1. India-CMS meeting, BARC 30 th July, 2010 1 MonoPhoton Study Umberto Berzano, Satyaki Bhattacharya, Sandhya Jain Yurii Maravin, Tia Miceli, Albert de Roeck, Shruti Shrestha, Mani Tripathi

2. India-CMS meeting, BARC 30 th July, 2010 2 Background sources Non beam backgrounds (TM, SS, MT,YM) Cosmics and Beam Halo  Z(-> )+  M ain irreducible background  W+  if the lepton isn’t detected  W(->l ) electron can fake photons or lepton is lost  Z(-> )+jet jet can fake a photon  Z(->ll)+  leptons faking photons or not detected  DY(ee) leptons faking photons and not detected   +jet jet wrong reconstruction or faking photons   when one photon goes missing  QCD multijets jet wrong reconstruction or faking photons

3. India-CMS meeting, BARC 30 th July, 2010 3 Outline Monte Carlo Studies Photon selection MET selection Selection table present status Z gamma background W background Common code status Data at 7TeV Spike Events Work on Isolation variables Single photon skim

4. India-CMS meeting, BARC 30 th July, 2010 4 Photon ID Criteria 22X mono-photon Photon ID ECalRecHit isolation < 10 GeV HCal Tower isolation < 7 GeV Hollow Cone track isolation < 9 GeV HadOverEM < 0.1 31X Egamma ID Loose Photon ID ECalRecHit isolation < 5+0.0045*photon p T GeV HCal Tower isolation < 5 GeV Hollow Cone track isolation < 9 GeV HadOverEM < 0.15 Tight Photon ID Loose Photon ID Sigma_ietaieta < 0.013 (in Barrel) Tighter Photon ID Tight Photon ID Hollow Cone track isolation < 5 GeV HadOverEM < 0.1  R<0.3  R<0.4 Barrel only

5. India-CMS meeting, BARC 30 th July, 2010 5 Photon ID optimization cross section (pb) 22X ID 31X Tight ID 31X Loose ID Tighter ID  Using a tighter cut on track isolation works very well on events with electrons and charged fraction in jets thus Tighter ID is clearly the best solution  It was verified with N S /√N B calculation as well both with complete selection and without the two vetoes (to have QCD events)

6. India-CMS meeting, BARC 30 th July, 2010 6 MET selection  Calo MET  TC MET  Particle Flow MET

7. India-CMS meeting, BARC 30 th July, 2010 7 MET recontruction impact cross section (pb) ADD MD1d2 Z(inv)  W(e ) QCDpT>80 Calo MET TC MET PF MET  MET algos are involved in MET and  cuts  Final cross sections as a function of MET algo  after full selection (with some relaxed cuts)  Strong suppression of QCD background while the signal is conserved using PF MET

8. India-CMS meeting, BARC 30 th July, 2010 8 Signal Selection in 3XX  HLT: HLT_Photon25_L1R  Primary vertex: |z| 4  p T  > 50 GeV/c  Barrel only  MET > 50 GeV (PF MET) – Still its behaviour is to be verified in high pt region   > 2.7  Tighter Photon ID  track veto: rejects events with a track above 20 GeV/c  jet veto: rejects events with a jet above 50 GeV

9. India-CMS meeting, BARC 30 th July, 2010 9 Cumulative cross sections start  HLTPVbarrelp T  >50 MET>50  >2.7  ID TVJV MD1d2 0.5190.4590.4340.3170.2030.1920.1910.1880.149 Z ( )+  0.6730.5840.5740.3710.3340.3050.2720.2640.2130.208 Z(ll)+  7.32.682.651.510.4550.0280.0220.0210.0090.007 Z( )+j 450038.3638.3222.116.335.554.360.1750.1310.127 DYee 1300578.2571.6321.968.00.1030.0820.07000 W+  23.28.097.984.920.9190.2280.1700.1620.0830.079 W(e ) 789952985221302325331.216.212.40.660.63 W(  ) 789995.6195.1655.0510.355.063.500.3810.126 W(  ) 7899314.5311.8165.115.88.05.80.6110.3240.315  +j p T >50 3277.82641.32637.31523.61260.83.5723.1522.8531.1030.266 QCD p T >30 6.0 10 7 1.2 10 6 5.3 10 5 5.8 10 4 459.6192.8000 QCD p T >80 92380021941821939312165541113375.8203.72.960.680 QCD p T >170 25470125011250079885916232.2117.90.5610.0240 QCD p T >300 1256892.1892.0688.0629.777.739.30.1320.0010 QCD p T >470 87.9873.07 66.1165.0816.197.940.01600  ’s are in pb

10. India-CMS meeting, BARC 30 th July, 2010 10 Cumulative cross sections start  HLTPVbarrelp T  >50 MET>50  >2.7  ID TVJV MD1d2 0.5190.4590.4340.3170.2030.1920.1910.1880.149 Z ( )+  0.6730.5840.5740.3710.3340.3050.2720.2640.2130.208 Z(ll)+  7.32.682.651.510.4550.0280.0220.0210.0090.007 Z( )+j 450038.3638.3222.116.335.554.360.1750.1310.127 DYee 1300578.2571.6321.968.00.1030.0820.07000 W+  23.28.097.984.920.9190.2280.1700.1620.0830.079 W(e ) 789952985221302325331.216.212.40.660.63 W(  ) 789995.6195.1655.0510.355.063.500.3810.126 W(  ) 7899314.5311.8165.115.88.05.80.6110.3240.315  +j p T >50 3277.82641.32637.31523.61260.83.5723.1522.8531.1030.266 QCD p T >30 6.0 10 7 1.2 10 6 5.3 10 5 5.8 10 4 459.6192.8000 QCD p T >80 92380021941821939312165541113375.8203.72.960.680 QCD p T >170 25470125011250079885916232.2117.90.5610.0240 QCD p T >300 1256892.1892.0688.0629.777.739.30.1320.0010 QCD p T >470 87.9873.07 66.1165.0816.197.940.01600 TV rejects events with charge leptons like W’s JV on QCD QCD events usually have low MET and not isolated photons  ’s are in pb

11. India-CMS meeting, BARC 30 th July, 2010 11 Signal region 300 pb -1 100 GeV130 GeV150 GeV170 GeV ADD MD1d217.5118.36.4 Z( )+  9.53.92.41.4 Z( )+jets5.50.600 W+  2.60.90.30.2 W(e )11.43.41.1 W(  )4.6000 W(  )5.41.3  +jet 0.60.1 0 N S /√N B 2.83.43.63.1 photon p T /MET thresholds are raised up together photon p T /MET > 150 GeV optimizes the signal to background ratio

12. India-CMS meeting, BARC 30 th July, 2010 12 Z gamma Background Data driven estimation: This background was estimated using diphoton sample with ratio of 1:5 at 10 TeV. We need to repeat this study at 7TeV. But the official Zgamma sample is still awaited. We have made this request, production is complete, still to be transferred to T2. This has the high contribution towards the background. To understand this background one needs to understand in detail the electron fake rate, where we have mainly two contributions: 1. Photon from electron brem. 2. Track mis-reconstruction. Work Ongoing to understand these contributions. W background

13. India-CMS meeting, BARC 30 th July, 2010 13 Status of Common code In addition to the various physics objects in the code, now we are trying to put variuos modules together to bring the whole analysis in one single code. PF Met, TCMet, CaloMet – all are part of the basic code. Swiss cross, rook fraction, e1/e9, rechit timing, e2/e25, e2/e9 – All variables for spike cleaning being implemented in the code. For Cosmics - a basic framework is also added to the code from Shruti.

14. India-CMS meeting, BARC 30 th July, 2010 14 Final considerations This is the status of mono-photon analysis with CMSSW_3_1_4, already moved to CMSSW_3_5_7. All the reprocessed backgrounds are available accept Z(inv) . Still to be transferred at T2. Signal samples are being produced now – had problems with official Sherpa production, seems to be “sorted out”. Exotica conveners being informed and we expect samples to be there sometime in August. New features and plans: Photon ID new official prescription – Effect of pixel seed requirement is to be studied, considering the effect of early conversions) Optimization of track and jet veto (using real data as well) – to be done. Ongoing Study on W background contribution – noticed double counting issues. (working on them now – not covered here anything with 357) Jet and electron faking photons studies – Ongoing. data-driven method for Z(inv)  at 7TeV – to be done.  using diphoton process as in the 22X note  feedback and common studies with V+  group

15. India-CMS meeting, BARC 30 th July, 2010 15 Present goals we have: MC Once all the samples are there, on suggestion from Greg and Conor,a complete MC based study from a new point of view is to be done. Here I present the idea: Signal is REAL PHOTON + REAL MET. Categerize samples: 1. Real photon + Real met 2. Fake Photon + Real met 3. Real Photon + Fake met 4. Fake Photon + Fake met Study every single background and its effect to each category seperately avoiding carefully all the double counting, all sources of fakes in detail from MC study.

16. India-CMS meeting, BARC 30 th July, 2010 16 Data at 7TeV Biggest Challenge is Spikes for our study. Swiss Cross is the most common recommended method to remove the spikes. Exotica Hotline saw some monophoton looking events few days before which passed swiss cross, but were found to be double spike events later. Timing, e2/e9 and e2/e25 plays an important role to remove these events. Reference: 'Monophoton events in Exotica Hotline' http://indico.cern.ch/conferenceDisplay.py?confId=100949 To be “worked on”: what can monophoton analysis achieve with early data? Say at 1pb-1?

17. India-CMS meeting, BARC 30 th July, 2010 17 Candidate Monophoton event – passing swiss cross.

18. India-CMS meeting, BARC 30 th July, 2010 Timing of second highest vs highest energy crystal Single Photon events in Data MC (Z(vv) gamma sample)

19. India-CMS meeting, BARC 30 th July, 2010 Z(vv) gamma sample With No pt cut on photon and pt >100 GeV

20. India-CMS meeting, BARC 30 th July, 2010 Isolation Variables

21. India-CMS meeting, BARC 30 th July, 2010 Isolation Variables vs # of tracks For the complete study, please see http://cmsdoc.cern.ch/~sandhya/Preliminary_plots/

22. India-CMS meeting, BARC 30 th July, 2010 Skimming work First Test EG skim based on HLT paths of photon published at T2_IT_Rome. /MinimumBias/sandhya-SkimTest1-5113a06033395584a6f8633cb59b003b/USER Various efficiencies were calculated before to decide on which HLT path could be taken and if we can put cuts on Isolation variables while making the skim itself. Everything is maintained here: https://twiki.cern.ch/twiki/bin/view/CMS/EGammaSkimming

23. India-CMS meeting, BARC 30 th July, 2010 23 Present Goals we have: Data 1. To keep an eye on Exotica hotline to study any signal looking like event. 2. Planning to make a study of different tunes with the data. We already have a framework set with the data at ~ 10 nb-1. Planning to have an Analysis note with this study in August/September with more data collected soon. 3. Data Driven estimation depends on how soon we have sufficient data to work on them.

24. India-CMS meeting, BARC 30 th July, 2010 24 Back up slides

25. India-CMS meeting, BARC 30 th July, 2010 25 Leading photon spectrum A signal region must be defined using both photon p T and MET Real photons from W processes - hard brem

26. India-CMS meeting, BARC 30 th July, 2010 26 Non Beam Backgrounds

27. India-CMS meeting, BARC 30 th July, 2010 27 QCD background with looser cuts No HLT & Pt cuts here. Tried to extrapolate with exponential didn't work, as exponential falls faster than the actual curve. No QCD events are left if we apply selection criteria

28. India-CMS meeting, BARC 30 th July, 2010 28 Identification of photons from Cosmics Using three discriminators  Matching photon candidates with the cosmic muon by propagating muon trajectory through the ECAL  Timing Information  Shower Shape Information For Photon MC data : /PhotonJet_Pt15/Spring10-START3X_V26_S09-v1/GEN-SIM-RECO For Photon Skim data : /EG/Run2010A-PromptReco-v2/RECO For Photon Skim Data, considered single photon trigger requiring at least one HLT photon with E T >20 GeV. Based upon CRAFT09 Data analysis, photon candidate is matched with muon in the limit of σ R <20. 2828

29. India-CMS meeting, BARC 30 th July, 2010 29 Shower Shape variables from CRAFT09 and Prompt PhotonMC 2929 Clear Discriminators

30. India-CMS meeting, BARC 30 th July, 2010 30 Distribution of Shower variables Photon Isolation criteria: (Tight cut, JSON file of good run list ~ 78nb-1) https://twiki.cern.ch/twiki/bin/viewauth/CMS/WorkBook?topic=PhotonID Comparison of unmatched photon(  R >20) candidates with Photon MC (Barrel photons within timing window of 5ns) 3030 2D Plot

31. India-CMS meeting, BARC 30 th July, 2010 31 Summary from Non Beam backgrounds Cosmics: We use Fisher discriminant technique to separate signal from background. Seprates 85% background retaining 95% signal ( from CRAFT09 analysis) From our earlier study, we expect 11 muons in a year corresponding to 2808 bunches in a second and 2 muons with 432 bunches in a year from CRAFT08 data( pointing muons above 150 GeV) (Refer AN-2009/191 and backup slides) Beam Halo : Talked by Tia in Vgamma group http://indico.cern.ch/conferenceDisplay.py?confId=99804 Now, we are working on integration every module to the common piece of Monophoton code. Right now, cosmics part is being worked on and then we go to Beam Halo.

32. India-CMS meeting, BARC 30 th July, 2010 32 Background datasets – 31X samples SampleCross section (pb)# events /PYTHIA6_ZInvgamma_7TeV_1E31_RECO0.673110k /ZinvisibleJets-madgraph/Summer09-MC_31X_V3_7TeV-v2/GEN-SIM-RECO45002.2M /Wgamma/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO23.2100k /Zgamma/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO7.3100k /Zee/Summer09-MC_31X_V3_7TeV_TrackingParticles-v1/GEN-SIM-RECO13002.5M /Wenu/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO78992M /Wmunu/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO78992M /Wtaunu/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO78992M /PhotonJet_Pt50to80/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /PhotonJet_Pt80to120/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /PhotonJet_Pt120to170/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /PhotonJet_Pt170to300/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /PhotonJet_Pt300to500/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /PhotonJet_Pt500toInf/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO 2723 446.2 84.43 22.55 1.545 0.0923 100k 110k 120k 1.0M 1.1M /QCD_Pt80/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /QCD_Pt170/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /QCD_Pt300/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO /QCD_Pt470/Summer09-MC_31X_V3_7TeV-v1/GEN-SIM-RECO 923800 25470 1256 87.98 3.2M 3.1M 3.3M 2.1M /DiPhotonBorn_Pt25to250/Summer09-MC_31X_V3_7TeV_TrackingParticles-v1/GEN-SIM-RECO /DiPhotonBorn_Pt250toInf/Summer09-MC_31X_V3_7TeV_TrackingParticles-v1/GEN-SIM-RECO 22.37 8.072 10 -03 747k 210k /DiPhotonBox_Pt25to250/Summer09-MC_31X_V3_7TeV_TrackingParticles-v1/GEN-SIM-RECO /DiPhotonBox_Pt250toInf/Summer09-MC_31X_V3_7TeV_TrackingParticles-v1/GEN-SIM-RECO 12.37 2.08 10 -04 539k 216k

33. India-CMS meeting, BARC 30 th July, 2010 33 Spring10 Datatsets – 35X samples Signal samples – Not ready, ZGToNuNu - done but “not” yet at T2

34. India-CMS meeting, BARC 30 th July, 2010 34 Non beam background - Cosmics This shows the energy deposit by superpointing muons in ECAL, which goes till 700 GeV. Though the statistics is less, but we need to know the rate of these cosmic muons so that once we have data over some interval of time, then we know how many such events we expect from the data. Only “Barrel” Muons

35. India-CMS meeting, BARC 30 th July, 2010 35 Cosmic rate calculation 1 second of collision means 4 X 10 7 collisions. For a window of 300 ps around the collision, cosmics will be accepted for 2 X 300 X 4 X 10 7 ps = 2.4 X 10 10 ps = 2.4 X10 -2 s If the rate of Cosmics is R Hz, then total cosmics collected in 1s for a window of 300ps around the collision = R X 2.4 X10 -2. Then further from this timing window, total cosmics faking photons =  RX2.4 X10 -2 ), where  is the efficiency of cosmics which passes the selection criteria.  R = 0.1414 X 10 -4 Hz ( for pointing muons above 150GeV). This means we will have 11 muons in a year corresponding to 2808 bunches in a second and 2 muons with 432 bunches in a year.

36. India-CMS meeting, BARC 30 th July, 2010 36 CMSSW : 3_5_7 Samples: /Wenu/Spring10-START3X_V26_S09-v1/GEN-SIM-RECO /Wgamma/Spring10-START3X_V26_S09-v1/GEN-SIM-RECO The goal is to move to tighter selection and see the effect of adding hasPixelseed requirement in the selection for these samples and understand the events after selection.

37. India-CMS meeting, BARC 30 th July, 2010 37 The event in Wenu after the selection All the 5 events are where its a high pt ISR photon from one of the quarks.

38. India-CMS meeting, BARC 30 th July, 2010 38 MET recontruction impact - I MET - photon pT

39. India-CMS meeting, BARC 30 th July, 2010 39 MET recontruction impact -II MET photon azimuthal separation

40. India-CMS meeting, BARC 30 th July, 2010 40 Cross section table after all cuts These are 5 events Signal Window: requiring a photon and met between 100 to 500 GeV.

41. India-CMS meeting, BARC 30 th July, 2010 41  x,  y and  z and  R from CRAFT09

42. India-CMS meeting, BARC 30 th July, 2010 42 Timing from real data from unmatched photons

43. India-CMS meeting, BARC 30 th July, 2010 43 Jet faking photon rate NO. NAME VALUE ERROR 1 p0 4.11481e-02 1.86561e-03 2 p1 -1.16008e+00 4.87904e-02 3 p2 -2.25920e-02 1.34382e-03 Fit function: par[0] + exp(par[1] + par[2]*x[0]) fake rate = N  tight / N  loose Common selection: HLT photon p T > 30 GeV | photon  | < 1.5 Calculated on QCD p T >30, 80, 170 GeV analysis ID H/E<0.5 Jet EM fraction can fake photon Another way: N T : an. ID N L : ISO Tk >9 (opposite cut)