Tracking POG Update Tracking POG Meeting March 17, 2009 Outline Recent accomplishments in Tracking POG - Reconstruction improvements for collisions - Analysis of CRAFT Data Upcoming Tasks Announcements Kevin Burkett (FNAL), Boris Mangano (UCSD)
Recent Tracking Improvements Track reconstruction has already been quite robust, based on MC studies, for tracks above ~1 GeV Recent focus has been on improving reconstruction for lower PT tracks and tracks displaced from the primary vertex Low PT tracks are important for early physics analysis Improves performance of particle flow Improved reconstruction of displaced tracks will impact reconstruction of conversions, V0s, and also some non-sm physics signatures Need to maintain fake rate, timing performance Already saw significant improvement from addition of 4th (pixelless seeded) and 5th (TOB-TEC seeded) steps Recent improvements from improved tuning of 4th and 5th steps 2
Improved Efficiency at Low PT Efficiency and fake rate vs. eta for particles produced with r<20 cm, z <60 cm Parameters in iterative tracking reoptimized during 310 cycle Improved efficiency and fake rate ttbar (ideal) before tuning after tuning (Must be balanced with reco time and event size) QCD 3-3.5TeV (ideal) K. Stenson 3
Effect of Retuning Iterative Tracking - PF Jets Graphs compare Gaussian fit resolution and histogram RMS End-cap PF Jet Resolution Before Retuning After Retuning Note improvement in tails (RMS Resolution) K. Stenson 4
Improved Reconstruction of K 0 s Improved reconstruction of displaced tracks by the TOB-TEC seeded step can be seen in the improved efficiency for reconstructing K 0 s in B 0 K 0 sμμ K 0 s efficiency vs. Radius in B 0 K 0 sμμ CMSSW 300 4-steps CMSSW 310 5-steps V0 producer was added to the standard reconstruction sequence in CMSSW 310pre2 K. Stenson, K. Ulmer, B. Drell, I. Tomalin, M. Konecki 5
Improved Efficiency at Low PT Improvements in seeding and optimization of iterative tracking allow lowering the PT threshold in pattern recognition to 300 MeV and below Significant improvement in efficiency at low PT Lowering thresholds more will need to be balanced with the timing budget for reconstruction Efficiency vs. PT in ttbar old new I. Tomalin, C. Saout, M. Konecki 6
CRAFT Analysis CRAFT Data is the first chance to study tracking with the magnetic field on, and to study the connection between tracks and the rest of the detector Commissioning of track reconstruction Cosmic reconstruction uses most of the same tools for pattern recongnition as default tracking Collision reconstruction can be tested on cosmics that pass close to the center of the detector Study of tracking resolution Measure resolution with real data using split tracks Understand tracker B-field relative to muon system 7
Measuring Tracking Efficiency Two methods used to measure tracking efficiency in cosmic data I. Standalone muon vs. tracker Given a standalone muon that points into the tracker, do we find a track? II. Tracker only (#1) Given a track found in one half half of the tracker, can we find a matching track in the other half of the tracker? III. Tracker only (#2) - NEW TOP y x BOTTOM Can we find matching tracks on two halves of the tracker, using an LHC-like inside-out navigation? 8
Tracking Efficiency Using Muons Use 1-Leg standalone muon collection Select very central muons by requiring STA muon to have: d0 <30 cm, dz <30 cm, η <1, 0.5< φ <2.5 Yields efficiency close to 100%, w/good agreement between data and MC Efficiency(%) CTF Cosmic RS CRAFT 99.5 99.3 99.1 MC 99.9 99.7 99.6 Data-MC -0.4-0.4-0.5 Efficiency Efficiency (Uncertainty ~ ±0.1) J. Andrea 9
Tracking Efficiency from Tracker Only Use same selection on probe as in muon-based efficiency d0 < 30 cm, dz < 30 cm η < 1, 0.5 < φ <2.5 ε(b T) (%) Data MC CTF 97.0±0.1 97.9±0.1 CosmicTF 93.9±0.1 93.3±0.1 RS 90.0±0.1 90.7±0.1 ε(t B) (%) Data MC CTF 95.6±0.1 97.6±0.1 CosmicTF 92.6±0.1 96.7±0.1 RS 90.5±0.1 91.4±0.1 Bottom efficiency agrees well with MC Top efficiency slightly lower Traced to dead region in TOB layer 4, and effect has now been verified in MC G. Cerati 10
Tracking Efficiency with Inside-Out Reco Test efficiency of collision-like reconstruction Inside-out reconstruction (CTF) with two halves reconstructed independently Track reconstructed on one side is tag, search for reconstructed track on opposite side Cuts on tag track: PT>10 GeV d0 <10 cm, dz <30 cm 10 #hits 25 11
Tracking Efficiency with Inside-Out Reco Results for pixel-less seeding CRAFT Data used is from period after latency fix Overall efficiency very high Differences between V4 and V9 reprocessing not fully understood Pixel-seeded tracking results still under study MC (Startup) MC (Ideal) CRAFT_V9 CRAFT_V4 Seeding Tracking Overall 12
Tracking Resolution Directly measure resolution on track parameters using CRAFT data N.Tran 13
Tracking Resolution Directly measure resolution on track parameters using CRAFT data Δdxy Δdz Δφ Δθ ΔpT Δ(1/pT) N. Tran 14
Multi-Track Events in CRAFT Data Recent results suggest that many multi-track events in the CRAFT data are due to overlapping reconstructed tracks Examine events with exactly two tracks Simple cleanup cuts: χ 2 /dof < 10 PT>1 GeV, #Hits >6 Observe correlations: φ2-φ1 θ1=θ2, φ1=φ2, dxy1=dxy2 θ1=π-θ2, Δφ=π, dxy1=-dxy2 θ2 θ1 Originally thought to be split tracks, but recent result points toward hit sharing -- still under study A. Graziano 15
CRAFT Tracking Analysis Summary Analyses of tracking performance using the CRAFT data have converged very nicely Efficiency measurements are in agreement, checked with Monte Carlo, and show good performance Resolution and efficiency of tracker and tracking looks good Results will be published in Strip Performance CRAFT note https://twiki.cern.ch/twiki/bin/view/cms/tkcraftnotestrip See Gaelle s summary from CRAFT http://indico.cern.ch/getfile.py/access? resid=0&materialid=slides&contribid=4&sessionid=0&subcontid=2&confid=50961 Details on the results shown today, plus others such as Tracker/Muon PT comparison (J. Andrea) and hit resolution (K. Ulmer) We will need to reproduce the results with CMSSW 310 in preparation for CRAFT09 16
Early Physics Analyses with Tracker Tracking will be important for many planned analyses with the data from the first collisions Charged Hadron Spectra Underlying Event Analysis Min Bias dn/deta analysis using Pixel Hits Charged Hadron Spectra from Tracklets (*TODAY*) Several analyses reviewed at 16 Feb Joint DPG-PH Meeting http://indico.cern.ch/conferencedisplay.py?confid=52680 Summary to be given by L. Fano later today in Tracker General Meeting 17
Tracking POG Priorities in Coming Months Completion of CRAFT Analyses Most results are nearly final and will go into CMS note A few remaining issues to follow-up Preparation for CRAFT 09 Need to help re-commission tracker Revalidate reconstruction in CMSSW 31X well in advance Ideally repeat CRAFT analyses with new release Are there any remaining improvements that could go in? Preparation for Collisions Need tools to measure tracking performance in data Efficiency, momentum scale, track parameter resolution, material budget (together with e/gamma POG) Validate beamline reconstruction in CMSSW 31X 18
Overview of Tracking POG Tasks Current tasks in Tracking POG can be divided into several general categories: Core tracking software tasks Maintenance and validation of software Documentation Measurement of tracking performance in data Preparation of early physics analyses Specific open projects (often related to other POGs) Contact Boris and Kevin if you are interested in any of the tasks listed on the next few slides 19
Tracking POG Tasks (I) Core tracking tasks: Validation of track reconstruction software Test of track reconstruction software on collision MC and cosmic data Verify performance in latest releases and in CRAFT09 data Optimization of reconstruction efficiency/fake rate for collisions Ongoing work in fairly good shape now Need feedback from other POGs Optimization of collision reconstruction timing Some timing studies performed already Performance probably OK for startup Create configurations where later iterative tracking steps are omitted? Release Validation Currently in good shape Add separate validation of each tracking step? Some fine-tuning of trackable hits, improvements for use with pileup 20
Tracking POG Tasks (II) Tracking performance in data Preparation of tools to measure tracking performance in data Some work has been done, but still much left Efficiency measurements for higher PT tracks overlap with muon group Need more work on lower PT tracking efficiency Tools should be tested with MC by the end of the summer Measurement of momentum scale Documentation CMS note on tracking performance Will be written in the coming months to document performance with final release prior to data-taking Documentation on twiki of tracking software See later announcement All developers should contribute 21
Tracking POG Tasks (III) Preparation for early physics analyses with tracker Many studies done in physics groups Significant changes in simulation,track reconstruction since then Repeat analyses with CMSSW 310 when available Specific open projects Improvements for tracking inside jets (includes DAF/MTF) First carefully reevaluate performance Eventually may need to implement splitting of clusters Improved PropagatorWithMaterial Correct propagation to vertex Correct implementation of passive material improvements in tracking for electrons address charge misidentification include vertex in fit 22
Documentation Online tracking documentation is in need of an update Review of documentation announced at March 2 meeting Reminder: Giuseppe reviewed much of the documentation with Kati Overview pages in pretty good shape, but descriptions of different algorithms, etc. need work It is the responsibility of the developers to help maintain the documentation We will soon announce a more detailed plan with a deadline for developers to update their documentation Please cooperate and update the pieces for which you are responsible!!! 23
Reminder of Some Technical Details Joint DPG-POG Meetings are every other Monday Working meetings usually on Friday (same week as joint DPG-POG Meeting) 1600-1800 CERN time Group twiki: https://twiki.cern.ch/twiki/bin/view/cms/trackingpog New hypernews: hn-cms-tracking 24
Today s Meeting 25