Search for the Higgs boson decays H → ee and H → eμ in pp collisions at √s = 13 TeV with the ATLAS detector

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Search for the Higgs boson decays H → ^{ee and H} → ^e μ in pp collisions at √

s = 13 TeV with the ATLAS detector

.TheATLASCollaboration^

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

Articlehistory:

Received24September2019

Receivedinrevisedform20November2019 Accepted6December2019

Availableonline10December2019 Editor: M.Doser

SearchesfortheHiggsbosondecaysH→^{ee andH}→^eμareperformedusingdatacorrespondingtoan integratedluminosityof139 fb⁻¹collectedwiththeATLASdetectorinpp collisionsat√

s=^{13 TeVat} theLHC. Nosignificantsignalsareobserved,inagreementwiththeStandardModelexpectation.Fora Higgsbosonmassof125 GeV,theobserved(expected)upperlimitatthe95%confidencelevelonthe branchingfractionB(^H→^ee)is3.6×¹⁰⁻⁴⁽³.5×^10−4)andonB(^H→^eμ)is6.2×¹⁰⁻⁵⁽⁵.9×^10−5).

Theseresults representimprovements byfactors ofaboutfive andsix onthepreviousbestlimits on B(^H→^ee)andB(^H→^eμ)respectively.

©^{2019TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense} (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

Thediscoveryofaheavy scalarparticlebyATLAS andCMS [1, 2] providedexperimentalconfirmationoftheEnglert–Brout–Higgs mechanism [3–8], which spontaneously breaks electroweak (EW) gaugesymmetryandgeneratesmasstermsfortheW andZ gauge bosons.Inthe StandardModel(SM) thefermionmassesaregen- erated via Yukawa interactions. The Yukawa couplings to third- generationfermions were determined by measurements ofHiggs boson production and decays [9–15], andfound to be in agree- mentwiththeexpectationsoftheSM.However,thereiscurrently noevidenceofHiggsbosondecaysintofirst- orsecond-generation quarksorleptons.

ThisLetter presentsthe first ATLAS searches for H→^{ee and} forthelepton-flavour-violating decay H→^eμ usingthe fullRun 2datasetofproton–proton(pp)collisions atacentre-of-mass en- ergyof √

s=^{13 TeV, withan integratedluminosity of 139 fb−1.} TheCMSCollaborationhaspreviouslyperformedsearchesforH→ ee [16] andH→^eμ[17] usingLHCRun1pp dataat√

s=^{8 TeV} correspondingtoanintegratedluminosityof19.7 fb⁻¹.

In the SM the H →ee branching fraction is given by GFmHm²_e/(4√

2πH)⁵×^{10−9,wherem}H andH aretheHiggs massandwidthrespectively.This branchingfractionis farbelow the sensitivity of the LHC experiments. Contributions from dia- gramsthat do not depend on the electron Yukawa coupling Yee and are non-resonant e.g. H→^eeγ, are expected to be signifi- cantlylarger,althoughstillmuchsmallerthanpresentsensitivity.

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

The LHCoffers thebest constrainton Yee [18], whichmaybe larger than predicted by the SM. The SM forbids lepton-flavour- number-violating Higgs boson decays. There are strong indirect constraintsontheoff-diagonalYeμ coupling,thestrongestderived fromlimitsonthebranchingfractionof μ →^eγ andtheelectric dipole momentoftheelectron [19].However, theseindirectcon- straintsassumeSMvaluesfortheasyetunmeasuredY_eeandYμμ Yukawa couplings. Searchingfor H→^eμ allows Yeμ to be con- straineddirectly.

BothanalysespresentedinthisLettercloselyfollowthesearch forthe SM Higgsboson decay H→μμ [20]. Thesignal is sepa- ratedfromthebackgroundprimarilybyidentifyinganarrowpeak in the distribution of theinvariant mass ofthe two leptons m_

corresponding to the mass of the Higgs boson of 125 GeV [21].

The backgroundinthe ee search isdominatedby Drell–Yan(DY) Z/γ^∗ production,withsmallercontributions fromtop-quarkpair (tt)¯ anddibosonproduction( Z Z , W Z andW W ).Intheeμsearch, a much smalleryield ofSM background events is expected. The DYbackgroundonlycontributesthroughdecaysof Z/γ^∗_→τ τ_→

eντνeμντνμ.Thustheproductionoftopquarks,dibosons(mainly through W W →^eνeμν_μ), W+^{jetsand multijetevents,withjets} misidentifiedasleptons,aremoreimportantthanintheee search.

2. ATLASdetector

The ATLAS experiment [22–24] at the LHC is a multipurpose particle detector with a forward–backward symmetric cylindrical geometryandanear4π coverageinsolidangle.¹ Itconsistsofan

1 ATLASuses aright-handedcoordinatesystemwith itsoriginat thenominal interactionpoint(IP)inthecentreofthedetectorandthez-axisalongthebeam

https://doi.org/10.1016/j.physletb.2019.135148

0370-2693/©^{2019TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense}(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP³.

innertrackingdetector(ID)surroundedbyathinsuperconducting solenoidprovidinga2 Taxialmagneticfield,electromagneticand hadroncalorimeters,andamuonspectrometer.

TheIDcoversthepseudorapidityrange|η|<2.5.Itconsistsof silicon pixel, silicon microstrip, and transition radiation tracking detectors. Lead/liquid-argon (LAr) sampling calorimeters provide electromagnetic (EM) energy measurements with high granular- ity.Asteel/scintillator-tilecalorimeterinthecentral pseudorapid- ityrange|η_|<1.7 measurestheenergies ofhadrons.The endcap and forward regions are instrumented withLAr calorimeters for boththe EMandhadronicenergymeasurements up to |η_|=4.9.

The muon spectrometer (MS) surrounds the calorimeters up to

|η|=².7 and is based on three large air-core toroidal supercon- ducting magnets with eight coils each. The field integral of the toroidsrangesbetween2.0and6.0 T macrossmostofthedetec- tor.Themuonspectrometerincludesasystemofprecisiontracking chambersandfastdetectorsfortriggering.

Atwo-leveltriggersystemisusedtoselectevents [25].Itcon- sists of a first-level trigger implemented in hardware and using a subset of the detectorinformation to reduce theevent rateto 100 kHz. This is followed by a software-based high-level trigger thatemploysalgorithms similartothoseusedofflineandreduces therateofacceptedeventsto1 kHz.

3. Simulatedeventsamples

Samples of simulated signal events with a Higgs boson mass of mH =125 GeV were generated as described below and pro- cessed through the full ATLAS detectorsimulation [26] based on GEANT4 [27]. Higgs boson production via the gluon–gluon fu- sion (ggF) process was simulated using the POWHEG NNLOPS program [28–35] with the PDF4LHC15 set of parton distribution functions(PDFs) [36]. TheHiggsboson rapidity inthesimulation was reweighted to achieve next-to-next-to-leading-order (NNLO) accuracy in QCD [37]. Higgs boson production via vector-boson fusion(VBF)andwithanassociatedvectorboson(V H )weregen- erated atnext-to-leading-order (NLO) accuracy inQCD usingthe POWHEG-BOXprogram [38–40].The Z H sampleswere simulated forprocesseswithquark–quarkinitialstates,andthesmallcontri- butionfromgluon–gluoninitialstatesisaccountedforinthenor- malisationofthe Z H cross section.The parton-level eventswere processed with PYTHIA8 [41] for the decay of the Higgs bosons into theee or eμ final states andto simulateparton showering, hadronisation and the underlying event, using the AZNLO set of tuned parameters [42]. All samples were normalised to state-of- the-art predictions using higher-order QCD and electroweak cor- rections [43–66].Theeffectsarisingfrommultiplepp collisions in thesameorneighbouringbunchcrossings(pile-up)wereincluded inthesimulationbyoverlayinginelastic pp interactionsgenerated withPYTHIA8usingtheNNPDF2.3LOsetofPDFs [67] andtheA3 set of tuned parameters [68]. Events were reweighted such that thedistribution ofthe average numberofinteractions per bunch crossingmatchesthatobservedindata.Simulatedeventswerecor- rectedtoreflecttheleptonenergyscaleandresolution,andtrigger, reconstruction,identificationandisolationefficienciesmeasuredin data.

Toevaluatetheuncertaintyinthebackgroundmodellinginthe ee channel,adedicatedfastsimulationforthedominantDY back- groundwas used to produce a sample of 10⁹ events, equivalent

pipe.Thex-axispointsfromtheIPtothecentreoftheLHCring,andthe y-axis pointsupwards.Cylindricalcoordinates(r,φ)areusedinthetransverseplane,φ beingtheazimuthalanglearoundthez-axis.Thepseudorapidityisdefinedinterms ofthepolarangleθasη= −^{ln tan}(θ/2).Angulardistanceismeasuredinunitsof R≡

(η)²+ (φ)^2.

to 40 times the integratedluminosity ofthe data. For thissam- ple, Z/γ^∗_{+ (}0,1)-jet events were generated inclusively at NLO accuracyusingPOWHEG-BOX [69] withtheCT10PDFset [70].Ad- ditional Z/γ^∗₊2-jeteventsweregeneratedwithALPGEN [71] at leading-orderaccuracywiththeCTEQ6L1PDFset [72].Theevents were interfaced to PHOTOS [73] to simulate QED final-state ra- diation. The effects ofpile-up and a fastparameterisation of the responseofthedetectortoelectronsandjets,usingsimplesmear- ingfunctions,wasthenappliedtothegeneratedevents.

4. Eventselection

Events are recorded using triggers that require either an iso- latedelectronoranisolatedmuonaboveatransversemomentum (pT)thresholdof26 GeV [25,74].Electronsarereconstructedinthe range|η_|<2.47 from clustersofenergydepositsinthecalorime- ter matched to a track in the ID [75]. Muons are reconstructed in therange |η|<2.5 by combiningtracks inthe ID eitherwith tracksintheMSor,for|η|<0.1,withcalorimeterenergydeposits consistentwithamuon [76].Theelectronsandmuonsarerequired tobeassociatedwiththeprimarypp collisionvertex,whichisde- fined asthecollisionvertexwithlargestsumofp²_T oftracks,and to beisolated fromother tracks [75,76]. Eacheventmustcontain eitherexactlytwo electrons oran electronandamuon. Onelep- tonmusthave p_T>27 GeV toensureahightriggerefficiencyand theothermustbeofoppositechargeandhavep_T>15 GeV.

Requirementsonjetsareusedinthisanalysistosuppressback- groundanddefinea categorythathasa highsensitivitytosignal produced intheVBFproductionmode.Jetsintherange|η_|<4.5 and pT>30 GeV are reconstructed from energy deposits in the calorimeter [77],usingtheanti-k_t algorithm [78,79] witharadius parameterof0.4.Trackinginformationiscombinedusingamulti- variatelikelihoodtosuppressedjetsfrompile-upinteractions [80].

Backgrounds with top quarks are suppressed by identifying b-hadrons andneutrinosinthefinal state.Jetsintherange|η_|<

2.5 containingb-hadronsareidentifiedasb-jetsusingamultivari- atealgorithmthatusescalorimeterandtrackinginformation [81].

Eventsarerejectedifthereisatleastoneidentifiedb-jet.Different workingpoints are usedforthe ee and eμchannels becausethe latterhasalargertop-quarkbackground.Forthe ee (eμ) channel theb-jet identificationefficiencyisabout60%(85%)witharejec- tionfactorofabout1200(25)forlight-flavourjets [82].Neutrinos produced insemileptonic top-quark decays escape detectionand lead tomissingtransversemomentum E^miss_T ,reconstructedasthe magnitudeofthevectorsumofthetransversemomentaofallcal- ibrated leptons andjetsand additionalID tracks associated with the primary vertex(soft term) [83].Backgrounds with significant E^miss_T aresuppressedby requiring E^miss_T /√

HT<3.5(1.75)GeV^1/2 fortheee (eμ)channel,where HTisthescalarsumofthetrans- versemomentaofleptonsandjetsand√

HT isproportionaltothe E^miss_T resolution.

Backgroundfromtheprocess H→γ γ,wherethephotonsare misreconstuctedaselectrons,isstudiedwithsimulatedeventsand found to contribute about0.07% inthe ee channel fora H→^ee branching fractionat theexpectedlimit. It isthereforeneglected intherestoftheanalysis.

Thesearchisperformedintherangeofdileptoninvariantmass 110<m_<160 GeV, which allows the background to be de- termined withanalyticfunctionsconstrainedby thesidebands to eithersideofthepotentialsignal.

The eventsample passing thebasicleptonselection isdivided into seven (eight) categories for the ee (eμ) channel that differ intheirexpectedsignal-to-backgroundratios,toimprovetheover- all sensitivityof the search. Thesecategories are based on those

usedinRef. [20], andarefoundtoprovidegoodsensitivityinthe presentanalyses.

First,a low-pT lepton category ‘Low p^_T’ is definedin the eμ

channel with events in which the subleading lepton has p_T<

27 GeV.Thisregionhasasignificantfractionofeventsinwhichei- therreconstructedleptonisofnon-promptoriginorisamisidenti- fiedphotonorhadron,hereaftercalledafakelepton.Theseevents arenotseparatedoutintheee channelbecausetherelativecontri- butionfromfakeleptonsissmaller.Acategoryenriched inevents fromVBF productionisdefinedfromtheremainingeventsbyse- lectingthosecontainingtwojetswithpseudorapiditiesofopposite signs,apseudorapidityseparation|ηj j|>3 andadijetinvariant massm_{j j}>500 GeV.

Events that fail to meet the criteria ofthe ‘Low p^_T’ and VBF categoriesareclassifiedas‘Central’ifthepseudorapiditiesofboth leptons are |η^|<1 or as ‘Non-central’ otherwise. For each of thesetwo categories, threeranges inthedileptontransversemo- mentum p^_T are considered: ‘Low p^_T’ (p^_T ≤^{15 GeV),‘Mid p}_T^’ (15<p^_T ≤^{50 GeV),and ‘High p}T’ (p^_T >50 GeV). These cate- gories exploit differences in the dilepton mass resolution, which is better for more central leptons, as well as differences in the expected signal-to-background ratio between the signal and the backgroundprocesses asfunction ofdilepton transverse momen- tumandrapidity.

5. Signalandbackgroundparameterisation

Analyticfunctionsareusedtodescribethem_distributionsfor boththesignalandthebackground.The H→^{ee andH}→^eμsig- nals consideredare narrow resonances witha massanda width set to the SM values of m_H =125 GeV and 4.1MeV respec- tively.Theobservedsignalshapesarethusdeterminedbydetector resolution effects and are parameterised as a sum of a Crystal Ballfunction (F_CB) [84] and a Gaussian function (F_GS) following Ref. [20]:

PS(m_) = ^fCB×^FCB(m_|^mCB,σCB,α,n) + (¹− ^fCB)×^FGS



m_|^mGS,σ_GS^S.

Theparameters αandn definethepower-lawtailoftheF_CBdistri- bution,whilemCB,mGS, σCB,and σ_GS^S denotethe FCB meanvalue, FGS meanvalue, FCB width,and FGS widthrespectively.Therela- tivenormalisationbetweenthetermsisgovernedbytheparame- ter f_CB.Theseparameters aredeterminedbyfittingthesimulated signalm_ distributionineachcategory.Intheee(eμ)channelthe signalmassresolutionvariesbetweenabout2.0GeV (2.3GeV) for thecentraland2.9GeV (3.0GeV)forthenon-centralcategories.

The background parameterisation for the ee channel follows Ref. [20] asthebackgroundisverysimilar.Them_eedistributionsin eachcategory aredescribed by asumofaBreit–Wigner function (FBW)convolvedwitha FGS,andan exponentialfunction divided byacubicfunction:

P_B(m_ee) = ^f× [^FBW(m_ee|^mBW,BW)⊗^FGS(m_ee|σ_GS^B)]

+ (¹−^f)×^{C eA·mee}/m³_ee,

where f representsthefractionofthe F_BW componentwheneach individualcomponent isnormalisedto unityand C isa normali- sationcoefficient.The σ_GS^B parameterineach category isfixed to thecorrespondingaveragemresolutionasdeterminedfromsim- ulated signal events. For all the categories, the FBW parameters are fixed tom_BW=⁹¹.2 GeV and BW=².49 GeV [85]. The pa-

rameters f and A and the overall normalisation are left free to bedeterminedinthefitanduncorrelatedbetweendifferentcate- gories.

ABernstein polynomial ofdegreetwo isused toparameterise the meμ distributionof the backgroundineach of the eightcat- egories in the eμ channel, with parameters uncorrelated across categories. The choice of background function is validated by an F-testconsideringBernstein polynomialsoffirst,second andthird degree.

The signal yield, which is allowed to be positive ornegative, is constrained using separate binned maximum-likelihood fits to theobservedm_ distributionsintherange110<m_<160 GeV inthetwo channels.Thefits areperformedusingthesumofthe signalandbackgroundmodels(‘S+^{B model’)andareperformed} simultaneouslyinallthecategories.Inadditiontothebackground- modelparametersdescribedearlier,thebackgroundnormalisation ineach categoryandthebranchingfractionofthesignal arefree parametersinthefit.

6. Systematicuncertainties

Thesignalexpectationissubjecttoexperimental andtheoreti- caluncertainties,whicharecorrelatedacrossthecategories.

The uncertainty in the combined2015–2018 integratedlumi- nosity is 1.7% [86], obtained usingthe LUCID-2detector [87] for the primary luminosity measurements. Other sources of experi- mentaluncertaintyincludethe electronandmuontrigger, recon- struction,identificationandisolationefficiencies [75,76], theb-jet identification efficiency [81], the pile-up modelling [88], the de- termination ofthe E^miss_T soft term [83], andthe jet energy scale andresolution [89].Theuncertainties intheelectronenergyscale andresolution [75] andinthemuonmomentumscaleandresolu- tion [76] affecttheshape ofthesignal distributionaswell asthe signalacceptance.

Thetotalexperimentaluncertaintyinthepredictedsignalyield in each ggF category is between 2% and 3% for the ee channel and between4% and6% forthe eμ channel. It is dominated by the luminosity, E^miss_T soft term and pile-up effects, and the last twocontributions arelargerintheeμ analysisduetothetighter E^miss_T /√

HT requirement. TheexperimentaluncertaintyintheVBF category isbetween7% and15% fortheee channel andbetween 6% and22% fortheeμ channel,dueto largercontributions from thejetenergyscaleandresolution.

Thetheoreticaluncertainties intheproductioncrosssection of the Higgs boson are taken from Ref. [43]. In addition, theoreti- calmodellinguncertaintiesaffectingtheacceptanceforthesignals are calculated separately for the ggF and VBF Higgs boson pro- ductionprocessesineachanalysiscategory.Theuncertaintyinthe acceptanceforthe V H processisneglected.Theeffectsofmissing higher-order termsin the perturbative QCD calculationsare esti- matedbyvaryingtherenormalisationandfactorisationscales.For theggF processtheuncertainties are approximatedastwocorre- latedsources that rangefromaround 1% to11% forthe different analysiscategoriesinboth channels. FortheVBF process theun- certaintiesintheacceptanceduetotheQCDscalesarefoundtobe small.Theeffectsofuncertainties inthepartondistributionfunc- tionsandthevalueof αSareestimatedusingthePDF4LHC15rec- ommendations [36] andfoundtobeverysmall.Theuncertaintyin themodellingofthepartonshower,underlyingevent,andhadro- nisation isassessedbycomparingthe acceptanceofsignal events showeredbyPYTHIAwiththatofeventsshoweredbyHERWIG [90, 91].Thetotalvariationsduetotheseuncertaintiesrangefromless than 1%to 11% fortheggF signalprocess andfrom1% to8% for theVBFsignalprocessdependingontheanalysiscategory.