Available on the CERN CDS information server CMS PAS HIG-16-011

CMS Physics Analysis Summary

Contact: cms-pag-conveners-higgs@cern.ch 2016/03/27

Search for resonant Higgs boson pair production in thebblνlν final state at

√s = 13 TeV

The CMS Collaboration

Abstract

A search for resonant Higgs bosons pair production decaying into bbWW with subse-quent WW decays into two leptons and two neutrinos, X→ HH→ bbWW→ bblνlν,is presented. The analysis is based on a sample of proton-proton collisions at

√s =

13 TeV at the LHC corresponding to an integrated luminosity of 2.3 fb−1. Masses areconsidered in the range between mX = 260 GeV and 900 GeV. The search focuses onthe invariant mass distribution of the b-jet pair, searching for a resonant-like excesscompatible with the H boson mass in combination with a boosted decision tree dis-criminant based on kinematic information. Data and predictions from the standardmodel are in agreement within systematic uncertainties. For mass hypotheses frommX = 500 GeV to mX = 900 GeV, the data are observed (expected) to exclude a pro-duction cross-section times branching ratio of a spin-0 particle from 174 to 101 (135 to75.8) fb.

1

1 IntroductionThe Higgs mechanism is an essential element of the standard model (SM) of particles and theirinteractions explaining the origin of mass and playing a key role in electroweak symmetrybreaking. However, the discovery of a Higgs boson (H) with a mass of around 125 GeV by theATLAS and CMS experiments [1, 2] opens the question of whether it is the SM Higgs boson orone of typically many Higgs bosons predicted by extensions beyond the standard model.

Extensions of the scalar sector of the SM predict the existence of additional Higgs bosons. Asimple scenario is the two Higgs doublet model (2HDM), where a second doublet of complexscalar fields is added to the minimal SM scalar sector lagrangian. The generic 2HDM poten-tial [3] has a large number of degrees of freedom. Assuming the preservation of the electro-magnetic gauge symmetry, the CP invariance on the bosonic sector of the theory, the choice ofthe custodial phase and the suppression of the tree-level flavour-changing neutral currents, thenumber of free parameters is reduced to six. These free parameters are the mass of the Higgsboson H (mH), the mass of the pseudoscalar A (mA), the masses of the charged Higgs bosons(mH±), the mass of the CP-even state X (mX), the ratio of the vacuum expectation values tan βand the mixing angle α between the two CP-even eigenstates. In case the new CP-even stateis massive enough (mX > 2mH) it can decay to a pair of Higgs bosons. Models inspired bywarped extra dimensions [4] predict the existence of new heavy particles X (with mX > 2mH),that can decay to a pair of Higgs bosons. Examples of such particles are radion (spin-0) [5–8]or the first Kaluza-Klein excitation of the graviton (spin- 2) [9, 10].

In this document we report on a search for resonant Higgs pair production, X → HH, whereone of the H decays as H→ bb, and the other as H→WW→ lνlν (where l is either an electronor a muon) using LHC proton-proton collision data at

√s = 13 TeV. The analysis focuses on

the invariant mass distribution of the b-jet pair, searching for a resonant-like excess compatiblewith the H boson mass in combination with a boosted decision tree discriminant based onkinematic information. The dominant background is tt production with smaller contributionsfrom Drell-Yan and single top processes. Figure 1 shows the schematic diagram of the resonantHH production with the subsequent Higgs and W boson decays, and the branching ratios ofthe SM Higgs boson.

This study is performed for the first time using LHC data.

2 CMS detector and simulationThe central feature of the CMS apparatus is a superconducting solenoid, 13 m in length and6 m in diameter, which provides an axial magnetic field of 3.8 T. The bore of the solenoid isoutfitted with various particle detection systems. Charged particle trajectories are measuredby silicon pixel and strip trackers, covering 0 < φ < 2π in azimuth and |η| < 2.5, where thepseudo-rapidity η is defined as η = − log tan(θ/2), with θ being the polar angle of the trajec-tory of the particle with respect to the beam direction. A crystal electromagnetic calorimeter(ECAL) and a brass/scintillator hadronic calorimeter (HCAL) surround the tracking volume;in this analysis the calorimetry provides high resolution energy and direction measurementsof electrons and hadronic jets. A preshower detector consisting of two planes of silicon sensorsinterleaved with lead is located in front of the ECAL at |η| < 1.479. Muons are measured ingas detectors embedded in the steel return yoke outside the solenoid. The detector is nearlyhermetic, allowing for energy balance measurements in the plane transverse to the beam di-rections. A two-tier trigger system selects the most interesting pp collision events for use inphysics analysis. A more detailed description of the CMS detector can be found elsewhere [11].

2 3 Event selection and background predictions

XH

H

g

g

b

b

`

ν

`

ν

XX)→BR(Hµµ γZ γγ ZZ cc ττ gg WW bb

YY

)→

BR

(H

µµ

γZ

γγ

ZZ

cc

ττ

gg

WW

bb

-810

-710

-610

-510

-410

-310

-210

-110

1

XXYY)→BR(HH

Figure 1: Left: Schematic diagram of the resonant HH production with the subsequent Higgsand W boson decays. Right: Branching ratios of the SM Higgs boson.

The background processes, in order of importance, are: tt, Drell-Yan, single top. Dibosonproduction, ttV production, as well as single SM Higgs production with subsequent decaysH → WW and H → bb, are also taken into account in the analysis chain even if they are notcontributing in a visible way. Other contributions, such as W + jets or QCD multijet eventswith jets misidentified as leptons, are negligible due to the tight dilepton selection. The dom-inant contribution, especially in the e±µ∓ selection, arises from tt production that mimics allthe objects in the final state (2 b-jets, 2 leptons, and 2 neutrinos) when both W bosons decay asW→ lν. All of the background contributions are taken directly from simulation.

Background simulation samples have been generated using MADGRAPH 5 versions 2.2.2.0and 2.3.2.2 [12], POWHEG 2 [13–17] and PYTHIA 8 [18, 19] version 8.205. The signal sampleshave been generated using MADGRAPH 5 version 2.2.2.0 and describe events at leading orderof gluon fusion production of a spin-0 or a spin-2 object decaying into two SM-Higgs bosonswith a mass of 125 GeV. One of the Higgs boson is required to decay into a pair of b-quarks,while the second one is required to decay to final states containing two leptons and two neutri-nos. This implies that the signal samples contain both H → Z(ll)Z(νν) and H → W(lν)W(lν)decay legs. The SM branching ratios are assumed, therefore the interference in between thetwo processes is the same as in the SM case. The leptons considered for the decay process areonly electrons and muons, while the three neutrinos flavour are considered. The consideredbranching fraction B(HH → bblνlν) is then 1.23%. The spin-0 and spin-2 objects have beengenerated using a narrow width, well below the experimental resolution in the analysis massrange.

3 Event selection and background predictionsWe collect events with a set of dilepton triggers, which require transverse momentum pT >17 GeV for the first lepton and pT > 12 GeV (8 GeV) for the second electron (muon). Weselect events with two oppositely charged leptons (e+e−, µ+µ−, e±µ∓), in which the electrons(muons) are required to have a pT greater than 20 GeV and 15 GeV (10 GeV), for the higherand lower pT lepton, respectively. Muons (electrons) in the pseudo-rapidity range |η| < 2.4(|η| < 2.5) are considered. A dilepton mass requirement of mll > 12 GeV is applied in the

3

selection.

Electrons, reconstructed by associating tracks with ECAL clusters, are identified using informa-tion on the cluster shape in the ECAL, track quality, and the matching between the track andthe ECAL cluster. Additionally, electrons from photon conversions are rejected. Muons are re-constructed from tracks found in the muon system, associated to tracks in the silicon trackingdetectors. They are identified based on the quality of the track fit and the number of associatedhits in the different tracking detectors. For both lepton flavours, the impact parameter withrespect to the reconstructed vertex with the largest p2

T sum of associated tracks (primary ver-tex) has to be below 0.5 mm in the transverse plane and 1 mm along the beam direction. Thelepton isolation, defined as the scalar pT sum of all particle candidates, excluding the lepton, ina cone around the lepton, divided by the lepton pT, is required to be < 0.04 (< 0.15) for elec-trons (muons). The lepton objects in the simulations are corrected for the residual differencesin between data and simulation.

Jets are reconstructed using a particle flow (PF) technique [20]. Candidates are clustered toform jets using the anti-kT clustering algorithm [21], implemented in the FASTJET package [22],with a distance parameter of 0.4. Jet energies are corrected for residual non uniformity andnon linearity of the detector response using corrections found with collision data [23]. Jets arerequired to have pT > 20 GeV, |η| < 2.4, and be separated from identified leptons by a distanceof√

∆φ2 + ∆η2 = ∆R > 0.3. The magnitude of the negative vector sum of all PF candidates isreferred to as Emiss

T . Corrections to the jet energy are propagated to the EmissT .

To identify jets originating from b quarks, the combined secondary vertex algorithm is used.Jets are considered as b-tagged if they pass the medium working point of the algorithm, whichprovides around 70% efficiency with a mistag rate less than 1% [24]. Correction factors areapplied to the selected jets in simulation to account for the different response of the combinedsecondary vertex algorithm in between data and simulation.

Among all possible dijet combinations fulfilling the previous criteria we select the two jets withthe highest sum of the combined secondary vertex output.

After the final object selection consisting of two opposite sign leptons and two b-tagged jets,four kinematic distributions (mll, ∆Rll, ∆Rjj, ∆φll,jj) are selected due to their separation powerbetween signal and background. In all of them, background dominated regions in which thepresence of signal is at the percent level, are present. By applying the following set of cuts:mZ −mll > 15 GeV, ∆Rll < 2.2, ∆Rjj < 3.1, and ∆φll,jj > 1.5, 95% of the background is removedwhile 86-97% of the signal is kept, for signal masses ranging from 400 to 900 GeV. These cutswere applied before further optimization by the boosted decision tree discriminant describedin the next section, as they remove significant background with negligible effect on signal.Figure 2 shows the pll

T and pjjT distributions for data and simulated events after requiring all the

cuts described in this section.

4 Signal extractionBoosted decision trees (BDTs) discriminants are used to further improve the signal-to-backgroundseparation. In a phase space dominated by tt production, the BDTs exploit information relatedto object kinematics. The set of variables provided as input to the BDTs is: mll, ∆Rll, ∆Rjj, ∆φll,jj,

pllT, pjj

T, min(∆Rj,l

), and MT, defined as MT =

√2pll

TEmissT (1− cos(∆φ(ll, Emiss

T ))). This set ofBDT inputs is similar to the one used in previous CMS studies targeting High Luminosity LHCscenarios [25].

4 4 Signal extraction

Eve

nts

/ 8.0

0 G

eV

0

50

100

150

200

250 Data t tSignal (1 pb) Drell-Yan

= 400 GeVXm Single top = 650 GeVXm VV = 900 GeVXm Vt t

Uncertainty SM Higgs

(13 TeV)-12.30 fbPreliminaryCMS

(GeV)T

Dilepton system p

0 50 100 150 200 250 300 350 400

Dat

a / M

C

0.6

0.8

1

1.2

1.4

channelsµe + eµ + ee + µµ

Eve

nts

/ 8.0

0 G

eV

0

20

40

60

80

100

120

140

160

180 Data t tSignal (1 pb) Drell-Yan

= 400 GeVXm Single top = 650 GeVXm VV = 900 GeVXm Vt t

Uncertainty SM Higgs

(13 TeV)-12.30 fbPreliminaryCMS

(GeV)T

Dijet system p

0 50 100 150 200 250 300 350 400

Dat

a / M

C

0.6

0.8

1

1.2

1.4

channelsµe + eµ + ee + µµ

Figure 2: The pllT (left) and pjj

T (right) distributions for data and simulated events after requiringtwo leptons, two b-tagged jets, and mZ−mll > 15 GeV. The simulated samples are normalizedusing a maximum likelihood fit, as described in details section 6.

All the lepton channels (e+e−, µ+µ−, e±µ∓) are merged during the BDT training. The tt, Drell-Yan and single top production, dominant SM processes in the selection, are considered as back-ground, and the spin-0 radion as signal. The BDT training and validation is performed in anindependent tt sample.

While the BDTs are trained using spin-0 signals, their performances are compatible with thoseusing spin-2 signals, allowing one single training for the two different spin hypotheses.

The mass range of the search is defined from mX = 260 GeV to mX = 900 GeV. We traintwo different BDTs within the mass range for signals with masses mX = 400 GeV and mX =650 GeV. The BDT trained at mX = 400 GeV is applied to signals with mX ≤ 450 GeV, and theBDT trained at mX = 650 GeV is applied to signals with mX ≥ 450 GeV. In this way we coverseveral signal kinematics in the range of interest. While the mX = 650 GeV training could beapplied in the whole mass range with some loss in the region between 260 and 450 GeV, theperformance of the mX = 400 GeV training is specific to its range of application. The two BDTdiscriminants are shown in Figure 3.

For each of the BDT outputs we define two regions (low-BDT-scores region and high-BDT-scores region) chosen to maximize the expected sensitivity in the high-BDT-scores region. Theboundary on the BDT output defining the regions is 0.1.

The final categories are defined from the two BDT regions in which we apply a cut on the mjjdistribution around the Higgs boson mass. The cut on the mjj mass is chosen to maximize theexpected sensitivity based on a cut-and-count method. We define the mjj peak region (mjj-P) as95 GeV < mjj < 135 GeV. The region outside this window (mjj < 95 GeV & mjj > 135 GeV) iscalled mjj side-band region (mjj-SB).

In summary, for each of the two BDTs, four final categories are defined as: high-BDT-scores &mjj-P, high-BDT-scores & mjj-SB, low-BDT-scores & mjj-P, and low-BDT-scores & mjj-SB.

5

Eve

nts

/ 0.0

2

0

20

40

60

80

100

120

140 Data t tSignal (1 pb) Drell-Yan

= 400 GeVXm Single top = 650 GeVXm VV = 900 GeVXm Vt t

Uncertainty SM Higgs

(13 TeV)-12.30 fbPreliminaryCMS

BDT output

0.6− 0.4− 0.2− 0 0.2 0.4 0.6

Dat

a / M

C

0.6

0.8

1

1.2

1.4

channelsµe + eµ + ee + µµ

Eve

nts

/ 0.0

2

0

20

40

60

80

100

120

140

160Data t tSignal (1 pb) Drell-Yan

= 400 GeVXm Single top = 650 GeVXm VV = 900 GeVXm Vt t

Uncertainty SM Higgs

(13 TeV)-12.30 fbPreliminaryCMS

BDT output

0.6− 0.4− 0.2− 0 0.2 0.4 0.6

Dat

a / M

C

0.6

0.8

1

1.2

1.4

channelsµe + eµ + ee + µµ

Figure 3: The BDT output distributions for different signal mass trainings, mX = 400 GeV(left) and mX = 650 GeV (right). The BDT trained at mX = 400 GeV is applied to signals withmX ≤ 450 GeV, and the BDT trained at mX = 650 GeV is applied to signals with mX ≥ 450 GeV.The simulated samples are normalized using a maximum likelihood fit, as described in detailssection 6.

5 Systematic uncertaintiesThe different sources of systematic uncertainties affect both normalization and shape of thebackground and signal expectations. We consider as experimental sources of systematic un-certainties the lepton selection (identification and isolation), trigger efficiency, jet energy scale,jet energy resolution, jet b-tagging, pileup, parton distributions, renormalization scale, factor-ization scale and LHC luminosity. In addition, we consider a systematic effect on the tt crosssection. The effects on the total yields in signal regions are summarized in Table 1. Shapevariations as consequence of the systematic uncertainties are considered when affecting the mjjdistribution.

The impact of the jet energy scale uncertainty is evaluated by shifting the jet energy correctionfactors for each jet up and down by one standard deviation, in both signal and backgroundpredictions. The uncertainties associated with the corrections for the b tagging efficienciesfor light-flavor and heavy-flavor jets are evaluated by varying these corrections by one stan-dard deviation. Lepton identification and trigger scale factors are computed with the “tag-and-probe” technique [26] and are applied to the simulated samples to match the performanceobserved in data.

Theoretical uncertainties on the cross section used to predict the tt background are propa-gated to the yield estimates. The magnitude of the uncertainties related to the parton dis-tribution functions for each simulated background process, is obtained using the replicas ofthe NNPDF 3.0 set. The uncertainties related to renormalization and factorization scale areextracted by multiplying or dividing them by two. In addition, we consider modeling uncer-tainties on the tt, Drell-Yan and single-top processes.

Systematic uncertainties due to the finite nature of simulation samples are taken into account.

6 5 Systematic uncertainties

Table 1: Summary of the systematic uncertainties and their impact range on total yields inall final categories and BDT trainings, for signal mX = 400 GeV, signal mX = 650 GeV, andbackground.

Source Sig. (mX = 400 GeV) Sig. (mX = 650 GeV) Background

Trigger efficiency 5.1 - 6.0% 6.7 - 7.4% 4.5 - 5.3%Jet b-tagging 4.9 - 6.5% 5.7 - 7.3% 5.1 - 6.0%Jet energy scale 1.6 - 3.0% 0.6 - 3.9% 1.0 - 3.6%Jet energy resolution 0.5 - 4.1% 1.8 - 3.5% 0.1 - 2.4%Electon ID & ISO 1.3 - 1.6% 1.3 - 1.7% 1.4 - 1.5%Muon ID & ISO 0.9 - 1.4% 1.0 - 1.1% 1.2 - 1.5%Pileup 0.4 - 1.8% 0.1 - 0.6% 0.5 - 2.2%Parton distributions 0.4 - 0.5% 0.2 - 0.5% 0.5 - 0.6%QCD scale 0.3 - 0.4% 0.2 - 0.4% 0.8 - 2.4%Luminosity 2.7%Signal MC stat. 1.4 - 2.4% 0.9 - 3.2% -

Affecting only tt (87.0 - 95.3% of the total bkg.)tt cross section - - 6.5%tt modeling - - 10%tt MC stat. - - 0.6 - 2.3%

Affecting only Drell-Yan (1.8 - 7.1% of the total bkg.)Drell-Yan modeling - - 30%Drell-Yan MC stat. - - 4.4 - 22.7%

Affecting only single top (2.5 - 4.6% of the total bkg.)Single top modeling - - 20%Single top MC stat. - - 6.6 - 24.4%

Affecting only other backgrounds (0.4 - 1.4% of the total bkg.)Other backgrounds MC stat. - - 3.5 - 24.6%

7

6 ResultsSelected events are classified into 4 categories, as described in section 4. Two of these categories(low-BDT-scores & mjj-P, and low-BDT-scores & mjj-SB) are background-dominated, and helpconstraining the background normalizations.

Signal efficiencies as a function of the X mass hypothesis in the four final regions defined in theanalysis, are shown in figure 4.

(GeV)spin-0Xm

300 400 500 600 700 800 900

Sig

nal s

elec

tion

effic

ienc

y (%

)

0

5

10

15

20

25νlνlb b→VV b b→ HH →

spin-0 X→gg

All regions

= 650 GeVX = 400 GeV BDT trained mX BDT trained m

-Pjj

high-BDT-scores & m -Pjj

high-BDT-scores & m

-SBjj

high-BDT-scores & m -SBjj

high-BDT-scores & m

-Pjjlow-BDT-scores & m -Pjjlow-BDT-scores & m

-SBjjlow-BDT-scores & m -SBjjlow-BDT-scores & m

13 TeVCMS Simulation

(GeV)spin-2Xm

300 400 500 600 700 800 900

Sig

nal s

elec

tion

effic

ienc

y (%

)

0

5

10

15

20

25νlνlb b→VV b b→ HH →

spin-2 X→gg

All regions

= 650 GeVX = 400 GeV BDT trained mX BDT trained m

-Pjj

high-BDT-scores & m -Pjj

high-BDT-scores & m

-SBjj

high-BDT-scores & m -SBjj

high-BDT-scores & m

-Pjjlow-BDT-scores & m -Pjjlow-BDT-scores & m

-SBjjlow-BDT-scores & m -SBjjlow-BDT-scores & m

13 TeVCMS Simulation

Figure 4: Signal efficiencies as a function of the X mass hypothesis in the four final regionsdefined in the analysis, for spin-0 (left) and spin-2 (right) hypotheses. The dashed line repre-sents the switch in between the two analysis strategies. The markers correspond to efficienciesevaluated on fully-simulated signal samples, while the efficiencies have been interpolated inbetween. The cyan corresponds to the high-BDT-scores & mjj-P, the blue corresponds to thehigh-BDT-scores & mjj-SB, the red corresponds to the low-BDT-scores & mjj-P, and the yellowcorresponds to the low-BDT-scores & mjj-SB. The black line, being the sum of all (mutuallyexclusive) regions, correspond to the full signal acceptance after preselection and as such iscontinuous.

We first perform a maximum likelihood fit in order to extract best-fit signal cross-section, whereall the nuisances parameters described in section 5 are free to float. These cross-sections are allcompatible with zero for all the signal mass hypotheses considered in this analysis.

Another maximum likelihood fit is performed to extract post-fit distributions of all distribu-tions shown in this note, along with post-fit uncertainties (figures 2 and 3). The four distribu-tions are fitted simultaneously, the signal strength fixed to zero and all nuisances parametersfloating. Post-fit uncertainties are computed from the uncertainties of all the nuisances param-eters, using the correlation between all the parameters.

Since no sign of new physics is found, we set limits on the production of resonant Higgs pairproduction cross-section. Upper limits at 95% confidence level (CL) are computed as functionof the X mass hypothesis, taking the data, background, and signal in the four final regionsdefined in the analysis, using the asymptotic CLs method [27]. Limits on the resonant Higgspair production cross section times branching fraction, HH → bbVV → bblνlν are shown inFigure 5.

8 References

(GeV)spin-0Xm

200 300 400 500 600 700 800 900

) (f

b)νlνlb

b→

BR

(HH

×

HH

) →

spin

-0 X

→(p

pσ

95%

CL

limit

on

210

310

410Observed 95% upper limit

Expected 95% upper limit

1 std. deviation±Expected limit

2 std. deviations±Expected limit

=1.0 TeV, kL=35)RΛradion (

±h=mH=m

A=125 GeV, m

h=0 GeV, m12m

)=0.10α-β=0.10, cos(β2HDM, tan

Type I

Type II

(13 TeV)-12.30 fbCMS Preliminary

(GeV)spin-2Xm

200 300 400 500 600 700 800 900)

(fb)

νlνlb b

→ B

R(H

H

× H

H)

→ sp

in-2

X→

(pp

σ95

% C

L lim

it on

210

310

410Observed 95% upper limit

Expected 95% upper limit

1 std. deviation±Expected limit

2 std. deviations±Expected limit

=0.3PlMRS1 KK graviton, kL=35, k/

(13 TeV)-12.30 fbCMS Preliminary

Figure 5: Expected and observed 95% CL upper limits on the the resonant Higgs pair produc-tion cross section times branching fraction for HH → bblνlν, computed using the asymptoticCLs method, for spin-0 (left) and spin-2 (right) hypotheses. The markers correspond to the limitevaluated on the fully-simulated mass points, while in between the signal prediction was inter-polated. The change in trend of the observed limits at 450 GeV, represented by the dashed line,corresponds to transition point of the analysis between the two BDTs (see section 4). Theoreti-cal cross-sections corresponding to 2HDM of Type I and Type II are overlaid on the spin-0 plot.Theoretical cross-sections corresponding to WED models for radions and RS1 KK gravitons arealso overlaid on the spin-0 and spin-2 plots respectively.

As already demonstrated by the best-fit cross-section, no excess is visible in the whole massrange of the analysis, the observed limits being compatible with the expected ones within 2standard deviations. The change of trend in the observed limits at 450 GeV corresponds to thetransition point of the analysis between the two BDTs, one optimized for low mass resonances,one for high mass resonances.

7 SummaryWe have presented a search for resonant Higgs pair production, X → HH, where one of the Hdecays as H→ bb, and the other as H→ WW→ lνlν using LHC proton-proton collision dataat√

s = 13 TeV, corresponding to an integrated luminosity of 2.3 fb−1. Masses are consideredin the range between mX = 260 GeV and 900 GeV.

Data and predictions from the standard model are in agreement within systematic uncertain-ties. For mass hypotheses from mX = 500 GeV to mX = 900 GeV, the data are observed (ex-pected) to exclude a production cross-section times branching ratio from 174 to 101 (135 to75.8) fb.

References[1] ATLAS Collaboration, “Observation of a new particle in the search for the Standard

Model Higgs boson with the ATLAS detector at the LHC”, Phys.Lett. B 716 (2012) 129.

References 9

[2] CMS Collaboration, “Observation of a new boson at a mass of 125 GeV with the CMSexperiment at the LHC”, Phys.Lett. B 716 (2012) 3061.

[3] G. C. Branco et al., “Theory and phenomenology of two-Higgs-doublet models”, Phys.Rept. 516 (2012) 1–102, doi:10.1016/j.physrep.2012.02.002,arXiv:1106.0034.

[4] L. Randall and R. Sundrum, “A Large mass hierarchy from a small extra dimension”,Phys.Rev.Lett. 83 (1999) 33703373.

[5] W. D. Goldberger and M. S. Wise, “Modulus stabilization with bulk fields”,Phys.Rev.Lett. 83 (1999) 49224925.

[6] O. DeWolfe, D. Freedman, S. Gubser, and A. Karch, “Modeling the fifth-dimension withscalars and gravity”, Phys.Rev. D 62 (2000) 046008.

[7] C. Csaki, M. Graesser, L. Randall, and J. Terning, “Cosmology of brane models withradion stabilization”, Phys.Rev. D 62 (2000) 045015.

[8] C. Csaki, J. Hubisz, and S. J. Lee, “Radion phenomenology in realistic warped spaceModel”, Phys.Rev. D 76 (2007) 125015.

[9] H. Davoudiasl, J. Hewett, and T. Rizzo, “Phenomenology of the Randall-Sundrum GaugeHierarchy Model”, Phys.Rev.Lett. 84 (2000) 2080.

[10] C. Csaki, M. L. Graesser, and G. D. Kribs, “Radion dynamics and electroweak physics”,Phys.Rev. D 63 (2001) 065002.

[11] CMS Collaboration, “The CMS experiment at the CERN LH”, JINST 0803 (2008) S08004.

[12] J. Alwall et al., “The automated computation of tree-level and next-to-leading orderdifferential cross sections, and their matching to parton shower simulations”, JHEP 07(2014) 079, doi:10.1007/JHEP07(2014)079, arXiv:1405.0301.

[13] P. Nason, “A New method for combining NLO QCD with shower Monte Carloalgorithms”, JHEP 11 (2004) 040, doi:10.1088/1126-6708/2004/11/040,arXiv:hep-ph/0409146.

[14] S. Frixione, P. Nason, and C. Oleari, “Matching NLO QCD computations with PartonShower simulations: the POWHEG method”, JHEP 11 (2007) 070,doi:10.1088/1126-6708/2007/11/070, arXiv:0709.2092.

[15] S. Alioli, P. Nason, C. Oleari, and E. Re, “A general framework for implementing NLOcalculations in shower Monte Carlo programs: the POWHEG BOX”, JHEP 06 (2010) 043,doi:10.1007/JHEP06(2010)043, arXiv:1002.2581.

[16] S. Alioli, S. Moch, and P. Uwer, “Hadronic top-quark pair-production with one jet andparton showering”, JHEP 01 (2012) 137.

[17] S. Alioli, P. Nason, C. Oleari, and E. Re, “Single-top production in the s- and t-channel”,JHEP 09 (2009) 111.

[18] T. Sjostrand, S. Mrenna, and P. Z. Skands, “PYTHIA 6.4 Physics and Manual”, JHEP 05(2006) 026, doi:10.1088/1126-6708/2006/05/026, arXiv:hep-ph/0603175.

10 References

[19] T. Sjostrand, S. Mrenna, and P. Z. Skands, “A Brief Introduction to PYTHIA 8.1”, Comput.Phys. Commun. 178 (2008) 852–867, doi:10.1016/j.cpc.2008.01.036,arXiv:0710.3820.

[20] CMS Collaboration, “Performance of the CMS missing transverse momentumreconstruction in pp data at

√s = 8 TeV”, JINST 10 (2015) P02006.

[21] M. Cacciari, G. P. Salam, and G. Soyez, “The anti-kT jet clustering algorithm”, JHEP 04(2008) 063.

[22] M. Cacciari, G. P. Salam, and G. Soyez, “FastJet user manual”, Eur. Phys. J. C 72 (2012)1896.

[23] CMS Collaboration, “Jet Energy Calibration in the 8 TeV pp data”, CMS-PAS-JME-13-0042015.

[24] CMS Collaboration, “Identification of b-quark jets with the CMS experimet”, JINST 8(2013) P04013.

[25] CMS Collaboration, “Higgs pair production at the High Luminosity LHC”, CMS PhysicsAnalysis Summary CMS-PAS-FTR-15-002, 2015.

[26] CMS Collaboration, “Performance of CMS muon reconstruction in pp collision events at√s = 7 TeV”, JINST 7 (2012) P10002, arXiv:1206.4071.

[27] A. Read, “Modified frequentist analysis of search resulst (the CLs method)”, TechnicalReport CERN-OPEN-2000-005, CERN, 2000.