Study of the hard double-parton scattering contribution to inclusive four-lepton production in pp collisions at √s = 8 TeV with the ATLAS detector

(1)

Article

Reference

Study of the hard double-parton scattering contribution to inclusive four-lepton production in pp collisions at √s = 8 TeV with the ATLAS

detector

ATLAS Collaboration AKILLI, Ece (Collab.), et al .

Abstract

The inclusive production of four isolated charged leptons in pp collisions is analysed for the presence of hard double-parton scattering, using 20.2 fb−1 of data recorded in the ATLAS detector at the LHC at centre-of-mass energy s=8 TeV. In the four-lepton invariant-mass range of 80

ATLAS Collaboration, AKILLI, Ece (Collab.), et al . Study of the hard double-parton scattering contribution to inclusive four-lepton production in pp collisions at √s = 8 TeV with the ATLAS detector. Physics Letters. B , 2019, vol. 790, p. 595-614

DOI : 10.1016/j.physletb.2019.01.062

Available at:

http://archive-ouverte.unige.ch/unige:114707

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Study of the hard double-parton scattering contribution to inclusive four-lepton production in pp collisions at √

s = ^{8 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:

Received28November2018

Receivedinrevisedform30January2019 Accepted30January2019

Availableonline4February2019 Editor: W.-D.Schlatter

Keywords:

Four-leptonproduction DoubleDrell–Yan Doubleparton-scattering Higgsgoldendecaychannel

Theinclusiveproductionoffourisolated chargedleptons inpp collisionsis analysedforthe presence ofharddouble-partonscattering,using20.2 fb⁻¹ofdata recordedintheATLASdetectorattheLHCat centre-of-massenergy√_s

=^{8 TeV.}În^thefour-leptoninvariant-massrangeof80<m4<1000 GeV,an artificialneuralnetworkisusedtoenhancetheseparationbetweensingle- anddouble-partonscattering basedonthekinematicsofthefourleptonsinthefinalstate.Anupperlimitonthefractionofevents originating from double-parton scattering is determined at95% confidence level to be fDPS=⁰.042, whichresultsinanestimatedlowerlimitontheeffectivecrosssectionat95%confidencelevelof1.0 mb.

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

1. Introduction

Theparton–partonscatteringattheoriginofhardprocessesin ppinteractions isaccompaniedbyproton-remnantfragmentsthat contributetothehadronicfinalstate throughtheso-calledunder- lyingevent.Asfirstpointedout bySjöstrandandvanZijl [1],one source of the underlying-event activity, particularly in the high- energyregime ofthe LHC,is multi-parton interactions (MPI):in- teractionsofpairs ofpartons fromthe interactingprotons which occursimultaneouslywiththehardprocess.Inhigh-energy ppinteractions, where the density of low-x partons is high, there is enoughenergytoproducehardmulti-partoninteractions.Thesim- plestexample ishard double-parton scattering(DPS), where two partonsfromeachprotoninteractwitheachotherleadingtoper- turbativefinalstates.

The interest in studying DPS is twofold. Firstly, the probabil- ity of occurrence of DPS and the potential correlations between theproducts ofthese two perturbative interactions providevalu- ableinformationaboutthe dynamicsof thepartonicstructure of theproton(seeRef. [2] andreferencestherein).Secondly,DPSpro- cesses mayalso constitute a background to reactions proceeding throughsingle-partonscattering(SPS).An exampleistheproduc- tion of fourcharged leptons in the ﬁnal state, addressedin this Letter. This reaction is dominated by the SPS production of two Z^(∗)bosons,followedbysubsequentleptonicdecays.The Z^(∗)no-

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

tation indicates the production of on- or off-shell Z bosons (Z andZ^∗),ortheproductionofoff-shellphotons(γ^∗).However,the fourleptonscouldalsobeproducedastheresultoftwoDrell–Yan processesoccurringsimultaneously,potentiallydistortingthemea- surementsofprompt-leptonproduction.

For a process pp→^A+^B+^X, ^the ^expected ^DPS ^cross ^sec- tionforproducingstates A andB intwoindependentscatterings,

σ_DPS^{A B},maybeestimatedfromthefollowingformula [3–5] (seealso Ref. [6] foradetailedderivation):

σ_DPS^{A B}=^k 2

σ_SPS^A σ_SPS^B σeff

, (1)

where σ_SPSÂ⁽^B⁾ denotes the production cross section of state A(B) in a single-parton scattering, the symmetry factor k depends on whetherthe two scatteringslead to thesame final state (A=^B, k=¹⁾ ôr^different^final ^states⁽Â=^B,^k=^2), ândσeff represents theeffectivetransverseoverlapareacontainingtheinteractingpar- tons.

For mostof theexisting measurements [7–21], σeff ﬂuctuates around 15 mb.However,fortheassociatedproductionofquarko- nia J/ψJ/ψ or J/ψϒ, σeff is systematically lower [22–25] than for all other investigated processes. This might indicate that σ_eff

isnot universalandthatthereare spatialfluctuationsofthepar- ton densitiesinthe proton,whichmayfavourcertain final states over others [26,27]. The concept of geometric fluctuationsin the spatial parton densities has also been invoked [28] to explain the collective phenomena observed in high-multiplicity proton–

https://doi.org/10.1016/j.physletb.2019.01.062

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

(3)

proton and proton–nucleus interactions [29–32]. In pp interactions at√

s=8 TeV [33], the doubleDrell–Yancontribution may add 0.3% to theyieldof fourleptons intheinvariant-mass range 80 < m₄ < 1000 GeV, using Eq. (1) with σeff=^{15 mb.} ^The latterisobtainedfromcalculationsofthe Drell–Yancrosssection in the phase space of the measurement in next-to-leading-order (NLO)QCDwithPowheg-Box[34–36].

SincedoubleDrell–Yanproductionisdrivenbyquark–antiquark annihilation,whilemostofthepreviouslyexplored DPSprocesses are driven by gluon–gluon scattering, andthe ﬁnal state offour chargedleptons constitutesthe goldenchannel forthe studiesof Higgsbosonproperties, H→^Z^(∗)^Z^(∗)→⁴,a studyofa possible DPScontributiontotheproductionoffourisolatedchargedleptons at√

s=8 TeV is warranted.The analysispresentedinthisLetter closelyfollowsaprevious analysisofthisﬁnal state [33],butex- tendsittoconsiderDPS.

2. ATLASdetector

The ATLAS detector [37] is a multipurpose particle detector with a forward–backward symmetric cylindrical geometry and nearlyfullcoverageinsolidangle.¹

Itconsistsofaninnertrackingdetector(ID)systemsurrounded by a superconducting solenoid, electromagnetic and hadronic calorimeters, and a muon spectrometer (MS) incorporating superconducting toroid magnets. During Run 1 of the LHC the ID consistedofapixeldetectorclosesttothebeam-pipe,followedby a siliconstrip detector anda transitionradiation tracker.This ID system,operatingina2 T axialmagneticﬁeld,providesthetrack- ingofchargedparticleswithinthepseudorapidityrange|η|<2.5.

The calorimeter system, which covers the range |η_|<4.9, includes in the barrel region a high-granularity lead/liquid-argon (LAr) barrel electromagnetic (EM) calorimeter (|η|<1.5) and a steel/scintillator-tile hadronic calorimeter(|η|<1.7). In the end- cap(1.5<|η_|<3.2)andforward(3.2<|η_|<4.9)regions,theEM calorimeterand thehadronic calorimeterare made ofLAr active layers with either copper or tungsten as the absorber material.

The muon spectrometer constitutes the outermost detector and includes fast trigger chamberscovering the region |η_|<2.4 and high-precisiontrackingchamberscovering|η|<2.7.Athree-level triggersystem [38] wasusedtoselecteventstoberecorded.

3. MonteCarloeventsamples

InSPSevents,thefour-leptoneventscorrespondtotheproduc- tion andsubsequent decay ofresonant Z orHiggs bosons, orto the productionof the continuum Z⁽^∗⁾Z⁽^∗⁾ system. Inthe case of DPS, thefourleptons aredecayproducts oftwo Z⁽^∗⁾bosons that areproducedintwodistinctparton–partonscatteringswithinthe sameppinteraction.

The Monte Carlo samples are unchanged with respect to Ref. [33]. The SPS q^q¯ →⁴ was simulated with the Powheg- Box (revision 2330) [34–36] Monte Carlo (MC) program, which isbasedonperturbative QCDcalculationsatNLO.Thefour-lepton productionthrough theqg initial state isincluded aspartofthe NLOcontributionstotheq^q¯ ^process.^The^partondistributionfunc- tions (PDFs) of the CT10NLO [39] set were used. The gg→⁴

1 ATLASusesaright-handed coordinatesystemwith itsoriginat thenominal interactionpoint(IP)inthecentreofthedetectorandthez-axisalongthebeam pipe.Thex-axispointsfromtheIPtothecentreoftheLHCring,andthe y-axis pointsupwards.Cylindricalcoordinates(r,φ)areusedinthetransverseplane,φ beingtheazimuthalanglearoundthez-axis.Thepseudorapidityisdeﬁnedinterms ofthepolarangleθasη= −^{ln tan}(θ/2).Angulardistanceismeasuredinunitsof R≡

(η)²+(φ)².

events corresponding tothe continuum Z^(∗)Z^(∗) production were generated withMCFM 6.1 [40] atleading order (LO) in QCD, using the CT10NNLO [41] set of PDFs,andthe cross sectionswere corrected for higher-order effects using the ratio of NLO to LO cross sections (the so-called K-factors) [42]. The on-shell Higgs boson production was simulatedwith Powheg-Box atNLO QCD, using the CT10NLO PDFs, in the case ofgluon–gluon fusion and vector-boson fusion, and with LO Pythia 8 [43] in the case of vector-boson associatedproduction (V H) andtop-pair associated production (tt H¯ ). The event yield of on-shell Higgs boson was normalised to the higher-order corrected cross section [44]. The eventswithoff-shellHiggsbosonproductionweresimulatedwith the LOMadGraph 5.1.5.12 [45] generator via vector-bosonfusion andvector-bosonscatteringprocesses,includingtheirinterference.

FortheLOPythia8andMadGraphgenerators,theLOversionof CTEQ6L1PDFs [46] wasused.

The MC generators listed above were interfaced to Pythia 8 for partonshowering, exceptMadGraph whichwas interfacedto Pythia 6 [47]. The underlying-event parameter values belong to theAU2 [48] tune.

TheDPSeventsthatcontributetothe4productionweresim- ulatedwithPythia8.175usingtheLOversionofCTEQ6L1PDFs.

Background events may originate from Z +^jets, ^t¯t, diboson (Z W, Zγ),triboson V V V (V =^Z,W), V H, and Z+^{top (t}¯t and t)processes.

The production of Z +jets events, including the light- and heavy-ﬂavourcontributions,wassimulatedwithAlpgen2.1.4 [49], using thePerugia2011C [50] tune. The Zγ production was mod- elledwithSherpa1.4.5 [51].Backgroundtt¯eventsweregenerated with Powheg-Box using the Perugia2011C tune. The Z H events, withsubsequentdecays Z→and H→^{V V}^∗ ^(with^two^leptons andtwoneutrinosortwoleptons andtwojetsintheﬁnalstate), were generatedwithPythia 8,usingtheAU2 tune.The Z W and t Z processeswere simulatedwithSherpaandMadGraph respectively,withthelatterusingtheAUET2Btune [52].Thebackground contribution from V V V andt¯t Z was modelled with MadGraph, usingthe AUET2Btune.TheMC generators forbackgroundsimu- lationusedtheLOversionoftheCTEQ6L1PDFset,exceptSherpa, whichusedtheCT10PDFset.

The largest contributions to the background, originatingfrom Z+^b_b¯ _{jets and} _t¯t production, were estimatedin Ref. [33] from therespectiveMCsamplesnormalisedtothedatainselectedcon- trolregions.Theremainingbackgroundcontributionsweredirectly extractedfromtheMCexpectations.

Additional pp interactions occurring in the same and neigh- bouring bunch crossings(pile-up)were alsosimulated,using the Pythia 8 MC generator, withthe A2 [53] tuneand MSTW 2008 LO [54] PDF set.The MC sampleswere reweighted to reproduce thedistributionofthemeannumberof ppinteractionsperbunch crossing observed in the data. The estimated number of events withtwo Z⁽^∗⁾ bosons producedin thesamebunch crossingwith lessthan1 cm separationalongthebeamaxisisnegligiblecom- paredtotheDPSexpectations.

Monte Carlo events were passed through the ATLAS detector simulation [55], which is based on the Geant4 [56] framework, and which includes simulation of the trigger selection. The MC eventswerereconstructedandselectedoﬄineusingthesamesoft- wareandselectionsasforthedata.

4. Eventselection

The dataset and the event selection are unchanged with re- specttoRef. [33].Theupdatedluminosityoftheanalysedsample is20.2 fb⁻¹.Theeventswereselectedonlineusingsingle-leptonor dileptontriggers.Thesingle-leptontriggerrequiredthetransverse

(4)

Fig. 1.The distribution ofthe four-lepton invariant mass, m4. The data (black dots) are compared with the sum of signal and background MC expectations (ﬁlledcolouredhistograms).AlsoshownistheexpectedcontributionofDPSfrom Pythia8.

energyoftheelectron candidateorthetransverse momentumof themuoncandidatetobeabove24 GeV.Thedielectrontriggerhad the same threshold of 12 GeV forboth electron candidates. The dimuon trigger required either two muons with transverse mo- mentumabove13 GeV oroneabove 18 GeV andtheother above 8 GeV.Anelectron–muontriggerwasalsousedwiththresholdsat 12 GeV forelectronsand8 GeV formuons.

Theﬁnal sample consists ofeventswith atleastfour leptons, whereeachlepton iseitheran electronoramuon. Thefourlep- tonsarerequired toformtwo same-ﬂavour (electronsormuons) opposite-charge (SFOC) lepton pairs. The pair with the invariant mass closer to the mass of the Z boson is called the leading pair, and the other pair is the sub-leading one. The invariant massoftheleading pairisrestrictedto therange50<m_leading<

120 GeV,while forthesub-leading pairthe massrequirementis 12<msub-leading<120 GeV. A J/ψ veto isapplied such that for anySFOC lepton combinationthe invariant massof thedilepton, m2,mustbegreaterthan5 GeV.Onlyeventswiththefour-lepton invariant mass in the range 80<m4<1000 GeV are selected.

The transverse momentum of dileptons, p_T⁺⁻, is required to be above 2 GeV. Selected leptons, ordered in descending order of transversemomentum, are required to havetransverse momenta pT above 20, 15, 10 (8 if muon), and 7 (6 if muon) GeV. The leptons areselected within thepseudorapidity range |η_|<2.5 in thecaseofelectronsand|η|<2.7 inthecaseofmuons.Inorder tohavewell-measuredleptons,aleptonseparationrequirementis imposed,such that thedistance betweenanytwo leptons in the

η–φ space, R,is requiredto fulfilthe condition R>0.1(0.2) forsame-flavour(different-flavour)leptons.Eacheventisrequired tohavethe triggeringlepton(s)matchedtoone ortwo ofthese- lectedleptons.

Thedata sample,after all selections,contains 476 events.The resulting data and MC distributions of the four-lepton invariant massareshowninFig.1.Forcompleteness,theﬁgurealsoincludes theDPScontributionof0.4 eventspredictedbythe Pythia 8.175 simulation.

5. DPSsignalextraction

TheassumptionthatinDPSthetwoscattersaredistinctimplies that, inthe DPSfour-lepton ﬁnal states,the two leptons of each dileptonwilltendtobebalancedin pTandthereforeback-to-back in the azimuthal angle φ, dueto the dominance of low-pT Z^(∗) production.IntheSPScase, theleadingandsub-leadingpairsare expectedtobalanceeachotherinp_T.

Based on the experience gained in the study of four-jet ﬁnal states [57], in order to distinguish between DPS events andSPS events,thedistributionsofthefollowingkinematicvariablesofthe fourleptonsareconsidered:

p_T_,_{i j}=|^pT,i+ ^pT,j| p_T_,_i+^pT,j

, φ_{i j}= |φ_i−φ_j|, yi j= |^yi−^yj|, i,j=¹,2,3,4, i=^j

_{i jkm}= |φi+^j−φ_k₊_m|, i jkm=1234,1324,1423.

(2)

Here, pT,iisthetransversemomentumcomponentofthei-thlep- ton(i=¹,2,3,4),andφi and yi are theazimuthalangleandthe rapidity ofthe i-thlepton,respectively.The angleφi+^j isthe azimuthal angleofthemomentum vectorcomposed bythe sumof momenta of leptons i and j. Leptons 1 and 2 form the leading dilepton.The leptonordering ischosen suchthat pT,1>pT,2 and pT,3>pT,4.

The distributions of the variables pT,12, φ13, y13, and 1234 are presented in Fig. 2(a)–(d). The distribution of p_T_,₁₂ peaks around 0.1 for simulated DPS events, while the simulated SPSeventsaremoreevenlydistributedacrosstherange[0,1].This demonstratesthat,asexpected,twoleptonscomingfromthesame Z candidateinDPSbalanceeachotherinpT,whileinSPSthepair- wise p_T balance is not dominant. This is againdemonstrated in the φ13 distribution,whereleptons 1and3are decorrelated in φ forDPS,whilefortheSPSeventstheseleading-pT decaylep- tonstendtobeback-to-backinφ,becausetheyoriginatefromthe two Z bosons,whichthemselves areexpectedtobe back-to-back inφ.The y₁₃distribution showsthatleptons associatedto dif- ferentdileptonstendtobemoreseparatedinrapidityinDPSthan in SPS. The back-to-back conﬁgurations of the two Z candidates in the caseof SPS, andtheir decorrelationin the caseof DPS is explicitlydemonstratedinthedistributionoftheazimuthal angle betweentwo Z candidates,1234.

ThedifferencebetweenthetopologiesofSPSandDPSeventsis usedtotrainanartiﬁcialneuralnetwork(ANN)todiscriminatebe- tweentheDPSandnon-DPSclasses,wherethelattercorresponds toSPSandbackgroundevents.

The training is performed with the ANN available in the ROOT [58] implementation of a feed-forward multilayer percep- tron. The Broyden–Fletcher–Goldfarb–Shanno supervised learning algorithm [59–62] isusedinthetraining.Theinputlayercontains 21neurons,correspondingtothevariableslistedinEq. (2),andthe outputlayerconsistsofoneneuron.Astheresultofoptimisingthe convergence andtheperformance ofthe ANN, aconﬁguration of 30and9neuronsisadoptedfortheﬁrstandsecondhiddenlayer, respectively.TheoutputoftheANN,ξDPS,isanumberdistributed between0and1,whichrepresentsthelikelihoodforan eventto belongtotheDPSclass.

The event weights are chosen such that during the training procedure the effective numbers of SPS qq-initiated¯ events, gg-initiated eventsandbackground Z+^b_b¯ jets eventsarein the ratio 1:¹:^1. ^The ^SPS gg-initiated events tend to spill over into the DPS signal region, anda better separation between the SPS andDPSclassesisachievedbyincreasingtheirweightinthemin- imisationoftheerrorfunction.Similarly,theeffectivecontribution of Z+^b_b¯ jets eventsisincreasedfor theANN training to distinguish them better from the DPS ones, as the kinematics of the Z +^bb¯ jets background subprocess has features similar to DPS.

The effectivenumbers ofeventsforDPS andnon-DPSeventsare equal.EachMC setissplit randomlyintotwo subsetshavingap- proximatelythesamenumberofevents.Onesubsetisusedforthe ANNtraining, whiletheotherisusedtovalidatetheperformance oftheANNandtodeterminethenumberoftrainingepochs,soas