1 / 64

Approval Talk for EWK-10-011 (part I): Measurement of Forward-Backward Asymmetry of Lepton Pairs

Approval Talk for EWK-10-011 (part I): Measurement of Forward-Backward Asymmetry of Lepton Pairs.

iren
Download Presentation

Approval Talk for EWK-10-011 (part I): Measurement of Forward-Backward Asymmetry of Lepton Pairs

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. Approval Talk for EWK-10-011 (part I):Measurement of Forward-Backward Asymmetry of Lepton Pairs N. Akchurin, G. Alves, A. Bodek, A. Bonato, Y. Chung, J. Damgov, P. de Barbaro, D. Green, A. Gritsan, Z. Guo, J. Han, S.W. Lee, K. Mishra, R. Rodrigues, Y. Roh, S. Tkaczyk, N. Tran, D. Vishnevskiy, E. Yazgan March 7, 2010

  2. PAS: EWK-10-011 (Part II)Measurement of the Weak-mixing Angle at CMS Part II to start on slide 27 Alessio Bonato, Andrei Gritsan, Zijin Guo, Nhan Tran Johns Hopkins University Efe Yazgan Texas Tech University on behalf of the EWK dilepton group March 7, 2011 2

  3. CADI: http://cms.cern.ch/iCMS/analysisadmin/cadi?ancode=EWK-10-010 • Supporting notes: 2010/455 (AFB in µµ), 2011/025 (AFB in ee), AN 2011/031 (Weinberg angle) • Twiki: https://twiki.cern.ch/twiki/bin/viewauth/CMS/ForwardBackwardAsymmetryOfDiLeptonPairs ARC members: Marco Dallavalle (Chair), Peter Timothy Cox, Marcello Mannelli, Claude Charlot, Kristian Allan Hahn

  4. Outline • AFB • Introduction and Motivation • Analysis Steps • Samples • Event Selection • Backgrounds • Results • Uncorrected Forward-Backward Asymmetry (AFB) • AFB with corrections (Unfolding, acceptance corrections and quark-direction corrections) • Combined (ee+µµ) AFB • Conclusions

  5. Introduction and Motivation • Forward-backward asymmetry • Fundamental SM measurement at √s = 7 TeV • First measurement in qqbarZ/g*µµ • Deviations from the SM prediction may indicate the existence of new neutral gauge bosons, quark-lepton compositeness, existence of SUSY particles or extra dimensions, … References to past work from D0, CDF, and CMS MC and data are given in the references in the PAS and in the notes.

  6. Forward-Backward Asymmetry in Di-Lepton events both vector and axial-vector couplings of electroweak bosons are present.  results in a forward-backward asymmetry in the number of Drell-Yan lepton pairs. A,B are proportional toweak isospin and charge of the incoming fermions.

  7. Dilution of Asymmetry • Asymmetry is diluted mainlydue to • Detector resolution, and QED FSR, • Acceptance, • Unknown quark/anti-quark directions. Wecorrected for theseeffects and demonstrate the agreement between the data with Standard Model predictions. Acceptance corrections depend on the Standard Model as represented by POWHEG and CT10 PDFs.

  8. Defining forward/backward • At the LHC quark/anti-quark directions are unknown (unlike Tevatron). • But the anti-quark at the LHC is a sea-quark,  on average, p(sea quark) < p(valance quark) • Di-lepton system is boosted along the quark direction. We correct for this effect in the final stage of the analysis by taking the `correct’ quark direction known at the generator level.

  9. Analysis Steps cosqCS* Uncorrected AFB Unfolding (back to the generator level before QED FSR, with acceptance cuts) Acceptance corrected AFB combined e+e-+µ+µ- Quark-direction corrected AFB

  10. Data Samples Muons • Only `good’ lumi sections and runs are used using the json file: Cert_136033-149442_7TeV_Nov4ReReco_Collisions10_JSON.txt  36 pb-1 Electrons MC samples are listed in the back-up slides.

  11. Event Selection – Muon Channel • Global and tracker muons • 2 opposite sign leptons • Impact parameter w.r.t. beam spot |dxy|<0.2 cm • Angle between muons (a-p)<-2.5 mrad (further cosmic muon rejection) • At least one muon hlt matched to the global fit • Run ≤ 147195  Mu9 • 147196 ≤ run < 148108  Mu11 • Run ≥ 148108  Mu15 • Global muon fit c2< 10 • pTµ>20 GeV, and |h| < 2.1 for both muons • Relative combined isolation (trk+HCAL) in R=0.3 < 0.15 • At least one pixel hit on the muon track • At least 10 strip hits • Number of used muon stations > 1

  12. Event Selection – Electron Channel • ET > 20 GeV (after energy scale correction) • 1.0045 in the barrel, 1.045 in the endcap • |h|<2.5, (1.4442<|h|<1.560 excluded) • WP80 selection criteria • Efficiency corrections applied as a function of pT and h (details in the back-up)

  13. Backgrounds • Backgrounds are small (<1% for each channel) • Electron channel: data-driven method • Muon channel: MC • Estimated background in each mass bin is subtracted for forward and backward events separately. • Even for the lowest mass bins, effect of background is small. 40<M<50 GeV e+e- e+e- 50<M<60 GeV e+e-

  14. Defining forward/backward: cosqCS* Collins-Soper frame µ- -qbar qCS* z’ qbar q PAS PAS Q: four-momentum of the di-lepton system Except for the last stage of the analysis (i.e. quark-dir. Corrections) we calculate the angle assuming quark direction = di-lepton direction.

  15. Uncorrected Asymmetry Statistical errors: • Measured and simulated AFB in 11 di-lepton mass bins. • Data points are weighted within a mass. • MC: POWHEG generated events passed through full CMS simulation with same selection cuts applied in data. • Good agreement between MC predictions and data • c2/d.o.f.=1.28 (muon channel), c2/d.o.f.=1.04 (electron channel) PAS PAS (Lyons 1986)

  16. Unfolding Observed mass spectra i  generator level mass bin j  observed mass bin RijFB < 0.007, RijBF < 0.005 Gen: pT > 20 GeV, |h|<2.1 before QED FSR Reco: all cuts Reco Mass [GeV] PAS e.g. RijFB gives the probability backward reconstructed event in bin j to be originated from a forward generated event in bin i. Gen Mass [GeV]

  17. Unfolding – Closure Test • Half of the MC events in the sample is used to derive the response matrices and the other half is used for testing. • Good agreement between unfolded and generated distributions. • Used 33 MC toy experiments 13k events each to evaluate pulls and correlations • We are implementing model independent techniques as well – to be completed after the Moriond conference.

  18. Unfolded Data • MC: Sub-set of generator level (before QED FSR) events with pT>20 GeV & |h|<2.1 for each mass bin defined by the reco level selection cuts. • Good agreement between unfolded data and MC predictions • c2/d.o.f.=1.47 (muon channel), c2/d.o.f.=0.76 (electron channel) PAS PAS

  19. Acceptance Corrections • Shapes of the cosqCS* distribution are different in different mass bins. • F/B asymmetry in acceptance comes from different shaped cosqCS* in different di-lepton mass bins. • For low mass bins acceptance cuts removes less forward events than backward events and vice versa for high mass bins. • The shapes are determined by the Standard Model and PDFs Black  no cut Blue  With acceptance cuts

  20. Acceptance Corrections for each mass bin i. Numerator: Sub-set of generator level (before QED FSR) events with pT>20 GeV & |h|<2.1 (for muons) and 2.5 (for electrons) defined by the reco level selection cuts. Denominator: All generator level events for each mass bin (still assuming Z-boost dir. = quark dir). • Using these corrections, we demonstrate the agreement between the data with the Standard Model predictions. • In this method, the acceptance corrections are dependent on the Standard Model as represented by POWHEG and CT10 PDFs. µµ

  21. Acceptance Corrections – Closure Test

  22. Acceptance Corrected Data • Good agreement between unfolded+acc. Corrected data and the generator level prediction. PAS PAS

  23. Dilution (or quark-direction) Corrections • Unknown quark directions  Large effect on AFB • Can be corrected by changing the sign of the angle for the cases when the Z is not boosted in the quark direction Non-diluted Diluted

  24. Quark Direction Corrected Data • Unfolded+acceptance+quark direction corrected data compared to AFB calculated in MC with correct quark directions and no cuts. • Good agreement-- even with dFB only as a function of mass. • c2/d.o.f.=1.47 (muon channel), c2/d.o.f.=0.76 (electron channel) PAS PAS

  25. Combined Measurement PAS PAS Combined results also agree with MC prediction. AFB from µµ and ee are consistent.

  26. Conclusions • We presented the measurement of the forward-backward asymmetry for opposite charge lepton pairs via Z/g* at √s = 7 TeV up to M(ll)=600 GeV. • AFB measurements demonstrate a good agreement with the Standard Model as presented by POWHEG and the CT10 PDFs.

  27. PAS: EWK-10-011 (Part II)Measurement of the Weak-mixing Angle at CMS Alessio Bonato, Andrei Gritsan, Zijin Guo, Nhan Tran Johns Hopkins University Efe Yazgan Texas Tech University on behalf of the EWK dilepton group March 7, 2011 27

  28. Outline • Introduction and motivation • Methodology • Event reconstruction and selection • Measurement of sin2W 28

  29. Introduction and motivation • The process, ppXl+l-, a rich channel for possible physics beyond the SM: extra dimensions, new gauge bosons, supersymmetry, etc. • The SM provides testing ground: ppZ/*l+l- • Information about couplings contained in angular distributions • Recall, for the SM Z (J=1), couplings include vector and axial-vector components: 1 = cV(W) and 2 = cA(W) • In developing angular formalism for dilepton resonances,we provide a measurement of the SM couplings and the Weinberg angle, sin2W. 29

  30. Introduction and Motivation _ Consider the process qq  Z/* l+l- and its differential cross-section: • dcos : sensitive to couplings Z/* uu, dd, l+l- (sin2W)and relative quark contributions (PDFs) • Relative contributions of Z/* dependent on mass • Multi-dimensional analysis increases sensitivity • Dimuon measurement unique to CMS */Z By studying the differential cross-section of the DY process, we can make precision measurements of SM parameters and PDFs; deviations may come from new physics in Z’/Z/*. 30

  31. Likelihood analysis of sin2W • Idea: per event multivariate likelihood function to extract maximal information from the event • Requires probability distribution function in observables mass, angle, rapidity: Psig(m,Y,cos; sin2W, PDFs) • Analytical approach, start with phenomenological model and introduce detector effects matched to data • In this analysis, single parameter likelihood fit of sin2W,assumes SM and PDFs well-established 31

  32. Experimental challenges • The LHC is a pp collider, the quark direction is unknown (dilution); choose reference frame based on boost (rapidity, Y) of the dilepton system • On average, quark carries more momentum than antiquark • Collins-Soper frame is used to reduce effect of boson transverse momentum • Detector acceptance reduces sensitivity to angular variables • Final state radiation (FSR) and detector resolution distort mass-angle correlations 32

  33. Likelihood analysis of sin2W • Building the probability density function • DY mass-angle distribution including partonic luminositiesPideal(m,Y,cos) • Include acceptance and efficiency: Pideal(m,Y,cos) Gacc(m,Y,cos) • Include detector resolution and FSR effects: [Pideal(m,Y,cos) R (m) ] G (m,Y,cos) • Model is built at LO; (N)NLO effects treated as corrections to model • Information about sin2W contained in the shapes of the multidimensional space Psig(m,Y,cos; sin2W, PDFs) = [Pideal(m,Y,cos) R (m) ] x Gacc(m,Y,cos) 33

  34. Data samples and event selection • Muon datasets • /Mu/Run2010A (run 136033- 144114) • /Mu/Run2010B (run 146428-149442) • Dec22 ReReco using recently approved latest tracker geometry and corresponding MC geometry • Integrated luminosity of 40 pb-1 using ‘noCalo’ JSON file • Triggers same as AFB but also include ‘DoubleMuon3’ trigger • Specific likelihood analysis cuts: • pT > 18,7 GeV, || < 2.4 • pT (CS) > 18GeV, || (CS) < 2.3 - for expanded analytical acceptance to increase sensitivity • pT (Z) < 25 GeV - to suppress NLO effects • Tracker isolation due to ‘noCalo’ dataset 34

  35. Background: dimuon channel Main backgrounds from +- and QCD with smaller contributions from WW,WZ, W+jets, ZZ, tt; total background < 1% for approval Expected number of background events: 36 Background shapes included in likelihood model (more later) 35

  36. Likelihood model Recall, built likelihood function includes sin2W dependence: Psig(m,Y,cos; sin2W, PDFs) = [Pideal(m,Y,cos) R (m) ] x Gacc(m,Y,cos) Pbkg(m,Y,cos) = P (m) P (Y) P (cos) Gacc(m,Y,cos) Pideal(m,Y,cos) in good agreement with LO Pythia MC: for approval Points: LO Pythia MC (gen. level) --- Lines: probability distribution function N.B. Probability distribution function in three-dimensional correlated space, 1D projections for illustration 36

  37. Acceptance and efficiency Acceptance + efficiency sculpts further the Y and cos distributions Lepton cuts: || < Ymax; pT > pT,min Acceptance conditions: |cos| < tanh(Ymax - |Y|); |cos| < [1-(2pT,min/m)2]1/2 Probability Density Function ~ P (m,cos,Y) x Ga(m,cos,Y) for approval Gacc(m,cos,Y) Before acceptance/after acceptance 37 Acceptance cuts in CS frame covers acceptance cuts in lab frame

  38. Acceptance and efficiency Further acceptance function to include efficiency effects and NLO corrections Probability Density Function ~ P (m,cos,Y) x Ga(m,cos,Y) x Gb(cos,Y) Efficiency vs. Y and cos Reflection of efficiency in (µ+) and (µ-) Gb(cos,Y) Add into the fit model a 2D description of the efficiency in Y and cos. Polynomial fit of efficiency in cos in bins of Y to construct 2D interpolation function. cos Y 38

  39. Resolution+FSR Account for resolution+FSR via convolution Probability Density Function ~[P (m,cos,Y) R (m)] x Gacc(m,cos,Y) Assume resolution function, R (m), unknown. Approximated by quadruple Gaussian, R4g(m), for analytical convolution. Fit full probability distribution function to the MC and obtain R4g(m) parameters from the fit Convolution of resolution function Generator level Resolution + FSR R4g(m) 39

  40. Background: prob. density func. Background probability density function built empirically in 1D slices times a 3D acceptance function: Pbkg(m,Y,cos) = P (m) P (Y) P (cos) Gacc(m,Y,cos) Expected background yield is 36 events for the 40 pb-1 sample (agreement between MC and data-driven estimates) Fix number of background events in fit to data 40

  41. Putting it all together: MC Final likelihood model on Powheg+Pythia CMS simulation PAS Result of 400 toy experiments including sig + bkg yields: sin2W = 0.2306  0.0004 (generated value: 0.2311) Mean expected statistical error per toy: 0.0078 41

  42. Toys, G.O.F., pull distribution Result of 400 toy experiments: Bkg contributions Poisson-varied, N events per toy constant mean: 0.2306 rms: 0.0078 Error from fit data: 0.0077 for approval mean: -0.02 rms: 0.96 Measure of goodness-of-fit data value 42

  43. Systematic uncertainties • Dominant backgrounds from FSR and resolution/alignment • FSR comparison in PHOTOS and Pythia • Variation of resolution function parameters and momentum scale in data 43

  44. Systematic Uncertainties • Other systematic uncertainties • LO model (ISR): vary generated value of sin2W and compare fit values • PDFs: comparison of MSTW and CTEQ PDF sets • Fit model: small bias from toy MC results • Background: vary QCD contribution by 50%, from comparison of data-drive with MC; EWK backgrounds expected to be modeled well by MC • Conservative estimates; certain cases statistics limited Total systematic errors are less than statistical errors 44

  45. Results with data Data fit central value kept blind until March 1st to avoid analysis bias PAS Fit result: sin2W = 0.????  0.0077 (stat.)  0.0036 (stat.) PDG value: 0.2312 Final cross-check: goodness-of-fit test yields good agreement with MC 45

  46. Results with data Data fit central value kept blind until March 1st to avoid analysis bias PAS Fit result: sin2W = 0.2287  0.0077 (stat.)  0.0036 (stat.) PDG value: 0.2312 Final cross-check: goodness-of-fit test yields good agreement with MC (-logL = -44507) 46

  47. Conclusions • A novel technique for measurement of the weak mixing angle via the likelihood method is presented • An analytical likelihood function is built including experimental effects and a single parameter fit of sin2W is performed • The results of the fit on 40pb-1 of data is: Fit result: sin2W = 0.2287  0.0077 (stat.)  0.0036 (stat.) 47

  48. Appendix: comparison of likelihood method with traditional template method 48

  49. Comparison of methods • What is the (statistical) improvement of the likelihood method over traditional methods for determining sin2W? • “template” method: D0 extracts sin2W from AFB by generating templates of AFB for many values of Weinberg angle, finds the most probable value of sin2W [PRL 101, 191801 (2008)] • Problem: we don’t have tons of MC at different values of sin2W to generate templates • Solution: we make a generator level study adding in FSR and realistic “fast smearing” • Not trivial to extract sin2W from unfolded AFB, requires proper statistical treatment and full correlation matrices; extract most probable sin2W value from raw AFB for convenient error estimation • More details of this on hypernews https://hypernews.cern.ch/HyperNews/CMS/get/EWK-10-011/11.html 49

  50. Comparison of methods • The setup: generate 1M events Powheg+Pythia+fast smear for different values of sin2W(0.2112 to 0.2512 in increments of 0.002) • Extract the raw AFB for each sample to use as “templates” • Run toys (~15k events per) with the sin2W = 0.2312 sample and find the most probable value and error on sin2W • Compare this with the likelihood fit method using the same toy experiments Fit to most probable value All templates Results of toy Result of one toy exp. Mean error on sin2W: 0.0014 (templates), 0.0080 (likelihood) Template method has errors factor of 1.4 larger than likelihood method Likelihood method equivalent to doubling the statistics! 50

More Related