Tracking and centrality in HI Sasha Milov (for the HI working - PowerPoint PPT Presentation

Tracking and centrality in HI Sasha Milov (for the HI working group) Heavy Ion readiness Walkthrough Sasha Milov HI walkthrough Sept 13, 2010 1

Outline • Centrality – Definition – Implementation – Errors • Tracking and vertexing: – Vertex method – Tracking optimization – Tracking performance with new parameters • Tracklets: – In p+p – In Pb+Pb – Readiness and requirements on Day-1 Sasha Milov HI walkthrough Sept 13, 2010 2

Centrality Sasha Milov HI walkthrough Sept 13, 2010 3

Centrality Glauber HIJING Peripheral collision Central collision • Impact parameter drastically changes the nature of the HI collision: – Smaller b imp  more mass (N wn )  more violent (e.g.: N ch , E T ) is the event. • Definition: – All b imp are ordered from 0% (central) to 100% (most peripheral). Centrality is the percentile this event belongs to. • b imp cannot be directly measured. – Assuming that many global observable (e.g.: N pixels , N tracks , ∑HCAL, etc…) monotonously (not even linearly!) depends on b imp the centrality is the same up to fluctuations. • Other centrality parameters: – N part , N coll , Are of overlap, eccentricity… Sasha Milov HI walkthrough Sept 13, 2010 4

Centrality Implementation • HICentrality package provides centrality in 1% bins • Based on observables: • Total energy on the EM level as calculated from cells • Total number of silicon pixel clusters • Hijing truth • Calibration: • private, latest 15.6.9.3. file HICentralityCalibrations.root is in repositiory and It needs to be copied to run area by standard get_files python utility. • Validation: • HIValidation package, runs in Athena rel. >= 15.9.0 (dev) Sasha Milov HI walkthrough Sept 13, 2010 5

MB sample efficiency About 6% of the Pb+Pb events at 2.75TeV are n+n collisions. For those events the trigger efficiency is the same as in p+p Using ATLAS (900 GeV) paper trigger efficiency is very close to 100% and vertex reconstruction efficiency is 67% for the lowest n BS sel bin. Putting an estimated numbers together the MB sample loss is 0.7*0.15(n ch ≈ n BS sel ) *0.06~1% Sasha Milov HI walkthrough Sept 13, 2010 6

Centrality uncertainties counts Accuracy of the N part determination comes from 3 main factors: • Model uncertainties • Detector response fluctuations • Trigger efficiency In PHENIX (see left) the latter was the dominant factor: efficiency = 92 ± 3% we expect ~99% Second important was the detector Real PEHNIX data response fluctuations. ATLAS has wider coverage and increased particle production Model uncertainties contribute 1-3% to N part . Sasha Milov HI walkthrough Sept 13, 2010 7

Vertex and Tracking Sasha Milov HI walkthrough Sept 13, 2010 8

Vertex Default vertex reconstruction method (chi2CutMethod=2) was taking too long time in HI events. In addition due to low luminosity we practically have no pile-up. Vertex efficiency ~99% (64 events out of 5k. MinBias Vertex finding accuracy in z is very good about 15um in central) Central Sasha Milov HI walkthrough Sept 13, 2010 9

Tracking optimization We are re-using the p+p definitions • Truth track: • mc_barcode -> at(truth_track_0pt) > 0 • mc_barcode -> at(truth_track_0pt) <200000 • mc_charge -> at(truth_track_0pt) > 0 • fabs(mc_gen_eta -> at(truth_track_0pt)) > 2.5 • mc_gen_pt-> at(truth_track_0pt)) > min_pt • Reconstructed track: • fabs(trk_d0_wrtPV -> at(reco_track))<1.5 • fabs(trk_z0_wrtPV -> at(reco_track)*sin(theta))<1.5 • fabs(trk_eta -> at(reco_track)) < 2.5 • trk_pt-> at(reco_track)> min_pt • Associated track: • trk_mc_index -> at(reco_track) > 0 • trk_mc_probability -> at(reco_track) > 0.5 • Efficiency: – Associated / Primary • Fake Rate: – (Reconstructed – Associated) / Reconstructed Sasha Milov HI walkthrough Sept 13, 2010 10

Centrality based on truth For the optimization we use truth based centrality because a) we do not know what multiplicity will be in real centrality bins b) the detector effects are determined by occupancy, not directly by the Pb+Pb b imp Sasha Milov HI walkthrough Sept 13, 2010 11

Tracking optimization For each setting we checked efficiency and fake rates in p T , rapidity, and centrality. And also CPU/memory use. Sasha Milov HI walkthrough Sept 13, 2010 12

Optimization Summary Tracking param. p+p Setup HI Setup HI optimized minPT 500 MeV 1000 MeV Tested to 500 MeV minClusters 7 9 9 minSiNotShared 4 7 7 maxShared 3 2 2 maxHoles 3 2 1 maxPixelHoles 2 2 0 maxSctHoles 2 2 1 maxDoubleHoles 1 1 1 radMax 600 600 600 roadWidth 20 20 20 seedFilterLevel 2 1 1 Xi2max 15 6 4 Xi2maxNoAdd 35 10 10 radStep 2 2 2 maxSize 20000 20000 20000 maxSizeSP 1500 4000 4000 mindRadius 10 10 10 maxdRadius 270 270 270 maxdZver 5 5 5 maxdZdRver 0.02 0.02 0.02 maxdImpact 10 10 10 maxdImpactPPS 1.7 1.7 1.7 13

New working point OLD point NEW point Sasha Milov HI walkthrough Sept 13, 2010 14

Tracklets Sasha Milov HI walkthrough Sept 13, 2010 15

Tracklets in p+p Tracklets using B and 1 st (2 nd ) layers of Pixel detector allow effectively reduce the threshold down to 100 MeV, and thus get (almost) the full dNch/d  measurement. p T >500MeV Tracklets were tested on 0.9TeV and 7TeV MC and on the real data with p+p and produce good results. 16 Sasha Milov HI walkthrough Sept 13, 2010

Tracklets in HI Tracklets also work very well in HI: • The occupancy of pixel remains low in the most central events • Proximity to the vertex greatly diminishes the uncertainly due to strange particle decays. Tracklets are the ideal method to get Day-1 multiplicity result in HI Issues: • fake rate (can be removed by background subtraction techniques) • particles with p T < 100 MeV. Both can be easily solved with the 0-field to get the result faster. Sasha Milov HI walkthrough Sept 13, 2010 17

Conclusions • Centrality – Centrality determination is implemented and working. – Might need calibration with real data. • Tracking optimization for HI – Vertex method changed, it works very well, very efficient. – Tracking works well in HI. The efficiency is ~70 in full  range – Decrease by 15% in the most central events. – Fakes reach ~5% in the most central events after optimization of the tracking parameters, predominantly at high  • Tracklet analysis – Main goal is to reduce the threshold to 100 MeV and make the multiplicity measurement. – Works well in p+p and shows stable results in HI MC. – Would produce a very fast result with a zero field run. Sasha Milov HI walkthrough Sept 13, 2010 18

BACKUPS

Centrality Implementation • HICentrality package – provides measure of centrality in percent bins • Based on observables – Total energy on the em level as calculated from cells – Total number of silicon pixel clusters – Hijing truth • Calibration – Calibration histograms accessed via COOL reference from files distributed over the Grid in conditions datasets – Latest calibration done with Hijing minbias private reconstruction in rel. 15.6.9.3 • Validation – Validation of centrality variables and bins available in HIValidation package – Checking distribution of centrality bins, comparing to Hijing truth • Running with new calibration in Athena rel. >= 15.9.0 (dev) Sasha Milov HI walkthrough Sept 13, 2010 20

• HICentrality algorithm is now run by HIRecExample heavy ion reconstruction configuration by default. The doHIGlobalVars and doHICentrality flags are turned on. • The default calibration file HICentralityCalibrations.root is stored in repositiory and installed during the build. It needs to be copied to run area by standard get_files python utility. • In HICentrality algorithm the HICentralityData object is created and registered with StoreGate under the name of HICentrality for each event. The centrality bins are calculated with 100x1% precision and a default 10x10% schema is setup in centrality object. • 1% bins are available by methods of HICentralityData class: GetImpactParameterBin(), GetNwoundedBin(), GetNcollBin(), GetCaloCellEnergyBin(), GetNumberOfSiClustersBin() • selected schema bins are available by other methods: GetImpactParameterBinBySchema(), GetNwoundedBinBySchema(), GetNcollBinBySchema(), GetCaloCellEnergyBinBySchema(), GetNumberOfSiClustersBinBySchema() Sasha Milov HI walkthrough Sept 13, 2010 21

Some basic properties Sasha Milov HI walkthrough Sept 13, 2010 22