1 / 34

MET Performance in Early Data

MET Performance in Early Data. People involved so far Ideas for early data projects and connections to physics groups Goals for winter / summer 2010. MET Tasks. Validate MET reconstruction performance early on, contribute to MET ‘sign off’

judd
Télécharger la présentation

MET Performance in Early Data

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. MET Performance in Early Data People involved so far Ideas for early data projects and connections to physics groups Goals for winter / summer 2010

  2. MET Tasks • Validate MET reconstruction performance early on, contribute to MET ‘sign off’ • Keep in mind: MET comes in different flavours (cf. backup slides) • 2 different MET versions at EM scale, just using calorimeter info – simplest version (cell based and topo-cluster based) • 2 different calibration schemes (local and global) for calorimeter signals • Cryostat and dead material terms, muon term • Refined MET (‘RefFinal’) would be the final step, it uses the energies of the reco objects directly (tuned by the reconstruction groups) plus muon and cryo term • Minimum bias and jet-stream data is first playground for MET • Assume real MET is zero, look at resolution, scale, tails, preferred directions • Medium term milestone would be validation in W/Z events • Obvious physics groups connections are SM W, top, SUSY Early data Late data D.Berge - CAT Physics Meeting

  3. People so far • Johan Lundberg: • Jet punch-through using muon spectrometer hits • MET resolution • SUSY group connection (planned) • Max Baak: • MET performance user data to provide ‘MET goodness’ estimate to ATLAS approved by MET experts (planned work) • SUSY group connection (planned) • David Berge: • MET ideal Monte Carlo performance • MET tails due to jet mis-reconstructions • MET in different calibration schemes • MET strategy for early data • Top/SUSY group connection (planned) • Daniel Froidevaux • Everything + W physics (trigger, analysis) D.Berge - CAT Physics Meeting

  4. Early data expectations Stolen from Laurent’s talk Truth-level dijet plot, PDF reweighting applied, from https://twiki.cern.ch/twiki/bin/view/AtlasProtected/DijetMassAndAngularDistribution 1 event per pb-1 900 GeVminbias SumET reach D.Berge - CAT Physics Meeting

  5. MET calibration with first collisions Example 10 TeVminbias MC We will be looking at such plots to evaluate the different MET terms! • Compare performance at different stages (MET at EM scale, with calibration, with muon term, dead material terms, etc) straight away to MC performance • Actual data curves will look much worse due to bad channels missed in the masking, wrong dead material corrections, displaced beams, beam-beam backgrounds • Getting it to look Gaussian with a mean MET close to zero is probably already quite a task! S.Resconi D.Berge - CAT Physics Meeting

  6. Tails in early data Example: J6 at 10 TeV Sublead jet eta • Plot to compare data versus MC immediately • E.g. additional dead material in real life could mean additional spikes in jet eta for large MET • Learn about regions in detector that produce MET tails, possibility to feed this back into early analysis (ignore events with jets in problematic eta regions) • Learn where calibration helps, and where other terms might be relevant • E.g. the crack at 1.5 is better with calibration added, worse again with muon and cryo term added! Difference here due to applying calibration, and then adding cryostat and muon term D.Berge - CAT Physics Meeting

  7. Jets punching through and seen in the muon spectrometer J6 at 10 TeV, MS hits versus lead-jet rec. eta Entries scaled to 1 pb-1 • Investigated so far impact of punch through on jet resolution and fake MET • Immediate handle on MET quality by cutting on muon spectrometer hits, but also important for data / Monte Carlo comparisons and additional correction for jet energy (latter point not at all covered in ATLAS!) • We are about to implement this into the common MET performance tools and make it available to the whole community Material in front of muon spectrometer (detector paper) D.Berge - CAT Physics Meeting

  8. MET goodness • The MET performance package is designed to be the tool that everybody uses to assess the MET performance for his analysis • We want to add a collection of variables in form of user data to this, along with reference cut sets people can apply for their analysis • I.e. for your SUSY analysis you would apply a cut and only use MET in events which pass the ‘loose / medium / tight’ MET goodness cut set • Thereby the MET performance people have a way of communicating which parts of the MET reconstruction they think is validated, and we also collect variables we know are relevant for the MET goodness in one place • Supported by the main MET responsibles • List of variables will include Calo timing, jet EM fraction, number of cells in MET, Df (MET,lead jets), eta lead jets, jet energy layer fractions, MS hits behind jets, calo-track MET consistency, MS non-pointing tracks, calo-track jets consistency, H1-LC MET consistency, etc. D.Berge - CAT Physics Meeting

  9. Topics to work on • Connection to physics groups • SUSY: tails and MET goodness cuts • Top: early scale, robustness, MET goodness cuts, MET (L1) trigger as orthogonal alternative to lepton triggers • W: early scale, MET (L1) trigger as orthogonal alternative to lepton triggers • Some technical aspects: • Prepare for easy MC comparison of data plots (quickly average together Jx samples) • Technical aspects of how to access data quickly (which format, where), and be able to rerun MET reco on ESDs • Study how much selection we actually need for balanced dijet events • Implement user-data stuff in MET perf package, make sure we can run it quickly over the perfDPDs / full ESDs • Medium term (1-20pb-1) interests: • Insitu validation of MET in events with real MET (W decays, 2300 events after selection per channel per pb-1!) • balance against hadronic recoil (Z decay, 200 events after selection per channel per pb-1!) • Rapid commissioning of the XE35 trigger chain (in particular L1_XE30) to be used as orthogonal trigger to estimate the e20_loose trigger efficiency for W measurement ( as early as 1-2 pb-1!) D.Berge - CAT Physics Meeting

  10. Winter Plots from 1 nb-1 at 900 GeV to 20 pb-1 at 7 TeV Summer Missing ET distribution versus Monte Carlo (QCD / minbias), raw distribution and applying quality cuts successively Maybe allow for overall normalisation difference, not shape • Missing ET resolution versus SumET from data (again balanced dijets and minbias), compared to Monte Carlo performance MET insitu resolution and scale from W->lnu MET bias against hadronic recoil pt from Z->ll D.Berge - CAT Physics Meeting

  11. Backup: MET Talk CAT Meeting 7 September D.Berge - CAT Physics Meeting

  12. Hadronic Calibration Workshop 2009: MET Summary • Workshop link: https://web.lip.pt/atlas-workshop/index.php • Agenda page: http://indico.cern.ch/conferenceOtherViews.py?view=standard&confId=48780 • Review panel report: http://indico.cern.ch/materialDisplay.py?sessionId=23&materialId=0&confId=48780 • Contributed material submitted beforehand, sessions mostly overview talks and discussions of submitted material • A lot of focus on jet reconstruction, calibration, energy scale • Less on missing ET (less people, less efforts) • Even less on data model/access, triggering D.Berge - CAT Physics Meeting

  13. Recap: Local hadronic vs. H1 calibration LC P.Loch D.Berge - CAT Physics Meeting

  14. Recap: Local hadronic vs. H1 calibration H1 • Local hadronic calibration: • Reconstruct the locally deposited energy from a cluster first, including classification of electromagnetic or hadronic origin • Apply corrections for dead material / out of cluster energy • Then the reconstructed energy is on average the isolated particle energy • Only then the jet finding is run, and jet corrections are applied • H1 style calibration: • Provide EM scale input to jet finding • Find jets on EM scale • Apply weights to recover hadronic scale (i.e. all corrections in 1 step) • Apply jet corrections D.Berge - CAT Physics Meeting

  15. Recap: MET reconstruction I S.Resconi D.Berge - CAT Physics Meeting

  16. Recap: MET reconstruction II S.Resconi D.Berge - CAT Physics Meeting

  17. Recap: MET reconstruction III S.Resconi D.Berge - CAT Physics Meeting

  18. Cosmic clean-up / cell masking session • A lot of work shown looking at cosmics and possible rejection using different calorimeter variables (Jet EM fraction, calo timing, track jets versus calo jets), BUT most obvious, muons, missing at the workshop • Many variables looked at, by many many people, would need coordination and coherence now • Beam backgrounds attempted to look at with last year’s data • No stats • Cannot trust simulations here • Need real collisions data! D.Berge - CAT Physics Meeting

  19. Session MET calibration with first collisions • Plan: go from base to final to refined MET, evaluate at each step with random trigger (before collisions) and then with collision data • Collision data: from minbias / QCD dijets (normally no real MET), to events with real MET (Z->tautau, W->lnu) ‘08 random trigger data From S.Resconi’s talk D.Berge - CAT Physics Meeting

  20. Session MET calibration with first collisions Example J4 250k events 150nb production xsec • Again, immediate test of MET resolution in events without real MET • Look at the tails from poorly measured jets From S.Resconi’s talk D.Berge - CAT Physics Meeting

  21. Session MET calibration with first collisions Punch-throughs Johan Tails etc. Different calibration schemes LC deals better with dead material, but slightly worse overall H1 LC D.Berge - CAT Physics Meeting

  22. Session MET calibration with first collisions Punch-throughs Johan Tails etc. Different calibration schemes LC deals better with dead material, but slightly worse overall D.Berge - CAT Physics Meeting

  23.   Session in-situ MET performance and closure with 10-100 pb-1 R.Teuscher’s talk D.Berge - CAT Physics Meeting

  24.   Session in-situ MET performance and closure with 10-100 pb-1 R.Teuscher’s talk Example: observable affected by beamspot displacement MET phi distribution affected by beamspot displacement, crossing angle, dead material, dead channels, hot cells D.Berge - CAT Physics Meeting

  25.   Session in-situ MET performance and closure with 10-100 pb-1 R.Teuscher’s talk Reminder: MET validation in events with real MET D.Berge - CAT Physics Meeting

  26.   Session in-situ MET performance and closure with 10-100 pb-1 R.Teuscher’s talk D.Berge - CAT Physics Meeting

  27.   Session in-situ MET performance and closure with 10-100 pb-1 R.Teuscher’s talk D.Berge - CAT Physics Meeting

  28.   Session in-situ MET performance and closure with 10-100 pb-1 • Z->ee,mumu projection method • Check balance of lepton pt versus hadronic recoil by investigating MET along and perpendicular lepton system Origin of bias clarified! CELL_OUT term suffers from low-energy particles losses! D.Berge - CAT Physics Meeting

  29.   Session in-situ MET performance and closure with 10-100 pb-1 • Much more material on ‘neutrinofication’ and MET from W transverse mass shown • Check agenda • Many approaches to validate MET with W/Z data • Attempts going on to channel / focus this work in Jet/EtMiss group • Collaboration with SM analysis groups D.Berge - CAT Physics Meeting

  30. Summary: MET homework • Cosmic clean-up: • most obvious approach: use the muon spectrometer • A lot of ideas/studies exploring calo timing, jet shapes, etc., needs coordination • Cell masking: L1/HLT approach to masking cells / channels disconnected from Jet/EtMiss offline reconstruction approach • Calibration of MET CellOut term (cells outside rec. objects) under revision • Studies to correct for low-energy particle loss using tracking ongoing • MET and jet-scale corrections: needs to be sorted out, up-to-now all jet-level corrections ignored to avoid double counting (cf. next page) D.Berge - CAT Physics Meeting

  31. Summary: MET homework D.Berge - CAT Physics Meeting

  32. Backup: Jet/EtMiss Meeting 29 September D.Berge - CAT Physics Meeting

  33. J6 correlations All events Require true MET>50 GeV Require true MET>100 GeV 82 events per 10pb-1 beyond 50 GeV difference LC tends to be larger LC tends to be larger True MET >100 GeV, LC TopoObj versus H1 CorrTopo+Cryo Event-by-event difference, no true MET cut SumET correlation all events D.Berge

  34. Tails – local hadronic versus H1 missing Et J6 All events, MET resolution • Asymmetric tail towards positive (rec-true) more pronounced for LC • Beyond 100 GeVDMET, 60% more events in LC scheme • 12816 events for LC versus 8077 for H1 • For 10pb-1: 49 events for LC, 31 events for H1 D.Berge

More Related