Three different triggers are employed to select hard electron, muon, and soft lepton events. Due to overwhelming QCD multi-jet backgrounds, a trigger which directly tags a soft lepton cannot be used, but alternatively we use a EmissT trigger to select soft lepton events. Also for the hard lepton triggers, not only a lepton trigger but also additional conditions are required so as to reduce the trigger rate. It is important to check if the trigger efficiencies are sufficient enough for the kinematic selections required in the following analysis.
Electron trigger : Electron trigger fires when an electron is found in an event. But it is suffered from the mis-firings from fake electrons and the trigger rate is overwhelmed by them. We require large EmissT at trigger level to reduce the background.
A trigger calledEF e24vh medium1 EFxe35 tclcwtags the events which have
• at least onemedium++quality electron withp.T>25 GeV,
• EmissT larger than 80 GeV.
EF e24vh medium1 EFxe35 tclcwhas a few % efficiency loss for large lepton pT. To compen-sate the efficiency loss,EF e60 medium1is added in the trigger menu, which fires for the electron with pT>60 GeV.
Figure 24 shows the electron trigger efficiency as a function of offline EmissT (corresponding to EmissT defined in Section 3.5). EmissT used in the trigger decision is different from the offlineEmissT : the trigger EmissT is based on a different calibration scheme and muons are not added up when calculating momentum balance. As a result, a slight topology dependence is expected. The dif-ference betweent¯t(red) andW+jets (blue) gives the rough size of the topology dependence. To select a similar topology as our Signal Regions, the following kinematic selections are applied:
mT>60 GeV,pTjet1>80 GeV,pTjet3>40 GeV and exactly 1 electron. The efficiency calculation is per-formed based on a method in which the events are selected by an orthogonal trigger and the trial trigger is tested on those events to measure the efficiency.
Trigger efficiency (trial trigger)= Number of events passing both the trial and orthogonal triggers Number of events passing the orthogonal trigger
(59) For the orthogonal trigger, an isolated electron trigger is used here.
The trigger efficiency becomes fully efficient for EmissT >80 GeV and the difference between data and Monte Carlo (tt,¯ W+jets) is below 1%, thus negligible.
[GeV]
miss
ET
0 20 40 60 80 100 120 140 160 180 200 220
Trigger Efficiency
0 0.2 0.4 0.6 0.8 1 1.2
Data t tW+jets
Figure 24: Trigger efficiency ofEF e24vh medium1 EFxe35 tclcwas a function ofETmiss.
Muon trigger : Muon trigger fires when a muon is found in an event. The muon trigger is also suf-fered from fake muons from QCD multi-jet background. To reduce fake trigger rate, we use a EmissT +muon+jet trigger:EF mu24 j65 a4tchad EFxe40 tclcw, which selects the events with
• at least one muon with pT>25 GeV,
• EmissT larger than 120 GeV,
• at least one jet with pTjet>80 GeV.
Figures 25 show the efficiency plots, which are measured in the same way as the Electron case with an isolated muon trigger being used for the orthogonal trigger. The left plot shows EmissT depen-dence and the right plot shows jet pT dependence. For the left plot, exactly 1 muon,mT>60 GeV, pTjet1>80 GeV andpTjet3>40 GeV are required, while for the right plot, exactly 1 muon,mT>60 GeV andEmissT >150 GeV are required. The trigger is fully efficient after requiringEmissT >120 GeV and pTjet1>100 GeV, where the difference between data and Monte Carlo is below 1%.
[GeV]
miss ET
0 20 40 60 80 100 120 140 160 180 200 220
Trigger Efficiency
0 0.2 0.4 0.6 0.8 1 1.2
Data t tW+jets
[GeV]
jet1 pT
20 40 60 80 100 120 140 160 180 200
Trigger Efficiency
0 0.2 0.4 0.6 0.8 1 1.2
Data t t W+jets
Figure 25: Trigger efficiencies ofEF mu24vh j65 a4tchad EFxe40 tclcwas functions ofETmiss(left) andpTjet(right).
EmissT trigger : Lepton pT thresholds of the lepton triggers are too high to tag events in the soft lepton analysis. Instead, triggers based onEmissT are used. Two different triggers are combined since the lowest un-prescaled triggers were changed during 2012 data taking period, which are summarized in Table. 8. SinceEF xe80T tclcw loosetrigger does not use the first three bunches in a bunch train, the luminosity becomes slightly lower for the trigger. The loss is 10%, so a scaling factor of 0.9 is multiplied to the first 2.1 fb−1of data, leading to an luminosity of 1.9 fb−1. As a result, the integrated luminosity for the soft lepton analysis becomes 20.1 fb−1.
For both of the triggers, the efficiencies reach to a plateau after requiring the following condition,
• EmissT larger than 200 GeV.
Since no serious discrepancy is observed between data and Monte Carlo above this threshold and no significant topology dependence is found, no further corrections or the uncertainty is introduced for this trigger.
Figure 26 shows the trigger efficiency curve ofEF xe80 tclcw loosewith the following condi-tions: exactly 1 electron,mT>60 GeV, pTjet1>80 GeV and pTjet3>40 GeV. An isolated electron trig-ger is used for the orthogonal trigtrig-ger.
The triggers mentioned above are summarized in Table 9.
[GeV]
miss
ET
0 50 100 150 200 250 300
Trigger Efficiency
0 0.2 0.4 0.6 0.8 1 1.2
Data t tW+jets
Figure 26: Trigger efficiencies ofEF xe80 tclcw looseas a function ofEmissT .
Trigger delivered lumi. rescaled lumi.
EF xe80T tclcw loose 2.1 fb−1 1.9 fb−1 EF xe80 tclcw loose 18.2 fb−1 18.2 fb−1
Table 8: The lowest un-prescaled triggers and their delivered and rescaled luminosities.
Channel Trigger Luminosity Comments
Electron EF e24vh medium1 EFxe35 tclcw 20.3 fb−1 Fully efficient with : orEF e60 medium1 Electron pT>25 GeV
EmissT >80 GeV Muon EF mu24 j65 a4tchad EFxe40 tclcw 20.3 fb−1 Fully efficient with :
Muon pT>25 GeV
ETmiss>120 GeV at least 1jet withpTjet>80 GeV Soft Lepton EF xe80T tclcw loose 20.1 fb−1 Fully efficient with :
orEF xe80 tclcw loose ETmiss>200 GeV.
Table 9: Summary of the triggers. The “Comments” show the minimum offline selection needed in the analysis.
4 Data and Monte Carlo simulation
4.1 Data samples
This thesis uses the data recorded by the ATLAS detector in proton-proton collision at the center-of-mass energy of √s=8 TeV. These events were collected from April 5th to December 6th in 2012. The data taken in the period yields a total integrated luminosity of 21.3 fb−1. To assure a good data quality, all detectors are required to be turned on and working without a fatal problem. The detector condition is recorded for each LumiBlock and the LumiBlocks flagged as problematic are eliminated, leaving an integrated luminosity of9
L dt = 20.3±0.6 (2.8%) fb−1. Figure 27 shows the LHC delivered, ALTAS recorded, and good-for-physics integrated luminosities.
Day in 2012
-1fbTotal Integrated Luminosity
0 5 10 15 20 25
1/4 1/6 1/8 1/10 1/12
= 8 TeV s
Preliminary ATLAS
LHC Delivered ATLAS Recorded Good for Physics
Total Delivered: 22.8 fb-1
Total Recorded: 21.3 fb-1
Good for Physics: 20.3 fb-1
Figure 27: LHC delivered (green), ATLAS recorded (yellow), and good-for-physics (blue) luminosities.
The x-axis shows the date in 2012.
4.1.1 Luminosity measurement
The ATLAS detector has several algorithms and sub-detectors [32] to monitor the luminosity. They mostly count the number of reference eventsµvisand calculate the luminosity using the cross-section of the eventσvisas
L= µvisnbfr
σvis . (60)
The unknown parameterσvis is calibrated beforehand by an absolute luminosity measurement,van der Meer(vdM) scan. vdM scan, or often referred to as beam separation scan, measures spacial proton density profile of bunches by shifting the impact offset between two colliding beams. Also the absolute number of protons in a bunch is measured by monitoring the beam current at the same time. With these information, the absolute luminosity is calculated as
L= nbfrn1n2
2πΣxΣy
, (61)
wheren1,n2are the number of protons contained in two colliding bunches andΣx,Σyare the root-mean-square of transverse proton distribution.
By equating Eq. 60 and Eq. 61,σvisis obtained as, σvis=µvis2πΣxΣy
n1n2 . (62)
Here we have only measurable values in the equation and henceσvisis determined.