1 / 48

DiPhoton + MET: Towards Unblinding of the 5 fb -1 Analysis

In the quest for unblinding the analysis, this study by Bruce Schumm on di-photon events and missing transverse energy (MET) introduces the A, B, C analysis strategy for optimization. The analysis includes thumbnail sketches, background estimates, and a request for unblinding the A and B analyses. Significant improvements like reverse electron-photon overlap criteria, conversion pixel requirements, and conversion categories are discussed. The study focuses on optimizing the signal selection criteria and cross-section sensitivity to sparticle masses. Several optimization strategies were suggested based on photon ET thresholds and analysis objectives. The challenges and solutions related to MET issues were examined, with discussions on CALO and energy calibration methods. The glossary explains essential terms like LocHadTopo, MetRefFinal, and Simplified MetRefFinal used in the study. The background estimates, pseudo-photon control samples, and signal blinding concerns for MET >100 GeV are elaborated. The study also delves into control samples like QCDg, QCDgg, and gg for background estimation. Towards enhancing the analysis, the study advocates for utilizing LocHadTopo and improving the MET_LocHadTopo calibration strategy for better results.

darrellm
Download Presentation

DiPhoton + MET: Towards Unblinding of the 5 fb -1 Analysis

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. DiPhoton + MET: Towards Unblinding of the 5 fb-1 Analysis Bruce Schumm 20 Feb 2012 • Reminder about 1 fb-1 analysis • New strategy: A B C analyses • Optimization • MET Blues • Background estimates • Request for unblinding of A, B analyses only

  2. 1 fb-1 Analysis: Thumbnail Sketch • (First-order) signal selection straightforward: •  2 tight isolated photons with ET  25 GeV • ETmiss  125 GeV • Optimization based only on ETmiss cut value • Optimization geared towards high-mass Gluino for broad range of bino masses (50 GeV to Mgluino)

  3. Also: Limit of  > 145 TeV set on SPS8 SUSY Breaking scale Analysis background-limited; sparticle cross section goes as ~M-9  reoptimize!

  4. Analysis Improvements Overlap Criteria - Reverse /e overlap criterion (have medium e kill ) - e  fake rate goes from 0.049-0.168 to 0.025-0.075 - 13% signal efficiency loss Conversion pixel requirements - Conversion tracks can have no pixel hits - e  backgrounds reduced by 43% (42% for events with one conversion; 61% for events with two conversions) - 13% signal efficiency loss Conversion categories - Division into three conversion categories suggested - No benefit if background is ~0

  5. Optimization: General Principles • Two scales characterize GMSB production/decay • Missing energy (mostly bino mass; also bino boost)  MET • Total energy, including photons (sparticle mass)  HT • Analysis A: High-mass sparticle, high-mass bino • Large MET, moderate HT • Analysis B: High-mass sparticle, low-mass bino • Moderate MET, large HT • Analysis C: SPS8 (direct gaugino production; sparticles too heavy) • Moderate-to-large MET; NO HT (most like 1 fb-1 analysis) • Also: 1 fb-1 backgrounds observed to have photons close to MET •   (photon to MET) cut explored for A,B,C analsyes

  6. Photon Et Optimization (1 fb-1 Analysis) Helenka Choose Cut of 50/50

  7. Full Optimization of A, B, C Points Dan • Optimization strategies: • All use photon ET > 50 GeV. • A: ETmiss, HT, and  • Use GGM (Mg,MB) = (900,800) • B: ETmiss, HT (signal  too small) • Use GGM (900,50) • C: ETmiss,  (low mass scale in production so no HT) • Use SPS8; =170 TeV

  8. C

  9. Optimization of A, B, C Points: Results Optimization results: All use photon ET > 50 GeV. Figure of merit rather flat in   use either 0.5 or no cut Drop unjustified significant digits

  10. MET Issues Martin, Dan, Helenka • MET_LocHadTopo • Local Hadronic calibration applied to all topo clusters • MET_RefFinal • Based off of objects • Local Hadronic calibration used for jet objects • MET_SimplifiedRefFinal • Based off of objects (Tight photons w/ pT>20 GeV used) • EMJES calibration used for jet object • Modified version of SimpRefFinal • Based off of objects (Loose photons w/ pT>20 GeV used) • EMJES calibration used for jet objects • Doesn’t exist on D3PD • ETmiss Glossary: • LocHadTopo: Used for prior diphoton+MET analyses; not directly ATLAS-supported • Simplified MetRefFinal: SUSY-group recommended (reconstruct ETmiss from separately-treated objects) • MetRefFinal: ATLAS (but non SUSY) supported; available on SUSY D3PDs though. (Use jets with local hadronic calibration) _ Not usable for 5fb-1 analysis; explore for future.

  11. SimpMetRefFinal: Initial “Photon Fix”

  12. Sample used: +jet MC

  13. Background: Definition of “QCD” Control Samples We model the ETmiss distribution of selected events with no intrinsic missing energy via two control samples: QCDg and QCDgg. A “control photon” satisfies the loose, but not the tight, selection requirement for two shower quantities: the shower shape in the shower core {fracs1} and the shower width {weta1} in the first sampling of the electromagnetic calorimeter. QCDg has one such control photon; QCDgg has two.

  14. LocHadTopo

  15. Simplified MetRefFinal !! Must also correct loose photons!

  16. QCDg, QCDgg, and gg – MET_SimpRefFinal No resemblance of control sample eTmiss shape to that of the signal 2/16/2012 20

  17. QCDg, QCDgg, and gg – MET_LocHadTopo N.B.: Studies with +Njets MC suggest high-ETmiss tails a bit larger for SimpMetRefFinal  Propose to use LocHadTopo for ETmiss 2/16/2012 21

  18. BACKGROUNDS E- Control Sample (Penn) QCDg Control Sample (DESY) Overlap ? (DESY) Missed? (SCIPP)

  19. EW Background from e- control sample Brig, Jack N.B.: “Signal” is e-

  20. Scale factors from Ze/Zee Multiply these by “signal” numbers on previous page

  21. Pseudo-photon control sample (Peter, Martin) SIGNAL BLINDED FOR MET > 100 GEV • QCDg distribution provides shape and tails • scale to signal (gamma-gamma) in low-MET region (ETmiss<20 GeV) SIGNAL BLINDED FOR MET > 100 GEV SIGNAL BLINDED FOR MET > 100 GEV

  22. Cross-check with MET distribution from Zee (background real ?)

  23. No QCDg above 100 GeV  Less than 1 event at 95% CL

  24. Overlap Between QCDg and EW backgrounds Martin

  25. Missed Backgrounds (?) Dan • Since our “QCD” backgrounds are estimated by normalizing control samples (QCDg, Zee) to low-MET signal, they should be comprehensively accounted for • “EW” (W,top) backgrounds estimated via e- control sample  Assumes all W,top contributions have at least one e fake • Is this true? If not, what is character of the “missed” component? • Might the “missed” component in fact be incorporated into the QCD control sample (pseudo-photon) estimate?

  26. According to MC, what fraction of EW background is due to e fakes? MC

  27. Of the 25% that is “missed” 18.9% + 47.5% = 66.4%  2/3 is expected to be reflected in the pseudo-photon sample This 2/3 may well be the source of the “EW-contamination” in the pseudo-photon sample (cross-check underway This component is neither missed nor double-counted!  Add “QCD” and “EW” backgrounds linearly (values and errors)

  28. Background Summary *Includes a 0.2 event contribution from “irreducible” backgrounds (two real photons) that are negligible in A and B regions; needs to be check for C

  29. A word on the 5 fb-1 reach (not fully-optimal analyses) A: ggm_900_800   -> LL: 142.4, sig: 16.9, #S: 21.3, #B 0.01 ggm_1000_800 -> LL: 39.8, sig: 8.9, #S: 7.1, #B 0.01 ggm_1100_800 -> LL: 10.8, sig: 4.7, #S: 2.4, #B 0.01 ggm_1200_800 -> LL: 3.4, sig: 2.6, #S: 0.9, #B 0.01 C: sps8_170 -> LL: 22.2, sig: 6.7, #S: 11.9, #B 0.9 sps8_180 -> LL: 14.5, sig: 5.4, #S: 8.9, #B 0.9 sps8_200 -> LL: 6.6, sig: 3.6, #S: 5.3, #B 0.9 sps8_220 -> LL: 2.7, sig: 2.3, #S: 3.0, #B 0.9 sps8_240 -> LL: 1.2, sig: 1.5, #S: 1.8, #B 0.9 B: ggm_900_50   -> LL: 45.1, sig: 9.5, #S: 10.3, #B 0.05 ggm_1000_50 -> LL: 13.6, sig: 5.2, #S: 4.0, #B 0.05 ggm_1100_50 -> LL: 2.7, sig: 2.3, #S: 1.2, #B 0.05 ggm_1200_50 -> LL: 0.4, sig: 0.85, #S: 0.3, #B 0.05 Limits for 1 fb-1 about 820 GeV for analysis-A-like scenario and ~145 TeV for SPS8 trajectory

  30. Fire and Brimstone • Verbatim 2011 analysis  150-200 GeV increase in limits • 2012 analysis improvements  250-400 GeV increase • CMS did better than ATLAS with 1 fb-1 because of • More favorable interpretation of GGM model • Luck • but per fb-1, our analysis was more sensitive • Dark matter, light Higgs(?)  SUSY (??) • Possibility never greater; if something is there, don’t want to miss it! • Critical to unblind right away.

  31. BACKUP

  32. GamNp1 diphoton sel LocHadTopo One extra event for MET>100GeV MeanRefFinal  MeanLocHadTopo RefFinalSimpSUSY Diphoton+MET meeting - 09.02.2012 - Helenka Przysiezniak / LAPP 42

  33. GamNp2 diphoton sel LocHadTopo One extra event for MET>90GeV MeanRefFinal  MeanLocHadTopo RefFinalSimpSUSY Diphoton+MET meeting - 09.02.2012 - Helenka Przysiezniak / LAPP 43

  34. GamNp3 diphoton sel LocHadTopo One extra event for MET110 GeV MeanRefFinal  MeanLocHadTopo RefFinalSimpSUSY Diphoton+MET meeting - 09.02.2012 - Helenka Przysiezniak / LAPP 44

  35. GamNp4 diphoton sel LocHadTopo MeanLocHadTopo  MeanRefFinal Both have event for MET200GeV RefFinalSimpSUSY Diphoton+MET meeting - 09.02.2012 - Helenka Przysiezniak / LAPP 45

  36. GamNp5 diphoton sel MeanLocHadTopo  MeanRefFinal LocHadTopo Few extra events for MET80GeV RefFinalSimpSUSY Diphoton+MET meeting - 09.02.2012 - Helenka Przysiezniak / LAPP 46

  37. QCDg, QCDgg, and gg – MET_RefFinal 2/16/2012 47

More Related