under conditions as close as possibleto those in the final experiment. Forthis the Gamma Irradiation Facility at CERN has been used which was built in order to test trackingchambers in a high energy particle beam in the presence of an ajustable high intensityγ-source. In4325
any tested module of four CSC planes two layers were used as reference for the track location whilethe other two were considered the devices under test. The intensity of the Gamma backgroundcould be adjusted up to 2000 Hz/cm2. The results are shown in Fig. 115. Compared to zerobackground, the resolution degrades from 45µm to 65µm at 1 kHz/cm2, while the inefficiency ofa single CSC layer increases from 4% to 10%. The 10% inefficiency at 1 kHz/cm2 is acceptable,4330
as the probability for two or more layers to be inefficient in the same event is very low. At higherhit rates, however, the reconstruction efficiency for tracksegments may become marginal, and anupgrade of the CSC system to 2×4 layers, as was foreseen in the original proposal, may become a
– 162 –
necessity.
Figure 115.CSC resolution and efficiency in a high rate test.
6.5 Trigger chambers4335
The trigger chambers of the muon system provide fast information on muon tracks traversing thedetector, allowing the L1 trigger logic to recognise their multiplicity and approximate energy range.The main requirements for the trigger system are:
• discrimination on muon transverse momentum
• bunch crossing identification4340
• fast and coarse tracking information to be used in the higherlevel trigger stages
• second coordinate measurement in the non-bendingφ−projection to complement the MDTmeasurement, see section 6.2.
• robustness towards random hits due ton/γ−background in the experimental hall
The trigger detectors must provide acceptance in the range|η | ≤ 2.4 and over the fullφ-range. This4345
poses a considerable challenge to the design of the trigger system as resolution requirements inbarrel and end-cap are quite different, an obvious reason being that muon momenta, correspondingto a givenpT , are strongly increasing withη . At |η |= 2.4, for example, p is about 5.8 times largerthan pT , while the integrated bending power is only about twice the value as atη = 0. This leadsto the necessity of an increased andη -dependent granularity in the end-cap trigger system, if the4350
pT -resolution is to match the one in the barrel. The fact that the three trigger layers in the end-capare outside the magnetic field, seeing no curvature, and thattheir respective distances are smallerthan the ones in the barrel (Fig. 116) also calls for a finer granularity of the end-cap trigger readout.
Another difficulty for end-cap triggering comes from the strong inhomogeneities of the mag-netic field in the region 1.3 ≤ η ≤ 1.65 as can be seen in Fig. 12. In this ”transition region” the4355
superposition of the fields of barrel and end-cap toroid leads to a complex field geometry withlarge field components inφ (the ’non-bending’ plane) and strong inhomogenieties of the integrated
– 163 –
bending power, which in two locations inη /φ is close to zero. In this angular region all tracksare nearly straight, similar to tracks with very high momentum. In order to avoid high fake triggerrates, this region must be excluded from the trigger by a masking algorithm, which again calls for4360
a fine readout granularity to keep the resulting trigger losses to a minimum.
RPC1
RPC2
RPC3BOL
BML
BIL
EIL
EML
TGC1TGC2
TGC3
Barrel Toroid
End-cap Toroid
Shielding
high-p T
low-p T
low-p T
high-p T
0 5 10 15 m
10
5
Figure 116.Schematics of the muon trigger system.RPC2 and TGC3 are the reference planes (’pivot’)for barrel and end-cap, respectively.
Taking these constraints into account,two different technologies have been selectedfor barrel (|η | ≤ 1.05) and end-cap (1.05≤|η | ≤ 2.4). In the barrel Resistive Plate Cham-4365
bers (RPC’s) are used due to good spatial andtime resolution as well as adequate rate ca-pability. RPC’s have no wires, which largelysimplifies constructions and makes chambersinsensitive to small deviations from planarity.4370
Being located in the comparatively homoge-neous field of the barrel toroid and havingsufficient spacing between the three triggerlayers (see Table 37), RPC’s give sufficienttrigger selectivity even with moderate chan-4375
nel count, i.e. spatial resolution.In the end-cap region Thin Gap Wire
Chambers (TGC’s) have been selected, op-erating as multi-wire proportional chambers,providing good time resolution and high rate capability. Their spatial resolution is mainly deter-4380
mined by the readout channel granularity which can be adjusted to the needs by wire ganging.TGC’s have demonstrated a high level of reliability and robustness in previous experiments.
To reduce the probability of accidental triggers, caused byrandom combinations of convertedγs, the coincidence condition in both types of trigger chambers is established separately in theη -andφ-projection, a valid trigger requiring a coincidence of both. This also suppresses fake triggers4385
from ”curling” tracks, i.e. multi-MeV electrons fromγ-conversions, spiralling in the magneticfield, potentially creating correlated hits in the trigger chambers. In barrel and end-cap the threelayers of trigger chambers are implemented as shown in Fig. 116. In the barrel two layers areplaced on top and below the MDT’s of the middle layer, while the third one is located close to theouter MDT layer (RPC1-RPC3). In the end-cap the three layersare in front and behind the second4390
MDT wheel (”Big wheel”). The trigger information is generated by a system of fast coincidencesbetween the three chambers along the trajectory of the muon particle. Each coincidence patterncorresponds to a certain deviation from straightness, i.e.curvature of the track, which is usedas criterion for the track to have passed a predefined momentum threshold. The ”deviation fromstraightness” is the deviation of the slope of the track segment between two trigger chambers from4395
the slope of a straight line between the interaction point and the hit in a reference layer called ”pivotplane”, which is the second layer in the barrel (RPC2) and thelast layer in the end-cap (TGC3), asillustrated in the figure. For the low (high)-pT trigger in the barrel, for example, the slope betweenRPC2 and RPC1 (RPC3) is compared to the slope between the interaction point and RPC2.
To assure full acceptance down to the low-momentum limit, the trigger chambers have regions4400
– 164 –
Figure 117.Cross section through the upper part of the barrel; RPC’s arecoloured or cross-hatched.
of overlap with adjacent chambers and between the barrel andend-cap regions. As this may causedouble counting of tracks leading to fake two-muon triggers, algorithms are in place as part of thetrigger logics to treat these overlap regions, either within the barrel or end-cap trigger logic, or asa part of the muon interface to the central trigger processor7.2.2.3. Details on the coincidencesystem and readout logics are given in [163] and [164].4405
6.6 Resistive Plate Chambers
The trigger system in the barrel consists of three cylindrical layers around the longitudinal axis,referred to as the three trigger stations. Fig. 117 shows a standard barrel sector and the locationof the RPC’s (shaded) relative to the MDT’s. The large lever arm between inner and outer RPCallows to trigger selectively on high momentum tracks in therange 9-35 GeV (high-pT trigger),4410
while the two inner chambers provide the low-pT trigger in the range 6-9 GeV. Chamber countand radial position of the RPC’s are given in Table 37. Each station consists of two independentdetector layers, measuringη andφ each. A track going through all three stations thus deliverssixmesurements inη andφ, subsets of four of them being available for the low- and high-pT triggerrespectively. This redundancy in the track measurement allows the use of a 3-out-of-4 coincidence4415
for the trigger in both projections, which rejects fake tracks from noise hits and greatly improvesthe trigger efficiency in the presence of small chamber inefficiencies.
Naming of the RPC’s is identical to the one in the MDT’s, a RPC in a small sector of the middlelayer thus being called a BMS. To denote a RPC/MDT pair in the outer layer the term ’station’ isused, while the RPC/MDT/RPC packages in the middle layer aresometimes called ’superstations’.4420
– 165 –
small sector large sector
Name units chamb’sr locat. to pivot max. z units chamb’sr locat. to pivot max. z
RPC 1 148 84 7820 545 9362 149 94 6800 678 9147
RPC 2 148 84 8365 9362 149 94 7478 9147
RPC 3 176 92 10229 1864 12847 192 96 9832 2354 12267
472 260 490 284
Table 37. Segmentation of the RPC system in 544 chambers and 962 units (126 chambers are made fromonly one unit). Col. 4 and 9 give the radial positions of the RPC stations (in mm), while col. 5 and 10 givethe distance to the pivot station (RPC 2), relevant for the pT resolution. Details for the additional 78 ’special’RPC’s which do not have a MDT partner, are not listed.
6.6.1 Principle of operation
The RPC is a gaseous parallel-plate detector without wires.Two resistive plates, made from plasticlaminate15, are kept parallel to each other at a distance of 2 mm by insulating spacers. The electricfield between the plates of about 4.9 kV/mm, allows avalanches to form along the ionising tracks4425
towards the anode. The signal is read out via capacitative coupling to metallic strips, perpendicularto each other, on both sides of the detector, i.e. on the outerface of the resistive plates. Thegas used is a mixture of C2H2F4/Iso-C4H10/SF6 which combines relatively low operating voltage,non-flammability and low cost, while providing a sufficient plateau for safe operation. The mainoperating parameters of the RPC are given in Table 38.4430
RPC’s can be operated both in avalanche and streamer mode. Inthe high background envi-ronment encountered at the LHC, the avalanche mode offers the benefit of higher rate capabilityand rate-independent time resolution and has therefore been selected as our operation mode. At thenominal operating voltage of 9.7 kV a signal with a width of about 5 ns is generated by the trackwith a streamer probablity of less than 1%.4435
The small jitter of the RPC signal in avalanche mode with respect to the passage of the particleis due to the primary electrons not having to drift to a regionof amplification, as is the case inall types of wire chambers. In the strong and uniform electric field inside a RPC cell all primaryelectron clusters form avalanches simultaneously, producing onesingle signal instantaneously afterthe passage of the particle. The charge multiplication in each avalanche continues until its arrival4440
at the anode plane and, therefore, the gas gain of each avalanche depends on the distance of theprimary cluster from the anode. The total signal charge is thus dominated by the few clustersproduced at the largest distances from the anode. As a consequence, the charge distribution inRPC’s is wider than in wire detectors where all primary clusters are approximately amplified withthe same gain.4445
15phenolic-melaminic plasic laminate
– 166 –
E-field in gap 4.9 kV
Gas gap 2 mm
Gas mixture C2H2F4/C4H10/SF6
(94.7/5/0.3)
Readout pitch ofη - andφ-strips 23–30 mm
Detection efficiency per layer ≥98.5%
Eff. incl. spacers and frames ≥97%
Intrinsic time jitter ≤2 ns
Jitter incl. strip propag. time ≤10 ns
Rate capability ∼1 kHz/cm2
Streamer probability ≤1%
Table 38.RPC chamber parameters and performance.
6.6.2 Mechanical structure
A RPC trigger chamber is made of two rectangular detectors, contiguous to each other, called”units”. Each unit has two independent resistive plate structures, called ”gas volumes”, whichare each read out by two orthogonal groups of pick-up strips.A track traversing a chamber isthus measured twice in each projectionη andφ. The trigger logic uses both projections, where a4450
minimum of 3 out of the 4 possible signals is required to definea valid track.
Figure 118. Cross section through a RPC, where two units are joint to forma chamber. Each unit has twogas volumes (red arrows), supported by spacers (blue), fourresistive electrodes (white) and four readoutplanes, reading the transverse and longitudinal direction. The sandwich structure, hashed, is made frompaper honeycomb.φ-strips are in the drawing plane,η strips perpendicular to it.
The structure of the gas volumes (i.e. 1/2 unit) is identicalfor all RPC’s (Fig. 118): tworesistive plates (plasic laminate, 2 mm) with a volume resistivity of 1010 Ωcm, enclose a gas gapof 2 mm, the correct distance of which being assured by a series of insulating spacers. The outside
– 167 –
surface of the resistive plates is coated with a thin layer ofgraphite paint (100 kΩ per sq.) to assure4455
the HV and ground connection of the resistive electrodes. The graphite electrodes are insulatedfrom the pick-up strips by means of PET films (200µm), glued to the graphite surfaces. Thepick-up strips, outside the PET layers, are bonded on polystyrene plates (3 mm) and connected toamplifiers. The outside surface of the polystyrene plates carries an aluminum sheet for grounding.It should be noted that a significant fraction of the primary signal, i.e. the discharge in the gas gap,4460
is available on the pick-up strips for readout, inspite of the signal having to propagate through thegraphite layer. The reason for this is the limited conductivity of the graphite layer as well as thefast rise time of the primary signal.
Each RPC unit is thus made of two detector layers (i.e. gas volumes) and four readout strippanels. The detector layers are interleaved with three support panels from light-weight paper hon-4465
eycomb16 (40 kg/m3) and are held in position by a solid frame of aluminum profileswhich givesthe required stiffness to the chamber. The total thickness of a RPC unit with two gas volumes,support panels and aluminum covers is 96 mm. The two units forming a chamber have an overlapregion of 65 mm to avoid dead areas for curved tracks. For technical reasons the gas volumes ofthe RPC units are segmented in the longitudinal direction bya narrow (8 mm) gap, which creates4470
a small region of inefficiency at the centre of the unit (inφ-direction). Thus, each RPC unit hasfour and each chamber eight independent gas volumes. The readout strips in the longitudinal di-rection are also segmented in two at the centre of the unit, each half-strip being separately readout at the narrow edge of the unit. This reduces the time jitter of the signals to 8–12 ns as well asthe capacitance seen by the amplifier. Detailed informationon the assembly procedure is available4475
in [49].
All standard RPC’s, as listed in Table 37, are assembled together with a MDT of equal di-mensions in a common mechanical structure (’common support’). Because of the confined radialspace, the common supports had to form a solid, light-weightstructure, holding the two chambertypes with minimum clearance and avoiding constraining forces among them.4480
There is a number of small RPC chambers (”special RPC’s”), not paired to MDT’s, which areused around the magnet ribs and in the feet region, where MDT’s cannot be installed because oflack of space. RPC’s, requiring less space, are used in theseregions to keep the acceptance loss ata minimum.
6.6.3 Signal path and readout electronics4485
A RPC operating in the avalanche mode produces signals of 5 nsFWHM with a time jitter of 1.5 ns.To preserve this high inherent precision the pick-up stripsmust be high quality transmission lineswith low attenuation, terminated at both ends with the characteristic impedance.
The layout of a readout strip plane is shown in Fig. 119. The strips with a width of 25-35 mmconsist of 17µm copper on a 200µm PET foil glued on a 3 mm plate of rigid polystyrene, which4490
is covered, on the outside, by 50µm PET and 17µm copper as ground reference. The strips areseparated by a 2 mm gap with a 0.3 mm thin ground strip at the centre for improved decoupling.This sandwich structure creates a wave resistance of about 25 Ω for the strips, slightly dependingon the width.
16BOL chambers have a reinforced structure using aluminium plates (2 mm) and aluminum honeycomb.
– 168 –
7 mm
3 mm
Pick-Up Strips
Front-End Board
Polystyrene Foam
PET Film
Aluminum GroundedPlane
Figure 119.Layout of a RPC readout strip plane.
R P C 3A S D
R P C 2A S D
R P C 1A S D
h / Fh i g h - p tC M
h / Fl o w - p tC M
O R
a c t i v eS P L I T
h i g h - p tP A D
l o w - p tP A D
S e c t o rL o g i c
R O D
O n d e t e c t o r O f f d e t e c t o r
M U C T P I
a c t i v eS P L I T
t r i g g e r p a t hR / O p a t h
R O B
Figure 120.Scheme of RPC readout logics.
Amplifiers serving eight strips are mounted in such a way as tokeep the insensitive area along4495
the frames of the RPC to a minimum and are therefore soldered directly to the strip boards, withoutthe intermission of a connector pair. While this prevents exchange of faulty amplifiers in the field,the known reliability of GaAs-ASICs seems to keep the risk oflosing readout channels during theexperiment at a minimum.
The frequency response of the GaAs-amplifers has a maximum at 100 MHz, well adapted4500
to the rise time of the chamber signals. The front end electronics input is coupled to the stripswith a transformer integrated in the printed circuit allowing to match the signal polarities ofη -andφ-strips, which are opposite to each other. The voltages defining the thresholds of the com-parators are supplied by external Digital-to-Analog-Converter-units (DACs), located in racks in theexperimental hall controlled via the DCS.4505
As mentioned above, the time jitter due to the strip length islimited by the segmentation ofthe longη -strips to a value of 8-12ns.
The shortφ-strips in the two units of a RPC chamber are ganged together behind the pream-plifier by an OR-gate, such that two physical strips form a logical strip of twice the length. This isdone to avoid unnecessary granularity in theη -direction and to adapt the RPC readout segmenta-4510
tion to the trigger sector segmentation as discussed in section 7.2.2.
In order to be used for the trigger, a signal from a RPC has to becompared with those in thetwo other RPC’s along the path of the particle, i.e. in the same sector and tower, a task whichis accomplished by a system of fast coincidence units close to the chambers. Thus, coincidencesbetween strips in RPC1 and RPC2 are used to create the low-pT trigger. A high-pT trigger requires4515
hits in all three trigger stations: RPC1 and RPC2 must fulfillthe low-pT conditionanda confirminghit must be found in RPC3. The inclusion of the low-pT trigger condition in the high-pT triggerdefinition leads to more robustness against fake triggers compared to a simple coincidence betweenRPC3 and only one inner RPC station. The corresponding logics scheme is shown in Fig. 120. Asfor example RPC1 is used in the low- as well as in the high-pT trigger, the output is duplicated in4520
a fast, active splitter unit to serve both coincidence systems.
– 169 –
In order to form the two trigger types, the coincidence unitsfunction in the following way. Inthe first stage a system of programmable delays is used to compensate for timing differences dueto propagation delays in strips and cables. In a second stage, all signals coming from groups ofstrips in two corresponding RPC’s are put in coincidence to each other, for example, eachη -strip4525
from a region in RPC1 with eachη -strip in the corresponding region of RPC2, forming a matrixof strip-strip coincidences. For each track only one coincidence among the n× n possible oneswill be activated, defining the slope of the track in theη -plane. In a third step the difference of thetrack’s slope to the one of an infinite momentum track is determined and compared to the maximumallowed value, as defined by the trigger threshold. On a positive result a fast brief message is sent,4530
via some intermediate stages to the Muon-to-Central-Trigger-Processor-Interface (MuCTPI), whilemore detailed information, containing, for example, the strip numbers, is prepared for readout tothe Readout Driver (ROD). The coincidence logic is independently implemented for theη andfor the φ projection. Both projections must deliver a valid coincidence in order that the triggercondition be fulfilled. An in-depth discussion of the trigger logic is given in Ref. [163] and [165].4535
6.7 Thin gap chambers
6.7.1 Introduction
2000
4000
6000
8000
10000
12000
6000 8000 10000 12000 14000 16000
I
M1
M2 M3S L
=2.70
=1.05
=2.40
=1.92
Z (mm)
R (
mm
)
low PT
hi PT
pivot plane
end-cap
forward
S L
DL-LL01V01
Figure 121. Location of TGC’s (black) in the innerand middle layer of the end-cap arranged around theMDT’s (cross-hatched). TGC wheel M1 consists oftriplet, M2 and M3 of doublet chambers.
Thin Gap Chambers (TGC’s) provide twofunctions in the end-cap of the muon spec-trometer: the muon trigger capability and the4540
determination of the second, azimuthal coor-dinate to complement the measurement of theMDT’s in the bending direction. The mid-dle layer of the MDT’s in the end-cap (EM-wheel) is complemented by seven layers of4545
TGC’s, while the inner layer (EI-wheel) iscomplemented by only two layers. The az-imuthal coordinate in the outer MDT wheel(EO) is obtained by the extrapolation of thetrack from the middle layer.4550
The radial, bending coordinate is mea-sured by the TGC wire groups, the azimuthalcoordinate by the radial strips. The TGC’sneed good time resolution to tag the beamcrossing with high efficiency (≥ 99%) and4555
fine granularity to provide a sufficiently sharpcutoff in the momentum of the triggeringmuon. To match the granularity to the required momentum resolution, the size of the wire groupsvaries from 4 to 20 as a function ofη , corresponding to a variation in width from 7.2 to 36 mm. Thealignment of wire groups in consecutive layers is staggeredto optimise the position resolution for4560
a given number of electronics channels. The radial strips are staggered in a similar way to achievean azimuthal granularity of 2-3 mrad, seen from the interaction point.
– 170 –
1.8 mm
1.4 mm
1.6 mm G-10
50 µm wire
Pick-up strip
+HV
Graphite layer
Figure 122. TGC structure showing anode wires,graphite cathodes, G-10 layers and a readout strip,orthogonal to the wires.
Figure 121 shows a longitudinal cutthrough the end-cap. TGC’s are located inthe EI-wheel (marked I) and in the EM-wheel4565
(marked M1–M3). The location of the MDTin a small (S) and large sector (L) are shownfor reference.
A detailed, up-to-date description of theTGC chamber system is given in [166] and4570
[164]. For the mechanical construction theMuon-TDR [49], for trigger aspects the L1-TDR [163] can be consulted. A detailed list-ing of all relevant construction parameters isgiven in the TGC parameter book [167].4575
6.7.2 Principle of operation
TGC’s are multi-wire proportional chamberswith the characteristics that the wire-to-wire distance (1.8 mm) is larger than the wire-to-walldistance (1.4 mm), see Fig. 122. With a highly quenching gas mixture of CO2 and n-C5H12 (n-pentane), this cell geometry allows for operation in saturated mode, i.e. with a very high gas gain4580
(∼ 106), which makes pulse heights highly uniform, nearly independent of the amount of primaryionisation.
This provides a number of practical advantages for this application:
• small sensitivity to mechanical deformations; thus a 0.14 mm change in gap size (5%) leadsto only a 15% change in pulse height.4585
• the signal height from a traversing particle shows little dependence on the incident angle upto about 40.
• the pulse height distribution shows very small Landau tails: less than about 2% of the of thepulse height measurements are contained in the tails of the distribution, i.e., have amplitudesmore than 2σ above the mean value of a Gaussian fit. As a consequence, the response to4590
low-energy neutrons (1-10 MeV), which form a large background, is similar to the one ofminimum-ionising particles.
• the highly quenching gas prevents the occurrence of streamers in all operating conditions.
The high electric field around the TGC wires and the small wire-to-wire distance leads tovery good time resolution for the large majority of the tracks. Only tracks at normal incidence4595
passing midway between two wires have much longer drift times due to the vanishing drift fieldin this region. This effect was already discussed in the context of the CSC’s which have a similarcell geometry, see section 6.4.2. In the TGC wheels, however, the angle of incidence for tracksemerging from the interaction point will always be greater than 10, such that a part of the trackwill be outside of the low field region. Including the variation of the propagation time on wires and4600
strips, signals arrive with 99% probablity inside a time window of 25 ns.The main operational parameters of the TGC’s are summarisedin Table 39.
– 171 –
Parameter Design value
Gas gap 2.8± 0.14 mm
Wire pitch 1.8± 0.10 mm
Wire diameter 50 µm
Wire potential 3100± 100 V
Operating plateau600 V
Gas mixture CO2 + n -pentane (55%/45%)
Gas amplification106
Table 39.Main TGC chamber parameters.
6.7.3 Mechanical structure
The seven layers of TGC’s in the middle layer (”Big Wheel”) are arranged in one triplet and twodoublets as shown in Fig. 121.4605
Figure 123. Cross-section of a TGC triplet and doublet.The triplet has three wire layers but only two strip layers.The dimensions of the gas gaps are enlarged with respect tothe other elements.
The trigger detectors, forming cir-cular disks, are mounted in two con-centric rings, an external one, called”outer” or ”end-cap”, covering the ra-pidity range 1.05 ≤ |η | ≤ 1.92 and a4610
central one, called ”inner” or ”forward”,covering the rapidity range 1.92 ≤|η | ≤ 2.4. Figure 123 shows the crosssection of a TGC triplet and doublet.The structure consists of wire planes4615
(anode), cathode planes, strip planes,shields and honeycomb support struc-tures. The cathode planes consist of1.8 mm thick G-10 plates, graphite coated on the inside, i.e.facing the wires (∼ 1 MΩ/sq.) andwith copper cladding on the other side. Two of the copper layers in triplet and doublet are seg-4620
mented into readout strips to read the azimuthal coordinate(marked ’Cu Strips’ in the figure). Thesolid copper layers serve as DC grounds, while the segmentedones, being connected to ampli-fiers, are virtual grounds for the wire signals. On the outside of the triplet and doublet chambersthe honeycomb stiffeners are covered by 0.5 mm thick G-10 plates for rigidity and mechanicalprotection.4625
A gas volume containing a wire plane and two cathodes is called chamber, while the entiretyof three or two chambers in a triplet or doublet is called unit. In the outer ring four or five chambersin triplet and doublets, respectively, are mounted in the way of a ladder forming ”modules”. Amodule covers 7.5 in the outer and 15 in the inner ring, see Fig.to be put later. In the outer ringtwo modules are linked together side by side to form a ”set”, covering 15. Altogether there are4630
– 172 –
EM wheel (”Big wheel”) EI wheel
M1 triplet M2 doubletM3 doublet I doublet total
modularity: Fwd EndcFwd Endc Fwd Endc Fwd Endc
sets/octant 3 3 3 3 3 3
mod’s/set 1 2 1 2 1 2
units/module 1 4 1 5 1 5
chamb’s/unit 3 3 2 2 2 2
units per set 1 8 1 10 1 10
octant 2 24 3 30 3 30
side 24 192 24 240 24 240 24 21 789
system 48 384 48 480 48 480 48 42 1578
chambers per set 3 24 2 20 2 20
octant 9 72 6 60 6 60
side 72 576 48 480 48 480 48 42 1794
system 144 1152 96 960 96 960 96 84 3588
Table 40.TGC modularity. Each wheel consists of an inner (forward) and an outer (end-cap) part, having adifferent azimuthal segmentation. The modules cover 15 in the forward and 7.5 in the end-cap part.
744 units in each of the Big Wheels, corresponding to 1488 chambers. The TGC wheels in theinner wheel (EI chambers) have a slightly different geometrical structure, containing 45 units and90 chambers on each side. A summary of the TGC segmentation isgiven in Table 40.
6.7.4 Signal path and readout electronics
Fig. 124 gives the architecture of the TGC readout. The upperpart shows the coincidence scheme4635
for the seven wire layers and the six strip layers in the EM wheel, the lower part the one for thefour wire and strip layers in the EI wheel. In both TGC wheels,wire and strip signals are processedseparately up to the sector logic. As can be seen, EI/FI information can be used in the triggerdecision in the sector logic.
The data flow, starting with the primary wire and strip signals is as follows. After amplifi-4640
cation in the front-end amplifiers signals are time-alignedand synchronised to the beam crossingfrequency on the ”patch-panel”-boards. In the next station(”slave boards”) the three signals ofthe triplet are combined to form a 2-out-of-3 coincidence, while the four signals of the two EM-doublets form a 3-out-of-4 coincidence. As already mentioned, this provides immunity againstchamber inefficiencies and improves rejection of fake tracks, caused by background or noise hits.4645
Track segments which have passed this early filter are forwarded to the coincidence matrix, which
– 173 –
KTR001V03
3 / 4 Coin. Matrix
triplet
innerdoublet
pivotdoublet
2 / 3 Coincidence
L1B De-rand
L1B De-rand
18
Patch Panel
BID, OR
BID, OR
BID, OR
BID, OR
BID, OR
BID, OR
BID, OR
ASD32
Wire Doublet Slave Board
Wire Triplet Slave Board
R
3 / 4 Coin. Matrix
triplet
innerdoublet
pivotdoublet
L1B De-rand
L1B De-rand
ϕ
20
Patch Panel
BID, OR
BID, OR
BID, OR
BID, OR
BID, OR
BID, OR
Strip Doublet Slave Board
Strip Triplet Slave Board
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
ASD32
wire
strip
18
16
Wire High-pT Board
R, R
ϕ, ϕ
δ
δ
OR
SelectorH-Pt
L
H
Strip High-pT Board
SelectorH-Pt
L
H
SelectorH-Pt
L
H
SelectorH-Pt
L
H
CTP
EI wire
FI wire
EI strip
FI strip
L1B De-rand
4
Patch Panel
BID
BID
BID
BID
ASD32
FI Slave Boards
ASD32
ASD32
ASD32
4OR
OR
L1B De-rand
4
Patch Panel
BID
BID
BID
BID
ASD16
EI Slave Boards
ASD16
ASD32
ASD32
2OR
OR
Hits Selector
Sector Logic
R- ϕ Coin.
MU
CT
PI
TTC
F/E Electronics
Figure 124. The TGC trigger and readout architecture, combining signals from the wire planes of a unit atan early stage.
do a fast search for signal coincidences corresponding to muon tracks in a certain momentum range.From there signals are transferred to the higher levels of the trigger system.
A detailed description of the ATLAS trigger system is given in Section 7 of this article. TheTGC readout is discussed in [164]4650
6.8 Commonalities in the muon system
6.8.1 The gas supplies in the muon system
While gas mixtures and operating conditions in the subsystems of the muon spectrometer are quitedifferent, being adapted to the requirements of the four chamber types, there are also a lot of
– 174 –
commonalities among them from an operational point of view.4655
• The gases must be circulated in a closed loop system (mainly for cost reasons). This requiresa complex recuperation and re-mixing system, using specialised purifiers to clean the gasbefore recirculation into the supply system. For safety andspace reasons all recuperationsystems are housed on the surface, requiring tight pressureand flow control for the transferto the underground experimental area.4660
• Distribution of the gas in the experimental cavern needs a high degree of modularity to beimmune against leaks in any given supply line. Because of theconsiderable hydrostatic pres-sure variations along the height of the hall (22 m), pressureregulation has to be segmentedaccordingly. Therefore, gas distributers (racks) are located at four different levels in the gal-leries of the experiment each one supplying chambers in a±3 m range of relative elevation.4665
• The flow rate in the gas distribution branches must be supervised by a correspondingly largenumber of flow- and pressure meters for control of gas circulation (purity), operating pres-sure and leak detection. Precise pressure control is most important for CSC-, RPC- and TGCchambers, where operation is defined at 1.5 mb above the outside pressure and where signif-icant overpressure might lead to the destruction of the chamber. While this is not the case,4670
of course, for the MDT’s, they also need precise pressure control for reasons of drift timeand gain stability. For the operation of the CSC’s, RPC’s andTGC’s, on the other hand, drifttime and gas gain stability are not as important.
Besides these commonalities there are many differences dueto the particularities of the gases usedin each system. The high flammability of the TGC gas mixture, for example, requires stringent4675
safety precautions, the outside of the TGC’s being constantly flushed with CO2 in a separate, closedgas circuit. The return gas is monitored for traces on n-pentane, which could point to leaks in theprimary distribution system.
The main characteristics of the four gas supply systems are summarised in Table 41. A detailedpresentation of the recirculation and distribution systems is given in Table [49].4680
6.8.2 Electronics services and power consumption
The segmentation of the electrical services of the muon system follows the structure of the detectorwhich is charaterised by a high level of modularity, consisting of electrically independent chambers.The service architecture aims to match this modularity in order to limit data loss in case a givensupply channel should fail. Power supply channels were kept’floating’ to avoid ground loops4685
which might create noise coupling between near-by chambersand would also expose the readoutto noise pick-up from sources in the ATLAS electrical environment. To further prevent pick-up, alllow-voltage (LV) supply cables are shielded by a copper meshenclosing the cable wires and beingconnected only atoneend to the return wire.
For cost reasons, however, the maximum supply modularity could not be maintained. Thus,4690
every two MDT chambers are supplied from one LV supply channel. In contrast, the segmentationof the MDT high voltage (HV) supplies is finer, with each multilayer of three (four) tube layersbeing supplied by a separate channel. Close to the chamber a passive HV-split is installed (’splitter
– 175 –
MDT CSC RPC TGC
gas mixture Ar-CO2 Ar-CO2-CF4 C2H2F4-C4H10-SF6 CO2-n-pentane
composition 93/7 30/50/20 94.7/5/0.3 55/45
gas gain 2×104 4×104 107 106
flammability – – low high
operating pressure (atm) 1 1 1 1
volume at op. press. (m3) 800 1.1 14 16
gas exch./day 1 4 2 2
flow rate at op. pres. (m3/h) 33 0.2 1.2 1.3
Table 41.Main characteristics of the muon gas supply systems.
box’) to supply each tube layer with an independent cable. Incase of shorted tubes (broken wire),the corresponding tube layer can be disconnected, while theother layers would continue to work.4695
The total power consumption of the MDT system is about 30 kW inbarrel and end-cap each.Only one 5 V supply line is needed for each MDT chamber with an average consumption of about50 W.
The CSC supplies have a supply channel per chamber and need more than one voltage forthe front-end electronics. The total power consumption is about 16 kW. The dissipated heat is4700
evacuated by a water cooling system.The RPC’s have separate LV supplies for the analog front-endand the digital electronics,
operating at -5 V and 3.3 V, respectively. A third voltage of -2 V is needed to supply the wired-ORs (’pull down voltage’), combining two aligned physical wires into a ’logical’ wire. A numberof reference voltages is supplied to each chamber to define the discriminator thresholds for the4705
ASD boards. The total power consumption of the RPC system is about 60 kW. HV for the RPC’s(9.7 kV) needs one channel for each gas gap. To limit the required number of primary supplychannels, splitter boxes are used, which are able to safely handle voltages up to 15 kV.
All power supply channels for the MDT barrel, the CSC’s and RPC’s are housed in racks inthe UX15 experimental hall and have been certified for tolerance to radiation and magnetic field.4710
The TGC chambers and the MDT in the big wheel (EO wheel) are supplied from racks directlymounted on the wheel structure (’mini-racks’) in order to limit cable lengths.
The TGC’s, like the RPC’s, have separate LV supplies for analog and digital readout electron-ics. The segmentation of the TGC’s in three sets per octant ismirrored by the supplies.
Services for the small wheel (EI wheel), containing MDT’s, CSC’s and TGC’s, are routed4715
through a flexible support structure (’cable Schlepp’). When the EI wheel is moved along thez-direction to give access to calorimeter and inner detector, most service channels don’t need in-terruption, reducing access time and minimising the risk ofdamage to the small wheel chamberoperation.
– 176 –
7. Trigger, Data Acquisition, and Controls4720
7.1 Introduction to event selection and data acquisition
As described in section 1, the trigger consists of three levels of event selection: Level-1 (L1),Level-2 (L2) and event filter. The L2 and event filter togetherform the High-Level Trigger (HLT).The L1 trigger is implemented using custom electronics, while the HLT is almost entirely based oncommercially available computers and networking hardware. A block diagram of the trigger and4725
data acquisition system is shown in Fig. 125.
Figure 125.Block diagram of the ATLAS trigger and data acquisition systemsystems.
The L1 trigger searches for signatures from high-pT muons, electrons/photons, jets, andτ -leptons decaying into hadrons. It also selects topologies with large missing transverse energy (Emiss
T )and large total transverse energy. It uses reduced-granularity information from a subset of detec-tors: the Resistive Plate Chambers (RPC) and Thin-Gap Chambers (TGC) for high-pT muons, and4730
– 177 –
all the calorimeter sub-systems for electromagnetic clusters, jets,τ -leptons,EmissT and large total
transverse energy. The maximum L1 accept rate, which the detector readout systems can handleis 100 kHz, and the L1 decision time must be less than 2.5µs.
The L2 trigger is seeded by Regions-of-Interest (RoI). These are regions of the detector, wherethe L1 trigger has identified possible trigger objects within the event. The L2 trigger uses RoI infor-4735
mation on coordinates and types of signatures to limit the amount of data that it must transfer fromthe detector readout. It uses dedicated algorithms to analyse full-granularity and full-precision datafrom all of the detectors. The L2 trigger reduces the event rate to below 3.5 kHz, with an averageevent treatment time of approximately 10 ms.
The event filter uses offline analysis procedures on fully-built events to further select events4740
down to a rate, which can be recorded for subsequent offline analysis. It reduces the event rate toapproximately 200 Hz, with an average event treatment time of order one second.
The HLT algorithms use the full granularity and precision ofcalorimeter and muon cham-ber data, as well as the full data from the inner detector, to refine the trigger selections. Betterinformation on energy deposition improves the threshold cuts, while track reconstruction in the in-4745
ner detector significantly enhances the particle identification (for example distinguishing betweenelectrons and photons).
The event selection at both L1 and L2 primarily uses inclusive criteria, for example high-ET
objects above defined thresholds. The one exception is the L2selection of events containing thedecay of a B-hadron, which requires the reconstruction of exclusive decays into particles with low4750
momentum.The data acquisition system receives and buffers the event data from the detector-specific
readout electronics at the L1 trigger rate (up to 100 kHz). The data transmission is performedover point-to-point ReadOut Links (ROL’s). On request, it subsequently moves the detector datacorresponding to RoI’s to the L2 trigger, and, for those events fulfilling the L2 selection criteria,4755
it performs event-building at the L2 trigger accept rate of about 3.5 kHz. The assembled eventsare then moved by the data acquisition system to the event filter, and the events selected there aremoved to permanent event storage.
In addition to controlling movement of data down the triggerselection chain, the data acqui-sition system also provides for the configuration, control and monitoring of the ATLAS detector4760
during data-taking. Supervision of the detector hardware (gas systems, power-supply voltages, etc.)is provided by the Detector Control System (DCS).
Section 7.2 presents the design, algorithms, and implementation of the L1 trigger. The HLTand data acquisition are described in section 7.3, which gives an overview of the flow of eventsthrough the system, a brief description of the main system components, and the algorithm imple-4765
mentation and performance expected for initial operations. The implementation and capabilitiesof the DAQ/HLT are presented in section 7.4. Finally, the Detector Control System is describedin section 7.5.
7.2 The L1 trigger
The L1 trigger makes the initial event selection based on information from the calorimeters and4770
muon detectors. The calorimeter selections are based on reduced-granularity information from allthe calorimeters (electromagnetic and hadronic; barrel, end-cap and forward). The L1 Calorimeter
– 178 –
Trigger (L1Calo) aims to identify high-ET electrons and photons, jets, andτ -leptons decaying intohadrons, as well as events with largeEmiss
T and large total transverse energy. A trigger on the scalarsum of jet transverse energies is also available. For the electron/photon andτ triggers, isolation4775
can be required. The information for each bunch-crossing used in the L1 trigger decision is themultiplicity of hits for a number of programmableET thresholds (between 4 and 16 thresholds) perobject type.
The L1 muon trigger is based on signals seen in the muon trigger chambers: RPC’s in thebarrel and TGC’s in the end-caps. The trigger searches for patterns of hits consistent with high-pT4780
muons originating from the interaction region. The logic provides six independently-programmablepT thresholds. The information for each bunch-crossing used in the L1 trigger decision is themultiplicity of muons for each of the sixpT thresholds.
The overall L1 accept decision is made by the Central TriggerProcessor (CTP) by combiningthe information for different object types. Trigger menus can be programmed with up to 2564785
distinct items, each item being a combination of requirements on the input data. Fig. 126 is ablock diagram of the L1 trigger. The trigger decision, together with the 40.08 MHz clock and othersignals, is distributed to the detector front-end and readout systems via the Timing, Trigger andControl (TTC) system, using an optical-broadcast network to connect up to 1024 destinations toeach source.4790
Calorimeters Muon Detectors
Muon Trigger
Detector Front-ends
Calorimeter Trigger
Timing, Trigger andControl distribution
Central Trigger Processor
JetE.M.Tau
ET
ET
Level-2 Trigger
Regionsof Interest
Figure 126.Block diagram of the L1 trigger.
While the L1 trigger decision itself is based only on the multiplicity of trigger objects (or flagsfor global quantities), information about the geometric location of trigger objects is retained. Uponthe event being accepted by the L1 trigger, both the muon and the calorimeter trigger processorssend this information as RoI to the L2 trigger (see section 7.3.2), where it is used to seed theselection performed by the HLT.4795
– 179 –
An essential requirement on the L1 trigger is that it must uniquely identify the bunch-crossingof interest. The very short (25 ns) bunch-crossing intervalmakes this a challenging task. In the caseof the muon trigger, the physical size of the muon spectrometer implies times-of-flight comparableto the bunch-crossing interval. For the calorimeter trigger, a serious complication is that the widthof the calorimeter signals extends over many bunch-crossings.4800
While the trigger decision is being formed, the informationfor all detector channels has to beretained in pipeline memories. These memories are often contained in custom electronics placedon or near the detector, where radiation levels are high and access is difficult. In the interest ofcost and reliability, it is desirable to keep the pipeline length as short as possible. The L1 latency,which is the time from the proton-proton collision until theL1 trigger decision, must therefore be4805
kept as short as possible. The design of the trigger and front-end systems requires the L1 latencyto be less than 2.5µs, with a target latency of 2.0µs, leaving 0.5µs contingency. About 1µs ofthis time is accounted for by cable-propagation delays alone. To achieve this aim, the L1 trigger isimplemented as a system of purpose-built hardware processors, which are described in more detailbelow.4810
7.2.1 Calorimeter trigger
L1Calo [168] is a pipelined digital system designed to work with about 7200 analogue triggertowers of reduced granularity (0.1 × 0.1 in ∆η × ∆φ) from the electromagnetic and hadroniccalorimeters. It sends the results for each LHC bunch-crossing to the CTP approximately 1.5 µsafter the event occurs, resulting in a total latency for the L1Calo chain of about 2.1 µs, well within4815
the allowed envelope.The L1Calo system is located off-detector in the USA15 side cavern. Its architecture, shown
in Fig. 127, consists of three main sub-systems. The pre-processor digitises the analogue inputsignals, then uses a digital filter to associate them with specific bunch-crossings. Finally, it uses alook-up table to do a pedestal subtraction, fine-tune the transverse-energy calibration, and ignore4820
small noise pulses. The data are then transmitted to both theCluster Processor (CP) and Jet/Energy-sum Processor (JEP) sub-systems in parallel. The CP sub-system identifies electron/photon andτ /hadron candidates withET above the corresponding programmable threshold and satisfying, ifrequired, certain isolation criteria. The JEP receives jetL1Calo trigger elements, which are 0.2×0.2 sums in∆η ×∆φ, and uses these to identify jets and to produce global sums ofscalar and4825
missing transverse energy. Both processors count the multiplicities of the different types of triggerobjects. The CP and JEP send these feature multiplicities, as well as transverse-energy thresholdinformation, to the CTP for every bunch-crossing.
When there is a Level-1 Accept (L1A) decision from the CTP, the stored data from theL1Calo sub-systems is read out to the data acquisition system: this includes input data, inter-4830
mediate calculations and trigger results in order to allow full monitoring and verification of theL1 trigger functionality. These data can also provide useful diagnostics for the LHC machine andthe ATLAS detectors.
The types and positions of jet,τ /hadron and electromagnetic cluster candidates are also col-lected and sent to the RoI builder (see section 7.3.7) for useby the L2 trigger.4835
The L1Calo architecture is relatively compact, with a minimal number of crates and cablelinks. This helps in reducing the latency. A number of multiple-use hardware modules, each
– 180 –
fulfilling several roles in the system, were developed in order to reduce hardware costs and designeffort, as well as reduce the number of spares required.
LAr(em)
Tile/LAr(had)
ET,ET
ET sum
Ex,Ey
ETsums
Declustering
Jets
Counting
Jet-finding
RoIs
FIFO, BCID
Look-up table
BC-mux
~7000 analogue links
Level-1 Central Trigger Processor (CTP)
Level-1Muon
Trigger
OnDetector
InTriggerCavern
Declustering
Cluster Processor
PPMs
9-bit jet elements
10-bit serial links:400 Mbit/s (~10 m)
8-bit trigger towers
JEMs CPMs
Cluster-finding
Jet/Energy Processor
To RODs(Level-2)
CP
RoIs
(e / and h
e / h
To RODs (DAQ)
To RODs (DAQ)
JEP
PPrTo RODs
(DAQ)
2x2 sum
twisted pair, <70 m
Preprocessor10-bit FADC
160 Mbit/sbackplane
80 Mbit/sbackplane
Calorimeters
Analogue Sum
Receiver
CountingCMMs
Figure 127.Architecture of the L1 calorimeter trigger.
7.2.1.1 The analogue front-end. Analogue signals from trigger towers in all the calorimeters4840
are sent to the USA15 cavern on 16-way twisted-pair cables. These cables are specially routed tominimise their length, and hence the trigger latency; they range from about 40 m to 70 m long. The16-pair long trigger cables from the tile calorimeter to theUSA15 cavern carry both the trigger-tower signals for L1Calo and signals from the rear layer, which could be used if needed to helpreduce backgrounds in the L1 muon trigger. The two kinds of signals are separated using patch-4845
panels upstream of the L1Calo receivers.
All of the trigger-tower signals arrive at 64-channel receiver modules. The main function ofthese is to adjust the gains, in order to provide transverse energy rather than energy where necessary,and to compensate for differences in energy calibration andsignal attenuation in the long cables.
– 181 –
The receivers reshape the signals, and contain linear, variable-gain amplifiers controlled by DAC’s.4850
Receiver outputs are sent as differential signals on short twisted-pair cables to the pre-processor. Afurther function of the receivers is to monitor a programmable selection of analogue input signals.
7.2.1.2 The pre-processor The pre-processor consists of eight 9U VMEbus crates. Four cratesprocess electromagnetic trigger towers and four process hadronic towers. Each crate contains six-teen pre-processor modules that can each receive four analogue cables on the front panel and pro-4855
cess sixty-four analogue input signals.
The pre-processor receives and processes analogue trigger-tower signals from the electromag-netic and hadronic calorimeters. The granularity of these signals is reduced compared to thefull calorimeter data. This is done by analogue summing at the detector of variable numbersof calorimeter cells, ranging from a few up to sixty. Differential signals first enter 16-channel4860
analogue-input daughter modules, which convert them to single-ended signals with appropriatebias and gain for digitisation. The main signal processing is performed by sixteen multi-chip mod-ules, each of which processes four trigger towers. Four 10-bit FADC’s digitise the signals witha sampling frequency of 40.08 MHz. Fine adjustment of the timing of each digitisation strobe isperformed by a four-channel ASIC, which provides programmable delays in steps of 1 ns across4865
the 25 ns LHC clock period. The digitised values are then sentto a custom pre-processor ASIC.
The pre-processor ASIC performs a series of tasks. It synchronises the timing of the fourinputs, to compensate for different times-of-flight and signal path-lengths. It then assigns signalsto the correct bunch-crossing, as detailed below. A look-uptable is used to carry out pedestalsubtraction, apply a noise threshold, and do a final transverse-energy calibration, resulting in 8-4870
bit trigger-tower values. Finally, it performs bunch-crossing multiplexing for the CP and it sumsthe four values into 0.2× 0.2 jet elements for the JEP. Two 10-bit LVDS serialisers operatingat 400 Mbit/s transmit the processed trigger-tower data to the CP, while a third serialiser sends thesummed 9-bit jet elements to the JEP.
The pre-processor ASIC reads out data to the data acquisition upon receiving a L1A signal.4875
The readout data are taken from pipeline memories at two stages: the raw digitised values from theFADC’s, and the 8-bit processed trigger-tower data from thelook-up tables. Data from the bunch-crossing of interest, as well as a programmable number of bunch-crossings around it (typically up tofive in all), allow monitoring of pulse shapes at the FADC’s, and of the bunch-crossing identificationand energy calibration at the look-up table outputs. These readout data are sent to a rear-transition4880
module behind the back-plane, where they are serialised andsent to the data acquisition readoutover an optical fibre.
Bunch-crossing identification Trigger pulses from both the electromagnetic and hadroniccalorimeters have widths of several bunch-crossings, so itis essential that trigger-tower signals areassociated with the correct bunch-crossing. This is a crucial requirement not only for normal-size4885
pulses, but also for saturated pulses (above about 250 GeV) and for pulses as small as possible(down to 2-3 GeV if possible). The pre-processor ASIC is capable of identifying a signal’s bunch-crossing using three different methods, which provides ample redundancy for consistency checksduring commissioning.
– 182 –
For normal, unnsaturated signals, a digital, pipelined andfinite-impulse-response filter pro-4890
cesses five consecutive FADC samples, by multiplying the samples with pre-defined coefficientsand summing the resulting values. A subsequent peak-finder attributes the maximum value of thissum to the corresponding bunch-crossing. The working rangeof the method spans from smallsignals (energy depositions of a few GeV) up to the near-saturation level of around 250 GeV.
For saturated signals, two consecutive samples are compared to a low and a high threshold,4895
making use of the finite peaking-time (approximately 50 ns) of an analogue input signal. Thus,detection of a leading edge allows attribution of the virtual peak to a specific bunch-crossing. Thismethod is valid from around 200 GeV up to the maximum energy range of the calorimeters.
A third method, provided for consistency checking, uses comparators with programmablethresholds on the analogue input daughter modules to present a rising-edge signal to the multi-chip4900
modules. Given the known peaking time, bunch-crossing identification can be performed usingan appropriate programmed delay in the pre-processor ASIC.The validity of this method beginswell above the comparator threshold and extends up to the full energy range. There is thus a largeoverlap with the two previous methods, allowing consistency checks between the methods to beperformed during commissioning.4905
The finite-impulse-response-filter output is presented to the look-up table to extract a cali-bratedET value for the trigger tower. If the bunch-crossing identification criteria are met, thisvalue is sent to the CP outputs. In the case of saturation, thetower is assigned the maximum 8-bitvalue of 255 GeV. For the JEP outputs, any 0.2×0.2-sum jet element, which contains a saturatedtrigger tower or which has a 9-bit sum in overflow, is assignedthe maximum 9-bit value of 512 GeV.4910
A tower or jet element with a maximum value is understood to besaturated by the CP and/or JEPsub-systems. Any event where a saturation condition occurswill produce a L1A signal, and theRoI information sent to the L2 trigger will be flagged by saturation bits. All data words sent tothe CP and JEP systems are also accompanied by an odd-parity bit for error detection.
Bunch-crossing multiplexing The data from the pre-processor module consist of four 8-bit4915
trigger towers per multi-chip module to the CP, and one 9-bit2× 2 sum jet element to the JEP.To economise on the number of links needed, it was noted that the bunch-crossing identificationalgorithm is essentially a peak-finding scheme. This means that an occupied bunch-crossing willalways be followed by an empty (zero) one. This allows two trigger towers being sent to the CPto share a single serial link. For the JEP, where a sum of four towers is transmitted, this does not4920
apply.
Trigger towers are paired at the pre-processor ASIC output stage. When a tower has a non-zero value its value is sent to the link, along with a flag bit that indicates which of the two towersis being transmitted first. On the next bunch-crossing, a value (or zero) for the other tower of thepair is sent to the link, again with a flag bit. In this case, theflag bit is used to indicate whether4925
the second tower’s value belongs to the same bunch-crossingas that of the first tower or to thefollowing bunch-crossing. This scheme is called bunch-crossing multiplexing. By using bunch-crossing multiplexing, data transmission to the CP sub-system is achieved with only two links permulti-chip module instead of four.
Output signal fan-out and pre-compensation The pre-processor multi-chip module sends4930
– 183 –
its outputs to the CP and JEP on 10-bit, 400 Mbit/s electricalLVDS serial links. These are routedthrough a daughter-card on the pre-processor module, wherethe required fan-out is performed.This fan-out is needed to provide the trigger algorithms with overlapping data between detectorquadrants in azimuth. The data pass through the back-plane to 11 m long shielded parallel-pair ca-bles. Measures had to be taken to improve signal-driving capabilities, since observed signal attenu-4935
ation and distortion from the cables may compromise data integrity. Using a RC pre-compensationnetwork to compensate for cable signal losses, bit-error rates of less than 10−14 have been achievedover 15 m of cable, which meets the design requirements.
7.2.1.3 The cluster and jet/energy-sum processorsThe CP and JEP sub-systems share manyarchitectural features and some common hardware. The jet algorithm in the JEP and the elec-4940
tron/photon andτ /hadron cluster algorithms in the CP both perform feature searches in overlap-ping, sliding windows. Therefore, a large amount of data duplication between processor modulesis required. Both sub-systems divide the calorimeters intofour azimuthal quadrants, with eachprocessor module within a quadrant covering a slice in pseudorapidity and 90 in azimuth. Over-lapping data from neighbouring azimuthal quadrants are provided by duplicated serial links from4945
the pre-processor. Within each quadrant, modules only needto share input data with their nearestneighbours, over short (roughly 2 cm) point-to-point back-plane links. This architecture minimisesthe number of cable links from the pre-processor, and the back-plane fan-out is simplified. Signaltiming is also fairly straightforward and the overall system is quite compact.
The CP is a four-crate system, with fourteen Cluster Processor Modules (CPM’s) in each crate4950
covering one calorimeter quadrant. The JEP is contained in two crates, each containing 8 Jet EnergyModules (JEM’s) from two opposing quadrants (16 JEM’s total) in each crate. Results from theprocessor modules are brought to two Common Merger Modules (CMM’s) in each crate: thesesum the data to produce crate-level results. The CMM’s also perform the system-level summationof data from the different crates, and transmit the final results to the CTP.4955
The electron/photon andτ /hadron triggers extend out to|η | < 2.5, which is the fiducial limitfor precision measurements with the inner detector and electromagnetic calorimetry. The jet trig-ger extends out to|η | < 3.2. TheEmiss
T and total transverse-energy triggers include the forwardcalorimetry, in particular to provide adequateEmiss
T performance, which means that they extendto |η | < 4.9. This implementation also allows the FCal to be used for forward-jet triggers.4960
7.2.1.4 The cluster processor module The electron/photon trigger algorithm [169], shown in Fig.128,identifies 2× 2 clusters of trigger towers, in which at least one of the fourpossible two-towersums (1×2 or 2×1) of nearest-neighbour electromagnetic towers exceeds a pre-defined threshold.Isolation-veto thresholds are set for the 12-tower surrounding ring in the electromagnetic calorime-ter, as well as for the 2×2 hadronic-tower sum behind the cluster and the 12-tower hadronic ring4965
around it. All these thresholds are programmable.
Theτ /hadron algorithm uses the same basic elements to select narrow hadronic jets. Each ofthe four possible two-tower sums (1×2 or 2×1) of nearest-neighbour electromagnetic towers isadded to the 2×2 hadronic-tower sum directly behind and the result is compared to a pre-definedthreshold. Isolation veto thresholds are set separately for each of the surrounding 12-tower rings in4970
both the electromagnetic and hadronic calorimeters.
– 184 –
The isolation thresholds for both algorithms are fixed values, rather than ratios of isolationenergy to cluster energy. This simpler approach was chosen on the basis of physics studies, whichshowed that the expected isolation sums are relatively insensitive to shower energies. In practice,high-energy clusters will generally have looser isolationcriteria to maximise the efficiency for4975
possible low-rate exotic signal processes, while lower-energy clusters will have stricter isolationcriteria in order to minimise the rates at the expense of a limited loss of signal.
These algorithms are run over all possible 4×4 windows, which means that the windows over-lap and slide by steps of 0.1 size in bothη andφ. This implies that an electron/photon orτ /hadroncluster can satisfy the algorithm in two or more neighbouring windows. Multiple-counting of clus-4980
ters is avoided by requiring the sum of the four central electromagnetic plus the sum of the fourcentral hadronic towers to be a local maximum with respect toits eight nearest overlapping neigh-bours. In order to avoid problems in comparing digital sums with identical values, four of theeight comparisons are ‘greater than’ while the other four are ‘greater than or equal to’, as shownin Fig. 129. The location of this 2× 2 local maximum also defines the coordinates of the elec-4985
tron/photon orτ /hadron RoI.
The CPM identifies and counts clusters satisfying sets of threshold and isolation criteria. Eightthreshold sets are reserved for electron/photon triggers,while eight further threshold sets can eachbe used for either electron/photon orτ /hadron triggers. Each CPM receives and deserialises inputdata on eighty LVDS cables from the pre-processor modules, brought in to the rear of the module4990
through back-plane connectors. These data correspond to 4× 20 electromagnetic and hadronictrigger towers inη andφ, respectively. Twenty serialiser FPGA’s receive the data and multiplexthem to 160 MHz. The data are then shared between neighbouring modules via the back-plane,and finally fanned out to eight CP FPGA’s, which perform the clustering algorithms. The serialiserFPGA’s also store the input data in pipelines for eventual readout to data acquisition upon reception4995
of a L1A signal.
The eight CP FPGA’s each service eight overlapping 4×4 windows. Pipelines implementedin each one of them save output data for readout to the data acquisition and also save cluster typesand coordinates for readout as RoI’s to the L2 trigger.
Two hit-multiplicity FPGA’s collect and sum the cluster multiplicities from the CP FPGA’s, for5000
reporting to the crate-level merging of CP results. These multiplicities are transmitted via the back-plane as two data streams, each containing 3-bit multiplicities for eight of the 16 cluster sets, plusone bit of odd parity. If more than seven instances of a cluster type are identified, the multiplicityis reported as seven.
Two additional Readout Controller (ROC) FPGA’s collect input data from the serialiser FPGA’s,5005
RoI data from the CP FPGA’s, and output data from the hit-multiplicity FPGA’s upon reception ofa L1A signal, and transmit them to readout driver modules serving the data acquisition and the L2trigger on two optical fibres from the front panel of the module.
7.2.1.5 The Jet/Energy module (JEM) The JEM works with jet elements that are the sum of2×2 trigger towers in the electromagnetic calorimeters addedto 2×2 trigger towers in the hadronic5010
calorimeters. The jet algorithm identifiesET sums within overlapping windows consisting of 2×2,3×3 or 4×4 jet elements, corresponding to window sizes of 0.4, 0.6 or 0.8 inη andφ, as shown inFig. 130, and these sums are then compared to pre-defined jet energy thresholds. Multiple-counting
– 185 –
Vertical SumsΣ
Σ Horizontal Sums
Σ Σ
Σ
Σ
Electromagneticisolation < e.m.isolation threshold
Hadronic isolation< inner & outerisolation thresholds
Electromagneticcalorimeter
Hadroniccalorimeter
Trigger towers (∆η × ∆φ= 0.1 × 0.1)
De-cluster/RoI region:local maximum
Electron/photon Algorithm
Tau/hadron algorithm adds 1x2/2x1 e.m. clustersto 2x2 inner hadronic region. Isolation usese.m. and outer hadronic ‘rings’.
Figure 128. Electron/photon andτ /hadron trigger algo-rithms.
R
>
>
>
>
Figure 129. ET local-maximum testfor a cluster/RoI candidate. Theη -axis runs from left to right, and theφ axis from bottom to top.
of jet candidates is avoided by requiring the window to surround a 2×2 cluster whose sum is a localmaximum, with the same definition as for electron/photon andτ /hadron clusters. The location of5015
this 2×2 local maximum also defines the coordinates of the jet RoI. All three jet window sizes arecalculated by default, and eight independent sets of jetET threshold and window size combinationsare available for trigger menus.
The energy-summation algorithm produces sums ofET , Ex andEy, and uses the system-levelsums of these to report on four total-ET and eightEmiss
T thresholds to the CTP.5020
Each JEM receives and deserialises data from 88 LVDS links, corresponding to 4× 11 jetelements inη andφ, for both the electromagnetic and hadronic calorimeters. Four input FPGA’sreceive the data, sum the electromagnetic and hadronic parts of each jet element to 10-bit values,and multiplex these sums to 80 MHz for fan-out to the main processor units on the same and neigh-bouring modules. Pipelines in each input FPGA save input data for readout to the data acquisition5025
upon reception of a L1A signal.
The jet and energy-summation algorithms are implemented intwo large main-processor FPGA’sper JEM. The main processors are also responsible for reporting results to the crate-level merging,as well as pipelining of data acquisition and RoI information for readout. The jet output of eachJEM is a data stream consisting of eight 3-bit jet multiplicities and one odd-parity bit. The energy5030
– 186 –
output is also a data stream containing the values ofEx and Ey, each compressed from 12 bitsto a 8-bit quad-linear scale (6-bit mantissa plus two multiplier bits). The energy output is alsoaccompanied by an odd parity bit.
A single readout-controller FPGA collects input data from the input FPGA’s, and output andRoI data from the main processor FPGA’s, for readout to readout driver modules serving data5035
acquisitionand the L2 trigger on two optical fibres from the front panel of the module.
Window 0.6 x 0.6 Window 0.8 x 0.8
De-cluster/RoI can bein 4 possible positions De-cluster/RoI must
be in centre position(to avoid 6x6, and 2 jets/window)
Window 0.4 x 0.4
Figure 130. Jet trigger algorithms, based on 0.2× 0.2 jet elements and showing RoI’s (shaded). In the0.6×0.6 case, there are four possible windows containing a given RoI.
7.2.1.6 The common merger module (CMM) Two modules in each CP and JEP crate carry outcrate-level merging of results received from the crate’s processor modules. In the CP crates, eachmerger module is responsible for calculating 3-bit clustermultiplicities for eight of the 16 elec-tron/photon andτ /hadron cluster definitions. In the JEP crates, one merger module produces 3-5040
bit multiplicities for the eight jet definitions, while the other produces sums ofET , Ex and Ey.Each CMM receives data from the crate’s 14 CPM’s or 16 JEM’s, over point-to-point links on thecrate back-plane.
The CMM carries out all of these merging functions by using different firmware versions.Each CMM receives up to 400 bits of data per bunch-crossing from the crate’s 14 CPM’s or5045
16 JEM’s. A large FPGA performs crate-level merging. Parallel LVDS cable links between thesub-system crates bring all crate-level results to one CMM of each type, which is designated as thesystem-merger CMM. A second FPGA on the CMM carries out the system-level merging.
7.2.1.7 The processor back-plane Physically, the processor back-plane is a monolithic back-plane of 9U height and hosts up to 21 single-width modules. Itis populated almost entirely with5050
2 mm hard-metric connectors, with 1148 signal and ground pins in each JEM/CPM and CMM po-sition. Two signal layers provide point-to-point links between neighbouring processor modules forinput data fan-in/fan-out. A CANbus on one of the layers provides an interface to the DCS. An-other four layers provide connections from each CPM or JEM tothe two CMM’s at the right andleft of the processor modules. The remaining two layers distribute TTC signals from the timing-5055
and-control module. To conserve pins, the VMEbus is reducedto only the 43 signal lines neededto provide A24D16 slave access to the modules.
– 187 –
The LVDS serial-input and merger-interconnect cables are connected to the rear of the pro-cessor back-plane and passed through it to the modules in front. This results in a system with fewcables on the front panels of the modules, with as a consequence fewer recabling errors and less5060
cable damage over the lifetime of the experiment. The crate also provides extended geographicaddressing. Upon power-up, address pins tell each module its location in the crate, as well as thecrate’s location in the system. The modules use this information to set unique VMEbus, CANbus,and TTC addresses. Modules with multiple FPGA configurations can also use this information toautomatically load the appropriate configuration.5065
7.2.1.8 The Read Out Driver (ROD) The trigger system has two separate readout systems.Input, output, and some intermediate data from each module are read out to the data acquisitionsystem and at the same time the CP and JEP sub-systems report feature types and coordinates asRoI data to the L2 trigger.
The readout system has been designed to handle one bunch-crossing of RoI data and up to5070
five bunch-crossings of data acquisition data per event at a L1A rate of up to 100 kHz. A commonapproach has been adopted in all L1Calo sub-systems for dataacquisition and RoI readout.
On each module to be read out, readout FIFO’s on each processor FPGA or ASIC are readout as 40.08 MHz serial streams to a readout controller FPGA for timing alignment. This passesthe serial streams in parallel to the inputs of a G-link transmitter, which transmits them serially at5075
800 Mbit/s over optical fibres to a ROD.
A common ROD module is used by both the data acquisition and RoI readout sub-systemsto gather and report data from the pre-processor modules, CPM’s, JEM’s, and CMM’s, using dif-ferent firmware configurations in five large input FPGA’s for different readout tasks and modules.The ROD is a 9U-module residing in a standard VME64x crate. Ithas 18 G-link receivers, which5080
pass their parallel outputs to the FPGA’s for data compression, zero suppression, and some datamonitoring. The ROD also contains four transmitters on a rear-transition module for passing com-pressed event data to the data acquisition and RoI readout buffers. Header generation and routing ofdata to the different outputs is carried out by a switch controller FPGA, whose settings depend onthe type and source of data being read out. In addition, a further large FPGA provides monitoring5085
capability on a sample of readout data.
7.2.2 Muon trigger
The L1 muon trigger is based on dedicated finely segmented detectors (the RPC’s in the barrel andthe TGC’s in the end-caps, as described in detail in section 6.5) with a good enough timing accuracyto provide unambiguous identification of the bunch-crossing containing the muon candidate.5090
The trigger in both the barrel and the end-cap regions is based on three trigger stations. Thebasic principle of the algorithm is to require a coincidenceof hits in the different chamber layerswithin a road, which tracks the path of a muon from the interaction point through the detector.The width of the road is related to thepT -threshold to be applied. A system of programmablecoincidence logic allows concurrent operation with a totalof six thresholds, three associated with5095
the low-pT trigger (threshold range approximately 6-9 GeV) and three associated with the high-pT
trigger (threshold range approximately 9-35 GeV). The trigger signals from the barrel and the muon
– 188 –
end-cap trigger are combined into one set of six threshold multiplicities for each bunch-crossing inthe Muon to CTP Interface (MUCTPI), before being passed on tothe CTP itself.
PAD
ROI
Barrel Large
Sector
Barrel Small
Sector∆η=0.1
∆η=0.2
η=0
Trigger Large Sector (SL)
Trigger Large Sector (SL)
Trigger Small Sector (SL)
Trigger Small Sector (SL)∆φ=0.1
∆φ=0.2
φ
CMη
CM
∆η=0.2
∆φ=0.1
∆η=0.1
∆φ=0.2
Figure 131.Schema (left) and segmentation (right) of the L1 muon barreltrigger.
7.2.2.1 Muon barrel trigger5100
Trigger signals The muon trigger for the barrel regions [170] makes use of dedicated RPCdetectors (see chapter 6). The RPC is a gaseous detector providing a typical space-time resolutionof 1 cm× 1 ns and a rate capability of about 1 kHz/cm2. As shown on the left side of Fig. 131,the RPC’s are arranged in three stations. The two Barrel Middle (BM) stations, RPC1 and RPC2,are arranged on either side of the MDT BM stations at approximately 7.8 m and 8.3 m radial5105
distance from the interaction point, respectively. The RPC3 Barrel Outer (BO) station, mounted onthe inside (large sectors) or outside (small sectors) of theMDT BO stations, is located at a radialdistance of about 10.2 m. Each station is made of two RPC doublets, one in theη projection andone in theφ projection. Both planes are used in the trigger. Theη -strips are parallel to the MDTwires and provide the bending view of the trigger detector. Theφ-strips are orthogonal to the MDT5110
wires and provide the second coordinate measurement. Theφ-strips are also needed for the patternrecognition. The RPC’s are organised in several modules, and their dimensions have been chosento match those of the corresponding MDT chambers. In most stations the RPC’s are composed oftwo units along both the azimuthal and the beam direction. Toavoid dead areas between adjacentunits, the active zones of neighboring RPC’s are partially overlapped inη .5115
Trigger algorithm The trigger algorithm operates in the following way: if a track hit isgenerated in the second RPC doublet (the pivot plane), a search for the same track is made in thefirst RPC doublet, within a road whose centre is defined by the line of conjunction of the hit in thepivot plane with the interaction point. The width of the coneis a function of the desired cut onpT :the smaller the cone, the higher the cut onpT . The system is designed so that three such low-pT5120
thresholds in each projection can be applied simultaneously. The algorithm is performed in boththe η and theφ projections to avoid accidental triggers from low-energy particles in the cavern.For the same reason a 3 out of 4 coincidence of the four layers of the two doublets is required. Theη andφ trigger information is combined to generate the RoI’s to be sent to the higher trigger levels.
– 189 –
The high-pT algorithm makes use of the low-pT trigger built from hits in RPC1 and RPC2,5125
and of the information generated in the RPC3 station. The algorithm operates in a similar way tothe low-pT one. The centre of the road is determined in the same way as forthe low-pT trigger, anda 1 out of 2 coincidence of the RPC3 doublet and of the low-pT trigger pattern result is required.As for the low-pT trigger, threepT thresholds operate simultaneously, resulting in a total ofsixthresholds reported to the central trigger logic for each event. Again, the trigger information inη5130
andφ is combined to form RoI’s to be sent to the L2 trigger.
System implementation The trigger scheme for the barrel muon trigger is implemented incustom-built electronics mounted either directly on the RPC detectors, or located outside the mainexperimental cavern. A schema of the trigger signal and readout chain is shown in Fig. 132. Signalsfrom the RPC detectors are processed in Amplifier-Shaper-Discriminator (ASD) boards attached5135
to the chambers at the end of the RPC strips. In the low-pT trigger, for each of theη and theφ projections, the RPC signals of the RPC1 and RPC2 doublets are sent to a coincidence matrixboard containing a coincidence matrix chip. This chip performs most of the functions of the triggeralgorithm and of the readout. At this stage the signals are aligned, the coincidence operations areperformed, and the threepT thresholds are applied. The coincidence matrix board produces an5140
output pattern containing the low-pT trigger results for each pair of RPC doublets in theη andφprojections. The information of the two adjacent coincidence matrix boards in theη projection,and the corresponding information of the two coincidence matrix boards in theφ projection, arecombined together in the low-pT Pad Logic (Pad) board. The four low-pT coincidence matrixboards and the corresponding Pad board are mounted on top of the RPC2 detector stations. The5145
low-pT Pad board generates the low-pT trigger result and the associated RoI information. This in-formation is transmitted to the corresponding high-pT Pad board, which collects the overall resultsfor low-pT and high-pT .
Figure 132.Schema of the trigger signal and readout chain of the L1 barrel muon trigger.
In the high-pT trigger, for each of theη andφ projections the signals from the RPC3 doublet,and the corresponding pattern result of the low-pT trigger, are sent to a coincidence matrix board,5150
very similar to the one used in the low-pT trigger. This board contains the same coincidence matrixchip as the low-pT board, programmed for the high-pT algorithm. The high-pT board producesan output pattern containing the high-pT trigger results for a given RPC doublet in theη andφprojection. The information of two adjacent coincidence matrix boards in theη projection and the
– 190 –
corresponding information of the two coincidence matrix boards in theφ projection are combined5155
in the high-pT Pad logic board. The four high-pT coincidence matrix boards and the correspondingPad board are mounted on top of the RPC3 detector.
The high-pT Pad board combines the low-pT and high-pT trigger results. The combined infor-mation for each bunch-crossing is sent via optical links to sector logic boards located in the USA15counting room. Each sector logic board receives inputs fromseven (six) low-pT (high-pT ) Pad5160
boards, combining and encoding the trigger results of one trigger sector. The sector logic boardsends the trigger data for each bunch-crossing to the Muon Central Trigger Processor Interface(MUCTPI, see section 7.2.2.3), located in the USA15 counting room.
For events that are selected by the L1 trigger, data are read out from both the low-pT and thehigh-pT Pad boards. These data include the RPC strip pattern and someadditional information used5165
in the L2 trigger. The readout data for events accepted by theL1 trigger are sent asynchronously toROD’s located in the USA15 counting room and from there to Readout Buffers (ROB’s). The datalinks for the readout data are independent of the ones used totransfer partial trigger results to thesector logic boards.
System segmentation and latency From the trigger point of view the barrel is divided into5170
two halves,η < 0 andη > 0, and within each half-barrel 32 logically identical sectors are defined.The correspondence between these logical sectors and physical chambers is indicated in the pictureon the right of Fig. 131. The Barrel Large chambers and the Barrel Small chambers of both middleand outer RPC stations are logically divided in two to produce two large sectors and two smallsectors per half-barrel octant. Inside a sector, the trigger is segmented in Pads and RoI’s.5175
A large sector contains six Pad regions, while a small sectorcontains seven Pad regions. Theregion covered by a Pad is about 0.2×0.1 in ∆η ×∆φ. Inside the Pad the trigger is segmented intoRoI’s. An RoI is a region given by the overlap of anη coincidence-matrix and aφ coincidence-matrix. The dimension of the RoI’s is about 0.1× 0.1 in ∆η ×∆φ. The total number of Pads is7×2×32 for the small sectors and 6×2×32 for the large ones, giving 416 Pads altogether. Since5180
one Pad covers fours RoI’s, the total number of RoI’s is 1664.To avoid losing efficiency due to uncovered regions in the trigger system, different parts of
the system overlap. However, this overlap can cause double-counting of muon candidates. Inthe barrel trigger system, overlap is treated and solved at three different levels. Within a Padregion the Pad Logic removes double-counting of tracks between the four RoI’s of the region. In5185
addition, if it is found that a trigger was generated in a zoneof overlap with another Pad region,this trigger is flagged as ‘border’ trigger and any overlap will be solved later on. The sector logicthen prevents double-counting of triggers within a sector.Triggers generated in zones of overlapbetween different sectors are flagged by the sector logic andsent to the MUCTPI, which preventsdouble-counting between sectors.5190
The latency of the muon barrel trigger is about 2.1 µs, well within the allowed envelope.
7.2.2.2 Muon end-cap trigger
Trigger signals The muon trigger for the end-cap regions [171] is based on signals providedby TGC detectors. Thin-gap chambers are similar to multi-wire proportional chambers but with ananode wire pitch greater than the anode-cathode distance. The time resolution is not as good as for5195
– 191 –
RPC’s, but enough to provide an efficiency greater than 99% for bunch-crossing identification forthe 25 ns gate of ATLAS. Crucial for the end-cap region of ATLAS is their larger rate capabilityof more than 20 kHz/cm2. The TGC’s are arranged in nine layers grouped into four planes inz(see also section 6). The TGC inner station at|z| ∼ 6.9 m consists of one plane of doublet units.At |z| ∼ 14 m seven layers are arranged in one plane of triplet units (M1, closest to the interaction5200
point) and two planes of doublet units (M2, M3). The doublet forming the plane farthest fromthe interaction point in each end-cap is referred to as the pivot plane, and its chamber layout andelectronics are arranged such that, to a good approximation, there are no overlaps or holes in thisplane. For triggering, the TGC’s cover a pseudorapidity range 1.05 < |η | < 2.4, except for theinnermost plane which covers a range 1.05 < |η | < 1.9. Each trigger plane consists of a wheel5205
of eight octants of chambers symmetric inφ. Each octant is divided radially into the ‘forwardregion’ and the ‘end-cap region’. Anode wires of TGC’s are arranged in the azimuthal direction andprovide signals forR information, while readout strips orthogonal to these wires provide signalsfor φ information. Both wire and strip signals are used for the muon trigger. Signals from twowire-planes and two strip-planes are read out from the doublet chambers, and signals of three wire-5210
planes but only two strip-planes are read out from the triplet chambers. Anode wires are groupedand fed to a common readout channel for input to the trigger electronics, resulting in wire-groupwidths in the range between 10.8 mm and 36 mm. Each chamber has32 radial strips, and thus thewidth of a strip is 4 mrad (8 mrad for the forward region). Wiregroups are staggered by half a wiregroup between the two planes of a doublet station, and by one third of a wire group between each5215
of the planes of a triplet station.
2000
4000
6000
8000
10000
12000
6000 8000 10000 12000 14000 16000
I
M1
M2 M3S L
=2.70
=1.05
=2.40
=1.92
Z (mm)
R (
mm
)
low PT
hi PT
pivot plane
end-cap
forward
S L
DL-LL01V01
1m
Trig
ger S
ecto
r
Sub-sectors: 15 x 4 = 60
Sub-sectors: 49 x 4 = 196
Sub-sector
2.5
2.0
1.5
η = 1.0
49
15
∆φ = 0.1
4 4 4
44 4
444
HFU001V03
Figure 133.Schema (left) and segmentation (right) of the L1 muon end-cap trigger.
Trigger algorithm The scheme of the L1 muon end-cap trigger is shown on the left handside of Fig. 133. The trigger algorithm extrapolates pivot-plane hits to the interaction point, to con-struct roads following the apparent infinite-momentum pathof the track. Deviations from this pathin the preceding confirming trigger planes are related to themomentum of the track. Coincidence5220
signals are generated independently forRandφ. A 3-out-of-4 coincidence is required for the dou-blet pair planes of M2 and M3, for both wires and strips, a 2-out-of-3 coincidence for the triplet
– 192 –
wire planes, and a 1-out-of-2 coincidence for the triplet strip plane. The final trigger decision in themuon end-cap system is done by merging the results of ther −φ coincidence and the informationfrom the EI/FI chambers. As theη −φ coverage of the EI/FI chambers is limited, the coincidence5225
requirements depend on the trigger region, in order to keep auniform efficiency in the end-capregion. Six sets of windows are constructed around the infinite-momentum path, corresponding tothree different high-pT and three different low-pT thresholds. Trigger signals from both doubletsand the triplet are involved in identifying the high-pT candidates, while in case of the low-pT can-didates the triplet station may be omitted to retain high efficiency, given the geometry and field5230
configuration of a specific region.
System implementation The trigger scheme outlined above is implemented in purpose-builtelectronics, partly mounted on and near the TGC chambers, and partly located outside the mainexperimental cavern. A schema of the trigger signal and readout chain is shown in Fig. 134. Thewire and strip signals emerging from the TGC’s are fed into anASD boards physically attached5235
to the edge of a TGC and enclosed inside the TGC electrical shielding. Each ASD board handles16 channels. From the ASD boards signals are routed to a patch-panel, which also receives timingsignals from the TTC system. Bunch-crossing identificationis performed at this stage, and physicaloverlaps of TGC chambers are handled. Each patch-panel receives signals from up to 18 ASDboards, routes DCS and other control and monitoring signals, and provides power to the next stage,5240
the slave boards, which hold the coincidence and readout circuits. Slave boards are placed on theaccessible outer surfaces of the TGC wheels: the electronics for the two doublets are mountedon the outside of the outer doublet wheel M3 and those for the triplets on the inner surface ofthe triplet wheel M1. The EI/FI slave boards are installed inracks located near the EI/FI wheels.Signals from the doublet and triplet slave boards are combined to identify high-pT track candidates5245
in coincidence boards combining all three trigger stations, so-called high-pT boards, located indedicated mini-racks around the outer rim of the triplet wheel. Wire (R-coordinate) and strip (φ-coordinate) information is still treated separately at this point. Signals from high-pT boards aresent to sector logic boards containing anr −φ coincidence unit and track selectors to select thehighest-pT coincidences. The sector logic also receives directly the signals from the EI/FI slave5250
boards and incorporates them into the trigger logic. The sector logic boards are located outside themain cavern in the service cavern USA15. The resulting trigger information for 72 separate triggersectors per side is sent to the MUCTPI.
Full-information data sets are read out through the data acquisition system in parallel with theprimary trigger logic. For readout purposes the slave boards of one or more trigger sectors are5255
grouped into local DAQ blocks. Each slave board is connectedto a so-called star switch, whichmanages the data collection for a local DAQ block. From the star switch, the data are passed on tothe ROD’s located in the USA15 counting room, and from there to ROB’s.
System segmentation and latency The right-hand side of Fig. 133 shows one pivot-planeoctant to demonstrate the trigger sector segmentation of the system. The pivot plane is divided5260
into two regions, end-cap (|η | < 1.9) and forward (|η | > 1.9). The end-cap region is divided intosix trigger sectors inφ, where a trigger sector is a logical unit that is treated independently in thetrigger. Trigger sectors are constructed to be projective with respect to the interaction point, and
– 193 –
Figure 134.Schema of the trigger signal and readout chain of the L1 muon end-cap trigger.
therefore may cross chamber boundaries. The forward regionis divided into three trigger sectors.There are hence 48 end-cap trigger sectors and 24 forward trigger sectors per end-cap of TGC5265
detectors. Each trigger sector consists of independent sub-sectors corresponding to eight channelsof wire groups and eight channels of readout strips, 148 for each end-cap trigger sector and 64 foreach forward trigger sector. The trigger sub-sectors correspond to the RoI communicated to thefollowing trigger levels in case a collision event is accepted by the first level trigger.
The latency of the muon end-cap trigger is about 2.1 µs, well within the allowed envelope.5270
7.2.2.3 Muon to Central Trigger Processor Interface (MUCTPI)
Functional overview The results of the muon barrel and end-cap trigger processors thatform the input to the MUCTPI provide information on up to two muon-track candidates per muontrigger sector. The information includes the position andpT threshold passed by the track candi-dates. The MUCTPI combines the information from all the sectors and calculates total multiplicity5275
values for each of the sixpT thresholds. These multiplicity values are sent to the CTP for eachbunch-crossing. For each sector either all muon candidatesmay be taken into account, or only thecandidate with the highestpT per sector. In forming the multiplicity sums, care has to be takento avoid double-counting of muon candidates in regions where trigger chambers overlap. As de-scribed above, many cases of overlaps are resolved within the barrel and end-cap muon trigger5280
processors. The remaining overlaps to be treated by the MUCTPI are those inφ direction betweenneighbouring barrel trigger sectors and between barrel andend-cap trigger sectors. The maximumoverall multiplicity is seven candidates. Larger multiplicities will appear as a multiplicity of seven.
Additional functions of the MUCTPI are to provide data to theL2 trigger and to the data ac-quisition system for events selected at L1. The L2 trigger issent a selectable subset of all muon5285
– 194 –
candidate information that form the RoI’s for the L2 processing. The muon RoI’s sent to L2 areordered, according to decreasingpT . The data acquisition receives a more complete set of infor-mation, including in addition the computed multiplicity values.
System implementation and latency The MUCTPI is divided into a number of buildingblocks which are housed in one 9U VMEbus crate. The differentfunctionalities of the MUCTPI5290
are implemented in three types of VMEbus modules which are connected to each other via an activeback-plane, and controlled by a commercial CPU unit acting as VMEbus master.
A total of 16 octant input boards each receive data corresponding to an octant in the azimuthaldirection and half the detector in theη direction. They form muon-candidate multiplicities for thisregion, correctly taking into account the overlap zones between barrel sectors and between barrel5295
and end-cap sectors. There is no overlap between muon trigger sectors associated with differentoctant boards. The interface board to the central trigger processor collects the multiplicity sumsfor the six pT thresholds over the custom back-plane described below. Thesums are transmittedto the CTP for each bunch-crossing. The interface board to the central trigger processor is alsoresponsible for distributing time-critical control signals to the rest of the MUCTPI system. The5300
readout driver of the system sends candidate information tothe data acquisition and the L2 triggerfor each accepted event. All modules are connected via a custom-built back-plane. It contains twocomponents: an active part forms the total candidate multiplicities by adding the multiplicities ofthe input boards; a passive part contains a bus system to transfer data from the input boards and theinterface board to the central trigger processor to the readout driver of the system.5305
The latency of MUCTPI is contained in the latency numbers forthe barrel and end-cap muontrigger systems quoted above, to which it contributes 0.2 µs.
7.2.3 Central trigger processor
7.2.3.1 Functional overview The Central Trigger Processor [172] (CTP) receives triggerinfor-mation from the calorimeter and muon trigger processors, which consists of multiplicities for elec-5310
trons/photons,τ /hadrons, jets, and muons, and of flags for total and missing transverse energy,and total jet transverse energy. Additional inputs are provided for special triggers such as a filled-bunch trigger based on beam-pickup monitors, and a minimum-bias trigger based on scintillationcounters. Up to 372 signals can be connected to the input boards of the CTP; however, only upto 160 can be transmitted internally. The selection of the signals used from all signals available at5315
the input boards is programmable. The currently foreseen input signals listed in Table 42 do notexceed the internal limit of 160 and will therefore all be available in parallel.
In the next step the CTP uses look-up tables to form trigger conditions from the input signals.Such a condition could be, for example, that the multiplicity of a particular muon threshold hasexceeded one, i.e. at least two muons in this event have passed this threshold. For such an event,5320
this trigger condition would be set totrue. Further trigger conditions are derived from internallygenerated trigger signals: two random triggers, two prescaled clocks, and eight triggers for pro-grammable groups of bunch-crossings. The maximum number oftrigger conditions at any onetime is 256.
The trigger conditions are combined to form up to 256 triggeritems, where every trigger5325
condition may contribute to every trigger item. An example for a trigger item would be that the
– 195 –
Cable origin Number of Bits Trigger Information
Muon Processor 6 thres.× 3 bits muon multiplicities
Cluster Processor 1 8 thres.× 3 bits electron/photon multiplicities
Cluster Processor 2 8 thres.× 3 bits electron/photon orτ /hadron multiplicities
Jet/Energy Processor 18 thres.× 3 bits jet multiplicities
Jet/Energy Processor 22×4 thres.× 2 bits forward jet multiplicities for each side
Jet/Energy Processor 34× 1 bit total transverse-energy sum
8× 1 bit missing transverse-energy sum
CTP_CAL 28 bits up to 28 input bits for additional trigger in-puts from beam pick-ups (see section 8.10),beam condition monitors (see section 3.4.1),luminosity detectors (see sections 10.3.3 and10.3.1), zero-degree calorimeters (see sec-tion 10.3.2), and others.
Table 42.Trigger inputs to the CTP of the L1 trigger. The number of bitsimplies the maximum multiplicitythat can be encoded, i.e. up to 7 for 3 bits. Multiplicities larger than this value will be set to the possiblemaximum, in this case 7.
following conditions have been fulfilled: at least two muonshave passed a particular threshold,and at least one jet has passed a particular threshold. Furthermore each trigger item has a mask,a priority (for the dead-time generated by the CTP), and a prescaling factor. The L1A signalgenerated by the CTP is the logical OR of all trigger items.5330
The CTP provides an 8-bit trigger type word with each L1A signal, which indicates the type oftrigger and can be used to select options in the event data processing in the front-end electronics andreadout chain. The CTP sends, upon reception of each L1A signal, information about the triggerdecision for all trigger items to the L2 trigger and the data acquisition. For monitoring purposes theCTP provides bunch-by-bunch scalers of inputs and, integrated over all bunches, scalers of trigger5335
inputs and trigger items before and after prescaling.
In addition to its function in the selection chain, the CTP isalso the timing master of theexperiment. The clock signal synchronised to the LHC beams arrives at the CTP, and is distributedfrom here together with the L1A and other timing signals to all other sub-systems.
7.2.3.2 System implementation and latency The CTP consists of several different modules5340
which are housed in a single 9U VMEbus crate. A diagram is shown in Fig. 135. In addition toVMEbus, the CTP modules use custom busses for the synchronised and aligned trigger inputs (PIT-bus, where PIT = pattern in time), for the common timing and trigger signals (COMbus), and forthe sub-detector calibration requests (CALbus). PITbus, COMbus and CALbus are implementedon the same custom-built back-plane installed in the CTP crate.5345
– 196 –
Figure 135.Diagram of the Central Trigger Processor of the L1 trigger.
Six different module types are employed in the CTP system. The timing signals from theLHC are received by the CTP machine interface module (CTP_MI), which can also generate thesesignals internally for stand-alone running. It controls and monitors the internal and external busysignals. The CTP_MI sends the timing signals to the COMbus. The CTP input module (CTP_IN)receives trigger inputs from the muon and calorimeter trigger processors and other sources. The5350
CTP_IN selects and routes trigger inputs to the PITbus, after synchronising them to the clock signaland aligning them with respect to the bunch-crossing. The CTP core module (CTP_CORE) receivesthe 160 trigger inputs from the PITbus. It combines them and additional internal triggers usingseveral look-up tables to form up to 256 trigger conditions.The conditions are combined usingcontent-addressable memories to form up to 256 trigger items. The trigger masks, prescales and5355
dead-time generation following the forming of the trigger items are implemented in this module.The CTP_CORE sends the trigger results to the COMbus. It alsoacts as the readout driver of thesystem, sending information to the L2 trigger and the data acquisition for each accepted event.The CTP monitoring module (CTP_MON) receives the 160 trigger inputs from the PITbus andmonitors their behavior on a bunch-by-bunch basis. The CTP output module (CTP_OUT) receives5360
the timing and trigger signals from the COMbus and fans them out to the sub-detectors, where theyare received by local trigger processors [173]. The CTP_OUTreceives back from the sub-systemsthe busy signals, which are sent to the COMbus, and calibration trigger requests, which are routedto the CALbus. The CTP calibration module (CTP_CAL) time-multiplexes the calibration requestson the CALbus and sends them via a front-panel cable to the CTP_IN. The CTP_CAL also contains5365
external inputs for beam pick-up monitors, minimum-bias scintillators, and test triggers.
The latency of the CTP is contained in the latency numbers forthe barrel and end-cap muontrigger systems and the calorimeter trigger system quoted above, to which it contributes 0.1 µs.
Receiving the timing signals from the accelerator and distributing the ATLAS timing signalsfrom the local trigger processors to the detector electronics is accomplished using the Timing,5370
– 197 –
trigger, and Control system (TTC). The ATLAS TTC system is based on the optical fan-out systemdeveloped within the framework of RD12 [174], which allows signals to be distributed from onesource to up to 1024 destinations. The system can be partitioned and sub-detectors can run withthe central ATLAS timing and trigger signals, or independently with their own specific signals.
7.3 HLT/DAQ5375
7.3.1 Overview
As explained in section 7.1, the main components of the HLT/DAQ are: readout, L2 trigger,event-building, event filter, Configuration, Control and Monitoring, and information services. Anoverview of the event selection performed by the HLT is givenin section 7.3.2. Here the move-ment of data from the detectors to the HLT and subsequently tomass storage is described. The5380
main features of each component are described below.
A block diagram of the DAQ/HLT is shown in Fig. 125. The movement of events from thedetector to mass storage commences with the selection of events by the L1 trigger. During thelatency of the L1 trigger selection, up to 2.5µs, the event data is buffered in memories locatedwithin the detector-specific front-end electronics. On selection by the L1 trigger the event data is5385
transferred to the DAQ/HLT system over 1600 ROL’s, having first transited through the detector-specific ROD’s. The 1600 event fragments are received into the 1600 ROB’s contained in theRead-Out System (ROS) units where they are temporarily stored and provided, on request, to thefollowing stages of the DAQ/HLT system.
For every selected event, the L1 trigger sub-systems (calorimeter, muon, and CTP) also pro-5390
vide the RoI information on eight ROL’s, a dedicated data path, to the RoI Builder where it isassembled into a single data structure and forwarded to the L2 supervisor (L2SV). As its namesuggests, the L2SV marshals the events within the L2 trigger. It receives the RoI’s, assigns theevents to one of the L2 trigger’s processing units (L2PUs) for analysis and receives the result of theL2PUs analysis.5395
Using the RoI information, requests for event data are made to the associated ROS’s. Thesequence of data requests is determined by the type of RoI identified by the L1 trigger and theconfiguration of the L2 trigger processing, i.e. the order ofitems in the trigger menu and the orderof the algorithms per trigger item. The result, accept or reject, of the analysis is returned to theL2SV which subsequently forwards it to the DataFlow Manager(DFM). In addition to sending5400
the result of its analysis to the L2SV, an L2PU also sends a summary of the analysis that it hasperformed to a L2 trigger-specific ROS.
The DFM marshals the events during the event-building. For those events which were foundnot to fulfil any of the L2 selection criteria, the DFM informsall the ROS’s to expunge the asso-ciated event data from their respective ROB’s. Each event that has been selected by the L2 trigger5405
is assigned by the DFM to an event-building node (called SFI). The SFI collects the event datafrom the ROS’s and builds a single event-data structure, theevent. An SFI can be building morethan one event at a time and the requests to the ROS’s for theirdata are dispatched according to analgorithm that ensures the quantity of data being received by the SFI does not exceed its availableinput bandwidth. The assembled event-data structure is sent to the event filter for further analysis.5410
– 198 –
On completing the building of an event an SFI notifies the DFM,which subsequently informs allthe ROS’s to expunge the event data from their respective ROB’s.
The event filter, in addition to the selection, classifies theselected events according to a pre-determined set of event streams and the result of this classification is added to the event structure.Selected events are subsequently sent to the output nodes (SFO’s) of the DAQ/HLT system. Con-5415
versely, those events not fulfilling any of the event filter selection criteria are expunged from thesystem. The events received by an SFO are stored in their local file system according to the classi-fication performed by the event filter. The event files are subsequently transferred from the SFO’sto CERN’s central data-recording facility.
7.3.2 HLT event selection5420
The selection of events by the HLT containing high transverse-energy electrons is seeded by thecandidate electromagnetic clusters found by the L1 trigger. Using the full-granularity data fromthe calorimeters, the transverse energy and position measurements are re-evaluated. The improvedprecision allows more stringent threshold cuts inET to be applied. Following this initial step, theleakage into the hadronic calorimeter is evaluated and variables related to the transverse shower5425
shape in the electromagnetic calorimeter are used to perform preliminary particle identification.For those candidate electromagnetic clusters that are consistent with an electron, track reconstruc-tion is performed within the RoI in the inner detector. Subsequently, the association between theelectromagnetic cluster and a candidate track is performedby matching the clusterη andφ posi-tion with theη andφ values of candidate tracks extrapolated to the calorimetersurface. If more5430
than one candidate track is found within the RoI, the track closest inη andφ to the electromag-netic cluster is selected. For those events for which an electromagnetic cluster can be successfullyassociated to an inner detector track, the ratio of the measured transverse energy and the transversemomentum of the corresponding track is used for particle identification.
The steps to select events containing high transverse-energy photons are similar to those of5435
the electron: a transverse shower shape in the electromagnetic calorimeter consistent with that ofa photon, including the requirement of isolation; and a vetoon the association, inη andφ, of theshower with a charged track in the inner detector (after verification that the candidate photon didnot convert into ane+e− pair within the volume of the inner detector).
The selection of high transverse-momentum muons is seeded by the muon RoI’s identified by5440
the L1 trigger. In a first step the re-identification of the RPC(or TGC) data associated to eachRoI is performed, and the RPC (or TGC) data are subsequently used to seed the identification oftrack segments in the MDT chambers. This first step also results in a more precise estimate ofthe transverse momentum, allowing a more stringent selection to be performed. The direction ofthe track segments in the muon spectrometer are then used to seed track reconstruction within the5445
inner detector. Subsequently, the association between thetrack segment in the muon spectrometerand candidate track segments in the inner detector is performed by matching, at a common surface,theη andφ values. Further selection is then made on the combined track’s transverse momentum.For candidate high transverse-momentum ( >20 GeV/c) a selection is also made on the muon’sisolation, i.e. the muon does not lie within a jet, using the electromagnetic and hadronic calorimeter5450
information.
– 199 –
The selection ofτ ’s in their hadronic decay modes is seeded by an isolated electromagneticcluster found by the L1 trigger. The transverse energy is re-evaluated and more stringent thresh-olds are applied both on the cluster and the calorimeter isolation fraction. If the cluster width isconsistent with aτ cluster then track reconstruction is performed in the innerdetector within the5455
RoI. Tracks within the narrow candidateτ signal cone are summed and required to match to thecalorimeter cluster inη andφ. Tracking isolation in the region outside the candidateτ signal conebut within the RoI is also required. The system mass, charge,and track multiplicity are also usedfor τ identification.
The identification and selection of jets is similar to that ofτ ’s, but they currently use no track-5460
ing or calorimeter isolation information.
The identification of b-jets, with|η | < 2.5, is initiated by the tracks associated to candidatejets. A selection is made on the value of the impact parameter, the presence of a secondary vertices,and the selection of low-pT leptons.
Within |η | < 4.9, the signature forEmissT will be based on the full calorimeter data and the5465
information from candidate muons.
7.3.3 Control
The overall control of the ATLAS experiment covers the control and monitoring of the operationalparameters of the detectors and experiment infrastructure, as well as the coordination of all detector,trigger and data acquisition software and hardware associated with data-taking. This functional-5470
ity is provided by two independent, complementary and interacting systems: the data acquisitioncontrol system, and the Detector Control System (DCS). The former is charged with controllingthe hardware and software elements of the detectors and the DAQ/HLT needed for data-taking,while the DCS handles the control of the detector equipment and related infrastructure. The DCSis described in section 7.5.5475
The DAQ/HLT system and detector systems are composed of a large number of distributedhardware and software components that in a coordinated manner provide for the data-taking func-tionality of the overall system. Likewise, their control and configuration is based on a distributedcontrol system. The control system has two basic components: the process manager and the runcontrol.5480
On each computer a process management daemon waits for commands to launch or interruptprocesses. On the reception of such commands it interrogates the access manager and the resourcemanager to ascertain whether the requested operation is permitted. It is a task of the access managerto indicate whether the requester is authorised to perform the operation, while it is a task of theresource manager to check that the resources are available to perform the operation.5485
A hierarchical tree of run controllers, which follows the functional de-composition into sys-tems and sub-systems of the experiment, steers the data acquisition by starting and stopping pro-cesses and by carrying all data-taking elements through a finite state machine, which ensures thatall parts of ATLAS are in a coherent state. Similarly to what happens for the commands to theprocess manager, run control commands also have to be authorised by the access manager. In addi-5490
tion to implementing a global finite state machine and managing the lifetime of processes, the runcontrollers are further customised according to the sub-system they are in charge of (e.g. they can
– 200 –
be extended with sub-states, or perform specific actions during transitions). One example of cus-tomised controller is the root controller, the starting point of the run control tree, which retrievesthe run number from the run number service before starting any new run and drives luminosity5495
block changes during the data-taking.Another fundamental aspect of the control is the diagnosticand error recovery system. Several
aspects of it are integrated into the run control processes.Errors raised by any data-taking nodeenter the error reporting system and can be reacted upon. Thediagnostic system can launch a setof tests to understand the origin of the reported problem andthe recovery system can then take5500
corrective actions. These aspects of the control have been implemented using an expert system.The knowledge base and the rules of operation differ for the various parts of the trigger and dataacquisition system and are developed in strict collaboration with the sub-system experts.
7.3.4 Configuration
The description of the hardware and software of the experiment required for data-taking are main-5505
tained in configuration databases. A configuration is organised as a tree of linked segments ac-cording to the hierarchy describing the DAQ/HLT and detector systems. A segment defines a welldefined sub-set of the hardware, software, and their associated parameters. The organisation of thedata is described by common object-oriented database schema which may be extended to describethe properties of specific hardware and software.5510
To support the concurrent access to the configuration data bythousands of applications and tonotify control applications of changes to the configurationdata during run time, remote databaseservers are deployed to allow access times to the configuration data do not scale with the number ofdeployed applications. An additional server, DBproxy, is deployed to cache the results of queriesto the relational databases, e.g. the conditions database.5515
Partitioning Partitioning refers to the ability to operate subsets of theATLAS detector inparallel and disjoint, thus facilitating the concurrent commissioning and operation of subsets of thedetector. Once two or more partitions have been commissioned they may then be operated togetheras a single partition. Partitions are in this way combined into a fully integrated and operationaldetector.5520
A partition maps to a TTC partition, therefore defining the subset of detector componentswithin a partition, see section 7.2. In addition, the staticpoint-to-point connections between thedetector ROD’s and the ROS’s uniquely associates a set of ROS’s to a partition. Other componentsof the DAQ/HLT (i.e. event-building nodes, event filter nodes and SFO’s) connected by multi-layered networks and can therefore be assigned to a partition as required by the operations to be5525
performed. The management of resources, e.g. event-building nodes, between partitions is achievedby the resource manager.
The RoI Builder drives the input to the L2 trigger and, from anoperational perspective, canonly be operated as a single unit. Thus the L2 trigger cannot be partitioned. Analogously, theCTP can only be operated as a single unit, therefore the complete L1 trigger may not be operated5530
in more than a single partition. However, to facilitate calibration and checking of L1Calo inputsignals, it is possible to operate a partition that consistsof the L1Calo and the liquid argon and/ortile calorimeters but independent of the CTP.
– 201 –
7.3.5 Monitoring and information distribution
The monitoring component provides the framework for the routing of operational data produced5535
by the DAQ/HLT components and its analysis. Operational data ranges from physics event datato histograms and the values of parameters. The routing of operational data is performed by theinformation, on-line histogramming and event monitoring services. The information service pro-vides the distribution of the values of simple variables. Consumers of the information are able tosubscribe to notifications of changes to one or more information items. It also provides a means for5540
any application to send commands to any of the information providers, specifically for the controlof information flow, e.g. an application may ask a particularprovider of information to increase thefrequency at which it publishes a particular piece of information. Complementing the exchange ofthe values of simple variables, MRS transports messages among trigger and data acquisition appli-cations. Messages may be used to report debug information, warnings or error conditions. MRS5545
allows association to a message of qualifiers and parameters. Moreover, receivers of messages areable to subscribe to the service to be notified about incomingmessages and apply filtering criteria.
The On-line Histogramming Service (OHS) extends the functionality of the information ser-vice to histograms, in particular raw and ROOT histograms. Within the DAQ/HLT there are manyinstances of the same application, e.g. L2PUs, active at anyone time producing histograms. Via the5550
OHS, a gatherer application sums histograms of the same typeand in turn publishes, via the OHS,the resulting histograms. The visual presentation of histograms is based on ROOT and Qt, andallows for the presentation of reference histograms, fitting, zooming and the sending of commandsto histogram providers.
The event filter processing application is based on the off-line computing framework. The sub-5555
stitution of the selection algorithms with a monitoring or calibration algorithm allows for monitor-ing and/or calibration tasks based on the offline computing framework to operate on-line, receivingevents from the SFI’s. It is also possible to configure these applications to receive events from theevent monitoring service. The latter provides a framework to enable the sampling and distributionof event data as it flows through the DAQ/HLT system. Monitoring applications are able to request5560
event fragments according to the values of elements in the event fragment, e.g. trigger and/or sub-detector type, from a specific sampling point, e.g. a particular ROS (part of an event) or SFI (acomplete event). Examples of monitoring applications using this service are the event dump andevent display.
To complement the viewing and analysis of histograms by an operator, a data quality monitor-5565
ing framework provides the automatic comparison of recently acquired data to reference data (e.g.reference histograms), statistical checks and alarm generation. More specifically, user-supplied al-gorithms and or reference data are used to automatically analyse the large quantitates of monitoringdata, and generate alarms when deviations from the specifiedcriteria occur.
7.3.6 Readout system5570
As described in section 7.3.1, the ReadOut System (ROS) receives event data from the detectorROD’s via 1600 ROL’s. The ROL has a homogeneous design and implementation, based on the S-LINK interface. It allows for the transmission of 32-bit data at 40.08 MHz, i.e. up to 160 Mbyte/s,and implements flow control and error detection [175]. ROB’sare the buffers located at the receiv-
– 202 –
0
2
4
6
8
10
12
14
MuP
rec
MuP
rec
MuTrig
Em
Cal
Em
Cal
Em
Cal
Em
Cal
Em
Cal
Em
Cal
Em
Cal
Em
Cal
HadC
al
TR
T
TR
T
TR
T
SC
T
Pix
els
RoI
request
rate
(kH
z)
Figure 136. Expected average data request rate perROS.
Figure 137. L1 trigger accept rate versus L2 triggeraccept rate.
ing end of the ROL’s, there being one ROL associated to one ROB. Three ROB’s are physically5575
implemented on a module called ROBIN and up to six ROBIN’s canbe located in a ROS which isimplemented on a server-class PC. The ROS provides for the multiplexing of up to 18 ROL’s to thesubsequent components of the DAQ/HLT, i.e. L2 trigger and event-building, reducing the numberof connections by a factor of approximately ten.
A request by an L2PU for data involves, on average, one or two ROB’s per ROS, whereas5580
the requests for data from the event-building nodes concerns the event data from all the ROB’sof a ROS. In either case, the ROS replies to the requester witha single data structure. At thenominal L1 trigger rate of 100 kHz, and an average of 1 kbyte received per ROL, the ROS is ableto concurrently service up to approximately 20 kHz of data requests from the L2 trigger, up to3.5 kHz of requests from event-building nodes, and expunge events on request from the DFM. The5585
rate of data requests received by a specific ROS depends on theη −φ region of the data it receivesover the ROL’s and from which detector it receives data, i.e.a ROS that receives data from theliquid-argon calorimeter barrel region is solicited for data more frequently than a ROS associatedto the barrel MDTs. The expected average rate of data requests as a function of 143 ROS’s in thefinal system is shown in Fig. 136. Fig. 137 shows the maximum L1trigger rate sustainable by the5590
worst-case ROS for different values of the L2 trigger’s acceptance, i.e. the event-building rate. Thefigure also shows the expected operating conditions.
ROBIN The ROBIN component provides the temporary buffering of theindividual eventfragments produced by the ROD’s. As mentioned above it receives event fragments at the L1trigger accept rate, i.e. up to 100 kHz, and subsequently buffers these data for the duration of the5595
L2 trigger decision and, for approximately 3% of the events,the duration of the event-buildingprocess. In addition, it services requests for data at up to rates of approximately 20 kHz. As aconsequence of the rates that have to be supported, the ROBINis a custom designed and builtPCI-X mezzanine [176].
All functions related to the receiving and buffering of event fragments from three ROL’s, i.e.5600
operations occurring at up to 100 kHz, are realised in an FPGA. A PowerPC is used to imple-ment the functions of memory management, servicing of data requests, control and operational
– 203 –
monitoring.
7.3.7 L2 trigger
The L2 trigger is achieved by the combined functionality of the RoI Builder, L2SV, L2PU and5605
Pseudo ROS. The RoI Builder receives, on eight point-to-point links, the RoI information from thedifferent sources within the L1 trigger and merges them intoa single data structure. It is thus at theboundary between the L1 and L2 trigger systems and operates at the level-1 trigger rate, i.e. up to100 kHz, see section 7.3.7. The single data structure containing the RoI data is transmitted by theRoI Builder over the ROL’s to the L2SVs. As described in section 7.3.1, the L2SVs marshal the5610
events through the L2 trigger.
The principal component of the L2 trigger is of course the L2 processing farm on which theevent selection (the L2PU) is executed. The system is designed to provide an event rejection factorof thirty, with a average event latency of approximately 10 ms, using only the data located in theRoI’s, i.e. 1-2% of the full event. The number of L2PUs performing the physics selection per node5615
is configurable. On the hardware currently deployed, see section 7.4, there are eight L2PUs pernode, i.e. an L2PU per processing core of the node, thus the average event latency per L2PU is40 ms.
Not all data within a RoI is requested from the ROS’s in a single request. As described in sec-tion 7.3.2, the first step of the L2 selection is to confirm the hypothesis of the L1 trigger. This5620
results in only requesting the data in the RoI from the calorimeters in the case of electromagneticclusters. Subsequent requests are made for the data from theinner detector. In this manner thefull data set associated to the RoI is only transferred for those events which fulfill the complete L2trigger selection criteria, i.e. the amount of data transferred between the ROS’s and the L2 triggeris minimised for those events which are rejected.5625
The event filter selection is seeded by the results of the L2 trigger’s analysis. The transmissionof a summary of the L2 trigger’s selection is achieved by the deployment of a L2 trigger-specificROS, the pseudo ROS. At the end of its event analysis the L2PU sends to the pseudo ROS infor-mation that summarises the results of its analysis. Subsequently, the pseudo ROS participates inevent-building as any other ROS within the system, its eventdata being the L2 trigger’s summary5630
analysis. In this way, the results of the L2 trigger’s analysis are built into the final event and thusavailable to the event filter selection.
The failure of one of more L2PUs during run time does not incursystem down time. Thesystem continues to operate at a reduced rate while the failed application, the L2PU, is restartedunder the supervision of the Run Control.5635
Region of Interest Builder The RoI Builder [177] is one of only three custom-built com-ponents within the DAQ/HLT system. It is a 9U VMEbus system which includes a single-boardcomputer for the purpose of configuration, control and operation monitoring. It is composed of aninput stage and an assembly stage. The input stage consists of three input cards that each receiveand buffer, over three ROL’s, the RoI data from eight L1 trigger sources, namely: three CP boards,5640
three JEP boards, the MUCTPI, and the CTP.The eight RoI fragments are subsequently routed over a custom back-plane to builder cards
in the assembly stage where they are assembled into a single data structure (RoI record). The
– 204 –
assignment of each event to a specific builder card for assembly is based on a token-passing mech-anism between the builder cards. Each builder card has four ROL’s which are used to transfer the5645
assembled RoI records to up to four L2SVs according to a round-robin algorithm.
Detector calibration using RoI. The calibration of the muon MDT chambers requires largedata samples within a well-defined time window to establish the relationship between the driftpath and measured time as a function of time. This measurement has to be made from the dataof the MDTs alone using candidate tracks, and is based on an iterative procedure starting from a5650
preliminary set of constants.The L1 trigger rate for high transverse-momentum muons is approximately 12 kHz. For these
candidate events the first step of the L2 trigger selection isthe reconstruction of tracks in the muonsystem. To facilitate the calibration of the MDTs, each L2PUcan be configured to additionallywrite, to a pre-defined buffer, the data of the candidate tracks and the results of its analysis, i.e.5655
the RoI information and the results of track fits. Another application retrieves groups of eventsfrom the buffer and sends them to a calibration server. The calibration server receives the groupsof events from all L2PUs, stores them to a local disk, and subsequently sends them to a remotecalibration farm (Tier 2 in the LHC computing model) for processing. The function of moving RoIdata to specific processing nodes dedicated to calibration is shown schematically in Fig. 138.5660
Figure 138.Block diagram shown the movement of RoI data to dedicated calibration nodes.
7.3.8 Event-building
The event-building functionality is provided by the DFM, ROS’s and SFI’s. The SFI is the applica-tion which collects the event data from the ROS’s and assembles a single formatted data structure.An SFI is configured with a randomised list of the ROS’s withinthe system, which is used to definethe order in which data requests are sent to the ROS’s. This results in the randomisation of the5665
traffic pattern in the underlying network and hence improvednetwork performance. To meet therate requirements a number of SFI’s work in parallel, each instance building a number of events
– 205 –
concurrently. Each SFI informs the DFM of its readiness to receive events, and the DFM allocatesevents to the SFI’s so as to ensure that the load is balanced across all available SFI’s.
The default behaviour of the SFI is to collect all the event data associated to a L1 identifier into5670
a single formatted data structure. It is also possible to configure the SFI’s to only build the singledata structure from a subset of the available event data, using as a basis the L1 and or L2 triggertype. To this end each SFI is configured with the information that allows it to associate the L1 orL2 trigger type to the subset of ROS’s whose event data data should be collected, thus realising amore efficient use of the available resources. For example, events can be built which have been5675
selected by the L1 trigger purely to facilitate the monitoring of the tile calorimeter and only requirethe data from the tile calorimeter.
In the event that a requested ROS data fragment is not received within a configurable timebudget, the outstanding data fragment can be re-requested.Only if several consecutive requests areun-fulfilled does the SFI abandon the inclusion of the missing data and assemble an incomplete5680
event. After an event has been moved to the event filter the SFImarks its buffers for re-use.If for whatever reason the buffers of the SFI become full, theSFI informs the DFM, i.e. exertsback-pressure, which subsequently suspends allocating events to the specific SFI until the SFI re-indicates its availability.
The event-building system is designed to function in the case of failure of one or more SFI5685
nodes. In this situation, the DFM ceases to assign events to the failed SFI’s. Once the failed nodesbecome re-available they can be re-integrated into the event-building system without the systemincurring down time.
7.3.9 Event filter
The event filter is a processing farm; on each processing nodea configurable number of indepen-5690
dent processing tasks receive and process events. Unlike the L2 trigger, these tasks are standardATLAS event reconstruction and analysis applications. Thephysics selection process is describedin section 7.3.2. For those events passing the selection criteria, a subset of the data generated duringthe event analysis is appended to the event data structure, enabling subsequent analysis at Tier 0,1, 2 and 3 to be seeded by the analysis of the event filter. An integral part of the selection process5695
is the classification of the events according to the ATLAS physics streams, see section 7.3.10. Tothis end, for those events that fulfill the selection criteria, a tag is added to the event data structureidentifying into which physics stream the event has been classified.
The failure of one of more event-filter processing tasks or ofa complete node during run-timedoes not provoke any system down-time. The system continuesto operate at a reduced rate while5700
the failed application, or node, is restarted under the supervision of the run control. To ensure thatno events are lost during such failures, each event on arrival in the event filter is written to diskstorage. On the restart of the failed application or of the node itself, an attempt can be made tore-analyse the event or accept the event without analysis.
7.3.10 Event output5705
The main functionality of the event-filter output nodes (SFO’s) is to receive events, which havepassed the event filter selection criteria, interface the DAQ/HLT to CERN’s central data recording
– 206 –
facility, and de-couple the data-taking process from possible variations in the central data recordingservice.
The SFO maintains, locally, a set of files into which it records events. The set of files maps5710
to the ATLAS defined inclusive data streams: electrons, muons, jets, photons,EmissT andτ ’s, B-
Physics. Each event is recorded in one or more files accordingto the stream classification made bythe event-filter processing task. The overlap, in terms of rate, between the inclusive data streamsare shown in Table 43.
Table 43.Overlap (Hz) between the data streams at a luminosity of 1033 cm−2s−1.
Stream e µ Jet γ ET & τ B-physics
e 38±8.8 0.1±0.0057 0.0039±0.0015 0.0077±0.0022 3.1±0.04 (8.6±4.5)×10−6
µ − 40±10 0.022±0.015 0.0034±0.002 0.3±0.075 1.0±0.57
Jet − − 54±8 0.11±0.03 0.7±0.4 0±0
γ − − − 40±7 0.055±0.013 0±0
ET & τ − − − − 26±6.8 (3.9±3.9)×10−6
B-physics − − − − − 1.6±0.7
In addition to the data streams mentioned above, a subset of the events is also written to5715
calibration streams and an express stream. The express stream is a subset of the events selected bythe event filter that have undergone an additional selectionstep, e.g. higher thresholds, and will beanalysed with high priority. The calibration streams contain events which have passed the eventfilter selection criteria, they contain only a subset of an event and are selected as being useful fordetector calibrations. These streams have a full overlap with the streams given in Table 43.5720
In the eventuality of a prolonged failure, up to 24 hours, in the transmission of data to CERN’scentral data recording service, the system is equipped withsufficient local storage capacity to bufferall events locally. Under normal operating conditions, this storage capacity is un-used and allowsthe system to operate at a peak event rate of up to 400 Hz.
7.4 Implementation and capabilities of the DAQ/HLT5725
Most of the DAQ/HLT functionality is implemented on commodity, rack-mountable, server-classPC’s. The PC’s run Scientific Linux and are interconnected bymulti-layer Gigabit Ethernet net-works, one for control functionality and another for data movement. The majority of PC’s havesimilar specifications (e.g. two CPU sockets, two gigabit ethernet connections, support for IP-MIv2.0), except for the number and type of CPUs implemented and the amount of memory. The5730
main features per component and the number of nodes deployedfor initial operations are given inTable 44. A few components, the RoI Builder, ROL and ROBIN, are, however, implemented incustom hardware.
The ROS PC’s are installed in standard ATLAS 52U racks, whileall other PC’s are installedin standard 47U or 52U server racks. The number of racks for each component type is given in5735
Table 44. In addition to the PC’s each rack also contains a local file server and two gigabit ethernetswitches. The later form part of the multi-layered Gigabit Ethernet network which implements
– 207 –
Component# Nodes# Racks# CPUs/nodeMemory (Gbyte) Type of CPU
ROS 149 16 1 0.512 3.4 GHz Irwindale
SFI 48 3
2 2 2.6 GHz Opteron 252DFM 121
L2SV 10
HLT 1116 36 8 8 Xeon E5345 2.33 GHz
SFO 6 2 2 4 Xeon E5130 2.0 GHz
Monitoring 324
4 8 Xeon E5160 3.0 GHz
Operations 20 2 4 Xeon E5130 2.0 GHz
Table 44. The main data-acquisition system components to be deployedfor initial operation: the readoutsystem (ROS), the event-building node (SFI), the dataflow manager (DFM), the L2 supervisor (L2SV), thehigh-level trigger (HLT) and the event filter output nodes (SFO).
the control and data networks. Each rack is also equipped with a water-cooled heat-exchanger,specific to the horizontal airflow within a rack, that provides 9.5 kW of cooling power. The numberof 1U PC’s per rack is typically just over thirty, constrained by cooling power, power distribution5740
(particularly in-rush current) and weight limits.
All PC’s are booting over the network using the preboot execution environment, the local fileservers providing the role of boot servers for the: kernel and boot image files; ATLAS softwareand user related information. They also provide the scratchspace to the nodes they serve, typicallythirty nodes, and are also used to host various central services related to the configuration and5745
control of the trigger and data acquisition system. A central file server holds the master copyof the software, which, in conjunction with booting over thenetwork, ensures the uniformity ofthe software used throughout the system. The unique installation of the operating system and theATLAS software is distributed from the central file servers to the local file server on a daily basis,though this synchronisation can occur at any time.5750
For initial operations, the DAQ/HLT system will be fully configured in the area of configu-ration, control and monitoring functionality. The operations PC’s are used to provide the variouscentral services for configuring and controlling the trigger and data acquisition system (e.g. runcontrol, error logging). The monitoring PC’s are used to monitor the system and sampled eventdata.5755
The initial system will also support full detector readout,over the 1600 point-to-point ROL’s,into the ROS’s at up to a L1 trigger rate of 75 kHz. The number ofROL’s, ROBIN’s and ROS’s perdetector TTC partition are given in Table 45. Also given in this table is the size of event data perL1 trigger for each part of the detector.
As described in section 7.3.8, the event-building functionality is performed by a set of SFI’s5760
and scales linearly with the number of SFI’s, each SFI contributing 60 Hz and approximately
– 208 –
TTC Partition # ROD’s # ROL’s # ROS’s Data per L1A signal (kB)
Inner detector
Pixel
B layer 44 44 4
60Disks 24 24 2
Layers 1-2 64 64 6
SCT
End-cap A 24 24 2
110End-cap C 24 24 2
Barrel A 22 22 2
Barrel C 22 22 2
TRT
End-cap A 64 64 3
307End-cap C 64 64 3
Barrel A 32 32 6
Barrel C 32 32 6
Calorimetry
Tile
Barrel A 8 16 2
576Barrel C 8 16 2
Ext. barrel A 8 16 2
Ext. barrel C 8 16 2
LAr
EM barrel A 56 224 20
48
EM barrel C 56 224 20
EM end-cap A 35 138 12
EM end-cap C 35 138 12
HEC 6 24 2
FCal 4 16 2
Muon spectrometer
MDT
Barrel A 48 48 4
154Barrel C 48 48 4
End-cap A 48 48 4
End-cap C 48 48 4
CSCEnd-cap A 16 16 2
256End-cap C 16 16 2
Level 1
Calorimeter
CP 4 8 1
28 ?JEP 2 8 1
PP 8 32 3
Muon RPCBarrel A 16 16 2
12Barrel C 16 16 2
Muon TGCEnd-cap A 8 8 1
6End-cap C 8 8 1
MUCTPI 1 1 1XXX
CTP 1 1 1
Total 928 1586 148 1500
Table 45. Numbers of readout links (ROL’s), readout buffer modules (ROBIN’s) and readout systems(ROS’s) per detector TTC partition.
– 209 –
90 Mbyte/s to the total event-building rate and aggregate bandwidth. For initial operations, forty-eight SFI’s are deployed allowing a sustained event-building rate of approximately 2.0 kHz, for anaverage event size of approximately 1.5 Mbyte/s.
In addition to the features given in Table 44, the PC’s for theSFO functionality are each5765
equipped with three RAID controllers each managing eight 500 Gbyte SATA II disks. The threesets of disks are operated as a circular buffer: while eventsare being written to the event streamsof one set of disks, a second set of disks is used to send data toCERN’s central data recordingservice. In this configuration, a single RAID controller does not perform both writing and readingoperations simultaneously, thus maximising the throughput of an SFO. The deployed set of SFO’s5770
fulfil the final design specifications: a sustained output bandwidth of 300 Mbyte/s and a peak rateof 600 Mbyte/s, thus for an average event size of 1.5 Mbyte this gives a sustained event rate of200 Hz.
Of the thirty-six HLT racks available for initial operations, eight racks (248 nodes) are dedi-cated to the event filter selection while the remaining twenty-eight racks (868 nodes) can be config-5775
ured to perform either the L2 trigger or event filter selection, i.e. the amount of computing powerapportioned to the L2 trigger and or the event filter will be adjusted according to the conditions.The baseline apportioning of these nodes envisages nine racks (279 nodes) for the L2 trigger andtwenty-seven racks (837 nodes) for the event filter. On the assumption that algorithm processingtimes and rejection rates match those given in the HLT/DAQ Technical Design Report, this system5780
should handle a L1 trigger rate of approximately 40 kHz, i.e.fifty per cent of the final design spec-ification. Although it is too early to confirm these assumptions, the latest studies indicate that theyshould be attainable.
7.5 Detector control system
In order to enable coherent and safe operation of the detector, a Detector Control System (DCS) has5785
been defined and implemented [178]. It has the task to set up the detector hardware in a selectedstate and to continuously monitor its operation. The overall architecture is shown in Fig. 139. TheDCS consists of two parts: the front-end systems and the back-end control stations.
The front-end connects to the detector hardware, and the equipment to supervise ranges fromsimple sensors to complex devices like software-controlled power supplies. A small set of commer-5790
cial devices has been selected as a standard, for example crates and power supplies. A general pur-pose I/O concentrator has been developed, called Embedded Local Monitor Board (ELMB) [179].It comprises a multiplexed ADC (64 channels with 16-bit resolution), 24 digital I/O lines and a se-rial bus to drive external devices. The ADC part of the ELMB board can be configured for varioustypes of sensors. A micro-controller pre-processes the data (e.g. calibration, threshold detection)5795
and then transfers them via CANbus to the back-end. The ELMB is designed and tested to beradiation tolerant to a level of about 1 Gy per year and can hence also be used in the detector itselfat places shielded by the calorimeter.
The back-end is organised in three layers: the global control stations with human interfacesin the ATLAS control room for overall operations, the sub-detector control stations for stand-5800
alone operation of a subdetector, and the local control stations for process control of sub-systems.All stations are PC’s running the commercial controls software PVSS-II [180], which providesthe supervisory functions needed, such as data analysis with the possibility to trigger pre-defined
– 210 –
Figure 139.Architecture of the DCS.
procedures, data visualisation and archiving, and interconnection of all stations via the network,thus forming a distributed system. This software system hasbeen chosen for all LHC experiments5805
under the framework of the Joint Controls Project, which provides on top of PVSS-II a set of toolsand software components for the standardised devices.
The local control stations read, process, and possibly archive the data from specific sub-systems. They execute the commands coming from the layers above and can also implementclosed-loop control.5810
The sub-detector control stations provide full stand-alone operation of a subdetector. Theycomprise a user interface in order to supervise and coordinate the different sub-systems. The de-tailed controls functions of the subdetectors are described in previous chapters of this paper.
The global control stations in the top layer provide all functions needed in the ATLAS con-trol room to run the detector as a whole. The Operator interface presents the full detector with a5815
finite state machine tool which models the hierarchical organisation of the detector down to indi-vidual devices. It allows navigation in this tree-like structure for both showing detector status andexecution of high-level commands. A typical display is shown in Fig. 140, giving the status ofsub-systems and trend-plots of critical quantities. The status system has similar functionality as theoperator interface, but without the possibility of commands. The data viewer provides selection5820
and plotting of all data available in DCS. The alarm system collects and displays all parameterswhich are outside of pre-defined ranges, classified in three severities: warning, error, and fatal. Thefunctions discussed so far are normally restricted to the ATLAS Controls Network, but can on re-quest be opened to selected stations on the internet. General status information is openly availablevia the Web. An information server collects data from systems other than the subdetectors and5825
publishes them for use by the subdetectors.
Each subdetector has one sub-detector control station and an additional one called commoninfrastructure control has been set up to supervise the common environment and services. Each
– 211 –
Figure 140.Graphical operator’s interface, showing status information of sub-systems and data plots.
of the five geographical zones defined for the common infrastructure control (the three countingrooms underground, the cavern of the experiment, and the computer rooms upstairs) has a local5830
control station attributed to it. In each zone, the environmental parameters such as temperatures,humidity and pressure are monitored, and for the electronics racks the operational parameters aremonitored and the electricity distribution is controlled.A large network of I/O points, consisting ofabout 100 ELMBs, has been installed in the detector volume. It reads the data of radiation monitors,the movement sensors of the FPIAA system, and some 200 temperature sensors positioned at the5835
support structure of the barrel toroid. Read-out capacity for future upgrades is available.Apart from data read by its local control station, the commoninfrastructure control uses also
data collected via the network by the information server in the global-control-station layer fromseveral systems. Data include summary information and status of the electricity distribution for theexperiment, the operational parameters of all subdetectorgas systems and of the liquid-argon and5840
helium cryogenics systems. All status information of the magnets is read in this way. And finally,the data to be exchanged in either direction between ATLAS and the LHC accelerator will use thismethod.
The DCS slso comprises common software tools and packages, used by subdetector controlsand the common infrastructure control. A configuration database stores the settings needed for5845
the different operational modes of the detector. All statusinformation and measured data can betransferred to the ATLAS-wide conditions data base (COOL).Another software package allowsthe synchronisation and information exchange between DCS and the data acquisition system.
Emphasis has been put on using common software packages (most notably PVSS-II) and stan-dardised hardware devices (e.g. the ELMB is used by all subdetectors) for the control of all subde-5850
tectors and the common infrastructure. This has enabled full integration of all detector componentsinto one system, and it provides the shift operator with a coherent interface to the experiment.
– 212 –
8. Integration and installation
8.1 Introduction
The ATLAS experiment is located in Switzerland at point 1 on the LHC ring, directly opposite5855
the main entrance to the CERN site. The soil conditions here were favourable for the excavationof the huge cavern needed to accommodate the large ATLAS experiment. Once the experimentwas approved, work began on the preparations for the infrastructure: civil engineering, electricalpower distribution, cooling water, ventilation, etc. Thiswork was handled by the CERN engineer-ing support groups responsible for each of the infrastructure items in liaison with the technical5860
coordination team of ATLAS. In parallel to the work on the infrastructure, the ATLAS technicalcoordination team carried out the studies on the assembly, integration, layout, support structures,services (pipes and cables), safety issues and access requirements/means for the experiment.
The construction of the component parts for ATLAS was distributed over many institutionsaround the world. The components had then to be brought to CERN in a timely manner, a consid-5865
erable challenge in itself in terms of the size, complexity and fragile nature of many of the detectorcomponents. In most cases, final assembly and testing was made at CERN on the surface, priorto installation underground. ATLAS technical coordination was in charge of monitoring the pro-duction at these various institutions and then of organising and implementing the assembly of theexperiment together with the people responsible for the production of the different sub-systems.5870
The experimental cavern is 100 metres underground and the experiment is almost as largeas the underground cavern in which it is housed. The challenges encountered have been similarto those relevant for the assembly of aship in a bottle, but on a much larger scale. Supervisingthe construction of such a complex project meant that ratherformal and uniform management andengineering tools had to be used to monitor and document the progress of the project and ensure5875
that items arrived on time and satisfied the requirements. Ensuring that all the pieces of the puzzlefitted together turned out to be a particularly difficult challenge, since the physics goals and thegeometry of the experiment require minimal clearances between neighbouring parts. One of themost stringent requirements of the ATLAS detector is to ensure hermetic coverage over most of thesolid angle: installing the detector had therefore to be made to great accuracy, in order to guarantee5880
optimal performance in terms of coverage.After a brief description of the organisation of the technical coordination team and of the pro-
cesses and tools used to fulfill the infrastructure, integration and installation tasks (section 8.2),this chapter briefly describes the mechanical integration (section 8.3), the overall infrastructure andservices at point 1 (section 8.4), the support and access structures (section 8.6), safety issues (see5885
section 8.5), the detector installation process (section 8.7), the detector opening and access scenar-ios (section 8.8), the beam pipe (section 8.9) and the interfaces to the LHC machine (section 8.10).
8.2 Organisational issues
The ATLAS project involves many people that are spread around the world. It has also generateda huge, complex and multi-disciplinary amount of data that needs to be organised and shared in an5890
easy and transparent way. In order to help manage the design,production and installation phasesof the ATLAS project, the technical coordination team has developed organisational processes andcomputing tools [181], which are described briefly below.
– 213 –
8.2.1 Organisational processes
The design phase of the project required the production of drawings, schedules and specifications5895
for procurement. Numerous meetings took place to enable theevolution of the project towards theprocurement and production phase. A number of review processes were included during this phaseof the project. The goal of these reviews was to evaluate the feasibility and technical validity ofthe proposed designs. In addition to these reviews, and before launching the production of majoritems, internal design reviews and production-readiness reviews were organised. Then, during the5900
production phase of the sub-systems, production-advancement reviews were implemented to checkprogress and compare it with the production milestones in the schedule. In case of specific andmajor technical issues, experts were called in from within and outside the collaboration to solvesuch problems in a timely way.
Tool Function
Engineering data management system(EDMS)
Structured storage and retrieval of engineering data
CERN drawing directory (CDD) Processing of technical drawings
Project progress tracking (PPT) Regular Web-based notification and reporting system
Equipment management database (EMD)and manufacturing and test folder (MTF)database
Traceability of all equipment installed in the cavern
Rack wizard Configuration of electronic rack connectivity(from detector to counting room)
ATLAS 3D editor Monitoring of ATLAS sub-system displacements(also uses survey data)
Cable database Assist cable installation team(labels, routing, connector specifications, etc)
Table 46. List and function of the various computing tools used by ATLAS technical coordination duringthe installation of the experiment
The ATLAS Technical Management Board (TMB) meets on a monthly basis and provides a5905
forum for regular reporting of the status and problems in allareas relevant to the work at point 1to the collaboration scientists and engineers. During these meetings, the installation schedule isdiscussed and proposed future strategies are agreed.
During the installation of ATLAS, the progress and status ofthe work was monitored on aweekly basis in dedicated meetings. These meetings were held with each main sub-system and the5910
minutes and action items from each meeting are stored in EDMS.
8.2.2 Organisational computing tools
The Web-based tools shown in Table 46 were used to assist in the communication and organisation
– 214 –
process.
8.3 Mechanical integration5915
The mechanical integration process had to address both static issues related to installation andsurvey of major detector components and dynamic issues related to detector placement, movementsof parts during installation (see section 8.8.2) and for access and maintenance (see section 8.8).This section deals with the mechanical integration aspectsrelated to installation.
The mechanical integration process defined the overall experimental layout, where each nested5920
sub-system has its well-defined shape and position and has nooverlaps with any other sub-system.This integration process included several steps:
• Initial input for the positioning of the sub-systems provided by the physicists;
• Determination of the space needed for access and services;
• Definition of the mechanical envelopes, as described below;5925
• Definition of the overall three-dimensional layout of the ATLAS detector using most of thetools described in Table 46;
• Continuous checking procedures and frequent identification and resolution of conflicts be-tween sub-systems using in particular the computer-aided design packages listed in Table 46.
8.3.1 Envelopes (individual, global, dynamic)5930
The concept of envelope defines the space allocated to each part of each sub-system. Three typesof envelope have been created, individual, global and dynamic, and they are defined as follows:
• The individual envelope is the space allocated for the manufactured object. There is alreadysome contingency space added to the nominal design drawing envelope, in order to take intoaccount fabrication and assembly tolerances.5935
• The global envelope includes, in addition to the individualenvelope, some space dedicated tothe inaccuracy of the positioning inside the experiment andthe deformations applied duringinstallation and operation.
• The dynamic envelope includes, in addition to the global envelope, space for deviations anddeformations during displacements (e.g. during access) ofthe object inside the experiment.5940
After the manufacture and installation of each sub-system,the envelopes were checked andcompared to the measurements performed by the survey team. Envelopes have been created as 3Dobjects with the help of various CAD systems. During the project it has been necessary to upgradeto more recent CAD software packages and this has required the translation of many hundreds ofmodels from one system to another with the associated checksthat this translation is made correctly.5945
All this work on modelling and conflict-checking has been most important in order to facilitate theinstallation process and avoid cost and schedule problems between real objects during installation.
– 215 –
8.3.2 Survey and placement strategy
The survey team is responsible for checking that each of the detector components conforms to themanufacturing dimensional tolerances, is placed within its approved space envelope, and conforms5950
to its initial position after opening and closing.
The task of surveying for the ATLAS experiment has been a verychallenging one due to thesize, nature, complexity and global scale of the work. The survey team has been following theevolution of ATLAS from the manufacturing/assembly phase to the installation phase at point 1.Adding to the complexity is the fact that ATLAS is being assembled in a relatively small cavern5955
and thus every centimetre was important. The aim was to optimise any space made available oncean installation was complete. Thus, as soon as the cavern wasdelivered to ATLAS by the civilengineers, and before any infrastructure was installed, anexhaustive scanning was carried out inorder to check the as-built work. This work has been an important ingredient in the optimisation ofthe layout of the experiment.5960
8.3.2.1 Survey reference grid in the cavern The nominal beam line was defined and used duringthe installation and positioning of the detectors in the cavern. It is defined by the best-fit alignmentline of the low-beta quadrupole magnets, located at a distance of 30 m from either side of theinteraction point. This reference line can deviate from thereal beam line by as much as 2-3 mm. Itis determined by the reference sockets in the tunnel, from which the machine elements are installed.5965
The final control is carried out on the machine elements themselves (inner focusing quadrupoletriplets included) relative to each other. A range of spatial uncertainty within 0.5 mm to 1.2 mm at1 sigma was estimated for any fiducial mark with respect to thenominal beam line depending onthe location of the given target.
The datum (interaction point, radial orientation of the colliding beams and reference plane)5970
is given by the initial geometry in the tunnel and the final positioning of the low-beta quadrupolemagnets. The survey grid reference in the cavern is linked tothe machine geometry by stan-dard geometrical measurements and permanent monitoring systems. These include hydrostaticand wire positioning capacitive sensors, implemented in the survey galleries and joining the low-β quadrupoles via the cavern and radial tubes [182]. The reference grid will thus be monitored5975
throughout the lifetime of the experiment.
8.3.2.2 Stability measurements of the floor and the bed-plates Civil-engineering calculationsindicated possible vertical floor movements of up to 6 mm settlement due to the loading of theexperiment and a 1 mm per year lift due to excavation heave. ATLAS has a very limited adjustmentcapability once the detector elements have been placed in-situ. A placement strategy was therefore5980
developed to position all elements within the best achievable mechanical tolerance, relatively to theinteraction point and the nominal beam line [183].
To monitor these predicted movements, regular and periodical measures have been carried outon about 20 reference marks embedded in the cavern floor. The measurements are referred to themachine levelling and deep references in the tunnel, full reports have been given since the first5985
measurements in August 2003.
In addition to these measures, a permanent hydrostatic system has been implemented in theATLAS support feet bed-plates. It consists of six capacitive sensing stations monitoring the water
– 216 –
plane in two 25 m long tubes, 55 mm in diameter, parallel to thebeam and linked by a transversaltube. Two additional stations have been installed in the extreme trenches, recognised as stable5990
zones, and linked to the bed-plate system. Altogether, thisis equivalent to a reference water planeof 75 m length, inspected by eight sensors attached to the structure (bed-plate and stable floor)within an accuracy of better than 20 microns.
The results from the measurements on the floor, as displayed in Fig. 141, show that, after aninitial stable period, during which presumably the heave ofthe floor was balanced by the loading,5995
a global heave of the floor of up to 1.2 mm happened in the central part of the cavern betweenMarch 2004 and March 2006. By August 2006, approximately 85%of the total load had beeninstalled and, over the past 12 months, the cavern floor seemsto have settled more or less and theupward heave is no longer visible.
Figure 141. Vertical movements of the ATLAS cavern floor, as measured as afunction of time in variousreference points since August 2003 (see text). Thez axis represents the longitudinal axis along the beamand positivezpoints towards side A of the cavern. Thex axis represents the radial axis perpendicular to thebeam and positivex points towards the centre of the LHC ring. The beam interaction point is at the origin(x = 0, z = 0). The labels used for the reference points are the following: C23-7 represents a point on side Cat a distance of 23 m longitudinally from the interaction point and of 7 m radially opposite to the centre ofthe LHC ring. In contrast, A23+10 represents a point on side Aat the same longitudinal distance from theinteraction point, but at a distance of 10 m radially towardsthe centre of the LHC ring. The points labelledB are located right in the middle of the cavern, nominally atz = 0. Quality of figure to be improved. Fewerlines should be shown.
The hydrostatic system in the bed-plates gives immediate movements of the supporting struc-6000
ture with an accuracy of a few microns and has been used to monitor local movements when
– 217 –
inserting the calorimeters. It will now be operated continuously to monitor movements in real timeduring data-taking.
8.3.2.3 Placement of ATLAS sub-systemsThe ATLAS detector had to be assembled under-ground in the experimental cavern mostly because of the nature of the barrel toroid magnet struc-6005
ture, which is 26 m long and 20 m in diameter. As discussed above, once assembled and cabled,the experiment and its main components cannot in practice beadjusted anymore relatively to thenominal beam line. Careful consideration had therefore to be given to the optimal placement ofsub-systems during the assembly process. Many sub-systemswere prepared on the surface, whileothers were prepared in the experimental cavern, prior to their installation in their final position.6010
The aim was to place all the detectors such that they will all be at their appropriate positions relativeto the nominal beam line, once the experiment is complete. Inthe process of defining the initialplacements, the following input had to be taken into account:
• floor movement in the ATLAS experimental cavern (see above);
• deflections of the barrel toroid structures as they were loaded with the muon chambers and6015
the services passing throughout ATLAS;
• deviations from as-built dimensions with respect to the nominal ones for neighbouring sub-systems;
• deviations from nominal of the relative placements of different components, which wereinstalled as one assembly. For example, the solenoid and theLAr barrel electromagnetic6020
calorimeter were assembled on the surface inside their common vacuum vessel with a non-negligible relative error with respect to their theoretical placement. When placing the wholeassembly in the cavern, the priority was therefore given to the solenoid regarding the relativeimportance of nominal placement for ATLAS operation;
• uncertainties on the position of the interaction point fromthe uncertainties on the closed-orbit6025
calculations and on the placement of the machine components.
Once an assembly of different sub-systems was completed, the relative position of its compo-nents was usually fixed and could not be changed anymore. As the most prominent example, thebarrel calorimeter assembly consists of: the barrel tile calorimeter, itself assembled from 64 indi-vidual tile modules (see section 5.3.1), the barrel LAr cryostat (see section 5.1.2), which includes6030
the solenoid (see section 2.1.1), the barrel LAr electromagnetic calorimeter (see section 5.2.2) andthe support rails for the inner detector, located on the inner bore of the cryostat.
The strategy for the placement of a given assembly was the following:
1. determine the best position for the assembly, taking intoaccount the input described above;
2. after the initial placement was completed, survey the assembly as installed and perform ad-6035
justments wherever possible to come as close as possible to the optimal position;
3. once a major component has been placed, update the envelope drawings to possibly takeadvantage of the as-installed envelopes of the various components. In most cases, the com-ponents respected the assigned envelopes, and it was possible to recuperate some space,
– 218 –
which was used either to increase the stay-clear area between assemblies and/or to optimise6040
further the position of subsequent assemblies. In rarer cases, such as the barrel and end-capinner detector assemblies, the distance between the assemblies had to be increase by 5 mmwith respect to nominal. These deviations will have to be accounted for in the final detectordescription of the as-installed experiment.
Assembly Component Placement accuracy Comments
Feet Support rails and barrel toroid
Rails 2 mm Support barrel and end-cap systems
Barrel toroid 10 mm (all coils) Magnetic axis aty = -8 mm
Barrel system Solenoid 1 mm Magnetic axis aty = -0.5 mm
Barrel system EM calorimeter 3 mm Axis at y = -2.2 mm
Barrel system Cryostat 2 mm Axis aty = 1.5 mm
Barrel system Tile calorimeter 3 mm Axis at y = -1 mm
ID barrel SCT 1 mm
ID barrel TRT 1 mm
ID end-cap side A SCT
ID end-cap side A TRT
ID end-cap side C SCT
ID end-cap side C TRT
ID pixel
ID beam-pipe
Table 47. Placement accuracy in transverse plane for different components of the ATLAS barrel system.The actual positions iny are given as the most important illustration of the priorities set in the placementstrategy between conflicting requirements from different components of the barrel system.Missing numbersfor ID to be put in for final draft.
Table 47 shows the placement accuracies achieved in the transverse plane for the main com-6045
ponents of the ATLAS barrel system relative to the nominal beam line. Table 47 also shows theexpected offsets in the vertical direction, illustrating the priority given to the placement of thesolenoid as close as possible to the nominal beam line.
The placement accuracy along the beam line is also in the few mm range, but there are ex-ceptions in some cases due to conflicts in envelopes between neighbouring sub-systems. The com-6050
ponents of a given assembly are grouped together under the same heading to indicate clearly thattheir relative position with respect to each other was determined prior to final installation. Theserelative positions are in many cases more precise than the overall placement accuracy of the as-sembly, as for example in the case of the SCT and TRT, or even asin the case of the pixels withrespect to the ID barrel, for which special care was taken to adjust the fixation points between the6055
two sub-systems on the surface after assembly of the complete ID barrel in order to make sure thatthe geometrical axis of the pixel sub-system will be within 100 microns of the nominal beam line.
– 219 –
In a similar way, albeit less complete because of installation work still ongoing in the end-capregions of the experiment, Table 48 summarises the positioning status of the end-cap calorimeters,the small wheels, the big wheels, the end-cap toroid magnetsand the end-wall chambers. All the6060
end-cap detectors can in principle be adjusted relatively to the nominal beam line, when the ATLASexperiment is in the open position. It is intended to keep therelative position of all the detectorsthe same after opening and closing of the experiment. To thisend, an active optical system hasbeen installed, which will provide a precise monitoring (20microns in the transverse plane and100 microns along the beam) of the relative position of thesecomponents with respect to the barrel6065
system.
Assembly ComponentPlacement accuracy Comments
End-cap toroid Side A 6 mm Align magnetic axis with barrel toroid
End-cap toroid Side C 6 mm Align magnetic axis with barrel toroid
End-cap calorimeter side A EMEC 2 mm FCal determines overall placement
End-cap calorimeter side A HEC 2 mm
End-cap calorimeter side A FCal 1 mm Align FCal symmetrically around beam-pipe
End-cap calorimeter side A TileCal 3 mm
End-cap calorimeter side C EMEC 2 mm FCal determines overall placement
End-cap calorimeter side C HEC 2 mm
End-cap calorimeter side C FCal 1 mm Align FCal symmetrically around beam-pipe
End-cap calorimeter side C TileCal 3 mm
Small wheel side A TGC
Small wheel side A MDT
Small wheel side C TGC
Small wheel side C MDT
Big wheel side A TGC1
Big wheel side A MDT
Big wheel side A TGC2
Big wheel side A TGC3
Big wheel side C TGC1
Big wheel side C MDT
Big wheel side C TGC2
Big wheel side C TGC3
End-wall MDT side A
End-wall MDT side C
Table 48.Placement accuracy in transverse plane for different components of the ATLAS end-cap systems.Missing numbers and comments to be put in for final draft. Format to be improved. Possibly merge withprevious table once numbers are all there.
– 220 –
8.3.2.4 Monitoring of the placement of the experiment The placement strategy explainedabove has been quite successful and all major ATLAS components are located well within theinitial target of being aligned to within a few mm from the nominal beam line at the start of data-taking.6070
More detailed information on the location of individual modules is kept in the various databasesand will be updated as movements are monitored over time. As an example, Fig. 142 shows therelative locations in 3D of the individual barrel tile calorimeter modules.
It will also be necessary to monitor the global movements of the experiment to understandpossible future deviations in position between the experiment and the actual beam. In the case of6075
such deviations at the level of 1 mm, the beams can be adjustedand steered, using the field in thetriplet magnets and/or adjusting the position of the last magnet. In the event of larger movements,it will be necessary to adjust the magnets over the last 300 m on either side of the interaction point.It has been agreed with the accelerator experts that such an exercise can be done once every threeto four years.
Figure 142. Illustration of individual barrel-system components in 3D, as obtained from the installationdatabase.This is a placeholder (with only the tile calorimeter) for a complete figure illustrating the as-installed displacements of the major components of the ATLAS barrel system.
6080
8.4 Infrastructure and detector services
This section describes the infrastructure for the surface and underground areas and the services,fixed and mobile, which are connected to the ATLAS experiment[184].
– 221 –
8.4.1 Civil engineering
Figure 143. Layout of surface buildings and of access shafts to the ATLASunderground cavern at point 1.This is a placeholder which needs labelling of buildings after which the corresponding text will be amendedand probably shortened.
The civil engineering works for the ATLAS experimental areastarted in November 1997 whilst6085
the previous accelerator (LEP) was still in operation and the situation remained so until the endof 1999. The works included the excavation and concreting oftwo new shafts, two new largecaverns plus the linking galleries, and the erection at the surface of six new buildings, as shown inFig. 143.
The underground works included the excavation and concreting of the following:6090
• The PX14 and PX16 shafts, respectively 18 m and 12.6m in diameter, both 60 m deep.
• The PX15 shaft was an existing shaft that required concreting only
• The main experimental cavern UX15 (50 m long, 30 m wide and 35 mhigh), which housesthe ATLAS detector. Due to the continued operation of LEP in 1999, the vault had to beconcreted before the support walls were put in place, leading to a non-standard anchoring6095
technique of the 7000 tonne roof of the UX15 cavern. UX15 was delivered to CERN for theinstallation of the infrastructure in June 2003.
• The counting room and services cavern USA15, that houses theelectronic racks and servicesthat need to be close to the detector and is accessible duringmachine runs.
– 222 –
• The seven linking galleries for personnel circulation and the distribution of services between6100
the UX15 and USA15 caverns.
The new surface buildings include the following:
• SX1 is a steel frame building located on top of the main shaftsand housing the travellingcranes, maximum 280 t capacity. Used as short term storage ofdetector components, priorto their lowering into the experimental cavern. It also actsas a shelter from the external6105
environment and allows better control of the conditions in the cavern. This building wasthe first to be structurally achieved to allow the infrastructure works to take place in dryconditions.
• SDX1, located on top of the personel access shaft, is used forthe personnel and materialaccess control to the underground areas. It also contains the uninterruptible power supplies6110
for the detector services, an electrical sub-station, and the DAQ room.
• SUX1, the ventilation building, containing the water chillers and air-conditioning units, forthe underground areas and the new surface buildings.
• SCX1, the main ATLAS control building, used during the detector assembly phase by TCnas an engineering design office.6115
• SH1, a concrete building as it contains the noisey cryogeniccompressors and other cryogenicequipment
• SF1, the new coolings towers for the final dissipation of heatrecovered from all the under-ground areas and surface buildings via the water cooling circuits.
• SY1, the first building at the entrance to the site used by CERNsite control personnel.6120
8.4.2 Electrical power distribution
The total electrical power required by the ATLAS site at point 1 is about 13 MW, which has resultedin the installation of a new 66 kV substation, the reorganisation of the existing 18 kV sub-station,and a new 3.3 kV sub-station. Some 21 new transformers with a total power of 42 MVA havebeen installed both on the surface and underground to bring the 3 kV or 400 V/230 V power to all6125
ATLAS systems.There are different power networks to deal with the appropriate levels of continuous and safe
operation:
• Assured power: the normal power is backed up by two diesel generators of 1 MVA, eachlocated on the surface;6130
• Secured power: the technical solution is as for assured power, but it is not interrupted whenemergency stop buttons are activated;
• Un-interruptible power: these very critical systems are fed by battery back-ups with a totalpower of approximately 700 kW.
– 223 –
8.4.3 Air-conditioning and cooling systems6135
The global environmental requirements in the bulk of the ATLAS underground cavern are drivenby general considerations and by more specific ones from the muon spectrometer chambers. Thetemperature should remain stable at 25±3C and the humidity should be between 25% and 60%.The cooling and ventilation installation was one of the firstinfrastructure items to be installed,once the surface buildings had been handed over to CERN in order to rapidly provide ventilation to6140
the underground areas. This installation included new cooling towers in SF1, the liquid chillers inSUX1, 16 air-handling units in SUX, SDX, SX1 and USA15 and kilometres of water pipes and air-ducts. The system is able to cope with emergency cases in fast-extraction mode (see section 8.5).
Figure 144.Air ducts installed for ventilation in the shaft and experimental cavern.
One of the most challenging tasks was the installation of airsupply and extraction ducts in thePX14 shaft, as shown in Fig. 144, and on the vault of the UX15 cavern, this system having the task6145
of extracting the 200 kW of heat released into the air of the cavern by the ATLAS detector. Theair-conditioning system has been up and running since summer of 2004. Provision has been madefor thermal screens to be incorporated inside the experiment to prevent too large a temperaturegradient across the inner layer of muon chambers.
Table 49 summarises the characteristics of the main coolingsystems used for extracting the6150
heat from the detector itself (silicon sensor leakage currents and ionisation in the TRT gas), fromthe electronics on-detector and off-detector, and from dissipation in the power cables. Most cool-ing systems are leakless by design and the inner detector haschosen fluorinert systems to minimise
– 224 –
risks to the detector in case of leaks. In the case of the pixeland SCT sub-systems, a novel andcomplex evaporative system has been designed and brought into operation to minimise the amount6155
of material devoted to cooling pipes, fluids and connectors on the detector itself. The detector-specific cooling systems are installed in the USA15 side cavern and have been brought into opera-tion in 2005 and 2006.
System Medium Capacity Channel count Operating temperature
Tile calorimeter Water 55 kW 24 cooling loops supplying 17 to 22C
256 tile fingers
LAr calorimeter Water 250 kW 24 cooling loops supplying 17 to 22C
60 electronic boxes
Diffusion pumps Water 50 kW 12 cooling loops supplying 14 to 19C
26 diffusion pumps
Muon spectrometer andWater 300 kW 26 cooling loops supplying CSC’s 17 to 22C
general-purpose stations plus racks for small and big wheels
ID evaporative C3F8 60 kW 4 distribution areas supplying -30 to 10C
(pixel and SCT) 204 cooling channels
ID mono-phase (TRT) C6F14 70 kW 4 distribution areas supplying 14 to 22C
176 cooling channels
ID mono-phase (cables)C6F14 80 kW 32 distribution manifolds 14 to 22C
placed all over the detector
Table 49.Overview of main characteristics of the major cooling systems operating in the ATLAS cavern.
8.4.4 Gas distribution
The ATLAS detector requires a variety of gases for its normaloperation. Table 50 summarises the6160
characteristics of the main gas systems used for normal operation.The surface gas building was designed for the storage, distribution, and mixing of inert and
flammable gases in accordance with the CERN regulations. Thebuilding was completed in 1997.Large quantities of liquid gases are stored, both inside andoutside this building, in particularN2,CO2,Ar,Xe,He,C2H2F4,CH4,C4H10,nC5H12. The building is also fitted with an anti-deflagration6165
roof, mechanical ventilation, permanent gas extraction and gas detection, and an alarm system.The mixing room contains the mixing systems for the different sub-detectors. The TRT, MDT’s,CSC’s and RPC’s use non-flammable gases. Only the muon TGC’s will use flammable gas (CO2-nPentane).
The underground installation consists of a large network ofstainless-steel pipes, which convey6170
the gases to the gas room located in the USA15 cavern and then to the gas racks in the main cavern.From the gas racks, many kilometres of pipes have been installed. They connect to all the differenttypes of muon chambers and to the inner detector.
More details required here concerning the operation of the various gas systems.
– 225 –
System Gas mixture Pressure (bar) Volume (m3) Channel count Impurity limits
TRT active gas Xe/CO2/O2 1.005 2.5 (detector) 48 for barrel < 100 ppmCF4
(normal operation) 70/27/3 2.0 (gas system) 14 per end-cap
TRT active gas Ar/CO2/CF4 < 100 ppmH2O
(cleaning mode) 70/26/4
TRT ventilation CO2 1±0.001 2.9 48 < 100 ppmH2O
(flushed to atm.) (barrel)
TRT cooling CO2 1±0.001 6.0 4 < 1%N2
(closed loop) (end-cap) < 100 ppmH2O
SCT ventilation N2 1.004 4 < 350 ppmH2O
(flushed to atm.)
ID ventilation CO2 1.0005 13 < 1%N2 in ID
(flushed to atm.)
MDT active gas Ar/CO2/H2O 3 (abs.) 800 112 for barrel < 100 ppmO2
93/7/(300–1000 ppm) 81 per end-cap < 100 ppmH2O
CSC active gas Ar/CO2/CF4 3 2 32 < 100 ppmO2
30/50/20
RPC active gas C2H2F4/i−C4H10 2 18 128 < 1%O2
97/3 < 100 ppmH2O
TGC active gas CO2/n-Pentane 2 16 128 for active < 5 ppm at entrance
55/45 56 forCO2 purge of n-pentane liquefier
Table 50.Overview of main characteristics of the gas systems operating in the ATLAS cavern.Need to addmaximum allowed leakage rate to atmosphere. Numbers available for ID but not yet for muon chambers.
8.4.5 Cryogenic systems6175
The ATLAS detector includes two independent systems requiring cryogenic technologies: the su-perconductive magnets and the liquid argon calorimeters. The cryogenic systems for the magnetsand the LAr detectors have each been divided into three parts:
1. External cryogenics, which comprise all the equipment needed to provide the required cool-ing capacity at given temperature levels, including refrigeration plants and infrastructure.6180
This equipment is located on the surface in the SH1 building and underground in the USA15cavern. Also, six large helium gas storage tanks (3m diameter, 21 m long) have been installedon the surface behind the SX1 building.
2. Proximity cryogenics, which is all the equipment linkingthe internal cryogenics to the ex-ternal cryogenics. This equipment is located in the main experimental cavern, on the HS sur-6185
rounding structure.
– 226 –
3. Internal cryogenics, which contain all the devices located inside the system concerned (mag-nets or liquid argon calorimeter).
The cryogenics systems for the magnets are described in somedetail in section 2.1.4.1, so thissection is devoted to a brief description of the cryogenics systems for the LAr calorimeters.6190
Figure 145.Layout of the underground external and proximity cryogenics lines for the LAr calorimeters.
The primary cooling source for the liquid argon calorimeterinstallation is a 20 kW nitrogenrefrigerator, which operates at 80 K. The compressor station is placed in the SH1 surface buildingand the cold box in the USA15 underground side cavern. The high- and low-pressure gas linesconnecting these two items pass through the PX-15 shaft. Thecold box delivers its cooling powerto a 15,000 litre phase-separator dewar placed in the main experimental cavern. Figure 145 shows6195
the underground layout of the proximity and external cryogenics for the LAr calorimeters (shownas if installed alone for convenience). One can clearly see the fixed cryogenic lines supplyingthe barrel calorimeter at the top and also the cryogenics lines in the flexible chains, which supplythe two end-cap calorimeters and which follow them wheneverthey have to move for access andmaintenance of the detector.6200
Two 50,000 litre liquid nitrogen storage tanks placed on thesurface will supply, in the case ofproblems with the nitrogen refrigerator system, liquid nitrogen to the phase-separator dewar via a283 meter long transfer line. A cryogenic centrifugal pump circulates the liquid nitrogen from thephase-separator dewar through the thirteen heat exchangers placed in the liquid argon cryostats.Each cryostat has been equipped with a valve box, which regulates the mass flow and pressure of6205
– 227 –
the liquid passing through each of the individual heat exchangers. These valve boxes are placed onthe steel structure (HS) that surrounds the experiment.
The gaseous nitrogen coming from the heat exchangers is returned to the phase-separatordewar and from there returned to the ANRS or, in case it is not in function, vented to the surfacethrough a 120-metre long gas line. The three cryostats, placed at the heart of the experiment, are6210
linked by large-diameter argon lines to their individual expansion vessel placed on the HS structure.The liquid/gaseous argon boundary of each of the cryostats is located in these expansion vessels.
The need to move the calorimeter end-cap cryostats over a 12 mdistance required the im-plementation of a movement system for the argon and nitrogenlines connecting these cryostatswith their expansion vessels and the nitrogen regulation valve boxes. These movement systems are6215
located on the HS structure (see section 8.6) and are described in section 8.8.The 83 m3 of liquid argon present in the cryostats can, in the event of problems, be emptied by
gravity into two 50 m3 argon storage tanks placed at the lowest point of the experimental cavern.A DN500 safety valve line collects the any gas coming from thepressure safety valves placed onthe cryostats or storage tank volumes venting it to the surface.6220
8.4.6 Racks and cables
Prior to the start of the civil engineering work, provision had to be made for the distribution ofcables from the experiment to the electronic racks in the counting rooms. This included:
• the arrangement of the side cavern USA15 with provision for 250 racks on two floors,equipped with a 2.5 MW water-cooling system;6225
• two main cable distribution galleries, which connect the main cavern to the counting roomwith provision for 44 cable trays, each 600× 100 mm2 in size;
• the distribution of these 44 cable trays on the HS supportingstructure around the perimeterof the experiment and to the HO supporting structure on the end-walls of the cavern (seesection 8.6).6230
At the time, the detailed cabling and rack needs were not known. The consolidated data fromthe various sub-systems was provided only much later and thecorrect provisions were thereforemade just before the start of the installation work. The following provisions were made in additionto the initial ones:
• locations for 70 racks in another side cavern (the US15 cavern) together with about 20 holes,6235
each 300 mm in diameter, through the 2 m-thick connecting wall for the passage of cables,with the associated cable-tray distribution system in the existing false floor and with a dedi-cated cooling system of 500 kW capacity;
• locations for 100 racks in the main cavern, supported from the HS steel structure, whichsurrounds the experiment;6240
• locations for 100 racks in a self-contained room in the SDX surface building with a 500 kW to-tal water-cooling capacity and a dedicated 100 kW air-conditioning system.
– 228 –
The various electronics units for the detector are thus installed in racks, implanted in USA15,UX15 and US15 in the underground areas, and in the DAQ room on the surface. The total numberof racks (electronics, gas and water) for ATLAS amounts to approximately 500. The Rack Wizard6245
tool mentioned in section 8.2 will be an essential tool to monitor the evolution of the racks withtime. The power requirements, specific contents and the connections will be constantly updated asthey evolve: this is required not only for the maintenance ofthe experiment and the understandingof its evolution on the long term, but also to meet INB (Installation Nucléaire de Base) regulations(see section 8.5).6250
Figure 146. Quantities of cable and flexible-pipe bundlesinstalled by the cabling team.
It was necessary to design newtypes of racks with respect to what ex-isted in the previous projects, due to theincreased power consumption, whichtherefore required more cooling capac-6255
ity. It was also found useful to increasethe width of some racks, leading to theproduction of 4 types of standard racks(from 46 to 57 units in height and from900 to 1000 mm in depth, the width be-6260
ing 600 mm). Since some of these racksare installed rather close to the ATLAStoroidal magnets, where the stray fieldis about 1000 Gauss, magnetic shield-ing was developed in order to protect the cooling motors. Newplastic fans were also adopted to6265
avoid eddy currents.
Most of the installed cable trays are made of stainless steelto minimise the perturbations tothe magnetic field (see section 2.2.2). A total of about 50,000 cable bundles, 3000 flexible pipesand 3500 metallic pipes (see Fig. 146) were installed over a period of two years (May 2005 toMay 2007). Many additional proximity cables were installedby the individual sub-systems. Space6270
had to be found to route the large quantity of cables and pipesof the inner detector and barrelcalorimeter systems through the muon spectrometer; this was accomplished by routing most ofthem radially outwards atz = 0 and at fixed azimuthal locations, as illustrated in Fig. 147. All therelevant cable and pipe data are stored in the cable databaseand constantly updated for the samereasons as those described for the racks above.6275
8.4.7 Drag-chains and mobile services
Many of the ATLAS sub-systems need to move away from the run position in order to allow accessinto the experiment. In addition, the calorimeter end-capsneed to remain in a cold bath of liquidargon, hopefully for the duration of the experiment. The toroid end-caps are cooled with liquidhelium and these are also to remain cold during the movement to avoid the lengthy cool-down and6280
warm-up periods (20 to 40 days).
In order to satisfy the above requirements it was decided to use so-calleddrag chainsthatallow for the services to be supported in a flexible structure. They are used as follows:
– 229 –
Figure 147. Drawing of layout and routing of cables and services for the ATLAS barrel system.Not clearwhether this is the best figure to illustrate the complexity of this issue.
• For each calorimeter end-cap there are three chains, as shown in Fig. 148, non-standard com-mercial products, each around 30 m-long with parts in stainless steel, specifically developed6285
for the application. They also have a force-assist system that enables the chain to be pulledback into its stored position when the calorimeter is being closed. Two of the three chains areat 45 downwards with respect to the beam axis. This was a particularly challenging task.
• For each muon inner station (small wheel) there are four chains, each about 3 m in lengthinside the experiment.6290
• For each toroid end-cap there are two chains. These are in aluminium and have been designedspecifically for the task, unsupported over a 9 m length at 24 mfrom the floor.
8.5 Built-in safety features
Fire and cryogenic fluid leaks are the main risks in the underground areas of ATLAS. In additionto standard fire fighting means, such as portable fire extinguishers and hose reels, it was decided6295
to implement a foam extinguishing system in the vault of the cavern. This foam system will, ifnecessary, be used in an extreme case to protect the experiment and the CERN firemen in the eventof a fire getting out of control. The system consists of 12 large blowers installed in the vault of theUX15 cavern which are fed by a mixture of water and detergent and can fill-up the cavern in lessthan 15 minutes, suffocating any fire. Since this foam has only a 1/1000 water content, personel6300
– 230 –
trapped in the foam should survive without problems until the foam settles (approximately onehour). Also, tests have shown that the foam does not penetrate into electronics racks.
To limit the dangers caused by leaks of cryogenic fluids, three large trenches have been builtin the floor of the cavern. In case of a leak, the cryogenic liquids and the cold and heavy gaseswill be contained in these trenches. Access is restricted tothese areas and there is an oxygen6305
deficiency detection system installed. In normal conditions, air is permanently extracted from thelowest point of these trenches. If a leak is detected, the gasextraction can be increased to a massiverate of 90,000 m3/h.
Figure 148.The end-cap calorimeter on side A in itsfully open position with all three drag chains and theflexible LAr fill-line connected.
All simulations have shown that the coldvapour cloud would then be contained in gen-6310
eral in the trenches. In the most dramatic case,with a massive leak of liquid argon spillingfrom the calorimeters, the cold vapour cloudwould not develop higher than 5 m above thecavern floor.6315
In order to be safe, even in this worst casescenario, the normal circulation galleries arelocated 10 m above the cavern floor (in par-ticular the gangways used by visitors) and allemergency evacuation routes point upwards6320
away from the dangerous zones.A number of gas, oxygen deficiency haz-
ard (ODH) and smoke detectors and sniffershave been installed to trigger on any poten-tially dangerous situation.6325
This safety section is still incomplete.There needs to be a few paragraphs on INB!
8.6 Support and access structures
Introduction missing.
8.6.1 Feet and rail system6330
The feet and rail system is shown in Fig. 149 shortly after completing the installation in the pit andbefore the lowering of the first barrel toroid coil. This system is the main support, the back-bone,of the ATLAS detector. It is made of nine pairs of feet, bound by girders that altogether support thetwo bottom coils of the barrel toroid magnet. On top of these feet are two rails, and their supports,on which the central part of ATLAS can slide. The total load, which the feet and rail system has6335
to cope with is about 6000 tonnes (of which approximately 1000 tonnes correspond to the barreltoroid, which is only supported from the feet). The feet and rail systems are mounted on bed-plates,which give the detector its 1.23% incline with respect to thecavern floor, an angle which matchesthe inclination of the LHC accelerator tunnel.
Since the toroid coils are placed inside the feet, there is a strong requirement for the material6340
to be non-magnetic. In addition, the total deformation was to be kept minimal, and stress well
– 231 –
Figure 149.The ATLAS feet and rail system after installation and prior to the installation of the first barreltoroid coil. Also shown are the blue steel surrounding structures (HS and HO), and, in the background, oneof the orange HF trucks.
below the elastic limit. Low-carbon austenitic stainless steel was chosen for its good mechanicalproperties and very low magnetic permeability. One of the main technical issues has been to pro-duce non-magnetic welds for such a huge amount of welded joints (up to 15 tonnes of filler metalin total).6345
In order to obtain a very precise and reproducible geometrical path of the loads, during themovement of the sub-detectors on their air-pad movement systems (see section 8.8.2), and also topreserve the integrity of the beam vacuum system (see section 8.9), the flatness requirement on therails was one millimetre over their total length (more than 25 m), and 0.2 mm over any length ofone metre. The maximum deflection of the system remains below1 mm during the movement of6350
the loads.
The integration with the barrel toroid magnet and with the muon spectrometer required numer-ous interactions with the respective communities. The requirement to preserve maximum accep-tance for the muon spectrometer resulted in special chambers in the region of the feet and additionalchambers alongside the rails. Numerous improvements in thefeet design were introduced to cope6355
with constraints from muon alignment, with various designsof the muon support rails, as well aswith the barrel toroid magnet instrumentation and contact surfaces.
– 232 –
8.6.2 Trucks
The so-called HF trucks are normal steel structures, which are placed directly below the two shaftsof the experimental cavern. They allow for the main components of the experiment to be low-6360
ered underground using the surface cranes (with a total capacity of 280 tonnes) and remain theretemporarily before they are moved into their final position,either by using the cavern travellingcranes (as for the barrel toroid magnet coils) or by using theair-pad movement systems (as for thecalorimeters).
These structures also have the role of supporting the end-cap toroid magnets, as well as the6365
end-cap calorimeters, in the opening sequences of the ATLASdetector (see section 8.8). Duringinstallation, they also support the forward shielding, whereas they will only support part of it whenit will be inserted in its final location. They are therefore able to cope with the 1000 tonnes of maxi-mum static load from the barrel calorimeter and they have to allow for the translation of the end-captoroid magnets and of the forward shielding away from the beam (total weight of 400 tonnes). For6370
such movements, air-pads will be attached to the base of these structures to allow them to slideover the cavern floor.
8.6.3 Surrounding structures (HS and HO)
Figure 150. The blue support structures (HS on the sides and HO at the endsof the main cavern) at thebeginning of ATLAS installation. The arches which now connect the two sides of HS at the top of theexperimental cavern were left out at the time for the installation of the barrel toroid magnet.
– 233 –
The blue HS structures, which surround the ATLAS detector, as shown in Fig. 150, havedual roles of providing personnel access to the periphery ofthe detector and to support all the6375
equipment that has to be located close to the detector: proximity cryogenics, electronic racks, gas-distribution racks, electrical switchboards, services distribution lines (gas, water, coolants, power).These structures were the most tricky to assemble, since they are very close to the detector incertain places and since they had to be assembled in two stages: the large pillars and gangwaysup to a height of 20 m were installed at the same time as the HO structures, but the top of the6380
HS arches was installed only after the completion of the installation of the barrel toroid magnetand with very little margin left. The two structures, which span distances of more than 20 m inthree dimensions had to match each other to within less than 2cm. The last arch was finally andsuccessfully installed at the end of 2005.
The main role of the HO structures, which are to be found at thetwo ends of the cavern, are6385
to support the end-wall muon chamber stations. They also serve as a useful means of assemblingthe sectors of the muon big wheels, before they are hung from the rail system. These structuresalso serve as viewing platforms for the thousands of visitors to the ATLAS cavern . Approximately1000 tonnes of normal steel have been used to build these 13-storey-high structures.
8.6.4 Muon barrel access structures6390
The purpose of the aluminium access platforms inside the barrel muon spectrometer is multi-fold:
• permanent access inside the barrel toroid is required, so that the muon chambers and theirservice connections can be accessed in a short intervention(for example, to disconnect thegas supply to a specific chamber;
• permanent access to the patch-panels of the inner detector (PP2’s), so that they can also be6395
accessed in a short intervention
• an emergency exit (through sector 1) in case of access to the barrel calorimeter;
• access for the installation of the barrel muon chambers;
• access to the vacuum pumps of the barrel toroid.
8.6.5 Big wheels6400
The muon spectrometer (see section 6.1) includes four big moving wheels at each end, each wheelmeasuring 25 metres in diameter (see Fig.151). Of the eight wheels in total, six will be composed ofthin gap chambers (TGC) for the muon trigger system and the other two will consist of monitoreddrift tubes (MDT’s) to measure precisely the position of themuons. The so-called big wheelscomprise aluminium structures, which support the muon end-cap chambers. These big wheels6405
resemble bicycle wheels and are made of sectors, which have been pre-assembled on the surfaceprior to their transport to the cavern, where they were assembled on the end-wall HO structures.Once one of the wheels is been completed on the HO structure, it is lifted onto the traction system,which allows it to move longitudinally towards the barrel toroid magnet and reach its final positionin the closed configuration of the experiment. It is important to note that the big wheels in their6410
– 234 –
final position need to be inclined with a slope of 1.23% with respect to the vertical to account forthe angle between the horizontal cavern floor and the inclination of the machine tunnel.
Figure 151. One of the assembled TGC big wheels in the ATLAS cavern. The chambers are fixed to analuminium structure, which was pre-assembled into sectorson the surface and then assembled as a completewheel in the cavern itself.
8.7 Detector Installation
The installation of the detector can be sub-divided into sixmain phases [185], which are brieflydecribed below. The main cavern has three zones: side A (towards the airport) , B (barrel) and6415
side C (towards the Jura and Charlie’s pub in Saint-Genis). The barrel toroid magnet, once installed,occupies the central region of the cavern leaving the two sides, A and C, for the lowering andassembly of the remaining large detectors and magnets.
8.7.1 Phase 1: infrastructure in the main cavern, feet and rails
Side A, Barrel and side C6420
The experimental cavern was handed over to ATLAS in May 2003.The first operation was toinstall the general infrastructure (metallic structures around the cavern walls, temporary electricityand lighting, ventilation ducts, and the overhead travelling cranes).
With the steel structures installed, the first elements of the ATLAS detector to be broughtdown were the bed-plates, which were bolted to the concrete cavern floor. After the bed-plates,6425
the stainless steel support feet, 18 in total, were lowered one by one and installed. The main
– 235 –
rails were installed and surveyed once positioned on the feet. The feet provide the mechanicalsupport for most of the ATLAS sub-systems, namely the barreltoroid magnet, the calorimeters, themuon chambers, the toroid end-cap magnets, the services andthe access structures, amounting toabout 5000 tonnes.6430
8.7.2 Phase 2: barrel calorimetry and barrel toroid
Side A: barrel toroidThe first barrel toroid magnet coil was delivered to Point 1 inOctober 2004. The coil with its
weight of 100 tonnes and total length of 25 m, was lifted by thesurface crane, tilted with hydraulicwinches, lowered, in an almost vertical orientation, through the 18 m diameter shaft down into the6435
cavern. It was then turned back to the horizontal orientation, before lowering it onto the temporarysupports (see Fig. 150). From there, it was picked up by the two 65 tonnes underground travellingcranes and put into its final position inside the ATLAS feet. Once the magnet coils were in position,the aluminium struts and girders were installed so that the next coil could be attached to them.This process was repeated until the completion of the assembly. In parallel with the barrel-toroid6440
assembly, the first 100 muon barrel chambers were installed in between the struts/voussoirs and theATLAS feet.
Side C: barrel calorimetersThe lower part of the tile calorimeter was lowered in March 2004. Individual tile calorimeter
modules were then assembled together, one by one, until 32 ofthe 64 modules were completed.6445
The LAr barrel calorimeter cryostat was then lowered into this half-cradle in October 2004, asshown in Fig. 152. The tile module assembly was then continued until the mechanical assembly ofthe full barrel calorimeter was completed.
Barrel: completion of barrel toroid and calorimeter instal lationThe last aluminium girder was put in place in September 2005,completing the mechanical6450
assembly of the barrel toroid magnet structure. Then the hydraulic jacks, which were supportingthe complete structure during the assembly, were released.At this moment the load was transferredfrom the extensive temporary supporting structure (used during the magnet assembly) to the supportfeet at the bottom. The temporary support structure was thencut and removed to give space for thebarrel calorimeter, which was moved inside the magnet in October 2005.6455
8.7.3 Phase 3: end-cap calorimeters and muon barrel chambers
Side C: end-cap calorimeterWith the barrel calorimeter installed inside the bore of themagnet, the space on side C was
now vacant for the assembly of the first end-cap calorimeter.This assembly was very was similarto that of the barrel and was finished in January 2006. It was then moved inside the magnet in6460
February 2006, once the installation of the services (pipes, cables etc.) was completed.
Side A: barrel muon chambersOn side A, the first of the 656 barrel muon chambers was installed in February 2006. When
the assembly of the second end-cap calorimeter started in March 2006 it had to be carried out inparallel with the muon-chamber installation. The second end-cap calorimeter was mechanically6465
completed in May. It stayed outside the BT for a further two months for the installation of the
– 236 –
Figure 152. Lowering of the barrel LAr calorimeter down to the cavern in October 2004. The first barreltoroid coil can also be seen on a temporary support platform before it is installed in the cradles of the feet.
services while the side C calorimeter was moved to the run position and the magnetic field of thesolenoid was switched on and measured in June ’06.
8.7.4 Phase 4: muon big wheels, inner detector and completion of muon barrel
Side C: big wheels6470
In April 2006, work started on the first end-cap muon middle station (most often referred to asa big wheel) with the mounting of the tooling on the end-wall structure (HO). The first sector wasinstalled in July 2006. Work progressed with an average rateof two sectors per week and this firstwheel was mechanically completed in September 2006. After installing the services, the wheel wasreleased from the end-wall structures and moved against thebarrel magnet in November 2006.6475
In March 2007, the second of four big wheels was completed andthe first wheel was thenopened for the lowering of the inner detector end-cap C. Alsoall the remaining barrel muon cham-bers were installed before closing the end of the barrel on side C with the completed big wheels.Side A
After finishing the solenoid field mapping, the barrel section of the inner detector was lowered6480
and installed inside the bore of the barrel cryostat in August 2006. While the work on the connec-tions of the inner detector services continued, the end-capcalorimeter A was moved partially insideto allow space for the completion of the muon barrel chambers. By the end of December 2006,600 chambers or 90% of the total, had been installed (see Fig.153).
– 237 –
Figure 153. View of installed barrel muon spectrometer stations and of installed calorimeter end-cap onside A.
In January 2007, the preparations for the muon big-wheel assembly started. The first sector6485
was installed in March 2007, because the end-cap calorimeter needed to be moved to the openposition to allow the lowering of the first inner detector end-cap.
Barrel: inner detector end-caps and pixelsThe installation of the barrel muon chambers continued in parallel during the assembly of the
first muon big wheel. Also the services installation for the inner detectors, calorimeters and muon6490
chambers continued. In May and June 2007, the two inner detector end-caps and the pixel detectortogether with the central VI section of the beam-pipe (see section 8.9) were successively loweredinto the pit, as shown in Fig. 154.
8.7.5 Phase 5: end-cap toroid magnets and muon small wheels
The two end-cap toroids were successfully lowered onto the trucks in June-July 2007, as illustrated6495
in Fig.155. The muon end-cap inner stations or small wheels are in the process of being assembledto the shielding disks on the surface.
8.7.6 Phase 6: beam-pipe and forward shielding
The last elements to be installed will be the beam-pipe and the forward shielding and this willrequire that all the sub-systems are progressively moved into their closed positions along the beam6500
axis. During this time, there will be no access to regions in between the sub-systems.
– 238 –
Figure 154. View of barrel calorimeter and inner detector end-flange after installation of the first innerdetector end-cap in early June 2007 (left). This was followed shortly thereafter by the installation of thesecond inner detector end-cap and of the pixel detector withthe central VI section of the vacuum pipe(right).
The last paragraphs above will have to be updated for the finaldraft based on the knowledgewe will have in early October 2007.
8.8 Access and detector opening
8.8.1 Access scenarios6505
This section needs some fleshing out and some pointers to INB regulations, safety to personnel,etc.
The three main access scenarios depend on the duration of theshut-down period and are de-fined as follows:
1. Short accesses are typically of the order of a few hours. Such accesses can be provided6510
immediately after the machine shut-down. They can happen ona daily basis, but are notscheduled. Therefore, no detector components are moved andthe access shaft to the surfaceis not opened (there is therefore no crane access through theshaft). All magnetic fields willstay on.
2. Standard accesses have a duration from a few weeks to five months. The shorter ones will6515
be based on the needs of the ATLAS sub-systems. In agreement with the other sub-systems,the other LHC experiments and the LHC machine, such accessescan be provided for a shortperiod. Standard access is also considered as the standard configuration during the annualLHC shut-down for a period of approximately five months.
During such accesses, the access shaft is opened so that crane access to the surface is possi-6520
ble. The removable elements of the forward shielding (see section 3.2) are brought up to thesurface, while the muon big wheels, the end-cap toroids, thesmall wheels and the end-capcalorimeters can be moved along the beam axis. The beam-pipeis left in place, but at atmo-spheric pressure. All magnetic fields are turned off. A maximum of ten persons are allowedinside at each end of the experiment.6525
– 239 –
3. Long accesses are dedicated to the inner detector and small-wheel removal and installation.Such accesses are also for non-standard interventions, which require a break of the beam-pipe. Their duration is the same as that of the LHC annual shut-down (of the order of fivemonths), but their frequency is expected to be much lower andwill be related to requestsof the experiment for a detector upgrade or for a major maintenance operation. In addition6530
standard accesses, the beam-pipe is dismantled and one of the end-cap toroids is movedsideways. A second truck is installed along the axis of the experiment in order to moveback the corresponding small wheel and lift it to the surface. The corresponding end-capcalorimeter is moved back so that sufficient access is possible to the inner detector. Allmagnetic fields are turned off. The number of people allowed access is defined according to6535
the evacuation plan of the cavern and the detailed operations, which need to be performed.
8.8.2 Movement system
During access, a number of sub-systems move into their position on air-pads: the end-cap calorime-ters, the muon small wheels, the shielding disks and the end-cap toroid magnets. The equipmentfor each detector movement system is basically the same: in the closed configuration, the detectors6540
rest on hydraulic cylinders called blocking jacks. They areequipped with nuts so that the load canbe transferred to solid feet, without the need for oil pressure. During movement, the load is trans-ferred from the blocking jacks to the air-pads, which consist of two main components: a rubberair-skirt, which allows the lifting of the detector on a film of air, and a hydraulic jack, which allowsfor the height to be adjusted to a set limit during the movement. Thus, the detector can slide on the6545
rails of the experiment with a very low friction factor of 0.01. The number of air-pads underneatha sub-system will depend on its weight. They are grouped so that the load is supported by threeiso-static points. The movement itself is provided by two hydraulic cylinders, parallel to the rails,and the detectors are moved step by step according to the stroke of the cylinders.
Because of the sensitivity of the detectors to vibrations, shocks, or tilt, the movement must be6550
smooth and very well controlled. Moreover, the clearance between detectors and the beam-pipe isonly about 15 mm, a distance of similar size to that of the air-pad lift. Therefore a compensation ofthe pneumatic action has been implemented, so that the sub-system under air-lift is not raised bymore than 5 mm. Four height sensors, located on each mobile sub-system, provide feedback to thecontroller, which drives the hydraulic valves of the air-pads.6555
The movement of the sub-systems is further complicated because of the services connected tothem through the drag-chains, as described in section 8.4. Some of these chains are equipped withtheir own movement system, therefore it is necessary to monitor these movements with respect tothose of the main movement system.
8.9 Beam-pipe6560
The beam vacuum system represents the main interface between the experiment and the LHC ma-chine. It must therefore fulfill a dual set of requirements:
• the ATLAS requirements, particularly excellent transparency to particles, limited beam-gasbackgrounds and conformity with environmental constraints, in terms of radiation, electro-magnetic noise and thermal behaviour;6565
– 240 –
• the accelerator requirements, namely safe operation of themachine, adequate beam apertureand static and dynamic vacuum conditions compatible with the ultimate LHC performance.
Figure 155. Lowering of the first end-cap toroidmagnet onto the truck on side A in June 2007. Oneof the TGC big wheels can be seen on the right of thepicture.
The ATLAS beam vacuum system con-sists of seven beam-pipes of 38 m total length,spanning the distance between the two TAS6570
collimators located at each end of the cavern.They are bolted together with flanges to forman ultra-high vacuum system, which can befully baked out in-situ. The central chamber,called vacuum inner detector (VI), is centred6575
about the interaction point. It has an innerdiameter of 58 mm and is constructed fromberyllium metal with a thickness of 0.8 mm.The remaining six chambers are installedsymmetrically on both sides of the interac-6580
tion point and named after the detector, whichsupports them: VA (vacuum argon end-cap),VT (vacuum toroid end-cap) and VJ (vac-uum forward shielding). They are constructedfrom thin-walled stainless steel tubes with di-6585
ameters increasing progressively from 60 mmto 80 mm and finally to 120 mm. Chambersinside different detectors are mechanically de-coupled by vacuum bellows, which also serveto absorb thermal expansion during bakeout.6590
The VI chamber was integrated into thepixel detector on the surface, and installed aspart of the pixel package. It is aligned on thebeam axis using a system of laser and CCDcameras, which measure the chamber defor-6595
mation. The VA chambers are centred insidethe warm bore of the LAr end-cap cryostatsby sliding supports, which allow the detector to move longitudinally along the beam-pipe. Spe-cial minimised ultra-high-vacuum flanges, with only 35% of the volume of a standard flange, havebeen developed to pass through the bore. The VT chambers are held by retractable jack supports6600
on rails in the forward shielding. These can be adjusted fromthe back-face of the end-cap toroidor fully retracted to allow the end-cap toroids to move longitudinally along the beam-pipe. TheVJ chambers are cantilevered from the forward shielding located on the cavern wall, inside a coni-cal support designed to fit inside the opened end-cap toroid.The flanges between the VJ chambersand the TAS collimators are remotely actuated from the outside of the forward shielding, because6605
of the high activation expected in this region at design luminosity.
This supporting system is conceived to allow ATLAS to rapidly move to a standard access
– 241 –
without the need to open the beam vacuum to air and hence re-activate the NEG system (see below).However, the chambers are not able to support the stresses induced by offsets expected duringopening whilst under vacuum. The chambers will therefore bevented to neon gas at atmospheric6610
pressure, purified to the ppb level by a specially developed gas-purifying system mounted on side Aof the HO structure. Neon is not pumped by the NEG system, so the beam vacuum system can bequickly made operational at the end of a short intervention by simply re-pumping the neon gas.
The main pump used to eliminate desorbed gasses in the systemis a non-evaporable getter(NEG) film sputtered onto the whole of the inner surface of thebeam-pipe. After activation by6615
heating the beam-pipe to∼ 200C, this NEG film gives a very high distributed pumping speedfor chemically active gasses [186]. Chemically-inert gasses not pumped by the NEG system areremoved by two minimised sputter-ion pumps [187] at± 3.8 m and by larger pumps at± 19 mfrom the interaction point.
The whole length of the vacuum system is permanently equipped with a mass-minimised sys-6620
tem of heaters, thermocouples and insulation which allow the NEG system to be re-activated annu-ally. This bakeout system consists of polyimide foil heaters wrapped with silica aerogel, polyimidetape and foil for the chambers. Flexible bellows, pumps and transitions are equipped with semi-permanent flexible heating jackets.
Significant optimisation of the forward beam-pipe chambersis planned for operation of the6625
LHC machine at high luminosity, as discussed in section 3.5.Stainless steel will be replaced byaluminium or other low-Z materials wherever possible to minimise both the background radiationin the muon chambers and access problems due to beam-pipe activation.
8.10 Interface to the LHC machine
Table 51 gives some of the basic beam properties for some of the interesting configurations envis-6630
aged for machine operation. Experience from previous colliders shows that the machine-inducedbackground in the experiments is very hard to predict. A number of different factors intervene in acomplex manner:
• the local vacuum pressure as well as the vacuum at more distant places, such as the arcs,affects the halo entering the experiment;6635
• inefficiencies of both the betatron and momentum-cleaning systems and the detailed settingsof the collimators will also heavily influence the observed background levels;
• other factors, which have a direct impact on the beam halo, are of course the total beamcurrent, the beam tune shift and the orbit positions.
It is therefore of prime importance to the experiment to define reliable background indicators6640
and to communicate them to the main control room.These background indicators must be contin-uously available to the operating crew for monitoring on a fill-by-fill level. Such signals have tobe available before stable beam conditions have been reached during the setting-up phase of themachine and they must therefore be available in the experiment independently of the main dataacquisition. The ATLAS BCM system (see section 3.4.1) meetsthese requirements and will be6645
used in this context.
– 242 –
Machine operation configuration Nominal 75 nsRoman pots
Number of bunches 2808 936 43
Number of protons per bunch (1011) 1.15 0.9 0.1
Bunch spacing 25 75 2025
β function (m) 0.55 1–11 2625
Crossing angle (µrad) 285 250 0
Peak luminosity (cm−2s−1) 1034 1033 1028
Table 51.Main operational parameters of the LHC machine for a few configurations: the nominal one (left),the initial one with a bunch-spacing of 75 ns (centre), and the specialised one for forward physics (right).
For safe and optimal operation of both the LHC machine and theATLAS experiment, the twoparties will continuously exchange information about their overall status as well as about the statusof relevant individual sub-systems. This data exchange will be used to synchronise actions duringthe different states of operation, to provide online feedback on tuning operations, and to allow rapid6650
reaction to errors and a quick and efficient process of understanding of their causes.
The LHC communicates to ATLAS the total beam intensity, as well as the individual bunches,the average 2-dimensional beam size, the average bunch length, the luminosity at the four interac-tion points, the average beam loss, and the average horizontal and vertical positions.
ATLAS reports to the LHC a measurement of the total luminosity, the luminosity per bunch,6655
three average rates, rates for individual bunches, and the position and size of the luminous region.This information allows the machine to optimise and monitorthe beam and collision conditions atATLAS.
The 40 MHz bunch clock of the LHC and a revolution pulse is transmitted from the LHCradio-frequency system at Point-4 to ATLAS over a total length of 14 km of optical fibre. Once6660
received in the ATLAS counting room, these signals are fine-adjusted in phase and subsequentlydistributed via the L1 central trigger processor to all ATLAS sub-systems.
ATLAS receives for each beam one signal from a beam-positionmonitor, which is located175 m upstream of ATLAS. These signals provide a precise timing references in order to monitorthe phase of the LHC clock with respect to the bunches. In addition, they serve as inputs to the6665
L1 trigger, for which they provide a time reference with respect to the abort gap of the LHC bunchtrain.
When in operation, the LHC machine undergoes a sequence of operational modes: filling,ramping, adjust, stable beams, unstable beams, are a few examples. The current machine opera-tional mode is received by ATLAS via software, which is appropriate in most cases for synchro-6670
nising ATLAS operation with LHC operation. Before two statetransitions, a hand-shake protocolbetween the LHC and ATLAS is used: the LHC operators request from ATLAS confirmation be-fore going into e.g. the state (adjust mode), where the low-β squeeze and other adjustments takeplace. A similar protocol is used before a scheduled beam dump by the LHC operator.
– 243 –
There are two flags related to the machine modes, which are transmitted from the LHC to6675
ATLAS through a fast, safe and reliable hardware link: the modes labelled as "stable beams" andas "movable devices allowed". These flags are used by the position-beam interlock of the roman-pot system (see section 10.3.3) and for the control of their insertion or extraction.
A fail-safe and reliable beam interlock system is installedaround the LHC ring, with severalsystems giving permission for beams. The absence of a beam-permission signal leads to an im-6680
mediate beam dump: the safe extraction of the beam from the LHC in less than 300µs. WithinATLAS, there are three different systems, which give beam permission: the ATLAS beam interlocksystem, the spectrometer magnet system and the roman-pot-position beam interlock system. TheATLAS beam interlock system takes inputs from the beam conditions monitor (see section 3.4.1)and possibly other detectors and gives beam permission onlywhen background conditions allow6685
a safe operation of the detector. In addition, software interlocks are foreseen, corresponding toinstances when the roman-pot-position beam interlock system gives permission for beams if the"movable devices allowed" flag is true or if the roman pots arein their end-switch position.
If the "stable beams" flag is false and the roman pots are not intheir end-switch position, theyare moved out of the beam. In addition, injection into the LHCis inhibited via a software signal6690
in the cases when the roman pots are not in their end-switch position. This is implemented insoftware, and a future implementation in hardware is possible.
– 244 –
9. Expected performance of the ATLAS detector - 55 pages
9.1 Introduction, 1 page
Since the publication of the ATLAS Detector and Physics Performance Technical Design Re-6695
port [188] in 1999, all the detector components of the experiment have been constructed andintegrated and most of them have been installed. A detailed understanding of their features (ge-ometry, amount of material and placement accuracy) has beenachieved over this period, togetherwith extensive measurements in test-beams, culminating in2004 with the large-scale combinedtest-beam (CTB) effort, which has led to an improved detector description and understanding of6700
the detector performance, and also to a first set of detailed calibration and alignment procedures,essential to the initial understanding of the detector performance and to the extraction of the firstphysics results. The main results obtained from these measurements and from their comparisonto the detailed simulation of the detector (used both for theCTB and for ATLAS itself) will bepresented below wherever relevant.6705
Comments on CTB results to be covered in the relevant places.Main figures illustrating:
• Test-beam setup (in introduction above)
• Alignment of inner detector (in tracking)
• Alignment of muon chambers (in tracking)6710
• Electron measurements with ID and EM calo (in egamma)
• Photon measurements with ID and EM calo (in egamma)
It is assumed that the EM calo stand-alone calibration and uniformity is discussed in thecalorimeter chapter and also that the combined hadronic measurements (pions) for barrel andend-cap are summarised in the calorimeter chapter.6715
End of comments on CTB results.Over the same period of about 7 years, a large and modular suite of software tools for simula-
tion and reconstruction has been developed and integrated into the ATLAS computing model andfirst full-scale exercises of the operation of this computing model have begun.
In the context of the preparation for the first data-taking run, the whole ATLAS collaboration6720
has joined forces to simulate and reconstruct an experimental setup very close to that installed in theATLAS pit, with its imperfect alignment and placement accuracies, its incomplete knowledge of themagnetic field and with its material distortions with respect to the detector description implementedin the Monte Carlo simulation. A large number of high-statistics samples of Monte-Carlo eventshave been run through the complete ATLAS simulation, reconstruction and analysis chain to assess6725
the readiness of the overall system to cope with the initial data. Results from this data challenge,in particular from its calibration and alignment component, will be also presented below.
Results to be included:Alignment results on the scale of ATLAS for ID, muons and combined (?)Overall calibration of EM and hadronic calorimeters6730
– 245 –
Determination of mass scale for ATLAS, E/p measurements, etc. (will this be ready?)
The purpose of this chapter is thus to present a brief synthesis of the main performance featuresof the ATLAS experiment, as expected today from this latest round of simulations and as validatedwherever possible using CTB measurements.6735
Section 9.2 describes the most important tracking performance features of the ATLAS de-tector, both for the inner detector and for the muon spectrometer, as well as for the combinedmeasurements of muons using both systems. Sections 9.3 (electrons and photons), 9.4 (hadronicjets), 9.5 (missing transverse energy), 9.6 (hadronicτ -decays), 9.7 (tagging of heavy flavours)and 9.8 (trigger) describe the expected performance of the overall ATLAS detector to trigger, re-6740
construct, identify and measure the major final-state objects over the required range of energies formost of the physics channels of interest at the LHC. Finally,Section 9.9 presents preliminary ideason how the overall ATLAS energy scale and alignment will be extracted in terms of general toolsand of in situ physics samples used.
– 246 –
9.2 Tracking (17 pages)6745
This section is devoted to the expected tracking performance in ATLAS and its powerful but com-plex magnet system (see section 2). In the inner detector andits solenoidal field, all charged particletracks with transverse momentumpT > 0.5 GeVare fully measured, albeit with limited efficiencyat low momenta because of the large amount of material in the inner detector (see section 4.8and Fig. 53). The overall expected performance of the inner detector is described in section 9.2.1.6750
In the muon spectrometer and its toroidal field, all muons with transverse momentum above afew GeV as they exit from the calorimeter are reconstructed over a solid angle very similar to thatof the inner detector. The overall expected performance of the muon spectrometer, both stand-aloneand combined with that of the inner detector, is presented insection 9.2.2.
9.2.1 Charged particles in the inner detector (11 pages)6755
The intrinsic measurement performance expected for each ofthe inner detector sub-systems isdescribed in Section 4.1. This performance has been studiedextensively over the years, both beforeand after irradiation of production modules, and also more recently, during the full slice test of theATLAS 2004 combined test beam and in a series of cosmics testsin 2006 in the surface building atpoint 1. The results have been used to update and validate themodeling of the detector response in6760
the Monte Carlo simulation.
This section describes the expected performance of the inner detector in terms of tracking,alignment, vertexing and particle identification.
9.2.1.1 Tracking performance for single particles and jets The expected performance of thetracking system for reconstructing single particles and particles in jets is determined from a precise6765
modeling of the individual detector response as illustrated above and of the geometry and passivematerial in the simulation. Great care has been taken to precisely determine the overall materialbudget of the Inner Detector (see section 4.8 for details andfigures).
Figure 156 shows thepT resolution as a function of|η | for 1 and 50GeV muons and thepT
resolution as a function ofpT for different slices in|η |. The transverse and longitudinal impact6770
parameter resolutions for charged pions are shown in Figs. 157 and 158. Figure 159 shows the effi-ciency and fake rate as a function of|η | for single muons and for pions in jets fromH → bb decays.
9.2.1.2 Alignment of the inner detector An accurate and detailed understanding of the align-ment of the inner detector and of its residual systematics will be a crucial ingredient for the initialunderstanding of the performance of the experiment. This section briefly explains the strategies6775
adopted to achieve this difficult task and the current statusin this rapidly developing area, whereresults from test-beam measurements and from alignment algorithms applied to large portions of arealistically misaligned inner detector have been recently obtained.
Alignment in combined test beam (CTB) A number of alignment strategies were tested,optimised and partially validated on the small set-up used in the CTB. Some of the results are6780
shown in Fig. 160 and summarised in Table 52.
– 247 –
pT resolution versus|η | for different values ofpT
pT resolution as a function ofpT for different
slices in|η |
Figure 156. Transverse momentum resolution as a function of|η | for muons withpT = 1 and 50GeV(left) and as a function ofpT for different slices in|η | (right).
d0 resolution vs eta for 1 and 50 GeV pions d0 resolution vs pt for different slices in eta
Figure 157. Transverse impact parameter (d0) resolution for pions withpT = 1 and 50GeV as a functionof |η | (left) and as a function ofpT for different slices in|η | (right).
(z0 sin theta) resolution vs eta for 1 and 50 GeV
pions
(z0 sin theta) resolution vs pt for different slices in
eta
Figure 158. Modified longitudinal impact parameter (z0 × sinθ) resolution for pions withpT = 1 and50GeV as a function of|η | (left) and as a function ofpT for different slices in|η | (right).
Alignment in ATLAS The various alignment algorithms are in the process of quitestringenttests using high-pT muon tracks simulated using a significantly misaligned geometry representationof the inner detector [189]. The current status of these studies is shown in Figs. 161 and 162.
– 248 –
Efficiency and fake rate vs eta for muons and pions
in jets
Efficiency and fake rate vs DeltaR from jet axis for
pions in jets (for NewTracking and iPatrec)
Figure 159. Efficiency and fake rate as a function of|η | for single muons and for charged pions in jetsfrom H → bb decays (left). Also shown (right) are the efficiency and fakerate for the same pion sample asa function of the distance∆R to the jet axis for two different reconstruction strategies.
... ... ...
... ... ...
Table 52.Main results obtained from CTB in terms of inner detector alignment
Evolution of some of the alignment parameters of
one of the algorithms as a function of iteration
number
Momentum resolution obtained in the CTB for one
alignment algorithm compared to simulation and
to the result with the initial alignment
Figure 160. For charged pions of various momenta in the combined test beam, evolution of some align-ment parameters from one algorithm versus iteration number(left). Also shown (right) is the momentumresolution obtained after alignment compared to simulation and to the result from the initial set of alignmentconstants.
9.2.1.3 Vertexing performance Vertexing tools constitute an important component of the higher-6785
level tracking algorithms. The primary vertex resolutionsfor WH with H → bb (mH = 400GeV)and forH → γγ (mH = 110GeV) with and without pileup are shown in Fig. 163. The secondaryvertex resolutions along the direction of flight forτ hadronic 3-prong decays fromZ → ττ and forB→ J/ψK0
s are shown in figure 164.
9.2.1.4 Particle identification, reconstruction of electrons and photon conversions The re-6790
construction of electrons and of photon conversions is a particular challenge for the inner detector,since most electrons lose a significant fraction of their energy and approximately 40% of the pho-
– 249 –
Plot of eigenvalues obtained for the full inner de-
tector system (including TRT and beam spot con-
straint?)
Z mass peak after various stages of the alignment
Figure 161. The eigenvalue spectrum obtained for the alignment of the full inner detector without externalconstraints (left) and theZ → µµ mass resolution at various stages of the alignment (right).
Show the E/p ratio for electrons and positrons us-
ing various misalignment scenarios
Show linearity obtained after alignment with re-
spect to external muon momentum measurement.
Figure 162. A comparison of the inner detector momentum scale obtained after internal alignment forelectrons and positrons (left), for which theE/p distribution is shown, and for muons measured in the muonspectrometer (right), for which the ratio of the measured momenta is shown.
Primary vertex resolution in Rphi for Hbb and Hgg Primary vertex resolution in Z0 for Hbb and Hgg
Figure 163. Primary vertex resolution in the transverse plane (left) and along the beam (right), as expectedfor WH with H → bb (mH = 400GeV) and forH → γγ (mH = 110GeV). The results are shown withoutand with pileup at a luminosity of 1033 cm−2 s−1.
tons convert into an electron pair in the inner detector material before reaching the calorimeter, asillustrated in Fig. 165. The transition radiation tracker plays a specific role to identify electrons and
– 250 –
decay length resolution for tau to 3 pi decay length resolution for B to JPsiKs
Figure 164.Secondary vertex resolution along the direction of flight for Z → ττ 3-prong decays (left) andfor B→ J/ψK0
s decays (right).
to reconstruct track segments from secondary electrons from photon conversions or from electrons6795
having radiated a large fraction of their energy in the silicon layers.Figure 166 shows the performance expected as a function of energy of various particles and
the rejection expected against pions as a function of the pion energy for two typical values of theelectron efficiency. The expected performance in terms of bremsstrahlung recovery for promptelectrons and for reconstruction of electrons from photon conversions is shown in Fig. 167.6800
Fractional energy as a function of|η | lost by 25
GeV electrons in ID material
Photon conversion probability in ID material as a
function of|η |
Figure 165. Fraction of energy lost on average by electrons ofpT = 25 GeV as a function of|η | (left)within the pixel/SCT part of the tracker (dashed) and withinthe inner detector volume (solid). Also shown(right) is the probability for a photon ofET = 60GeV to convert within the same regions.
– 251 –
The onset curve for transition radiation, as ex-
pected from simulation and as measured from
CTB data
Pion efficiency vs pion energy for 80% and
90% electron efficiency for the combined particle-
identification in the TRT using transition radiation
and time-over-threshold.
Figure 166.The onset curve (left) for transition radiation as expectedfrom simulation and as measured fromthe combined test beam. The expected overall particle identification performance of the TRT in ATLAS isillustrated (right) through the pion efficiency shown as a function of pion energy for 80% and 90% electronefficiency.
E/p as reconstructed before and after brem recov-
ery for 10 GeV and 25 GeV electrons using both
DNA and calo brem recovery. Include GSF if fea-
sible.
Conversion reco efficiency versus radius, if pos-
sible separated between tracking and vertexing. If
feasible, add also figure showing E/p for converted
photons from pi0 bgd versus photon signal.
Figure 167. Expected E/p distributions (left) for electrons of 10 GeV and 25 GeV transverse momentumbefore and after bremsstrahlung recovery (see text). Efficiency for reconstructing photon conversions as afunction of conversion radius (right) for photons of 20 GeV and 60 GeV transverse energy.
– 252 –
9.2.2 Muon measurements, 6 pages
9.2.2.1 Introduction Muon measurements in ATLAS are a combination of accurate measure-ments in the muon spectrometer, which is also used to efficiently trigger on muons over a widerange of energies and over|η | < 2.7, as described in detail in section 6.5, and in the inner detec-tor. The latter provides the best measurement at low to intermediate momenta, whereas the former6805
takes over above 30GeV. The toroidal field guarantees excellent momentum resolution even at theedge of the acceptance inη , but a complex grid of Hall probes and optical alignment sensors isrequired to reconstruct the actual toroid coil positions (see section 2.2) and to monitor very accu-rately and more or less in real time the displacements of the chambers with respect to the coils (seesection 6.3.4).6810
This section describes the alignment validation results obtained in the CTB and the muonreconstruction performance expected in ATLAS in terms of momentum resolution, track-findingefficiency and mass resolution for selected channels.
9.2.2.2 Calibration and alignment In the case of combined test-beam measurements using a setof barrel muon spectrometer stations, Figure 168 shows track sagittae as measured in the middle6815
station relative to straight tracks from the inner to outer stations before and after the correctionsfrom the optical alignment system have been applied.
Track sagittae measured in CTB with artificially
induced distortions.
Track sagittae measured in CTB with dispersion
of 20 microns after implementing corrections from
optical alignment system.
Figure 168. For the barrel muon chambers in the CTB, track sagittae measured as the displacement of themiddle-station track segment to the straight-line track crossing from the inner station to the outer station viathe middle one. The results are shown for artificial distortions introduced in the relative chamber alignment,before (left) and after (right) applying the corrections derived from the optical alignment system.
9.2.2.3 Muon performance Figure 169 illustrates, forW+W− → µνµν the relative rate of oc-currence as a function of|η | of combined track reconstruction versus stand-alone muon spectrom-eter track reconstruction. In the case of a successful combination between muon spectrometer and6820
inner detector for the reconstructed muon, Fig. 170 shows the resulting pulls for the main recon-structed quantities, namely curvature,η andz.
Figs. 171 and 172 show respectively the momentum resolutionand track reconstruction effi-ciency expected after full calibration and alignment have been performed. The results are shownas a function ofpT andη . Figure 173 shows the impact of random chamber misalignments on the6825
width of reconstructedZ → µµ decays as well as the dimuon mass spectrum expected for such
– 253 –
decays. Figure 174 shows the sensitivity of the muon reconstruction to the cavern background atlow luminosity.
Moving towards physics performance, Fig. 175 shows the overall reconstruction efficiency formuons from various physics channels and also shows the difficult regions in the case ofZ → µµ6830
decays. Finally, Fig. 176 shows the expected mass resolutions for two typical cases of Higgs-bosondecays to muons.
Illustrate for wide spectrum of high-pT muons the
relative fractions of stand-alone versus combined
tracks found
Figure 169. Relative frequency as a function of|η | (left), with which muon tracks fromW+W− events aresuccessfully reconstructed in the muon spectrometer and combined with the inner detector. A small numberof tagged muon tracks, identified as such by the presence of a single segment in the muon spectrometer, canalso be seen.
Pull in curvature Pull in eta Pull in z
Figure 170.Quality of combined muon reconstruction for a sample of muons of pT = 100GeVover the fullη -coverage of the muon spectrometer. Shown are the pulls on the measured curvature (left), pseudorapidityη (centre) and position along the beam extrapolated back to the beam line (right).
9.2.2.4 Muon performance in situ Some results on extraction of performance using tag andprobe techniques with Z→ µµ decays.
– 254 –
Resolution versuspT averaged overηResolution versus|η | (averaged overφ?) for 20,
100 and 1000 GeVpT muons
Figure 171. Muon momentum resolution as a function ofpT and averaged overη (left) and as a functionof |η | for various values ofpT (right).
Efficiency versuspT averaged overη (include
cavern background)
Efficiency versusη (averaged overφ?) for 6, 20
and 100 GeVpT muons (include cavern back-
ground)
Figure 172.Muon reconstruction efficiency as a function ofpT and averaged overη (left) and as a functionof η for various values ofpT (right).
Width of Z → µµ for random 1 mm and 1 mrad
misalignments of barrel toroid chambers
Dimuon mass obtained for stand-alone and com-
bined tracks forZ → µµ
Figure 173. Impact of random chamber misalignments on width of reconstructed dimuon mass spectrumfor Z → µµ decays (left). Also shown is the distribution of the reconstructed dimuon mass for stand-aloneand combined muon tracks (right).
– 255 –
Fake muon rates with cavern background
at 1033 cm−2 s−1 as a function of pT for
Z → µµ , tt and for muons in jets
Figure 174.Fake muon rates as a function ofpT for various physics channels and for the cavern backgroundexpected at a luminosity of 1033 cm−2 s−1.
Efficiency versuspT (or |η |?) for muons from
Z → µµ , tt and for muons in jets and for different
levels of cavern background
Efficiency map versusη andφ for Z → µµ in-
cluding cavern background at 1033 cm−2 s−1
Figure 175. Efficiency for reconstructing muons from various physics processes for different levels ofcavern background (left). The efficiency for muons fromZ → µµ decays is also shown as a function of|η |andφ (right).
Invariant mass distribution obtained for
H → µµµµ decays withmH = 130GeV
Invariant mass distribution obtained forA → µµ
decays withmA = 300GeV
Figure 176. Expected mass resolution for Higgs-boson decays to muons. Shown are the distributions forH → µµµµ with mH = 130GeV (left) and forA → µµ with mA = 300GeV (right).
– 256 –
9.3 Electron and photon identification and measurements (8 pages)6835
This section is sub-divided into three parts:
• the calibration and stand-alone performance of the EM calorimeter;
• the expected performance for electron and photon identification, reconstruction and mea-surements;
• strategies for validation of the performance in situ.6840
9.3.1 Calibration of the electromagnetic calorimeter
Description of the calibration procedure: formulae, samples used, definitions.Description of the different corrections, of their magnitude and of the order in which they are
applied.Illustrative plots ( can be thinned down afterwards):6845
• Energy precorrection used for position corrections.
• Example of S-shape, averaged over eta=0-0.4.
• deta (reco-true) vs eta: scatter plot and fit, also showing eta bins. Four plots, for EMB1,EME1, EMB2, EME2. Also enlargements of edge regions.
• Energy dependence of S-shape: before and after corrections, deta vs eta and just deta, for a6850
few energies.
• Cluster size dependence of S-shape.
• Electron/photon dependence for S-shape.η -position resolution: EMB1, EME2, EMB2, EME2.
• phi offset vs eta, before and after correction.6855
• phi position resolution.
• energy dependence of phi offset.
• electron/photon dependence of phi offset.
• cluster size dependence of phi offset.
• phi modulation binning6860
• phi modulations + fit, before and after correction. for electrons and photons, barrel andendcap, for a few different energies.
• same thing for eta modulations.
• energy ratio vs eta for different energies, cluster sizes, and particle types.
• energy resolution vs energy, with fits.6865
– 257 –
9.3.2 Electron and photon reconstruction and identification
Description of electron and photon candidate definitionDescription of baseline cut (blablaba about shower shapes,tracker, and combined recoDescription of LL, Hmatrix, (neural net?)Discussion of relative performance with reference to tableand figures6870
Tables: Rejections for photons and electrons as a function of cuts and comparisons with multi-variate analyses
Figures:
• H → ZZ∗ → 4l mass plot
• H → γγ mass plot6875
9.3.3 Strategies for validation of performance in situ
Figure: sensitivity of energy flow in EM calorimeter to material in inner detectorFigure: inter-calibration in situ withπ0 → γγ andZ → eedecaysFigure: sensitivity of E/p distribution to material of inner detector (if available)Figure: mapping of material in inner detector using photon conversions (if available)6880
– 258 –
η-5 -4 -3 -2 -1 0 1 2 3 4 5
Ele
ctro
nics
Noi
se (
MeV
)
1
10
210
310
FCal1FCal2
FCal3
HEC1HEC2
HEC3HEC4
PSEM1
EM2
EM3Tile1
Tile2Tile3
η-5 -4 -3 -2 -1 0 1 2 3 4 5
Pile
-Up
Noi
se (
MeV
)
1
10
210
310
410
FCal1FCal2
FCal3
HEC1HEC2
HEC3HEC4
PSEM1
EM2
EM3
Figure 177. Expected electronic noise for the ATLAS calorimeter subsystems as a function of|η | (left).Expected pile-up noise at a luminosity of 1034 cm−2 s−1 as a function of|η | and for each calorimetersubsystem (right).
Pseudorapidity
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
(G
eV)
nois
eσ
0
2
4
6
8
10
12
14
Jets from towers
Jets from topo-clusters
E (GeV)
0 100 200 300 400 500 600 700
310×
E (GeV)
0 100 200 300 400 500 600 700
310×
Num
ber
of c
ells
0
5000
10000
15000
20000
25000| < 0.7η0 < |
TowerNoiseTool
CaloTopoClusters
| < 0.7η0 < |
TowerNoiseTool
CaloTopoClusters
Figure 178. Average noise contribution to cone 0.7 jets reconstructed from towers (open circles) andtopoclusters (full circles) as a function of|η | (left). Total number of cells contributing to cone 0.7 jetsbuilt from towers (full circles) and topoclusters (triangles) versus jet energy (right).
9.4 Jet reconstruction, 8 pages
9.4.1 Jet clustering and reconstruction
9.4.2 Particle jets
9.4.3 Calorimeter jets
The value of the width of the electronic noise is obtained by summing in quadrature the contri-6885
butions of the electronic noise (Fig. 177, left), which varies by large factors from one calorimetersystem to another, and from pile-up (Fig. 177, right), whichdepends on the luminosity. The overallnoise contribution is decreased by more than a factor of two if three-dimensional clusters are usedinstead of towers to build e.g.∆R = 0.7 cone jets, as shown in Fig. 178 (left). This decrease ofthe noise is explained by the decrease in the number of cells composing the jets, as illustrated in6890
Fig. 178 (right).
9.4.4 Jet calibration
The effects which affect the jet calibration may be divided in two classes : detector-driven ef-
– 259 –
Sigma(E) versus E for pions at eta = 0.35 Sigma(E) versus E for pions at eta = 2.45 (or 1.3?)
Figure 179. Energy resolution for H1 calibrated pions reconstructed using towers (open circles) andtopoclusters (full dots) for|η |= 0.35 (left) and|η |= 2.45 (right).
fects (noise, non-compensation, cracks, dead material, magnetic field effects, pile-up) and physics-driven effects (underlying event, showering effects, clustering).6895
9.4.4.1 Calibration at the level of particle jets Unless otherwise specified, the results presentedbelow include the effect of electronic noise but do not include any pile-up effects.
For the two calorimeter clustering algorithms considered here (towers and topoclusters) andfor the H1 calibration, Figs. 179 and 180 show respectively the expected energy resolution andlinearity for charged pions at two values of|η |. Fig. 181 shows for pions of fixed transverse energy6900
the expected energy resolution as a function of|η |. The same simulation tools applied to pions inthe combined test-beam setup have been shown to reproduce reasonably well the experimental data(see section 5.6), but further work is required to achieve anaccurate description of hadronic showersover the full range of energies available and for all calorimeters (Accordion+Tile, Accordion+HECand FCAL).6905
Figs. 182 and 183 show respectively the expected jet energy resolution and linearity at twovalues of|η |, for two cone sizes and, as a reference, for the total energy deposited in the calorimeter.Figure 184 shows the fitted stochastic and constant terms obtained for the jet energy resolutionusing different cone sizes.
9.4.4.2 Calibration at the level of parton jets A detailed understanding of the absolute jet en-6910
ergy scale will require the use of many high-statistics datasamples, such as jets balanced by ahigh-pT Z boson or photon decaying into leptons orW → j j decays in reconstructed top-quarkdecays, as shown in Figs. 185 and 186.
9.4.5 Jet reconstruction performance
In this section, the jet reconstruction performance is briefly summarised in terms of jet reconstruc-6915
tion efficiency for different algorithms and physics use-cases.
9.4.5.1 Jet reconstruction efficiency Figure 189 shows the jet reconstruction efficiency obtainedas a function of the jetET andη for both algorithms (towers and three-dimensional clusters).
– 260 –
Energy linearity versus E for pions at eta = 0.35Energy linearity versus E for pions at eta = 2.45
(or 1.3?)
Figure 180.Energy linearity for H1 calibrated pions reconstructed using towers (open circles) and topoclus-ters (full dots) for|η | = 0.35 (left) and|η | = 2.45 (right).
Sigma(E)/E versus eta for pions of 20 GeV ET (or
perhaps 5 GeV?)Sigma(E)/E versus eta for pions of 50 GeV ET
Figure 181. Energy resolution as a function of|η |, for charged pions ofET= 20 GeV (left) and 50 GeV(right). The curves correspond to the energy resolution expected with a stochastic term of 50% and a constantterm of 3%.
Sigma(E) versus E for jets at eta = 0.35 (0.7 cone
from towers and topo and total Ecalo)
Sigma(E) versus E for jets at eta = 2.45 (0.7 cone
from towers and topo and total Ecalo)
Figure 182. Energy resolution for H1-calibrated jets reconstructed using the total energy deposited in thecalorimeter (triangles) and using a cone of size 0.7 based ontowers (open circles) and topoclusters (fullcircles). The results are shown for|η | = 0.35 (left) and for|η | = 2.45 (right).
– 261 –
Linearity versus E for jets at eta = 0.35 (0.7 cone
from towers and topo and total Ecalo)
Linearity versus E for jets at eta = 2.45 (0.7 cone
from towers and topo and total Ecalo)
Figure 183. Energy linearity for H1-calibrated jets reconstructed using the total energy deposited in thecalorimeter (triangles) and using a cone of size 0.7 based ontowers (open circles) and topoclusters (fullcircles). The results are shown for|η |= 0.3 (left) and for|η |= 2.45 (right).
Stochastic term for fittted jet energy resolution in
three cases (total deposited energy, 0.4 cone and
0.7 cone).
Constant term for fittted jet energy resolution in
three cases (total deposited energy, 0.4 cone and
0.7 cone).
Figure 184.Statistical (left) and constant (right) terms of the fitted jet energy resolution as a function of|η |,for the total energy deposited in the calorimeter (full circles) and for cone 0.7 reconstruction (open circles)and cone 0.4 reconstruction (full squares), based on topoclusters.
9.4.5.2 Low-pT jet vetoing In searches for specific exclusive final states, the requirement that noadditional jet be present in the event is often used as a powerful tool to reject certain backgrounds.6920
Figure 190 shows, in the case of a central jet veto (|η | < 2.0) and of a Higgs-boson signalproduced through vector-boson fusion, the expected efficiency for such a veto as a function of thejet ET threshold. Also shown is the expected veto efficiency fortt events, which are one of thedominant backgrounds in this channel.
9.4.5.3 Forward jet tagging In the context of production through vector-boson fusion, jet tag-6925
ging in the forward calorimeters is also an important tool toseparate signal from background.Figure 191 illustrates the expected performance in this difficult region as a function of|η | and as afunction of jetET .
– 262 –
Figure 185.For a sample ofγ+jet events reconstructed using a cone size of 0.7, distribution of relative trans-verse momentum balance between the photon and the leading jet, pbalance
T = (p jetT − pγ
T)/pγT , as a function
of (p jetT + pγ
T)/2. The distributions are shown for calibrated calorimeter jets (full red circles), particle jets(blue triangles) and partons (black squares).
W → j j mass spectrum reconstructed in selected
top-quark decays
Figure 186. Invariant mass distribution of dijet system forW → j j decays selected in top-quark decays, asobtained from reconstructed jets (open circles) and from particle jets (full circles).
– 263 –
Invariant mass distribution obtained forH → bb
decays withmH = 100GeV
Invariant mass distribution obtained forH → hh→
bbbb decays withmH = 300GeV
Figure 187. Reconstructed invariant mass distribution of two b-jets from H → bb decays withmH = 100 GeV (left), obtained from reconstructed jets (open circles) and from particle jets (fullcircles). Reconstructed invariant mass distribution of four b-jets fromH → hh → bbbb decays withmH = 300GeV (right), obtained from reconstructed jets (open circles) and from particle jets (full circles).
Jet energy scale accuracy obtained versus jetET
Figure 188. Precision with which the jet energy scale can be determined using the bootstrap method (seetext) as a function of the jetET .
Efficiency versusET for reconstructing jets from
towers in variousη -regions
Efficiency versusET for reconstructing jets from
topoclusters in variousη -regions
Figure 189.Jet reconstruction efficiency as a function of jetET for jets reconstructed from towers (left) andtopoclusters (right) in variousη -regions.
– 264 –
Jet veto efficiency versusET (VBF H → WW→
lν j j with mH = 800GeVandtt background)
Figure 190. Comparison of the jet veto efficiency in the central region (|η | < 2.0) as a function of the jetET threshold for a heavy Higgs-boson signal and for the dominant background fromtt pairs.
Jet tagging efficiency versus|η | in forward region
(for a fixed jet energy of 1 TeV?)
Jet tagging efficiency versusET in forward region
(for fixed η = 4.5?)
Figure 191. Jet tagging efficiency as a function of|η | (left) and of jetET (right) in the forward region forvarious jet reconstruction algorithms.
– 265 –
9.5 Missing transverse energy, 4 pages
A very good measurement of the missing transverse energy (EmissT ) is a crucial requirement for the6930
study of many physics channels in ATLAS. A goodEmissT measurement in terms of resolution and
linearity is important for e.g. the reconstruction of the top-quark mass from semileptonic decaysor the reconstruction of the Higgs-boson mass when it decaysto τ -leptons. On the other hand,an incorrect measurement ofEmiss
T could significantly enhance for example the backgrounds fromQCD multi-jet events to a possible signal from supersymmetry or the backgrounds fromZ → ll6935
decays accompanied by high-pT jets to a possible signal from Higgs-boson decay into two leptonsand two neutrinos. This section describes briefly the reconstruction and calibration ofEmiss
T in AT-LAS (section 9.5.1), illustrates the expected performancewith a few examples (sections 9.5.2, 9.5.3and 9.5.4) and finally concludes with a discussion of the possible sources of fakeEmiss
T in ATLAS(section 9.5.5).6940
9.5.1 Reconstruction and calibration ofEmissT
This section describes the various contributions to the reconstruction ofEmissT and also summarises
the method used to obtain the best estimate ofEmissT for a given choice of reconstructed objects to
be used for a specific analysis.
9.5.1.1 Contribution from calorimetry and noise suppression6945
9.5.1.2 Contribution from cryostats
9.5.1.3 Contributions from muons
9.5.1.4 Refined calibration ofEmissT
9.5.2 Evaluation ofEmissT performance
The EmissT performance is evaluated by comparing the final reconstructed and calibrated value6950
of EmissT with the trueEmiss
T for a variety of processes of interest. Of paramount importance insuch a comparison are not only the modulus and direction ofEmiss
T , but also theEmissT resolution
and possible tails in the measured quantities together withthe contributions from the different con-tributions discussed above.
The summary plots showing theEmissT linearity and the resolution for many different channels6955
are shown in Figs. 192 and 193.
9.5.3 Measurement ofEmissT direction
Figure 194 shows the angular resolution, obtained from the r.m.s. of the∆φ distribution betweentrue and reconstructedEmiss
T as a function of the trueEmissT in the event.
9.5.4 Use ofEmissT for mass reconstruction6960
The reconstructedEmissT vector can clearly be used to improve the overall reconstruction of final-
state topologies with only one neutrino in the final state (e.g. in tt events with one hadronic andone semi-leptonic top-quark decay). But, as was demonstrated originally in [190], one can evenuse the reconstructedEmiss
T vector inZ → ττ andA → ττ decays with several neutrinos in the
– 266 –
EtMissTruth (GeV)0 20 40 60 80 100 120 140
(E
Tm
issT
ruth
-ET
mis
sRec
)/E
Tm
issT
ruth
-0.2
-0.15
-0.1
-0.05
-0
0.05
0.1
0.15
0.2
shift vs ETmiss
MET_Topo H1
MET RefinedMET_Topo LHMET_Topo LHTopoCalib
MET_Topo H1
MET RefinedMET_Topo LHMET_Topo LHTopoCalib
MET_Topo H1
MET RefinedMET_Topo LHMET_Topo LHTopoCalib
MET_Topo H1
MET RefinedMET_Topo LHMET_Topo LHTopoCalib
Channel ETmiss
Ztautau 1206 = 17.3
Wmunu 1206 = 33.
Wenu 1206 = 33.4
ttbar 1206= 68.5
SU3 1206= 279.4(140)
Sumet (GeV)0 100 200 300 400 500 600 700
Ex(
y)M
iss
Res
ol
0
2
4
6
8
10
12
14
16
18
20
Refined Calibration: Ex(y)Miss Resol vs Sumet Sumet Ztautau: fit 0.53
Sumet Wenu: fit 0.48
Sumet Wmunu: fit 0.54 Sumet Zee: fit 0.46
(TDR) Sumet 0.46
Figure 192. Linearity of reconstructedEmissT as a function of the trueEmiss
T for different channels (left).Resolution of the two components of theEmiss
T vector as a function of the total transverse energy measuredin the calorimeters for different samples in the case of low to medium values of the total transverse energy(right).
Sumet (GeV)0 500 1000 1500 2000 2500 3000 3500 4000
Ex(
y)M
iss
Res
ol
0
10
20
30
40
50
60
Refined Calibration: Ex(y)Miss Resol vs Sumet Sumet J7: fit 0.64 Sumet J5: fit 0.52
Sumet J4_30mic_12.0.31: fit 0.50 Sumet J3: fit 0.50 Sumet J2: fit 0.49 Sumet J1: fit 0.50 Sumet J6: fit 0.56
(TDR) Sumet 0.46
Sumet (GeV)0 200 400 600 800 1000 1200 1400 1600 1800 2000
Ex(
y)M
iss
Res
ol
0
5
10
15
20
25
30
35
40
Refined Calibration: Ex(y)Miss Resol vs Sumet Sumet J6_1mm: fit 0.54
Sumet SU3: fit 0.58
Sumet tt_1mm: fit 0.57
(TDR) Sumet 0.46
Figure 193.Resolution of the two components of theEmissT vector as a function of the total transverse energy
measured in the calorimeters for different samples in the case of high values of the total transverse energy.The samples used are standard QCD dijet and multijet events (left) and specific samples fromtt productionand supersymmetry (right).
final state to reconstruct the invariant mass of theττ pair under certain simplifying assumptions.6965
The results of such a procedure are shown in Fig. 195 for the reconstruction ofZ → ττ decaysand ofA → ττ decays withmA = 450GeV in the case of a supersymmetric A boson.
9.5.5 FakeEmissT
9.5.5.1 Sources of fakeEmissT
9.5.5.2 FakeEmissT from muons6970
9.5.5.3 FakeEmissT from Calorimeter Figure 196 shows the expected rates of true (i.e. originat-
ing from neutrinos) and reconstructedEmissT for a sample of high-pT dijet events. For events with
a reconstructedEmissT larger than 50GeV, Fig. 197 shows the distribution of|η | for the leading
reconstructed jet in the event.
– 267 –
EtMiss (GeV)0 50 100 150 200 250 300 350
) (r
ad)
ϕ(σ
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
qqνtt -> bb l
EtMiss (GeV)0 20 40 60 80 100
(ra
d)ϕσ
0.2
0.4
0.6
0.8
1
1.2
1.4
N.5146ττZ->
Figure 194. Angular resolution inφ of reconstructedEmissT as a function of trueEmiss
T , as obtained fortt events (left) andZ → ττ events (right).
0 20 40 60 80 100 120 140 160 1800
10
20
30
40
50
60
70
80
lep-had allrec TTmassphi cut
Invariant mass distribution obtained forA → ττ decays with
mA = 450GeV
Figure 195. Reconstructed invariant mass of the pair ofτ leptons with oneτ -lepton decaying to a leptonand the other one decaying to hadons. The results are shown for Z → ττ decays (left) and forA → ττ de-cays (right) withmA = 450GeV in the case of a supersymmetric A boson.
Miss (GeV)TE
100 200 300 400 500 600
1
10
210
310
Fake MET
True MET
Figure 196. For dijet events with a transverse energy between 400 and 600GeV, expected distribution oftrueEmiss
T compared to theEmissT distribution obtained from reconstruction.
– 268 –
FakeEmissT and calorimeter cracks
Figure 197. Pseudorapidity distribution of the leading reconstructedjet for dijet events with a transverseenergy between 400 and 600GeV and with a reconstructedEmiss
T larger than 50GeV.
– 269 –
9.6 Hadronic τ -decays (5 pages)6975
Hadronic decays ofτ -leptons will clearly play an important role at the LHC, especially as probesfor new phenomena spanning a wide range of theoretical models. Based on this motivation, varioustools have been developed to efficiently reconstruct and identify these decays, whilst providingthe required large rejection against the usually overwhelming backgrounds from hadronic jets.The parallel and equally difficult task of triggering on these decays as inclusively as possible is6980
addressed in section 9.8.
9.6.1 Reconstruction and identification of hadronicτ -decays
The overall performance performance of two differentτ -identification algorithms is presented inthis section. The first algorithm [191] relies on calorimeter-based seeds for reconstructingτhadronic decays, whereas the second algorithm [192] relieson track-based seeds and an energy-6985
flow approach. Two specific performance aspects of particular interest for the reconstruction ofhadronicτ -decays are mentioned below before moving to the overall performance.
9.6.1.1 Efficiency and quality of track reconstruction Fig 198 shows the efficiency for recon-structing tracks from single-prong and three-prongτ -decays as a function of thepT of theτ -leptonin the case ofτ -leptons from W and Z boson decays.6990
(GeV)trackT
p0 5 10 15 20 25 30 35 40 45 50
effic
ienc
y
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
events τ τ → and Z ν τ → W
single-prong decays
three-prong decays
events τ τ → and Z ν τ → W
trackη 0 0.5 1 1.5 2 2.5
effic
ienc
y
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
events τ τ → and Z ν τ → W
three-prong decays =1-2 GeV trackT
p
=5-6 GeV trackT
p
=15-25 GeV trackT
p
events τ τ → and Z ν τ → W
Figure 198. Reconstruction efficiency forπ± tracks as a function of track transverse momentum (left) andof |η | (right) for events fromW → τν andZ → ττ samples.
9.6.2 π0 subclusters in single-prong decays
For single-prong decays and for the algorithm combining thetrack measurement with the topologi-cal clustering of energy measurements in the first and secondcompartments of the electromagneticcalorimeter, Table 53 illustrates how efficientlyπ0 subclusters can be reconstructed in events con-taining aρ or a1 meson compared to those only containing a charged pion. Fig.199 (left) shows6995
the resolution obtained by the algorithm for reconstructing the visible transverse energy from theτ -decay. Fig. 199 (right) shows the reconstructed invariantmass of theρ → π±π0 candidate, asobtained from the track and the leading electromagnetic subcluster, for the various single-prongfinal states.
– 270 –
Table 53. For single-prongτ -lepton candidates inW → τν andZ → ττ events, expected multiplicities ofelectromagnetic subclusters.
decay mode no EM clusters1 EM cluster >= 2 EM clusters
all τ → hadν 27.0% 39.8% 33.2%
τ → π±ν 66.4% 20.0% 13.6%
(49.3% of all) (10.1% of all) (8.7% of all)
τ → ρν 16.3% 50.9% 32.8%
(29.2% of all) (62.0% of all) (50.3% of all)
τ → a1(→ 2π0π±)ν 9.8% 37.4% 52.8%
(6.7% of all) (17.4% of all) (27.5% of all)
τT
)/ E visτT- Etrk+clus
T( E
-1 -0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
0.08
0.1
events ν τ → W hist101443Entries 4284Mean -0.02323RMS 0.1656Underflow 0Overflow 0.003268
/ ndf 2χ 0.001437 / 3Constant 0.20422± 0.09219 Mean 0.12646± -0.02078 Sigma 0.19805± 0.05127
hist101443Entries 4284Mean -0.02323RMS 0.1656Underflow 0Overflow 0.003268
/ ndf 2χ 0.001437 / 3Constant 0.20422± 0.09219 Mean 0.12646± -0.02078 Sigma 0.19805± 0.05127
ν ρ → τ > 9 GeV
Ttrack p
> 1 GeV topoForTausT E
(at calo2)> 0.0375 cl, trk
R∆
topoForTaus qualif, multi clus
events ν τ → W
(GeV)cluster+track m0 0.5 1 1.5 2 2.50
20
40
60
80
100
120
140
160
180
200
events ν τ → W hist101436Entries 3982
Mean 0.7056
RMS 0.2346
Underflow 0
Overflow 28
ν ρ → τ ν ± π → τ
ν) ± π 0π 2 → a1(→ τ > 9 GeV
Ttrack p
> 1 GeV topoForTausT E
(at calo2)> 0.0375 cl, trk
R∆ > 15 GeV
ρT E
topoForTaus qualif, multi clus
events ν τ → W
Figure 199. For single-prong decays fromW → τν events, resolution obtained for reconstructed visibletransverse energy (left) and distribution of invariant mass reconstructed from track and subcluster in electro-magnetic calorimeter for various final states fromτ -decay (right).
9.6.3 Identification of hadronic τ -decays and rejection of jets7000
Figure 200 shows the expected performance of the combined calorimetry and inner detector foridentifying hadronic decays ofτ leptons while rejecting most of the hadronic jets. The rejection iscomputed with respect to truth jets reconstructed with the standard cone algorithm.
9.6.4 Initial physics with τ leptons
Table 54 gives the expected numbers of signal and backgroundevents for an integrated luminosity7005
of 100 pb−1 and for topologies corresponding toW → τν , Z → ττ andtt → τν b j jb decays.Figure 201 (left) shows the expected distribution for the charged track multiplicity for acceptedW → τν candidates withEmiss
T > 40 GeV and after vetoingW → eν events which otherwisedominate the single-prong sample. Figure 201 (right) showsthe reconstructed visible mass for
– 271 –
efficiency0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
reje
ctio
n
210
310
410
510
NN optimisation: 1P (reco+id)
= 10 - 20 GeVvisibleTE
= 20 - 40 GeVvisibleTE
= 40 - 60 GeVvisibleTE
= 60 - 100 GeVvisibleTE
NN optimisation: 1P (reco+id)
∈0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
R
210
310
410
-jets and qcd-jets, MHLlhτ, for ∈R vs <28.5T15.0<p
<61.5T43.5<p
<133.5T88.5<p
<334.5T217.5<p
no noise
noise + 2 sigma cut
Figure 200. Expected performance for track-based (left) and calo-based (right) algorithms. Shown is therejection against jets as a function of efficiency for hadronic τ -decays for various ranges of the visibletransverse energyEvisible
T .
Table 54.Expected numbers of events with 100pb−1 collected with 1031 luminosity
Process trigger signal evtbgd evt comments
W → τν , τ → had ν
Z → ττ , τ → had ν , τ → e ν
tt → τν b j jb
lepton-τ candidates, with the expected contributions from signalZ → ττ events and from the7010
dominantbb background.
– 272 –
tracks N0 1 2 3 4 5 60
500
1000
1500
2000
2500
3000
-1 Events for 100 pb
> 40 GeV )miss
T final off-line ( E
events ν τ → W events τ τ → Z
QCD events
-1 Events for 100 pb
(GeV)-visτe, m0 20 40 60 80 100 120 1400
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
x BR [pb]σ : τ τ → Z h2011Entries 7
Mean 59.52
RMS 20.88
Underflow 0
Overflow 0.1292
Integral 7.718
x BR [pb]σ : τ τ → Z
Figure 201. Distribution of charged track multiplicity expected for reconstructedτ -candidates withET > 15 GeV andEmiss
T > 40 GeV (left). Shown are the expected contributions from theW → τνsignal and the various backgrounds for an integrated luminosity of 100 pb−1. Visible invariant mass dis-tribution for lepton-τ pairs fromZ → ττ decays for leptons withpT > 20 GeV andτ -candidates withET > 15GeV and for an integrated of 100 pb−1.
– 273 –
9.7 Flavour tagging (4 pages)
9.7.1 b-tagging using tracking and vertexing
- common assumptions: no pile-up, perfect alignment, efficiency, noise
- samples7015
- basic info about b-jets:pT , track multiplicity, etc
9.7.1.1 Impact parameter of tracks
- track selection for b-tagging
- relation to tracking section
- jet and signing7020
- link to figure 202
a0 for tracks inb,c and light jets
- tt events (sample 5200)
- purifiedc and light jets
- standard btag quality cuts on tracks
a0/σ for tracks inb,c and light jets
- tt events (sample 5200)
- purifiedc and light jets
- standard btag quality cuts on tracks
Figure 202. Signed impact parameter distribution (left) and impact parameter significance distribution(right) for b-jets, c-jets and light jets.
9.7.1.2 Secondary vertices
- description of inclusive secondary vertex reconstruction
- impact of primary vertex finding
- link to figure 2037025
– 274 –
Vertex mass forb and light jets
- tt events (sample 5200)
- purified light jets
- one curve for b-jets
- one curve for light jets
Energy fraction forb and light jets
- tt events (sample 5200)
- purified light jets
- one curve for b-jets
- one curve for light jets
Vertex mass forb and light jets
- tt events (sample 5200)
- purified light jets
- one curve for b-jets
- one curve for light jets
Figure 203. Secondary vertex variables: mass (left), energy fraction (middle) and number of two-trackvertices (right) for b-jets and light jets.
9.7.1.3 Likelihood-ratio taggers
- description of the method
- brief description of all taggers
- link to figure 204
jet weight forb, (c) and light jets for IP2D
- tt events (sample 5200)
- purifiedc and light jets
jet weight forb, (c) and light jets for IP3D+SV1
- tt events (sample 5200)
- purifiedc and light jets
Figure 204. Jet weight distribution for IP2D tagger (left) and IP3D+SV1(right) for b-jets, c-jets and lightjets.
9.7.1.4 Labelling and purification7030
- labeling for b,c,tau,light,(gluon)
- event topologies and the need for purification
- link to figure 205
– 275 –
R versusεb for IP3D+SV1 tagger
- tt events (sample 5200)
- WH120 events
- no purification
- one curve for light
- one curve for gluon jets
- one curve for c-jets
R versusεb for IP3D+SV1 tagger
- tt events (sample 5200)
- WH120 events
- purifiedc and light jets
- one curve for light
- one curve for gluon jets
- one curve for c-jets
Figure 205.Rejection versus b-jet efficiency without purification (left) and with purification (right) for ttbarand WH with IP3D+SV1 tagger.
9.7.1.5 Expected performance
- link to figure 2067035
- problems at lowpT and/or highη
- problems at highpT
- 2D binning, remaining differences across samples
9.7.2 Soft lepton tagging
- interest of soft lepton tagging7040
- brief description of soft muon tagging
- brief description of soft electron tagging
- link to figure 207
9.7.3 Impact of alignment
- description of misalignment scenarios7045
- link to figure 208
– 276 –
R versus jetpT
- tt events (sample 5200)
- purified light jets
- one curve for IP2D
- one curve for IP3D+SV1
R versus jetη
- tt events (sample 5200)
- purified light jets
- one curve for IP2D
- one curve for IP3D+SV1
Figure 206. Rejection of light jets as a function of jetpT (left) and jetη (right), for IP3D+SV1 tagger anda b-tagging efficiency of 60%.
R versusεb for soft muon tagger
- WH120 events
- one curve for default
- one curve for each cavern background scenario (2 max.)
- BR and ID efficiency included (i.e.ε ∼ 10%)
R versusεb for soft electron tagger
- WH120 events
- one curve for default
- one curve for each pile-up scenario (2 max.)
- BR and ID efficiency included (i.e.ε ∼ 10%)
Figure 207.Rejection of light jets versus b-tagging efficiency for the soft muon tagger (left) and soft electrontagger (right).
9.7.4 Estimation of performances from data
9.7.4.1 Measuring b-tagging efficiency
- System8 method description
- trigger and rates to get the muon jet sample7050
- results and systematics
– 277 –
R versusεb for JetProb tagger
- tt events
- purified light jets
- one curve for default
- one curve for each misalignment scenario (2 maximum)
R versusεb for IP3D+SV1 tagger
- tt events
- purified light jets
- one curve for default
- one curve for each misalignment scenario (2 maximum)
Figure 208.Rejection of light jets versus b-tagging efficiency for the JetProb tagger (left) and the IP3D+SV1tagger (right) for various misalignment scenarios.
- other approaches:prelT templates
- other approaches: tag counting
9.7.4.2 Estimating mistag rates
- negative tag rates7055
- corrections: heavy flavor contamination, hyperons and interactions
- results and systematics
9.7.5 Calibration
9.7.5.1 JetProb calibration
- track classes7060
- trigger and rates needed fora0/σ negative side calibration
- impact of miscalibration
9.7.5.2 Calibration of likelihood ratio-based taggers
- b-jet p.d.f. from semi-leptonictt
- light jet p.d.f. (how?)7065
- impact of miscalibration
– 278 –
9.8 Trigger performance (8 pages)
9.8.1 Overview
Short recall (refer to) TDAQ Architecture, emphasising coupling to software flexibility (RoI mech-anism, trigger elements and steering). L2 and EF as complementary stages of selection, with their7070
own specific aspects. Remind about inner detector full scan at L2 and EF and its use for B-triggersand possibly minimum-bias triggers.
9.8.2 Selection strategy
• Description of the selection strategy and explanation of the context (inclusive triggers, 1033
for physics, 1031 for commissioning, etc)7075
• Overview of different signatures
• Mention evolution of B-trigger strategy from 1031 to 1033. Also mention B-physics will bepursued based on di-muon triggers up to 1034.
Typical L1 menu Typical HLT menu
Figure 209.L1 and HLT Menu and rates + efficiency, rejection, timings, etc.
9.8.3 Trigger menus
• Menu table for 1033 (L1 and HLT with rates, efficiencies, timing, etc) plus of course explica-7080
tive text
• Cover associated physics signals and give sample composition whenever possible
• For timing performance, use results from most recent Technical Run, add some commentsabout global optimization, choice of variables, ROI size, etc. Minimise overlap with TDAQchapter performance section.7085
9.8.4 Examples of trigger performance
Illustration, through selected examples, of trigger performance results for some items of the menutable (including study of pre-scale factors, etc). For performance results known to be sensitive to
– 279 –
pile-up and/or cavern background even at 1033, results should be shown when the effect is signif-icant (e.g. muon fake rates) and at least checked when the effect is believed to be small using the7090
samples which are quickly becoming available now.
• Electrons and photonsSome words about photon conversions and brem recoveryOptimisation of trigger signaturesFigure 210: for e.g. e25i inclusive trigger, example of trigger efficiency versus|η | and pT7095
(comparing ideal and misaligned geometry).
e25i trigger efficiency versus|η | e25i trigger efficiency versuspT
Figure 210.Example of trigger efficiency versus|η |(left) andpT (right) of isolated electrons for e25i triggeritem (comparing ideal and misaligned geometry).
• τ -leptonsFigure 211: efficiency curves for L1/L2/EF versus visibleET of τ -leptonTable: L2 and EF : signal efficiency/background rates going through the steps of one menu(tau35i for example)7100
τ L1 trigger efficiency versusET τ HLT trigger efficiency versusET
Figure 211.Example of L1 (left) and HLT (right) trigger efficiency versusET for tau35i trigger item.
• JetsBest opportunity to show pre-scale strategy, as shown in Figure 212.
– 280 –
Expected L1 jetET spectrum for the prescale strat-
egy envisaged at 1033 cm−2 s−1
Figure 212.Example of pre-scale strategy for inclusive jets at a luminosity of 1033 cm−2 s−1.
• MuonsFigure 213: EF muon rates.
τ L1 trigger efficiency versusET
Figure 213.Expected EF output rate for muons as a function ofpT at a luminosity of 1033 cm−2 s−1.
• B-physics7105
If possible, to the extent that studies are completed after the summer, include theB →J/ψ(µµ) andB→ Dsπ triggers as examples of B-physics triggers.
9.8.5 Trigger commissioning
Description of the trigger commissioning strategy, using some results from the 1031 menu. Explainhow to benefit from lower thresholds, dielectron low masses,the tau10MET20 slice forW → τν7110
and the single muon trigger for initial B-physics. Refer to the relevant performance sections.As an example, explain in detail commissioning of the L1 muontrigger (Figure 214).
9.8.6 Trigger efficiency from data
Explanation of methodologies to extract trigger efficiencies from data.Figure: Turn on curve for one/two menu items of trigger efficiency vsET(comparing MC and7115
method)
– 281 –
Trigger commissioning menu at a luminosity of
1031cm−2s−1
Figure 214.Trigger menu envisaged for commissioning at a luminosity of1031cm−2s−1.
Trigger efficiency for electrons fromZ → eede-
cays as measured from data using the e25i trigger
item
Trigger efficiency for muons fromZ → µµ de-
cays as measured from data using the mu20i trig-
ger item
Figure 215.Efficiency as measured from data for electrons (left) and muons (right).
– 282 –
9.9 Overall ATLAS energy scale and alignment (4 pages)
To be done later once scope of all previous sections has been finalised in terms of calibration,alignment, and in situ measurements.
– 283 –
10. Outlook and conclusions7120
This section will only really be written for the final draft toput in a snapshot of the status at thetime of publication. Comments should be limited to the broadoutline below in terms of scope andbalance and perhaps also to the section on forward detectorswhich is fleshed out more already inthis first draft.
The ATLAS experiment is now preparing actively to take first data with colliding beams in7125
summer 2008. A few components of the ATLAS detector described in the various Technical DesignReports over the past ten years remain to be installed in the spring of 2008, but commissioning ofthe detector and readout systems, and preparation of offlinecomputing and analysis are alreadynow becoming the major current activities.
To operate at the nominal LHC luminosity of 1034 cm−2 s−1, a few additional detector hard-7130
ware elements need to be installed also after the 2008 run, while for initial running they are notneeded. The detector systems in the forward direction, which were not part of the original set ofATLAS Technical Design Reports, have matured over the last years and most of them will alreadybe installed for the 2008 run, but certain components will nevertheless need to be completed inthe coming years. These actions will then complete the ATLASdetector system as it has been7135
envisaged for operation at the LHC over a number of years at design luminosity.
After a few years of operation at design luminosity, the innermost pixel-detector layer, the so-called vertexing layer, will need to be replaced due to radiation damage. The preparations for thisreplacement are already well under way. This replacement could be perceived as the first step in apossible upgrade of much more major scope, if the LHC machinewere to be upgraded to operate7140
at even higher luminosities by up to a factor of 10. If such were to be the case, the ATLAS detectorwill need substantial improvements, including a complete replacement of the inner detector, andR&D efforts towards such an upgrade have already been launched with direct guidance from theATLAS collaboration management and technical coordination.
10.1 From end 2007 until start-up (1 page)7145
The following activities need to be completed to prepare theATLAS detector for data-taking in2008:
• installation of missing hardware components. The installation of the ATLAS detector in itsunderground cavern began in 2003, as described in section 8.7. The detector is now almostcomplete and only a few elements remain to be installed before data-taking begins in 2008:7150
Here will be described the ATLAS components still to be installed at the time of submissionof the paper (small wheels and some end-wall chambers).
• completion of detector commissioning. All of the ATLAS detector sub-systems have beentested and operated as much as possible on the surface and in test-beam setups, and alsoto a certain extent in the underground cavern after installation. The flux of muons from7155
cosmic rays provides a useful test of systems such as the calorimeters, inner detector andmuon chambers to check alignment, calibration and the integration of all the readout anddata-collection systems into the ATLAS data acquisition.
– 284 –
Here will be added text concerning the Mn commissioning weeks and the status of globalATLAS operation at the time the paper is submitted.7160
• preparation of computing and offline analysis. In parallel with the activities at point 1, theGRID computing infrastructure and organisation are being commissioned and tested. Large-scale exercises of the data-processing, software and computing infrastructure have been per-formed successfully and more are planned in the build-up towards data-taking.
Here will be added some text about the global status in this area and in particular about the7165
FDR. Do these items belong here? Should one even discuss thisitem at all?
10.2 From start-up till operation of complete detector at design luminosity(1/2 page)
• Upgrading the trigger capacity (describe the plans for installing more HLT processors)
• Completing the muon spectrometer and the forward shielding
10.3 Forward detector systems (2-3 pages)7170
In addition to the main ATLAS detector, three smaller systems are being built to provide goodcoverage in the very forward region [193]. These are closelyconnected to the luminosity deter-mination in ATLAS, but are also foreseen to study forward physics. If ordered according to theirdistance from the ATLAS interaction point, the first system is a Cerenkov detector called LUCID.LUCID is the main luminosity monitor in ATLAS and is located at a distance of 17 m from the7175
interaction point. The second system is a zero-degree calorimeter (ZDC), which is located at adistance of 140 m from the interaction point. This corresponds to the location, where the LHCbeam-pipe is divided in two and the ZDC is located between thebeam pipes just after the splitinside an absorber. The most remote detector is the so-called ALFA system. ALFA consists ofscintillating-fibre trackers located inside roman pots at adistance of 240 m from the ATLAS inter-7180
action point. Additional detectors are being considered atan even further distance of about 420 mfrom the interaction point, but they are just mentioned in this brief introduction since no decisionhas been taken yet concerning them.
In the following, a brief description is given of the three forward systems currently underconstruction.7185
10.3.1 The LUCID detector
LUCID (or LUminosity measurement using Cerenkov Integrating Detector) will be the main lu-minosity monitor in ATLAS at the LHC design luminosity. Its main purpose is to detect inelasticpp scattering in the forward direction, both in order to measure the integrated luminosity of theATLAS runs and to monitor online the instantaneous luminosity and beam conditions. Potentially,7190
LUCID could also be used for diffractive studies, for example as a rapidity-gap veto.The main requirements for this detector are:
• good acceptance to minimum-bias events;
• sufficient time resolution to measure individual bunch-crossings;
– 285 –
• ability to count particles.7195
The LUCID detector is schematically depicted in Fig. 216. Itconsists of aluminium tubesfilled with C4F10 surrounding the beam-pipe and pointing towards the ATLAS interaction point.The Cerenkov light emitted by a traversing particle is reflected down the tube and read out byphoto-multipliers. The signal amplitude from these photo-multipliers can be used to distinguishthe number of particles per tube and the fast time response provides unambiguous measurements7200
of individual beam crossings. A small-scale LUCID, dedicated purely to luminosity monitoring,has been validated in beam tests. The LUCID detector itself will be installed in time for the start-upof the LHC. Based on the measured performance of this initialdetector, an optimised and upgradeddetector, including a larger number of tubes, is foreseen tobe installed at the same time as the LHCwill be upgraded towards reaching its design luminosity of 1034 cm−2 s−1.7205
Figure 216.Schematic structure of the LUCID detector. Also shown is oneof its associated photomultipli-ers.This figure is only a placeholder.
10.3.2 The ZDC detector
The second forward system in ATLAS is the zero-degree calorimeter (ZDC), which will measureneutral particles at a 0 polar angle. The ZDC has a central role in the ATLAS heavy-ionphysicsprogram, where it will be used to measure the centrality of the collisions, the luminosity, as wellas to provide certain triggers. It will, however, also be of importance both in thepp program to7210
study forward particle production and for the tuning of the LHC machine, for which it can be usedto determine the location of the interaction point and the beam crossing angle.
The ATLAS ZDC consists of six tungsten/quartz calorimeter modules, schematically shownin Fig. 217. The light collected from the quartz fibres is readout by photomultipliers. In addition,
– 286 –
the ZDC is equipped with horizontal quartz rods, parallel tothe beam, in order to determine the7215
location of the showers in the plane perpendicular to the beam. The ZDC has been extensivelytested and will be installed in time for the start-up of the LHC. An upgrade is foreseen after aboutone year of running, when additional space in the absorber inwhich it will be located will becomeavailable.
Figure 217. Location and schematic layout of the ZDC detector located ata distance of 140 m from theATLAS interaction point.This figure is only a placeholder.
10.3.3 The ALFA detector7220
The ALFA detector (orAbsolute Luminosity For ATLAS) consists of scintillating fibre trackerslocated in roman pots at a distance of 240 m on each side of the interaction point. The romanpots (see Fig. 218) allow the detectors to approach as close as possible to the beams inside theLHC beam-pipe and the main purpose of ALFA is to measure elastic proton scattering at lowangles. This is primarily to determine the absolute luminosity in ATLAS, but also other physics7225
– 287 –
Figure 218. Schematic view of the support mechanics for one of the ALFA detectors and of its locationat a distance of 240 m from the ATLAS interaction point (left). Also shown (right) is one of the as-builtstructures, which will house the scintillating fibre trackers.
studies are foreseen such as measuring the totalpp cross-section, measuring elastic scatteringparameters and potentially also using the tagged protons for diffractive studies.
To achieve an optimal precision in the luminosity measurement, the specification is to measureelastic scattering in the Coulomb interference region, which requires a measurement of scatteringangles down to about 3µrad. In order to reach such small angles, the LHC has to run with special7230
so-called high-β∗ optics (see Table 51), but, even with such beam optics, the detectors have to belocated at a radial distance of only 1-2 mm from the beam. The main requirements on the fibretracker are:
• a spatial resolution of about 30µm,
• no significant inactive edge region and7235
• insensitivity to the radio-frequency noise from the LHC beams and to the vacuum in theroman pots.
The highβ∗ runs correspond to a very low instantaneous luminosity and for this reason no radiation-hard technology has had to be adopted for this specific set of detectors.
– 288 –
Given these requirements, ATLAS has chosen a scintillatingfibre tracker, as shown in Fig. 219.7240
Prototype detectors of the ALFA tracker have been validatedin beam tests at DESY [194] andCERN [195], together with the front-end electronics and theso-called overlap detector alignmentsystem. The tests have shown an adequate performance for theluminosity measurement and thefull ALFA system is foreseen to be installed during the shutdown between 2008 and 2009.
Figure 219.Schematic layout of one of the ALFA scintillating fibre trackers and its trigger (left) and pictureof one of the prototypes (right).This figure is only a placeholder.
10.3.4 B-layer replacement (1 page)7245
• Reminder about radiation-hardness limits and sensor lifetimes;
• Replacement of B-layer (a few words about possible differences in the detector hardware andoperational characteristics).
10.3.5 Super-LHC upgrade (1-1.5 page)
Upgrading the LHC luminosity by a factor of 10 should substantially increase the physics potential7250
of the experiments and feasibility studies and R&D for such an upgrade are being pursued both forthe machine and the detectors.
Such an upgrade will however demand substantial changes in the current experimental hard-ware. A decision about the scope and time-scale of such an upgrade can be made after a fewyears of LHC and detector operation, considering both physics results and machine/detector per-7255
formances at that point. Nevertheless the experimental changes are being planned by specific R&Don the major components where substantial changes will mostlikely be required.
The major changes foreseen are the following:
• the inner detector sub-systems all need to be replaced because the as-built detector hasbeen designed many years ago to survive 10 years of operationat the design luminosity7260
of the LHC, which was already a formidable challenge at the time;
– 289 –
• the various interfaces to the LHC machine (beam-pipe, forward shielding and the machineelements in the vicinity of the detector) will have to be re-optimised and substantially up-graded;
• the calorimeter electronics and readout in general will very likely need substantial improve-7265
ments to handle the increased rates;
• the forward calorimeters and the end-cap muon chambers willalso need some upgrades.
Each of the above items will be fleshed out somewhat for the final draft.
10.3.6 Concluding remarks
– 290 –
11. Bibliography: main references and complete list of back-up papers7270
References
[1] ATLAS physics TDR.
[2] ATLAS paper(s) about x-dim.
[3] R. Ruber et al., Construction of the ATLAS central solenoid, Nucl. Instrum. Meth.A888 (2007)801–820 (NOT YET PUBLISHED).7275
[4] A. Dudarev et al., Construction of the ATLAS toroid magnet system, Nucl. Instrum. Meth.A888(2007) 801–820 (NOT YET PUBLISHED).
[5] H. H. J. ten Kate, Superconducting Magnet System for the ATLAS Detector at CERN, IEEE Tr.Appl. Superconductivity 9 (1999) 841.
[6] H. H. J. ten Kate, The ATLAS Superconducting Magnet System: Status of Construction &7280
Installation, IEEE Tr. Appl. Superconductivity 16 (2006) 499.
[7] ATLAS Collaboration, Magnet System Technical Design Report, CERN/LHCC/97-18 (1997).
[8] ATLAS Collaboration, Barrel Toroid Technical Design Report, CERN/LHCC/97-19 (1997).
[9] ATLAS Collaboration, End-cap Toroid Technical Design Report, CERN/LHCC/97-20 (1997).
[10] ATLAS Collaboration, Central Solenoid Technical Design Report, CERN/LHCC/97-21 (1997).7285
[11] A. Yamamoto et al., Progress in the ATLAS Central Solenoid, IEEE Tr. Appl. Superconductivity 10(2000) 353.
[12] S. Mizurnaki et al., Fabrication and Mechanical Performance of the ATLAS Central Solenoid, IEEETr. Appl. Superconductivity 12 (2002) 415.
[13] R. Ruber et al., Quench Characteristics of the ATLAS Central Solenoid, IEEE Tr. Appl.7290
Superconductivity 16 (2006).
[14] R. Ruber, Ultimate performance of the ATLAS Superconducting Solenoid, Paper presented at Appl.Superc. Conf. 2006, Seattle, USA.
[15] A. Dael et al., Progress in the Design of the Barrel Toroid Magnet for the ATLAS Experiment andAssociated R&D at CEA-Saclay and INFN-Milano, Proc. MT-15,Beijing, China, Oct 1997.7295
[16] P. Védrine et al., Manufacturing and integration progress of the ATLAS barrel toroid magnet atCERN, IEEE Tr. Appl. Superconductivity 14(2) (2004) 491-494.
[17] J.-M. Rey et al., Cold mass integration of the ATLAS Barrel Toroid Magnets at CERN, IEEE Tr.Appl. Superconductivity 16 (2006).
[18] P. Vedrine et al., Completion of the Manufacturing of the ATLAS Barrel Toroid Magnet at CERN,7300
IEEE Tr. Appl. Superconductivity 16 (2006).
[19] P. Miele et al., The ATLAS magnet test facility at CERN, IEEE Tr. Appl. Superconductivity 11(2000) 1713.
[20] A. Dudarev et al., First full-size ATLAS barrel toroid coil successfully tested up to 22 kA at 4 T,IEEE Tr. Appl. Superconductivity. 15(2) (2005) 1271-1275.7305
[21] A. Dudarev et al., On-Surface Test of the ATLAS Barrel Toroid Coils: Overview, IEEE Tr. Appl.Superconductivity 16 (2006).
– 291 –
[22] A. Dudarev et al., Quench Behaviour of the ATLAS Barrel Toroid Coils, IEEE Tr. Appl.Superconductivity 16 (2006).
[23] J.-J. Rabbers et al., Experimental and theoretical investigation of the ramp losses in the conductor7310
and coil casing of the ATLAS Barrel Toroid Coils, IEEE Tr. Appl. Superconductivity 16 (2006).
[24] A. Foussat et al., Assembly Concept and Technology of the ATLAS Barrel Toroid, IEEE Tr. Appl.Superconductivity 16 (2006).
[25] D. E. Baynham et al., Eng. Status of the End Cap Toroid Magnets for the ATLAS Experiment, IEEETr. Appl. Superconductivity 10 (2000).7315
[26] D. E. Baynham et al., ATLAS End Cap Toroid Cold Mass and Cryostat Integration, IEEE Tr. Appl.Superconductivity 16 (2006) 537.
[27] D. E. Baynham et al., ATLAS End Cap Toroid Integration and Test”, Paper presented at Appl.Superconductivity Conference 2006, Seattle.
[28] P. Miele et al., ATLAS magnet common cryogenic, vacuum,electrical and control systems, IEEE Tr.7320
Appl. Superconductivity 14 (2004) 504-508.
[29] L. Chevalier et al., Field map calculations for ATLAS, Nucl. Instrum. Meth.A888 (2007) 801–820(NOT YET PUBLISHED).
[30] ATLAS Collaboration, Muon Spectrometer Technical Design Report, ATLAS TDR 10,CERN/LHCC/97-22 (1997).7325
[31] T. Nikitina, Simulations Numériques pour le Calorimètre du Projet Atlas: ElectromagnétismeNon-linéaire et Homogénéisation, Thèse de Doctorat, Université de Franche-Comté (2005).
[32] T. Nikitina and F. Bergsma, A Program to Calculate the ATLAS Magnetic Field, Atlas Internal NoteATL-MAGNET-2001-002 (2001).
[33] TOSCA, program developed by Vector Fields Limited, England.7330
[34] F. Bergsma, Proc. 13th International Magnetic Measurement Workshop, Stanford Linear AcceleratorCenter, 19-23 May 2003.
[35] S. Snow et al., Field mapping of the ATLAS central solenoid, Nucl. Instrum. Meth.A888 (2007)801–820 (NOT YET PUBLISHED).
[36] W. Kozanecki et al., Field mapping of the ATLAS barrel toroid, Nucl. Instrum. Meth.A888 (2007)7335
801–820 (NOT YET PUBLISHED).
[37] M. Bosman et al., Estimation of Radiation Background, Impact on Detectors, Activation andShielding Optimization in ATLAS, ATLAS Internal Note ATL-GEN-2005-001.
[38] A. Romaniouk, TRT straw counting rate from photon background at LHC, ATLAS Internal NoteATL-INDET-96-149.7340
[39] S. Ahlen et al., Measurement of ATLAS MDT neutron sensitivity and development of a simulationmodel, ATLAS Internal Note ATL-MUON-98-236.
[40] R. Engel, J. Ranft and S. Roesler, Hard diffraction in hadron-hadron interactions and inphotoproduction, Phys. Rev.D52 (1995) 1459–1468.
[41] R. Engel, Photoproduction within the two-component dual parton model: amplitudes and7345
cross-sections, Z. Phys.C66 (1995) 203–214.
– 292 –
[42] R. Engel and J. Ranft, Hadronic photon-photon interactions at high energy, Phys. Rev.D54 (1996)4244–4262.
[43] T. Sjostrand et al., Pythia 6.1, Comput. Phys. Commun.135(2001) 238.
[44] C. Zeitnitz and T.A. Gabriel, The GEANT-CALOR Interface and Benchmark Calculations for Zeus7350
Calorimeters, Nuc. Instrum. MethodsA349 (1994) 106.
[45] T.A. Gabriel et al., CALOR89: A Monte Carlo Program Package for the Design and Analysis ofCalorimeter Systems, ORNL/TM-11185.
[46] R. Brun and F. Carminanti, CALOR89: A Monte Carlo Program Package for the Design andAnalysis of Calorimeter Systems,CERN Program Library LongWriteup W5013, September 1993.7355
[47] A. Ferrari, P.R. Sala, A. Fasso and J. Ranft, FLUKA: A multi-particle transport code (Programversion 2005), CERN report CERN-2005-010.
[48] N.V. Mokhov, The MARS Code System Users’s Guide, Fermilab-FN-628 (1995).
[49] The Muon Spectrometer Technical Design Report (TDR), CERN/LHCC/97-22 (31/05/1997).
[50] C. Cheplakov et al., Irradiation test of readout chain components of the ATLAS liquid argon7360
calorimters, ATLAS note ATL-LARG-99-018.
[51] C. Leroy, Irradiation tests of ATLAS liquid argon forward calorimeter (FCAL) electronicscomponents, ATLAS note ATL-LARG-2002-003.
[52] A. Amorim et al., Study of the effect of the radiation on the TILECAL Barrel Hadron calorimeter tobe used in ATLAS, ATLAS note ATL-TILECAL-93-021.7365
[53] A. Karyukhin et al., Study of molded scintillator radiation hardness, ATLAS noteATL-TILECAL-94-025.
[54] M. David et al., Low dose rate effect in scintillating and WLS fibers by ionizing radiation, ATLASnote ATL-TILECAL-96-078.
[55] M. David et al., Radiation damage in WLS fibers, ATLAS note ATL-TILECAL-98-151.7370
[56] G. Lindstrom, Radiation damage in silicon detectors, Nuc. Instrum. MethodsA512 (2003) 30.
[57] R. Ichimyia, Radiation qualification of electronics components used for the ATLAS level-1 muonendcap trigger system, ATLAS note ATL-DAQ-CONF-2005-003.
[58] V. Cindro, et al., The ATLAS Beam Conditions Monitor, tobe submitted to JINST.
[59] R. The CERN-RD42 Collaboration, CVD Diamond RadiationDetector Development,7375
http://rd42.web.cern.ch/RD42/ and CERN/DRDC 94-21, May 1994.
[60] Element Six Ltd., KingŠs Ride Park, Ascot, Berkshire SL5 8BP, UK.
[61] H. Frais-Kolbl, E. Griesmayer, H. Kagan and H. Pernegger, A Fast Low-Noise Charged-ParticleCVD Diamond Detector, 2004.
[62] M. Niegl, Concept and Implementation of an FPGA-based Data Recorder and Processor for the7380
ATLAS BCM, Bachelor Thesis, University of Applied SciencesWiener-Neustadt, June 2007.
[63] ATLAS Collaboration, Inner Detector: Technical Design Report, CERN/LHCC/97-016 and 97-017(1997).
[64] The ATLAS Collaboration, ATLAS Pixel Detector: Technical Design Report, CERN/LHCC/98-013(1998).7385
– 293 –
[65] Construction and Performance of the ATLAS Pixel Detector, in preparation for Nucl. Methods A.,time scale unknown, editor: M. Gilchriese.
[66] A. Ahmed et al., The Silicon Microstrip Sensors of the ATLAS SemiConductor Tracker, Nucl.Instrum. Meth. Res.A578 (2007) 98–118.
[67] Pixel p-spray insulation technical reference..7390
[68] Systematic tests of pixel and SCT sensor sub-samples..
[69] I. Gorelov et al., Nucl. Instrum. Meth., A481 (2002) 204; G. Alimonti et al., ATLAS Internal Note,ATL-INDET-2005-006 and ATL-INDET-2005-007 (2005)..
[70] SCT beam tests of sensors pre- and post-irradiation..
[71] TRT Drift Tubes, in preparation for Nucl. Methods A., time scale unknown, editors: A. Romaniouk7395
and K. McFarlane.
[72] A. Abdesselam et al., The Barrel Modules of the ATLAS SemiConductor Tracker, Nucl. Instrum.Meth. Res.A568 (2006) 642–671.
[73] F. Campabadal et al., Design and Performance of the ABCD3TA ASIC for readout of silicon stripdetectors in the ATLAS semiconductor tracker, Nucl. Instrum. Meth. Res.A552 (2005) 292–328.7400
[74] Y. Unno et al., Application of Cu-polyimide flex circuitand Al-on.glass pitch adapter for theATLAS SCT barrel hybrid, Nucl. Instrum. Meth. Res.A541 (2005) 286–294.
[75] R.J. Apsimon et al., Application of advanced thermal management technologies to the ATLAS SCTbarrel module baseboards, Nucl. Instrum. Meth. Res.A565 (2006) 561–571.
[76] A. Abdesselam et al., The ATLAS SemiConductor Tracker End-Cap Module, Nucl. Instrum. Meth.7405
Res.A575 (2007) 353–389.
[77] J. Grosse-Knetter, The ATLAS Pixel Detector, Nucl. Instrum. Meth. Res.A568 (2006) 252–257.
[78] I. Peric et al., The FEI3 Readout Chip for the ATLAS PixelDetector, Nucl. Instrum. Meth. Res.A565 (2006) 178–187.
[79] R. Beccherle et al., MCC: The Module Controller of the ATLAS Pixel Detector, Nucl. Instrum.7410
Meth. Res.A492 (2002) 117–133.
[80] TRT Electronics, in preparation for Nucl. Methods A., time scale unknown, editor: H. Williams.
[81] T. Akesson et al., Implementation of the DTMROC-S ASIC for the ATLAS TRT Detector in a0.25µm CMOS Technology, Nucl. Sci. Symposium Conference Record,2002 IEEE1 (2002)549–553.7415
[82] A. Abdesselam et al., The Optical Links of the ATLAS SemiConductor Tracker, to be submitted toNucl. Instr. Methods Res. A., editor: A. Weidberg..
[83] M.L. Chu et al., The Off-Detector Opto-Electronics forthe Optical Links of the ATLASSemiconductor Tracker and Pixel Detector, Nucl. Instrum. Meth. Res.A530 (2004) 293–310.
[84] K.E. Arms et al., ATLAS Pixel Opto-Electronics, Nucl. Instrum. Meth. Res.A554 (2005) 458–468.7420
[85] GOL: A Radiation Tolerant Gigabit Serializer for LHC Data Transmission, Proc. LEB2001, eds. P.Moreira et al..
[86] A. Abdesselam et al., The Data Acquisition and Calibration System for the ATLAS SemiconductorTracker, to be submitted to Nucl. Instr. Methods Res. A..
– 294 –
[87] A. Abdesselam et al., ROD/BOC/TIM of the ATLAS SCT Detector, Nucl. Instr. Methods Res. A,7425
time scale unknown.
[88] Evolution of the TRT Backend and the new TRT-TTC Board, Proc. LECC 2005, eds. P. Lichard etal..
[89] A. Abdesselam et al., Power Supply and Distribution of the ATLAS SCT Detector, Nucl. Instr.Methods Res. A, time scale unknown.7430
[90] A. Abdesselam et al., Mechanical Structure and Performance of the ATLAS SCT Detector, Nucl.Instr. Methods Res. A, time scale unknown, editor: E.Perrinet al.
[91] R. Nickerson et al., Robotic Mounting of ATLAS Barrel SCT Modules, Nucl. Instrum. Meth. Res.A568 (2006) 686–691.
[92] P. W. Phillips, Functional Testing of the ATLAS SCT Barrels, Nucl. Instrum. Meth. Res.A5707435
(2007) 230–235.
[93] H. Pernegger, Integration and Test of the ATLAS Semiconductor Tracker, Nucl. Instrum. Meth. Res.A572 (2007) 108–112.
[94] TRT Barrel, in preparation for Nucl. Methods A., time scale unknown, editors: C. Wang and H.Ogren.7440
[95] TRT End-Cap, in preparation for Nucl. Methods A., time scale unknown, editor: C. Rembser.
[96] A. Abdesselam et al., The Detector Control System of theATLAS SemiConductor Tracker duringMacro-Assembly and Integration, ATLAS Note ATL-COM-INDET-2007-010.
[97] S. Agostinelli et al., GEANT4 - A Simulation Toolkit, Nucl. Instrum. Meth. Res.A506 (2003)250–303.7445
[98] J. Allison et al., GEANT4 Developments and Applications, IEEE Trans. Nucl. Sci.53 (2006)270–278.
[99] T. Cornelissen, CTB Tracking: Track Reconstruction for the Testbeam and Cosmics, ATLAS noteATL-COM-INDET-2006-003.
[100] F. Akesson et al., ATLAS Tracking Event Data Model, ATLAS Note ATL-SOFT-PUB-2006-004.7450
[101] P. Bruckman et al., Global Chisq Aprroach to the Alignment of the ATLAS Silicon TrackingDetectors, ATLAS Note CERN-ATL-INDET-PUB-2005-002..
[102] P. Bruckman, Alignment Strategy for the ATLAS Inner Detector, submitted to Nucl. Inst. MethodsA..
[103] C. Escobar, Alignment of the Inner Detector of the ATLAS Experiment, submitted to IEEE Trans.7455
Nucl. Science, 2006..
[104] Combined Alignment Method 1, in preparation, time scale unknown..
[105] ATLAS Collaboration,Liquid Argon Calorimeter Technical Design Report, CERN LHCC-96-041.
[106] D.Axen et al.,Signal feedthroughs for the ATLAS barrel and endcap calorimeters Rev. Sci. Instrum.(2005)76063306.7460
[107] B. Aubert et al.,Development and construction of large size signal electrodes for the ATLASelectromagnetic calorimeterNucl. Instrum. Methods A (2005)539558-594.
– 295 –
[108] M.L. Andrieux et al.,Construction and test of the first two sectors of the ATLAS barrel liquid argonpresamplerNucl. Instrum. Methods A (2002)479316-333.
[109] N.N. et al.,Construction, Assembly and Tests of the ATLAS electromagnetic End-Cap Calorimeter.7465
[110] B. Aubert et al.,Construction, assembly and tests of the ATLAS electromagnetic barrel calorimeterNucl. Instrum. Methods A (2006)558388-418.
[111] B.Belhorma et al.,Evaporating short-circuits in the ATLAS liquid argon barrel presampler 006ATL-LARG-PUB-2005-003; ATL-COM-LARG-2005-006; CERN-ATL-LARG-PUB-2005-003.
[112] ATLAS Collaboration,Tile Calorimeter Technical Design Report, CERN LHCC-96-042, ATLAS7470
TDR,15 December 1996.
[113] ANL-TR to identify the wealth of engineering design and analysis, report in preparation.
[114] Master Plate Production for the Tile Calorimeter, ANL-TR-99-04.
[115] Summary paper of tile construction- report in preparation.
[116] P.Adragna et al.,The ATLAS hadronic tile calorimeter : from construction toward physics, IEEE7475
Trans. Nucl. Sci. 53 (2006) 1275-81.
[117] M.J.Varanda,The Tile Hadronic Calorimeter for the ATLAS Experiment, IEEE TRans. on Nucl. Sci.,Vol 48, No 3, 2001, page 367.
[118] F.Bosi et al., Nucl. Instrum. Meth.A (2002)485311-321.
[119] C.Cardeira et al.A Robot for Fiber Insertion in a ProfileATL-TILECAL-PUB-2005-007.7480
[120] D.M.Gingrich et al.,Construction, Assembly and Testing of the ATLAS Hadronic End-capCalorimeterATL-LARG-PUB-2007-009; CERN-ATL-COM-LARG-2007-006, 2007 JINST 2P05005.
[121] ATLAS collaboration,ATLAS calorimeter performance, Technical Design Report, CERNLHCC-96-040.7485
[122] J. Ban et al.,Cold electronics for the liquid argon hadronic end-cap calorimeter of ATLASNucl.Instrum. Methods A (2006)556158-168.
[123] N.N. et al.,The ATLAS Forward Calorimeter.
[124] N.N. et al.,ATLAS liquid argon calorimeter front end electronics system.
[125] J.Ban et al.,ATLAS liquid argon calorimeter back end electronics2007 JINST 2 P06002.7490
[126] N.N. et al.,Electronic calibration board for the ATLAS liquid argon calorimeter.
[127] N.N. et al.,ATLAS LAr Front End Board.
[128] P.Adragna et al,A PMT-Block test bench, Nucl. Instrum. Meth.A (2006)564597-607.
[129] M.Crouau et al,Characterization of 8-stage Hamamatsu R5900 Photomultiplier for the TILECalorimeter, TILECAL-NOTE 97-129 (1997).7495
[130] K.Anderson et al,Front-end Electronics for the ATLAS Tile Calorimeter, Proceedings of the FourthWorkshop on Electronics for LHC Experiments (LEB 98), Roma 1998, page 239.
[131] S.Berglund et al,The ATLAS Tile Calorimeter Digitizer, Proceedings of 5th Workshop onElectronics for the LHC Experiments (LEB 99), Snowmass, Colorado, (1999) pages 255-259.
– 296 –
[132] K.Anderson et al,ATLAS Tile Calorimeter Interface, Proceeding of the 8th Workshop on Electronics7500
for LHC Experiments, Colmar, France, (2002) pages 269-273.
[133] A.S.Cerqueira et al.,Analog System for Building the First-Level Triggering Signal Provided by theHadronic Calorimeter of ATLAS DetectorNucl. Instrum. Methods A (2007)570117-125.
[134] E.Fullana et al,Digital Signal Reconstruction in the ATLAS Hadronic Tile Calorimeter, IEEE Trans.Nucl. Sci. Vol 53, No. 4 , 2006.7505
[135] V.González et al.,Development of the Optical Multiplexer Board Prototype forData Acquisition inthe TileCal SystemIEEE Trans. On Nuclear Science, vol. 53, 2006.
[136] W.E.Cleland, E.G.Stern,Signal Processing Considerations for Liquid Ionization Calorimeters in aHigh Rate Environment, Nucl. Instrum. Meth.A (1994)338467.
[137] M. Aharrouche et al.,Response Uniformity of the ATLAS liquid argon electromagnetic calorimeter7510
submitted to NIM.
[138] J.Carvalho,Calibration and monitoring of the ATLAS Tile Calorimeter, Proceedings of the X11International Conference on Calorimetry in High Energy Physics, (2006) p385.
[139] E.Starchenko et al,Cesium Monitoring system for ATLAS hadronic calorimeter, Nucl. Instrum.Meth. (2002)494508-523.7515
[140] Reference to LVPS must be written - its essential.
[141] A.M.Henriques-Correia, A.Karioukhine,G. Schlager: Further performance tests of the coolingsystem for the calibration of the ATLAS Tilecal modules, ATLAS internal note,ATL-TILECAL-2002-014, CERN, 14 Nov 2002.
[142] G.Blanchot et al,ATLAS Cooling System LCS v.2 Full Scale Test, ATLAS Project Document7520
ATC-TL-ES-0001, 2001.
[143] J. Colas et al.,Position resolution and particle identification with the ATLAS EM calorimeterNucl.Instrum. Methods A (2005)55096-115.
[144] M. Aharrouche et al.,Energy linearity and resolution of the ATLAS electromagnetic barrelcalorimeter in an electron test-beamNucl. Instrum. Methods A (2006)568601-623.7525
[145] M. Aharrouche et al et al.,Characterization of the ATLAS electromagnetic liquid argon calorimeterswith muonsfirst final draft (25 may 2007).
[146] M.Aharrouche,The ATLAS liquid argon calorimeter: Construction, Integration, Commissioning andCombined Test beam resultscontribution to XIth Vienna Conference on Instrumentation2007.
[147] N.N. et al.,Electromagnetic Energy Measurement in LAr Calorimeter at the CTB 2004draft7530
foreseen before end july and should replace the preceding reference.
[148] B. Dowler et al.,Performance of the ATLAS hadronic end-cap calorimeter in beam testsNucl.Instrum. Methods A (2002)48294-124.
[149] A.E. Kiryunin et al,GEANT 4 physics evaluation with testbeam data of the ATLAS hadronic end-capcalorimeterNucl. Instrum. Methods A (2006)560278-290.7535
[150] C. Cojocaru et al.,Hadronic calibration of the ATLAS liquid argon end-cap calorimeter in thepseudorapidity region1.6 < |η | < 1.8 in beam testsNucl. Instrum. Methods A (2004)531481-514.
[151] The ATLAS Liquid Argon EMEC/HEC Collaboration,Muon Results from the EMEC/HECCombined Run corresponding to the ATLAS Pseudorapidity Region 1.6 <| η |< 1.8ATL-LARG-2004-006 (2004).7540
– 297 –
[152] P. Schacht for the ATLAS Endcap Collaboration,Recent Results from the ATLAS Endcap CombinedTest BeamProceedings of the XII International Conference on Calorimetery in High Energy Physics,Chicago, 5-9 June 2006, 338-349 and NIM paper to come.
[153] N.N. et al.,Electron Signals in the Forward Calorimeter Prototype for ATLAS.
[154] N.N. et al.,Electron and Pion Response of the ATLAS FCal detector from 2003 testbeam data.7545
[155] The MDT parameter book, available at the web pagehttp://atlas.web.cern.ch/Atlas/GROUPS/MUON/layout/MDT-parameter-book.html.
[156] G. Brandenburg, J. Chapman et al., The readout electronics of the MDT chambers in the ATLASexperiment, Nucl. Instrum. Meth.A888 (2007) 941–960 (NOT YET PUBLISHED).
[157] H. Boterenbrood, P. Jansweijer, G. Kieft, A. Konig, J.Vermeulen, T. Wijnen, The Read-Out Driver7550
for the ATLAS MDT Muon Precision Chambers, IEEE Trans. Nucl.Sci.53 (2006) 741–748.
[158] C. Guyot, J. Bensinger et al., The alignment of the ATLAS Muon spectrometer, Nucl. Instrum.Meth.A888 (2007) 981–1000 (NOT YET PUBLISHED).
[159] J. R. Bensinger, ATLAS muon endcap alignment, Nucl. Phys. Proc. Suppl.150(2006) 136–139.
[160] H. L. Groenstege, The RASNIK/CCD 3D alignment system,(15/12/1994), ATL-MUON-94-063,7555
ATL-M-PN-63.
[161] H.v.d. Graaf et al., RASNIK Technical System Description for ATLAS, (Sept. 2000), ETR 2000-04.
[162] V. Polychronakos et al., Construction of the ATLAS CSCchambers, Nucl. Instrum. Meth.A888(2007) 821–840 (NOT YET PUBLISHED).
[163] First-Level Trigger Technical Design Report (TDR), CERN/LHCC/98-14 (30/06/1998).7560
[164] L. Levinson, O. Sasaki et al., The TGC trigger system, Nucl. Instrum. Meth.A888 (2007)1021–1040 (NOT YET PUBLISHED).
[165] S. Veneziano et al., The RPC trigger system, Nucl. Instrum. Meth.A888 (2007) 1001–1020 (NOTYET PUBLISHED).
[166] G. Mikenberg et al., Construction of the ATLAS TGC chambers, Nucl. Instrum. Meth.A888 (2007)7565
861–880 (NOT YET PUBLISHED).
[167] The TGC parameter book, http://www.cern.ch/Atlas/project/TGC/www/design/tgc.xls.
[168] The Calorimeter Trigger system of the ATLAS First Level Trigger, Members of the CalorimeterTrigger Groups, Journal to be defined, Spring 2007.
[169] J. Garvey et al., 2003.7570
[170] The Barrel Muon Trigger system of the ATLAS First LevelTrigger, Members of the Barrel MuonTrigger Groups, Journal to be defined, Spring 2007.
[171] The Endcap Muon Trigger system of the ATLAS First LevelTrigger, Members of the Endcap MuonTrigger Groups, Journal to be defined, Spring 2007.
[172] The Central Trigger system of the ATLAS First Level Trigger, Members of the Central Trigger7575
Groups, Journal to be defined, Spring 2007.
[173] Missing citation info.
[174] Missing citation info.
– 298 –
[175] The ATLAS Read Out Links, Journal to be defined, Spring 2007.
[176] The final design of the Trigger/DAQ Read-Out-Buffer Input, Journal to be defined, Spring 2007.7580
[177] The Region of Interest Builder, Journal to be defined, Spring 2007.
[178] H. Boterenbrood, H.J. Burckhart, J. Cook, V. Filimonov, B. Hallgren, W. Heubers, V.Khomoutnikov, Y. Ryabov, F. Varela, Design and Implementation of the ATLAS Detector ControlSystem, proceedings 2003 IEEE Real Time Conference, Montreal, Canada, May 2003.
[179] B. Hallgren, H.Boterenbrood, H.J. Burckhart, H.Kvedalen, The Embedded Local Monitor Board in7585
the LHC Front-end I/O Control System, proceedings 7th Workshop on Electronics for LHCExperiments, Stockholm, Sweden, September, 2001.
[180] http://www.etm.at.
[181] ATLAS Technical Coordination team, The assembly of the ATLAS experiment at CERN: designstudies, management and organisation activities, in preparation, editor: M. Hatch.7590
[182] C. Lasseur, Alignment of the LHC machine and experiments - Part 2: the experiments, CERNEDMS document 330354, https://edms.cern.ch/document/330354/1..
[183] Synthesis of the survey measurements and positioningstrategy for the components of the ATLASdetector, in preparation, editors: C. Lasseur and D. Lissauer..
[184] ATLAS Technical Coordination team, The assembly of the ATLAS experiment at CERN:7595
infrastructure provided at CERN for the ATLAS experiment, in preparation, editor: F. Butin.
[185] ATLAS Technical Coordination team, The assembly of the ATLAS experiment at CERN:installation activities, in preparation, editor: G. Tappern.
[186] XX et al., The non-evaprable getter pumping systems for the LHC experiments, ....
[187] R. Veness et al., The ATLAS beam vacuum system, ....7600
[188] ATLAS Detector and Physics Performance Technical Design Report, CERN/LHCC/99-15 (1999).
[189] A. Ahmad et al., Inner detector as-built detector description validation for CSC, ATLAS InternalNote ATLAS-INDET-INT-2007-002.
[190] R. K. Ellis et al.„ Nucl. Phys. B297 (1988) 221;L. DiLella in Proceedings of the Large Collider Workshop, edited by G. Jarlskog and D. Rein7605
(Aachen, 4-9 October 1990) CERN 90-10/ECFA 90-133, Vol. II,p. 530..
[191] M. Heldmann and D. Cavalli, An improved tau-Identification for the ATLAS experiment, ATLASPhysics Note ATL-PHYS-PUB-2006-008.
[192] E. Richter-Was and T. Szymocha, Hadronic tau identification with track based approach: theZ → ττ ,W → τν and dijet events from DC1 data samples, ATLAS Physics Note7610
ATL-PHYS-PUB-2005-005.
[193] ATLAS, ATLAS Forward Detectors for Measurement of Elastic Scattering and LuminosityDetermination (ATLAS TDR, in preparation); ATLAS Forward Detectors for LuminosityMeasurement and Monitoring (CERN-LHCC/04-10, LHCC I-014); Zero Degree Calorimeters forATLAS (CERN-LHCC/07-01, LHCC I-016).7615
[194] S. Asket al., Nucl. Instr. and Meth.A568 (2006) 588–600.
[195] F. Anghinolfiet al., Hadron beam test of a scintillating fibre tracker system forelastic scattering andluminosity measurement in ATLAS, JINST2 P07004(2007) 588–600.
– 299 –
