Feasibility of a cross-section measurement for J/ψ->ee with the ATLAS detector

Transcription

1 Feasibility of a cross-section measurement for J/ψ->ee with the ATLAS detector ATLAS Geneva physics meeting Andrée Robichaud-Véronneau

2 Outline Motivation Theoretical background for J/ψ hadroproduction Analysis method Cross-section measurement Systematic uncertainties October 15th, 2010 ATLAS Geneva physics 2

3 Measuring J/ψ->ee in ATLAS When? In the early days of ATLAS while the instantaneous luminosity is low and the trigger is not too highly prescaled Basically now Why? They provide a good signal for low energy calibration and performance assessment of the EM calorimeter and the ID ( standard candle ) important for new physics To help understanding the QCD aspects of the production of J/ψ mesons at hadron colliders Because we can! A cross-section measurement could be obtained early to compare with existing results from J/ψ->μμ at 7 TeV October 15th, 2010 ATLAS Geneva physics 3

4 The charmonium spectrum Particle Mass (MeV) BR(-> J/ψ) J/ψ χ c % χ c % χ c % ψ' % J/ψ is part of a larger family of cc mesons called charmonium October 15th, 2010 ATLAS Geneva physics 4

5 J/ψ production types Different types of production: Direct: pp->j/ψ Indirect: pp->χ c ->J/ψγ Direct and indirect form the prompt production Non-prompt: pp->bb->j/ψ Tevatron measured them separately could be difficult at LHC Photon from indirect radiative decays has an energy of ~400 MeV Non-prompt production can be selected using the pseudo proper time of the B-meson reconstructed vertex 0 = L xy M p T c Generator level October 15th, 2010 ATLAS Geneva physics 5

6 J/ψ production models vs data CDF run I CDF run II Need more data measurements to decide on the fate of the COM and/or of future models COM Colour Octet Model (a)-(b)-(c)-(d) Colour Singlet Model (a) + (b) October 15th, 2010 ATLAS Geneva physics 6

7 J/ψ->ee cross-section in ATLAS The feasibility of a cross-section measurement for J/ψ->ee in ATLAS was studied: Analysis performed on simulation only using 10 TeV as E cm Simultaneous measurements are performed: Prompt J/ψ cross-section Non-prompt J/ψ cross-section More challenging than J/ψ->μμ October 15th, 2010 ATLAS Geneva physics 7

8 Monte Carlo samples at 10 TeV Samples Lumi (pb -1 ) Cross-section (μb) # events pp->j/ψ->ee k bb->j/ψ->ee k Drell-Yan->ee k filtered minimum bias M bb->ex M cc->ex k Truth jet filter with p T > 6 GeV on minimum bias (small bias after selection) Multilepton Filter: 2 electrons with p T > 3 GeV and η < 2.7 on signal samples and Drell-Yan Single electron PythiaB filter on heavy flavour samples (p T > 5 GeV and η < 2.5) Minimum bias is statistically limited, even after filter October 15th, 2010 ATLAS Geneva physics 8

9 Event selection 2e5_medium trigger: 2 trigger electron objects with E T >5 GeV and medium electron identification level 2 electrons found in the event with: E T >5 GeV η < 2.47 and not in the calorimeter transition region in η ΔR > 0.1 between the electron clusters in the pair (to reject duplicates) ΔR < 0.03 between the EF trigger and the offline object Opposite charge tracks refitted to a common vertex using VKalVrt χ2 < 6 for the common vertex refit 2 < m ee < 4 GeV (signal region: 2.5 < m ee < 3.5 GeV) Tight identification cuts Choose leading J/ψ in the event October 15th, 2010 ATLAS Geneva physics 9

10 Background sources Heavy flavours (dominant after selection) True electrons from bb and cc (also fakes and electronfake pairs) are significant, also in the mass region of interest (from 2 to 4 GeV) A non-negligible fraction survives the tightest identification cuts Minimum bias (excluding heavy flavours) Source of true background electrons (photon conversions, Dalitz decays, light meson decays) and fakes. Hadron fakes contribution controlled using tight identification cuts Drell-Yan pairs True good isolated electron pairs Cannot be rejected, but has a low cross-section October 15th, 2010 ATLAS Geneva physics 10

11 Signal efficiencies Prompt Non-prompt = # of selected events # of generated events Non-prompt efficiency lower than prompt one Shape difference explained by hadronic activity in nonprompt events October 15th, 2010 ATLAS Geneva physics 11

12 Effect of identification cuts cuts applied sequentially cuts applied individually R had and R η have the largest difference in rejection between the two samples. R cut = N n 1 N n N n 1 October 15th, 2010 ATLAS Geneva physics 12

13 Trigger efficiencies Prompt Non-prompt = # of events passing E T and cuts and 2e5 and R trigger offline # of events passing E T and cuts Trigger-offline matching: EF_2e5:ΔR < 0.03 and L1_2EM3: ΔR < 0.16 The selection can be improved to have a more similar response for the two signal contributions October 15th, 2010 ATLAS Geneva physics 13

14 Mass fit and resolution Invariant mass fitted with Crystal Ball function + linear background Fit results: Mass = 3.082±0.008 GeV Resolution = 136±8 MeV October 15th, 2010 ATLAS Geneva physics 14

15 Background subtraction Assuming a linear background for the mass region of interest, use sideband subtraction to remove background. This was verified for 2 < m ee < 4 GeV. Signal region: 2.5 < m ee < 3.5 GeV Sidebands: 2.0 < m ee < 2.5 GeV and 3.5 < m ee < 4.0 GeV Give negative weights to events in sidebands and positive weights to events in signal region to compensate. October 15th, 2010 ATLAS Geneva physics 15

16 Control plots Means of distributions are well centered around 0 after subtraction October 15th, 2010 ATLAS Geneva physics 16

17 Ratio of non-prompt-to-prompt Ratio of non-prompt-to-prompt events using the displacement of the non-prompt vertex wrt the primary vertex Compute the pseudo-proper time τ 0. 0 = L xy M p T c October 15th, 2010 ATLAS Geneva physics 17

18 Ratio of non-prompt-to-prompt τ 0 distribution fitted to the sum of a δ-function and an exponential, smeared with a normalised Gaussian R expected = 0.214±0.004 Fit result R for signal only = 0.25±0.01 Linear correction applied to fit result Corrected value of R = 0.21±0.03 The full correction is taken as a systematic uncertainty October 15th, 2010 ATLAS Geneva physics 18

19 Cross-section calculation N: inclusive number of events ϵ p and ϵ np : efficiencies for prompt and non-prompt R: ratio of non-prompt-to-prompt L: integrated luminosity p = N p p L = N 1 R p L np = N np np L = NR 1 R np L Obtained fiducial cross-sections: σ p =106±4 nb, σ np = 28±1 nb. Pythia predictions are σ p =105.5±0.4 nb, σ np =28.3±0.3 nb. This results validates the method used for the crosssection calculation October 15th, 2010 ATLAS Geneva physics 19

20 Differential cross-section Quantities plotted as function of reconstructed p T N p and N np distributions unfolded to generator level p T using bin-by-bin unfolding (RooUnfold) October 15th, 2010 ATLAS Geneva physics 20

21 Differential cross-section Prompt Non-prompt Using expressions defined previously Cross-section computed here before the MultiLepton generator filter Discrepancy in σ np originates from R fit. October 15th, 2010 ATLAS Geneva physics 21

22 Systematic uncertainties Value shifts obtained from data Relative ariations in efficiency measured (in table) October 15th, 2010 ATLAS Geneva physics 22

23 Systematic uncertainties Δσ p /σ p Δσ np /σ np Luminosity 11% 11% Electron ID 6.3% 6.5% Cut variations 9.5% 9.1% Number of events 2% 2% Ratio of np-to-p 2% 12% Theory 11% 11% Total 19.5% 22.7% E T and m ee cuts have the largest impact on cut variations. R η and TRfrac are the largest contributors to the electron identification systematic uncertainty on the efficiency. Largest theoretical contributions (e.g. polarization) not considered here. Theory here means variation of the hard interaction scale October 15th, 2010 ATLAS Geneva physics 23

24 Systematic uncertainties Additionnal sources of systematic uncertainties are: Trigger efficiency Material effects Track p T resolution and scale Vertexing Polarization PDF uncertainty Parton shower, hadronisation and UE modelling All of them should be considered in a 7 TeV data measurement October 15th, 2010 ATLAS Geneva physics 24

25 Conclusions Using the method described here, we determine that the cross-section measurement is feasible. More challenging than J/ψ->μμ due to low p T issues such as duplicate clusters and reconstruction efficiency There is large theoretical uncertainty, but this is true for both leptonic decay channels The largest systematic uncertainties on the measurement can be controlled. Given the efficiency shape below 10 GeV, the total crosssection can only be measured above this threshold. σ p = 85±4 (stat.)±14 (syst.)±9 (theo.) nb σ np = 20±1 (stat.)±4 (syst.)±2 (theo.) nb In data, trigger and offline efficiencies can be obtained using Tag and Probe methods The fit method for the ratio of non-prompt-to-prompt events should be improved. October 15th, 2010 ATLAS Geneva physics 25

26 Backup slides October 15th, 2010 ATLAS Geneva physics 26

27 J/ψ history Existence of the c quark first predicted by the GIM mechanism in 1970 with a higher mass than previously observed quarks Discovery of a narrow resonance with a mass of 3.1 GeV by two experiments simultaneously (MIT and SLAC) in November 1974 (S. Ting et al, B. Richter et al). Given the name J by MIT and ψ by SLAC, hence the final name J/ψ October 15th, 2010 ATLAS Geneva physics 27

28 J/ψ->ee event selection in data In order to observe in first data, a tailored event selection was needed Usage of EM clusters seeded by topological clusters to increase efficiency by a factor 2 Chosen identification variables with good discriminating power from the first and second cluster layers and the tracks Apply duplicate removal on the tracks and the clusters (especially needed at low energy with the topo-seeded clusters Require electron track p T > 2 GeV and dielectron p T > 0.5 GeV Using least biasing unprescaled triggers (or a combination of several prescaled triggers) to obtain the highest possible statistics ATL-COM-PHYS J/ψ is also seen in standard clustering, but with less statistics (factor of ~2) October 15th, 2010 ATLAS Geneva physics 28

29 J/ψ->ee event selection in data October 15th, 2010 ATLAS Geneva physics 29

30 J/ψ->ee observation in ATLAS Cuts modified to reach a lower p T threshold I performed checks which shows that we expect more OS than SS Good agreement between 7 TeV data and MC ATLAS, work ATLAS, in progress work in progress ATLAS, work in progress ATLAS, work in progress ATLAS, work in progress October 15th, 2010 ATLAS Geneva physics 30

31 J/ψ->ee observation in ATLAS ATLAS, work in progress ATLAS, work in progress Dielectron invariant mass using energy from the EM calorimeter cluster and direction from the ID track Fitted using a Crystal Ball function to take the low mass radiative tail into account Lower mass than expected due to non-optimal cluster calibration October 15th, 2010 ATLAS Geneva physics 31

32 Duplicate electron clusters We observed and studied the rate of duplicate electron clusters The rate for low energy electrons with original clustering (2.4%) was higher than for high energy electrons (0.8%) Changes in clustering helped reducing the rate: Reduced seed cluster size Reduced seed E T threshold (3.0 to 2.5 GeV) Improved duplicate removal procedure October 15th, 2010 ATLAS Geneva physics 32

33 Electron reconstruction Two algorithms: Cluster-seeded: optimized for high-p T Track-seeded: optimized for low-p T A comparative study was performed Cluster-seeded electrons gives a better efficiency than track-seeded electrons for electrons with E T >5 GeV Used by the trigger Cluster-seeded sufficient for this analysis October 15th, 2010 ATLAS Geneva physics 33

34 Electron identification Standard in ATLAS, based on several cuts for discriminating variables in E T and η bins Three levels of cuts Loose: Using only shower shapes in the middle layer of the calorimeter Medium: On top of loose, adding first layer shower shapes and track matching cuts Tight: On top of medium, adding more track matching cuts and using the TRT electron identification power (also rejecting conversions using b-layer hits) Recently optimized in time for data-taking Optimization based on H->llll and Z->ee October 15th, 2010 ATLAS Geneva physics 34

35 Electron identification ϵ Reco ϵ Loose ϵ Medium ϵ Tight 71.0% 61.0% (85.9%) 58.3% (82.1%) 44.6% (69.7%) Overall efficiencies: within truth acceptance (wrt reconstruction) Truth acceptance = E T >5 GeV and η < 2.47 October 15th, 2010 ATLAS Geneva physics 35

36 Trigger efficiencies October 15th, 2010 ATLAS Geneva physics 36

37 Prompt signal reweighting The prompt signal sample contains 4 contributions as forced decays. When decays are forced, Pythia sets their BR to 1. Event weights are applied to correct for the BR and are calculated as: Process Event weights pp->j/ψ χ c0 ->J/ψγ χ c1 ->J/ψγ χ c2 ->J/ψγ w i = j BR i before filter BR j f j where i,j are the four processes and the fractions of events (f) are calculated before the ML filter. October 15th, 2010 ATLAS Geneva physics 37

38 Generator filter efficiencies October 15th, 2010 ATLAS Geneva physics 38

39 Unfolding October 15th, 2010 ATLAS Geneva physics 39

40 Cut variation systematics October 15th, 2010 ATLAS Geneva physics 40

41 Number of events systematics October 15th, 2010 ATLAS Geneva physics 41

42 Acceptance systematics October 15th, 2010 ATLAS Geneva physics 42

43 Differential cross-section October 15th, 2010 ATLAS Geneva physics 43