Measurement of exclusive $\gamma \gamma \rightarrow l^{+}l^{-}$ production in proton–proton collisions at $\sqrt{s}=7$ TeV with the ATLAS detector

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Measurement of exclusive γ γ _{→ }



production in proton–proton collisions at √

s = ^{7 TeV with the ATLAS detector}

.ATLASCollaboration^

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

Articlehistory:

Received23June2015

Receivedinrevisedform27July2015 Accepted28July2015

Availableonline31July2015 Editor:W.-D.Schlatter

ThisLetterreportsameasurementoftheexclusiveγ γ_{→ }⁺⁻(=^e,μ)cross-sectioninproton–proton collisions atacentre-of-massenergyof7 TeV bytheATLASexperimentattheLHC,basedonaninte- gratedluminosityof4.6 fb⁻¹.Fortheelectronormuonpairssatisfyingexclusiveselectioncriteria,afit tothedileptonacoplanaritydistributionisusedtoextract thefiducialcross-sections.Thecross-section in theelectronchannel isdetermined tobe σ_{γ γ→}^excl._e+e⁻=⁰.428 ± ⁰.035(stat.) ± ⁰.018(syst.)pb for a phase–space region with invariant mass of the electron pairs greater than 24 GeV, in which bothelectrons havetransverse momentumpT>12 GeV andpseudorapidity|η_|<2.4.Formuonpairs withinvariantmassgreaterthan20 GeV,muontransversemomentumpT>10 GeV andpseudorapidity

|η_|<2.4,thecross-sectionisdeterminedtobeσ_{γ γ}^excl_→μ^. _+μ−=0.628 ± ⁰.032(stat.) ± ⁰.021(syst.)pb.

Whenprotonabsorptiveeffectsduetothefinitesizeoftheprotonaretakenintoaccountinthetheory calculation,themeasuredcross-sectionsarefoundtobeconsistentwiththetheoryprediction.

©^{2015CERNforthebenefitoftheATLAS}Collaboration.PublishedbyElsevierB.V.Thisisanopen accessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

Aconsiderablefractionofproton–proton(pp)collisionsathigh energies involve reactions mediated by photons. This fraction is dominated by elastic scattering, with a single photon exchange.

Quasi-real photons can also be emitted by both protons, with a variety of final states produced. In these processes the pp colli- sion can be then considered as a photon–photon (γ γ) collision.

At the LHC, these reactions can be studied at energies well be- yond the electroweak energy scale [1]. The cross-section of the pp(γ γ)→ ⁺⁻X processhasbeenpredictedtoincreasewithen- ergy[2]andconstitutesanon-negligiblebackgroundtoDrell–Yan (DY)reactions[3].

Theexclusivetwo-photonproductionofleptonpairs(pp(γ γ)→

⁺⁻pp,referredtoasexclusive γ γ→ ⁺⁻)canbecalculatedin theframeworkofquantumelectrodynamics(QED)[4,5],withinun- certaintiesoflessthan2%associatedwiththeprotonelasticform- factors.Exclusivedileptoneventshaveacleansignaturethathelps discriminate them from background: there are only two identi- fiedmuonsorelectrons,withoutanyother activityinthecentral detectors,andtheleptonsareback-to-backinazimuthalangle.Fur- thermore,due tothe very smallphoton virtualities involved,the incidentprotonsarescatteredatalmostzero-degreeangles.Conse- quently,themeasurementofexclusive γ γ → ⁺⁻ reactionswas proposed forpreciseabsoluteluminosity measurement athadron

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

colliders [5–8]. However, this process requires significant correc- tions (oftheorderof20%)duetoadditionalinteractionsbetween theelasticallyscatteredprotons[9,10].

At hadron colliders exclusive γ γ → ⁺⁻ events have been observedinep collisionsatHERA[11],inpp collisions¯ attheTeva- tron[12–14]andinnucleus–nucleuscollisionsatRHIC[15,16]and theLHC[17].Theexclusivetwo-photonproductionofleptonpairs in pp collisions attheLHC was studiedrecentlyby the CMScol- laboration[18,19].

ThisLetterreportsameasurementofexclusivedileptonproduc- tioninpp collisionsat√

s=^{7 TeV.The}measurementofexclusive dilepton production cross-section is compared to the QED-based predictionwithandwithoutprotonabsorptivecorrections.

2. TheATLASdetector

The ATLASexperiment[20]attheLHCisamulti-purposepar- ticledetectorwitha forward–backwardsymmetriccylindricalge- ometry andnearly 4π coverage in solid angle.¹ It consistsofin- ner tracking devices surrounded by a superconducting solenoid,

1 ATLAS usesaright-handedcoordinatesystemwithitsoriginatthe nominal interactionpointinthecentreofthedetectorandthe z-axiscoincidingwiththe axisofthebeampipe.Thex-axispointsfromtheinteractionpointtothecentreof theLHCring,andthey-axispointsupward.Thepseudorapidityisdefinedinterms ofthepolarangleθasη_{= −}ln tan(θ/2),andφistheazimuthalanglearoundthe beam pipewith respecttothe x-axis.Theangular distanceis definedasR=

(η)²+ (φ)².Thetransversemomentumisdefinedrelativetothebeamaxis.

http://dx.doi.org/10.1016/j.physletb.2015.07.069

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

electromagnetic andhadroniccalorimeters,andamuonspectrom- eter. The inner detector (ID) provides charged-particle tracking in the pseudorapidity region |η_|<2.5 and vertex reconstruc- tion. It comprises a silicon pixel detector, a silicon microstrip tracker,andastraw-tubetransitionradiationtracker.TheIDissur- roundedby a solenoid that produces a 2 T axial magnetic field.

Lead/liquid-argon(LAr)samplingcalorimetersprovideelectromag- netic(EM)energymeasurements withhighgranularity. Ahadron (iron/scintillator-tile)calorimetercoversthecentralpseudorapidity range|η_|<1.7.Theend-capandforwardregionsareinstrumented withLArcalorimetersforboththeEMandhadronicenergymea- surements up to |η|=⁴.9. The muon spectrometer (MS) is op- erated in a magnetic field provided by air-core superconducting toroidsandincludestrackingchambersforprecisemuonmomen- tummeasurementsup to|η|=².7 and triggerchamberscovering therange|η_|<2.4.

Athree-leveltriggersystemisusedtoselectinterestingevents.

The first level is implemented in custom electronics and is fol- lowedbytwosoftware-basedtriggerlevels,referredtocollectively astheHigh-LevelTrigger.

3. Theoreticalbackgroundandeventsimulation

Calculationsofthecross-sectionforexclusivetwo-photonpro- ductionofleptonpairsinpp collisionsarebasedontheEquivalent Photon Approximation (EPA) [4,5,21–24]. The EPA relies on the property that the EM field of a charged particle, here a proton, movingathighvelocitybecomes moreandmoretransversewith respecttothedirectionofpropagation. Asa consequence,anob- serverinthelaboratoryframecannotdistinguishbetweentheEM fieldofa relativisticprotonandthetransverse componentofthe EMfield associatedwithequivalent photons. Therefore,usingthe EPA,thecross-sectionforthereactionabovecanbewrittenas

σ_pp^EPA_{(γ γ)→+−pp=}^x P(x1)P(x2)σ_{γ γ→}+⁻(m²_+−)dx1dx2,

where P(x1)and P(x2) arethe equivalentphoton spectraforthe protons, x1 and x2 are the fractions of the proton energy car- ried away by the emitted photons and m_+⁻ is the invariant massofthe leptonpair.Thesevariablesare relatedby m²_+−/s= x1x2wheres isthepp centre-of-massenergysquared.Thesymbol

σ_{γ γ→}+⁻ refersto thecross-sectionfortheQEDsub-process. As discussedpreviously,thephotonsarequasi-real,whichmeansthat theirvirtuality Q² isverysmallcomparedtom²_+−.Inthiskine- matic region the EPAgives the same predictions asfull leading- order(LO)QEDcalculations[4,5].

In the reaction pp(γ γ)→ ⁺⁻X the protons scattering can be: elastic, X = ^pp; single-dissociative, X = ^{p X; or double-} dissociative, X=^{XX (thesymbols X, X denoteanyadditional} final state produced in the event). Unless both outgoing protons are detected, the proton dissociative events form an irreducible backgroundtothefullyelasticproduction.

Such photon-induced reactions, in particular exclusive γ γ →

⁺⁻production,requiresignificantcorrectionsduetoprotonab- sorptive effects. These effects are mainly related to pp strong- interactionexchangesthataccompany thetwo-photon interaction andthat leadtotheproductionofadditionalhadronsinthefinal state.Recent phenomenologicalstudiessuggest that theexclusive

γ γ_{→ }⁺⁻ cross-section issuppressedby a factorthat depends on the mass and rapidity of the system produced [10]. For the kinematic range relevant for this measurement the suppression factor is about 20%. This factor includes both the strong pp ab- sorptivecorrection(∼^8%suppression)andthephoton–proton(γp) coherence condition (bγp>rp, wherebγp is the γp impact pa- rameterandr_p thetransversesizeoftheproton).

Simulated event samples are generated in order to estimate the background and to correct the signal yields for detector ef- fects. Thesignaleventsamplesforexclusive γ γ_{→ }⁺⁻ produc- tion are generated using the Herwig++ ^{2.6.3 [25] Monte Carlo} (MC) event generator, which implements the EPA formalism in pp collisions. The dominant background, photon-induced single- dissociativedileptonproduction,issimulatedusing Lpair 4.0[26]

with the Brasse [27] and Suri–Yennie [28] structure functions for proton dissociation. For photon virtualities Q²<5 GeV² and masses of the dissociating system, m_N<2 GeV, low-multiplicity states fromthe productionanddecays of resonances are usu- ally created. For higher Q² or mN, the system decays to a va- riety of resonances, which produce a large number of forward particles. The Lpair package is interfaced to JetSet 7.408 [29], where the Lund [30] fragmentation model is implemented. The Herwig++ and Lpair generators do not include any corrections toaccountforprotonabsorptiveeffects.

Fordouble-dissociativereactions, Pythia 8.175[31]isusedwith the NNPDF2.3QED [32] parton distribution functions (PDF). The NNPDF2.3QEDsetuses LOQEDandnext-to-next-to-leading-order (NNLO)QCDperturbativecalculationstoconstructthephotonPDF, startingfromtheinitialscaleQ₀²=^{2 GeV2.Dependingonthemul-} tiplicityofthe dissociatingsystem, the default Pythia 8 stringor mini-string fragmentation model is used for proton dissociation.

The absorptiveeffectsindouble-dissociativeMC eventsare taken into account using the default multi-parton interactions model in Pythia 8[33].

The Powheg 1.0 [34–36] MC generator is used with the CT10[37]PDFtogenerateboththeDY Z/γ^∗→^{e+e−and Z}/γ^∗→ μ⁺μ⁻ events. It is interfaced with Pythia 6.425 [38] using the CTEQ6L1 [39] PDF set and the AUET2B [40] values of the tun- ableparameterstosimulatethepartonshowerandtheunderlying event (UE). These samples are referred to as Powheg+^Pythia. The DY Z/γ^∗_→τ⁺τ⁻ process is generated using Pythia 6.425 together with the MRST LO* [41] PDF. The transverse momen- tum of lepton pairs in Powheg+^{Pythia samples is reweighted} to a Resbos [42] prediction, whichis found to yield good agree- mentwiththetransversemomentumdistributionof Z bosonsob- servedindata[43,44].Theproductionoftop-quarkpair(tt) events¯ is modelled using MC@NLO 3.42 [45,46] and diboson (W⁺W⁻, W^±Z , Z Z ) processesaresimulatedusing Herwig 6.520[47].The eventgeneratorsusedtomodelZ/γ^∗,t¯t anddibosonreactionsare interfacedto Photos 3.0[48]tosimulateQEDfinal-stateradiation (FSR)corrections.

Multiple interactions per bunch crossing (pile-up) are ac- countedforby overlayingsimulatedminimum-biasevents,gener- atedwith Pythia 6.425usingtheAUET2Btune andCTEQ6L1PDF, andreweighting the distribution ofthe averagenumber ofinter- actions per bunch crossing inMC simulation to that observed in data.Furthermore,the simulatedsamplesareweighted such that thez-positiondistributionofreconstructedpp interactionvertices matchesthedistributionobservedindata.TheATLASdetectorre- sponseismodelledusingtheGEANT4toolkit[49,50]andthesame eventreconstructionasthatusedfordataisperformed.

4. Eventreconstruction,preselectionandbackgroundestimation

The datausedin thisanalysiswerecollected during the 2011 LHCpp runatacentre-of-massenergyof√

s=^{7 TeV.Afterappli-} cationofdata-qualityrequirements,thetotalintegratedluminosity is 4.6 fb⁻¹ with an uncertainty of 1.8% [51]. Events from these pp collisions are selected by requiring atleast one collision ver- texwithatleasttwo charged-particle trackswith pT>400 MeV.

Events are then required to have at least two lepton candidates (electronsormuons),asdefinedbelow.

Eventsinthe electronchannelwere selectedonlineby requir- inga single-electronordi-electrontrigger.Forthesingle-electron trigger,thetransversemomentumthresholdwasincreasedduring data-taking from20 GeV to 22 GeV inresponse to the increased LHC instantaneous luminosity. The di-electron trigger required a minimumtransversemomentumof12 GeV for eachelectroncan- didate.Electroncandidatesarereconstructedfromenergydeposits in the calorimeter matched to ID tracks. Electron reconstruction uses track refitting with a Gaussian-sum filter to be less sensi- tive to bremsstrahlung losses and improve the estimates of the electron track parameters [52,53]. The electrons are required to have a transverse momentum p^e_T>12 GeV and pseudorapidity

|η^e|<2.4 with the calorimeter barrel/end-cap transition region 1.37<|η^e|<1.52 excluded. Electron candidates are required to meet“medium”identificationcriteriabasedonshower shapeand track-qualityvariables[54].

Eventsin themuon channelwere selected onlinebya single- muonordi-muontrigger,withatransversemomentumthreshold of18 GeV or10 GeV,respectively.Muoncandidatesareidentified by matching complete tracks inthe MS totracks in the ID [55], andare required tohave pμ

T >10 GeV and |η^μ_|<2.4.Only iso- latedmuonsareselectedbyrequiringthescalarsumofthe pT of thetrackswithp_T>1 GeV inaR=⁰.2 conearoundthemuon tobelessthan10%ofthemuon pT.

Di-electron (di-muon) events are selected by requiring two oppositely charged same-flavour leptons with an invariant mass m_e+e⁻ >24 GeV for the electron channel and mμ⁺μ⁻ >20 GeV for the muon channel. After these preselection requirements 1.57×¹⁰⁶ di-electron and2.42×^{106 di-muon candidate events} arefoundinthedata.

The background to the exclusive signal includes contributions fromsingle- anddouble-proton dissociative γ γ_{→ }⁺⁻ produc- tion, as well as Z/γ^∗, diboson, tt and¯ multi-jet production. The contribution from γ γ →^{W+W− and} γ γ →τ⁺τ⁻ processes is considered negligible. Single- anddouble-dissociative background contributionsareestimatedusingMCsimulations.Theelectroweak ( Z/γ^∗,diboson) andtop-quarkpair backgroundcontributionsare alsoestimatedfromsimulations andnormalisedtotherespective inclusive cross-sections calculated at high orders in perturbative QCD(pQCD),asinRef.[56].Scalefactorsareappliedtothesimu- latedsamplestocorrectforthesmalldifferencesfromdatainthe trigger, reconstruction and identificationefficiencies for electrons andmuons[54–56].MCeventsarealsocorrectedtotake intoac- countdifferencesfromdatainleptonenergy,momentumscaleand resolution[55,57].

The multi-jet background is determined using data-driven methods, similarly to Refs. [44,58]. For the e⁺e⁻ channel, the multi-jet sample is obtained by applying the full nominal pres- election but requiring the electron candidates to not satisfy the medium identification criteria. For the μ⁺μ⁻ channel, it is ex- tractedusing same-charge muon pairs that satisfy the remaining preselectioncriteria.Thenormalisationofthemulti-jetbackground isdetermined byfittingthe invariantmass spectrumoftheelec- tron(muon) pairinthe datato asumofexpectedcontributions, includingMCpredictionsofthesignalandtheotherbackgrounds.

5. Exclusiveeventselectionandsignalextraction

Inorder toselectexclusive γ γ→ ⁺⁻ candidates, avetoon additionalcharged-particle trackactivityisapplied.This exclusiv- ity veto requires that no additional charged-particle tracks with p_T>400 MeV beassociatedwiththedileptonvertex,andthatno additionaltracksorverticesbefoundwithina3 mmlongitudinal isolation distance, z^iso_vtx, from the dilepton vertex. These condi- tions are primarily motivated by the rejection of the Z/γ^∗ and

multi-jetevents,whichtypicallyhavemanytracksoriginatingfrom thesamevertex.

The charged-particle multiplicity distribution in Z/γ^∗ MC events is reweighted to match the UE observed in data, follow- ing the same procedure as in Ref. [59]. Uncorrected Z/γ^∗ MC modelsoverestimatethecharged-particlemultiplicitydistributions observed in data by 50% for low-multiplicity events. In order to estimate the relevant weight, the events in the Z -peak region, defined as 70 GeV<m_+⁻ <105 GeV, are used. This region is expectedto includea largeDY component. The correctionproce- durealsoaccountsfortheeffectoftracksoriginatingfrompile-up and ID track reconstruction inefficiency. The requirement of no additional tracks associated with the dilepton vertex completely removesmulti-jet,tt,¯ anddibosonbackgrounds.

The z^iso_vtx distribution forevents withno additionaltracks at thedileptonvertexispresentedinFig. 1(a).Thestructureobserved atsmallz^iso_vtx valuesisduetothevertexfindingalgorithm,which identifies the vertexastwoclosevertices inhigh-multiplicityDY events:thetwo-trackvertexformedfromtheleptontracksandthe vertexfromtheUE tracks.The3mm cutsignificantly suppresses theDY background,atthecostofa26% reductioninsignalyield.

The inefficiency is relatedto tracksandvertices originatingfrom additionalpp interactions.

Contributionsfromthe DY e⁺e⁻ and μ⁺μ⁻ processescan be furtherreducedbyexcludingeventswithadileptoninvariantmass in the Z -peak region. The invariant mass distribution of muon pairs foreventssatisfyingtheexclusivityveto(exactlytwo tracks at the dilepton vertex, z^iso_vtx>3 mm) is presented in Fig. 1(b) (where theexcluded Z -peak regionis indicatedby dashed lines).

The figure shows that the MC description of the mμ⁺μ⁻ distri- bution is satisfactory. To further suppressthe proton dissociative backgrounds,theleptonpairisrequiredtohavesmalltotaltrans- verse momentum (p^_T^+− <1.5 GeV). This is shown in Fig. 1(c), whichdisplaysthedi-muontransversemomentumdistributionfor events outsidethe Z region that satisfy theexclusivity veto. The p^_T^+− resolution below 1.5 GeV is approximately0.3 GeV for the electronchanneland0.2 GeV forthemuonchannel.

Theresultofeachstepoftheexclusiveselectionappliedtothe data,signalandbackgroundsamplesisshowninTable 1.Afterall selection criteria are applied, 869events remain forthe electron channel,and2124eventsareselectedinthemuonchannel.From simulations,approximatelyhalfareexpectedtooriginatefromex- clusive production.The number ofselected events inthe data is belowtheexpectationfromthesimulation,withanobservedyield thatisapproximately80%ofthesumofsimulatedsignalandback- groundprocesses(seediscussioninSection7).

Afterthe final exclusive eventselection, thereis still a signif- icant contaminationfromDY, single- anddouble-dissociativepro- cesses.Scalingfactorsforsignalandbackgroundprocessesareesti- matedbyabinnedmaximum-likelihoodfitofthesumofthesim- ulateddistributionscontainedintheMCtemplatesforthevarious processes,tothemeasureddileptonacoplanarity(1− |φ⁺⁻| /π) distribution. The fit determines two scaling factors, defined as the ratiosof thenumberof observedto thenumber ofexpected eventsbasedon theMCpredictions,fortheexclusive(R^excl.)and single-dissociative (R^s-diss.) templates.Thedouble-dissociativeand DY contributionsare fixedtotheMC predictionsinthefitproce- dure.Contributionsfromotherbackgroundprocessesarefoundto benegligible.

Fig. 2 shows the e⁺e⁻ and μ⁺μ⁻ acoplanarity distributions in data overlaid with the result of the fit to the shapes from MC simulations for events satisfying all selection requirements.

The resultsfromthe bestfitto thedata fortheelectron channel are: R^exclγ γ→^. e⁺e⁻=⁰.863±⁰.070(stat.)forthesignalscalingfactor and R^s-dissγ γ→^.e⁺e⁻ =⁰.759±⁰.080(stat.) forthe single-dissociative