Search for the associated production of the Higgs boson with a top quark pair in multilepton final states with the ATLAS detector

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Search for the associated production of the Higgs boson with a top quark pair in multilepton final states with the ATLAS detector

.ATLAS Collaboration

^

a r t i c l e i n f o a b s t ra c t

Articlehistory:

Received19June2015

Receivedinrevisedform29July2015 Accepted31July2015

Availableonline5August2015 Editor:W.-D.Schlatter

A search for the associated production of the Higgs boson with a top quark pair is performed in multileptonfinalstatesusing20.3 fb⁻¹ofproton–protoncollisiondatarecordedbytheATLASexperiment at √

s=^{8 TeV at the LargeHadron Collider. Five final states, targeting the decays H}→^{W W∗,} τ τ, and Z Z^∗,are examinedfor the presenceof theStandard Model (SM)Higgs boson:twosame-charge lightleptons (e orμ)withoutahadronically decayingτ lepton; threelightleptons;twosame-charge lightleptons withahadronically decayingτ lepton; fourlightleptons;and one lightleptonandtwo hadronically decaying τ leptons. No significant excess of events is observed above the background expectation.Thebestfitforthet¯t H productioncrosssection,assumingaHiggsbosonmassof125 GeV, is2.1⁺₋¹₁^._.⁴₂timestheSMexpectation,andtheobserved(expected)upperlimitatthe95%confidencelevel is4.7(2.4)timestheSMrate.Thep-valueforcompatibilitywiththebackground-onlyhypothesisis1.8σ; theexpectationinthepresenceofaStandardModelsignalis0.9σ.

©^{2015CERNforthebenefitoftheATLAS}Collaboration.PublishedbyElsevierB.V.Thisisanopen accessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

The discovery of a new particle H with a mass of about 125 GeVinsearchesfortheStandardModel(SM)[1–3]Higgsbo- son [4–7] at the LHC was reported by the ATLAS [8] and CMS [9] Collaborations in July 2012. The particle has been observed in the decays H→

γ γ

[10,11], H→ ^{Z Z∗} →⁴ [12,13], and H→^{W W∗}→ 

ν



ν

[14,15],andevidence hasbeen reportedfor H→

τ τ

[16,17], consistent with the rates expected for the SM Higgsboson.

TheobservationoftheprocessinwhichtheHiggsbosonispro- ducedinassociationwithapairoftopquarks(t¯t H )wouldpermit adirectmeasurementofthetopquark–Higgsboson Yukawacou- plingin aprocess that is tree-levelatthe lowestorder, whichis otherwise accessible primarily through loop effects. Having both thetree- and loop-level measurements wouldallow disambigua- tion of newphysics effects that could affectthe two differently, such asdimension-six operators contributing to the gg H vertex.

This letter describes a search for the SM Higgs boson in the t¯t H production mode in multilepton final states. The five final statesconsideredare:twosame-charge-signlightleptons(e or

μ

) withnoadditionalhadronicallydecaying

τ

lepton;threelightlep- tons;twosame-sign lightleptons withonehadronically decaying

τ

lepton;fourlightleptons;andonelightleptonwithtwohadron- ically decaying

τ

candidates. Thesechannels are sensitive to the Higgs decays H→^{W W∗,}

τ τ

, and Z Z^∗ produced in association

 E-mailaddress:atlas.publications@cern.ch.

with a top quark pair decaying to one ortwo leptons. Asimilar searchhasbeenperformedbytheCMSCollaboration[18].

Theselectionsofthissearcharedesignedtoavoidoverlapwith ATLASsearchesfortt H in¯ H→^bb¯ [19]andH→

γ γ

[20]decays.

Themainbackgroundstothesignal arisefromt¯t productionwith additionaljetsandnon-promptleptons,associatedproductionofa top quark pair anda vector boson W or Z (collectively denoted tt V ),¯ andother processeswheretheelectronchargeisincorrectly measured or wherequark or gluon jetsare incorrectly identified as

τ

candidates.

2. ATLASdetectoranddataset

The features of the ATLAS detector [21] mostrelevant to this analysisare briefly summarized here. Thedetector consistsofan inner tracking detector system surrounded by a superconduct- ing solenoid, electromagnetic and hadronic calorimeters, and a muonspectrometer.Chargedparticlesinthepseudorapidity¹range

|

η

_{| <}2.5 arereconstructedwiththeinnertrackingdetector,which is immersedin a2 T magnetic field parallel to the detectoraxis

1 TheATLASexperimentusesaright-handedcoordinatesystemwithitsoriginat thenominalinteractionpoint(IP)inthecentreofthedetector,andthez-axisalong thebeam line.Thex-axispoints fromtheIPtothecentreoftheLHCring,and they-axispointsupwards.Cylindricalcoordinates(r,φ)areusedinthetransverse plane,φbeingtheazimuthalanglearoundthez-axis.Observableslabelled“trans- verse”areprojectedontothe x– y plane.Thepseudorapidity isdefinedinterms ofthepolarangleθasη= −ln tanθ/2.Thetransversemomentumisdefinedas pT=psinθ=p/coshη,andthetransverseenergyEThasananalogousdefinition.

http://dx.doi.org/10.1016/j.physletb.2015.07.079

0370-2693/©^{2015CERNforthebenefitoftheATLAS}Collaboration.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

andconsistsofpixelandstripsemiconductordetectorsaswellasa straw-tubetransitionradiationtracker.Thesolenoidissurrounded byacalorimetersystemcovering|

η

|<4.9,whichprovidesthree- dimensional reconstruction ofparticle showers. Lead/liquid-argon (LAr) sampling technology is used for the electromagnetic com- ponent. Iron/scintillator-tile sampling calorimeters are used for the hadroniccomponentfor |

η

_|<1.7, andcopper/LAr andtung- sten/LArtechnologyis usedfor|

η

_|>1.5.Outside thecalorimeter system, air-core toroids provide a magnetic field for the muon spectrometer.Three stationsofprecision drifttubes andcathode- stripchambersprovideameasurementofthemuontrackposition andcurvatureintheregion|

η

|<2.7.Resistive-plateandthin-gap chambersprovidemuontriggeringcapabilityupto|

η

_|=2.4.

This search uses data collected by the ATLAS experiment in 2012 ata centre-of-mass energy of √

s=^{8 TeV. All events con-} sidered were recorded while the detector and trigger systems were fullyfunctional; theintegratedluminosity of thisdatasetis 20.3 fb⁻¹.

3. Crosssectionsforsignalandbackgroundprocesses

The cross section for the production of t¯t H in pp collisions has been calculated at next-to-leading order (NLO) in quantum chromodynamics (QCD) [22–26]. Uncertainties on the cross sec- tionareevaluatedbyvaryingtherenormalizationandfactorization scales by factors of two and by varying the input parton dis- tribution functions (PDF) of the proton. A Higgs boson mass of mH=^{125 GeV is assumed;thisgivesapredictedt}¯t H production crosssectionat√

s=^{8 TeV of129+}₋⁵12(scale) ±^{10(PDF)fb[27].}

ThisassumedHiggs bosonmass isconsistent withthecombined ATLASandCMSmeasurement[28].

In this letter the associated production of single top quarks with a Higgs boson is considered a background process and set to the Standard Model rate. The production of t Hqb and t H W istakenintoaccount. IntheStandardModel theserates arevery smallcomparedtott H production.¯ Theseprocessesare simulated with the same parameters as used by the ATLAS tt H ,¯ H→

γ γ

search [20]. The cross sections for both are computedusing the MG5_aMC@NLO generator[29]atNLOinQCD.Fort Hqb,therenor- malizationandfactorizationscalesaresetto75 GeVandthepro- cessiscomputed inthe four-flavourscheme, yielding

σ

(t Hqb)= 17.2⁺₋⁰₁^._.⁸₄ (scale)⁺₋¹₀^._.²₉ (PDF)fb.Fort H W ,dynamicfactorizationand renormalizationscales are used, andthe process is computedin the five-flavour scheme; the result is

σ

(t H W)=⁴.7⁺₋⁰₀^._.⁴₃ (scale) +⁰.8

−⁰.6(PDF)fb.Theinterferenceoft H W productionwithtt H ,¯ which appearsat NLOfort H W in diagramswithan additionalb-quark inthefinalstate,isnotconsidered.

The production of t¯t W and tt¯(Z/

γ

^∗)→^t¯t⁺⁻ yield multi- lepton final states with b-quarks and are major backgrounds to the t¯t H signal. For simplicity of notation the latter process is referred to as t¯t Z throughout this letter with off-shell Z and photon componentsalso includedexcept wherenotedotherwise.

The tt W process¯ includes both t¯t W⁺ and t¯t W⁻ components.

Next-to-leading-order cross sections are used for t¯t W [30] and tt Z¯ [31]. The MG5_aMC@NLO generatoris usedto reproducethe QCDscaleuncertaintiesofthesecalculationsanddetermineuncer- tainties due to the PDF. For tt W production¯ the value 232±²⁸ (scale)± ^{18(PDF)fbisused,andfort}¯t Z production² thevalueis 206±^{23 (scale)} ±^{18 (PDF)fb.}

2 TheNLOcrosssection isonlyevaluatedfor tt Z production¯ with on-shellZ.

Thecrosssectionobtainedfort¯t(Z/γ^∗)productionincludingoff-shellZ/γ^∗ con- tributionsinaleading-ordersimulationisscaledbyaK -factorof1.35obtainedas theratioofNLOandLOon-shellcrosssections.The K -factordiffersfromthatof Ref.[31]duetoadifferentchoiceofPDF.

The associated productionof a single top quark and a Z bo- son is a subleading background for the most sensitive channels.

The cross section has been calculated at NLO for the t- and s-channels [32]. The resulting values used in this work are 160

± ^7(scale)± ^{11(PDF)fbfort Z and76}±^4(scale)±^5(PDF)fb fort Z .¯ The crosssection fortheproductionoft W Z iscomputed atleading order(LO) usingthe MadGraph v5 generator [33]and foundtobe4.1fb.

Thecrosssectionforinclusiveproductionofvectorbosonpairs W W , W Z ,and Z Z iscomputedusingMCFM[34].Contributions fromvirtualphotonsandoff-shell Z bosonsareincluded.Theun- certainties on the acceptance for these processes in the signal regions (which favour productionwithadditional b- or c-quarks) dominate over the inclusive cross-section uncertainty (see Sec- tion7.2)andsothelatterisneglectedintheanalysis.

The inclusive tt cross¯ section is calculated at next-to-next- to-leading order (NNLO) in QCD which includes resummation of next-to-next-to-leadinglogarithmic (NNLL)softgluontermsusing Top++ [35], yielding 253⁺₋¹³₁₅ pb for √

s=^{8 TeV. The} single-top- quark samples arenormalized tothe approximateNNLO theoret- ical cross sections [36–38] using the MSTW2008 [39] NNLO PDF set.Theproductionof Z→ ⁺⁻+^{jets andW}→ 

ν

₊jets isnor- malizedusingNNLOcrosssectionsascomputedbyFEWZ[40].

4. Eventgeneration

Theeventgeneratorconfigurationsusedforsimulatingthesig- nal and main backgroundprocesses are shown in Table 1. Addi- tionalinformationisgivenbelow.

The tt H signal¯ eventsimulationsamplescontainall Higgsbo- son decays with branching fractions set to values computed at NNLO inQCD [26,66–69].The factorization(

μ

F) andrenormaliza- tion(

μ

R) scalesaresettom_t+^mH/2.Higgsbosonandtopquark massesof125 and172.5 GeV, respectively,are used.Thesesam- plesarethesameasthoseusedbyotherATLASt¯t H searches[19, 20].

ProductionofsingletopquarkswithHiggsbosonsissimulated as follows.For t Hqb,events are generated atleading orderwith MadGraphinthefour-flavourscheme.Fort H W ,eventsaregener- atedatNLOwith MG5_aMC@NLO inthefive-flavourscheme.Higgs bosonandtopquarkmassesaresetasfortt H production.¯

The main irreducible backgrounds are production oft¯t W and t¯t Z(tt V¯ ).Forthett W process,¯ eventsaregeneratedatleadingor- der withzero,one, ortwo extra partonsin thefinal state, while fortt Z zero¯ oroneextrapartonisgenerated.Theimportantcon- tributionfromoff-shell

γ

^∗/Z→ ⁺⁻isincluded.Thet Z process issimulatedwiththesamesetup,withoutextrapartons.

For diboson processes,the full matrixelement for⁺⁻ pro- duction, including

γ

^∗ andoff-shell Z contributions, is used. The Sherpaqq and¯ qg samples includediagramswithadditionalpar- tons in the final state atthe matrix-element (ME) level, andin- cludeb- andc-quarkmasseffects. Sherpa wasfoundtohavebetter agreementwithdatathan Powheg forW Z ,whilethe Sherpa and Powhegdescriptionsof Z Z productionaresimilar.

A tt¯+jets samplegenerated withthe Powheg NLO generator [61] isused; the top quark massis set to 172.5 GeV. Smallcor- rections to the t¯t system and top quark pT spectra are applied based on discrepancies in differential distributions observed be- tweendataandsimulationat7 TeV[70].Double-countingbetween thet¯t andW t singletopproductionfinalstatesiseliminatedusing thediagram-removalmethod[71].

Samples of Z → ⁺⁻+^{jets and W} → 

ν

+jets events are generated with up to five additional partons using the Alp- gen v2.14[65] leading order(LO) generator.Samples are merged withmatrixelement-partonshoweroverlapsremoved usingMLM

Table 1

Configurationsusedforeventgenerationofsignalandbackgroundprocesses.Ifonlyonepartondistributionfunctionisshown,thesameoneisusedforboththematrix element(ME)andpartonshowergenerators;iftwoareshown,thefirstisusedforthematrixelementcalculationandthesecondforthepartonshower.“Tune”refersto theunderlying-eventtuneofthepartonshowergenerator.“Pythia 6”referstoversion6.425;“Pythia 8”referstoversion8.1;“Herwig++”referstoversion2.6;“MadGraph”

referstoversion5;“Alpgen”referstoversion2.14;“Sherpa”referstoversion1.4;“gg2ZZ”referstoversion2.0.

Process ME generator Parton shower PDF Tune

tt H¯ HELAC-Oneloop[41,42] Pythia8[43] CT10[44]/CTEQ6L1[45,46] AU2[47]

+^{Powheg-BOX[48–50]}

t Hqb MadGraph[33] Pythia8 CT10 AU2

t H W MG5_aMC@NLO[29] Herwig++[51] CT10/MRST LO**[52] UE-EE-4[53]

tt W¯ + ≤2 partons MadGraph Pythia6[54] CTEQ6L1 AUET2B[55]

tt¯(Z/γ^∗)+ ≤1 parton MadGraph Pythia6 CTEQ6L1 AUET2B

t(Z/γ^∗) ^{MadGraph Pythia}6 CTEQ6L1 AUET2B

q^q¯,qg→^{W W},W Z Sherpa[56] Sherpa CT10 Sherpadefault

qq→qqW W , qqW Z , qq Z Z Sherpa Sherpa CT10 Sherpadefault

qq¯,qg→^{Z Z Powheg-BOX[57] Pythia8 CT10 AU2}

gg→^{Z Z gg2ZZ[58] Herwig[59] CT10 AUET2[60]}

tt¯ Powheg-BOX[61] Pythia6 CT10/CTEQ6L1 Perugia2011C[62]

s-, t-channel, W t single top Powheg-BOX[63,64] Pythia6 CT10/CTEQ6L1 Perugia2011C

Z→ ⁺⁻+ ≤5 partons Alpgen[65] Pythia6 CTEQ6L1 Perugia2011C

W→ ν+ ≤^{5 partons Alpgen Pythia6 CTEQ6L1} Perugia2011C

matching[72].Productionofb- and c-quarksisalsocomputedat matrix-elementlevel,andoverlapsbetweenMEandpartonshower productionarehandledbyseparatingthekinematicregimesbased ontheangularseparationofadditionalheavy partons.The result- ing“light”and“heavy”flavoursamplesarenormalizedbycompar- ingtheresultingb-taggedjetspectrawithdata.

Allsimulated sampleswith Pythia 6 and Herwig [59] parton showering use Photos 2.15 [73] to model photon radiation and Tauola1.20[74]for

τ

decays.The Herwig++ samplesmodelpho- ton radiation with Photos but use the internal

τ

decay model.

Samplesusing Pythia 8.1and Sherpa usethosegenerators’internal

τ

leptondecayandphotonradiationgenerators.For Herwig sam- ples,multiplepartoninteractionsaremodelledwith Jimmy[75].

Showered andhadronized events are passed through simula- tions of the ATLAS detector (either full GEANT4 [76] simulation or a hybrid simulation with parameterized calorimeter showers and GEANT4 simulation of the tracking systems [77,78]). Addi- tional minimum-bias pp interactions (pileup) are modelled with the Pythia 8.1generatorwiththe MSTW2008LOPDF setandthe A2tune [79]. Theyare addedto the signal andbackgroundsim- ulatedeventsaccordingto the luminosity profileof therecorded data,withadditionaloverallscalingtoachieveagoodmatchtoob- servedcalorimetry andtrackingvariables.The contributionsfrom pileup interactions both within the same bunch crossing as the hard-scatteringprocess and in neighbouring bunch crossings are includedinthesimulation.

5. Objectselection

Electron candidates are reconstructed from energy clusters in the electromagnetic calorimeterassociated with reconstructed tracksintheinner detector.Theyare requiredtohave|

η

cluster|<

2.47.Candidatesinthetransitionregion1.37<|

η

cluster|<1.52 be- tween sections of the electromagnetic calorimeter are excluded.

A multivariate discriminant based on shower shape and track informationisused todistinguish electrons fromhadronicshow- ers [80,81]. Only electron candidates with transverse energy E_T greater than 10 GeV are considered. To reduce the background fromnon-promptelectrons,i.e. fromdecaysofhadrons(including heavy flavour) produced in jets, electron candidates are required tobe isolated. Twoisolation variables,based oncalorimetricand trackingvariables,arecomputed.The first(E^cone_T ) isbasedonthe sumof transverse energies ofcalorimeter cells within a cone of radius R≡

(φ)²+ (

η

)²=⁰.2 around the electron candi- datedirection. Thisenergysumexcludescellsassociatedwiththe

electron and is corrected for leakage from the electromagnetic shower and ambient energy in the event. The second (p^cone_T ) is definedbasedontrackswith p_T>1 GeV withinacone ofradius R=⁰.2 aroundthe electron candidate. Both isolation energies are separately required to be lessthan 0.05×^ET. The longitudi- nal impact parameter of the electron track with respect to the selectedeventprimaryvertex, multipliedby thesineofthepolar angle,|^z0sinθ|^{,isrequired tobe lessthan1 mm. The transverse} impactparameterdividedbytheestimateduncertaintyonitsmea- surement, |^d0|/

σ

(d0),mustbe lessthan4.Iftwoelectrons closer than R=⁰.1 are selected, only the one with the higher pT is considered. An electron is rejectedif, after passing all the above selections,itlieswithinR=⁰.1 ofaselectedmuon.

Muon candidatesare reconstructed bycombininginner detec- tor tracks withtrack segments or full tracks in the muon spec- trometer [82]. Only candidates with |

η

_|<2.5 and pT>10 GeV are kept. Additionally, muonsare required to be separated by at leastR>0.04+ (^{10 GeV})/pT,μ fromanyselectedjets(see be- low fordetails onjetreconstruction andselection).The cutvalue is optimizedto maximize the acceptanceforprompt muonsat a fixed rejection factor for non-prompt andfake muon candidates.

Furthermore, muons must satisfy similar E^cone_T and p_T^cone isola- tion criteria asfor electrons, with both required to be lessthan 0.10×^pT.Thevalueof|^z0sinθ|^{isrequiredtobelessthan1mm,} while|^d0|/

σ

(d0)mustbelessthan3.

Hadronicallydecaying

τ

candidates(

τ

had)arereconstructedus- ingclustersintheelectromagneticandhadroniccalorimeters.The

τ

candidates are required to have pT greater than 25 GeV and

|

η

_|<2.47. The number of charged tracks associated with the

τ

candidatesisrequiredtobeoneorthreeandthecharge ofthe

τ

candidates, determined from the associated tracks, must be ±^1.

The

τ

identification uses calorimeter cluster and tracking-based variables, combinedusing a boosteddecisiontree(BDT) [83].An additionalBDT whichusescombinedcalorimeterandtrackquan- tities is employed to reject electrons reconstructed asone-prong hadronicallydecaying

τ

leptons.

Jetsare reconstructedfromcalibrated topologicalclusters[21]

built from energy deposits in the calorimeters, using the anti-kt algorithm [84–86] with a radius parameter R =⁰.4. Prior to jet finding, a local cluster calibration scheme [87,88] is ap- plied to correct the topological cluster energies for the effects of non-compensating calorimeterresponse, inactive material and out-of-cluster leakage. The jets are calibrated using energy and

η

-dependent calibration factors, derived from simulations, to the meanenergyofstableparticles insidethejets. Additionalcorrec-

tions to account for the difference between simulation anddata are derived fromin-situ techniques[89,90]. Afterenergy calibra- tion,jetsarerequiredtohavep_T>25 GeV and|

η

|<2.5.

Toreducethecontamination fromjetsoriginatingin pp inter- actionswithinthesamebunchcrossing(pileup),thescalarsumof the pT oftracks matchedtothe jetandoriginatingfromthe pri- maryvertexmust beatleast 50%ofthe scalarsumofthe pT of alltracksmatchedto thejet.Thiscriterionisonlyappliedto jets withp_T<50 GeV (thosemostlikelytooriginatefrompileup)and

|

η

_|<2.4 (toavoidinefficiencyattheedgeoftrackingacceptance).

The calorimeter energy deposits from electrons are typically alsoreconstructed asjets;in orderto eliminatedouble counting, anyjetswithinR=⁰.3 ofaselectedelectronarenotconsidered.

Jetscontaining b-hadronsareidentified(b-tagged)via amulti- variatediscriminant[91]that combinesinformationfromtheim- pactparameters ofdisplacedtrackswithtopologicalpropertiesof secondaryandtertiarydecayverticesreconstructedwithinthejet.

The working point used for this search corresponds to approxi- mately70%efficiencytotagab-hadronjet,withalight-jetmistag rateof≈^{1% and a charm-jetrejectionfactorof5,asdetermined} forb-tagged jetswith p_T of 20–100 GeV and |

η

|<2.5 insimu- latedt¯t events.Toavoidinefficienciesassociatedwiththeedgeof the tracking coverage, only jetswith |

η

_|<2.4 are considered as possible b-tagged jetsin this analysis. The efficiency andmistag ratesoftheb-taggingalgorithmaremeasuredindata[91,92]and correctionfactorsareappliedtothesimulatedevents.

6. Eventselectionandclassification

All events considered in this analysis are required to pass single-lepton (e or

μ

) triggers. These achieve their maximal plateauefficiencyforlepton p_T>25 GeV.

This analysis primarily targets the H→^{W W∗ and}

τ τ

decay modes.Consideringthe decayofthett system¯ aswell, theset¯t H eventscontaineitherW W W W bb or¯

τ τ

W W bb.¯ Thestrategyisto targetfinalstatesthatcannotbeproducedint¯t decayalone—i.e., threeormoreleptons,ortwosame-signleptons—thussuppress- ingwhatwouldotherwisebethelargestsinglebackground.

Theanalysiscategoriesareclassifiedbythenumberoflightlep- tons and hadronic

τ

decay candidates. The leptons are selected usingthecriteriadescribedearlier.Eventsareinitiallyclassifiedby countingthe numberoflight leptons with pT>10 GeV. At least onelightleptonisrequiredtomatchaleptonselectedbythetrig- gersystem. After initial sorting into analysiscategories, in some casesthelepton selectioncriteriaare tightenedby raisingthe pT threshold,tighteningisolation selectionsorrestrictingtheallowed

|

η

_|range,asexplainedinthefollowingper-categorydescriptions.

Theanalysisincludes fivedistinctcategories: twosame-sign light leptonswithno

τ

had(20

τ

had),threelightleptons(3),twosame- signlight leptons andone

τ

had (21

τ

had), four light leptons (4), andone lightlepton andtwo

τ

had (12

τ

had). The categorieswith

τ

had candidates targetthe H→

τ τ

decay;the othersareprimar- ily sensitive to H→^{W W∗ witha very small} contribution from H→^{Z Z∗.The}contributionstoeachcategoryfromdifferentHiggs bosondecaymodesare showninTable 2.Theseselectioncriteria ensurethataneventcanonlycontribute toasinglecategory.The contaminationfromgluonfusion,vectorbosonfusion,andassoci- atedV H productionmechanismsfortheHiggsbosonispredicted to be negligible. Summed over all categories, the total expected numberof reconstructed signal events assuming Standard Model tt H production¯ is10.2,correspondingto0.40%ofallproducedt¯t H events.Thedetailedcriteriaforeachcategoryaredescribedbelow.

Table 2

Fractionoftheexpectedtt H signal¯ arisingfromdifferentHiggsbosondecaymodes ineachanalysiscategory.Thesix20τhadcategoriesarecombinedtogether,asare the two4categories.Thedecayscontributingtothe“other”columnaredomi- nantlyH→μμandH→^bb.¯ Rowsmaynotaddto100%duetorounding.

Category Higgs boson decay mode

W W^∗ τ τ Z Z^∗ Other

20τhad 80% 15% 3% 2%

3 74% 15% 7% 4%

21τhad 35% 62% 2% 1%

4 69% 14% 14% 4%

12τhad 4% 93% 0% 3%

6.1. 20

τ

hadcategories

Selected events are required to includeexactly two light lep- tons, which must have the same charge. Events with

τ

had can- didates are vetoed. To reduce the background from non-prompt leptons,theleading(subleading)leptonisrequiredtosatisfy pT>

25 (20)GeV,and themuon isolation requirements are tightened to E^cone_T /pT<0.05 and pTcone/pT<0.05.The angularacceptance of electroncandidates isrestricted to|

η

|<1.37 in order tosup- presst¯t backgroundeventswherethesignoftheelectroncharge is misreconstructed,asthe chargemisidentificationrateincreases athighpseudorapidity.

In order to suppressthe lower-multiplicity tt¯+^{jets and t}¯t W backgrounds,eventsmustincludeatleastfourreconstructedjets.

In order to suppress diboson and single-boson backgrounds, at leastone ofthesejetsmust beb-tagged.The selectedeventsare separatedby lepton flavour (e^±e^±,e^±

μ

^±,and

μ

^±

μ

^±) andnum- ber ofjets(exactly four jets,at leastfivejets) intosixcategories withdifferentsignal-to-backgroundratio,resultinginhigherover- allsensitivitytothet¯t H signal.

6.2. 3category

Selectedeventsare requiredto includeexactly threelightlep- tons with total charge equal to ±^{1. Candidate events arising} from non-prompt leptons overwhelmingly originate as opposite- signdileptoneventswithoneadditionalnon-promptlepton.As a result, thenon-promptlepton isgenerallyoneofthetwoleptons withthesamecharge.Toreduce thesebackgrounds,ahighermo- mentumthresholdpT>20 GeV isappliedtothetwoleptonswith thesamecharge.No requirementsareimposedonthenumberof

τ

_had candidates.In order to suppressthet¯t+^{jets andt}t V back-¯ grounds,selectedeventsarerequiredtoincludeeitheratleastfour jetsofwhichatleastonemustbeb-tagged,orexactlythreejetsof whichatleasttwoareb-tagged.Tosuppressthett Z background,¯ eventsthatcontainanopposite-signsame-flavourleptonpairwith thedileptoninvariant masswithin 10 GeVofthe Z massareve- toed.Eventscontaininganopposite-signleptonpairwithinvariant massbelow12 GeV arealsoremovedtosuppressbackgroundfrom resonancesthatdecaytolightleptons.

6.3. 21

τ

hadcategory

Selected events are required to include exactly two light leptons, with the same charge and leading (subleading) pT >

25 (15) GeV, and exactly one hadronic

τ

candidate. The recon- structedchargeofthe

τ

had candidatehastobeoppositetothatof thelightleptons.Inordertoreducet¯t+^{jets andt}t V backgrounds,¯ events must include atleast four reconstructed jets. In order to suppress diboson and single-boson backgrounds, at least one jet must be b-tagged.To suppressthe Z→ ⁺⁻+jets background, eventswithdielectroninvariantmasswithin10 GeVoftheZ mass arevetoed.

Fig. 1. Thespectrumofthenumberofjetsexpectedandobservedinthet¯t Z (left)andt¯tW (right)validationregions(VR).Thehatchedbandrepresentsthetotaluncertainty onthebackgroundpredictionineachbin.The“non-prompt”backgroundsarethosewithaleptonarisingfromahadrondecayorfromaphotonconversionindetector material.Rareprocessesincludet Z ,t¯tW W ,triboson,t¯ttt,¯andt H production.Theoverlaidredlinecorrespondstothett H signal¯ predictedbytheSM.(Forinterpretationof thereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)

6.4.4categories

Selectedevents are requiredto include exactlyfour light lep- tons with total charge equal to zero and leading (subleading) pT>25 (15) GeV. No requirements are applied on the number of

τ

hadcandidates.Inordertosuppressthett¯+^{jets andt}¯t V back- grounds,the selectedevents are requiredto includeatleast two jetsofwhichatleastone mustbe b-tagged.Tosuppressthett Z¯ background,eventsthatcontainanopposite-signsame-flavourlep- tonpairwithdileptoninvariantmasswithin10 GeVoftheZ mass are vetoed. In order to suppress background contributions from resonances that decay to light leptons, all opposite-sign same- flavourleptonpairsarerequiredtohaveadileptoninvariantmass greater than 10 GeV. The four-lepton invariant mass is required tobebetween100and500 GeV,whichgiveshighacceptancefor t¯t H , H→^{W W∗}→ 

ν



ν

, butrejects Z→⁴ and high-masstt Z¯ events.Selectedeventsare separatedby thepresence orabsence ofasame-flavour,opposite-signleptonpairintotwocategories,re- ferredtorespectivelyastheZ -enrichedandZ -depletedcategories.

Inbothcasesthe Z massvetoisapplied,butbackgroundeventsin theZ -enrichedcategorycanarisefromoff-shellZ and

γ

^∗→ ⁺⁻ processeswhileinthe Z -depletedcategorythesebackgroundsare absent.

6.5.12

τ

_hadcategory

Selectedeventsarerequiredtoincludeexactlyonelightlepton withpT>25 GeV andexactlytwohadronic

τ

candidates.The

τ

had candidates must have opposite charge. In order to suppress the t¯t+^{jets andt}¯t V backgrounds,eventsmustincludeatleastthree reconstructedjets. Inorderto suppressdibosonandsingle-boson backgrounds,atleastoneofthe jetsmustbeb-tagged.Thisfinal stateis primarilysensitive to H→

τ

⁺

τ

⁻ decays,allowing useof the invariant mass of the visible decay products of the

τ

had

τ

had system(m_vis) asa signal discriminant.Signal eventsare required tosatisfy60<m_vis<120 GeV.

7. Backgroundestimation

Important irreducible backgrounds include tt V and¯ diboson productionandare estimatedfromMC simulation.Validation re- gions enriched in these backgrounds are used to verify proper modellingofdatabysimulation.Reduciblebackgroundsaredueto non-prompt lepton production and electron charge mis- identification, andareestimatedfromdata,withinput fromsim- ulation in some categories. In the 12

τ

had category the primary concernisfake

τ

hadcandidates,whicharemodelledusingsimula- tionandvalidatedagainstadata-drivenestimate.

7.1. tt V and¯ t Z

The primary backgrounds withprompt leptons stemfrom the production of t¯t W and t¯t Z . The tt W background¯ tends to have lower jet multiplicity than the signal and so the leading contri- bution comes from eventswith additional high-pT jets; it is the majortt V contribution¯ inthe 20

τ

had categoriesandcomparable to tt Z in¯ the 21

τ

had category. Thet¯t Z process hassimilar mul- tiplicity to the t¯t H signal but can only contribute to the signal categorieswhen the Z boson decaysleptonically, so the on-shell contributioncanberemovedbyvetoingeventswithopposite-sign dilepton pairs with invariant mass near the Z pole. This is the larger of the two tt V contributions¯ for the 3, 4, and 12

τ

had categories.Thet Z processmakesasubleadingcontributiontoboth channels.Avalidationregionisusedtoverifythemodellingoft¯t Z usingon-shell Z decays. Agreementis seen within thelarge sta- tistical uncertainty. No region of equivalent purityand statistical powerexistsforttW production;¯ neverthelesstheexpectationsare cross-checked witha validation region defined with the 20

τ

had selectionexceptwithtwoormoreb-taggedjetsandeithertwoor threejets,wherethet¯t W purityis≈^{30%,andarefoundtobecon-} sistent within uncertainties. Thespectra ofthenumberof jetsin thesevalidationregionsare showninFig. 1.

Uncertaintieson the t¯t V background contributions arise from boththeoverallcrosssectionuncertainties(seeSection3)andthe acceptance uncertainties. The latter are estimated by comparing particle-level samples after showering produced by three differ- ent pairs of generators: a) the nominal MadGraph LO merged

Table 3

Expectedandobservedyieldsineachchannel.UncertaintiesshownarethesuminquadratureofsystematicuncertaintiesandMonteCarlosimulationstatisticaluncertainties.

“Non-prompt”includesthemisidentifiedτhadbackgroundtothe12τhad category.Rareprocesses(t Z ,t¯tW W ,tribosonproduction,ttt¯¯t,t H )arenotshownasaseparate columnbutareincludedinthetotalexpectedbackgroundestimate.

Category q mis-id Non-prompt tt W¯ t¯t Z Diboson Expected bkg. t¯t H(μ=1) Observed

ee+ ≥5 j 1.1±0.5 2.3±1.2 1.4±0.4 0.98±0.26 0.47±0.29 6.5±1.8 0.73±0.14 10

eμ+ ≥^{5 j 0.85}±^{0.35 6.7}±^{2.4 4.8}±^{1.2 2.1}±^{0.5 0.38}±^{0.30 15}±^{3 2.13}±^{0.41 22}

μμ+ ≥^{5 j – 2.9}±^{1.4 3.8}±^{0.9 0.95}±^{0.25 0.69}±^{0.39 8.6}±^{2.2 1.41}±^{0.28 11}

ee+^{4 j 1.8}±^{0.7 3.4}±^{1.7 2.0}±^{0.4 0.75}±^{0.20 0.74}±^{0.42 9.1}±^{2.1 0.44}±^{0.06 9}

eμ+^{4 j 1.4}±^{0.6 12}±^{4 6.2}±^{1.0 1.5}±^{0.3 1.9}±^{1.0 24}±^{5 1.16}±^{0.14 26}

μμ+4 j – 6.3±2.6 4.7±0.9 0.80±0.22 0.53±0.30 12.7±2.9 0.74±0.10 20

3 – 3.2±0.7 2.3±0.7 3.9±0.8 0.86±0.55 11.4±2.3 2.34±0.35 18

21τhad – 0.4⁺₋⁰₀^._.⁶₄ 0.38±0.12 0.37±0.08 0.12±0.11 1.4±0.6 0.47±0.08 1

12τhad – 15±5 0.17±0.06 0.37±0.09 0.41±0.42 16±5 0.68±0.13 10

4Z -enr. – ¹⁰⁻³ ³×^{10−3 0.43}±^{0.12 0.05}±^{0.02 0.55}±^{0.15 0.17}±^{0.02 1}

4Z -dep. – ¹⁰⁻⁴ ^{10−3 0.002}±^0.002 ²×^{10−5 0.007}±^{0.005 0.025}±^{0.003 0}

sample versus an equivalent LO merged sample generated with Sherpa 2.1.1, to account for ME-parton shower matching effects;

b)theLOmerged Sherpa sampleversusa Sherpa+OpenLoops[93]

NLO sample, tocompare LOmerged andNLO acceptance;andc) MG5_aMC@NLO with Pythia 8 parton shower versus Herwig++

partonshower,tocomparepT-orderedversusangular-orderedpar- ton showers. Eachofthesevariations isinput independentlyinto the final fit. When summed in quadrature they have an im- pactof5–23%depending onthe categoryandbackgroundsource (t¯t W versus tt Z ).¯ Uncertainties arising from changes in the ac- ceptance dueto the choice of QCD scale and PDF are also eval- uated; these have an impact of 1.3–6.7% for scale and 0.9–4.8%

for PDF.

7.2. Otherpromptleptoncontributions

Other backgrounds with prompt leptons arise from multibo- son processes (W Z , Z Z ,and triboson production) in association with heavy-flavour jets, or with a misidentified light-flavour jet.

ThemainprocessaffectingthefinalresultisW Z+^{jets.Validation} regionswiththreeleptonsincludingaZ candidateandeitherzero orone b-taggedjet are studied. Thenumber ofjetsin W Z+^0b events is reproduced well in the highly populated bins (upto 4 jets),leading to theconclusion that thejet radiation spectrumis wellmodelled.Thedominantuncertaintyonthepredictioninthe signalregion isexpectedtoarise fromthe W Z+^{b crosssection.}

Dataconstrain thiscomponentwithroughly 100%uncertainty.As aresulta 100%uncertaintyisassignedto the W Z+^{b crosssec-} tion, givinga 50% uncertainty on the total W Z yield, correlated across categories. The cross sections for production of W W +^b and Z Z+^{b arealsoassigned50%}uncertainties;thesehavenegli- gibleimpactonthefinalresult.

7.3. Chargesignmisidentification

The process e^±→^e±

γ

_→e^±e⁺e⁻ occurring in detector ma- terial can result in an electron produced with nearly the same momentum as the parent electron but with opposite charge. In these cases the observed electron has opposite charge to that ofthe primary electron (charge mis-id).The analogousprocesses

μ

^±→

μ

^±e⁺e⁻ and

μ

^±→

μ

^±

μ

⁺

μ

⁻havenegligible ratesforthe selected events.The t¯t and Z/

γ

^∗_{→ }⁺⁻+jets events that un- dergo this process contribute to 20

τ

had in the ee and e

μ

cate- gories. As electrons pass through more material at high |

η

|^{, the} chargemis-idrateincreasesaswell,andsotheelectron|

η

|<1.37 requirement significantly reduces the impact of this background.

Thechargemis-idrateduetotrackcurvaturemismeasurementfor electronsandmuonsisnegligible.

The charge mis-id probability is determined by a maximum- likelihoodfitusing Z→^{ee events}reconstructedassame-signand as opposite-sign pairs, as a function of electron

η

and pT. This probability functionisthenappliedtoasample ofeventspassing the20

τ

hadselectionexceptthattheleptonpairisrequiredtobe oppositesign.Thechargemis-idprobabilityfromtherelativelylow momentum Z daughtersisextrapolatedtohigherpT usingscaling functions extractedfrom Monte Carlosimulations. The dominant uncertaintyisduetothestatisticalprecision ofthechargemis-id probabilitydetermination,andis≈^{40% inthesignalregions.}

7.4. Non-promptlightleptons

A significant backgroundarises from leptons not produced in decays of electroweak bosons (non-prompt leptons), which can promote(forexample)asingle-leptontt event¯ intoa20

τ

had cat- egoryoradileptont¯t eventtothe3or21

τ

hadcategories.These backgrounds in the signal regions are expected to be dominated byt¯t orsingletopquarkproductionwithleptonsproducedinde- cays ofheavy-flavourhadrons.Productionoftt with¯ anadditional photon whichconvertsinthe detectormaterialisa subdominant contribution.Withthetightobjectselectionrequirementsapplied in this analysis, almost all reconstructed electron and muon ob- jects correspond to real electrons and muons; the fraction aris- ing from incorrect particle identification is negligible. Estimates ofthesebackgrounds areobtainedfromdata.Eachchannel hasa slightlydifferentprocedure,motivatedbythespecificeventtopol- ogyandthestatisticalpoweravailableinthecontrolregions. The methods are discussed below,andthe expectednon-prompt lep- ton contributions to thevarious categoriesare showninTable 3.

In the following, a tight lepton isa lepton that passesthe nom- inal selection, a sideband lepton isdefined asa lepton candidate which satisfies different criteria than the tight lepton selection (identification selection, isolation, or p_T), and (sideband) control regions either require one or more sideband leptons to replace a tight lepton in the signal region selection, or have the same lepton selection as the signal region but different jet require- ments.

7.4.1. 20

τ

hadcategories

The non-prompt lepton yields in the signal regions are es- timated by extrapolating from sideband control regions in data whichare enrichedint¯t non-promptcontributions.Forelectrons, sidebandobjects areselected byinverting theelectronidentifica- tion and isolation requirements; formuons the sideband objects havelow transverse momentum,6<pT<10 GeV,butotherwise areselectedthesamewayasnominalmuons.Transferfactorsare used toextrapolatefromeventswithonetight andonesideband

lepton,but which otherwise passthe signal region selections, to thesignalregionswithtwotightleptons.Thesetransferfactorsare determinedfromadditionaldatacontrolregions(tight+sideband andtwotightleptons)withlowerjetmultiplicity (1≤ⁿjet≤^{3 for} electrons,2≤ⁿjet≤^{3 formuons).Inallregionstheexpectedcon-} tribution from processes producing prompt leptons is subtracted beforeextractingtransferfactorsorusingtheyieldsforextrapola- tion.Forchannelswithelectrons,thechargemis-idbackgroundis alsosubtracted,andadileptonmassvetoisappliedinthecontrol regionstosuppresscontributionsfrom Z→^{e+e− decays.A cross-} checkonthemuonestimate,usingan extrapolationinmuoniso- lationinsteadofmuon p_T,agreeswellwiththenominalprocedure andprovidesadditionalconfidenceintheestimate.

Thesystematicuncertaintiesonthisprocedureareestimatedby checkinga)itsabilitytosuccessfullypredictthenon-promptback- groundintt simulation¯ andb)thestabilityofthepredictionusing datawhenthe selection ofthecontrol regions isaltered. Forthe former,differentpartonshower andb-hadron decaymodels were checked,aswas the result ofremoving the b-tagged jet require- ment.Inaddition,forelectrons,theeffectsofrelaxingthepseudo- rapidityrequirement to|

η

|<2.5 and ofraising the p_T threshold werestudied.Thesechecksshowstabilityatthe25–30%level,lim- itedbythestatisticalprecisionofthesimulations.Thestabilityin datais checked by alteringthe pT required for theb-tagged jet, applyingarequirementon missingtransversemomentum³ E^miss_T , extractingthetransferfactorsonlyfromeventswiththreejets,or (formuons)using10–15 GeV muonsasthesidebandobjects.This checkshowsstabilityofthepredictionsto14%formuonsand19%

forelectrons.Additionalsystematicuncertaintiesintheprediction arisefromthestatisticaluncertaintiesontheyieldsinthecontrol regionsandthesubtractionofpromptandchargemis-idcontribu- tions.Theoveralluncertaintiesonthenon-promptyieldprediction inanygivencategoryrangefrom32%to52%,andcorrelationsbe- tween the categories due to uncertainties in the transfer factors areincludedinthefit(seeSection9).

7.4.2. 3category

Sidebandleptonsaredefinedbyreversingtheisolationrequire- mentforelectronsandmuonsand,forelectrons,requiringthatthe candidatefailthetightelectronidentificationdiscriminantrequire- mentoftheanalysisbutpassa looserselection.The non-prompt leptoncontributioninthesignalregionisestimatedbyextrapolat- ingfromdataregionswithtwotightandonesidebandlepton,us- ingtransferfactorsestimatedfromMonteCarlosimulation.These eventstypicallycontaintwopromptopposite-signleptonsandone non-promptlepton,whichnecessarilymustbeofthesamesignas one ofthe prompt leptons. Thereforethe non-prompt lepton es- timationprocedure isapplied onlyto thetwo same-sign leptons.

The simulation-derived transfer factor is validatedin a region of lower jet multiplicity (2≤ⁿjet≤^{3 and exactly one b-tagged jet).}

Goodagreementisobservedinthisvalidationregionbetweenthe prediction (11.8±².3) and the observed yield (9.8±⁴.9 events afterpromptbackgroundsubtraction). Systematicuncertainties in theprocedurearederivedbystudyingtheagreementbetweendata andsimulationinthevariablesusedfortheextrapolation,whichis

≈^{20%forbothelectronsandmuons.Additional}uncertaintiesarise fromthe statisticaluncertainties on the yields in the control re- gionsandinthet¯t simulation.

3 Thisiscalculatedusingcalorimeterenergydeposits,calibratedaccordingtoas- sociatedreconstructedphysicsobjects,andalsoincludingthetransversemomenta ofreconstructedmuons.

7.4.3. 21

τ

hadcategory

Reconstructing twosame-signlight leptonsfromt¯t production or similar sources requires that one of the light leptons is non- promptorhasitschargemisidentified.Inthe21

τ

hadcategory,the charge mis-id contributionis negligible andtheprimary concern is non-prompt light leptons. Around half of the

τ

had candidates in these events come from W →

τ ν

decays, while the remain- der arise from misidentified light-quarkor gluon jets. Regardless ofwhetherthe

τ

hadcandidateisafake,thereisalsoanon-prompt light lepton.Due tothis fact,sidebands inthe light-leptonselec- tioncriteriaareused,analogouslytothe20

τ

hadand3categories.

Sincetheratioofrealandfake

τ

hadcandidatesissimilarinthesig- nalandallcontrolregions,fake

τ

had candidatesarenotaccounted forseparately; thesmallvariations in theratiointhe control re- gions are found to have negligible impact on the total estimate inthe signalregion. Inorderto maintain similarorigincomposi- tion of the non-prompt leptons, the ET isolation requirement is inverted, the pT isolation requirement is relaxed, and for elec- tronstheidentificationcriteriaarealsorelaxedtoalooserworking point. The low jet multiplicity region 2≤ⁿjet≤^{3 isused to de-} termineatransferfactorfromsidebandto tightleptonselections.

The expectednon-promptlepton yieldinthe signalregion isob- tained by usingthis transferfactor to extrapolate froma control region withthe same jet selection as thesignal region butwith one tight and one sideband light lepton. The procedure is vali- dated by checking that it correctly reproduces the signal region yield expected in t¯t simulations. The assigned systematic uncer- tainty (27%) isdominated by thestatistical precision ofthistest.

Theoveralluncertaintyonthenon-promptbackgroundprediction is dominatedby the limitedstatistics ofthe highjet multiplicity controlregion.

7.4.4. 4category

The non-prompt lepton contribution in this category is ex- pectedtobenegligibleandisestimatedtobe^{10−3 eventsinthe} Z -enriched sample and^{10−4 events inthe} Z -depletedsample.

Inbothcasesthisrepresents^{2% ofthetotalbackgroundexpecta-} tion.Theseestimatesareobtainedusingthetransfer factorsfrom the3channelandappropriatecontrolregionswithtwolooselep- tonsandrelaxedjetmultiplicityrequirements.

7.5.

τ

hadmisidentificationinthe12

τ

hadcategory

The nominal estimate for the fake

τ

had yield is derived from tt simulation.¯ To obtaina sufficientlylarge sample size,fast sim- ulation using parameterized calorimeter showers is used. At all preselection stages the simulationis found to give an acceptable description of thett background,¯ both inkinematic distributions andtotalyield.Thisestimateiscross-checkedwiththedata-driven methoddescribedbelow.

Ofthetwo

τ

hadcandidates,oneisopposite insigntothelight lepton (OS) andthe other hasthe same sign(SS). The SS candi- dateisalmostalwaysafake

τ

had,whilethelightleptonisprompt andthe OS

τ

had candidateis oftenreal (≈^{30%). A sideband}

τ

had isdefinedasacandidatepassingalooseidentificationBDTselec- tion butnot thenominaltight one. Assuming the

τ

had candidate fakeprobabilitiesare notcorrelatedbetweenjetsidentifiedasOS and SScandidates, control regions can be used to predict yields inthesignalregion.Therearethreecontrolregions,dependingon whetheronlytheOS,onlytheSS,orboththeOSandSS

τ

hadcan- didates are sidebandobjects. The two regions with sideband OS

τ

had candidates are used to obtain the transfer factor for the SS

τ

_had candidate,which is then applied to theregion witha tight OSandsidebandSScandidatetoobtainthepredictionforthesig- nal region where both are tight. The transfer factoris measured