CMS Tracking Performance Results from early LHC Running

CMS Tracking Performance Results from early LHC Running CMS PAPER TRK-10-001 L. Spiegel, K. Stenson, M. Swartz 1

First Collisions Tracking Paper Provide description of tracker, tracking algorithm, and performance for summer conferences Based on already approved PAS TRK-10-001 Add description of tracker (Sec 2) Modify description of triggers and data samples (Sec 3) Add description of pixel operating conditions, calibration, and performance (Sec 4.1) Add description of strip operating conditions, calibration, and performance (Sec 4.2) Expand PAS description of tracking algorithm (Sec 5) Condense description of tracking performance from PAS (Sec 6) Initiated/defined/reviewed by Tracker Editorial Board, Tracker DPG, Tracking/B-tag POG Conveners F. Palla, G. Hall, S. Nahn, D. Kotlinski, V. Chiochia, M. Manelli, J. Cole, G.M. Bilei, D. Contardo, C. Delaere, T. Speer, K. Burkett, A. Venturi, W. Adam, A. Rizzi Three editors L. Spiegel (new strip sections) K. Stenson (expanded tracking and condensed tracking performance) M. Swartz (new pixel sections) Same ARC as PAS: J. Alexander, V. Krutelyov, M. Pioppi Reports on the work of many collaborators from Tracker, Tracking/B-tag POGs 2

Preface Paper was written in February of 2010 All results based on December 2009 running at 900 GeV and 2360 GeV Results are non-uniform (understanding of pixel and strip systems were more/less advanced in many aspects) - subsystem results are not parallel Pixels and strips are different technologies with different readout schemes relevant parameters are different (eg. S/N meaningful for strips, not so much for pixels) non-parallel results Results that rely on detailed tracker simulation are not included eg. no tracking efficiency results, material studies, etc more and detailed papers in the near future Improved results based on 2010 data will be published in the coming year? 3

Pixel Detector Standard description of pixel system in Section 2 Subsection 4.1.1 documents December operating conditions Bias voltages and temperature fraction of live channels = 98.4% Retuned readout thresholds lower than previously published CRAFT2008 effective in-time thresholds somewhat larger than test-pulse thresholds average layer efficiencies > 99% new analysis agrees with result reported in dn/dh paper Subsection 4.1.2 documents the initial overall timing scan More recent fine scan results tested but still not in use Cluster size [pixel] 5 4 3 2 CMS 1 Endcap Pixel Barrel Pixel -5 0 5 10 15 Clock phase 3 [ns]

Pixel Section - Cluster Multiplicity Arbitrary scale -1 10 0.9 TeV CMS data MC -2 10-3 10-4 10 1 1.5 2 2.5 3 3. 1 1.5 2 2.5 3 3. log (Number of Clusters) log (Number of Clusters) 10 10 Compare measured cluster distribution at 900 GeV with Pythia/Geant ( Atlas tune) all triggers on left veto on BSC beam-gas trigger bits on right rough agreement is observed similar agreement with layer occupancies pixels working as expected in zeroth order discrepancies likely due to underlying event model 5

Pixel Section - Normalized Cluster Charge CMS 2009 s=900 GeV CMS 2009 s=900 GeV Number of clusters / 1[ke] 14000 12000 10000 8000 6000 10 4 3 10 10 2 MC DATA Barrel Number of clusters / 1[ke] 9000 8000 7000 6000 5000 4000 10 4 3 10 10 2 MC DATA Endcap 4000 10 1 0 10 20 30 40 50 60 70 80 90 100 3000 2000 10 1 0 10 20 30 40 50 60 70 80 90 100 2000 1000 0 10 20 30 40 50 60 70 80 90 100 normalized cluster charge [ke] 0 10 20 30 40 50 60 70 80 90 100 normalized cluster charge [ke] Scale cluster charge by track length to sensor thickness shows first results, work still in progress measured barrel width is larger than simulation charge smearing from incomplete gain corrections/imperfect calibration? ~5% scale shift in endcaps gain or threshold issues? electronic saturation not modeled correctly 6

Pixel Section Lorentz Angles Published CRAFT 2008 results are obsolete due to voltage and temperature changes. Note that the detailed Pixelav simulation is actually what is calibrated Correct Generic Reco/Simulation for implant focusing effects Generate Template profiles CRAFT2009 is only calibration for endcaps Deploy new grazing angle technique for barrel uses 2D segmentation of pixels Improved low PT tracking enables cluster size technique for barrel drift ["m] 100 80 60 40 20 0 Offset -2.359! 0.030 tan( L ) 0.399! 0.000 900 GeV data, B = 3.8T linear fit Grazing Angle Technique 0 50 100 150 200 250 300 depth ["m] Sample Detector Technique Meas tanql Pixelav tanql CRAFT 2009 Barrel Cluster size 0.409±0.002 0.407±0.002 CRAFT 2009 Endcap Cluster size 0.081±0.005 0.080±0.004 Min Bias 2009 Barrel Grazing Angle 0.399±0.001 0.401±0.001 MIn Bias 2009 Barrel Cluster size 0.409±0.002 0.411±0.005 7

Pixel Section Resolution Update CRAFT 2008 barrel module overlap analysis using Min Bias 2009 data. Coordinate Clusters Meas Resolution Pixelav Resolution local x (f) all 12.8±0.9mm local x (f) no double pixels 12.7±2.3mm 14.1±0.5mm local y (z) all 32.4±1.4mm local y (z) no double pixels 28.2±1.9mm 24.1±0.5mm Remove small charge clusters and edge clusters from the analysis Consider the effect of double-size pixels on the result double size pixels span Readout Chip boundaries including them worsens resolution wrt typical tracks (twice as probable in overlaps) Pixelav results don t include them Overlap x-resolutions are poorer than typical tracks populate the edges of the a-angle acceptance where resolution is poorer expect x-resolution < 10mm from Pixelav for most hits 8

Silicon Strip Detector Standard description of strip tracker in Section 2 Subsection 4.2 documents operating conditions Bias voltage, temperature of sensors and hybrids Distinction between peak and deconvolution operating modes of the APV25 A transition was made to deconvolution toward the end of the Early Collision period. Timing much more critical in deconvolution (hence the timing scan) Total integrated strip efficiency given at 97.8% Consistent with 2.2% bad modules and <0.2%(!) construction bad channel rate No detailed layer information. Documented in CRAFT Histogram and result collections https://twiki.cern.ch/twiki/bin/view/cms/trkapprovedresults900gev Approved pixel, strip, and tracking results (includes 2360 GeV data as well) https://twiki.cern.ch/twiki/bin/view/cms/trk10001 Staging area for plots to be approved for TRK-10-001 No Lorentz angle results as they were not sufficiently mature at the time when the paper was written. 9

Strip Section S/N Plots shown for 2.36 TeV deconvolution mode illustrating the Landau nature of the signal and the difference between thin and thick sensors. The deco mode admits more noise than the peak mode, but there is no reason to expect changes in the S/N with changes in the center-of-mass energy. 10

Strip Section Timing Scan Most of the running was done in Peak mode Timing taken from CRAFT Reference to CMS note 2008/007 A fine scan was made close to the end of the collision period run 123977, TOB layer 3 delays were implemented just before transition to Deconvolution mode (run 124275) No Peak data with fine scan delays 10 ns timing adjustment S/N reduced by 2.5% in Peak mode 11

Strip Section de/dx Strip tracker benefits from the large dynamic range in the APV25 analog based readout Up to 3 normal incidence MIPs in Early Collision running but will eventually be expanded to 7-8 MIPs Pixels are including in tracking but not in the de/dx determination Upper plot shows bands for kaons and protons and to a lesser extent deuterons. Fit is based on the proton band. Deuterons suffer from ADC saturation which is enhanced by the relatively larger η values. Pythia does not produce deuterons at all and there is some evidence that GEANT underproduces deuterons 12

Tracking results Section 5 gives a brief description of track reconstruction Section 6 gives tracking results. All of these results come from December 2009 data and were approved for the TRK-10-001 PAS. Data/MC comparison of track variables (#tracks, hits/track, c 2, p T, h, f, d 0, d z ) Primary vertex resolution K S, L 0, X ±, K * (892) ± masses and resolution K S and L 0 lifetime de/dx tested on L 0 and used on f -> K + K Conversion mass, p T, and probability Nuclear interaction vertex position vs r and z b-tagging variables 3D IP significance and 3D vertex separation significance Plus the 4-track vertex event display (approved elsewhere) 13

Tracking distributions 14

Primary vertex resolution Split one primary vertex into two groups of tracks which must originate from the same point. Fit the two independent vertices and extract resolution. 15

V 0 mass distribution Data mass (MeV) MC mass (MeV) PDG mass (MeV) K S 497.68 ± 0.06 498.11 ± 0.01 497.61 ± 0.02 L 0 1115.97 ± 0.06 1115.93 ± 0.02 1115.683 ± 0.006 Data s (MeV) MC s (MeV) K S 7.99 ± 0.14 7.62 ± 0.03 L 0 3.01 ± 0.08 2.99 ± 0.04 16

V 0 lifetimes Data and MC are split into bins of ct and a fit for the yield is performed in each bin. Divide MC yields by true (exponential) distribution to obtain correction factor. Correct data and fit for lifetime. For K S find 89.8 ± 2.1 ps compared to PDG of 89.53 ± 0.05 ps For L 0 find 271 ± 20 ps compared to PDG of 263.1 ± 2.0 ps. uncorrected data corrected data 17

X ± and K*(892) ± Take L 0 and try to vertex with a detached p to find a X. Take K S and combine with a primary p ± to find a K * (892) ±. 18

de/dx Left (right) plot is de/dx of low (high) momentum track in L 0 decay Hard track is always the proton follows proton curve f peak seen with de/dx cut on kaons; no f seen when fail de/dx cut 19

Conversions and interactions Photon conversion invariant mass, p T, and probability Radius and z positions of interaction vertices 20

b-tagging variables and 4-track detached vertex 21

Next Steps ARC Report (J. Alexander) and approval? Collaboration-Wide Review http://cms.cern.ch/icms/analysisadmin/get?analysis=trk-10-001-paper-v2.pdf Please send comments to TRK-10-001 hypernews Thanks to F. Palla/Tracker Editorial Board, Tracker DPG, Tracking/B-Tag POGs, ARC Big thanks to the many unnamed contributors to the paper! 22