CMS Tracker Performance Francesco Palmonari (INFN Pisa) on behalf - - PowerPoint PPT Presentation

6 december 2011 francesco.palmonari@cern.ch 1

CMS Tracker Performance

Francesco Palmonari (INFN Pisa)

• n behalf of the CMS Collaboration

8th International "Hiroshima" Symposium on the Development and Application of Semiconductor Tracking Detectors

6 december 2011 francesco.palmonari@cern.ch 2

Outline

The CMS silicon tracker Tracker Operations in 2011 Strips and Pixels status Tracker Performances Conclusions Please note: Laser Alignment → A.Perieanu Radiations damages effects → C.Barth

6 december 2011 francesco.palmonari@cern.ch 3

The CMS detector at the LHC

LHC Point5: 100 m underground in the comune de Cessy (pays de Gex - FR)

← CERN site

5.7 fb-1 delivered and 5.2 fb-1 recorded by CMS during pp collisions in 2011 Almost 90 μb-1 delivered to CMS during ions collisions as of 28.11.2011 3.8 Tesla magnetic field

6 december 2011 francesco.palmonari@cern.ch 4

The CMS silicon tracker

Silicon Strips Pixels

9.6 M channels 66 M channels CMS Tracker 198 m2 silicon area + 1.1 m2 silicon area = σ(pt)/pt ~ 1-2% (pt~100 GeV) 10 layers (TIB TOB) 3 layers IP resolution: ~10-20μm 12 disk (TID TEC) 2 disks (pt = 100-10 GeV)

~ 4 stereo hits

pitches: 80-180 μm thickness: 320-500 μm NB: 2 sensors-modules are

• nly in TOB and TEC

TIB TID TOB TEC Bpix

r = 4,7,11 cm

Fpix

z = 35,47 cm

6 december 2011 francesco.palmonari@cern.ch 5

Tracker operations in 2011

Standard operations:

• Periodic calibration: pedestals, noise, optical receiver offsets
• Periodic bias scan to monitor the radiation damage:
• leakage current increasing everywhere as expected

pixels: strips: measured Vdep decrease as expected - iLeak measured on board of modules (dcu) and from caen power system

• no Vdepchange for >5fb-1 (expecting < -10 V)

CMS Preliminary

6 december 2011 francesco.palmonari@cern.ch 6

Tracker operations in 2011

CMS data taking efficiency: ~91% in pp ~96% in ions** ** = as of the 04.12.2011 Luminosity lost by category (pp) ~9% in 1935 runs

28% because of tracker DAQ: 26% (18% strips and 8% pixels) POWER: 2% (1% strips; 1% pixels) Please note: (strips+pixel) constitute the 70.6% of the CMS DAQ (480/680 FEDs)

Uptime: > 98% with stable conditions

• Strips cooling: stable coolant 4 Celsius; leaks controlled to <0.7 kg/day
• Pixels cooling: stable coolant 7 Celsius
• Power/Detector Control System (DCS): stable had <1% PSU replaced
• Electronics: stable.
• Data Quality Monitoring (DQM): stable. prompt feedback provided to Operations Coordinators

Full reconstruction chain monitored online and offline with: raw data – Digis – on track/off track clusters - track reconstruction histograms

6 december 2011 francesco.palmonari@cern.ch 7

Strips detector status

In this map: permanent defects not used not used good good TIB: 94.3% TID+-: 98.1% TOB: 98.2% TEC+:98.8% TEC-:99.1% 5 loops with passive cooling only.

NB: possibility to recover 0.5% of the channels during the shutdown LS1

total alive channels fraction: 97.79%

6 december 2011 francesco.palmonari@cern.ch 8

Pixel detector status

Active channels fraction: 96.9% (Barrel 98.4 % ; Disks 92.8 %) percentage of dead channels: 3.1% location of the dead pixel in the occupancy maps:

NB: ~3% of Fpix missing due to the loss of an Analog Opto Hybrid (AOH) in 2010

6 december 2011 francesco.palmonari@cern.ch 9

Cluster reconstruction (strips)

S/N distributions from on-track clusters after correcting for the path length. Distributions below use the signal taken in deconvolution in the inner (thin) and

• uter (thick) barrel sensors.
• S/N = (cluster signal) / (noise of strips)
• Signal maximized after fine delay adjustment (next slide)

320 μm sensors 500 μm sensors Two possible signal readout modes:

• deconvolution: standard mode for physics

weighted sum of 3 signal samples

• peak: full signal sample; in this mode the

noise is ~0.3 lower but time resolution is worst

6 december 2011 francesco.palmonari@cern.ch 10

Timing optimization

Random Time Delay Scan: measure the signal position w.r.t the nominal sampling point deviations (from 0) are within 1 ns signal profile has the expected 12 ns width Pixels: delay scanned maximizing the

• n track cluster size;

choose phase delay in order to have enough room within the efficiency plateau

6 december 2011 francesco.palmonari@cern.ch 11

Hit resolution

Hit resolution depends on sensor thickness and strip pitch: the minimum value is reached for an angle corresponding to optimal charge sharing Obtained hit resolutions: Strips: 15 μm to 45 μm pitch degrees

Layers μm 0-10 TIB 12 80 16.0±3.5 TIB 34 120 27.9±2.9 TOB 1234 183 41.3±3.8 TOB 56 122 24.5±2.7

Pixels: 9 μm to 35 μm with 2 independent method:

• Overlaps (shown in the plot)
• Hit triplets

6 december 2011 francesco.palmonari@cern.ch 12

Tracks reconstruction

Using Kalman filter technique for high track density environment: Seeding→Pattern recognition→Track fitting

• Seeding: pixel hit triplets or pixel/strip hit

pairs with constraint from the beam spot

• Iterative tracking (7 iterations)
• at each iteration remove track-assigned hits

and relax seed cuts

Track parameters agreement with MC samples (Pythia 8 and GEANT4)

Event selection: good collision events defined by trigger and vertex selections criteria Track selection: σ(pt)/pt < 5% and dz <10σ

6 december 2011 francesco.palmonari@cern.ch 13

Alignment

• Increased local precision w.r.t. 2010
• Control of the weak modes with mass

constrain Z → μ+ μ−

• accounting for sensor bows and kinks

(from 56k parameters to 200k)

• correcting for time dependent large

(< 30 μm) pixel volumes movements as part of the PV validation

6 december 2011 francesco.palmonari@cern.ch 14

PV resolution and efficiency

PV resolution: depends on the number PV efficiency depends on the

• f tracks used and their pt num. of tracks of the PV cluster

Data driven “split” method Fakes tracks excluded (by ptcut)

provide PV-resolution[number of tracks] split method used within cluster

6 december 2011 francesco.palmonari@cern.ch 15

Tracking performance (pp collisions)

Data driven technique to measure the tracking efficiency (μ shown as example): embed simulated track in Min. Bias Data

→ test if track is still reconstructed

Event average pile-up (PU) increased going from ~5 to ~10 in sept.2011

But tracking and vertexing showed no performance degradation

6 december 2011 francesco.palmonari@cern.ch 16

Particles identification

Modules Analog readout allow particles identification using the energy loss information (at least 10 hits/deposits along tracks for dE/dx): 1) a synchronization pulse is used as measure for the electronic gain 2) all signals are normalized to a default value of that pulse 3) the MIP are used to equalize the sensors response by applying particle gain calibration factors 4) these factors are applied so that all sensors and readout chains have equal behavior and the measured charge can be used for particles ID exploiting dE/dx dE/dx validated with Λº→pπ decays

Where lower momentum particle is always the π

6 december 2011 francesco.palmonari@cern.ch 17

Conclusions

The CMS tracker is part of a marvelous detector:

• The CMS tracker is a stable and reliable system providing expected performance for

S/N, hit resolution and track reconstruction. It is also used for particle identification.

• Such performance are mandatory to produce outstanding physics analysis results

The CMS tracker ready to sustain the 2012 data taking where a further increase of the luminosity to integrate is foreseen

6 december 2011 francesco.palmonari@cern.ch 18

Thanks !

References

• 1. CMS Luminosity Collision Data, https://twiki.cern.ch/twiki/bin/view/CMSPublic/LumiPublicResults.
• 2. The CMS Collaboration, The CMS Experiment at the CERN LHC, JINST 3S08004 (2008).
• 3. CMS Tracker Detector Performance Results, https://twiki.cern.ch/twiki/bin/view/CMSPublic/DPGResultsTRK.
• 4. The CMS Collaboration, Tracking and Primary Vertex Results in First 7 TeV Collisions, CMS Physics

Analysis Summary, CMS PAS TRK-10-005 (2010).

Special thanks to:

Laura, Erik, Andrea, Adrian, Derek, Matthew, Gordon, Petra, Victor, Daniel, Karl, Frank

6 december 2011 francesco.palmonari@cern.ch 19

Backup... detector ;)

6 december 2011 francesco.palmonari@cern.ch 20

backup

!! CMS 2011 RECORDS !!

6 december 2011 francesco.palmonari@cern.ch 21

backup

S/N ratio last year:

6 december 2011 francesco.palmonari@cern.ch 22

backup

Track reconstruction: seeding → pattern recognition → track fitting explained Hit resolution: the overlap method Obtained via the comparison between: measured and predicted hit position from track fitting in the overlap regions (within the same tracker sub structure in

• rder to minimize effects of track extrapolation

and amount of material transversed)

6 december 2011 francesco.palmonari@cern.ch 23

backup

Hit resolution: the pixel triplets method Select tracks with hits in 3 pixel layers.

• Redefine track:
• curvature from full tracker



• position and anglesfrom hits 1 and 3
• analytic code from J. Gassner 1996 (ETH Zürich, H1)
• Interpolate to middle layer:
• residual between track and hit

6 december 2011 francesco.palmonari@cern.ch 24

backup

Hit resolution in the pixels: the riplet method results

6 december 2011 francesco.palmonari@cern.ch 25

backup

PV reconstruction: we can count 20 of them in this event