Evidence for the production of three massive vector bosons with the ATLAS detector

(1)

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Evidence for the production of three massive vector bosons with the ATLAS detector

.TheATLASCollaboration

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

Articlehistory:

Received26March2019

Receivedinrevisedform28August2019 Accepted2September2019

Availableonline10September2019 Editor:M.Doser

A search forthe production ofthree massive vectorbosons inproton–proton collisions is performed usingdataat√

s=^{13 TeV}^recorded^with^the^ATLAS^detector^at^the^Large^Hadron^Colliderⁱⁿ^the^years 2015–2017,correspondingtoanintegratedluminosityof79.8 fb⁻¹.Eventswithtwosame-signleptons (electronsormuons)andatleasttworeconstructedjetsareselectedtosearchforW W W→ννqq.

Eventswith threeleptons withoutanysame-flavouropposite-sign leptonpairs areused to searchfor W W W→ννν,whileeventswiththreeleptonsandatleastonesame-flavouropposite-signlepton pairand oneormorereconstructedjetsare usedtosearch forW W Z→νqq.Finally,events with fourleptons are analysed tosearchfor W W Z→νν and W Z Z→^qq.Evidenceforthe joint productionof threemassive vector bosonsisobserved withasignificance of4.1 standard deviations, wheretheexpectationis3.1standarddeviations.

©²⁰¹⁹^TheÂuthor(s).^Published^byÊlsevier^B.V.^Thisîsânôpenâccessârticleûnder^the^CC^BY^license (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

Thejointproductionofthreevectorbosonsisarareprocessin theStandardModel (SM).Studies oftribosonproductioncan test thenon-AbeliangaugestructureoftheSM theoryandanydevia- tionsfromtheSM predictionwouldprovidehintsofnewphysics athigherenergyscales [1–4].Tribosonproductionhasbeenstud- iedat the LargeHadron Collider (LHC)using proton–proton(pp) collisiondatatakenat√

s=^{8 TeV for}^processes^such^as γ γ γ [5], Wγ γ [6,7],Zγ γ [8,7],W Wγ andW Zγ [9,10],andW W W [11].

Thisletterpresents theﬁrst evidenceforthe jointproduction ofthreemassive vector bosonsin ppcollisions using thedataset collected with the ATLAS detector between 2015 and 2017 at

√s=^{13 TeV.} ^At ^leading ^order^(LO) ⁱⁿ^quantum chromodynamics (QCD),theproductionofthreemassivevectorbosons(V V V,with V =^W,Z)canproceedviatheradiationofeachvectorbosonfrom afermion,froman associatedboson productionwithaninterme- diateboson(W, Z/γ^∗ or H) decayingintotwo vectorbosons,or fromaquarticgaugecouplingvertex. RepresentativeFeynmandi- agramsareshowninFig.1.

Twodedicatedsearchesareperformed,onefortheW^±W^±W^∓ (denotedas W W W)process andoneforthe W^±W^∓Z (denoted asW W Z)andW^±Z Z (denotedasW Z Z)processes.Tosearchfor the W W W process, eventswith two same-sign leptons with at leasttwojetsresultingfromW W W→ννqq(=^e,μ,including

τ_→νν) or threeleptons resultingfrom W W W →ννν are

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

consideredandarehereafterreferredtoastheννqqandννν

channels,respectively.TosearchfortheW W ZandW Z Z (denoted as W V Z) processes, events withthree or four leptons resulting fromW V Z→νqq,W W Z→νν,andW Z Z→^qqare used. Selection criteria are chosen in order to ensure there is no overlapbetweendifferentchannels. Adiscriminant that sepa- ratesthe W W W or W V Z signalfromthebackgroundisdefined in each channel. The discriminants are combinedusing a binned maximum-likelihood fit, which allows the signal yield and the backgroundnormalisationstobeextracted.Thecombinedobserv- ableisthesignalstrengthparameter μdefinedastheratioofthe measured W V V cross section to its SM expectation, where one commonratioisassumedforW W W andW V Z.

2. TheATLASdetector,dataandsimulationsamples

TheATLASdetector [12–14] isamulti-purposeparticledetector comprisedofaninnerdetector(ID)surroundedbya2 Tsupercon- ductingsolenoid,electromagnetic(EM)andhadroniccalorimeters, anda muon spectrometer (MS) withone barrel andtwo endcap air-coretoroids.TheIDconsistsofasiliconpixeldetector,asilicon microstrip detector,anda transitionradiation tracker,andcovers

|η_|<2.5 in pseudorapidity.¹ The calorimeter system covers the

1 ATLASuses aright-handedcoordinatesystemwith itsoriginat thenominal interactionpoint(IP)inthecentreofthedetectorandthez-axisalongthebeam pipe.Thex-axispointsfromtheIPtothecentreoftheLHCring,andthe y-axis points upward.Cylindricalcoordinates(r,φ)areusedinthetransverseplane,φ https://doi.org/10.1016/j.physletb.2019.134913

0370-2693/©²⁰¹⁹^TheÂuthor(s).^Published^byÊlsevier^B.V.^Thisîsânôpenâccessârticleûnder^the^CC^BY^license(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP³.

(2)

Fig. 1.Representative Feynman diagrams at LO for the production of three massive vector bosons, including diagrams sensitive to triple and quartic gauge couplings.

pseudorapidityrange |η_|<4.9.The MS providesmuon triggering capabilityfor|η|<2.4 andmuonidentiﬁcationandmeasurement for|η|<2.7.Atwo-leveltriggersystem [15], usingcustomhard- warefollowedbyasoftware-basedtriggerlevel,isusedtoreduce theeventratetoanaverageofaround1 kHzforoﬄinestorage.

The data used were collected between 2015and 2017 in pp collisionsat√

s=^{13 TeV.}Ônlyêvents^recorded^withâ^fullyôper- ationaldetector andstablebeamsare included. Candidateevents areselectedbysingleisolated-lepton(eor μ)triggerswithtrans- versemomentumthresholdsvaryingfrompT=^{20 GeV to}^{26 GeV} (dependingontheleptonflavour andrunperiod)orsingle-lepton triggers with thresholds of pT=^{50 GeV for} ^muons ând ^pT= 60 GeV forelectrons.Duetothepresenceoftwo,threeorfourlep- tonsinthefinalstate,thesesingle-leptontriggersarefullyefficient forthetribosonsignalsinthesignalregionsdefinedinSections4 and5.Theresultingtotalintegratedluminosityis79.8fb⁻¹.

Signal andbackground processes were simulatedwith several MonteCarlo(MC) eventgenerators, whilethe ATLAS detectorre- sponsewasmodelled [16] withGeant4[17].Theeffectofmultiple pp interactions in the same and neighbouring bunch crossings (pile-up) was included by overlaying minimum-bias events sim- ulated with Pythia 8.186 [18] interfaced to EvtGen 1.2.0 [19], referred toasPythia 8.1in thefollowing, andusingthe A3[20]

setoftuned MCparameters, oneachgenerated eventinall samples. Triboson signal events [21] were generated using Sherpa 2.2.2 [22–24] with the NNPDF3.0NNLO [25] parton distribution function(PDF)set,whereallthreebosonsareon-mass-shell,using a factorised approach [26]. Events with an off-mass-shell boson through W H→^{W V V}^∗ ^and ^{Z H}→^{Z V V}^∗ ^were ^generated ^using Powheg-Box 2 [27–32] interfaced to Pythia 8.1 for the W W W analysis, while for the W V Z analysis only Pythia 8.1 was used.

ThegeneratorwasinterfacedtotheCT10[33] (NNPDF2.3LO[34]) PDF andtheAZNLO[35] (A14 [36])setof tunedMC parameters fortheW W W (W V Z)analysis.Bothon-mass-shellandoff-mass- shellprocessesweregeneratedatnext-to-leadingorder(NLO)QCD accuracy [37–40] and are included in the signal deﬁnition. The expected cross sections for W W W and W W Z production are 0.50 pb and0.29 pb, respectively,with an uncertainty of∼¹⁰^%, evaluated by varying parameters inthe simulationrelated tothe renormalisation and factorisation scales, parton shower and PDF sets.

Diboson(W W,W Z,Z Z) [26],W/Z+γ [21] andsingleboson (W/Z+jets) [41] production,aswellaselectroweakproductionof W^±W^±+^{2 jets,}^{W Z}+^{2 jets,}^and^{Z Z}+^{2 jets,}^were^modelled^us- ingSherpa2.2.2withtheNNPDF3.0NNLOPDFset.Inordertoim- provetheagreementbetweenthesimulatedandobservedjetmul- tiplicitydistributionsforthe W Z→ν and Z Z→events, ajet-multiplicitybasedreweighting was appliedtothesimulated W Z and Z Z samples. Top-quark pair events (tt)¯ were generated using Powheg-Box 2 [42] interfaced to Pythia 8.230 [43]

beingtheazimuthalanglearoundthebeampipe.Thepseudorapidityisdeﬁnedin termsofthepolarangleθ asη= −^{ln tan}(θ/2).Angulardistanceismeasuredin unitsofR=

(η)²+(φ)².

and EvtGen 1.6.0. The NNPDF3.0NLO PDF set was used for the matrix-element calculation, while the NNPDF2.3LO PDF set was used for the showering with the A14 set of tuned parameters.

Other background processes containing top quarks were generated with MadGraph5_aMC@NLO [44] interfaced to Pythia 8, at LO (t¯tγ, t Z, t¯t W W, and ttt¯¯t) or at NLO (tt W¯ , t¯t Z, and t¯t H),withMadGraph5_aMC@NLOinterfacedtoHerwig[45] (t W Z and t W H) or with Powheg-Box 2 [46] interfaced to Pythia 6 (t W).

3. Objectdeﬁnitionsandselectioncriteria

Selected events are required to contain at least one reconstructed primary vertex. If more than one vertex is found, the vertexwiththelargest p²_T sumofassociatedID tracksisselected astheprimaryvertex.

Electrons are reconstructed as energy clusters in the EM calorimeterthatarematchedtotracksfoundintheID.Muonsare reconstructed by combining tracks reconstructed in the ID with tracksortracksegments foundintheMS. Leptonsneedtosatisfy pT>15 GeV and have|η|<2.47 forelectrons (electrons within thetransitionregionbetweenthebarrelandendcapcalorimeters, 1.37<|η_|<1.52,areexcluded)and|η_|<2.5 formuons. Leptons are required to be consistent with originating from the primary vertexbyimposingrequirementsonthetransverseimpactparam- eter, d0, its uncertainty, σd₀, the longitudinal impact parameter, z0, and the polar angle θ. These requirements are |^d0|/σd0 <5 and |^z0×^sinθ|<0.5 mm for electrons, and |^d0|/σd0 <3 and

|^z0×^sinθ|<0.5 mm for muons. Electrons have to satisfy the likelihood-based“Tight”qualitydefinitionwhichresultsinefficien- ciesof58% at E_T=⁴.5 GeV to88% atE_T=100 GeV [47].Forthe W W W (W V Z) analysis, muonsare requiredto pass the “Medi- um”(“Loose”)identificationcriteriawhichresultsinefficienciesof approximately96% (98%)formuonsfroma Z→μμsample [48].

To reject jetsmisidentiﬁed asleptons orleptons from hadron decays (including b- and c-hadron decays), referred to as “non- prompt” leptons in the following,leptons are requiredto be iso- lated from other particles in both the calorimeters and the ID.

Theleptonisolation conesizeisatmostR=⁰.2,exceptforthe muon isolationintheID,whereitisatmostR=⁰.3.Electrons are requiredto pass the “Fix (Loose)” isolation requirement [49]

andmuons arerequired topass the “Gradient”(“FixedCutLoose”) isolation requirement [48] for the W W W (W V Z) analysis. The identification andisolation requirements formuonsare more re- strictive in the W W W analysis because a larger contamination fromnon-promptleptonsisexpected.TheelectronFix(Loose)isolation requirement results in an efficiency above 95% [47]. The muon isolationefficiencyisabove 90%(99%)fortheGradientiso- lation criteria for muons with pT of 25 GeV (60 GeV), and the FixedCutLooseefficiencyisabove95% [48].

A dedicatedboosteddecision tree(BDT), termed“non-prompt lepton BDT” [50], is used to reject leptons likely to originate fromheavy-ﬂavourdecays.Inaddition,electrons havetopassthe

“chargemisidentiﬁcationsuppressionBDT” [49] torejectelectrons

(3)

likely to have the electric charge wrongly measured. The non- prompt lepton BDT uses isolation and b-tagging information de- rivedfromenergydepositsandtracksinaconearoundthelepton direction. The charge misidentification suppression BDT uses the electrontrack impact parameter, thetrack curvature significance, the cluster width and the quality of the matching between the cluster andits associated track.Leptons passing all requirements listedabovearereferredtoas“nominal”leptons.Thecombination ofisolation, non-prompt lepton BDT andcharge misidentification suppressionBDT criteria resultsin an efficiencyof 65% (89%)for electrons with pT of 20 GeV (80 GeV) [47]. The combinationof isolation and non-prompt lepton BDT results in an efficiency of 65%(99%) formuonsthat havetheGradient isolation with p_T of 20 GeV (100 GeV),andanefficiencyof74%(99%)formuonsthat havetheFixedCutLooseisolationwithpTof20 GeV (80 GeV) [48].

Jetsarereconstructed fromcalibratedtopologicalclustersbuilt fromenergydepositsinthecalorimeter [51] usingtheanti-ktalgo- rithmwitharadiusparameterof0.4 [52,53] andcalibratedusing thetechniquesdescribedinRef. [54].Jetcandidatesarerequiredto havep_T>20 GeV and|η|<2.5.Torejectjetslikelytobe arising frompile-up collisions, an additional criterion using the jet ver- textagger [55] discriminantis appliedforjetswith pT<60 GeV and|η|<2.4.Jetscontainingb-hadrons(b-jets)areidentiﬁedbya multivariate discriminantcombining informationfrom algorithms usingsecondaryverticesreconstructedwithinthejetandtrackim- pactparameters [56,57], withan eﬃciency of 85% (70%) for the W W W (W V Z)analysis.

The missing transverse momentum, whose magnitude is denoted E^miss_T , is deﬁned as the negative vector sum of the p_T of all reconstructed and calibrated objects in the event. This sum includes a term to account for the energy from low-momentum particlesthat arenot associatedwithanyofthe selectedobjects, andiscalculatedfromIDtracksmatchedtothereconstructedpri- mary vertex in the event [58]. The sum also includes jets with

|η_|>2.5 and pT>30 GeV.

Theobjectreconstruction andidentiﬁcationalgorithms do not always resultinunambiguous identiﬁcations.An overlapremoval algorithmisthereforeapplied. Electronssharinga trackwithany muons are removed. Any jet within R<0.2 of an electron is removedandelectrons withinR<0.4 ofanyremainingjetsare removed. Jets with lessthan three associated tracks and within R<0.2 ofamuonareremoved,andmuonswithin R<0.4 of anyoftheremainingjetsareremoved.

Atleastonereconstructed“trigger”leptonwithaminimumpT is required to match within R<0.15 a lepton with the same ﬂavourreconstructed by thetriggeralgorithm.The thresholdsfor the trigger (other) leptons are 27 GeV (20 GeV) for the W W W analysis,andfrom21 GeV to27 GeV (15 GeV),dependingonthe runperiodandleptonﬂavour,forthe W V Zanalysis.

4. AnalysistargetingW W W

Theexperimental signatureofthe ννqq processisthe presence of two same-sign leptons, E^miss_T , and two jets. The signa- tureof the ννν process is the presence ofthree leptons and E^miss_T .Toreducethebackgroundcontributionsfromprocessesthat havemorethantwo(three)leptonsintheννqq(ννν)channel a “veto lepton” definition is introduced. Compared with the nominallepton selection criteriadescribed inSection 3, the veto lepton pT threshold islowered to7 GeV, andthe isolation, non- promptleptonBDT,charge misidentificationsuppressionBDT, and impact parameter requirements are removed. For veto electrons, the likelihood-based Loose identification definition [49] is used.

For veto muons, the Loose identiﬁcation deﬁnition [48] is used, andthepseudorapidityrangeisextendedto|η|<2.7.

To select ννqq candidates, events are required to have ex- actlytwonominalleptonswithpT>20 GeV andthesameelectric charge, at least two jets, and no identiﬁed b-jets. Four regions are considered, based on the lepton ﬂavour, namely ee, eμ, μe, and μμ,whereeμdenotes thehighest-pT (leading) leptonbeing an electron, while μe denotes the leading lepton beinga muon.

Events with an additional veto lepton are removed. The invari- ant mass of the dilepton system is required to be in the range 40<m<400 GeV.Theupper masslimitreduces thecontribu- tionfromtheW Z+^jets^process.^The^leading(sub-leading)jetmust have pT>30(20)GeV and|η_|<2.5. Thedijetsystem, formedby thetwojetswiththelargestp_T,isrequiredtohavem_{j j}<300 GeV and|ηj j|<1.5,wherem_{j j} isthe dijetinvariant massandηj j is the pseudorapidity separation between the two jets. The cuts appliedonthe dijetsystemmainlyreduce thecontributionsfrom thesame-sign W W vectorboson scatteringprocess.Additionally, intheeefinalstate, E^miss_T isrequiredtobeabove55 GeV andm mustsatisfym<80 GeV orm>100 GeV, toreduce contami- nationfrom Z→êe ^where^the^charge ôfône êlectronîs^misiden- tified.Thism cutisnot appliedinthe μμfinal state,since the muonchargemisidentificationrateisfoundtobenegligible,noris itappliedintheeμand μefinalstates,wherethecontamination fromZ eventsissmall.

Toselectνννcandidates,eventsarerequiredtohaveexactly threenominalleptons with pT>20 GeV and noidentiﬁedb-jets.

Eventswithanadditionalvetoleptonareremoved.Toreducethe contributionfromtheW Z→ν process,eventsarerequiredto haveno same-ﬂavouropposite-sign (SFOS)lepton pairs,andthus only μ^±e^∓e^∓ande^±μ^∓μ^∓eventsareselected.

AmajorbackgroundoriginatesfromtheW Z+^jets→ν+^jets process, contributing to the ννqq channel when one lepton is notreconstructedoridentiﬁed,ortotheννν channel,whena Z bosondecaysintoapairof τ leptonsbothofwhichdecaytoan electronormuon.Simulationisusedtoestimatethisbackground.

The W Z+^jets^modellingîs^tested ⁱⁿâ ^{W Z}^-dominated ^validation regiondefinedbyselectingeventswithexactlythreenominallep- tonswithoneSFOSleptonpair.Inaddition,eventsarerequiredto havenob-jets reconstructed,E_T^miss>55 GeV andthetrileptonin- variant mass m>110 GeV. Data and simulation agree in this validationregion, asshowninFig. 2(a)forthe leading lepton pT distribution.

ContributionsfromSMprocessesthatproduceatleastonenon- prompt lepton are estimated using a data-driven method as de- scribed inRef. [59] by introducing“fake” leptons. The definitions of nominal and fake leptons are mutually exclusive. Fake electrons have to satisfy the likelihood-based Medium [49] but fail theTightidentification, andtheisolation, non-promptleptonBDT andchargemisidentificationsuppressionBDTrequirementsarere- moved.Fakemuonshavetheimpactparameterrequirementsloos- ened to|^d0|/σd0<10,andboth isolation andnon-promptlepton BDT requirementsareremoved.Additionally, theyhavetofailthe nominalmuondefinition.Simulationshowsthat thet¯t process is the dominantcontributorofeventswithfake leptons, withmore than90%intheννqqchannelandmorethan95%intheννν

channeloriginatingfromthisprocess.Eventscontainingone(two) nominallepton(s)andonefakeleptonwithpT>20 GeV arescaled by a “fake factor” to predict the non-prompt lepton background contribution in the ννqq (ννν) channel. The fake factor is theratioofthenumberofnon-promptleptons passingthenomi- nalleptoncriteriaoverthenumberpassingthefakeleptoncriteria.

Its value is derived from two t¯t-enriched regions selected with twoorthreeleptons(noSFOSleptonpairs)andexactlyoneb-jet.

Oneofthesame-signleptonspasseseithernominalorfakelepton criteria, while the other lepton(s) must pass the nominal lepton