1. Dottorato di Ricerca in Fisica XIX Ciclo Universita' degli Studi di Torino – 14 Dicembre 2006 – Simulation, Performance and Local Track Reconstruction of the Drift Tube Detectors of the CMS Experiment. Candidato G. Cerminara

2. December 14, 2006 2 Overview The Large Hadron Collider (LHC) –motivation and experimental challenges The Compact Muon Solenoid (CMS) experiment –overview of the detector –tracking strategy –the muon spectrometer and the Drift Tubes (DT) Simulation of the DT response Local Reconstruction with the DT detectors –synchronization and calibration procedures –reconstruction of points and track segments –performance Analysis of DT test-beam data –validation of the DT simulation –performance Conclusions

3. G. Cerminara December 14, 2006 3 Standard Model and Higgs Boson Standard Model (SM) of Particle Physics very successful theory: –tested and confirmed with very high accuracy (quantum level) up to E ~100 GeV The mechanism which gives masses to all particles is unexplained: –A missing piece: the Higgs Boson Either it is found below ~1 TeV, or new physics must appear (e.g. with no Higgs, WW scattering violates unitarity) The SM appears as an “effective” theory at electroweak scale SM extensions –Higgs Mechanism valid up to given energy scale  : a more general theory should be valid above –Possibly a wider symmetry, broken at the scale  –All candidates predict a rich phenomenology to appear below scales of ~1 TeV New physics is expected below ~ 1 TeV but several issues are still open....

4. G. Cerminara December 14, 2006 4 Large Hadron Collider (LHC) LHC: pp-machine designed to discover these new physics: –High Luminosity L L =2x10 33 cm -2 s -1 (1 st period ~ 2008) L H =10 34 cm -2 s -1 (design luminosity) –High Energy collisions √s = 14 TeV Difficult experimental environment: –Low S/B → trigger selection is crucial:  pp inel ~ 55mb → event rate O(GHz)  signal = O(pb) → event rate O(10 -1 Hz) –Muons → a clear signature for many processes in the LHC hadronic environment 40 MHz → 1 BX every 25 ns

5. G. Cerminara December 14, 2006 5 Large Hadron Collider (LHC) LHC: pp-machine designed to discover these new physics: –High Luminosity L L =2x10 33 cm -2 s -1 (1 st period) L H =10 34 cm -2 s -1 (design luminosity) –High Energy collisions √s = 14 TeV Difficult experimental environment: –Low S/B → trigger selection is crucial:  pp inel ~ 55mb → event rate O(GHz)  signal = O(pb) → event rate O(10 -1 Hz) –Muons -> a clear signature for many processes in the LHC hadronic environment 40 MHz → 1 BX every 25 ns some processes with muons in the final state

6. G. Cerminara December 14, 2006 6 The Compact Muon Solenoid General purpose collider experiment → ~ 4  detector design CMS strength: –very precise electromagnetic calorimeter (ECAL) –redundant muon system (tracking and trigger) Design based on the choice of 4T superconducting solenoidal magnet Total weight : 12,500 t Overall diameter : 15 m Overall length : 21.6 m Magnetic field : 4 Tesla

7. G. Cerminara December 14, 2006 7 CMS Tracking Strategy Superconducting Solenoid: –4T magnetic field Bending in the transverse plane (  ) –independent tracking inside and outside the coil: inner silicon tracker muon spectrometer in the iron return yoke Muon spectrometer –key role in the CMS trigger –good p T resolution for standalone measurement –becomes fundamental for resolution at high p T B=4T for r <3m B~1.8 T in the Iron Yoke

8. G. Cerminara December 14, 2006 8 The Muon Spectrometer Muon spectrometer uses 3 types of gas detectors with trigger capabilities –Barrel & Endcaps: Resistive Plate Chambers (RPC) (|  | < 2.1) good time resolution:  t ≈ 2 ns → BX assignment –Endcaps: Cathode Strip Chambers (CSC) (0.8 < |  | < 2.4)  x ≈ 100 – 240  m / layer –Barrel: Drift Tubes (DT) pseudorapidity coverage: |  | < 1.2 4 stations of chambers 250 chambers → O(10 5 ) channels  x ≈ 200  m / layer

9. G. Cerminara December 14, 2006 9 Drift Tube Chambers Each DT chamber is composed by: –2 superlayers (SL) measuring the bending coordinate → r-  SLs –1 SuperLayer (SL) measuring the track angle w.r.t. the beam line → r-z SLs Each SL is a quadruplet of cell layers staggered by half a cell –layer structure allows to generate a trigger within the chamber (autotrigger) Working principle: –Convert a drift-time into a drift-distance –Build track segments within the chamber r-  SLs r-z SL (no in MB4) Honeycomb spacer

10. G. Cerminara December 14, 2006 10 Drift Tube Cell Drift cell: 13 x 42 mm 2 cell –Ar/CO 2 (85%/15%) gas mixture: good quenching properties and saturated drift velocity –Field shaping obtained with central stripes: good linearity of space-time relation: v drift ~ 54  m/ns maximum drift time ~ 390 ns

11. G. Cerminara December 14, 2006 11 Outline From now on I will focus on my contribution to the DT software Simulation of the DT response –behaviour of the cell Local Reconstruction with the DT detectors –synchronization and calibration procedures –reconstruction of points and track segments –performance Analysis of DT test-beam data –validation of the DT simulation –performance of the local reconstruction

12. G. Cerminara December 14, 2006 12 Outline Simulation of the DT response –behaviour of the cell Local Reconstruction with the DT detectors –synchronization and calibration procedures –reconstruction of points and track segments –performance Analysis of DT test-beam data –validation of the DT simulation –performance of the local reconstruction

13. G. Cerminara December 14, 2006 13 Simulation of the Cell Response The simulation of the cell response must: –model the interaction of crossing particles with the detector material: bending in the magnetic field energy loss production of secondary particles –reproduce the response of the cell to ionizing particles crossing the gas volume: the drift properties depend on: –track impact angle , –magnetic field in the sensitive volume B parametrization of the drift cell based on a detailed GARFIELD simulation: t drift,  p,  n = f (x, , B) ] GEANT 4 Allows to reproduce statistically the arrival time distribution   - ray

14. G. Cerminara December 14, 2006 14 Effect of the Track Impact Angle Large impact angles occur especially in r-z superlayers: –  grows with the pseudorapidity . The effect of the impact angle must be considered in the simulation  = 41.6 o  = 56.5 o muon sample with 10 < p T < 100 GeV/c tails due to secondary particles

15. G. Cerminara December 14, 2006 15 Effect of the Track Impact Angle (II) The track with ≠ crosses many isochronal lines: –higher effective drift velocity –effect of cell non-linearities enhanced –lower resolution  = 0  ≠ 0 B = 0 Non-linearities are more important close to the anode The effective drift velocity grows with the angle anod e cathod e example:

16. G. Cerminara December 14, 2006 16 Effect of the Magnetic Field The DT chambers are placed in the return yoke: –the space should ideally be field-free but residual field is present in the iron gaps between the wheel and at the end of the coil The effect of the residual magnetic field is therefore considered in the simulation 4T4T Magnetic Field B Z DT chambers

17. G. Cerminara December 14, 2006 17 Effect of the Magnetic Field The DT chambers are placed in the return yoke: –the space should ideally be field-free but residual field is present in the iron gaps between the wheel and at the end of the coil The effect of the residual magnetic field is therefore considered in the simulation IP Wh. 0Wh. 1Wh. 2Wh. 0Wh. 1Wh. 2 iron gaps muon sample with 10 < p T < 100 GeV/c

18. G. Cerminara December 14, 2006 18 Effect of the Magnetic Field (II) The magnetic field modifies the shape of the drift lines: B wire → effect on and linearity B norm → only effect on B drift → effects are negligible –GARFIELD parametrization: t drift,  p,  n = f (x, , B wire, B norm ) B wire ≠ 0 anod e cathod e example:

19. G. Cerminara December 14, 2006 19 The Drift Time Distribution Drift time distributions obtained with the GARFIELD parametrization – effects of  and B on v drift and non-linearities Further time contributions added to tdrift to simulate a TDC measurement –time-of-flight (TOF) of the muon from the intereaction point (IP) –time for the signal propagation along the anode wire –time pedestals (trigger latency, cable path, electronic delays....) T max ~ 390ns

20. G. Cerminara December 14, 2006 20 Outline Simulation of the DT response –behaviour of the cell Local Reconstruction with the DT detectors –synchronization and calibration procedures –reconstruction of points and track segments –performance Analysis of DT test-beam data –validation of the DT simulation –performance of the local reconstruction

21. G. Cerminara December 14, 2006 21 Reconstruction in DT Chambers Local reconstruction in DT chambers is performed in three steps: –1D hits @ cell level: the drift time is converted in a drift distance from a wire (with intrinsic Left/Right ambiguity) –2D segments @ r-  and r-z superlayers: 1D hits are used to fit 2D segments independently in r-  and r-z SLs r-  SLs: up to 8 hits in the fit r-z SL: up to 4 hits in the fit –3D segments @ chamber level: the two projections are combined to build a 3D segment in the chamber (which will be used in the track fit) The pattern recognition + segment building allow to: –minimize the effect of soft secondaries –know three-dimensional position of the ionizing event –measure the position and direction → improved resolution with respect to 1D hits

22. G. Cerminara December 14, 2006 22 From TDC Times to Distances Measurement of drift distances at the cell level: The measurement of the drift distance x requires the knowledge of: –time pedestals needed to extract t drift from t TDC : synchronization procedure to evaluate the pedestals –space-time relation for the drift electrons: two reconstruction algorithms implemented: –constant drift velocity x = v drift t drift requires calibration of the drift velocity v drift –parametrization of the cell behaviour x = f (t drift, , B wire, B norm ) Simulatio n DAQ TDC measurement t TDC = t drift + delays x x drift distance x = x(t drift )

23. G. Cerminara December 14, 2006 23 From TDC Times to Distances Measurement of drift distances at the cell level: The measurement of the drift distance x requires the knowledge of: –time pedestals needed to extract t drift from t TDC : synchronization procedure to evaluate the pedestals –space-time relation for the drift electrons: two reconstruction algorithms implemented: –constant drift velocity x = v drift t drift requires calibration of the drift velocity v drift –parametrization of the cell behaviour x = f (t drift, , B wire, B norm ) Simulation DAQ TDC measurement t TDC = t drift + delays x x drift distance x = x(t drift )

24. G. Cerminara December 14, 2006 24 Synchronization Procedure TDC measurements need to be synchronized to extract the drift time information: t TDC = t drift + TOF + t prop_wire + offset(electronic delays, trigger latency...). Synchronization procedure consists of: –determination relative offset between channels → t 0 due to different path in the read-out electronics hardware procedure: test pulses are sent to the front-end –determination of the absolute value of the pedestal (SL by SL) → t trig accounts for the average TOF, t prop_wire and offset. software (off-line) procedure: directly measured from the TDC time distribution (no external reference) –further corrections for TOF and t prop_wire applied event by event as soon as the 3D position of the hit is known (during segment building) time pedestal

25. G. Cerminara December 14, 2006 25 t tri g Determination of the t trig Pedestal The absolute value of the pedestal (t trig ) is measured directly from the TDC time distribution (SL by SL): –fully automatic fitting procedure is needed to handle t trig computation for ~3 superlayers x 250 chambers must be robust against Time Box distortions such as noise, after pulse peaks... –fit of rising edge of the time box with the integral of a Gaussian the beginning of the rising edge estimated with: t trig = - k  k → tuning of the t trig to minimize the residuals on the reconstructed distance

26. G. Cerminara December 14, 2006 26 From TDC Times to Distances Measurement of drift distances at the cell level: The measurement of the drift distance x requires the knowledge of: –time pedestals needed to extract t drift from t TDC : synchronization procedure to evaluate the pedestals –space-time relation for the drift electrons: two reconstruction algorithms implemented: –constant drift velocity x = v drift t drift requires calibration of the drift velocity v drift –parametrization of the cell behaviour x = f (t drift, , B wire, B norm ) Simulation DAQ TDC measurement t TDC = t drift + delays know n x x drift distance x = x(t drift )

27. G. Cerminara December 14, 2006 27 Reconstruction with Constant v drift good linearity of the drift field → hit distance from the wire reconstructed as: x = v drift t drift –v drift dependence on B and  only accounted “on average” through the calibration procedure calibration with the meantimer technique –from geometric considerations: –different meantimer relations for different track angles and cell patterns average meantimer computed (weighted average) –the drift velocity is estimated as: –the procedure also allows to estimate the uncertainty associated to each reconstructed distance maximum drift time of electrons crossing a semi-cell (L/2) v drift = L/2 L/2

28. G. Cerminara December 14, 2006 28 Reconstruction with the Cell Parametrization Time-to-distance relation within the cell parametrized with GARFIELD: x = f (t, , B wire, B norm ) –the information about , B wire, B norm is not available at the level of individual hit → an iterative procedure is used: step 1step 1: hit at the cell level: –  : pointing to the IP –B estimated at the mid-point of the wire step 2step 2: hit used to build 2D segment (r-  or r-z views) –  : determined by the fit and used as input to the parametrization step 3step 3: hit used to build 3D segment (r-  and r-z) –hit position along the wire determined → better estimation of B –After each step the distance is re-computed and the fit is re-done. The parametrization accounts for: –the impact angle of each track and local variations of the B field –cell non-linearities

29. G. Cerminara December 14, 2006 29 step 1 step 2 step 3 simulation muons 10 < p T < 100 GeV/c Performance of the Reconstruction The resolution improves at each step: –Using GARFIELD parametrization: –At the step 3 the resolution is below 170  m for all the superlayers (r-  and r-z) r-  superlayers (bending coordinate) residuals on the reconstructed distance from the wire

30. G. Cerminara December 14, 2006 30 step 1 step 2 step 3 simulation muons 10 < p T < 100 GeV/c Performance of the Reconstruction the resolution improves at each step: –Using GARFIELD parametrization: r-  superlayers (bending coordinate) residuals on the reconstructed distance from the wire asymmetric tail due to  -rays + TDC dead time   - ray dead time (150 ns) v drift t dead- TDC

31. G. Cerminara December 14, 2006 31 Parametrization vs Constant v drift GARFIELD parametrization improves the resolution w.r.t. constant v drift –especially where  and B are big → linearities more important (eg. SLs r-z in wheels ±2) Calibration of the time pedestal more critical for the parametrization –systematic bias in the t trig → wrong parametrization of the non-linearities simulation muons 10 < p T < 100 GeV/c anod e cathod e anod e cathod e residuals on the reconstructed distance vs distance from the wire (Step 1) parametrizatio n constant v drift  ~ 274  m  ~ 453  m

32. G. Cerminara December 14, 2006 32 Performance of Segment Reconstruction The segment reconstruction allows to: –measure the impact angle of the muon track –improve the resolution on the track position w.r.t. signle layer r-  coordinates (bending) in 3D segments  ~ 63  m  ~ 0.5 mrad residuals on the segment position residuals on the segment angle simulation muons 10 < p T < 100 GeV/c

33. G. Cerminara December 14, 2006 33 Outline Simulation of the DT response –behaviour of the cell Local Reconstruction with the DT detectors –synchronization and calibration procedures –reconstruction of points and track segments –performance Analysis of DT test-beam data –validation of the DT simulation –performance of the local reconstruction

34. G. Cerminara December 14, 2006 34 DT Test on Muon Beam In October 2004 two DT chambers tested on a muon beam at CERN. Goals: –test of chamber functionalities –test of Level-1 trigger electronics (local and regional trigger) –comparison between real data and official CMS simulation (CMSSW) –test of the local reconstruction and calibration procedures Setup of the Test Beam (TB): –two chambers on 40MHz bunched muon beam –different beam energies and angles –with/out iron absorbers between chambers  MB 1 MB 3 schematic top view No B field available!

35. G. Cerminara December 14, 2006 35 Simulation and Event Selection The Test Beam setup has been simulated with the official CMS simulation –geometry and material description –different running conditions Data selected to discard effects not interesting for the comparison –double-muon events –pile-up from previous bunch crossing in the spill Output of the simulation compared with the real data

36. G. Cerminara December 14, 2006 36 Comparison of Hit Multiplicity simulation vs data Comparison of hit multiplicity (simulation vs data) is a test of: –GEANT thresholds for the production of secondary particles –handling of secondary particles by the GARFIELD parametrization –weight of effects not considered in the simulation (eg. after-pulses, inefficiencies, pile-up from other BX, electronic effects....) –good agreement in the shape of the distribution –smaller # of measurements in the simulation w.r.t. real data (-4% and -6%) iron slabs no iron slabs # of measurements per event in one chamber higher multiplicity E = 100 GeV

37. G. Cerminara December 14, 2006 37 Secondary Particles and Meantimer Meantimer allows to estimate the contribution of  -rays and other secondaries to the # of measurements –  -ray → lower meantimer, example:  -ray in layer 4 → t 4 ↓ → T MAX ↓ –The meantimer peak corresponds to T MAX = (L/2) / v drift the simulation well reproduces the average drift velocity (@ all angles) t4t4 E = 100 GeV 0 o impact angle iron slabs no iron slabs secondaries T MAX → v drift

38. G. Cerminara December 14, 2006 38 Drift-time Distributions Good agreement of the drift-time distributions –the simulation well reproduces: cell non-linearities (@ all angles) time response of the cell  = 0 o  = 10 o  = 30 o  = 20 o E = 150 GeV – No Iron Slabs impact angle 

39. G. Cerminara December 14, 2006 39 Local Reconstruction of TB Data Comparison of the reconstruction on real and simulated data: –test the simulation of the cell resolution for different track angles –compare the performance of the GARFIELD parametrization and of the constant v drift on real data Same calibration procedure applied to real and simulated data –calibration of v drift –calibration of the time pedestals (t trig ) Residuals on reconstructed distance computed without external tracking system –comparison of reconstruction with segment extrapolation

40. G. Cerminara December 14, 2006 40 Cell Resolution Comparison of the resolution on the distance from the wire –the width of residual on the reconstructed distance is in good agreement @ all angles resolution well reproduced effect of the track angle accurately simulated  = 0 o  = 30 o real data simulatio n real data simulatio n  ~ 185  m  ~ 250  m

41. G. Cerminara December 14, 2006 41 Cell Resolution (II) Residuals on the reconstructed distance vs distance from the wire –very good agreement in the behavior of the residuals in the various cell regions –non-linearities at large angle well reproduced  = 30 o real data simulatio n

42. G. Cerminara December 14, 2006 42 Parametrization vs Constant v drift Test beam data used to test the performance of the reconstruction using the GARFIELD parametrization –comparison with performance of constant drift velocity The parametrization allows to improve the resolution from 3% to 10% depending on the impact angle width of the residuals obtained with the two algorithms improve the resolution @ large impact angles

43. G. Cerminara December 14, 2006 43 Parametrization vs Constant v drift The improvement in the resolution obtained taking into account the cell non-linearities  = 10 o parametrizatio n constant v drift  = 30 o parametrizatio n constant v drift

44. G. Cerminara December 14, 2006 44 Further Tests of Local Reconstruction Summer 06: Magnet Test and Cosmic Challenge –combined data taking of all CMS sub-detectors with final read-out and trigger electronics DT reconstruction running on line for DQM purposes Calibration algorithms used by the whole user community DT segments one of the first muons bending in the CMS magnetic field HCAL hits ECAL hits tracker hits

45. G. Cerminara December 14, 2006 45 Further Tests of Local Reconstruction Summer 06: Magnet Test and Cosmic Challenge –combined data taking of all CMS sub-detectors with final read-out and trigger electronics DT reconstruction running on line for DQM purposes Calibration algorithms used by the whole user community DT segments one of the first muons bending in the CMS magnetic field HCAL hits ECAL hits tracker hits The DT reconstruction and calibration code demonstrated to fit the requirements of: on-line reconstruction user community

46. G. Cerminara December 14, 2006 46 My Work in the Muon Project The CMS Muon Project is a wide international collaboration My work within this group is focused on various aspects of DT operations –DT chamber production development of software needed during production phase –DT chamber test and commissioning shifts for test and cosmic data taking to commission DT chambers development of software tools to analyse data –Development and maintenance of code for DT local reconstruction in the framework of official CMS Reconstruction Code (ORCA–CMSSW) the code needed to use DT chambers for physics –Data Quality Monitoring (DQM) applications –DT Cosmic Challenge and Magnet Test DQM tasks and data taking

47. G. Cerminara December 14, 2006 47 Conclusions The muon reconstruction plays a key role for the CMS physics programme Reliable simulation of the DT chamber response –tested and compared against real test-beam data → good agreement Robust local reconstruction and calibration –tested on real data: test-beam chamber commissioning Magnet Test & Cosmic Challenge –fits the requirement of different and user community...DT reconstruction is ready for data taking!

48. Backup Slides

49. G. Cerminara December 14, 2006 49 DT Commissioning in CMS After the production chambers are shipped from local sites to CERN DTs are then installed in CMS and they are tested. Goals of the test are: –certify that the chamber is operational Test of the chamber electronics (MiniCrate) Test of chamber performance with cosmic muons (1 chamber at a time in auto-trigger mode) –test (and update!) CMS reconstruction software to run on real data!

50. G. Cerminara December 14, 2006 50 Local Reconstruction in the DTs Hit position in the cell estimated from TDC measurement: two reconstruction algorithm implemented in ORCA code –constant drift velocity in the whole cell –time-to-distance relation within the cell parametrized with GARFIELD: x(t) = f(t, , B wire, B norm ) the information about , B wire, B norm is not available at the level of individual hit → an iterative procedure is used Fit 2D segments separately in R-  (8 layers) and R-z SLs (4 layers) –L/R ambiguities solved by best  2 –Update x(t) using information on impact angle  and refit Combine R-  and R-z 2D segments in a 3D segment –Update x(t) using best knowledge on B wire, B norm and refit –Resolution in R-  plane: position ~ 100  m, direction ~1mrad up to 12 hits/station

51. G. Cerminara December 14, 2006 51 Local Reconstruction in the CSCs and RPCs Cathode Strip Chambers: –  coordinate measured by strips charge distribution on a cluster of adjacent strips fitted with “Gatti” function to determine the centroid –r coordinate measured by wires read out in bunches to limit number of channels → limited precision –Use the measured points to fit a 3D linear segment Resolution: 120 – 250  m for the bending coordinate Resistive Plate Chambers: –Measure 3D points clustering the strips: up to 6 points in the barrel and 4 in the endcaps Uncertainties: L / √12 L  ~ O(1 cm) and L z ~O(100 cm)  (r)(r) up to 6 hit/station

52. G. Cerminara December 14, 2006 52  Reconstruction Software Almost same algorithms for HLT and off-line reconstruction HLT muon reconstruction performed in two logical steps (goal: reject events as soon as possible): –Level-2: uses the muon system data only –Level-3: uses data from muon system + tracker hits Track reconstruction using Kalman Filter –off-line vs HLT → different seeds for Kalman Filter and access to calibration data –Base ingredients for reconstruction: objects locally reconstructed in the detectors (RecHits) reconstruction “on demand” in a region of interest defined from the track extrapolation

53. G. Cerminara December 14, 2006 53 Track Reconstruction Track reconstruction with a Kalman Filter –recursive method to fit a discrete set of data independently of the number of measurements available –determine a generic state vector (= position and momentum + covariance matrix) Kalman Filter → initial seed: –HLT reconstruction: seeded by L1-trigger candidates (external seeding: faster) 4 best muons from the Global Trigger: –p T, position, angle, BX and quality –L1 p T resolution: 17 – 22 % depending on  –Efficiency: ~ 97 % –Off-line reconstruction: seeded by local segment reconstruction (internal seeding: no dependency on L1)

54. G. Cerminara December 14, 2006 54 Level-2 Reconstruction (Standalone) Initial state extrapolated from track seed Hits to be included in the fit are looked for iteratively inside-out, at each step: –extrapolate the trajectory to next layer of chambers –perform local reconstruction in the chambers on the path –if RecHits are compatible (  2 test) update the state vector using the Kalman filter –DTs: 3D segments are used for the fit –CSCs, RPCs: 3D hits are used (inhomogeneous magnetic field) The obtained state (at the last station) is used to perform the actual fit going outside-in. –Final estimate at the innermost muon station Trajectory cleaning, ghosts suppression and “smoothing” Vertex extrapolation and constraint –Extrapolate the track to the point of closest approach to the beam including the nominal interaction point with its spread (  = 20  m) in the fit

55. G. Cerminara December 14, 2006 55 Level-3 Reconstruction (Global) Start defining a Region of Interest (ROI) in the tracker using standalone (Level-2) track (with vertex constraint) –efficiency and performance crucially depend on Level-2 reconstruction Regional reconstruction performed in the ROI –Seeds generated from pixels or double sided  -strip detectors + vertex constraint –Again Kalman filter (inside-out) up to 30 tracks grown in parallel tracks are discarded if no hit in more than 4 consecutive layers –Trajectories which share more than half of the hits are selected on the basis of their  2. Tracks are fitted using also “standalone” reconstructed muons in the spectrometer. Great improvement in resolution with respect to “standalone” reconstruction

56. G. Cerminara December 14, 2006 56 HLT Performance: Efficiency HLT efficiency for muons with p T = 5 - 100 GeV/c p T (GeV/c) L1 Trigger accepatance |  | < 2.1 Overall efficiency ~ 96 % –Low energy muons → low efficiency because of large multiple scattering –Efficiency drops (~90%) in the cracks between wheels HLT acceptance is limited to |  | < 2.1 because no L1 trigger electronics on ME 1/1 low p T → multiple scattering cracks between wheels ||||

57. G. Cerminara December 14, 2006 57 HLT Performance: Resolution Good resolution Tails under control (very important for trigger rates) Big improvement using tracker hits  = 0.013  = 0.018  = 0.12  = 0.17 BarrelEndcapsLevel-2 Level-3 Resolution 1/p T (single  5-100 GeV/c) Level-2 Level-3 resolution x 10

58. G. Cerminara December 14, 2006 58 Off-line Performance: Resolution Standalone (spectrometer only) Global (spectrometer + tracker)  dependency due to solenoidal B field High p T muons (~1TeV): –showering in the chambers → difficult Local Reconstruction –energy loss → bias New reconstruction strategies under study  p T /p T barrelendcapbarrelendcap

59. G. Cerminara December 14, 2006 59 CMS Trigger Design Innovative (No Level-2 dedicated hardware) multilevel trigger design: –Level-1 Trigger: implemented on dedicated hardware Calorimeter and muon data (coarse granularity) Dedicated hardware → minimum dead time –Input from detector: 40 MHz –Output to DAQ: ~100 kHz –High Level Trigger (HLT): software running on a farm of commercial processors Uses as much as possible “off-line quality” data –Output: max rate for storage O(100) Hz 1 event ~ 1MB 40 MHz 100 kHz 100 Hz General selection tool: p T thresholds on particles

60. G. Cerminara December 14, 2006 60 Magnetic Field Superconducting Solenoid –r = 3m, L=14m –B = 14T within the solenoid –B ~ 1.8T in the iron return yoke Great bending power Independent measurement inside / outside A lot of material within chambers Field measurement: –During Magnet Test (2006) Rotating arm instrumented with Hall and NMR probes: –  r = 20 cm,  z = 5 cm –NMR probes inside the solenoid for on-line monitoring 4T4T Magnetic Field B Z

61. G. Cerminara December 14, 2006 61 Muon System Alignment Chamber alignment is fundamental –chamber resolution ~100  m –movements due to B on /B off : O(1cm)! Optical alignment system –rigid structures + optical links (LED, laser, CCD) –link system for alignment with tracker –performance:  r  ~150  m (same sector)  r  ~210  m (between sectors) Alignment with tracks –Problem: knowledge of material and magnetic field Only muons with p T > ~50 GeV/c are usefull

62. G. Cerminara December 14, 2006 62 L1 Trigger General Design Implemented on custom hardware –minimal dead time Synchronous, pipelined (25 ns) –delayed by 3.2  s = 128 BX including propagation (~1-2  s) Max output  max DAQ input –Design: 100 kHz; at startup: 50 kHz 2 Subsystems –Calorimeter Trigger –Muon Trigger –Result: jet, e/    jet candidates; E T miss,  E T No local decisions; selection by the “Global Trigger” –128 simultaneous, programmable algorithms, each allowing: Thresholds on single and multiple objects of different type Correlations, topological conditions, Prescaling

63. G. Cerminara December 14, 2006 63 L1 Muon Trigger Local” (chamber) level (ASICs) –Find segments in DT/CSC chambers “Regional” (subsystem) level (FPGAs) –RPC: Pattern Comparator (PACT) looks for predefined patterns –DT/CSC Track Finders combine segments into track; assign p T Global Muon Trigger –Combines candidates from DT, CSC,RPC Exploits complementarity of systems –Delivers 4 best muons to the Global Trigger Each with p T, position, angle, BX, quality –Efficiency: ~97% –p T resolution: 17-22% depending on  (muons from W decays)

64. G. Cerminara December 14, 2006 64 DTs, CSCs L1 Trigger DTCSC (strips) Local Trigger: build segmentsRegional Trigger: DT/CSC Track Finders Extrapolation Unit –Link segments using look-up tables Track Assembler: –link segment pairs to tracks –cancel fakes Assignment Units –p T, charge, , , quality FPGAs in the control roomASICs on detector or peripheral crates FIXME: remove CSCs

65. G. Cerminara December 14, 2006 65 RPC Level-1 Trigger Based on Pattern Comparator (PACT) –Look for predefined hit patterns in time coincidence At least 3 hits out of 4 stations 2 different groups of 4 stations in the barrel (6 stations in total) –Each hit pattern corresponds to a p T value –Hardware: ASICs Located in the counting room

66. G. Cerminara December 14, 2006 66 Trigger Rate ℒ  = 10 34 cm -2 s -1 30 Hz

67. G. Cerminara December 14, 2006 67 The Kalman Filter Kalman Filter is a recursive method for the fit of a discrete set of data –Provides track fitting independent of the number of measurement available PROBLEM: Determination of a generic state vector x (= position + momentum to a given surface) given a set of measurement z k. METHOD: –Start from a seed (= initial state vector + covariance matrix) –Each step k consists of two phases: propagation: predict an a priori state x k - obtained by projecting the previous state with its covariant matrix update: use the information from all measurement to update the state x k and the covariance matrix RESULT: –the result is the state on the surface of the last measurement station SMOOTHING: –Update the trajectory parameters of previous steps using all the measurement at every measured surface.