Physics Letters B 800 (2020) 135069
Contents lists available atScienceDirect
Physics Letters B
www.elsevier.com/locate/physletb
Searches for lepton-flavour-violating decays of the Higgs boson in
√ s = 13 TeV pp collisions with the ATLAS detector
.TheATLASCollaboration
a r t i c l e i n f o a b s t ra c t
Articlehistory:
Received13July2019
Receivedinrevisedform27August2019 Accepted13September2019
Availableonline4November2019 Editor:M.Doser
This Letter presents directsearches forlepton flavour violationin Higgs bosondecays, H→eτ and H→μτ,performed with the ATLASdetector at the LHC. The searches are based on adata sample of proton–proton collisions ata centre-of-mass energy √
s=13 TeV, corresponding to an integrated luminosityof36.1 fb−1.NosignificantexcessisobservedabovetheexpectedbackgroundfromStandard Model processes. The observed (median expected) 95% confidence-level upper limits onthe lepton- flavour-violatingbranchingratiosare0.47% (0.34+−00..1310%)and0.28% (0.37+−00..1410%)forH→eτandH→μτ, respectively.
©2019TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.
1. Introduction
The search for processes beyond the Standard Model (SM) is one of the main goals of the Large Hadron Collider (LHC) pro- gramme at CERN. A possible sign of such processes is lepton flavour violation (LFV) in decays of the Higgs boson [1,2]. Many beyond-SMtheories predict LFV decaysof the Higgsboson, such assupersymmetry [3,4], other modelswithmore thanone Higgs doublet [5,6], composite Higgs models [7], models with flavour symmetries [8] or warped extra dimensions [9–11] models and others [12,13].
In this Letter, searches for LFV decays of the Higgs boson, H→eτ andH→μτ,attheLHCwiththeATLASexperimentare presented.Studiesare basedonproton–proton(pp)collisiondata recordedin 2015–2016 at a centre-of-mass energy √
s=13 TeV.
Thedatasetcorrespondstoanintegratedluminosityof36.1 fb−1. PreviousATLASsearches [14,15] placedanupperlimitof1.04%
(1.43%)onthe H→eτ (H→μτ)branchingratio(B)witha95%
confidencelevel(CL)usingRun 1datacollectedat√
s=8 TeV,cor- respondingtoanintegratedluminosityof20.3 fb−1.TheCMSCol- laborationrecentlyprovided95%CLupperlimitsonthesebranch- ingratiosof0.61%and0.25%,respectively,usingdatacollectedat
√s=13 TeV,withanintegratedluminosityof35.9 fb−1 [16].
Thesearchespresentedhereinvolvebothleptonic(τ→ νν¯1) andhadronic(τ→hadrons+ν)decaysof τ-leptons,denoted τ and τhad respectively. The dilepton final state τ only consid- erspairsofdifferent-flavourleptons.Same-flavourleptonpairsare
E-mailaddress:atlas.publications@cern.ch.
1 Unlessexplicitly mentionedotherwise, leptons(denotedbyor )referto electronsormuons.
rejected due to the large lepton pair-production Drell-Yan back- ground.Twochannelsareconsideredforeachofthetwosearches:
eτμ and eτhad for the H→eτ search, μτe and μτhad for the H→μτ search. The analysisisdesignedsuch that anypotential LFVsignal overlapbetweenthe H→eτ and H→μτ searchesis negligible.Manymethodsarereusedfromthemeasurementofthe Higgsbosoncross-sectionintheH→τ τ finalstate [17].
The ATLAS detector2 is described in Refs. [18–20]. It con- sists of an inner tracking detector covering the range |η|<2.5, surrounded by a superconducting solenoid providing a 2T ax- ialmagneticfield,high-granularityelectromagnetic(|η|<3.2)and hadroniccalorimeters (|η|<4.9), anda muon spectrometer(MS) which coversthe range|η|<2.7 and includes fasttriggercham- bers(|η|<2.4)andsuperconductingtoroidalmagnets.
2. Simulationsamples
SamplesofMonteCarlo(MC)simulatedeventsareusedtoop- timizetheeventselection,andtomodelthesignal andseveralof the background processes. The samples were produced with the ATLAS simulationinfrastructure [21] usingthefull detectorsimu- lationperformedbythe Geant4 [22] toolkit.TheHiggsbosonmass wassettomH=125 GeV [23].The fourleadingHiggsbosonpro- ductionmechanismsareconsidered:thegluon–gluonfusion(ggF), vector-boson fusion (VBF) and two associated production modes
2 ATLASuses aright-handedcoordinatesystemwith itsoriginat thenominal interactionpointinthecentreofthedetectorandthez-axisalongthebeampipe.
Theazimuthalangleφrunsaroundthebeampipe,thepseudorapidityisdefinedin termsofthepolarangleθasη≡ −ln tan(θ/2).Angulardistanceintheη–φspace isdefinedas R≡
( η)2+ ( φ)2. https://doi.org/10.1016/j.physletb.2019.135069
0370-2693/©2019TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP3.
Table 1
Generatorsusedtodescribethesignalandbackgroundprocesses,partondistributionfunction(PDF)setsforthehardprocess,and modelsusedfor partonshowering, hadronizationandtheunderlyingevent(UEPS).Theordersofthetotalcross-sectionsusedtonormalizetheeventsarealsogiven.MoredetailsaregiveninRef. [17].
Process Generator PDF UEPS Cross-section order
ggF Powheg-Box v2 [26–30] NNLOPS [31] PDF4LHC15 [32] NNLO Pythia 8.212 [25] N3LO QCD + NLO EW [33–36]
VBF Powheg-Box v2 MiNLO [30] PDF4LHC15 NLO Pythia 8.212 ∼NNLO QCD + NLO EW [37–39]
W H, Z H Powheg-Box v2 MiNLO PDF4LHC15 NLO Pythia 8.212 NNLO QCD + NLO EW [40–42]
W/Z+jets Sherpa 2.2.1 [43] NNPDF30NNLO [44] Sherpa 2.2.1 [45] NNLO [46,47]
V V/Vγ∗ Sherpa 2.2.1 NNPDF30NNLO Sherpa 2.2.1 NNLO
tt¯ Powheg-Box v2 [26–28,48] CT10 [49] Pythia 6.428 [50] NNLO+NNLL [51]
Single t Powheg-Box v1 [52,53] CT10 Pythia 6.428 NLO [54–56]
(W H , Z H ),whiletheothersgivenegligiblecontributions andare ignored.Thecross-sectionsofallHiggsbosonproductionprocesses werenormalizedtotheSM predictions [24].TheLFVHiggsboson decaysaswell asthe H→τ τ andH→W W background decays weremodelledwithPythia 8 [25].Otherbackgroundprocessesin- volve electroweak productionof W/Z bosons via VBF, Drell–Yan productionof W/Z inassociation withjet(s) as well asdiboson, single top-quarkandtop-quarkpair(t¯t)production.TheMC gen- eratorsusedfortheSM H→τ τ cross-sectionmeasurement [17]
werealsoemployedhereforallbackgroundcomponents.Thegen- eratorsandpartonshower modelsusedtosimulatedifferentpro- cessesaresummarizedinTable1.
3. Objectreconstruction
The correct identification of H→ τ events requires recon- struction of severaldifferent objects (electrons, muons, and jets, includingthoseinitiatedbyhadronicdecaysof τ-leptons)andthe missing transverse momentum pmissT , whose magnitude is called EmissT .
Electrons are reconstructed by matching tracks in the in- ner detector to clusteredenergy deposits in the electromagnetic calorimeter [57]. Loose likelihood-based identification [58], pT>
15 GeV and fiducial volume requirements (|η|<2.47, excluding thetransitionregionbetweenthebarrelandtheendcapcalorime- ters 1.37<|η|<1.52) are applied.Medium identification, corre- sponding to an efficiencyof 87% at pT=20 GeV, is imposed for thebaselineelectronselection.
Muonsare identified by tracks reconstructed inthe inner de- tectorandmatchedto tracksintheMS. Looseidentification [59], pT>10 GeV and |η|<2.5 requirements are applied. Medium identification(efficiencyof96.1%formuonswith pT>20 GeV) is imposedforthebaselinemuonselection.
Isolation criteriaexploiting calorimeter and track-based infor- mation are applied to both electrons and muons. The gradient working point is used, featuring an efficiency of 90% (99%) ob- tainedforleptonswithpT>25 GeV (60 GeV)originatingfromthe Z→ process [58,59].
Jetsare reconstructed using the anti-kt algorithm [60] as im- plemented by the FastJet [61] package.The algorithm is applied to topologicalclustersof calorimetercells [62] with aradius pa- rameter R=0.4. Only jets with pT>20 GeV and |η|<4.5 are considered.Jetsfromother pp interactionsinthesameandneigh- bouringbunch crossings(pile-up)are suppressedusingjet vertex tagger(JVT) algorithms [63,64]. Jetscontaining b-hadrons (b-jets) areidentified by theMV2c20algorithm [65,66] in thecentral re- gion (|η|<2.4). A working point corresponding to 85% average efficiencydetermined forb-jets int¯t simulatedeventsis chosen, rejectionfactorsare2.8and28againstc-jetsandlight-flavourjets respectively.
The reconstruction of the object formed by the visible prod- uctsofthe τhad decay(τhad-vis)beginsfromjetsreconstructedby
the anti-kt jet algorithm with a radius parameter R=0.4. Infor- mation fromtheinnerdetectortracksassociatedwiththeenergy deposits in the calorimeteris incorporatedin the reconstruction.
Only τhad-vis candidates with pT>20 GeV and|η|<2.5 arecon- sidered.3 One or three associated tracks with an absolute total charge |q|=1 are required. An identification algorithm [67,68]
based on boosted decisiontrees (BDT) [69–71] is used to reject
τhad-vis candidates arising from misidentification of jets or from decays of hadrons with b- or c-quark content. Unless otherwise indicated, atightidentification(ID)workingpoint isusedforthe
τhad-vis, corresponding to an efficiency of 60% (45%) for 1-prong (3-prong)candidates.Jetscorresponding toidentified τhad-vis can- didates are removed from the jet collection. The τhad-vis candi- dates withonetrackoverlappingwithanelectroncandidatewith high ID score,as determined by a multivariate (MVA) approach, are rejected. Leptonic τ-decays are reconstructed as electrons or muons.
Events considered in the analysis are triggered with single- electron or single-muon triggers. The pT thresholds depend on theisolationrequirementanddata-takingperiod [72,73].Thelow- est trigger thresholds correspond to 25−27 GeV (electrons) and 21−27 GeV (muons).
4. Eventselectionandcategorization
Events selected in the τ channel contain exactly one elec- tronandonemuonofopposite-sign(OS)charges.Similarlyinthe
τhad channel, alepton anda τhad-vis ofOScharges arerequired, andevents withmorethan one baselinelepton are rejected. The selection criteriaaresummarizedinTable2fortheanalysiscate- goriesaswell asthecontrolregions(CRs),whicharedescribedin Section5.
Intheτ channel,1 and2denotetheleadingandsublead- ing lepton in pT,respectively. Events wheretheleading lepton is anelectron(muon)areusedinthesearchforH→eτμ (H→μτe).
A requirement on the dilepton invariant mass, equal to the in- variantmassoftheleptonandthevisible τ-decayproducts,mvis, reducesbackgroundswithtopquarks,andthecriterionappliedto the track-to-cluster pT ratioofthe electronreducesthe Z→μμ
background wherea muon deposits a large amountof energyin theelectromagneticcalorimeterandismisidentifiedasanelectron in the μτe channel. The contribution fromthe H→τ τ decayis reducedbytheasymmetricpT selectionofthetwoleptons.
In the τhad channel, the criterion based on the azimuthal separations of lepton–EmissT and τhad-vis–EmissT ,
i=,τhad-viscos φ (i,EmissT ), reduces the W + jets background whereas therequirementon| η(, τhad-vis)|reducesbackgrounds withmisidentified τhad-viscandidates.
Forbothchannelsofeachsearch,ab-vetorequirementreduces the single-top-quarkandt¯t backgrounds.Eventsare furthercate-
3 Thetransitionregioninηisexcluded,similarlytoelectrons.
The ATLAS Collaboration / Physics Letters B 800 (2020) 135069 3
Table 2
Baselineeventselectionandfurthercategorizationfortheτ andτhadchannels.Thesamecriteriaarealso usedforthecontrolregion(CR)definitionsintheτ channel(Section5),butonerequirementofthebaseline selectionisinvertedtoachieveorthogonaleventselection.ThereisnoCRintheτhadchannel.
Selection τ τhad
Baseline
exactly 1e and 1μ, OS exactly 1and 1τhad-vis, OS pT1>45 GeV pT>27.3 GeV pT2>15 GeV pτThad-vis>25 GeV,|ητhad-vis| <2.4 30 GeV<mvis<150 GeV
i=,τhad-vis
cos φ(i,EmissT ) >−0.35 peT(track)/peT(cluster) <1.2 (μτeonly) | η(,τhad-vis)| <2
b-veto (for jets with pT>25 GeV and|η| <2.4)
VBF
Baseline
≥2 jets, pjT1>40 GeV, pjT2>30 GeV
| η(j1,j2)| >3, m(j1,j2) >400 GeV
− pτThad-vis>45 GeV
Non-VBF
Baseline plus fail VBF categorization
mT(1,EmissT ) >50 GeV −
mT(2,EmissT ) <40 GeV −
| φ(2,EmissT )| <1.0 −
pτT/pT1>0.5 −
Top-quark CR inverted b-veto:
VBF and non-VBF ≥1 b-tagged jet (pT>25 GeV and|η| <2.4) Z→τ τCR inverted pT1requirement:
VBF and non-VBF 35 GeV<pT1<45 GeV
gorizedintoVBF(withafocusontheVBFproductionoftheHiggs boson)andnon-VBFcategories. TheVBFselectionisbasedonthe kinematics ofthe two jets withthe highest pT,where j1 andj2 denotetheleadingandsubleadingjetinpT,respectively.Thevari- ablesm(j1,j2) and η(j1,j2) stand forthe invariant mass and η
separationofthesetwojets.Thenon-VBFcategorycontainsevents failingtheVBFselection.Inthedileptonchannel,additionalselec- tioncriteriaareappliedtofurtherrejectbackgroundeventsinthis category.ThesecriteriaarealsolistedinTable2,wheremT stands forthe transversemass4 ofthetwo objectslisted inparentheses, and pτT represents the magnitude of the vector sum of pT2 and EmissT .Therequirementon pτT/pT1 reducesthebackgroundarising fromjetsmisidentifiedasleptons.TheVBFandnon-VBFcategories ineachoftheτ andτhad channelsgiveriseto foursignalre- gionsineachsearch.
TheanalysisexploitsBDTalgorithmstoenhancethesignalsep- arationfrom thebackground inthe individual searches,channels andcategories.Thecomponentsofthefour-momentaoftheanaly- sisobjectsaswellasderivedeventvariables(e.g.invariantmasses andangular separations) are the input variables ofthe BDT dis- criminant. Correlations between these input variables have been carefullychecked,highly correlatedvariables have beenremoved and the remaining ones are ranked according to their discrimi- nation power [74,75]. The list of variables isthen optimized, re- movingthelowest-rankedvariableswithmarginalcontributionto the sensitivity. The final list of variables is presented in Table 3 foreachchannelandcategory.TheinvariantmassoftheHiggsbo- sonreconstructedundertheH→ τ decayhypothesisexhibitsthe highestsignal-to-backgroundseparationpowerandithelpstodis- tinguishLFVsignalfromH→τ τ andH→W W backgrounds.For theτ channeltheinvariantmassisreconstructedwiththeMMC algorithm [76] and isdenoted bymMMC; forthe τhad channel it isreconstructedwiththe collinearapproximation [76] andis de- notedbymcoll.The analysisCRsare used tovalidate the levelof
4 ThetransversemassoftwoobjectsisdefinedasmT=
2pT 1pT 2(1−cos φ), wherepT iaretheindividualtransversemomentaand φistheanglebetweenthe twoobjectsintheazimuthalplane.
agreement between data and simulated distributions of the BDT scoreandinputvariables,aswellastheircorrelations.
5. Backgroundmodelling
Themostsignificantbackgroundsinthesearcharefromevents with Z→τ τ decaysorwith(singleorpair-produced)topquarks, especiallyintheτ channel,aswellasfromeventswithmisiden- tified objects, which are estimatedusing data-driven (d.d.) tech- niques.Therelativecontributionfrommisidentified objectstothe total background yield is 5–25% in the τ channel and 25–45%
in the τhad channel, depending on the search andthe analysis category. The shapes of distributions fromthe Z→τ τ andtop- quark (single-top-quark and tt)¯ processes are modelled by sim- ulation in both the τ and τhad decay channels. In the τ
channel,the relativecontributionsof Z→τ τ andtop-quarkpro- duction processes are20–35% and20–55%, respectively; thetop- quark background dominates in the VBF category. In the τhad
channel, the top-quark background fraction is 1–10%, while the Z →τ τ process contributes to 45–55% of the total background.
The individual contributionsare listed inTables 4and 5. Smaller backgroundcomponentsare alsomodelledby simulationandare grouped together: Z →μμ, diboson production, H→τ τ and H→W W .
GoodmodellingofthebackgroundisdemonstratedinFig.1for a selectionof importantBDT input variables.Detailsof theback- groundestimationtechniquesaregivenbelow.
5.1. τchannel
TwosetsofCRs,asdefinedinTable2,areusedtoconstrainthe normalizationof Z→τ τ andtop-quarkbackgroundcomponents.
These CRs inherit their definitions from the corresponding anal- ysiscategory but invertone requirementto ensure orthogonality with the nominal selection. The normalization factors are deter- mined during the statistical analysis by fitting the event yields in all signal andcontrol regions simultaneously. Foreach search, separate Z →τ τ normalizationfactors are usedfor theVBF and non-VBF categories. In the case of the top-quark background, in