CMS Silicon Strip Tracker: Operation and Performance

Size: px

Start display at page:

Download "CMS Silicon Strip Tracker: Operation and Performance"

Tyler Eaton
5 years ago
Views:

1 CMS Silicon Strip Tracker: Operation and Performance Laura Borrello Purdue University, Indiana, USA on behalf of the CMS Collaboration

2 Outline The CMS Silicon Strip Tracker (SST) SST performance during 2010 p-p collisions Status of the SST calibration Analysis Conclusion

3 CMS Silicon Strip Tracker (SST) The CMS SST is the largest silicon strip detector ever built Details Volume: 24.4 m 3 Active Area 198 m modules ~ 9.3 M read-out channels 3.8 T magnetic field 2.4 m TOB TIB TID TEC Pixel BARREL Double-sided modules Single-sided modules TIB: 4 layers with 320 µm Si sensors TOB: 6 layers with 500 µm Si sensors ENDCAP TID: 3 disks per side with 320 µm Si sensors TEC: 9 disks per side with 320 and 500 µm Si sensors in 4 inner rings and 3 outer rings

4 SST operation SST operation - bias voltage applied 300 V - working in 3.8 T magnetic field - analogue readout (pedestal and CM noise subtracted): deconvolution mode - coolant temperature ~ 4 ºC ~ room temperature for sensors and hybrids Modules can be readout in peak or in deconvolution Peak: readout single point from shaper Deconvolution: combine signal from 3 readout buckets to shorten pulse duration - standard operation mode needed to minimize pileup - higher noise compared to peak mode Pulse height [ADC count] Pulse shape Peak --- Deco Time [ns]

Data taking SST commissioned and studied with cosmic data in 2008/09 Successfully operated during collision data in 2009 Since 30 th March SST included in data taking after LHC reaches stable beam

5 Data taking SST commissioned and studied with cosmic data in 2008/09 Successfully operated during collision data in 2009 Since 30 th March SST included in data taking after LHC reaches stable beam condition 7 TeV proton-proton collisions Fraction of SST operational channels in 2010: 98.1% TIB/TID 96.25% TOB 98.8% TEC+ 98.8% TEC- 99.1% CMS integrated luminosity Stable beams period in 2010 CMS Efficiency 91% Only 8% of CMS downtime is due to SST L (nb -1 ) Date

Data Quality Monitoring (DQM) First check on SST performance is done using the DQM system - Online: give prompt feedback during data taking - Offline: analyze the full statistics and certify data The

6 Data Quality Monitoring (DQM) First check on SST performance is done using the DQM system - Online: give prompt feedback during data taking - Offline: analyze the full statistics and certify data The full reconstruction chain is monitored through histograms on Raw data (readout and unpacking error) Local reconstruction (Digi, Cluster, on/off track clusters) Global track parameters (Track reconstruction) Hit residuals (Alignment) Since SST has fine granularity, DQM produces ~300K histograms Specific tools provided to check the SST performance - summary histograms - automatic quality test - synoptic view of the detector Fraction of working detectors for each SST layer green corresponds to a fraction >95%) 8 th June 2010 IPRD10 L. Borrello CMS TEC- Silicon TEC+ Strip Tracker TIB Operation TID- and Performance TID+ TOB

reprocessing of data - high Signal-to-Noise (S/N) in agreement with deconvolution

7 Good performance Cluster reconstruction - noise from commissioning runs stable with time - low cluster occupancy ~10-5 few noisy channels identified and removed in the reprocessing of data - high Signal-to-Noise (S/N) in agreement with deconvolution readout S/N for on-track cluster corrected for the path length for the barrel SST in pp collision at 7 TeV

SST is operated in deconvolution mode Timing optimization Timing should be further optimized wrt peak readout mode Synchronization is achieved by scanning the clock phase In December 2009, fine delay

8 SST is operated in deconvolution mode Timing optimization Timing should be further optimized wrt peak readout mode Synchronization is achieved by scanning the clock phase In December 2009, fine delay scan was done only for one TOB layer Procedure repeated in April for all SST sub-systems with 7 TeV pp collisions Procedure Tracker operated in peak mode (used as a telescope for the measurement) - one layer is in deconvolution - the delay of layer under study is scanned in steps of 2ns over a window of [-25ns,+25ns] - best delay is obtained by maximizing the signal of the leading strip of cluster on tracks New timing offset uploaded preliminary results confirmed the expected improvements in S/N (~ 4%) 2009 result for TOB

Track reconstruction Track reconstruction done in 3 steps Seeding Pattern recognition Track fitting Seeding by pixel hit triplets or pixel/strip

fitting: track parameters estimate Procedure iterated 6 times At each iteration hits assigned to track are removed and seed cuts are relaxed Event

9 Track reconstruction Track reconstruction done in 3 steps Seeding Pattern recognition Track fitting Seeding by pixel hit triplets or pixel/strip hit pairs with constraint from the beam spot to identify track candidate Pattern recognition: track candidate propagation (Kalman filter) Track fitting: track parameters estimate Procedure iterated 6 times At each iteration hits assigned to track are removed and seed cuts are relaxed Event selection: good collision events: trigger and vertex selection Track parameters in good agreement with MC samples more info in the talk of Bernardini

10 OFFLINE CALIBRATION

11 Calibration workflow Standard procedure in CMS is based on - Almost full automatic procedure: prompt calibration on Express Stream - Prompt reconstruction after ~48 hours with updated conditions Two different data streams are used and results are uploaded to the condition database Channel status Identification of bad channels Hit efficiency, Lorentz Angle (LA) and Gain calibration Automatic production of the calibration n-tuple Simple script to perform calibration analysis Data taking in started with no delay between Express and Prompt reconstruction Results from calibration workflow used for Monitoring the detector status Reprocessing of data Realistic MC processing

12 Channel status calibration Procedure - channels with high occupancy are identified - run-by-run analysis automatically produced - results uploaded to the condition database - information of faulty components is taken into account in track reconstruction Procedure ready to be used for prompt reconstruction No prompt calibration loop yet Results: On average, ~ 0.1% bad channels identified and masked

13 Gain calibration A uniform response across the modules is needed for de/dx studies and Data/MC comparison Two gain measurements available From tickmark (height of the APV digital signal) to equalize the different readout chains From particle to equalize different response of the sensors Particle gain is calculated using 7 TeV collision runs For the first time, statistics is sufficient to compute the particle gain for single readout chip (APV) gain derived for nearly all of the APVs Particle calibration is also used in MC

Lorentz Angle Magnetic field produces a modification of the cluster charge spread on the detector strips the drift direction is tilted by the Lorentz angle (LA) Measurement of LA using cluster width

14 Lorentz Angle Magnetic field produces a modification of the cluster charge spread on the detector strips the drift direction is tilted by the Lorentz angle (LA) Measurement of LA using cluster width as a function of track impact angle - Lorentz drift θ L Track angle θ T Results from cosmic data TIB: tan(θ L ) = 0.07 ± 0.02 TOB: tan(θ L ) = 0.09 ± 0.01 An additional corrections is needed for deconvolution mode Fraction of the charge from sensor backplane does not reach strips in time for readout - Reconstructed hit position is biased - Effect on the Lorentz Angle - corrections measured using collision data and uploaded into the database - results validated by the alignment group and used in the reprocessing of data

Module Hit Efficiency The efficiency is calculated by finding track trajectories that pass through a given module and looking for a reconstructed hit (sensor edges are removed to avoid systematic due

15 Module Hit Efficiency The efficiency is calculated by finding track trajectories that pass through a given module and looking for a reconstructed hit (sensor edges are removed to avoid systematic due to track reconstruction) Almost all layers have an efficiency larger than 99.9% (excluding known bad modules) Few inefficient modules found - 2 in TIB, 1 in TOB, 8 in TEC- are masked offline for re-reconstruction - sources of inefficiency under investigation Hit efficiency per layer/disk for a 7 TeV run Stable behavior: No change of hit efficiency with time observed All Modules Good Modules

16 ANALYSIS

Hit resolution Hit resolution is calculated comparing the predicted position from the track fitting with the position of the hits Study is done considering the overlap region (same layer) to minimize

17 Hit resolution Hit resolution is calculated comparing the predicted position from the track fitting with the position of the hits Study is done considering the overlap region (same layer) to minimize - the amount of material between two layers - the effects of track extrapolation Hit resolution depends on sensor thickness and strip pitch - Minimum value achieved for an angle corresponding to optimal charge sharing Module TIB12 TIB34 TOB1-4 TOB56 Pitch (micron) Thickness (micron) Results 8 th June 2010 are IPRD10 excellent: measured L. Borrello resolution CMS is Silicon ~15 Strip (20) Tracker µm Operation in TIB and (TOB) Performance sensors

Particle Identification Energy loss in silicon strip sensors used for particle identification Relation among the particle momentum p and the de/dx

selected candidates shows kaon, proton and deuteron peaks - High purity tracks with 12 SST hits, p<2 GeV and de/dx>5 MeV/cm - good match between data and

18 Particle Identification Energy loss in silicon strip sensors used for particle identification Relation among the particle momentum p and the de/dx estimators can be used to evaluate the mass of the candidate de/dx = K m 2 /p 2 + C - proton line is used to extract the parameters K, C Mass plot of selected candidates shows kaon, proton and deuteron peaks - High purity tracks with 12 SST hits, p<2 GeV and de/dx>5 MeV/cm - good match between data and MC - suppression of Deuterons in MC 8 th June 2010 IPRD10 L. Borrello More info CMS in the Silicon talk Strip of Loic Tracker Quertenmont Operation and Performance

19 Reconstruction of known resonances Few examples Clear evidence of mass peaks mass and resolution in good agreement with MC K s : σ=8.0 MeV (MC 7.6 MeV) Λ: σ=3.0 MeV (MC 3.0 MeV) D 0 Kπ

20 Conclusions The Silicon Strip Tracker has being operated successfully within the CMS experiment at LHC since the first pp collisions events Tracker performance are excellent and stable High S/N Robust track reconstruction Low number of bad channels Expected hit resolution Efficient particle identification Tracker has optimal performance for the analysis of 7 TeV collision data

21 BACK-UP SLIDE

22 CMS experiment Large SC solenoid: 3.8T, L = 13m, d = 6m Muon spectrometer: DT, CSC, RPC Hermetic hadron calorimeter: η < 5 PbW0 4 EM calorimeter: η < 3 All-Silicon Tracker: η < 2.5

SST: Modules Modules are based on single-sided p+ strip on n-bulk devices 15148 detector modules 15 different sensor geometries and two different

23 SST: Modules Modules are based on single-sided p+ strip on n-bulk devices detector modules 15 different sensor geometries and two different thickness (320 and 500 um) Double-Sided modules are made of two module of the same type glued back to back (one of them been tilted, stereo module)

24 Silicon Strip Tracker Tracking efficiency: ε >99% (µ), ~90% hadrons Resolution: pt/pt ~ 1-2% (η<1.6)

APV readout mode Deconvolution is the nominal operation mode for collisions - fast signal shaping, reduce pile-up of events - Take 3 samplings of PEAK amplitude s i with

25 APV readout mode Deconvolution is the nominal operation mode for collisions - fast signal shaping, reduce pile-up of events - Take 3 samplings of PEAK amplitude s i with t = 25 ns - DECO amplitude = Σ s i w i (w i = weights) - Produce short pulse from three consecutive samples - τ D < τ P shorter time window for charge integration in DECO

SST: Commissioning The commissioning procedure is required to configure, synchronize and calibrate the various components of the readout system.

26 SST: Commissioning The commissioning procedure is required to configure, synchronize and calibrate the various components of the readout system. It consists of several independent steps performed on four partitions (TIB/TID, TOB, TEC-, TEC+) Check of connection Electrical cabling (1) Optical cabling (2) Internal timing Synchronization of all channels to include different fiber length (3) Chip parameter optimization Optical Gain (4) Analog baseline tune (5) Pulse shape tuning (6) Pedestal run Pedestal and noise value for DB upload (7) APV latency scan Synchronize tracker with LHC clock (8) Fine tuning of pulse shape sampling Tune to 1ns level (9) After this procedure, detector is ready for data taking ,6,8,9 4 7

Lorentz Angle The Lorentz angle is measured for each individual module - Lorentz drift θ L - Track angle θ t h is the detector thickness, θ L, p1 and p2 are the fit parameters p1 is the

27 Lorentz Angle The Lorentz angle is measured for each individual module - Lorentz drift θ L - Track angle θ t h is the detector thickness, θ L, p1 and p2 are the fit parameters p1 is the slope of the line divided by the ratio of thickness to pitch p2 is the average cluster size at the minimum TOB layer 4 profile plot of cluster size versus the tangent of the incidence angle

Charge collection deficit Charge collection is different in peak and deco readout mode - shorter integration time for DECO wrt PEAK Bias in reconstructed cluster position found investigating a

28 Charge collection deficit Charge collection is different in peak and deco readout mode - shorter integration time for DECO wrt PEAK Bias in reconstructed cluster position found investigating a discrepancy between PEAK and DECO alignment geometries from cosmic data Explanation: Charge from sensor backplane does not reach strips in time for readout effect larger in DECO and for 500 µm thick detectors Sensor width Pulse height [ADC count] Charge drift Pulse shape Cluster Peak --- Deco Time [ns] True rechit Reco rechit Track

Backplane and LA correction Effect on detector performance Fraction of cluster charge is not collected Reconstructed hit position extrapolated from cluster

local reconstruction Corrections for deconvolution mode - w (due to the lost charge) and a Lorentz drift <tan θ(d)> correction - corrections measured using

29 Backplane and LA correction Effect on detector performance Fraction of cluster charge is not collected Reconstructed hit position extrapolated from cluster barycenter is biased Effect on the Lorentz Angle A consistent alignment geometry for both peak and deco data is needed - corrections should be included in local reconstruction Corrections for deconvolution mode - w (due to the lost charge) and a Lorentz drift <tan θ(d)> correction - corrections measured using collision data - uploaded into the database to be used during local reconstruction - results validated by the alignment group and used in the reprocessing

Alignment 15148+1440 sensors 6 degree of freedom each O(10µm) accuracy Minimization hit/track residuals χ 2 track parameters

local minimization of sensor position, iterative, detailed track model Applied sequentially from large substructures to

30 Alignment sensors 6 degree of freedom each O(10µm) accuracy Minimization hit/track residuals χ 2 track parameters and sensor positions Two approaches: Millipede (II): Global minimization, simple track model Hits and Impact Points (HIP): local minimization of sensor position, iterative, detailed track model Applied sequentially from large substructures to sensor level 2010 cosmics and collision events used for present alignment: 1.5M cosmic tracks (p>4 GeV) 1.7M collision tracks (p>3 GeV)

Hit resolution: method Goal: Measure the silicon sensor resolution using data from collision tracks Use tracks which pass through two sensors of the same layer Minimizes effects of misalignment and

31 Hit resolution: method Goal: Measure the silicon sensor resolution using data from collision tracks Use tracks which pass through two sensors of the same layer Minimizes effects of misalignment and reduces the potential amount of detector material between hits Combine all tracker information without the hits from the layer under study to obtain the best track prediction Compare the difference in hit and predicted positions of the two overlaps and calculate the resolution for each pair of overlapping sensors separately

2010 results: charmed mesons Clear evidence of mass peaks 7 TeV

good agreement with MC pt(d * )>5 GeV pt(k,π)>0.

0 GeV 3<L xy /σ<20 Secondary Vertex: 3 tracks, pointing to PV

32 2010 results: charmed mesons Clear evidence of mass peaks 7 TeV collisions O(10 6 ) minimum bias events mass and resolution in good agreement with MC pt(d * )>5 GeV pt(k,π)>0.600 GeV pt(π s )>0.250 GeV pt(d 0 )>3 GeV pt(k)>1.25 GeV pt(π)>1.0 GeV 3<L xy /σ<20 Secondary Vertex: 3 tracks, pointing to PV L/σ> 7 p>1.5 GeV, pt>0.1 GeV D 0 Kπ D + Kππ M(Kπ)-PDG <25 MeV M(Kππ)-M(Kπ)-PDG <1.2 MeV

Tracking and Alignment in the CMS detector

Tracking and Alignment in the CMS detector Frédéric Ronga (CERN PH-CMG) for the CMS collaboration 10th Topical Seminar on Innovative Particle and Radiation Detectors Siena, October 1 5 2006 Contents 1