In order to search the FCNC signal which contains one b quark in the pattern, the b-tagging technique is used for identifying if the jet is originated from a b quark (short as b jet).
This identification exploited the relatively long life time of b quark, which makes the vertex of a b jet shift from the primary vertex. Therefore, high spatial resolution track is of great importance.
In this analysis, the Combined Secondary Vertex method is used for b jet iden-tification. First, the charged tracks in a jet are collected, and then the Trimmed Kalman Vertex Finder is exploited which selects the outliers of the tracks that can be used for secondary vertices reconstruction. Then the reconstructed secondary ver-tices’ parameters and the tracks’ parameters will be combined as a discriminating variable.
Figure 3.8: The discriminating variable of CSV for different types of jet. The solid, dotted, and dashed lines are the b, c and light quarks respectively.
Chapter 4
Data and Monte Carlo Samples
The data used in this analysis is collected with CMS detector from the LHC de-livered 2012 8 TeV proton-proton collision. The double electron and double muon triggered data are used with corresponding integrated luminosity of 19.5 f b−1. With the help of computing elements, it is also possible to generate the events which sim-ulate the proton collisions and the detector response. In event generation, the first step is the parton level generation carried out by MadGraph 5[6] event generator, which performs the tree level calculation of a physical process using the helicity method[7]. Besides the process in standard model, MadGraph is also capable of doing calculation through a user-defined model. Then the output will be sent to PYTHIA 6[8] generator for the fragmentation and hadronization. TAUOLA[9] is also used for the emulation of tauon decay. The generated event will be sent to the GEANT 4 detector simulator which simulates the detector response. The first step is called generation, and the second step is called simulation. We refer to this sample as MC sample.
The MC sample is generated separately of different processes, while in data all
kinds of processes are mixed. Therefore, if we combine all the MC sample, it is possible that it can represent the result in data, and we can further estimate the fractions and distributions of different processes with different physical variables.
Also, with MC sample we can determine the proper way to separate the signal process from the background and can have much higher statistics, for the event number in data is limited by the apparatus.
The FCNC MC signal MC sample contains the ttbar process where one of them decays to q (which may be c, cbar, u and ubar) and Z, while the other decays to b (or bbar) and W. Then the Z and W bosons decay to leptonic final states. Setting the branching ratio of this process to be 0.1%, we can then calculate the cross-section of this sample to be 0.047 pb. The sample contains equal amount of t to Zc and t to Zu events.
The other standard model background included in the study are the W, Z, WW, WZ, ZZ, ttbar, ttbarZ, ttbarW, and single top. Among them, the major contribution comes from WZ and ttbarZ events which have very similar pattern comparing to FCNC in the desired final states. The table below summarize the sample used in this analysis.
4.1 Pile-Up Treatment
The LHC collisions are of extremely high luminosity. In 2012, there are often more than 20 interactions per event; although these events are of relatively small momen-tum transfer, the resulting energy deposit will be non-negligible. Therefore, on one hand, it is of great importance to model the pile-up effect in the MC simulation, while on the other hand, it is very crucial to reduce noise from pile-up effect.
Data Int. Lumi. (pb−1) DoubleElectron(Muon) Run2012A-13Jul2012-v1 808.47 DoubleElectron(Muon) Run2012A-recover-06Aug2012-v1 82.14 DoubleElectron(Muon) Run2012B-13Jul2012-v1 4428.67 DoubleElectron(Muon) Run2012C-24Aug2012-v1 495.00 DoubleElectron(Muon) Run2012C-PromptReco-v2 6401.67 DoubleElectron(Muon) Run2012D-PromptReco-v1 7273.70
Total 19489.65
Table 4.1: The data used in this analysis and the corresponding integrated luminos-ity.
We use the number of interaction to describe the pile-up effect. For an event with high number of interaction, the pile-up is severe, while for an event with low number of interaction, the pile-up is not significant. The number of interaction is a probability function and in the 2012 Run period A, the most probable value is around 15. To model the pile up effect in the MC sample, the general approach is overlapping the low energy interactions several times with some shifts on the hard process. The probability function of the number of interaction varies with the luminosity; therefore, one first assign a probability function similiar to but not exactly the same as the distribution in data. Then for those MC events with a certain number of interaction, it is reweighted with the probability ratio of data over MC of that number of interaction.
Figure 4.1: An event after reconstruction. The green lines are the tracks of charged particles and the yellow dots are the vertices.
Figure 4.2: The pile-up distribution of the Run2012A data and the distribution assigned for MC production in 2012 summer. The division of these two gives the proper weights for different number of interaction.
Sample Cross Section (pb) Calculation Order
FCNC 0.047 assumed
TTJets 234 approx. NNLO
WJetsToLNu 37509 NNLO
DYJetsToLL 3503.7 NNLO
WWJetsTo2L2Nu 5.817 NLO
WZJetsTo2L2Q 2.467 NLO
WZJetsTo3LNu 1.189 NLO
ZZJetsTo2L2Nu 0.776 NLO
ZZJetsTo2L2Q 2.713 NLO
ZZJetsTo4L 0.213 NLO
T s-channel 3.79 approx. NNLO
Tbar s-channel 1.76 approx. NNLO
T t-channel 56.4 approx. NNLO
Tbar t-channel 30.7 approx. NNLO
T tW-channel 11.1 approx. NNLO
Tbar tW-channel 11.1 approx. NNLO
TTWJets 0.232 NLO
TTZJets 0.174 NLO
TBZToLL 0.0217 NLO
Table 4.2: The MC sample used in this analysis
Chapter 5
Event Selection
In this analysis, we look for the top pair events with one decaying to W+b and the other decay to Z+q, where q may be a charm or up quark, and W, Z decays to electrons or muons. Therefore, based on the decay channels of W, and Z, the search can be divided into four different channels: eee, eeµ, eµµ, µµµ. The tauon channel is not considered due to the small reconstruction efficiency.
The event selection steps can be summarized as below:
1. Choose events that fire double electron or double muon high level trigger (HLT).
2. Select good vertex, electrons, muons, and jets.
3. Reconstruct Z candidate with two oppositely charged same flavor leptons.
4. Reconstruct W candidate with the addition charged lepton and MET confining W’s mass.
5. Veto events with more than three good leptons.
6. Veto events with less than two jets.
7. Choose events with exactly one b-tagged jet.
8. Reconstruct top pair, one using Z+q and the other using W+b.
9. Select the pair with maximum transverse open angle.
10. Select the top pair in the top mass window.
5.1 High Level Trigger
The high level trigger used in this study are
1. HLT Ele17 CaloIdT CaloIsoVL TrkIdVL TrkIsoVL Ele8 CaloIdT CaloIsoVL TrkIdVL TrkIsoVL
2. HLT Mu17 TkMu8
3. HLT Mu17 Mu8
These triggers selects two leptons with one pt > 17 GeV and the other pt >
8 GeV as well as other calorimeter and tracker selection. For the double muon triggers, if either one fires or both, the event will be selected for further analysis.
The further selection in the analysis are more tighter to avoid any effect from HLT.
These trigger efficiency, which is defined as the number of di-lepton events that pass our preliminary selections and also fire the trigger over the total number of di-lepton events that pass our preliminary selections, are measured to be 99%, 98%, 91% and 93% for eee, eeµ, eµµ and µµµ channels, respectively. To prevent any bias from the trigger, we did this measurement by using the multi-jet triggered event which we assume to be non-relevant to lepton information. In these events, we look for events that has two same flavor leptons (electron or muon) that pass our lepton
and Z selections (the preliminary selections), then, the efficiency can be obtained by dividing the number of these events firing the di-lepton triggers over the total number of these events.