1 Preparing for First Data at LHC Wing To UCSB Physics Advancement Talk 11 September, 2008

Embed Size (px)

DESCRIPTION

3 New Collider and Detector The Large Hadron Collider ( Physicists’ New Toy ) Proton Proton Collider at Center of Mass Energy (CME) 14 TeV High CME leads to much higher production cross section due to PDF. σ (gg → 2TeV < fb. σ (gg → 14TeV ~ pb. High Luminosity, Start at peak L cm -2 s -1. Upto peak L cm -2 s -1. L(Tevatron) Peak ~ cm -2 s -1. Ideally can produce Start at ~ 1 pb -1 per day. Upto 100s pb -1 per day.

Citation preview

1 Preparing for First Data at LHC Wing To UCSB Physics Advancement Talk 11 September, 2008 2 Introduction: the need of Calibration, and the types of analysis that can be done in first run. Calibrating and Validation Electron Identification using Photon Conversions. A SUSY search analysis. Superpartner of the top (stop) search. Conclude with some notes and future plans. Outline of the talk 3 New Collider and Detector The Large Hadron Collider ( Physicists New Toy ) Proton Proton Collider at Center of Mass Energy (CME) 14 TeV High CME leads to much higher production cross section due to PDF. (gg 2TeV < fb. (gg 14TeV ~ pb. High Luminosity, Start at peak L cm -2 s -1. Upto peak L cm -2 s -1. L(Tevatron) Peak ~ cm -2 s -1. Ideally can produce Start at ~ 1 pb -1 per day. Upto 100s pb -1 per day. 4 CMS Inner Tracking System All Silicon Ionization Detectors. 13 layers in the barrel region, 9 disks in the endcap region. Inner most part are pixel while two outer parts are only strips along Z direction. EM Calorimeter Lead Tungsten. 22cm long (25 X 0 ) Electrons and Photons interacts with EM field of the nucleus of the Pb and W. Interaction causing a shower of photons and electrons that can be detected. Hadron Calorimeter Brass Absorber Russian artillery shells. 15 layers, each 5cm thick. Scintillation tiles to measure the hadrons energies. Muon Chambers 250 Drift Tubes (Ar & CO 2 ). 540 Cathode Strips Chambers. 610 Resistive Plate Chambers. 5 Calibration New new energy and new luminosity. Every component requires calibration and validation with data. One of them is the Electron Identification. Electrons are identified by they track it creates and the energy they deposite on the ECal. Electron ID variables EM Energy / Track P Had Energy / EM Energy Distance between the Track and ECal Cluster. 6 New Physics in First Data Challenges in First Data Low integrated luminosity, ~ 1 pb -1 day -1 Uncertaintities on measurements are high. E jet scale, MET, e identifical. Begin searches in high production and distinct signature. The two biggest new physics signals are either SUSY or gg H SM. H SM is hard to find because it decays to bbar and bbar has a large production cross section. SUSY with low mass parameters could be found within an integrated luminosity as low as 100 pb -1. 7 Calibration of Electron ID Method How to find them? Conversions Found? Single Photon Sample MinBias Sample QCD Jets Sample Predicting fake electrons. Look at the Electron ID variables 8 Calibration of Electron ID In start-up data we need to validate Electron ID variables. emE/TrkP and hadE/EE. (track, EBC) and (track, EBC). Need a sample of electrons from first data. Zee, pure but small production rate due to b cross section and small branching ratio. Photons are in every event from -zero decays. Find Photon Conversions with only Tracker variables. These conversion electrons can make an independent calibration and validation of Electron ID Variables. EBC Ecal Basic Clusters 9 Method: Finding Conversion with Tracks To find conversions in the pixel part of the tracker we use CTF Tracks seeded by first 2 layers of TIB (more material more conversions). Electrons from conversion have: small , cot(), and z0 displaced vertex so d0 will be non-zero. sum of the 3 Momenta of two tracks points away IP. At = 0. The tracks are right on top of each other. Displaced Vertex of ~ 20 cm in Y. Conversion occurred in TIB1. Fig. 9a&b: Conversion Candidate Tracks cm 10 Method: cuts used to find conversions Fig 10a: Single Photon Events (Blue) have small cot() compare to MinBias Evts (Red) Fig 10b: Conversion occurs off IP. The d0 will be non-zero for conversion electrons. By convention d0*charge is always positive for conversion electrons. Photons Events MinBias Events MinBias Events Photons Events Normalized d0*charge distribution of Tracks (cm) Normalized cot() Distribution of Tracks Find appropriate cuts variables using 10000, 30GeV Single Photons and compare the cut variables to 1M MinBias Events. 11 Fig 11a: Distance between two tracks at their minimal approach (TX). is arbitrary for two tracks since track_ changes. Use a derived variable Track Cross distance (TX). Method: Cuts used to find conversions TightLoose cot() d0*charge0.03 (cm) TX0.180 (cm)0.360 (cm) z (cm)0.520 (cm) P2B.0224 (cm).0448 (cm) Normalized Track Cross (TX) distribution of Tracks (cm) MinBias Events Photons Events Table 11b: Cuts on two tracks used to find conversions. Track parameter z0 is not well measured due to pixel less tracking Single Photons Events with Pt 2-30 GeV Every Track in event is an electron ECal Filter: 2 EBC of 5 GeV Conversions found. 606 tight conversions, 561 (92%) matched to SimTracks. 902 loose conversions, 798 (88%) matched to SimTracks. Unrealistic but it can be used to check our other results. Simulated Single Photons Match conversion tracks to SimTracks with the same Pt, and at vertex. Scroll between Slide 12 and 13 to see match-up in XY-plane cm/cm 13 Simulated Single Photons cm/cm 10 5 Single Photons Events with Pt 2-30 GeV Every Track in event is an electron ECal Filter: 2 EBC of 5 GeV Conversions found. 606 tight conversions, 561 (92%) matched to SimTracks. 902 loose conversions, 798 (88%) matched to SimTracks. Unrealistic but it can be used to check our other results. Match conversion tracks to SimTracks with the same Pt, and at vertex. Scroll between Slide 12 and 13 to see match-up in XY-plane 14 Simulated MinBias Lets try to find Conversions in MinBias Sample + Similar to first data at the LHC. + Fewer tracks per event. -Tracks are generally soft, most tracks dont reach ECal. -Low pt tracks scatters larger % of their momentum. -Needs to run large number of Events, so we use an EBC filter 2 x 2GeV Tight cuts for higher purity + Found only 12 conversions out of 10 6 MinBias Evts. + 7 (58%) Match both SimTrks. Loose cuts for higher statistics + Found only 24 conversions out of 10 6 MinBias Evts (50%) Match both SimTrks. Need to Simulate 100M MinBias to get good statistics at this rate. Need sample with more photons. 15 Simulated QCD Jets QCD Jet Events, 10 5 with Pt 50+ GeV More photons per event. -More & K can fake electron tracks. -Photons made from 0 will be inside Jets. -QCD Jet Conversions tight conversions,1265 (84%) matched to SimTracks loose conversions, 1752 (70%) matched to SimTracks is enough to start looking at Electron ID variables. 16 Results in QCD Jets Sample Fig 16a: Distribution of R in XY plane (Rxy) for conversion candidates. Conversion Rxy in Jet Events (cm) First Lets look at something simple. R in the XY-plane where the conversion occur. Structures seen ~ 5cm, 8cm, 11cm are each layer of Pixel Tracker. The first Layer of TIB is found at ~22cm 17 Results in QCD Jets Sample We need to connect the sample of electrons to Electron ID variables. First propagate the track to the radius of the ECal. Then search for the nearest EBC. The nearest EBC is sharply peaked around zero with small tail. E/P is peaked at 1 as expected for electrons. However left side has a large shoulder. These are not pure electrons. We need a method to remove the background from plots. Fig 17a: (track, EBC) for conversion in Jets Fig 17b: EM Energy / Trk Outer P for conversion in Jets 18 Background Prediction Predict the background by extrapolating from background dominate region into signal region. Use Cot() for example. Fig. 18a shows Cot() of all tracks. The blue is the signal region where the red is the background region. Zooming into the background region, we can see that the background is linear. Then we can simply extrapolate a straight line into the signal region. Fig. 18a: Cot() between 2 tracksFig. 18b: Cot() between 2 tracks zoomed 19 QCD Jet Sample with Bkg Subtraction Fig 19 a & b Background subtraction removed most of the conversion that occur at R < 5 cm and between 15 to 22 cm. Fig 19 c & d The left shoulder was also removed by the background prediction 20 QCD Jet Sample with Bkg Subtraction Fig 20 a & b: HadE/EE tail was reduced but still exist Fig 20 c & d: Most of the side band were remove by background subtraction in both and distributions. Fig 20 e & f: has a const cut at 0.05 which is seen here on the left. 21 Compare to Single Photon Events Fig. 21c: H/E Jet with bkg sub & photon events. Fig. 21a: ConvR Jets bkg_sub & photon events.Fig. 21b: E/P Jets with bkg_sub & photon events. Photon events have lower number of entries and histograms are rescaled accordingly. H/E still has a tail after Bkg subtraction. Conversion occurs inside Jets. Hadrons are sometimes right on top of electrons. Granularity of HCal ~ 4x of ECal. Overlap to nearby hadrons is more likely. ConversionR Photons 22 Compare to Single Photon Events Fig. 22a: (track, EBC) Jets with background subtraction & photon events. Fig. 22b: (track, EBC) Jets with background subtraction & photon events. The last two plots are for Delta Phi and Delta Eta. They each match fairly well in the core. But the Jets have more side band than the Photons just as like the other Electron ID variables. 23 The SUSY model used. The Signature. Find it with Jet Counting. Background prediction and reduction. Some MET and Total Energy Search. SUSY Outline 24 A SUSY Example: Stops Search Why Light Stops? SUSY predicts a left and a right type squark for every quark. Experiments will only produce a mixture of left and right type squark. The mixing is found by the following mixing matrix. The large mass of the top will cause the stops mass eigenstates to be highly mixed of the left and right type stops. We can parameterize this with a mixing angle theta () and find the mass eigenstate to be, 25 Gluino Stop Signature Both stop decay channels will give us 2 bs and 2 Ws. Looks like 4 tops. W decays to e or 20% of the time. 2 Jets at about 70% of the time. 10% to tau* which will decay ~1 jet. LSP will create addition Missing Et Large masses of gluinos requires SumEt to be large also. Leptons can come out with same sign. #Lep (e,)# Jet Min*# Jet Max* 0 (40%) (60%) (18%)68 3+ (2.7%)56 4 (0.16%)44 Table 25A: Leptons and Jets Counts for SUSY Decay. Fig 25A: Gluino Decay Chain Through stops. g 26 Jets Counting Method Using pythia to generate gg ~g~g and ffbar ~g~g. CME at 14TeV then rescale them to 1 fb -1 for comparison. Total production cross section is 6.4 pb, so well have 6400 events in 1 fb -1. If we require 1 muon/electron we lose about 70% of the events and almost all if we require two leptons. What are the expected background? 27 Jets Counting Background QCD and ttbar events are from Claudios ntuple sample, everything is normalized to 1 fb-1 for comparison From this plot, the best chance is at the 10 or 11 jets channel. The signal is still factor of 4-5 smaller than the background. To beat these background we should require leptons in the event. QCD Jets should have zero lepton ttbar will have two Ws which leads to 36% chance for 1+ lepton. Our signal has four Ws so it should have 59% chance for 1+ lepton. 28 Isolating and counting leptons Lepton Counting: 1 lepton, 2 leptons... should be easy right? CMS was designed to detect muons and electrons. But not every lepton comes from our initial hard scattering. Leptons can be made from secondary decays such as photon conversions. Need to find primary leptons that came from decay of tops. top -> Wb -> l b. Primary tracks will have small d0 and many valid hits. Isolation from other particles: track and energy Iso. For muon: good global fit between the tracker and muon chambers. For Electron: Low hadEt/emEt, emEt/TrkP ~ 1. 29 The Search for the Good Lepton Only go through isolation variable, the process is the same for electrons. Use QCD Jets to find bad leptons since ttbar is contaminated with good leptons. Then compare them Z->mumu for a pure muons sample. Fig 28A: d0 distribution Z->mumu has a large peak near zero, a cut at 0.01cm keeps 99% of signal. Rejects 36% background. Fig 28B: Iso03_sumPt/Pt distribution Again a cut at 0.1 keeps 98% of signal. Rejects 87% background. 30 The Search for the Good Muons VariableCutSignalBackground mu_d00.01cm mu_isoPt/Pt mu_isoemEt/Pt mu_isohadEt/Pt mu_gchi/ndof mu_#validhit TotalAll Table 29A: Muon isolation cuts used. 31 The Search for the Good Electrons VariableCutSignalBackground el_d00.01cm el_hOverE el_EoverPin0.7 dEtaOut dPhiOut el_#validhit el_tkIso/Pt el_tq_caloIso/Pt TotalAll Table 30A: Electron isolation cuts used. 32 Background with Good Leptons SignaturesSUSY signalttbar bkg 1 muon + 6 jets472 * = muon + 7 Jets396 * = muon +8 Jets218 * 0.82= muons + 5 jets60.8 * = muons + 6 Jets51.8 * = 5.25 SignaturesSUSY signalQCD bkg 1 muon + 6 jets472 * 0.862= electron + 6 Jets483 * = 125 Muons Isolation reduced QCD background. Electron also reduced QCD but costs half our signal. Stick to only muons for now. ttbar background is not reduced as much due to real leptons from top decays. Fig 31B: ttbar background nJets Fig 31A: QCD background nJets The best sig/bkg channels are the 1 muon 7+ jets or 2 muons 6+ jets channels. 33 Looking through ttbar The data will be a superposition of ttbar and SUSY in all the search channels. Need a way to tell the difference between ttbar and SUSY. We can look for kinetimatic regions where the background dominates the signal. SUSY events are usually central, due to large energy required to produce the gluinos. Look for forward events to predict the background by requiring the lepton to have high eta. 34 Looking through ttbar The plot of the right is the background subtracted nJet distribution. The prediction seems to be shifted to lower nJets. The lower right plot shows the difference betweeen nJet counts in the forward region compared to the central region. Low nJet bins are over estimated, while the high nJet bins are under estimated. This could be due to the difference between the central and forward region of the detector. We need a way to predict this effect. 35 Looking through ttbar To predict the difference between the two regions. Data driven background prediction. Top right plot shows the Rnj distribution of ttbar and SUSY. Then we can the predicted background in the central region, The nJet bins does not match well, ttbar maybe too forward to be predicted this way. 36 Same Sign 2 Lepton Channel If we use the 2 muons channel and also require them to be the same sign. SM same sign leptons are very small. The signal will jumps out on top of the the background in this case. 37 MET and SumET Bump Search There are other signatures for stop decays besides nJet distributions. One of the key SUSY signature is Missing Et. The plot on the right shows MET for signal (blue), ttbar (red) and their sum, sig+bkg, (black) in 1 muon + 6 jets events. MET for SUSY has a broad distribution. Neutrinos and LSP will contribute to MET, but their vector sum is randomized. However, the Sum_Et is not affected by randomization. We can play the same game to find the signal in SumEt with 1mu+6jets. 38 SumET Background Prediction Plot of the right shows the Predicted Signal (lavendar), and compare to the signal at generator level (blue). The predicted signal will not be seen at this level. Pulling the same trick as in the nJet Counting with same sign lepton. The ttbar background is reduced significantly so the signal could be seen. 39 Conclusion on SUSY Analysis Discovery is most likely through same sign lepton channel. ttbar seems to cause a great deal of trouble as a background. Analysis requires more refining. Gluino mass can be > 600 GeV will go down. Electron channels needs to be improved. Loosen cuts and use the crossed out variables to get better signal to background ratio. b-tagging tool could be useful once b-tagging has been validated. 40 Notes on Electron Identification Tracker Only Conversion Finder finds conversions in Single Photons, MinBias and Jet Events. Background prediction gives good fake subtraction in large data sample. Startup Data ~ 100 Hz ( 8M evts / day ) ~ 100 Conv/day. Real Data will have EGamma HLT Triggers First collision occurs on ____ / ____/2008. Find conversions within a few days afterward. Conversion finder needs to be tested in data. Photon conversions can also be used to map the density of the tracker material to make reconstructions and simulations better. 41 Conclusion and Plan SUSY