Contents lists available atScienceDirect
Physics Letters B
www.elsevier.com/locate/physletb
Search for Higgs boson decays into a pair of light bosons in the bb μμ
final state in pp collision 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:
Received3July2018
Receivedinrevisedform12October2018 Accepted19October2018
Availableonline21December2018 Editor:W.-D.Schlatter
AsearchfordecaysoftheHiggsbosonintoapairofnewspin-zeroparticles,H→aa,wherethea-bosons decayintoab-quarkpairand amuon pair,ispresented.The searchuses36.1 fb−1 ofproton–proton collision data at√
s=13 TeV recorded by the ATLAS experiment atthe LHC in 2015and 2016. No significantdeviationfromtheStandardModelpredictionisobserved.Upperlimitsat95%confidencelevel areplacedonthebranchingratio(σH/σSM)×B(H→aa→bbμμ),rangingfrom1.2×10−4to8.4×10−4 inthea-boson massrangeof20–60 GeV.Model-independentlimits are setonthe visibleproduction cross-sectiontimesthebranchingratiotothebbμμfinalstatefornewphysics,σvis(X)×B(X→bbμμ), rangingfrom0.1 fbto0.73 fbformμμbetween18and62 GeV.
©2018TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.
1. Introduction
The discovery of the Standard Model (SM) Higgs boson [1,2]
hasopenedup newavenuestosearch forphysics beyondtheSM (BSM)withperspectivesto search for non-SMor “exotic”decays ofthe Higgsboson.Suchsearchescould provideuniqueaccessto hidden-sectorparticlesthataresingletsundertheSMgaugetrans- formations [3]. ExoticdecaysoftheHiggsboson arepredictedby manynew-physicsmodels [3,4],includingthosewithanextended Higgssector [5–9],darkmatter(DM)models [10–14],modelswith afirst-orderelectroweakphasetransition [15,16] andtheorieswith neutralnaturalness [17,18]. Thesemodelshavealsobeen usedto explainthe observationsofa γ-rayexcess fromthe galacticcen- tre (GC) by theFermi LargeArea Telescope [19,20]. For example, amodelforthe GC γ-rayexcess was proposed inwhich30 GeV DMparticles pair-annihilate dominantly through a CP-oddscalar mediator that subsequently decays into SM fermions [13]. If the mediatorissufficientlylighter thantheSMHiggsboson(H) then H decayintothemediatorpaircanbeobservedattheLHC.
ExistingmeasurementsconstraintheBSMor“exotic”branching ratio(B)ofthe125 GeV Higgsbosondecaystolessthanapproxi- mately34% at95% confidencelevel [21].Duetothenarrowwidth (∼4 MeV) ofthe Higgsboson, even asmall non-SMcouplingof O(10−2)canleadtoO(10%)branchingratiointoBSMstates.This potentiallylargeB(H→BSM states)motivatesdirectsearchesfor exoticHdecays.
E-mailaddress:atlas.publications@cern.ch.
TheanalysispresentedinthisLetterperformsthesearchinthe bbμμ final state. The a-boson can be eithera scalar ora pseu- doscalarunderparitytransformations,sincethedecaymodecon- sideredinthissearchisnotsensitivetothedifferenceincoupling.
Assumingthatthea-bosonmixeswiththeSMHiggsbosonandin- heritsitsYukawacouplingstofermions,thelargestbranchingratio isexpectedtobetotheheaviestfermionsaccessiblebykinematics (2ma<mH), where ma and mH are the a-boson and Higgs bo- sonmasses.Forma10 GeV thismeansthea-bosonwoulddecay preferentially into bb.However, in models withenhanced lepton couplings such as the Type-III 2HDM [22], the a→μμ branch- ing ratio can also be relatively large. Additionally, the sensitivity of a givenchannel doesnot depend only on the expectedsignal ratein a particular model, butalso on theefficiency fortrigger- ing andreconstructingeventsofinterest. Thepresenceofaclean dimuonresonanceprovidesadistinctivesignaturethatcanbeused for triggering and precision mass reconstruction, which helps to suppressbackground.
Searches forthe Higgsboson witha mass of125 GeV decay- ingintotwo spin-zeroparticles, H→aa,havebeenperformedin various final states in ATLAS and CMS [23–29]. The CMS search with√
s=8 TeV datainthebbμμfinalstateset95%CLlimitson (σH/σSM)×B(H→aa→bbμμ)between2×10−4and8×10−4in thea-bosonmassrangeof25–62.5 GeV [25].InType-III2HDM+S scenariowithtanβ=2 [4],wheretanβ denotestheratioofthe vacuum expectation values of the two Higgs fields, these limits translateintoupperlimitson(σH/σSM)×B(H→aa)rangingbe- tween13%and50%.Someofthemoststringentlimitsup todate forType-III2HDM+S withtanβ=2 come from theCMSsearch with √
s=13 TeV data in the bbτ τ final state, setting the up- https://doi.org/10.1016/j.physletb.2018.10.073
0370-2693/©2018TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP3.
perlimitson(σH/σSM)×B(H→aa)between4%and26%inthe a-bosonmassrangeof15–60 GeV [28].
2. Dataandsimulation
The search presented inthis Letter is basedon the 36.1 fb−1 datasetof proton–protoncollisions ata centre-of-mass energyof
√s=13 TeV recorded by the ATLAS experimentat theLHC dur- ing2015and2016. TheATLASexperiment [30] is amultipurpose particle detector witha forward–backward symmetric cylindrical geometry anda near 4π coverage in solid angle.1 It consists of an inner tracking detector surrounded by a thin superconduct- ingsolenoidprovidinga 2 Taxialmagneticfield, electromagnetic andhadronic calorimeters, anda muon spectrometer. Events are collected withsingle-muontriggers requiringthe muon pT tobe above 24 or 26 GeV, depending on the data-taking period. The trigger efficiencyforthe signal events withthe muon pT onthe triggerplateauisabout 80%.
Simulatedevents are used to modelthe signal andSM back- grounds processes. Higgs boson production through the gluon–
gluon fusion (ggF) and vector-boson fusion (VBF) processes was modelled at next-to-leading order (NLO) using Powheg-Box v2 [31–33] interfacedwithPythia 8.186 [34] using theAZNLOsetof tunedparameters [35] forthesimulationofthebbμμdecayofthe Higgs boson,as well as forparton showering andhadronisation.
The ggF Higgs boson production rate is normalised to the total cross-sectionpredictedby anext-to-next-to-next-to-leading-order QCDcalculationwithNLOelectroweakcorrectionsapplied [36–40].
The VBF production rate is normalised to an approximate next- to-next-to-leading-order(NNLO)QCDcross-sectionwithNLOelec- troweakcorrectionsapplied [41–44].Fivemasspointsweresimu- latedintherangema=20–60 GeV instepsof10 GeV forbothggF andVBFproduction.
Sherpa2.2.1 [45] withthe NNPDF3.0[46] setofpartondistri- bution functions(PDF) was used forthe generationof Drell–Yan, W+jets anddiboson(W W,W Z,Z Z)backgrounds.Cross-sections were calculated at NNLO QCD accuracy for Z(∗)/γ∗+jets and W +jets production [47] andat NLO including LOcontributions withtwoadditional partonsforthedibosonprocesses [45,48,49].
Thett¯andsingle-top-quarksamplesweregeneratedwithPowheg- Box v2 [32] using the CT10 PDF set [50] interfaced with Pythia v6.428 [51] and the Perugia 2012 set of tuned parameters [52]
forthepartonshower.Themassofthetopquark(mt) wassetto 172.5 GeV.The parameterhdamp inPowheg,usedtoregulate the high-pTradiation,was settomt forimprovedagreementbetween dataandsimulationinthehighpTregion [53].Thecross-sectionof tt¯wascalculatedatNNLOinQCDincludingresummationofnext- to-next-to-leadinglogarithmic(NNLL)softgluonterms [54,55].The cross-section forsingle-top-quarkproductionwas calculated with theprescriptionsinRefs. [56,57].Theproductionoft¯tpairsinas- sociationwith W/Z bosons (denoted by tt V¯ ) was modelledwith samplesgeneratedatLOusingMadGraph5_aMC@NLOv2.2.2 [58]
andshoweredwithPythiav8.186.Thesamplesarenormalisedto NLOcross-sections [59,60].
Additional pp collisions generated with Pythia v8.186 were overlaid to model the effects of additional interactions in the sameandneighbouringbunchcrossings(pile-up)forallsimulated
1 TheATLASCollaborationusesaright-handedcoordinatesystemwithitsorigin atthenominalinteractionpoint(IP)inthecentreofthedetectorandthe z-axis alongthebeampipe.Thex-axispointsfromtheIPtothecentreoftheLHCring, andthey-axispointsupwards.Cylindricalcoordinates(r,φ)areusedinthetrans- verseplane,φbeingtheazimuthalanglearoundthebeampipe.Thepseudorapidity isdefinedintermsofthepolarangleθasη= −ln tan(θ/2).Angulardistanceis measuredinunitsofR≡
(η)2+(φ)2.
events.The pile-upsimulation usedthe A2setoftuned parame- ters [61] andtheMSTW2008LOPDFset [62].Allthesampleswere processed through the full ATLAS detector simulation [63] based onGEANT4[64] andprocessedwiththesamereconstructionalgo- rithmasusedfordata.
3. Selectioncriteria
Interaction vertices from proton–proton collisions are recon- structedfromatleasttwotrackswithtransversemomentum(pT) larger than 0.4 GeV, andare required to be consistent with the beamspot envelope.The primary vertex (PV) is identified as the onewiththelargest
p2T ofassociatedtracks [65].
Muoncandidatesarereconstructedusingtheinformationfrom the innerdetector andthemuon spectrometer [66].Theyare re- quiredtosatisfy“medium”identificationcriteria [66],bematched to the PV and have pT>7 GeV and |η|<2.7. Additionally, the muons must satisfy the following criteria: theprojected longitu- dinal impact parameter |z0sinθ| must be lessthan 0.5 mm and the ratio of the transverse impact parameter d0 to its estimated uncertainty σd0, |d0/σd0|, must be less than 3. Finally, the se- lected muons must fulfil requirements on the scalar sum of pT of additional inner detector tracks andon the sum of the ET of calorimeter topological clusters [67] ina cone of size R=0.2 around the muon to ensure they satisfy “tight” isolation crite- ria [66].Theserequirementsselectsignalmuonswithanidentifi- cationefficiencyof∼94%andisolationefficiencyrangingbetween
∼91%forma=20 GeV and∼95%forma=60 GeV.
Jets are reconstructed using the anti-kt algorithm [68] imple- mented inthe FastJet package [69] with a radiusparameter R= 0.4 appliedtotopologicalclustersofenergydepositsincalorime- tercells.Jetsfrompile-uparesuppressedwiththeuseoftracking information asdetailedinRef. [70].All selectedjetsare required to have pT>20 GeV, |η|<2.5 and must pass quality require- ments defined to minimise the impact of detector effects, beam backgroundsandcosmicrays.
Jetsconsistentwiththehadronisationofab-quark(b-jets)are identified using a multivariate discriminant [71,72]. This analysis usesthe77%b-jetidentificationefficiencyworkingpointforwhich thepurityoftheb-taggedsampleisapproximately95%,whilethe probability ofmisidentifying ajetinitiatedby acharmquarkasa b-jetisapproximately16%,asdeterminedfromasampleofsimu- latedt¯t events.
Inordertorejectnon-promptmuonsfromthedecayofhadrons within a jet, an overlap removal algorithm is applied. If a jet is found within R=0.4 of themuon candidate,the overlapisre- solved in the following way: if there are more than two tracks with pT>500 MeV associatedwiththejet then themuon isre- movedfromtheevent,otherwisethemuonisretainedandthejet isremoved.
Themissingtransversemomentum(EmissT )usedintheanalysis iscalculatedasthemagnitudeofthenegativevectorsum(−→pmiss
T )
ofthetransversemomentaofallselectedandcalibratedobjectsin the event andthe additional “soft” term that takes into account tracksnotassociatedwithanyofthetheseobjects [73].The“soft”
term is calculatedfrom inner detector tracksmatched to the PV andincludedtoachieveabetterEmissT resolution.
Eventsarerequiredtohaveexactlytwob-taggedjetswithpT>
20 GeV andexactly tworeconstructed muonsofopposite charge, withtheleadingmuonhavingpT>27 GeV tobeinthemaximum- efficiency regime of thetrigger andthe subleading muon having pT>7 GeV. The dimuon invariant mass (mμμ) isrequired to be between16 GeV and64 GeV.Theupperboundonmμμ isdefined by theassumptionthatthe125 GeV Higgsbosondecaysintotwo
Fig. 1.The(a)mμμ,(b)mbbbeforetheKLfit,(c)mbbμμbeforeand(d)mKLbbμμaftertheKLfitforeventsafterthepreselectionstage,butremovingtheupperboundonmμμ. Thet¯tcontributionismodelledwiththesimulatedsamplenormalisedtothetheoreticalcross-section.TheDrell–Yancontributionistakenfromdatatemplates(describedin thetext)andnormalisedtothetotalyieldpredictedbytheDrell–Yansimulation.ThesignaldistributionsforallfivesimulatedmaarealsoshownassumingtheSMHiggs bosoncross-section(includingggF,VBFandV H production)andB(H→aa→bbμμ)=10%.Thebranchingratiointhisandallsubsequentfiguresischosensoastogive goodvisibilityontheplot.
on-shellparticlesofequalmasses,whilethelowerboundismoti- vatedbythekinematicsofthea-bosondecays.Forlowervaluesof ma,mostofthesignaljetsfallbelowthereconstructionthreshold andthe jetstend tooverlapgeometricallyinthedetectorso that thesensitivityoftheanalysistothe H→aasignaldecreases.This setofselectioncriteriaisreferredtoasthe“preselection”.
Signal events are characterised by the invariant mass of the twob-jets(mbb)beingequal,withinthedetectorresolution,tothe dimuoninvariantmassandthefour-objectmass(mbbμμ)beingap- proximately125 GeV.Oneside ofthe H→aadecay(a→μμ)is measuredwithapproximatelytentimesbetterresolutionthanthe othersideofthedecay(a→bb),asshowninFigs.1(a)and1(b).
Akinematic-likelihood(KL)fit [74] exploitingthesymmetryof H→aadecaysisperformedtotest thecompatibilityofan event withthe mbbmμμ hypothesis and improve the mbbμμ resolu- tioninsignal events.The KLfit findsthe energies oftheleading (Eˆb1)andsubleading(Eˆb2) b-jets thatmaximisethelikelihoodfor aneventwithmeasuredleadingandsubleadingb-jetenergies Eb1 and Eb2 andwithdimuon invariant massmμμ. Thelikelihood is definedasfollows,
L=W(Eˆb1,Eb1)·W(Eˆb2,Eb2)·FBW(mKLbb,mμμ),
where mKLbb is the dijet invariant mass computed from the b-jet four-momenta corresponding to Eˆb1 and Eˆb2, W is the transfer functionoftheb-jets,andFBWisaBreit–Wignerfunctioncentred
onmμμwithawidththatissmallcomparedtothembbresolution.
Thetransferfunction W(Eˆb1,Eb1)isadoubleGaussianprobability density function derived from simulated events as a function of jet pT and ηusingthedifferencebetweentrueandreconstructed energies.Thefitdeterminesamaximum-likelihoodvalueofL(de- notedby ln(Lmax)),whichquantifieshowwellaneventfitstothe constraints.Theb-jet momentadeterminedby thefitareused to recomputethefour-bodymassdenotedmbbKLμμ.AsseeninFig.1(d), theresolutionofthemKLbbμμ distributionforthesignalisimproved by up to a factor of two compared to the pre-fit mbbμμ shown inFig.1(c), whilethe backgroundshapewithin thembbμμ signal peak remains almost unchangedwiththe yields risingby ∼20%.
This allows the analysis to place tighter constraints on the dif- ference between the reconstructed invariant mass of the bbμμ
systemandmH,rejecting morebackground eventsandobtaining highersignalsignificance.
Twocriteriabased on the kinematiclikelihood fitare applied to selectsignal-like events andreject background eventsthat do not fit the mbb=mμμ constraint well: |mKLbbμμ−mH|<15 GeV andln(Lmax)>−8.Finally,the EmissT <60 GeV requirementrejects a largeportion oftt¯ pairswhere bothtop quarksdecaysemilep- tonically,while retainingmostofthe signal events.Adding these threerequirementsafterthepreselection stagedefinesthesignal- enhanced region (SR). A search for a localised excess above the expectedbackgroundisperformedinmultiplemμμ binsoftheSR