%CERTIFY%

PileUpInDileptonTop

Introduction

While the exact beam conditions for first data are unknown, in order to deliver the 200pb-1 assumed in the pub note luminosities must approach 10^32. At this luminosity pile-up effects become important, and will be greatest at a bunch spacing of 450ns where on average there will be 4 interactions per bunch crossing. The following documents an effort to assess the effect of pile-up at this beam condition on the the top dilepton cross section. All samples have been simulated with 14.2.25.8, which does not take into account the average energy shift in CaloCell for pileup samples with 450ns. These results therefore might slightly underestimate effects.

Results are not finalized, and might change frequently

Strategy

Investigate the effects of Pile-up on each cut to determine an uncertainty in final acceptance. Since background events simulated with Pile-up are limited, we can start with all leptonic TTbar decays to get an initial estimate of pile-up effects.

Numbers are normalized 100pb-1 All units are in MeV (Except for multiplicity of course)

Cuts

see TopDileptonPubNote for definitions

Dataset for TTbar:mc08.105200.T1_McAtNlo_Jimmy.recon.AOD.e357_s462_r635/

Dataset for TTbar with Pileup:mc08.105200.T1_McAtNlo_Jimmy.recon.AOD.e357_s462_d150_r642/

Number of Events w/o Pile-up: 69983 Number of Events w/ Pile-up: 75571

Trigger

A slight decrease in the trigger rates is seen.

For all events

Trigger No Pile-up With Pile-up Percent Change
e15_medium 6227 +/- 50 6126 +/- 47 -1.6%
mu15 7787 +/- 57 7516 +/- 54 -3.5%

Two Os Leptons

Average Lepton Multiplicity

Investigate how pile-up changes electron reconstruction by looking at the multiplicity of selected leptons.

Top Decay Channel (From Truth) No Pile-up no Isolation Cut With Pile-up no Isolation Cut No Pile-up w/ Isolation Cut With Pile-up w/ Isolation
Electrons   # of Reco Electrons found in an event
l+l .46 .42 .46 .42
e+e .91 .85 .91 .85
e+mu .47 .43 .47 .43
mu+mu .00 .00 .00 .00
Muons # of Reco Muons found in an event      
l+l .56 .56 .56 .56
e+e .00 .00 .00 .00
e+mu .55 .54 .55 .55
mu+mu 1.12 1.12 1.12 1.12

To 1% the letpon id does not change even with an isolation cut.

Isolation Energy of selected Electrons

Investigating isolation a bit further we can see a small shift in the isolation cone energy, but the suggested cone (.2) is small so there is little changed in the above numbers.

  • All Selected Electrons:
    Reco_El_all.jpg

  • Electrons by Origin:
    Reco_El_type.jpg

Isolation Energy of Selected Muons

  • All Muons:
    Reco_Mu_all.jpg

  • Muons by Origin:
    Reco_Mu_type.jpg

MET

MET also changes very little

MET for all events in the examined datasets. MET.jpg

Two Jets

The following is the jet multiplicity for signal ttbar and background ttbar after the above cuts have been applied.

Signal: All dileptonic decay modes not including a tau decaying hadronicly Background: All other leptonic top decays. (No fully had.)

  • Jet_Multi_AfterCuts.jpg:
    Jet_Multi_AfterCuts.jpg

Overall Cut Flow

Cut No Pileup With Pileup Fraction of Previous Line with Pileup
Events 21706
2l 875 811 .040 .037
Met 755 704 .86 .87
2Jets 624 611 .83 .87
Z veto 601 592 .96 .97
Trigger 584 580 .97 .98

The only significant change is an increase in the efficiency of the 2jet cut.

Estimate of pile-up's effect on background

The most noticeable effect of pile-up is an increase in the jet multiplicity of events. This can strongly affect the acceptance of background events, particularly in events with a leptonicly decaying Z which would normally be removed by the 2 jet requirement.

Currently no ideal samples are ready for a study of this effect. The closest two are:

valid1.105144.PythiaZee.recon.AOD.e380_s492_r610 no Pile-up

valid1.105144.PythiaZee.recon.AOD.e380_s494_d150_r621 with Pile-up

valid1.105145.PythiaZmumu.recon.AOD.e380_s492_r610 no Pile-up

valid1.105145.PythiaZmumu.recon.AOD.e380_s494_d150_r621 with Pile-up

These data sets include "features" like a low Z invariant mass, and are not large enough to be statistically significant after a MET cut is applied. However, if we assume that jet-multiplicity is not correlated to MET we can get a first estimate on how background distributions change.

Start by looking at the Jet Multiplicities of Z->ee and Z->mumu with and without pile-up.

The following graphs are the normalized jet multiplicities of the above data sets after a 2os lepton selection and a pt cut of 20*GeV.

Zee_Jet_Multi_20.jpg

There is a very noticeable shift in jet multiplicity. Lets use this shift to calculate the systematic.

Fitting the Effect of Pile-up

The effect of pile-up can be measured directly from the above distributions if we make the assumption that pile-up's only effect will be to add additional jets, and it does this with a constant probability (This appears to hold well for Z+Jets). The probability of this can be measured as follows.

1.The total probability (P) of an event having any number of additional jets can be found by looking at the shift in the 0jet bin. Since events can only be removed from this bin, the difference in total events between pile-up and normal samples (properly scaled) is the probability.

2. The probability of 1jet being produced can be found by P1=((#PileUp)-(#No Pileup)*(1-P)). The (1-P) factor essentially removes all events from the original bin that would have been shifted, and the excess that remains must have come from events shifted by one jet from the zero bin.

3. P2 can be measured the same way in the second jet bin. P2=((#PileUp)-(#No Pileup)*(1-P))-P1*(Bin1) where now we have to subtract out events that were shifted by one jet from the one jet bin. The events that are left are those shifted by 2 from the zero jet bin

4. P3 P4 etc. can be measured in the same fashion by taking into account more and more bins.

Results

P Total=Total Probability of an event have 1 or more extra jets

P1=Probability of an event having exactly 1 additional jets

p2=Probability of an event having exactly 2 additional jets

Process P Total P1 P2 P3 P4
Z->ee 15% 10% 4% 1% 0 0
Z->mumu 21% 12% 6% 1% 1%

The effect on Muon is believed to be a little bit higher because there is no overlap region were jets are removed, so each muon event has a larger phase space.

Plan:Use these probabilities to correct background histograms. Recalculate cross section, take shift as the systematic.

Quick Method

Assume background distributions are similar, and adjust the 2-jet bins proportionately.

This simplest way to do this is to use the relative weights in the 2-jet bins. (Higher bins are small, Lower bins do not pass the selection).

P=2 jet bin with pile-up N=2 jet bin without pile-up

For Zee

P/N=2.0 +/- .3

For Zmumu

P/N=2.6 +/- .3

These numbers should be the same, and the error bars only just barely overlap which may be an indication that this treatment is insufficient, but for now take the average.

P/N=2.3 +/- .2

Now, I assume that this number holds over all channels, and after the missing ET cut. While this assumption presumably has large errors it can give a rough estimate of the effect of pile-up.

Using Akira's numbers from http://indico.cern.ch/getFile.py/access?contribId=3&resId=0&materialId=slides&confId=55641

After all cuts the most significant backgrounds follow

Background No Pile-up with Pile-up
Wt 6.50 14
WW 2.59 4
Ztautau 7.26 14
Wmunu 6.31 12
Wtaunu 1.15 2
Ttbar 16.83 17
Total 42.3 63
Signal 226.5  

No Pile-up: Sqrt(S+B)/S=7.2%

With Pile-up: Sqrt(S+B)/S=7.5%

Note: Total for background without pile-up is given to the accuracy in talk. (Which is why it differs from the straight sum)

Note: As seen above (in the signal section) the background from TTbar does not shift greatly compared to Z+jets.

Larger PT cuts on jet multiplicities are one of many options to improve the background from pile-up effects.

Zee_Jet_Multi_50.jpg


assumbly- Major updates:
-- JacobSearcy - 11 May 2009

%RESPONSIBLE% JacobSearcy
%REVIEW% Never reviewed

Edit | Attach | Watch | Print version | History: r13 < r12 < r11 < r10 < r9 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r13 - 2009-11-20 - PatrickJussel
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Atlas All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback