Search for exotic decays of a Higgs boson into undetectable particles and one or more photons

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Search for exotic decays of a Higgs boson into undetectable particles and one or more photons

.CMS Collaboration

^

CERN,Switzerland

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

Articlehistory:

Received1July2015

Receivedinrevisedform2December2015 Accepted7December2015

Availableonline11December2015 Editor:M.Doser

Keywords:

CMS Physics Higgs

A searchis presented for exotic decaysof aHiggsbosoninto undetectable particlesand one ortwo isolatedphotonsinppcollisionsatacenter-of-massenergyof8 TeV.Thedatacorrespondtoaninte- gratedluminosityofupto19.4 fb⁻¹collectedwiththeCMSdetectorattheLHC.Higgsbosonsproduced ingluon–gluonfusionand inassociationwith aZbosonare investigated,usingmodels inwhichthe Higgsbosondecaysintoagravitinoand aneutralinoorapairofneutralinos,followedbythedecayof theneutralinotoagravitinoandaphoton.Theselectedeventsareconsistentwiththebackground-only hypothesis,andlimitsare placedontheproductofcrosssectionsand branchingfractions.Assuminga standardmodel Higgsbosonproductioncrosssection, a95%confidencelevelupper limitissetonthe branchingfractionofa125 GeVHiggsbosondecayingintoundetectableparticlesand oneortwoiso- latedphotonsasafunctionoftheneutralinomass.Forthisclassofmodelsandneutralinomassesfrom 1to120 GeVanupperlimitintherangeof7to13%isobtained.Furtherresultsaregivenasafunction oftheneutralinolifetime,andalsoforarangeofHiggsbosonmasses.

©^{2015CERNforthebenefitoftheCMS}Collaboration.PublishedbyElsevierB.V.Thisisanopenaccess articleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

ThedetailedstudiesofthepropertiesoftheobservedHiggsbo- son[1–3]arekeycomponentsoftheLHCphysics program.Inthe standardmodel(SM) andfora givenmassoftheHiggsboson,all properties of the Higgs boson are predicted. Physics beyond the SM (BSM) mightlead to deviations fromthese predictions. Thus far,measurementsoftheHiggsbosons couplingstofermions and bosonsandofthetensorstructureoftheHiggsbosoninteraction withelectroweakgaugebosonsshownosignificantdeviations[4,5]

withrespecttoSMexpectations.

MeasurementsofHiggsboson couplings performedforvisible decay modesprovide constraints on partial decay widths of the Higgsboson toBSMparticles. Assumingthatthecouplingsofthe Higgsboson toW and Z bosonsare smallerthan the SM values, thisindirectmethodprovidesanupperlimitonthebranchingfrac- tionofthe125 GeVHiggsbosontoBSMparticlesof57%ata95%

confidencelevel (CL) [4,6].An explicit search forBSM Higgs bo- sondecayspresentsanalternativeopportunityforthediscoveryof BSMphysics.Theobservationofasizabledecaybranchingfraction oftheHiggsbosontoundetected(e.g.invisibleorlargelyinvisible)

 E-mailaddress:[email protected].

finalstateswouldbeaclearsignofBSMphysicsandcouldprovide awindowondarkmatter[7–10].

SeveralBSMmodelspredictHiggsbosondecaystoundetectable particlesandphotons. Incertainlow-scalesupersymmetry (SUSY) models, the Higgs bosons are allowed to decay into a gravitino (G) and a neutralino (

χ

₁⁰) or a pair of neutralinos [11,12]. The neutralinothen decaysintoaphoton andagravitino,thelightest supersymmetric particleanddark matter candidate.Fig. 1shows Feynman diagrams forsuch decaychains of theHiggs boson(H) produced bygluon–gluon fusion (ggH)orinassociation witha Z bosondecayingtochargedleptons(ZH).

Asthegravitinointhesemodelshasanegligiblemass[11,12], the remaining parameteris theneutralino mass.If its massis in therangem_H/2<m_χ₁⁰<m_H,withm_H=125 GeV themassofthe observedHiggsboson,thebranchingfractionB(^H→ 

χ

₁⁰G→

γ

GG) can be large. For mχ_1⁰<mH/2, the decay H→ 

χ

₁⁰

χ

₁⁰→

γ γ

^GG is expected to dominate. The samediscussion can be applied to heavyneutralHiggsbosonswithmasseslargerthan125 GeV.The lifetimeoftheneutralinocanbefiniteinsomeclassesofBSMsce- narios,leadingtotheproductionofoneormorephotonsdisplaced fromtheprimaryinteraction.

In the SM, the signature equivalent to the signal arises when theHiggsbosondecaysasH→^Z

γ

_→

ν νγ

¯ withabranchingfrac- http://dx.doi.org/10.1016/j.physletb.2015.12.017

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

Fig. 1. Feynman diagrams for the H→undetectable+γfinal state produced via ggH (left) and ZH (right).

tionof3×^{10−4.ThedecayH}→^Z

γ

hasbeenstudiedinZ→^e+e− and Z→

μ

⁺

μ

⁻ final states.Upper limits on the product of the crosssectionandbranchingfractionofaboutafactoroftenlarger thantheSMexpectationhavebeensetatthe95% CL[13,14].With theavailabledataset thesearchpresentedisnot sensitivetothis decay,butitissensitivetoenhancements intheHiggsbosonde- cayrates toundetectable particlesandphotonsarising fromBSM physics.

Variousbackgroundprocessesleadtothesignalsignaturesand areestimatedfromsimulationorfromcontrolsamplesindata.The dominantbackgroundprocessesare from

γ

+jets eventsanddi- bosoneventsintheggH andZH search,respectively.Detailsofthe backgroundestimationtechniquesarediscussedinSection 5.The strengthofthe ZH channelanalysisisan almostbackground-free selection leading to a larger sensitivity in the model-dependent interpretation. While both the ggH andthe ZH channels provide sensitivitytoBSMHiggsbosonsignatures,theggH channelallows amodel-independentinterpretationoftheresults.

Thisanalysispresents a firstsearch for decaysofa scalar bo- sontoundetectableparticlesandoneortwoisolatedphotons.The scalar boson is produced in ggH or in ZH. The data used corre- spondto an integratedluminosity ofup to 19.4±⁰.5 fb⁻¹ ata center-of-massenergy of √

s=^{8 TeV in 2012 collected withthe} CMSdetectorattheCERNLHC.

Theresultsofthesearcharepresentedintermsofthelow-scale SUSYbreakingmodelform_H=125 GeV andm_χ₁⁰ between1 GeV and120 GeV,andform_H between125 GeV and 400 GeV forthe examplecasewheremχ_1⁰=^mH−^{30 GeV.Theeffectofafinite}

χ

₁⁰ lifetime(

τ

_χ0

1)isstudiedfortheexamplecasewheremH=^{125 GeV} andmχ_1⁰=^{95 GeV.}

2. TheCMSexperiment

TheCMSdetector,definitionsofangularandspatialcoordinates, and its performance can be found in Ref. [15]. The central fea- tureofthe CMSapparatusisa superconducting solenoid, of6 m internal diameter, providing a magnetic field of 3.8 T. The field volumecontainsasiliconpixelandstriptracker,acrystalelectro- magnetic calorimeter(ECAL), anda brass and scintillator hadron calorimeter.Muonsare measured in gas-ionizationdetectors em- beddedinthesteelflux-returnyokeofthemagnet. Thefirstlevel oftheCMStriggersystem,composedofspecializedhardwarepro- cessors, is designedto selectthe mostinteresting events within 3 μs,usinginformationfromthecalorimetersandmuondetectors.

Ahigh-level triggerprocessorfarmisusedtoreducetheratetoa fewhundredeventspersecondbeforedatastorage.

Aparticle-flowalgorithm[16,17]isusedto reconstructall ob- servableparticlesintheevent.Thealgorithmcombinesallsubde- tector informationto reconstructindividual particlesand identify

them ascharged hadrons, neutral hadrons, photons, andleptons.

The missingtransverse energyvector E^miss_T is definedastheneg- ative vector sumof the transverse momenta ofall reconstructed particles(chargedorneutral)intheevent,withE^miss_T = |^EmissT |^.Jets are reconstructedusingtheanti-kT clustering algorithm[18]with a distance parameter of R=⁰.5, as implemented in the FastJet package [19,20]. A multivariate selection is applied to separate jetsfromthe primary interactionandthose reconstructed dueto energydeposits associatedwithpileup interactions [21].The dis- crimination is based on thedifferences inthe jet shapes, on the relative multiplicity of charged and neutral components, and on thedifferentfractionoftransversemomentumwhichiscarriedby the hardest components. Photon identification requirements and other procedures used in selecting events can be found in Sec- tion4.

3. Dataandsimulationevents

In the search for Higgs bosons produced in ggH, the trigger system requiresthe presenceof onehigh transverseenergy (Eγ

T) photon candidateandsignificant E^miss_T .Thepresence ofa photon candidate with Eγ

T >30 GeV is required within the ECAL barrel region (|

η

^γ_{| <}1.44). At thetrigger level E^miss_T is calculatedfrom calorimeter information, and is not corrected for muons. A se- lection requirement of E^miss_T > 25 GeV is applied. The efficiency of the trigger ismonitored and measured withtwo control trig- gers for the photon andthe E^miss_T trigger requirement. The data recordedwiththistriggercorrespond toan integratedluminosity of7.4 fb⁻¹ andwerepartoftheCMS“dataparking”programim- plemented forthe last partof the data takingat √

s=^{8 TeV in} 2012. Inthat program,CMSrecordedadditionaldatawithrelaxed triggerrequirements,planningforadelayedofflinereconstruction in2013afterthecompletionoftheLHCRun 1.

For the search for Higgs bosons produced in ZH, collision eventswere collected usingsingle-electronandsingle-muontrig- gers whichrequirethepresence ofan isolated leptonwith pT in excessof27 GeV and24 GeV,respectively.Alsoadileptontrigger wasused,requiringtwoleptonswithpTthresholdsof17 GeV and 8 GeV.Theluminosityintegratedwiththesetriggersat√

s=^{8 TeV} is19.4 fb⁻¹.

Several MonteCarlo (MC) eventgenerators are used to simu- late signal and backgroundprocesses. Thesimulated samplesare usedtooptimizetheeventselection,evaluateselectionefficiencies andsystematicuncertainties, andcompute expectedeventyields.

In all casesthe MC samplesare reweighted to matchthe trigger efficiencymeasuredindata.

The V

γ

, WZ, ZZ, VVV (where V represents W or Z bosons), Drell–Yan(DY) productionof qq¯→^Z/

γ

^∗,andqq¯→^{W+W− pro-} cessesare generatedwiththe MadGraph 5.1eventgenerator[22]

atleading-order(LO), thegg→^{W+W− processisgeneratedwith} theLO eventgenerator gg2ww 3.1 [23], andthe t¯t and tW pro- cesses are generated with powheg 1.0 at next-to-leading-order (NLO).Thesignal samplesarealsoproducedwith MadGraph.The crosssectionsatNLO orhigher ordersif availableare used fora given process to renormalize the MC event generators. All pro- cesses are interfaced to the pythia 6.4 generator [24] for parton showerandhadronization.

The CTEQ6L set of parton distribution functions(PDF) [25] is used forLO generators, while the CT10 [26] PDF set is used for NLO generators. Forall processes,the detector responseis simu- latedwithadetaileddescriptionoftheCMSdetector,basedonthe Geant4 package [27]. Additional pp interactions overlapping the eventofinterestindata,denotedaspileupevents,areaccounted forby simulating pp interactions with the pythia generator and addingthemtoeachMCsample.TheMCsamplesaretunedtore- producethe distribution inthe numberof pileupevents indata.

Theaveragenumberofpileupeventsisabout26forthecollected dataused in the ggH channel, and is about21 for the collected datausedintheZH channel.

4. Eventselection

Twostrategies are followedto isolate the Higgsboson events producedby ggH andby ZH fromthe backgroundprocesses.The signalcrosssectionsareseveralordersofmagnitudesmallerthan themajorreduciblebackgroundprocesses,whosecontributionsare greatlyreducedusingtheeventselectionsdescribedinthefollow- ingsub-sections.

4.1.EventselectionintheggH channel

IntheggH channel, eachselectedeventisrequiredto haveat leastonephotoncandidatewithEγ

T >45 GeV and|

η

^γ| <¹.44 us- ing a cut-based selection [28,29].To reduce the SM backgrounds arisingfromtheleptonic decaysofW andZ bosons,aleptonveto isapplied.Eventsare rejectedifthey haveone ormoreelectrons fulfillinga looseidentificationrequirement[30] and p^e_T>10 GeV and|

η

^e| <².5,excludingthetransitionregionof1.44<|

η

^e| ≤¹.57 since the reconstruction of an electron object in this region is not optimal. Similarly, events containing muon candidates with pμ

T >10 GeV,|

η

^μ| <².1,andwellseparatedfromthephotoncan- didaterequiringR(

γ

,

μ

)=

(

η

)²+ (φ)²>0.3 (wheretheφ isazimuthalangleinradians),arerejected. Inadditiontothese- lectionrequirementsdescribed above, the E^miss_T is requiredto be greater than 40 GeV. Thislevel of selection is referred to asthe preselection.Additionalselectioncriteriaareappliedtosearchfor new physics in either a quasi model-independent way or opti- mizedforaSUSYbenchmarkmodel.Inthischanneljetscanarise frominitial-stateradiation.Forboth search strategiesjetsare re- quiredto have p_T^j>30 GeV and |

η

^j| <².4. These jetsmust not overlapwiththephotoncandidatebelowR(

γ

,jet)<0.5.

In the model-independent analysis, events with two or more jetsare rejected.Foreventswithonejet theazimuthal anglebe- tweenthephotonandthejet(φ (

γ

,jet))isrequiredtobesmaller than 2.5.Thisselectionrequirementrejects thedominant

γ

₊jet background,wherethephotonandthejettendtobeback-to-back inthetransverseplane.

Inthemodel-dependentanalysisdevelopedforSUSYscenarios, no requirement is applied on jet multiplicity. In order to mini- mizethecontributionfromprocessessuchas

γ

+^{jets andmultijet} events,twomethodsareusedforidentifyingeventswithmismea- sured E^miss_T .The E^miss_T significance method[31] takesaccount of reconstructedobjects foreacheventandtheir known resolutions tocompute an event-by-eventestimate ofthelikelihoodthat the

observed E^miss_T isconsistentwithzero.Inaddition,aminimization method[29]constructsa

χ

²functionoftheform

χ

²

₌ ^

i=^objects

 (

p_T^reco

)

i

− (

^pT

)

i

( σ

_pT

)

_i



2

+

 

E^miss_x

σ

_Emiss x



2

+

 

E^miss_y

σ

_Emiss y



2

,

(1) where (pTreco)i are thescalar transverse momenta of the recon- structed objects, such as jets and photons that pass the above- mentioned identification criteria,the (

σ

pT)i are theexpectedres- olutionsineachobject,the

σ

_Emiss

x,y are theresolutions ofthe E^miss_T projection along the x-axisandthe y-axis, andthe(^pT)_i are the free parameters allowed to vary in the minimization of the

χ

²

function.TheE^miss_x,y termsarefunctionsofthefreeparameterspx,y,



E^miss_x,y

=

^Emiss_x,y ^,reco

+ 

i=objects

(

p^reco_x,y

)

i

− (

^px,y

)

i

.

(2)

In events with no genuine E^miss_T , the mismeasured quantities arere-distributedbackintotheparticlemomenta tominimizethe

χ

² value.EventsarerejectediftheminimizedE^miss_T (E^miss_T )isless than45 GeV andthechi-squareprobabilityislargerthan10⁻³.

Tofurther suppressmultijet backgrounds,events are not con- sideredifthescalarsumofthetransversemomentaoftheidenti- fiedjetsintheevent(H_T) isgreaterthan100 GeV. Anadditional requirementisapplied ontheangle(

α

)betweenthebeamdirec- tionandthemajoraxisofthesupercluster[28] inorderto reject non-promptphotonsthathaveshowerselongatedalongthebeam line.

Finally,thetransversemass,

m^γE

miss T

T

≡

2E^γ_TE^miss_T

[

¹

−

^cos

φ ( γ ,

E^miss_T

)],

formedbythephotoncandidate,E^miss_T ,andtheiropeningangle,is requiredtobegreaterthan100 GeV.Photonsfromthecontinuum Z

γ

backgroundhaveaharderspectrumthanthephotonsresulting fromtheHiggsdecayintheSUSY benchmarkmodelsconsidered.

Tofurther reduce the continuum Z

γ

background andformodels withm_H=^{125 GeV a cut of Eγ}_T <60 GeV isapplied. Forhigher massesthe cut is optimizeddepending oneach mass hypothesis goingfrom60 GeV upto200 GeV formH=^{400 GeV.}

Thelistofselectioncriteriausedinthemodel-independent and theSUSYbenchmarkmodelanalysesaregiveninTable 1,together withthecumulativeefficienciesrelativetothepreselectionforsig- nalandbackgroundprocesses.

4.2. EventselectionintheZH channel

TheleptonicdecaysoftheZ boson,consistingoftwooppositely charged same-flavor high-pT isolated leptons (e⁺e⁻,

μ

⁺

μ

⁻), are usedtotagtheHiggsbosoncandidateevents.Largemissingtrans- verseenergyfromtheundetectableparticles,atleastoneisolated high-ET photon,andlittleormoderatejetactivityarerequiredto selectthesignalevents.

Thedetailsoftheleptoncandidateselectionandmissingtrans- verse energy reconstruction are given in Ref. [32]. In addition, photon requirements based on a multivariate selection discussed in Refs. [28,33]have beenused. Thekinematic selection requires twoleptonswithpT>20 GeV andonephotonwithEγ

T >20 GeV.

Furthermore,thedileptonmassmustbecompatiblewiththatofa Z bosonwithin15 GeV ofthepolemass.

ToreducethebackgroundfromWZ events,eventsareremoved if an additional loosely identified lepton is reconstructed with p_T>10 GeV.Torejectmostofthetop-quarkbackground,anevent

Table 1

SummaryofggH selectionforboththequasimodel-independentanalysisandtheanalysiswiththeSUSYbenchmarkmodelwiththecumulativeefficienciesoftheselection requirementsrelativetothepreselectionforZγ→ννγ¯ ,γ+jet andforasignalinaSUSYbenchmarkmodelwithggH productionofaHiggsbosonwithmass125 GeV decayingintoaneutralinoofmass120 GeV andaphoton.

Selection requirements Model-independent SUSY benchmark model

Zγ→ννγ¯ γ+jet Zγ→ννγ¯ γ+jet m_χ0

1=120 GeV

Number of jets<2 0.909 0.769 – – –

φ(γ,jet) <2.5 radians 0.834 0.262 – – –

Transverse mass>100 GeV – – 0.867 0.292 0.829

HT<100 GeV – – 0.785 0.188 0.804

E^miss_T >45 GeV – – 0.761 0.071 0.743

Prob(χ²) <10⁻³ – – 0.626 0.033 0.467

E_T^misssignificance>20 – – 0.440 0.001 0.195

α>1.2 – – 0.390 0.001 0.165

E_T^γ<60 GeV – – 0.074 0.0002 0.106

Table 2

SummaryofZH selection.

Variable Selection

Leptons 2 leptons, pT>20 GeV

Photons 1 photon, E^γ_T>20 GeV

|^m−^mZ| <15 GeV

Anti b-tagging applied

Jet veto 0 jets with p_T^j>30 GeV

φ_,Emiss

T +E^γ_T >2.7 radians

|^pET^{missT +EγT}−^pT|/^pT <0.50

φ <2.25 radians

p_T >60 GeV

E^miss_T >60 GeV

isrejected ifit passesthe b-taggingselection (anti b-tagging) or if there is a selected jet with p_T larger than 30 GeV (jet veto).

Theb-taggingselectionisbasedonthepresenceofamuoninthe eventfromthesemileptonicdecayofabottom-quark, andonthe impactparametersoftheconstituenttracksinjetscontainingde- caysofbottom-quarks[34].Thesetofb-taggingvetocriteriaretain about95%ofthelight-quarkjets,whilerejectingabout70%ofthe b-jets.

The signal topology is characterized by a Z() system with large transverse momentum balanced inthe transverse plane by a E^miss_T + ^EγT systemfromthe Higgsboson decay.Toreject back- groundfromZ

γ

andZ+jets eventswithmisreconstructed E^miss_T the azimuthal angle φ_,Emiss

T +^EγT is required to be greater than 2.7 radians, the variable |^p_T^{EmissT +EγT} −^p_T|/^p_T ^{is required to be} smallerthan0.5,andtheazimuthalanglebetweenthetwoleptons

φ isrequiredtobesmallerthan2.25radians.Finally,p_T isre- quiredtobelargerthan60 GeV,andE^miss_T isrequiredtobelarger than60 GeV.Asummaryoftheselectionfortheanalysisisshown inTable 2.

The signal-to-background fraction depends on the |

η

^γ_|, the pseudorapidityofthephoton,withgreaterdiscriminationatlower values.Toexploit thiseffectandimprovesensitivity,theselected eventsare subdividedaccordingto whetherthephoton is recon- structed in the barrel or endcap regions, as explained in Sec- tion7.2.

5. Backgroundestimation

Thebackgroundestimation techniquesandthecomposition of allbackgroundsinthesearchwiththeggH andZH signaturesare discussed below. The yield for the irreducible background from H→^Z

γ

_→

ννγ

isnegligibleandisthereforeignoredintheanal- ysis.

5.1. BackgroundestimationintheggH channel

The dominantbackgroundforthe

γ

+^EmissT signalinthe ggH channel is the process

γ

+^{jet. Other SM} backgrounds include Z

γ

→

ν νγ

¯ ,W

γ

,W→^e

ν

,W→

μν

,W→

τ ν

,multijet,anddipho- tonevents.Backgroundeventsthatdonotarisefromppcollisions arealsoconsideredintheanalysis.Thesebackgroundscanbecat- egorizedbroadlyintothreecategories,asdescribedbelow.

5.1.1. Backgroundestimatesfromsimulation

The

γ

+jet process surviving the various E_T^miss selection re- quirements is one of the most significant backgrounds in this analysis due to the presence of an isolated photon andits large productioncrosssection.TheMCnormalizationofthisbackground is corrected using control samples indata for two eventclasses, events without jets and those with one or more jets. The con- trol samples in data are obtained using events collected with a prescaledsingle-photontriggerandwiththeE^miss_T requirementre- versed to ensure orthogonality to the signal phase space, where the contamination from other processes is minimal. Multiplica- tivecorrectionfactors(C )areobtainedafternormalizingtheevent yield in the simulation to match the data in the control region, separately foreventswithno jets(C=¹.7)andoneormorejets (C=¹.1).Thesecorrectionfactorsare usedtonormalizethesim- ulated

γ

+^{jet event yieldin thesignal region.An} uncertaintyof 16% is obtainedfor these correction factors based on the differ- ence between the corrected anduncorrected simulation and the relative fractionof nojet events (about10% ofthe eventsin the control region)andone ormorejet events.The backgroundpro- cessesZ

γ

_→ 

γ

andW→

μν

contributeonlyasmallfractionof thetotalbackgroundprediction,duetotheleptonvetoappliedat thepreselectionstage,andaremodeledusingsimulatedsamples.

To take intoaccount differences betweendata andsimulation due toimperfect MC modeling,various scale factors (SF) are ap- plied tocorrectthe estimatesfromsimulation.These SFsarede- fined by the ratio of the efficiency in data to the efficiency in simulationforagivenselection.The SFforphotonreconstruction andidentificationisestimatedfromZ→^{e+e− decays[35]andis} consistentwithunity.

5.1.2. Backgroundestimatesfromdata

Thecontaminationfromjetsmisidentifiedasphotons(jet→

γ

) isestimatedinadatacontrolsampleenrichedwithmultijetevents definedby E^miss_T <40 GeV.Thissampleisusedtomeasurethera- tioofthenumberofcandidatesthatpassthephotonidentification criteria to those failing the isolation requirements. The numera- torofthisratioisfurthercorrectedforthephotoncontamination duetodirectphotonproductionusingan isolationsideband. The

Table 3

SummaryofallrelativesystematicuncertaintiesinpercentforthesignalandbackgroundestimatesfortheHiggsmodel(model-independentinparenthesis)selectioninthe ggH analysis.

Source Signal Jet→γ Electron→γ γ+jet Zννγ Wγ

PDF 10 (0) – – – 4 (4) 4 (4)

Integrated luminosity 2.6 (2.6) – – 2.6 (2.6) 2.6 (2.6) 2.6 (2.6)

Photon efficiency 3 (3) – – 3 (3) 3 (3) 3 (3)

Photon energy scale±^{1% 4 (0.5) – – 4 (0.5) 4 (0.5) 4 (0.5)}

E^miss_T energy scale 4 (2) – – 4 (2) 4 (2) 4 (2)

Jet energy scale 3 (2) – – 5 (5) 3 (2) 3 (2)

Pileup 1 (1) – – 1 (1) 1 (1) 1 (1)

Zννγ normalization – – – – 3 (3) –

γ+jet normalization – – – 16 (16) – –

Wγnormalization – – – – – 3 (3)

Jet→γ – 35 (35) – – – –

Electron→γ – – 6 (6) – – –

corrected ratiois applied to data eventswhich pass the denom- inator selection and all other event requirements in the signal region.

Thesystematicuncertaintyofthismethodisdominatedbythe choice of the isolation sideband, and is estimated to be 35% by changing the isolation criteria in the sideband region definition.

The other sources of systematic uncertainty are determined by changingthe E_T^miss selectionforthe controlregion,andtheloose identificationrequirementsonthephotons,allofwhicharefound tobecomparativelysmall.

Events with single electrons misidentified as photons (electron→

γ

)areanothermajorsourceofbackground.Thisback- groundisestimatedwithatag-and-probemethodusingZ→^e+e− events[36].The efficiencytoidentify electrons(

_γe) isestimated in the Z boson peak mass window of 60–120 GeV. The ineffi- ciency(1−

γe)isfoundtobe2.31±⁰.03%.Theratio(1−

γe)/

γe, whichrepresentstheelectron misidentificationrate,is appliedto asample wherecandidates are requiredtohavehitsin thepixel detector,andisused toestimate the contamination inthe signal region.Themisidentificationrateisfoundtobedependentonthe numberofverticesreconstructedintheeventandthe numberof tracksassociatedwith theselected primaryvertex. Thedifference inthefinalyieldstakingthisdependenceintoaccountorneglect- ingit,usingtheinclusivemeasurementof

γe,islessthan5%.

5.1.3. Non-collisionbackgroundestimatesfromdata

The search is susceptible to contamination from non-collision backgrounds, which arise from cosmic ray interactions, spurious signalsin the ECAL,and accelerator-induced secondary particles.

The distribution of arrival-times of photons from these back- groundsisdifferentfrom thatofpromptphotonsproducedinhard scattering.Toquantifythecontaminationfromthesebackgrounds a fit isperformed to the candidate-time distribution usingback- grounddistributionsfromthedata[29].Thecontaminationdueto out-of-timebackgroundcontributionsisfoundtobelessthanone percentof the total backgroundand is thereforenot included in thefinaleventyield.

5.2.BackgroundestimationinZH channel

ProcessesthatcontributesignificantlytotheSMexpectationin theZH channelarelistedbelow.

5.2.1. Non-resonantdileptonbackgrounds

The contributions from W⁺W⁻, top-quark, W +^{jets, and} Z/

γ

^∗→

τ

⁺

τ

⁻ processes are estimated by exploiting the lepton flavor symmetry in the final states of these processes [37]. The branching fraction to the e^±

μ

^∓ final state is twice that of the e⁺e⁻ or

μ

⁺

μ

⁻ final states. Therefore, the e^±

μ

^∓ control region

isusedtoextrapolate thesebackgroundstothee⁺e⁻ and

μ

⁺

μ

⁻

channels. The methodconsiders differencesbetweentheelectron and muon identification efficiencies. The data driven estimates agreewellwiththenumberofbackgroundeventsexpectedwhen applying the same method to simulation. The small difference between the prediction and the obtained value using simulated eventsistakenasasystematicuncertainty.

The limitednumber ofsimulatedevents isalso considered as partofthe systematicuncertainty.Insummary,the totalsystem- aticuncertaintyisabout75%.Onlytwoeventswereselectedinthe datacontrolregion.

5.2.2. Resonantbackgroundwiththreeleptonsinthefinalstate The WZ→ 

ν

 process dominatesthe resonant backgrounds withthreeleptonsinthefinalstate.Theelectron→

γ

misidentifi- cationrateismeasuredinZ→^{e+e− eventsbycomparingthera-} tiosofelectron–electronversuselectron–photonpairsindataand insimulation,asdescribedinSection5.1.2.Theaveragemisidenti- ficationrateis1–2%withthelargervaluesathigher|

η

|^{γ .}

5.2.3. Resonantbackgroundwithtwoleptonsinthefinalstate

TheWZ→ 

ν

processwithfailuretoidentifytheleptonfrom W bosondecaysandtheZZ→²2

ν

processdominatethesetypes ofevents.Thejet→

γ

misidentificationrateismeasuredinasam- plecontainingamuonandaphoton.Thissampleisexpectedtobe dominatedbyjetsmisidentifiedasphotons,withsomecontamina- tion fromW/Z

γ

events,whichare subtracted inthe studyusing the simulated prediction. The misidentification rate is similar to theobtainedvaluesintheggH channel.

5.2.4. Resonantbackgroundwithnogenuinemissingtransverseenergy ThebackgroundfromZ

γ

orZ+jets eventsispredictedbythe simulation tobe about15% ofthe totalbackground.Severaldata regionsarestudiedtoverifythatthebackgroundisestimatedcor- rectly. A good agreement between data and simulation is found in all cases. A 50% uncertainty, the statisticaluncertainty of the controlregion,istakenforthesebackgroundsestimatedfromsim- ulation.

6. Summaryofsystematicuncertainties

Systematicuncertaintiesinthebackgroundestimatesfromcon- trolsamplesindataaredescribedinSection5.Asummaryofthe systematic uncertainties considered in each channel are listed in Tables 3 and 4.

Acommonsourceofsystematicuncertaintyisassociatedwith the measurement of the integrated luminosity, determined to 2.6% [38]. The uncertainties in the normalization of signal and

Table 4

SummaryofrelativesystematicuncertaintiesinpercentforthesignalandbackgroundestimatesintheZH analysis.

Source ZH Zγ or Z+^{jets WZ ZZ WW}+^top-quark

Integrated luminosity 2.6 – 2.6 2.6 –

Lepton efficiency 3.6 – 3.6 3.6 –

Photon efficiency 3.0 – – – –

Momentum resolution 0.5 – 1 1 –

E_T^missenergy scale 0.5 – 0.6 0.1 –

Jet energy scale 2 – 4 4 –

b-tagging 0.7 – 0.7 0.7 –

Underlying event 3 – – – –

PDF 7.1 – 6.3 7.7 –

Renorm. and factor. scales 7.0 – 10.7 6.5 –

Z/γ^∗→ ^+−normalization – 50 – – –

Non-resonant dilepton bkg. norm. – – – – 70

Jet→γ – – 30 30 –

Electron→γ – – 10 10 –

Amount of simulated events 3.5 60 10 30 40

simulation-basedbackgrounds areobtainedby varying therenor- malization and factorization scales, and the parton distribution functions[26,39–43].

Because the model-independent and model-specific selections differ significantly in the ggH channel, the systematic uncertain- tiesareevaluatedseparatelyforeachselection.Thephotonenergy scale uncertainty [28] of about 1% affects the signal and back- groundpredictionsby 4% forthe model-specific selectionandby 0.5%forthemodel-independentselection.Similarly,thejetenergy scaleuncertaintyaffectsthesignalandbackgroundpredictionsby 2–5%depending on theprocess andselection.Afterchangingthe photonorjetenergyscales,the E^miss_T isalsorecomputed.Inaddi- tion,thesystematicuncertaintyassociatedwiththejetenergyres- olutionandunclusteredenergy scaleare propagated tothe E^miss_T computation,andaffectthesignalandbackgroundpredictionsby 2–4%. As described inthe previous section, a 16% uncertaintyis appliedto the

γ

+jet normalization duetothedifference inthe jetmultiplicitydistributionbetweenthedataandbackgroundpre- dictionsinthe

γ

₊jet control region.Theuncertaintyduetothe pileup modeling is found to be 1%, and is estimated by shifting thecentralvalueofthetotalinelasticcrosssectionwithin itsun- certainty.

In the ZH channel, lepton-reconstruction and identification scalefactorsaremeasuredusingacontrolsampleofZ/

γ

^∗→ ^+− events in the Z peak region [36]. The associated uncertainty is about2%perlepton.Thephotonidentificationuncertaintyistaken to be 3% [33]. Theeffect ofuncertainties in jet energy scale and E^miss_T on the analysis is also considered. The uncertainty in the b-taggingefficiency is estimatedto be about 0.7% comparingin- clusive Z/

γ

^∗_→ ^+− samples in data and simulation. The total uncertainty in the background estimates in the signal region is 36%,whichisdominatedbythestatisticaluncertaintiesinthedata controlsamplesfromwhichtheyarederived.

Theimpactofthesystematicuncertainties intheggH channel is relatively important: the sensitivity isincreased by about 50%

if all the systematic uncertainties are removed, where the nor- malizationuncertainties onthe

γ

+^{jet and jet}→

γ

background processes dominate. The ZH channel is limited by the statistical uncertainty,andthe effectofthe systematicuncertainties reduce thesensitivitybylessthan10%.

Correlationsbetweensystematicuncertaintiesinthetwochan- nels are taken into account. In particular, the main sources of correlated systematicuncertainties are those inthe experimental measurements suchasthe integratedluminosity,photon identifi- cation,thejetenergyscale,andmissingtransverseenergyresolu- tion. Allother systematicuncertainties are uncorrelated between themgiventhedifferentsignalandbackgroundprocesses.

Table 5

Observedyieldsandbackgroundestimatesat8 TeV inthe ggH channelafterthemodel-independentselection.Statisti- calandsystematicuncertaintiesareshown.

Process Event yields

γ+jets (313±50)×10³

jet→γ (910±320)×10²

e→γ 10 350±620

W(→ ν)+γ 2239±111

Z(→νν¯)+γ 2050±¹⁰²

Other 1809±⁹¹

Total background (420±⁸²)×¹⁰³

Data 442×¹⁰³

7. Results

The results fromthe two searches andtheir combination are reported in this section. In the absence of deviations from the standard model predictions, the modified frequentist method, CL_s[44–46],isusedtodefinetheexclusionlimits.

7.1. Model-independentresultsintheggH channel

Because ofthevarietyofpossibleBSMsignalsthatcould con- tributetothisfinalstate,theresultsarepresentedforasignalwith themodel-independentselectiondescribed inSection4.Thetotal numberofobservedeventsandtheestimatedSMbackgroundsare summarized inTable 5,andfound to be compatiblewithin their uncertainties. Fig. 2showsthemγE^miss_T

T and E^miss_T distributionsfor themodel-independentselection.

Fig. 3 shows the observed and expected model-independent 95% CL upperlimitsfortheggH analysis ontheproduct ofcross section, acceptance, and efficiency for mγE^miss_T

T >100 GeV, as a functionofE^miss_T threshold.

7.2. Model-specificresultsintheggH channel

Imposingthemodel-specificselectiondescribedinSection4for the ggH channel, 1296 events are selected in data with a total estimated background of 1232±^{188. The yields for this selec-} tion are shownin Table 6andestimated forHiggsboson decays (H→ ^G

χ

₁⁰,

χ

₁⁰_{→ }G

γ

) assuming the ggH production rate forSM Higgsbosons anda 100%branching fractionforthisdecay. Fig. 4 showsthe transverseenergydistribution ofphotonsfordata,the background estimates, and signal after the model-dependent se- lection, except the upper selection on the photon, for the ggH channel.

Fig. 2. Them^γ_T^EmissT andE^miss_T distributionsfordata,backgroundestimates,andsig- nalafterthemodel-independentselectionfortheggH channel.Thebottompanels ineachplotshowtheratioof(data−^background)/background andthegrayband includesboththe statisticalandsystematicuncertaintiesinthebackgroundpre- diction.ThesignalisshownformH=125 GeV andm_χ0

1 =120 GeV anda100%

branchingfraction.

7.3.ResultsintheZH channelandcombinations

Atotalof foureventsare selectedwiththe searchin ZH.The background yield is estimated to 4.1±¹.8. The numbers ofob- served and expected events are shown in Table 7. The signal model assumes a SM ZH production rate and a 100% branching fraction to undetectable particles and one or two photons. The expected signal yield is larger for cases where mχ_1⁰ is smaller than mH/2 since there are two photons in the final state (H→



χ

₁⁰

χ

₁⁰→

γ γ

GG), and as a result the sensitivity improves for smallermasses.Good agreementbetweenthedataandtheback- ground prediction is observed. The transverse mass, m^E

miss T

T ≡

Fig. 3. Theexpectedandobserved95%CLupperlimitontheproductofcrosssec- tion,acceptance,andefficiency(σ(pp→γ+E^miss_T )A )form^γE

miss T

T >100 GeV,as functionoftheE^miss_T thresholdfortheggH channel.

Table 6

Observedyields,backgroundestimates,andsignalpredictionsat8 TeV intheggH channelfordifferentvaluesofthem_χ0

1 andfor differentcτ_χ0

1 ofthe χ₁⁰.These correspondtoB(^H→undetectable+γ)=^{100%,assumingtheSMcrosssectionat} thegivenmHhypothesis.Thecombinationofstatisticalandsystematicuncertainties isshownfortheyields.

Process Event yields

ggH(mH=125 GeV,m_χ0

1=65 GeV) 653±77

ggH(mH=^{125 GeV},m_χ0

1=^{95 GeV}) 1158±¹³⁷

ggH(mH=^{125 GeV},m_χ0

1=^{120 GeV}) 2935±³⁴⁹

ggH(mH=^{125 GeV},m_χ0

1=^{95 GeV})cτ_χ0

1=^{100 mm 983}±¹¹⁶

ggH(mH=^{125 GeV},m_χ0

1=^{95 GeV})cτ_χ0

1=^{1000 mm 463}±⁵⁵

ggH(mH=125 GeV,m_χ0

1=95 GeV)cτ_χ0

1=10 000 mm 83±10

ggH(mH=^{150 GeV},m_χ0

1=^{120 GeV}) 4160±⁴⁹¹

ggH(mH=200 GeV,m_χ0

1=170 GeV) 5963±704

ggH(mH=^{300 GeV},m_χ0

1=^{270 GeV}) 5152±⁶⁰⁸

ggH(mH=^{400 GeV},m_χ0

1=^{370 GeV}) 4057±⁴⁷⁹

γ+^{jets 179}±²⁸

jet→γ 269±94

e→γ 355±28

W(→ ν)+γ 154±15

Z(→νν¯)+γ 182±¹³

Other 91±¹⁰

Total background 1232±¹⁸⁸

Data 1296

2p_T p^E

miss T +E^γ_T

T [¹−^cos(φ_,Emiss

T +E^γ_T)]^,and|

η

^γ_|distributionsdis- criminate signal and background andare shownin Fig. 5 at the finalstepoftheselection.

The 95% CL upper limits are extracted from counting exper- iments in three categories: the model-specific selection in the ggH channel, and photons identified in the barrel and the end- capcalorimetersfortheZH channel.Resultsarecombinedusinga binned-likelihoodmethod.The95%CLupperlimitson(

σ

_B)/

σ

SM, where

σ

SM isthecrosssection fortheSM Higgsboson,are eval- uated for different mass values of

χ

₁⁰ ranging from 1 GeV to 120 GeV fortheindividualsearchesandtheircombinationandare showninFig. 6.Theupperlimitsformχ_1⁰<m_H/2 arenotshown fortheggH channelbecausethesensitivityisverylowduetothe combinationkinematicpropertiesandthecorrespondingselection;

inparticulartheE^miss_T andphotonp_Tvaluestendtobeoutsidethe