Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 1 TAU/JET/E T MISS TRIGGERS IN ATLAS
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 2 OUTLINE Tau/Jet/E T MISS trigger description : what is in the TDR and what is new. An example of an analysis : optimization of tau and E T MISS triggers for W at L=
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 3 TAU/JET/E T MISS SOURCES AND INTEREST FOR PHYSICS Standard Model: –inclusive W τν (Z ττ) production –QCD. SM and MSSM Higgs: – GeV SM Higgs: qqH(ττ) –A/H ττ –H + τν (m H + m top ) SUSY Compositness
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 4 The ATLAS trigger Level 1 (hardware): Defines Regions of Interest (RoI). Uses Calo cells and Muon chambers with reduced granularity. e/ , jet candidates. Level 2 (software): Seeded by LVL1 RoI. Full granularity of the detector Performs calo-track matching Event Filter (software): Offline-like algorithms. Refines LVL2 decision Full event building ~200 Hz ~2 kHz 2s2s 10ms 1s <75(100) kHz Execution time TIER 0 mass storage High Level Trigger
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 5 LVL1 Trigger Rates Selection at 2x10 33 cm -2 s -1 Rates (KHz) MU MU60.2 EM25I12.0 2EM15I4.0 J J J650.2 J60+xE600.4 TAU25+xE302.0 MU10+EM15I0.1 OTHERS (pre-Scales,calibration,…)5.0 TOTAL 25
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 6 HIGH LEVEL TRIGGER RATES SelectionPhysics coverage2x10 33 cm -2 s -1 Rates (Hz) Electron Higgs, new gauge bosons, extra dim., SUSY, W/Z, top e25i, 2e15i ~40 PhotonHiggs, SUSY, extra dim. 60i, 2 20i ~40 Muon Higgs, new gauge bosons, extra dim., SUSY, W/Z, top, B-Physics m20i, 2m10 2m6 with m B /m J/ ~50 Jets SUSY, compositness, resonances j400, 3j165, 4j110 ~25 Jet & E T miss SUSY, leptoquarksj70 + xE70 ~20 tau & E T miss Extended Higgs models (e.g. MSSM), SUSY 35 + xE45 ~5 Otherspre-scales, calibration, … ~20 Total~200 The rates for the HLT taken considering the EF performances equal to those one of the OFFLINE.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 7 LVL1 CALORIMETRIC SIGNAL PROCESSING TAU/JET/E T MISS triggers are calorimetric but they use two different processors
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 8 TAU TRIGGER
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 9 Hadronic Tau Trigger (I) Hadronic Tau Trigger (I) (ATL-COM-DAQ ) LVL1 trigger: look at 4X4 matrix of calorimetric towers ( = 0.1 x 0.1 each trigger tower). E T threshold for the central core (EM+Had) and isolation thresholds between core and 12 external towers for e.m. and had. calorimeters. + track multiplicity in the RoI second layer of EM calorimeter LVL2 trigger: look at the shower shape in the 2nd layer of e.m. calorimeter and at the track multiplicity inside the RoI defined at LVL1. Cut on the ratio between E T contained in a 3x7 cell cluster and E T in a 7x7 cell cluster and on track multiplicity For | |<2.5 as in the TDR
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 10 Hadronic Tau Trigger (II) Hadronic Tau Trigger (II) (ATL-COM-DAQ ) LVL3 (Event Filter) : look at the complete event. The variables of the offline algorithms are used as an approximation of the LVL3 trigger five variables: number of reconstructed tracks, within R = 0.3 of the candidate calorimeter cluster, between 1 and 3; cut on isolation fraction, defined as the difference between the E T contained in a cone size of R=0.2 and 0.1 normalized to the total jet E T ; cut on EM jet radius, an energy weighted radius calculated only in the e.m. calorimeter ; cut on EM energy fraction, defined as the fraction of the total jet energy in the e.m. calorimeter; threshold on the p T of the highest p T track. as in the TDR
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 11 EFFECT OF TRIGGER SELECTIONS E T coreEM iso HAD iso
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 12 TAU TRIGGER EVOLUTION ) from EMSamp2 Calo variables (more variables used than in the TDR) Tracking (# of tracks, charge,…) Final decision : matching of cluster and tracks, energy estimate LVL2 : Calorimeter based approach Perform tracking and obtain Calorimeter variables Final decision : matching of cluster and tracks, energy estimate with energy flow LVL2 :Tracking based approach New:studied for Very Low Lumi cm -2 s -1 Current approach For the LVL1 different RoI sizes are under study (timing, resolution and efficiency,…) Under developing an EF tracking based algorithm.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 13 JET AND E T MISS TRIGGER
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 14 identify hadronic jets using calorimetric data; classify them according to E T ; provide multiplicity of jets passing required threshold; provide the coordinates of the candidates to the LVL2; have an energy resolution as good as possible for high E T and low E T jets. JET ELEMENT : = 0.2 x 0.2 (now only one sample in depth) Algorithm : - 2x2 jet element cluster (0.4x0.4) to identify a jet RoI, it is a local E T maximum. - 4x4 jet element (or 3x3 or 2x2) trigger cluster to measure the jet E T. Trigger cluster size : -big enough to have a good energy resolution for high E T jets (containment) -not too big for low E T jets (noise and pileup) RoI size and step size : -spatial resolution and jet separation. For | |<3.2 LVL1 JET TRIGGER as in the TDR
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 15 ROI L2 Jet Jet Cone Size several iterations needed: timing is an important key. LVL2 JET RECONSTRUCTION LVL2 starts from LVL1 RoI information ( , φ location) Iterative cone algorithm (R=0.4) to calculate weighted , φ energy center. Possible granularities : cell-based, LVL1 trigger towers,…. Jet calibration (energy scale and resolution) has an important effect on trigger efficiency.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 16 JETS AT THE EVENT FILTER Results of the study: Jet reconstruction is better with a higher size of RoI, Higher size of RoI requires more time, Topological clusters are faster than calorimeter towers but Towers reconstruct better p T of jets Study of jet reconstruction at the EF : Size of Region of Interest (RoI): 16 (0.4x0.4), 32 (0.8x0.8), 64 (1.6x1.6) Different types of clusters: topological clusters or calorimeter towers Algorithms: Fast KT, Cone Dijets samples with 35 GeV < P T < 1120 GeV
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 17 E T MISS TRIGGER E T MISS is a global variable. LVL1 : Calorimeter energies summed into a map with a granularity = 0.2 x 0.2. Ex, Ey, E T, E T MISS are computed. E T MISS trigger is not a standalone trigger, but it will be used in association with jet or tau trigger. Rapidity coverage : critical for E T MISS trigger performances. For | |<5 Rapidity coverage Mean value of E T MISS All0.9 GeV | | < 55.6 GeV | | < 48.8 GeV | | < GeV as in the TDR For QCD events : trigger rate (KHz) E T MISS (GeV)
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 18 LVL2 possible strategy: Based on LVL1 Missing E T “ROI” (with scalar E T, ΣE x, ΣE y ) Based on LVL1 Jet ROIs - Use cell data for each RoI Based on Trigger Towers: - Refine with better calibration and replace saturated towers For all the above, add muons Possible Strategies for E T MISS Trigger at LVL2 and EF EF possible strategy: Using FEB header ΣE X, ΣE Y from RODs Using full cell data. For both of these, add muons.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 19 Algorithms for Tau/Jet/E T MISS triggers are still under development : not a final decision taken. ATLAS physics groups have started now to include trigger information in the simulations to perform analysis : a trigger part to perform a Trigger Aware Analysis has been added in the last releases of the ATLAS software. The Tau Trigger slice is going to be added now : no analysis available with the “true” simulation of the tau trigger. Trigger effects can be emulated : next slides will show an analysis.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 20 “Trigger aware analysis” from user perspective Trigger optimization and prospects for W with 100 pb -1 (few weeks of data taking at very low luminosity cm -2 s -1 ) Data samples: events W dijet events (J1-J2-J3) For topological studies ~10^8 events from fast simulation Daniel Froidevaux and Elzbieta Richter-Was
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 21 Why W ? Extract signal for most abundant source of -leptons as early as possible. This requires a performant and E T miss trigger from the very start! For L = , baseline plan is to trigger on 25I + XE30 at LVL1 (for a rate of about 2 kHz) and to raise the thresholds to 35i + xE45 at the HLT (for an output rate of about 5 Hz). Measurment of W / W e to confirm good understanding of trigger/reco/identification efficiencies E/p measurement in single-prong decay for calorimeter calibration. Assumed that trigger chain is fully operational and that the detector operates more or less as expected (especially in terms of E T miss performance). Expected rates for 100 pb -1 W hadron W e Z hadron .B (pb) 30i + xE35 ~ 15000~ ~ 1300 20i + xE25 ~ 60000~ ~ 3500 Efficiencies of ~ 80% for the trigger and of ~ 50% for the id/reco of hadronic decays were assumed.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 22 Algorithm to “emulate” LVL1 trigger seed “RoI” with topo-clusters, accept if E T > 5 GeV calculate energy in 2x2 and 4x4 towers of 0.1x0.1 ( x ) noise subtraction not applied, cells with negative energy suppressed from enegy counting energy in HAD (originaly at EM scale) multiplied by 1.25 remove overlapping “RoI” with iterative procedure, imposing separation by R > 0.3 missing energy taken from ”uncalibrated calo off-line”.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 23 More on LVL1-like RoI’s L1 tower multiplicity < <> = 1.5 multiplicity <> = 0.78 Total QCD E T spectrum signal E T spectrum R between RoI’s R between RoI’s
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 24 Energy resolution for signal RoI and threshold efficiency Threshold: 0.75 * 20 GeV 90% efficient at E T visible = 20 GeV 2x2 RoI 4x4 RoI
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 25 Energy resolution for background RoI and threshold efficiency Threshold: 0.75 * 20 GeV Factor 10 rejection at E T visible = 20 GeV 2x2 RoI 4x4 RoI 2x2 RoI
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 26 Isolation for EMTau RoI QCD signal 1.E T L1otherEM /E T L1core < E T L1otherHAD /E T L1core < E T L1core /E T L1tower > E T L1otherEM /E T L1coreEM < E T L1otherHad /E T L1coreHad < 1.0 L1core = 2x2 L1tower = 4x4 Full cirles: with threshold E T L1tower > 0.75 * 20 GeV Isolation very loose.... Factor 5 rejection at 80% efficiency but almost no improvement if E T threshold added.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 27 More on E T miss : we have only off-line available truth off-line Atlfast E T miss is calculated at EM scale, from calo only. E T miss, E T MissFinal, E T TruthNonInt, E T MissAtlfast
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 28 LVL1 trigger: “ E T L1tower 20 GeV Isol + E T miss 20 GeV ” ( this means E T L1tower 0.75 * 20 GeV and isol_1+ isol_2 ) Rates: 0.02 Hz signal 2* 10 5 events for 100pb Hz QCD bgd S/B ~ Concluding on LVL1-like selection Isolation criteria rather weak. We use E T miss from uncalibrated calo at EM scale... L=10^31
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 29 The LVL2 like selection: explore track-seeded reconstruction Algorithm for tauL2 : start from track with p T > 9 GeV accept if no more than 2 associated tracks in R 2 GeV store info on “track quality” of leading track for futher discrimination build energy from simplified Eflow (energy overestimated by 10%-20% mostly because noise not suppressed) calo identification variables from EM2 or all EM calo Same definition for E T MISS as at LVL1 Signal response = 1.16 = 0.17 Bgd response = 0.86 = 0.18
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 30 Concluding on LVL2-like selection LVL2 trigger: “E T L2tower 20 GeV + track quality + id EM2 + id all Cal + E T missFinal 20 GeV” Rates: 0.01 Hz signal 10 5 signal events “on tape”for 100pb -1 5 Hz QCD bgd S/B ~ Loose triger selection, now we have to supress bgd in off-line analysis L=10^31
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 31 Off-line analysis ~ 10^8 QCD events in fine p T bins E T miss > 60 GeV gives bgd: rejection 10^3, signal: accept 10% -> still 10^4 evt for 100pb-1, S/B ~ 0.2 without refined tau indentification Results with only fast-sim offline, E T miss has no instrumental tails ! ~ LVL2 thresh. We start offline analysis with S/B ~ and predicted ~10^5 signal events “on tape” Need rejection ~ 10^3 for effic ~ 50% or increase E T miss threshold How fast bgd is supressed with the off-line E T miss threshold.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 32 Now we can go back to plot from page 25: Atlfast Off-line 60 GeV
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 33 Verify what off-line tauid rejection is possible.... since some discriminantion power already explored when accepting LVL2 (calo+tracks) candidates. We take tauL2 candidate “on tape” (after LVL2 tauiD) and check efficiency for matching tau1P, tau3P identified with PDE-RS optimisation (one MVA technique among many) tau1P tau3P discriPDERS > % 6.5% 46% 3.0% % 3.5% 23% 1.0% % 1.0% 4% 0.2% signal bgd signal bgd pT = 20 – 40 GeV Now we can go back to full-sim samples After L2 track-based trigger, discrimination fairly flat as function of p T tau1p (tau3p) ; track-based offline algorithm to identify 1-prong (3-prong) tau decay.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 34 Summary: LVL1 and LVL2 selection (calo+tracks) emulated for W analysis With rather soft selection ETmiss > 20 GeV + EMTauRoI > 20 GeV estimated for 10^31: 60 Hz after LVL1 5 Hz after LVL2 For off-line analysis start with S/B ~ ~ 10^5 signal events accepted for 100pb-1 Increasing E T MISS threshold helps in the background rejection: at 60 GeV threshold, supression 10^2-10^3 at 10% efficiency. Offline tau selection has to do the final work to extract the signal. Low luminosity provides unique opportunity to study low energy hadronic signatures in ATLAS (in view of SUSY) : important possibility to verify the understanding of tauID and E T MISS reco before attacking “New Physcics”.
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 35 BACKUP SLIDES
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 36 Sistema calorimetrico di ATLAS Tile Calorimeters Electromagnetic Liquid Argon Calorimeters Forward Liquid Argon Calorimeters η=1.475 η=1.8 η=3.2 Hadronic Liquid Argon EndCap Calorimeters EM LAr | | < 3 : Pb/LAr X 0 3 sezioni longitudinali 1.2 = – 1% equal. Central Hadronic | | < 1.7 : Fe(82%)/scintillatore(18%) 3 sezioni longitudinali 7.2 = 0.1 0.1 End Cap Hadronic 1.7 < < 3.2 : Cu/LAr – 4 sezioni longitudinali < 0.2 0.2 Forward calorimeter 3 < < 4.9 : EM Cu/LAr – HAD W/LAr 3 sezioni longitudinali
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 37 Tau Trigger Rate
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 38
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 39 More on E T miss : we have only off-line available
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 40 S.Levy, HCP session, July 2005
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 41 Signal and background at 14 TeV cross-section (PYTHIA) signal ~ 10 x higher QCD bgd ~ x higher than in CDF. ( ERW, ATL-PHYS ) <> = 16.6 GeV <> = 22.6 GeV <> = 18.4 GeV spectrum rather soft for E T miss, p T vis signal (ATL-PHYS )
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 42 Results from past studies (ATLAS) E/p measurements for calibration of hadronic calorimeters C. Biscarat COM-CAL Rejection Hz rates predicted after HTL at 10^33.trigger-like: ETmiss > 35 GeV + pTjet > 20 GeV preselection: veto iso lepton, veto iso photon tau-jet selection: track with p T > 30 GeV single-track: veto if extra tracks p T >1 GeV in tau cone narrow-jet: calo isolation total id rejection: 10^5 Events for 100 pb -1 : | 5270 W | 3630 W | 320 QCD jets (bb) total id effic: ~ 25%
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 43 Off-line analysis We don’t have enough events to continue with full-sim samples. We have move to fast-sim samples to study topological selection only and to estimate how much bgd suppression is possible: ~ 10^8 QCD events in fine p T bins
Francesca Sarri, INFN e Università di Pisa MCWS, 25/10/2006 LNF 44 Off-line analysis ~ 10^8 QCD events in fine p T bins Vetoying any other jet ‘a la CDF’ gives 30% accept for signal 25% acept for bgd We have looked at few more distribributions.... nothing obvious to optimise on... We started offline analysis with S/B ~ and predicted ~10^5 signal events “on tape” Need rejection ~ 10^3 for effic ~ 50% or increase E T miss threshold