Download presentation
Presentation is loading. Please wait.
Published byHelen Foster Modified over 8 years ago
1
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano1 EtMiss in first ATLAS data Introduction First results from data MET at EM scale Comparison with MC MET cleaning and tails Towards the Final MET Conclusions Plans for 2010
2
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano2 Missing transverse energy (MET) based on 3 global variables: Scalar sum of transverse energy: E T Opposite of vectorial transverse energy sum projected on x, y-axis: METx, METy Module: MET=√METx 2 +METy 2, Direction: Miss =atan2(METy/METx) Missing transverse energy with 2009 data: Random events : = =0, = 0 Control calorimeter (noise, pedestal, …) Check tails in E T, MET distributions Minimum bias events : = =0, E T in [0,100] GeV Comparison data-MC Check tails in E T, MET distributions First performance plot: METx,y) vs E T MET Calibration ? With ½ million of Minimum Bias events, lots of things to do Introduction
3
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano3 3 Basic E T miss from all Calorimeter cells with two possible noise suppression approaches ( MET_Base, MET_Topo) Final E T miss adding calibration step plus contribution from muons and for dead material (MET_Final): –Different calibrations approaches: Global cell energy density calib (GC) and local hadron calib (LC) –Correction for energy lost in cryostat between EM and Had calorimeters (MET_Cryo) from jets –Contribution from muons Basic E T miss MET_Base MET_CorrTopo MET_Calib MET_Cryo MET_Muon MET_Final CaloCells MET_Topo |E|>2 noise TopoClusters MET_LocHadTopo NO calibration Global calibration Local calibration Global calibration Apply calibration: Final E T miss From Basic to calibrated MET In the following default is MET_Topo (EM Scale)
4
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano4 4 plots approved in December with low statistics After Xmas reprocessing: update the plots with all 2009 statistics+new plots→CONF Note 2009 RANDOM 2009 BEAM 1 “good” run ~4000 evts First results from LHC data ~35Kevts 900Gev ~12Kevts 2.36TeV 1 “good” run ~4000 evts 3 plots su 4 fatti da Milano D. Cavalli S. Resconi R.Simoniello
5
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano5 Can start to write 3 MET CONF notes Based on data from December reprocessing good run list + agreed selection: MET-CONF 1 : “ Jet/Etmiss data quality and jet/Etmiss cleaning” Common with Jets MET-CONF 2 : “ Etmiss performance based on calorimeter cells” (editors: D. Cavalli, P. Pralavorio) Update of approved MET_Topo (EM scale) + other understood plots: MET for random MET, METx, METy at 900 GeV and 2.36 TeV and METxy resolution curve + Time stability + different contributions to MET MET-CONF 3 : “ Etmiss performance based on tracks” Dead line for end of February MET CONF notes on first data Show these plots in the following
6
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano6 Apply a selection to increase minimum bias purity Trigger MBTS_1_1.AND. MBTS timing ( t A-C <10 ns).AND. LAr timing ( t A-C <5 ns) Keep only event if all jets (AntiKt D=0.6) have EM fraction: emf >0.1 Keep only event if all jets satisfy n90 ≥ 5 (n90= Ncells containing 90% of jet energy) Data: reprocessed DESD_COLLCAND (r988_p62) Toroid and Solenoid field on, and √s = 0.9, 2.36 TeV All selected runs should have data quality flag y+ Good Run List (JetETmiss GRL): statistics ~ 500Kevts Monte-Carlo: MC09 geometry, new tuning PYTHIA Minbias events: Single/Double diffractive (SD/DD) + Non diffractive (ND) DD/SD/ND = 6.4 / 11.7 / 34.4 mb All plots to be done with these requirements. JetEtMiss central production of DESD (and D3PD) ongoing… (not yet final plots in the following) Data and Selections to be used for MET CONF Note (Bad cell masking: 10 bad channels in EMBA, EMBC, HECC - still an unmasked sporadic noise cell in HEC [-1.9,1.9])
7
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano7 Redo performance plots with ~100 times more events (13 runs, Nevts=384726) 900 GeV data: METx, METy No tails (GRL after emf>0.1) Good agreement data-MC (only ND here): bulk of METx, METy distributions understood Similar results as for the approved plots Similar results at 2.36 TeV (see backup slides)
8
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano8 Redo performance plots with ~100 times more events (13 runs, Nevts=384726) No tails (GRL after emf>0.1) Good agreement data-MC (only ND): bulk of MET distribution understood Similar results at 2.36 TeV 900 GeV data: MET, SumET Better agreement with new MC respect to old one, but still some discrepancy between data and MC shape Similar results at 2.36 TeV Look in details at E T distribution Sum the transverse energy of ~2500 cells mainly from Endcap region
9
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano9 Investigate data-MC discrepancy by looking at different hypothesis : Physics ? Compare with PHOJET. Add Single and Double diffractive events: Only impact E T <5 GeV … To go further requires input from physics Trigger bias ? Incorrectly modelled in MC (affects the low E T regions) Will test orthogonal trigger selection Detector bias ? Main discrepancy in FCal (but random OK) Investigate support tube sagging impact … No impact on MET performance (see next slide) But can affect some physics analysis Random Data MC SumET
10
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano10 The resolution curve is not biased from tails (fit the core of METxy distributions) can use all stable runs data (20 runs, Nevts=792958) METx, METy Resolution vs SumET 2009 BEAM
11
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano11 Reprocessed Data: METphi (1) mc09_900GeV.105001.pythia_minbias.recon.AOD.e500_s655_s657_d257_r1023_tid104562_00 (2) mc09_2TeV.105001.pythia_minbias.recon.AOD.e499_s672_s673_d270_r1039_tid104683_00 Phi asymmetry mainly due to beam spot displacement Different beam spot in data and MC The 2 MC have different beam spot 900 GeV and 2.36 TeV data: MET_phi calculate MET from reconstructed vertex Include the correct beam spot position in MC: MC with same beam spot as data coming soon! Directly affects physics analysis
12
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano12 Study stability of MET quantities as a function of time Here show only one example (METx) Monte Carlo Data Results stable with time within +/-5% 2009 RANDOM 2009 BEAM Double Gaussian Tile Noise not in MC ≠ Beam spot position in data and MC 2009 BEAM MET time stability
13
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano13 New physics may produce high MET Need to control fake MET at a very high level (>10 -6 ) Main sources of Fake MET Hardware (noisy cells, problems during DAQ, …) Software (corrections for “bad” calorimeter regions) Physics (Cosmic background) Strategy (up to now): remove events if any “noisy” jet (emf<0.1, n90<5) Work started on alternative solutions: Detect fake Tile TopoCluster, use cluster timing After cleaning (with detector/jets), no remaining tails … with our statistics ! Before After removing “noisy” jets evts Event with jets MET tails
14
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano14 As no show stoppers from MET_Topo variables at EM scale Can look at the calibration … good agreement data-Monte-Carlo for all calib need more events with jets to study calibration In MB calibration is more difficult because of low pT particles which do not reach the calo… Can check other contributions to MET_Final (even if almost 0…) Good agreement data-MC, but need high statistics evts with jets (and muons) Calibration and other MET terms MET_Muon calculated from combined muons (momentum from Spectrometer only) Met_Cryo calculated from jets MET_Muon MET_Cryo
15
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano15 Based on all reconstructed physics objects (e/ , , b-jet, jet, ) + unused TopoClusters Most complex schema to apply after validation of reconstructed objects: After particle identification, decomposition of each object into constituent Calorimeter Cells Overlap removal done at cell level Cell calibration weights depend on the type of the reconstructed object (e/ , , b-jet, jet, …) they belong to MET_CellOut + MET_RefEle MET_Ref MET_RefTauMET_RefJetMET_RefMuo MET_Cryo MET_MuonBoy MET_RefFinal +++++ += Go back to constituent Calorimeter Cells apply overlap removal at Cell level Cell calibration weights dependent on the object add them to calculate partial terms Electrons Jets Muons Unused TopoClusters Taus Photons isEM=Medium pT> 10 GeV isEM=Tight pT> 10 GeV isTau=TightpT> 5 GeV MET_RefFinal
16
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano16 contribution to MET from different objects and from cells outside objects can be separated In MinBias evts ~no reco objets dominated by MET_CellOut (cells in TopoClusters unused in physics objects) and MET_RefJet to be done at EM scale for CONF note MET_RefFinal Good agreement data-MC MET_CellOut
17
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano17 4.10 5 (3.10 4 ) Minbias evts at 0.9 (2.36) TeV provide a first test of MET The algorithms in MET package work well and are robust. Work at EMscale with cells from TopoClusters : MET_Topo Scalar sum of transverse energy ( E T ) distribution first compare to MC: Discrepancy under investigation but not altering MET performance Missing transverse energy (METx, METy, MET): Good agreement data-MC for distribution and performance With detector + event cleaning (noisy jets), MET tails compatible with MC miss oscillation mainly due to beam spot displacement A look at MET calibration, and other terms entering final MET: Encouraging results … but much more data is needed to conclude ! No show stoppers but need more events with jets and events with real MET to continue ! Conclusions
18
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano18 news from Chamonix: The LHC will run at 3.5+3.5 TeV in 2010 and 2011, until the experiments collect an integrated luminosity of ~ 1 fb-1. Only a short technical stop is foreseen at the end of 2010- beginning of 2011. -- The 2010-2011 run will be followed by a long (~1 year) shut-down, to redo all splices and thus enable the machine to operate up to the design energy (7+7 TeV). Plans for 7 TeV ~10 pb-1 : QCD di-jets - MET calibration (compare GC and LC) 50-100pb-1: W production - set MET scale with W lnu 100-200pb-1: Z production - diagnostic plot in Z ll - set MET scale with Z Tautau Plans for 2010 Roadmap for MET Validation Before Collisions: Data Quality –Noise –Dead channels 2009 data: First collisions –From basic EtMiss: Minbias Dijets (calibration) –to Refined EtMiss: Use reco objects when valid –Surprises in data: pile-up, tails 2010 data: 10-100 pb -1 –Z ee, Z mumu –Neutrinofication –W lnu –Z tau tau –ttbar –MET Tails in SUSY Milano group will continue to be deeply involved in MET commissioning with 7 TeV, after the first very successful experience of work on data in Milano!
19
Very severe event selection Invariant Mass Peak Position(GeV) = 91.8 +/- 1.4 stat +/- 0.5 syst Systematic error on lepton and tau (scale and resolution) are small compared to the statistical error (0.4) Systematic error from fit stability (0.3) No background subtraction needed, but have to take into account error on background level …. Plots done on a statistics correspondent to 100pb -1 for signal & EW bkg At 7TeV need 140 pb-1 to have the same statistic → 8% precision in MET scale with 100pb-1 at 10 TeV Signal: 1.48 ± 0.051 evts/pb -1 → Analysis Acceptance 0.13% All backgrounds: 0.12± 0.03 evts/pb -1 → S/B = 12.3 MET scale from Z invariant mass
20
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano20 Backup
21
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano21 METx, METy Resolution in CMS worse by a factor ~1.3 METx, METy Resolution vs SumET ATLAS 2009, 900 GeV and 2.36 TeV data
22
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano22 We used Xmas reprocessed data and MC: - DESD_COLLCAND.r988_p62: (L1 MBTS. OR. offline MBTS.OR. LAR Endcap timing): - Bad cell masking: 10 bad channels in EMBA, EMBC, HECC 1) Stable runs at 900GeV (20 runs, Nevts=792958): 142383, 142195, 142194, 142193, 142191, 142190, 142189, 142174, 142171, 142166 142165, 142154, 142149, 142065, 142042, 141999, 141994, 141811, 141749, 141748 2) Runs from egamma GoodRunList at 900GeV (13 runs, Nevts=384726): 141749, 141811, 142149, 142154, 142165, 142166, 142171, 142174, 142189, 142191, 142193, 142195, 142383 egamma GRL (05/01/2010):all detectors at least yellow, and accepting the run with the toroid ramping as no impact on calorimeter/ID expected.. 3) New MC at 900GeV (Nevts=800K): mc09_900GeV.105001.pythia_minbias.recon.AOD.e500_s655_s657_d257_r1023_tid104562_00 4) Runs at 2 TeV (Nevts-25741): 142402 5) New MC at 2 TeV Nevts=100K): mc09_2TeV.105001.pythia_minbias.recon.AOD.e499_s672_s673_d270_r1039_tid104683_00 Etmiss results with Reprocessed Data at 900GeV and at 2TeV
23
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano23 Egamma final list of good run for 900 GeV From egamma share point (05/01/2010): - The good run list for 900 GeV has been modified requesting the tiles to be at least yellow (as for other detectors) and accepting the run with the toroid ramping as no impact on calorimeter/ID is expected. The selection has been done with : - find run 141000+ and dq lar,pix,sct,trtb,trte,til y+ and dq atlsol g and lhc beamenergy 449-451 - We have not updated the 2 TeV as SCT was never on. GR L proposed for the jetetmiss dESDs - jetetmiss_noindet_900GeV.xml = 900GeV runs w/o indet requirements - jetetmiss_windet_900GeV.xml = 900GeV runs w/ indet requirements - jetetmiss_woindet_2TeV.xml = 2.4 TeV runs w/o indet requirements (there are no 2.4 TeV runs w/ indet ok) There are additional good run lists that require in addition the jet and met DQ flags to be green, the output of which will be made available in the dESDs as per-event skim-decision objects.
24
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano24 How Jet/EtMiss will deal with yellow DQ flags There would be only Green or Red Jet/EtMiss DQ flags associated to a reprocessing tag ("Binary" option). A yellow flag would become green for a reprocessing tag if we have a way to fix this problem at the analysis level, otherwise it would become red. It would remain yellow in the head until the problem is properly fixed. For the December 09 reprocessing, the two main problems that caused yellow Jet/EtMiss DQ flags were: - an unmasked sporadic noise cell in the HEC [-1.9,1.9] - and bad energy correction due to missing module in Tile [1.0,1.2] and [-1.1,-1.8]. These two problems can be fixed by applying extra cuts. Therefore, all the Jet/EtMiss DQ flags for December 09 reprocessing were set to green.
25
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano25 Performance plots (1 run evts-25741) Good agreement data-MC: bulk of METx, METy distributions understood Similar results as at 900 GeV 2 TeV data: METx, METy
26
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano26 Good agreement data-MC: bulk of MET distribution understood Similar results as at 900 GeV 2 TeV data: MET, SumET Performance plots (1 run evts-25741) Better agreement with new MC respect to old one, but still some discrepancy between data and MC shape Similar results as at 900 GeV
27
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano27 Look at Event Em Fraction Use Event Em Fraction cut to suppress tails: calculated from the weighted sum of the jet Emfrac (in METPerf package): emfrac =(PS+EM+FCAL0)/(PS+EM+HEC+FCAL+Tile+Gap); can be calculated only in events with jets: 21544 evts over 792958 (2.7%) for data (Event Em Fraction >0.1 eff~ 91 %) 32634 evts over 800K for MC (4%) (Event Em Fraction >0.1 eff~ 96 %) Public results from cosmics studies ApprovedCosmicPlotsJetEtMiss Event EM Fraction for data and MinBias MC Peak at zero also in MC to be investigated (jet in cracks) ?
28
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano28 Compare GRL reprocessed data (384726 events), GRL reprocessd data emfrac cut (383770 events) and the new MC after emfrac cut (798618 events) – MC entries normalised to entries of cut reprocessed Data from GRL runs: METx, METy, MET, SET SumET distribution from data are now in better agreement with new MC Only a small improvement of tails after bad channels masking – tails suppressed by emf cut
29
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano29 Comparison of data with MC08 and MC09 MC Better agreement in SumET distribution with new MC (MC09), but still not very good Similar results at 2.36 TeV 900 GeV data: MET
30
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano30 GRL runs without cut in EM frac (implementing now the possibility to produce all METPerf plots after selection) Tails in HEC probably due to sporadic noise cell. All distributions in good agreement with MC except FCAL SUMET in different calorimeters SumET in different calos
31
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano31 -Very good agreement for EMB, EMEC, HEC and Tile -Physics close to noise in HEC and Tile -Disagreement data-MC comes mainly from FCal SumET in different calos after emf cut
32
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano32 Data higher than Monte-Carlo in all the subdetectors Ncells in TopoClusters in different calos
33
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano33 MET MET_Topo MET_CorrTopo MET_LocHadTopoMET_Final Vey similar ‘cause MET_Muon and MET_Cryo ~ 0 No CalibGC LC Best agreement between data and MC for LC As expected, MET increases in calibrated data
34
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano34 RefFinal Final Larger tails in RefFinal respect to Final
35
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano35 In MET_RefFinal contribution to MET from different objects and MET_CellOut can be separated In MET_RefFinal, cells belonging to different objects (e, jets,…) are calibrated differently In MinBias evts ~no rec. objets Dominated by MET_CellOut (MET_RefJet) Interesting to compare MET_CellOut, MET_RefJet with MC MET_RefFinal
36
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano36 Contributions of different reco objects to METx_RefFinal Tails in MET_RefFinal: due to MET_MuonBoy and MET_RefJet Cells in jets are treated in the same way in MET_Final and MET_RefFinal In MET_Final: MET_Muon calculated from Spectrometer tracks only Larger tails in MET_RefFinal due to MET_MuonBoy
37
ATLAS Italia 02/02/2010D. Cavalli – INFN Milano37 MET_Tracker now included in ATLAS software provide an alternative to Calo-Based MET (MET_Topo Good agreement data-MC Performance worse by 50% w.r.t. MET_Topo MET_Calo Tracker-based MET IDHits MET_Track “Good” Reconstructed Tracks Inner Detector (90 M channels) | |<2.5 From Primary Vertex Remove Conversions
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.