%TOATTENTION%

CSCObjectSelection of topview

CSCBulkSamplesStatus

PhysicsAnalysisWorkBook

Physics topics

top production

• 
• More precisely: 86% : 14% for gg and qq to tt production ?

w polarization measurement

ttbar spin correlation measurement

• ttbar produced in pairs, the spins of them are correlated. The to be measured factors is defined as and .
  • The spins are casted to selected quantization axises. Choose good axises, A can be maximized.
  • The spins are casted to each other. . It can be expressed as a sum of 3 independent casting in 3 dimension space .
  • and can be transformed to each other, according to the definition above.

Exist analysis

A) Semileptonic ttbar Events:
1) Leptons
   Number of Isolated Leptons: exactly 1
   Electrons must have a Pt higher (or equal) to 20GeV
   Muons     must have a Pt higher (or equal) to 20GeV
   Use Tight leptons
2) Jets
   Number of Jets:  at least 4
   Jets must have a Pt higher (or equal) to 30GeV
   Number of b-jets: at least 2
3) Missing
   Pt missing should be higher (or equal) to 20GeV
B) Trigger:
at level 1  EM25 (MU20 or MU40) for electrons (muons)  
at Level 2 and EF, E25I (MU20I) for electrons (muons).

• in pdf
• in ppt 1
• in ppt 2

For consistency in CSC T7

• Electrons must have a Pt higher (or equal) to *25GeV*
• Changes:

Qualities	e cut pt>20GeV	e cut pt>25GeV	comments
electron selection efficiency	51.2%	46.9%
4+ jets	59.3%	59.9%	jet which is close to selected electron deltaR<0.1 is considered as a electron. 4% electron as above is not selected in e events (2% in all events), and 28% events with 3 jets, means 28%*2%=0.56% events, electron helps to pass cut of 4+jets
all jets tag efficiency	26.6%	26.5%	more electron-jets come in to worsen the efficiency
ljet rejection	131	133	same as above
kinematic cut	9.77%	9.43%
quality cut	2.93%	2.81%
electron reconstruction efficiency	69.6%	70.2%	a little bit higher for high pt electron
electron purity	98.4%	98.8%	higher fake rate for low pt electrons

Assigned to	Due date	Description	State	Closed	Notify
ChengguangZhu	2007-12-08	purpose: 1.fix the bug on electron-jet removal 2.Use 2 steps on objects selection: preselection and final selection. procedure: 1.put pre and loose selection on e muon and jets before ttbar alg 2.preselect electron of all pt, abs(eta)<2.5, robust isolated , care about isolation in crack. robust loose isEM (check isEM on crack) 3.preselect muon of all pt, abs(eta),2.5, robust isolated, (isolation on crack), robust combined (depend on ID detector) 4.jets with pt>10GeV and abs(eta)<2.5, don't divide ljets and bjets 5.Set ntuple for preselected objects, not on all objects 6.do final and tight selection of them in ttbar alg 7.From preselected containers, use tight selection cut put them into new containers, set ntuple: nselelec nselmuon nseljets, nselbjets	openclosed			edit

Summanry of discussions:

• 2007 Dec/13 with p.p
  • Systematic error for w polarization:

1. from background: tau events and combinatoral background has diffirent shape compared to signal. the variation of 10% (change the quantity of the background level) of  tau event has no impact on the measurement (sn-2005), the cmbk is not valuated in that note, but its shape correlated with almost all object variation, means error from cmbk correlate with almost all other systematic errors 
2.from correction function: the error should be trivial, increasing the statistics of MC. This correction function (efficiency function later called) is taken for mixing signal and background. My logic is to substract the background from data and calculate the correction function with only signal. The argument is my method will bring more error? true. Is there any paper using the correction operate on mixed signal and background? I should understand how much error my method introduced.
3. more systematic errors: b jet miscalibration use variation of 3% but not inlcude the revaluating of etmiss.  etmiss miscalibration correlate with b jet calibraiton and (pile-up?),   etmiss resolution should be one more.

• • ttbar correlation

1. bias of A measurement too large, may not put in the CSC note.
2.truth info of A not readable. to look at MCNLO event if no-correlation data give correct resuts to check if my code is correct.

• 2007 Dec/14 with E.M
  • correction function for w polarization:

Correction function is applied on pure signal or mixed signal and background? parton level distribution of cospsi is only for pure signal, as

reco level distribution of data is mixed signal and background (tau event and cmbk), as In exist analysis, correction function is done unpon the 2 distribution above. The correction function is to correct the acceptance(cut etc..) and resolution effect. Same cuts operate on different physics channel, the acceptance is different, apparently, the selection efficiencies are different, background is rejected and signal is more reserved. cs should be as high as possible, and cb as low as possible. Assume: the MC same as data for signal and background, correct proportion. the upper correction function could be regarded as 2 parts: in parton level and , which is all data from parton level to reco level and . The points are signal in paton level has a expected shape on cospsi. tau events has a paton level shape too, cmbk and other background heavily denpends on the event reconstruction method, may has a specific shape or flat in parton level. We see that the distribution of them are totally different and correction function for each are different too. For a specific value of , for example, 0.2. In parton level, the quantity of events are

• • ttbar spin correlation A bias

To find why the bias are there: check code problem, check generator problem, ... should present in CSC note with a coclusion

Assigned to	Due date	Description	State	Closed	Notify
ChengguangZhu	2007-12-13	look for paper using correction funtion mixing bkgd and signal; check truth of MCNLO data for A and Ad	openclosed			edit

Plots (root code)

• Fake electron rate v.s. pt of electron:
  

Cospsi for w polarization

• • Fit cospsi of "ept>25G" data with old atlfast efficiency function (0.986879-0.415953*x+0.035223*x*x+0.507039*x*x*x):
    
  • with new atlfast efficiency function 0.975961+(-0.485711)*x+(0.105970)*x*x+(0.741176)*x*x*x from csc data 1206:
    
• with ptele>25GeV / truth --> efficiency function: 1.000415+(-0.432787)*x+(0.048137)*x*x+(0.764136)*x*x*x
  • without trigger: 1 F0 7.02706e-01 3.07945e-02 2 FL 2.96495e-01 1.72329e-02 3 FR 7.98366e-04 1.86842e-02
  • with L1 lepton trigger: 1 F0 7.03546e-01 3.21901e-02 2 FL 2.91245e-01 1.72329e-02 3 FR 5.20892e-03 1.95434e-02
  • with L2 lepton trigger: 1 F0 6.98176e-01 3.34059e-02 2 FL 2.90510e-01 1.86917e-02 3 FR 1.13142e-02 2.02790e-02
  • with EF letpon trigger: 1 F0 7.03270e-01 3.36210e-02 2 FL 2.91165e-01 1.88643e-02 3 FR 5.56531e-03 2.03344e-02

%ENDHIDE%

ttbar correlation A

• • ttbar correlation factor with old efficiency function: 1.09518-0.0719311*x-1.10887*x*x+0.42101*x*x*x+0.666338*x*x*x*x
    
  • with new atlfast efficiency function:1.095193+(0.236384)*x+(-1.191871)*x*x+(-0.255675)*x*x*x+(0.717809)*x*x*x*x
    
• with ptele>25GeV / truth -->efficiency function: 1.098214+(0.212162)*x+(-1.114520)*x*x+(-0.168282)*x*x*x+(0.416528)*x*x*x*x
  
  • without trigger: A: 0.794352 error: 0.176444
  • with L1 lepton trigger: A: 0.724979 error: 0.188435
  • with L2 lepton trigger: A: 0.723005 error: 0.193218
  • with EF letpon trigger: A: 0.692447 error: 0.19566
• with new data, see https://twiki.cern.ch/twiki/bin/view/Main/ZhuABias#Slide_1_Used_CSC_data

L1 trig:
correction fucntion: 1.079624+(0.149426)*x+(-1.221292)*x*x+(0.304894)*x*x*x+(1.180660)*x*x*x*x
A:  0.815048  error:  0.167271
L2 trig:
correction fucntion: 1.079624+(0.149426)*x+(-1.221292)*x*x+(0.304894)*x*x*x+(1.180660)*x*x*x*x
A:  0.731522  error:  0.170414
L3 trig:
correction fucntion: 1.079624+(0.149426)*x+(-1.221292)*x*x+(0.304894)*x*x*x+(1.180660)*x*x*x*x
A:  0.710314  error:  0.171958

%ENDHIDE%

Assigned to	Due date	Description	State	Closed	Notify
ChengguangZhu	2007-12-20	with trigger, A and Ad changes, to test what kind of events are triggered off, and have bias on A and Ad	openclosed			edit

ZhuABias

ZhuCSCSystematics

ttbar correlation Ad

• • ttbar correlation factor with old efficiency function: 0.999068-0.0661217*x
    
  • with new atlfast efficiency function: 0.978340+(-0.058341)*x
    
• with ptele>25GeV/ truth--> efficiency function: 0.976399+(-0.073418)*x
  
  • without trigger: Ad: -0.338016 error: 0.102039
  • with L1 lepton trigger: Ad: -0.317348 error: 0.107676
  • with L2 lepton trigger: Ad: -0.373814 error: 0.110614
  • with EF letpon trigger: Ad: -0.363767 error: 0.112203
• with new data, see https://twiki.cern.ch/twiki/bin/view/Main/ZhuABias#Slide_1_Used_CSC_data

L1_trig : 
correction fucntion: 0.986804+(-0.150406)*x
Ad:  -0.432209  error:  0.101379
L2_trig :
correction fucntion: 0.986804+(-0.150406)*x
Ad:  -0.409707  error:  0.104382
EF_trig :
correction fucntion: 0.986804+(-0.150406)*x
Ad:  -0.396277  error:  0.105264

%ENDHIDE%

Assigned to	Due date	Description	State	Closed	Notify
ChengguangZhu	2007-12-07	quality of the ttbar reconstruction; angle resolutions for spin correlations; resolution of cospsi PHI, PSI of pure signals, cmbks; use fit for PSI other than mean; distribution of Ad according ptele or l_pt; Fast measurement is corrected to truth value?; what is the differece beteen fast and full, which makes the full simulation so large deviation on measurement, can be corrected?	openclosed			edit

• ttbar spin correlation measurement value v.s. lepton pt
  • up: (-3)*cosphilep_lej/0.51:pt_l; right, project to pt_l;
  • down: (-9)*coslep*coshad_lej/0.51:pt_l, right: project to pt_l
    

• ttbar spin correlation measurement value v.s. lepton eta
  • up: (-3)*cosphilep_lej/0.51:eta_l; right, project to eta_l;
  • down: (-9)*coslep*coshad_lej/0.51:eta_l, right: project to eta_l
    

Parameter A with corrections

• for parameter A, correction from reconstruction efficiencya and e pt resolution is added.:
  
• logfile

root [0] .L method2_a.cxx
root [1] CheckObjects(1)
in Initialized() .....
in parAEfficiencyFunctionFromMC() .....
efficiency func for par A: 1.098214+(0.212162)*x+(-1.114520)*x*x+(-0.168282)*x*x*x+(0.416528)*x*x*x*x

in TruthParATest().....
The mean of coslep*cpshad_lej*(-9)/0.51 :0.332396 and error:0.0244047

in PerfectMCTest() .....
The mean of coslep*cpshad_lej*(-9)/0.51 :0.383894 and error:0.160355

in UnPerfectMCTest() .....
The mean of coslep*cpshad_lej*(-9)/0.51 :0.794352 and error:0.176444

in ReconstructionEffiencyDiffernece() .....


in ElectronPtResolutionDifference() .....

in ChooseCorrections() .....
all corrections: (((nselelec+nselmuon==1 && nselmuon==1)*m_reco_effi_diff_num(ptmuon_true,etamuon_true))+((nselelec+nselmuon==1 && nselelec==1)*e_reco_effi_diff_num(ptele_true,etaele_true)))*(((nselelec+nselmuon==1 && nselelec==1)*ElectronResolutionEffectDist(ptele_true))+(nselelec+nselmuon==1 && nselmuon==1))
in CorrectionsTransformToParameterA() .....
fit the correction to parameter A
Correction func for par A: 0.877674+(-0.015678*x)+(0.033812*x*x)
The mean of coslep*cpshad_lej*(-9)/0.51 :0.755641 and error:0.164871

Parameter Ad with corrections

• parameter Ad, correction with reco effi and e pt reso:
  
• logfile

root [0] .L method2_ad.cxx
root [1] CheckObjects(1)
in Initialized() .....
in parAEfficiencyFunctionFromMC() .....
efficiency func for par Ad: 0.976399+(-0.073418)*x

in TruthParATest().....
The mean of cosphilep_lej*(-3)/0.51 :-0.175966 and error:0.0143719

in PerfectMCTest() .....
The mean of cosphilep_lej*(-3)/0.51 :-0.166506 and error:0.0945965

in UnPerfectMCTest() .....
The mean of cosphilep_lej*(-3)/0.51 :-0.338016 and error:0.102039

in ReconstructionEffiencyDiffernece() .....


in ElectronPtResolutionDifference() .....

in ChooseCorrections() .....
all corrections: (((nselelec+nselmuon==1 && nselmuon==1)*m_reco_effi_diff_num(ptmuon_true,etamuon_true))+((nselelec+nselmuon==1 && nselelec==1)*e_reco_effi_diff_num(ptele_true,etaele_true)))*(((nselelec+nselmuon==1 && nselelec==1)*ElectronResolutionEffectDist(ptele_true))+(nselelec+nselmuon==1 && nselmuon==1))
in CorrectionsTransformToParameterA() .....
fit the correction to parameter A
Correction func for par A: 0.862251+(-0.007647*x)+(0.051789*x*x)
The mean of cosphilep_lej*(-3)/0.51 :-0.349955 and error:0.0948953