Higgs Boson Production via Gluon - Gluon Fusion And Decay Through: H → WW → lνlν Cutbased Analyses Of The Run Two ATLAS Data

(1)

Higgs Boson Production via Gluon - Gluon Fusion

And Decay Through: H → WW → lνlν

Cutbased Analyses Of The Run Two ATLAS Data

Kars Huisman 10366156

8th July 2016

Supervisors: Lydia Brenner and Wouter Verkerke

Second Assessor: Ivo van Vulpen

Faculty of Natural Sciences, Mathematics and Computer Sciences, University of Amsterdam

Paper of Bachelorproject Natuur- en Sterrenkunde, size 15 EC, executed between 30-03-2016

and 08-07-2016 at NIKHEF

(2)

Abstract

In this paper the likelihood (meaning likely or unlikely) of Higgs particle production via gluon-gluon Fusion and decay via H → WW → lνlν in the currently available (10% of projected total) run 2 data of the ATLAS detector is being investigated. For this purpose the data to Monte Carlo background (data/bkg) ratio is used as a distinguishing variable. Cuts were performed on different variables to optimize the expected signal to Monte Carlo background ratio and a control region for the WW background was made to calculate a normalization factor. This resulted in a normalized data/bkg ratio of 0.99 ± 0.05. The data/bkg ratio can be smaller than, equal to or bigger than 1, therefore it is not possible to say if the Higgs particle is being produced in run 2 data and decays via the mentioned channel.

(3)

Samenvatting

De theoretici Peter Higgs, Fran¸cois Englert en Robert Brout voorspelde in 1960 dat er een quantum veld moest bestaan waarin het later naar hem vernoemde Higgsdeeltje een excitatie is. Dit veld zorgt ervoor dat deeltjes massa hebben. In 2012 werd dit deeltje waargenomen in de LHC - detectoren (ATLAS & CMS) in CERN. Het Higgsdeeltje werd in de run 1 data gevonden. In dit onderzoek wordt de run 2 data van de ATLAS detector geanalyseerd om te kijken of het waarschijnlijk of onwaarschijnlijk is dat het Higgsdeeltje geproduceerd is. Het verschil tussen run 1 en 2 is dat de energie waarmee de protonen in de LHC op elkaar botsen groter is; 13 Tera Elektronvolt (TeV) voor run 2 t.o.v. 7/8 TeV voor run 1, het aantal deeltjes dat per oppervlakte eenheid botst kleiner is (ge¨ıntegreerde luminositeit) 3.2 inverse fentobarn (fb−1) voor run 1 t.o.v. 4.8/5.8 fb−1voor run 2 en dat de tijd tussen twee opeenvolgende protonenbundels kleiner is 25 nanoseconden t.o.v. 50 nanonseconden.

Om het Higgsdeeltje te vinden wordt er eerst gekeken naar karakteristieke vervalproduc-ten voor Higgsverval. In dit onderzoek zijn dat elektronen en muonen. Bij dit proces vervalt het Higgsdeeltje volgens de volgende schematische reactie: H → WW → lνlν. Het Higgsdeeltje vervalt hier naar twee W deeltjes met tegenovergestelde lading. Elke W deeltje vervalt vervolgens naar een elektron of muon (l) met een neutrino (ν). Zo kunnen er verschillende paren van vervalproducten gemeten worden: ee , em , me of mm -paren (e staat voor elektron en m voor muon).

Er vinden in de LHC ook vervalsreacties plaats die niet van een Higgs deeltje komen, maar wel deze karateristieke vervalproducten geven. Deze vervalreacties zijn achtergrond. Om zo veel mogelijk paren over te houden die van Higgsverval afkomstig zijn (signaal) wordt er signaal bewaard en achtergrond weggesneden. Dit proces van wegsnijden en bewaren wordt ookwel het uitvoeren van ’cuts’ genoemd. Zo houd je namelijk steeds meer signaal en steeds minder ruis over.

Het doel van dit onderzoek is om te kijken of het waarschijnlijk of onwaarschijnlijk is dat het Higgsdeeltje aanwezig is in de run 2 data. Dit wordt onderzocht door de ho-eveelheid gemeten paren (data) te vergelijken met wat er verwacht wordt aan gemeten paren (achtergrond) als er geen Higgs deeltje zou zijn. Er wordt hierbij gekeken naar de verhouding tussen deze twee getallen (data/achtergrond). Als er meer gemeten wordt dan verwacht dan is deze verhouding groter dan één. Zo niet dan is deze verhouding kleiner of gelijk aan één. In het eerste geval betekent het dat het waarschijnlijk is dat het Higgsdeeltje geproduceerd wordt in run 2 en in het tweede geval niet. Na het uitvoeren van cuts blijkt dat data/achtergrond verhouding 0.99 ± 0.05 is. Deze verhouding kan dus groter, kleiner of gelijk zijn aan één, daarom kan er nog niet gezegd worden of het waarschijnlijk of onwaarschijnlijk is dat het Higgs geproduceerd wordt in run 2.

(4)

1 Introduction

In 2012, the ATLAS (A Toroidal LHC ApparatuS) and CMS (Compact Muon Solenoid) collaborations detected the Higgs particle with a mass of 126.0 ± 0.4 (stat) ± 0.4 (sys) GeV/c2 _{[5] and 125.3 ± 0.4 (stat) ± 0.5 (sys) GeV/c}2 _{[7] respectively in run 1 of the}

Large Hadron Collider (LHC). In this paper, the currently available run 2 data (10% of the projected total) of the ATLAS detector will be analyzed. Run 2 differs from run 1 in terms of the available energy at which the proton-proton collisions occur in the ATLAS detector, 13 TeV instead of 7/8 Tev, in terms of integrated luminosity (events per meters squared), 3.2 fb−1instead of 4.8/5.8 fb−1and the spacing between the bunches is smaller 25 nanoseconds instead of 50 nanoseconds [3],[6].

The Higgs particle can be produced via a process called gluon-gluon Fusion (ggF). In order to find the Higgs particle one needs to look at signature decay products (electrons, muons and neutrinos). These particles come from the following Higgs decay reaction: H → WW → lνlν (See Chapter 2.3 for more). Here a Higgs boson decays to two oppositely charged W particles, which in turn decay to two oppositely charged leptons with associ-ated neutrino. Each lepton can be an electron or a muon.

The goal of this paper is to investigate if it is likely the Higgs particle is being produced via ggF and ultimately decays to two oppositely charged leptons in the run 2 data of the ATLAS detector. To determine if it is likely or unlikely the data to Monte Carlo (MC) background (data/bkg) ratio will be used as distinctive variable. In this paper cuts will be performed to optimize the expected signal to Monte Carlo background ratio (SBR) and the cuts will be compared to the cuts of run 1. Because the amount of run 2 data is small it is not expected it is likely the Higgs particle is being produced. Therefore, an advanced statistical analyses will not be done.

(6)

2 Theory

In this section the theory behind the origin of the Higgs particle, the composition of the ATLAS detector and the Higgs production reaction and Higgs decay channel is and some of the most important variables are is going to be explained.

2.1 Electroweak Symmetry Breaking in the Standard Model

All the standard model (SM) particles (fig. 1), the building blocks of matter, have mass, although the SM that is being described by symmetry theory does not predict these particles to have mass. The mechanism that allows W and Z particles to have mass is called the Brout-Englert-Higgs mechanism. This mechanism allows spontaneous elec-troweak symmetry breaking (EWSB) when the vacuum expectancy value is bigger than zero. It introduces a scalar field in which the Higgs-particle is an excitation. Spontaneous symmetry breaking allows all fermions to have mass, but this process is different than the process that gives W and Z particles mass because it involves an interaction between the Higgs - and the fermion - field [1].

Figure 1: The SM of elementary particles, the numbers in the squares are the mass, charge and spin of the particle( reading from top to bottom) [9].

(7)

2.2 ATLAS Detector

As mentioned in the introduction the search for the Higgs particle will start by looking at specific decay products; the before mentioned leptons and associated neutrinos. In this section there will be explained which parts of the ATLAS detector can detect these decay products and which quantities these parts can measure.

The ATLAS detector (fig. 2) has multiple layers of detectors which can reconstruct the interaction point, mass, charge and energy of a particle. The particles are identified via the amount of energy they deposit in the calorimeters and via the track they leave in the different detectors. The tracking chamber determines the trajectory, charge and mass of charged particles and reconstructs the primary and secondary vertices of these particles.

Figure 2: Schematic picture of the ATLAS detector and its components [10].

The second and third layer are calorimeters for electromagnetic (photons, electrons) and hadronic particles (protons, neutrons, pions) respectively. In fig. 3 the track that each particle leaves in the different parts of the detector and where they initiate a shower (de-posit their energy), is shown. The outer layer (muon spectrometer) can detect muons by the track they leave in this detector. This track can only be associated with muons, because other particles don not leave a track (neutrinos), are stopped by the rest of the detector (electromagnetic and hadronic particles) or have decayed before they can reach the muon spectrometer (τ - particles). Particles that can not be detected directly are tau particles and neutrinos. Tau particles live too short to reach the detector, they can be de-tected by looking at signature decay products (electrons, muons and hadrons). Neutrinos do not (or weakly) interact with the detector and can be detected by looking at missing transverse energy [1].

(8)

Figure 3: Schematic picture of the different detectors inside the ATLAS detector, the yellow line is the trajectory a particle leaves in the detector if the line splits up that means the particle decays [2].

At the LHC there are bunches of protons which are being collided with each other. One bunch contains in the order of 1011 _{protons [6] and the spacing between the bunches is}

25 nanoseconds [3]. This causes a lot of collisions to occur in a short time-span with a lot of decay products. In order to find decay products which are likely to come from Higgs-particle decay, a process called triggering is used. This means that events that meet certain selection criteria are saved. Events that do not meet the criteria are ignored.

One of the problems that detecting a large amount of particles gives, is pile-up. This simply means that it is hard to identify from which event a particle originally came. This can be due to the read-out time of a detector, which takes too long to register and therefore misses another particle hitting it, or it can be that two particles which are not related hit a detector simultaneously and the detector sees it as one high energy particle [6]. It can also be that particles coming from cosmic radiation are measured by the detector, this is accounted for by measuring what particles hit the detector when the LHC is turned off [4].

2.3 Higgs - production and - decay

There are four main Higgs production modes. These are, in order of importance, gluon -gluon Fusion (ggF), Vector Boson Fusion (VBF), Higgsstralung and production via two top or bottom quarks (fig. 4). In this paper the ggF production mode will be investigated. From fig. 4 it can be seen that there are no jets in the ggF diagram (VBF has 2).

(9)

Figure 4: Feynman diagrams of the four main Higgs productions. a) Gluon-gluon fusion , b) Higgs-stralung , c) Vector-Boson Fusion , d) production via two quarks [1].

The produced Higgs particle can further decays to two oppositely charge W bosons (fig. 5). From the two W bosons two oppositely charged leptons with associated neutrinos are pro-duced. Each lepton can be an electron or a muon. The branching ratio, the probability a particle decays via a certain mode, of the H → WW is 22% and of WW → lνlν is 10.5%, which makes it likely that the expected signal to Monte Carlo background ratio (SBR) is strong enough to detect the Higgs boson. Because of spincorrelations the angle between the created leptons, ∆Φll is relativly small [6].

The backgrounds for ggF are WW, VV (WZ,Wγ, Wγ* and ZZ), Drell-Yann (DY),t¯t, Wt, Z+Jets, W+jets and QCD, where WW is the biggest background [6]. For the Feynman diagrams of these backgrounds see Appendix[A]. All these backgrounds can decay to the final state of 2 leptons with associated neutrinos. But these backgrounds do not come from a decaying Higgs particle, so in order to find the Higgs particle these backgrounds need to be reduced or if possible cut away completely. This can be done by applying cuts (see Chapter 3.2).

The available energy in the rest frame of the Higgs particle (fig. 5) is 125 Gev. Assuming, for simplicity, that the W particles have a small velocity, the mass of the W particles need to add up to 125 Gev. This can only be the case if the W particles are created off-shell, meaning the mass of these particles can be larger than, smaller than or equal to the most likely rest mass of 80.2 Gev/c2_{. This will lead to a W particle with a high mass}

(10)

Figure 5: Decay mode from H → WW → lνlν. The Higgs boson decays to two oppositely charged W bosons which in turn decay to two oppositely charged leptons with associated neutrino. Each lepton can either be an electron or a muon

and one with a low mass. High mass W particles are more likely to decay to muon and low mass W particles are more likely to decay to electrons, because the available energy is higher and lower respectively. Therefore it is expected the amount of electron-muon, muon-electron final states are more likely than muon-muon, electron-electron final states. For this reason this paper will only investigate the final states consisting of electron-muon and muon-electron pairs.

2.4 Variabels

The following variables will be mentioned further on and are therefore listed below:

• Njet: The number of jets.

• Mtt: Invariant mass of a dilepton system containing two tau particles.

• Emiss

T : Missing energy in the transverse direction (MET); this is the plain

perpen-dicular to the direction of the beam.

• Pleadlepton_T : Transverse momentum of the lepton with the highest momentum (lead lepton) of the two leptons.

• Psubleadlepton_T : Transverse momentum of the lepton with the lowest momentum (sub-lead lepton) of the two leptons.

• Pll

T: Transverse momentum of the dilepton system: The sum of the transverse

mo-mentum vectors of the leading and sub-leading leptons.

• ∆Φll: Angle between the two leptons in the transverse plain (fig. 6).

• ηleadlepton_{: This quantity (eta) transforms according to formula (1) [8] and says}

(11)

direction of the beam (fig. 7). This variable is also known as pseudorapidity. • METDPhill: Angle between the dilepton - and the dineutrino - system in the

trans-verse plain (fig. 6).

• Mll: Invariant mass of the dilepton system.

• Mleadlepton_T : Transverse mass of the leading lepton. • Msubleadlepton_T : Transverse mass of the subleading lepton.

• MT: Transverse mass, invariant mass in the transverse direction of all particles. The

transverse mass is reconstructed via formula (2) and (3) [6].

η = − ln[tan(θ/2)] (1) MT = q (Ell T+ E miss T )2− Pll T + E miss T 2 (2) where E_Tll= q Pll T 2 + M2 ll (3)

(12)

Figure 6: A schematic drawing of the direction perpendicular to the beam and perpendicular to the direction of motion. The beam is the black dot in the middle and its going into the page, the angle between the two lines is the angle φ

Figure 7: A schematic drawing of the direction perpendicular to the beam and parallel to the beam’s direction. The beam direction is indicated by the black arrow. The angle between the dotted line and the direction of the beam is θ. In this drawing the dotted line is the direction a particle is moving in.

(13)

3 Method

3.1 Cuts

In order to find the Higgs particle the analyses focused on optimizing the expected signal to Monte Carlo background ratio (SBR) by performing (pre-selection) cuts on the vari-ables mentioned in section 2.4 except the transverse mass. The goal of the cuts was to maximize the SBR (meaning SBR is high as possible). This involved an iterative process where cuts were performed, the results were looked at and new cuts were made based on those results. The cut - based analyses was done on the stoomboot cluster at NIKHEF. The cutfile-code that was used for the (pre-selection) can be found in Appendix[B]. Pre - selection cuts were applied to cut out areas where the Monte Carlo distributions was not accurate in its prediction or remove events from a signal peak that did not have any-thing to do with Higgs-decay for example the Zveto cut which removes events from the Z -peak[6]. The figures show the histograms before and after the cuts were made. The order in which these histograms are shown is the same order in which the cuts were applied in the analysis. In general it is expected that the cuts applied in this analyses can be tighter than the cuts in run 1, because the amount of expected events is higher due to a higher available energy in run 2 than in run 1. The effect of the cuts on the SBR will be discussed in section 4.1.

1. Emiss_T > 35 GeV

This cut on the MET is a pre - selection cut, but is mentioned because MET is an important variable for indirect neutrino detection. The missing transverse energy cut for run 1 is Emiss

T > 20 GeV [6]. The cut this paper made is tighter because the

available energy in run 2 collisions is higher than run 1 which will cause neutrinos to have a higher energy.

Before After

Figure 8: These are the logarithmic histograms for the Emiss

T variable the with electron-muon,

muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and the missing transverse energy is on the X-axis in GeV.

(14)

2. Njet= 0

The ggF production channel distinguishes itself from the other main Higgs - pro-duction channels by having no jets. Therefore Njet = 0 which is the same cut on

this variable as in run 1.

Before After

Figure 9: These are the logarithmic histograms for the number Njet variable for the electron-muon,

muon-electron final states before and after the cuts were applied. In these histogram the number of events are on the Y-axis and number of jets on the X-axis.

3. Mtt < 20 GeV/c2

No cut was made in run 1 on this variable. This paper does cut on this variable to increase the SBR.

Before After

Figure 10: These are the logarithmic histograms for the Mtt variable with electron-muon,

muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and mass of the di-tau system on the X-axis in GeV/c2_.

(15)

4. 24 < Pll

T < 88 GeV/c

The lower boundary of this cut (Pll

T > 24 GeV/c) is looser than the run 1 cut of

Pll

T > 30 GeV/c [6]. The expected transverse momentum of the dilepton system is

expected to be higher since the available energy in run 2 is larger than in run 1. Therefore the cut made by this paper is not expected to be looser. However, a looser cut was applied to increase the SBR .

Before After

Figure 11: These are the logarithmic histograms for Pll_T variable with electron-muon, muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and transverse momentum of the dilepton system is on the X-axis in GeV/c.

(16)

5. 22 < Pleadlepton_T < 60 GeV/c

The lower boundary cut on this variable is the same pre - selection cut as in run 1 [6]. This is not expected because the higher available energy in run 2 will cause the transverse momentum of the leading lepton to be larger than in run 1. To increase the SBR the lower boundary is the same as in run 1 and an upper boundary is added.

Before After

Figure 12: These are the logarithmic histograms for Pleadlepton_T variable with electron-muon, muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and transverse momentum of the leading lepton is on the X-axis in GeV/c.

(17)

6. 10 < Psubleadlepton_T < 45 GeV/c

The lower boundary cut on this variable is the same pre - selection cut as in run 1 [6]. The pre-selection cuts are the same on the Psubleadlepton_T variable. This is not expected because the higher available energy in run 2 will cause the transverse momentum of the sub-leading lepton to be bigger than in run 1. To increase the SBR the lower boundary is the same as in run 1 and an upper boundary is added.

Before After

Figure 13: These are the logarithmic histograms for Psubleadlepton_T variable with electron-muon, muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and transverse momentum of the subleading lepton is on the X-axis in GeV/c.

(18)

7. ∆Φll< 2.0 rad

Because of the mentioned spin-correlations the angle ∆Φll is small. The run 1

cut on this variable is ∆Φll < 1.8 rad. The angle between the leading and

sub-leading momentum vectors is small and is expected to stay the same because only the size of the transverse momentum vectors increases this has no influence on the angle between the leptons. Therefore it is not expected the cut on variable changes. However it does change because it increases the SBR.

Before _After

Figure 14: These are the logarithmic histograms for ∆Φll variable with electron-muon, muon-electron

final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and angle between the two lepton is on the X-axis in rad.

(19)

8. -2.0 < ηleadlepton_{< 2.0 rad}

Run 1 does not cut on this variable, in the run 2 data there are more expected events which allows for more cuts to increase the SBR.

Before After

Figure 15: These are the logarithmic histograms for ηleadlepton _{variable with electron-muon,}

muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and angle between the leading lepton and the beam direction is on the X-axis in rad.

9. METDPhill < 2.35 rad

The run 1 cut on this variable is METDPhill < π/2 rad [6], this paper makes a tighter cut on this variable because relatively more background than signal is cut away as can be seen in fig. 16.

Before After

Figure 16: These are the logarithmic histograms for METDPhill variable with electron-muon, muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and angle between the dilepton system and the dineutrino system is on the X-axis in rad.

(20)

10. Mleadlepton_T < 130 GeV/c2

The run 1 paper does not cut on this variable. This paper does cut on the Mleadlepton_T variable to decrease to amount of expected background. The expected signal de-creases only by a tiny amount, see fig. 17.

Before After

Figure 17: These are the logarithmic histograms for Mleadlepton_T variable with electron-muon, muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and angle between the transverse mass of the leading lepton is on the X-axis in GeV/c2_.

11. Msubleadlepton_T < 110 GeV/c2

The run 1 paper does not cut on this variable. This paper (again) does cut on this variable to decrease the amount of expected background events while the decrease in the number of expected signal is small.

Before After

Figure 18: These are the logarithmic histograms for Msubleadlepton_T variable with electron-muon, muon-electron final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and angle between the transverse mass of the subleading lepton is on the X-axis in GeV/c2_.

(21)

12. Mll < 60 GeV/c2

The cut made in the run 1 paper is Mll< 55 GeV/c2[6]. This paper makes a cut on

this variable that is about the same. If the available energy increases the momentum and kinetic energy of the dilepton system will increase but this increase in kinetic energy will not have an effect on the rest mass. Therefore it is expected the cut on this variable stays the same which is almost the case.

Before After

Figure 19: These are the logarithmic histograms for Mll variable with electron-muon, muon-electron

final states before and after the cuts were applied. In these histograms the number of events are on the Y-axis and angle between the mass of the dilepton system is on the X-axis in GeV/c2_.

3.2 Control Region

A control region (CR) was made for the WW background, to correct the MC simulation after the cuts were applied in the signal region (SR), because it was the largest background. From the CR a normalization factor (NF) was calculated. The NF corrected the MC to fit to the measured data in the CR. In the CR the amount of expected Higgs signal events coming from ggF had to be smaller than one and the WW background had to be the largest background in the CR. The CR was made by using a reversed cut on the Mll

variable (appendix[B]).

3.3 Data to Monte Carlo background ratio

To determine if it was likely the Higgs particle was produced the data/bkg ratio was used as distinctive variable. The expected Higgs signal is not included in this MC background. If there was more data than expected this ratio could have been bigger than one. This meant there were more lepton pairs being produced than expected and it was likely the Higgs particle was produced. The data/bkg ratio could also be less than one or equal to one. If the data/bkg was less than one, taking the magin of error into account, then it was unlikely the Higgs particle was produced and there could be something wrong with the MC simulation because we got less data than expected. If the data/bkg is equal to one,

(22)

taking the magin of error into account, it was unlikely the Higgs particle was produced because you saw as many events as you expect there would be (data/bkg = 1), from a background only hypothesis.

(23)

4 Results

4.1 Control Region

Applying the cuts mentioned in the section method and turning off the blinding, the fol-lowing histograms are acquired:

Before Normalization After Normalization

Figure 20: These are the linear histograms for MT variable with electron-muon, muon-electron final

states before and after normalization. The number of WW background events increase after normal-ization. The number of events are on the Y-axis and the transverse mass on the X-axis in GeV/c2_.

From the CR a NF of 1.21 for the WW background is calculated. See appendix[B] for the calculation of the normalization factor and for the cutflow on which these results are based.

(24)

4.2 Cuts

In fig. 21 and fig. 22 the effect of each cut on the SBR and data/bkg ratio respectively is shown. Z-Veto and MET are pre-selection cuts.

Figure 21: In this figure the signal to Monte Carlo background ratio for the normalized data is shown on the Y - axis. On the X-axis the name of the cut is shown.

Figure 22: In this figure the data to Monte Carlo background ratio for the normalized data is shown on the Y - axis. The name of the cuts are shown on the X-axis

From figure fig. 21 it is clear that the SBR increases or stays about the same within the margin of error with each performed cut. From fig. 22 it is clear the margin of error on the data/bkg increases. The data/bkg ratio (after normalization) calculated from the SR is 0.99 ± 0.05.

(25)

5 Discussion

The data/bkg ratio of 0.99 ± 0.05 suggest that we can not conclude if it is likely or un-likely the the Higgs particle is being produced because the data/bkg ratio can be bigger than, equal to or smaller than one.

In the SR the number of expected signal events is 40.3 ± 0.4 and the normalized and expected number of background events in total is 461.1 ± 12.6 (Appendix[B], fig. 36) compared to 20 ± 4 expected signal events and 142 ± 16 total expected background events from run 1 in the H → WW → lνlν channel [5]. The measured number of events in the SR of this paper is 455 and the number of measured events in the SR of run 1 is 185 [5]. This gives a data to background ratio of 1.30 ± 0.15 and a normalized expected SBR of 0.14 ± 0.04 for run 1. The normalized expected SBR in this paper is 0.0874 ± 0.0033.

This difference in SBR can be explained by the fact that run 1 uses a Boosted Decision Tree (BDT) [6] (while this paper does not) which optimizes the SBR. This can explain why the SBR in run 1 is higher than the SBR in this paper. It can be assumed that using a BDT in this paper will give a similar SBR as in run 1. However the available energy, luminosity and bunch spacing in run 2 are different from run 1 which will lead to different distributions of the run 2 data, therefore it does not necessarily have to be the case that using BDT for the run 2 data will lead to a similar value for the SBR as in run 1. In conclusion, not using a BDT could explain the difference between the SBRs. In future data - analyses a BDT could be used to maximize the SBR.

Based on the literature [5],[7] the Higgs particle exists, therefore a data/bkg ratio bigger than one is expected, which was the case in run 1 . This expectation is not confirmed since the data/bkg ratio is 0.99 ± 0.05 the data/bkg ratio can also be smaller or equal to one. This deviation from what is expected comes from the fact that the margin of error is big. One can see from fig. 22 the margin of error gets bigger or stays the same with each cut. The error bars ultimately become so big that it can not be concluded if the data/bkg increased, decreased or stayed the same (comparing Zveto with Mll). The margin of error gets bigger because the amount of data gets smaller. Therefore more data will lead to a smaller margin of error.

Only if the margin of error would be 0.01 it would be possible to conclude it is unlikely the Higgs particle is being produced. The minimum margin of error on the data/bkg (for ZVeto, see Appendix[B]) is 0.01 and applying cuts will cause the error bars of the data/bkg to increase. It is likely the error bars will be bigger than 0.01 after applying cuts. Therefore it is necessary, that the value of the data/bkg ratio needs to increase or decrease in such a way that the data/bkg ratio is equal to, smaller or bigger than 1 within

(26)

the minimal margin of error in order to conclude if it is likely or unlikely the Higgs particle is being produced. This can also be accomplished by collecting more data or possibly by applying other cuts. It might also be possible to include the electron-electron and muon-muon channel in the analysis. This increases the amount of data to be analyzed, which may cause smaller error bars. However the expected signal for this channel is weak and therefore including it may add nothing but background. In conclusion, if the margin of error gets smaller and the data/bkg increases it may be possible to conclude if it is likely or unlikely the Higgs particle is being produced. That is why in future experiments more data should be collected to be able to conclude if it is likely or unlikely the Higgs particle is being produced.

(27)

6 Conclusion

The data/bkg ratio of 0.99 ± 0.05 suggest that it can not concluded if it is likely or unlikely the the Higgs particle is being produced because the ratio can be bigger than, equal to or smaller than one. More data needs to be collected to change the data/bkg ratio value and decrease its margin of error also a BDT could be used to optimize the SBR. If the margin of error gets smaller and the data/bkg increases it may be possible to see if it is likely or unlikely the Higgs particle is being produced.

(28)

7 Acknowledgments

I would like to thank Lydia Brenner. Her positive and happy personality made working on this project very interesting and fun. Questions could always be asked, even when she was swamped by her own work, and an understandable reply was always given. Finally, I would also like to thank my fellow bachelorstudents Eline van den Heuvel, Mart Pothast and Jordy Butter for the amazing time I had with them at the NIKHEF.

(29)

8 Literature

References

[1] Gadatsch, S. (2015). The Higgs Boson. PhD thesis, Physics, University of Amster-dam, Amsterdam.

[2] Nguyen, A. K. & Nguyen T. H. V., (2015). Was the higgs boson discovered? Down-loaded on the 21th of june 2016 from http://arxiv.org/abs/1503.08630.

[3] Pastore, F. (2016). Atlas run-2 status and performance. Elsevier, 270-272:3–7.

[4] The ATLAS Collaboration (2010). Studies of the performance of the ATLAS detector using cosmic-ray muons. Downloaded on the 5th of July 2016 from: hhttp://arxiv.org/abs/1011.6665.

[5] The ATLAS collaboration (2012). Observation of a new particle in the search for the Standard Model Higgs boson with the ATLAS detector at the LHC. Elsevier, 716(1):1–29.

[6] The ATLAS Collaboration (2014). Analyses of H → WW → lνlν and VBF production modes with 20 fb−1 and 4.5 fb−1 of data collected with the ATLAS detector at√s = 8 and 7 TeV. ATLAS note.

[7] The CMS collaboration (2012). Observation of a new boson at a mass of 125 GeV with the CMS experiment at the LHC . Elsevier, 716(1):30–61.

[8] Thomson, M. (2013). Modern Particle Physics. New York: Cambridge University Press.

[9] Universit¨at Z¨urich (2015). SM1. Downloaded on the 13th of june 2016 from http://www.physik.uzh.ch/groups/serra/images/SM1.png.

[10] University of California Santa Cruz. atlasdet. Downloaded on the 5th of July 2016 from: http://scipp.ucsc.edu/personnel/atlasdet.jpg.

(30)

9 Appendix[A]

This section shows all the Feynman diagrams belonging to the most important back-grounds.

9.1 t¯

t

Figure 23: Each created W particle can decay further to a leptons and associated neutrino

(31)

9.2 WW/WZ/ZZ

Figure 25: The created W particle can decay further to a leptons and associated neutrino

Figure 26: The created W particle can decay further to a lepton and associated neutrino and the created Z particle to two leptons

(32)

9.3 Wt

Figure 28: The created W particle can decay further to a leptons and associated neutrino. The created quark can be a down,strange or bottom quark.

(33)

9.4 Wγ

Figure 29: The created W particle can decay further to a leptons and associated neutrino, the photon to two leptons

(34)

9.5 Z+jets

Figure 31: The created Z particle and photon can decay further to two oppositely charged leptons.

Figure 32: The created Z particle can decay further to two oppositely charged leptons, this background is also known as Drell-Yan

(35)

9.6 W+jets

Figure 33: The created W particle can decay further to a lepton and associated neutrino.

9.7 QCD

(36)

10 Appendix[B]

SBR & data/bkg per cut

In fig. 35 the values for the normalized data/bkg, the amount of expected signal and unnormalized expected background events are shown. The plots in the subsection 5.1 are based on these numbers. These values are copied directly from the cutflow.

Figure 35: This figure shows the values for the normalized data to Monte Carlo background ratio, the amount of expected expected signal and unnormalized expected background events with margin of errors after eacht performed cut.

Normalization Factor & Code for the (pre-selection) cuts

From the cutflow that can be seen in fig. 36 we calculated a normalization factor (NF) with formule (3).

N F = [Data − BG]/W W (4)

In formula (3) Data is the amount of data measured in the CR, BG is the total estimated background events in the CR en WW is the amount of WW events in the CR. The numbers from unnormalized CR were plugged in (fig. 36). This gave a NF of 1.21 ± 0.04. This uncertainty on the data/bkg ratio is not processed in the normalized expected WW and total background uncertainties. In figures fig. 37 and fig. 38 the code for the pre-selection cuts and cuts made by this paper are shown respectively.

(37)

Figure 36: This figure shows the amount of estimated background events made by the Monte Carlo simulation and the measured data for the signal region( Mll< 60 GeV/c2) and control region (Mll>

60 GeV/c2) before and after normalization

(38)

Figure 38: This figure shows the cuts made for the signal region (Mll< 60 GeV/c2) and control region

Higgs Boson Production via Gluon - Gluon Fusion And Decay Through: H → WW → lνlν Cutbased Analyses Of The Run Two ATLAS Data