400 likes | 568 Views
QCD Event shape variables in pp collision at 900/2360 GeV. Introduction Minbias Models Data-set and event/track selection Event shape variables in Data and systematics Comparison of different models Conclusion. Introduction.
E N D
QCD Event shape variables in pp collision at 900/2360 GeV • Introduction • Minbias Models • Data-set and event/track selection • Event shape variables in Data and systematics • Comparison of different models • Conclusion
Introduction • Event shape variables widely used in ee/ep machine to tune MC (non-perturbative QCD effect) • An attempt to look in pp collider for better QCD model Variables under study : This study is based on only tracker information, with the assumption of zero mass and P_Z=0
Minbias models • DW (Rick Field, ref : hep-ph/0201192) :Mainly tuning of CDF underling events, charge multiplicity, scalar sum in transverse plane, tuning PYTHIA, e.g., • 'MSTJ(11)=3 ! Choice of the fragmentation function' (light quark in symmetric fuction and heavy quark with Peterson/SLAC functionm), D=4 • 'MSTP(81)=1 ! multiple parton interactions 1 is Pythia default', D=1 • 'MSTP(82)=4 ! Defines the multi-parton model', D=4 • 'PARP(82)=1.9 ! pt cutoff for multiparton interactions', D=2.0 GeV • 'PARP(83)=0.5 ! Multiple interactions: matter distrbn parameter', D=0.5 • 'PARP(84)=0.4 ! Multiple interactions: matter distribution parameter', D=0.4 • 'PARP(90)=0.25 ! Multiple interactions: rescaling power', D=0.16 • 'PARP(67)=2.5 ! amount of initial-state radiation', D=4.0 • 'PARP(85)=1.0 ! gluon prod. mechanism in MI, prob of additional interaction...', D=0.9 • 'PARP(86)=1.0 ! gluon prod. mechanism in MI' prob of additional interaction...', D=0.95 • 'PARP(62)=1.25 ! Effective cut-off Q or K_T ', D=1.0 GeV • 'PARP(64)=0.2 ! Transverse momentum evolution scale ..', D=1.0 • 'PARP(91)=2.1 ! 354 kt distribution', D=2.0 • 'PARP(93)=15.0 ! 355 Upper cut-off of K_T'), D=5.0
Minbias models • Professor’s model (ref : arXiv:0902.4403) : Based on DELPHI tuning • FSR and hadronisation from e+e− data • 'PARJ(1)=0.073 ! FLAV P(qq)/P(q) D=0.10 • 'PARJ(2)=0.2 ! FLAV P(s)/p(u) D=0.30 • 'PARJ(3)=0.94 ! FLAV (P(us)/P(ud))/(P(s)/P(d)) D= 0.4 • 'PARJ(4)=0.032 ! FLAV Suppression of spin 1 wrt spin 0, D=0.05 • 'PARJ(11)=0.31 ! FLAV prob of spin 1 (ud) meson, D=0.5 • 'PARJ(12)=0.4 ! FLAV Prob. a strange meson has spine 1, D=0.6 • 'PARJ(13)=0.54 ! FLAV Prob of spine 1 (in heavier meson) D=0.75 • 'PARJ(25)=0.63 ! FLAV Extra suppression factor for eta, D=1.0 • 'PARJ(26)=0.12 ! FLAV Extra suppression factor for eta', D=0.4 • 'MSTJ(11)=5 ! HAD Choice of the fragmentation function', Lund, but interpolation between Bowler and Lund ... D=4 (lund) • 'PARJ(21)=0.313 ! HAD', Gaussian width of parton Pt smearing inside hadron D=0.36 • 'PARJ(41)=0.49 ! HAD', 'a' of symmetri Lund fragmentation function, D=0.3 • 'PARJ(42)=1.2 ! HAD', 'b' of symmetri Lund fragmentation function, D=0.58 • 'PARJ(46)=1.0 ! HAD', rC id Bowler shape D=1.0 • 'PARJ(47)=1.0 !408 HAD', rB id Bowler shape D=1.0
Minbias models : Professors • ISR and MPI from pp-bar data • 'MSTP(81)=1 (2) ! MPI 21 is Pythia new set of MPI models', D=1 • 'MSTP(82)=4 (5) ! MPI model, structure', D=4 • 'PARP(82)=1.9 (2.0) ! MPI pt cutoff for multiparton interactions', D=2.0 • 'PARP(83)=0.6 (1.7) ! MPI matter distribution parameter', O(b) \prop exp(-b^d), D=0.5 • 'PARP(90)=0.22 (0.26) ! MPI rescaling power', D=0.16 • Professor’s tuned ISR, FSR parameters are default in PYTHIA • Also parameter’s are tuned for color reconnection, branching fractions (same is true for Perugia)
Minbias models • Perugia model (Ref : arXiv:0905.3418) • FSR and hadronisation from e+e− data (same as Professor’s) • ISR and K_T : Drell-Yan PT spectrum at Tevtron (√S=1800 + 1960 GeV) • Underlying events (UE) and Beam Remnants (BR) : Charge multiplicity (Nch), dNch/dPT, <PT>(Nch) in min-bias events at CDF • Energy Scaling : Nch in min-bias events at 200,540 and 900 GeV (UA5) + 630 + 1800 GeV (CDF) • FSR : • 'PARJ(81)=0.257 ! FSR', Lambda value in running alpha_S, D=0.29 • 'PARJ(82)=0.8 ! FSR', Invariant mass cut-off m_min in PYSHOW, D=1.0 • 'PARP(71)=2.0 ! FSR', .. maximum parton virtuality, D=4.0 • ISR : • 'MSTP(64)=3 ! ISR', Choice of Alpha_s and Q2 for space-like parton showers D=2 • 'PARP(64)=1.0 ! ISR', evolution scale multiplied by parp(62), D=1.0 • 'MSTP(67)=2 ! ISR', Colour coherence effect, D=2 • 'PARP(67)=1.0 ! ISR', Q2 scale of the hard scattering (MATP(32) is multiplied by this, D=4 • 'MSTP(70)=2 ! ISR', Regularision scheme when Pt->0, D=1 • 'MSTP(72)=1 ! ISR', Maximum scale for FSR stretch between ISR ... D=1
Data sets and event selection • Data : /MinimumBias/BeamCommissioning09-Jan29ReReco-v2/RECO • MC : /MinBias/Summer09-STARTUP3X_V8O_2360GeV_Jan29ReReco-v1/GEN-SIM-RECO /MinBias/Summer09-STARTUP3X_V8P_900GeV_Jan29ReReco-v1/GEN-SIM-RECO/MinBias/Summer09-STARTUP3X_V8K_900GeV_DW-v1/GEN-SIM-RECO/MinBias/Summer09-STARTUP3X_V8K_900GeV_P0-v1/GEN-SIM-RECO/MinBias/Summer09-STARTUP3X_V8K_900GeV_ProQ20-v1/GEN-SIM-RECO Due to crab problem, we do not have full sample
Event selection and track selection (tried to use similar crit of UE selection) Event selection : • Only one primary vertex • Vertex position, |Δr|<0.15 and |ΔZ|<15cm wrt nominal interaction vertex • Fraction of high purity tracks > 25% • Number of reconstructed tracks < 150 • BSC trigger bit 40,41 : HLT_MinBiasBSC, beamhalo veto and BPTX+ && BPTX− ,technical trigger bit 0) (for data only) Track selection : • Associated with primary vertex and weight in vertex fit> 0.2 • PT > 750 MeV, |η|<2.2 (study on the threshold, range as systematic) • quality(TrackBase::highPurity) • |d0|<0.1cm, |dz|<0.1cm wrt event vertex, ndf>=10, atleast one silicon layer hit and Track fit prob > 10−8 Finally, at least 3 selected tracks in the event
Run/Lumi/bunch selection //Good run/lumi/branch * (irun==124009 && ilumi>=1 && ilumi<=68 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124020 && ilumi>=12 && ilumi<=94 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124022 && ilumi>=69 && ilumi<=160 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124023 && ilumi>=41 && ilumi<=96 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124024 && ilumi>=2 && ilumi<=83 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124027 && ilumi>=24 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124030 && ilumi>=1 && ilumi<=31 && (ibrnc==2824||ibrnc==151||ibrnc==51)); * (irun==124230 && ilumi>=26 && ilumi<=68 && (ibrnc==51 ||ibrnc==151 ||ibrnc==232 || ibrnc==1042 || ibrnc==1123 || ibrnc==1933 ||ibrnc==2014 ||ibrnc==2824 ||ibrnc==2905)); • Again used the same events as it is in UE analysis • But also looked for events with Run # <123900 and (Early) • Lumi/bunch rejected in UE selection (Others) • 2360 data No luminosity, efficiency correction …, only comparison of shape
Comparison of primary vertex parameters Data and MC does not match in track multiplicity as well as primary vertex resolution
Comparison of track parameters in Data and MC • Impact parameters are better for MC sample • Track fit probability is also better in MC
Pt distribution of track for different |η| range • Track multiplicity has large difference in Data and MC • Nearly same as it is number of associated tracks with the primary vertex • Tracking efficiency is less in forward direction |η|<2.5 |η|<2.2 |η|<1.6 |η|<1.9 Not much dependency on eta : Perugia model has poorer matching
Pt distribution of track for different |η| range (zoomed) |η|<2.5 |η|<2.2 |η|<1.6 |η|<1.9 Not much dependency on eta : Perugia model has poorer matching
Pseudorapidity of track for different Pt threshold Pt>0.5 Pt>0.3 Pt>1.5 Pt>1.0 Better Data/MC matching with higher Pt threshold
Number of selected tracks for different Pt & |η| criteria Pt>0.5 |η|<2.5 Pt>0.5 |η|<2.0 Pt>1.0 |η|<2.5 Pt>1.0 |η|<2.0 Close match with higher Pt threshold, but different models vary different way. At high Pt threshold Perugia and Professor’s model show better matchings.
Comparison of track parameters in Data and MC Variables after all selection criteria • χ2/ndf : from ChisqTestX() of MC samples wrt to data • Tracking efficiency is less in forward direction • Pt distribution for Perugia minbias model is different than data/other models, but better interms of track multiplicity
Track parameters : Different data sets • Looks data in different run periods • Distributions in different datasets are nearly same • Bottom plots : Ratio wrt 1st plots, • Error is statistical only Top one in logarithamic scale, whereas ratio’s are in linear scale <χ2> : rms deviation of ratio from 1, normalised with error Pr : Prob (chisquare, ndf)
Track parameters : Different data sets • Distributions in different datasets are nearly same
Systematic • Change in distributions in different run • Divided data in four parts, • UE event selection excluding run 124330 • Run 124330 • Run 124009 – 124330, but lumi/bunch is not selected in UE • Initial runs, < 123900 • Change due to range in eta (arbitrary) • |η|< 2.0, 2.1, 2.2, 2.3, 2.4 • Change due to different Pt threshold (arbitrary) • Pt > 0.70, 0.75, 0.80 GeV Problem : Statistical error is not negligible, it is added quadratically with relative shift
Event Shape parameters : Different data sets • Distributions in different datasets are nearly same • Error in each bin for systematic : Maximum in three data sets, sqrt(shift^2 + error^2)
Event Shape parameters : Different data sets • Distributions in different datasets are nearly same
Event Shape parameters : Different |η| range • Not much variation due to different range of |η|
Event Shape parameters : Different |η| range • Not much variation due to different range of |η|
Event Shape Variables : Comparison of Data and MC All four models are deviating from data, need MC tuning. Perugia is slightly better than others
Event Shape Variables : Comparison of Data and MC All four models are deviating from data
Correlation of tracks : Comparison of Data and MC Binwidths for Pt distribution are not const −Ve for opposite sign tracks : All four models are deviating from data, Professor-Q has better agreement for ΔPt, whereas Perugia for ΔΦ
Correlation of tracks : Comparison of Data and MC −Ve for opposite sign tracks : All four models are deviating from data, again ProQ has better agreement for ΔP, whereas P0 for Opening angle
Difference in azimuthal angles of selected tracks Pt>0.5 |η|<2.0 Pt>0.5 |η|<2.5 Pt>1.0 |η|<2.5 Pt>1.0 |η|<2.0 −ve sign is for opposite sign tracks Perugia model shows better matching than others two
Difference in Pt of selected tracks Pt>0.5 |η|<2.0 Pt>0.5 |η|<2.5 Pt>1.0 |η|<2.5 Pt>1.0 |η|<2.0 Perugia is poorer in comparison with others, opposite to ΔΦ correlation
χ2 comparison of different models, with different Pt, |η| criteria 10 |η| bin from 2.5 to 1.6 and 18 Pt bin from 0.3 to 2.0 GeV Matching of Data/MC varies with selection, mainly on Pt, very less on eta Perugia model : Much more different than others
χ2 comparison of different models, , with different Pt, |η| criteria QCD event shape variables 10 Pt bins >[0.5, 0.6, 0.65, 0.7, 0.75, 0.8, 0.9, 1.0, 1.5, 2.0] GeV 5 |η| bin < 2.4 to 1.6 Perugia model is has better matching with data. For different threshold statistics are different, but variations in different models are different
Comparison of primary vertex parameters : 2360 GeV Data and MC does not match in track multiplicity as well as primary vertex resolution Beam position is much precise at high energy
Comparison of track parameters in Data and MC : 2360 GeV • Impact parameters and track fit probability are better for MC sample, but same for both beam energy
Comparison of track parameters in Data and MC : 2360 GeV Variables after all selection criteria • Expected increase in multiplicity and average Pt of track with beam energy
Event Shape variables : Comparison of Data and MC at 2360 GeV • Same discrepancy in Data and MC, what was seen at 900 GeV, again track multiplicity is major concern.
Conclusion • Data and MC differ in track multiplicity, which can effect the distributions of QCD event shape variables • Minbias models are not consistent with data, Perugia is closer to Data, but discrepancy is ~10% level (excluding normalisation factor) • Correlation of Pt and Φ of tracks are different in different models. • 2.36TeV data also show the same discrepancy.
Event Shape parameters : Different Pt threshold • Large dependency the threshold. Threshold mainly change the number of tracks in a events, which eventually change shapes
Event Shape parameters : Different Pt threshold • Large dependency the threshold. Threshold mainly change the number of tracks in a events, which eventually change shapes