Search for new phenomena in dijet mass and angular distributions from $\mathit{pp}$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Search for new phenomena in dijet mass and angular distributions from pp collisions at √

s = 13 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:

Received7December2015

Receivedinrevisedform18January2016 Accepted18January2016

Availableonline20January2016 Editor: W.-D.Schlatter

This Letter describes a model-agnostic search for pairs of jets (dijets) produced by resonant and non-resonant phenomena beyondthe Standard Model in3.6 fb⁻¹ ofproton–proton collisions witha centre-of-mass energy of√

s=13 TeV recorded bythe ATLAS detectorat theLarge HadronCollider.

The distributionof the invariant mass of the two leading jets is examined for local excesses above a data-derived estimate ofthe smoothly falling prediction ofthe Standard Model. The data are also comparedtoaMonteCarlosimulationofStandardModelangulardistributionsderivedfromtherapidity of the two jets. No evidence of anomalous phenomena is observed in the data, which are used to exclude, at95% CL, quantum black holes with threshold massesbelow 8.3 TeV, 8.1 TeV, or5.1 TeV in three different benchmarkscenarios; resonance massesbelow 5.2 TeV forexcitedquarks, 2.6 TeV in aW^ model, arangeofmassesstarting frommZ^=¹.5 TeV andcouplingsfrom gq=⁰.2 in a Z^ model; andcontact interactionswithacompositenessscale below 12.0 TeV and17.5 TeV respectively for destructive and constructive interference between the new interaction and QCDprocesses. These results significantlyextendtheATLAS limitsobtainedfrom8 TeVdata. Gaussian-shapedcontributions to the mass distribution are also excluded if the effective cross-section exceedsvalues rangingfrom approximately50–300 fbformassesbelow2 TeV to2–20 fbformassesabove4 TeV.

©^{2016CERNforthebenefitoftheATLAS}Collaboration.PublishedbyElsevierB.V.Thisisanopen accessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

The centre-of-mass energyof proton–proton(pp) collisions at theLargeHadronCollider(LHC)atCERNhasbeenincreasedfrom

√s=^{8 TeV to}√

s=^{13 TeV,openinganewenergyregime toob-} servation.

NewparticlesproducedinLHCcollisionsmustinteractwiththe constituentpartonsoftheproton.Consequently,thenewparticles can alsoproduce partonsin thefinal state. Finalstates including partonsoftendominateinmodelsofnewphenomena beyondthe StandardModel (BSM).The partonsshower andhadronize,creat- ing collimated jets of particles carrying approximately the four- momenta of the partons. The total production rates for two-jet (dijet) BSMsignalscanbe large,allowing searchesforanomalous dijetproductiontotestforsuchsignalswitharelativelysmalldata sample,evenatmassesthatconstitute significantfractionsofthe totalhadroncollisionenergy.

IntheStandardModel(SM),hadroncollisionsproducejetpairs primarilyvia2→^{2 partonscatteringprocessesgovernedbyquan-} tum chromodynamics (QCD). Farabove the confinementscale of QCD (≈^{1 GeV),jets emergefrom collisions withlarge transverse}

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

momenta, pT, perpendicular tothe directionof the incident par- tons. Forthe dataanalysedhere, QCD predicts asmoothly falling dijetinvariant massdistribution,m_{j j}.Newstatesdecayingto two jetsmayintroducelocalized excessesinthisdistribution.In QCD, duetot-channelpolesinthecross-sectionsforthedominantscat- teringprocesses, mostdijetproductionoccursatsmallangles θ^∗, definedasthepolarangleinthedijetcentre-of-massframe.¹Many theoriesofBSMphysicspredictadditionaldijetproductionwitha significantpopulationofjetsproducedatlargeangleswithrespect tothebeam;forreviewsseeRefs.[1,2].Thesearchreportedinthis LetterexploitsthesegenericfeaturesofBSMsignalsinananalysis ofthemj j andangulardistributions.

As is common, a rapidity y=^ln((E+^pz)/(E−^pz))/2 is de- fined for each of the outgoing partons, where E is its energy and p_z isthecomponentofitsmomentumalong thebeamline.² Each incoming parton carries a fraction (Bjorken x) of the mo-

1 Since, experimentally,thetwo partonscannotbedistinguished,θ^∗ isalways takenbetween0andπ/2 withrespecttothebeam.

2 ATLAS usesaright-handedcoordinatesystemwithitsoriginatthe nominal interactionpoint(IP)inthecentreofthedetectorandthe z-axisalongthebeam line.Thex-axispoints fromtheIPtothecentreoftheLHCring,andthe y-axis pointsupwards.Cylindricalcoordinates(r,φ)areusedinthetransverseplane,φ beingtheazimuthalanglearoundthez-axis.Thepseudorapidityisdefinedinterms

http://dx.doi.org/10.1016/j.physletb.2016.01.032

0370-2693/©^{2016CERNforthebenefitoftheATLAS}Collaboration.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

mentumoftheproton.Amomentumimbalancebetweenthetwo partonsbooststhecentre-of-massframeofthecollisionrelativeto thelaboratoryframealong thez direction by yB=^ln(x1/x2)/2= (y₃+^y4)/2, where y_B is the rapidity of the boosted centre-of- mass frame, x₁ and x₂ are the fractions of the proton momen- tumcarried by each partonand y3 and y4 are the rapidities of the outgoing partons inthe detectorframe. Differences between tworapidities areinvariant undersuch Lorentzboosts, hencethe followingfunctionoftherapidity difference y^∗= (^y3−^y4)/2 be- tweenthetwojets,

χ=^e2|y∗|∼¹+^cosθ^∗ 1−^cosθ^∗,

isthesameinthedetectorframeasinthepartoniccentre-of-mass frame.Inthecentre-of-massframe,thetwopartonshaverapidity

±^y∗.

Thevariable χ isconstructedsuchthatinthelimitofmassless partonscattering, andwhenonlyt-channel scatteringcontributes to the partonic cross-section, the angular distribution dN/dχ is approximatelyindependentof χ.Themeasuredshapesoftheob- served dN/dχ distributions differ fromthe parton-level distribu- tionsbecausetheobserveddistributionsconvolvetheparton-level distributionswithnon-uniformpartonmomentumdistributionsin x1 and x2.Restrictingtherangeoftwo-partoninvariant massand placinganuppercuton y_B reducesthesedifferences.

Priorsearches of dijetdistributions with lower-energy hadron collisions at the SppS¯ [3–5], the Tevatron [6,7], and the LHC at

√s=^{7–8 TeV [8–19] and recently at 13 TeV [20], did not find} BSMphenomena. This Letter presents an analysis of 3.6 fb⁻¹ of proton–protoncollisionLHC dataat√

s=13 TeV recorded bythe ATLAS detector, focusing on thedistributions ofmj j and χ with methodsbasedonthoseusedbyRefs.[17,19].

2. TheATLASdetector

TheATLASexperiment[21]attheLHCisamulti-purposeparti- cledetectorwithaforward–backwardsymmetriccylindricalgeom- etrywithlayers oftracking,calorimeter,andmuondetectorsover nearlytheentiresolidanglearoundthepp collisionpoint.Thedi- rectionsandenergiesofhigh-pT hadronicjetsaremeasuredusing silicontrackingdetectorsandstraw tubesdetecting transitionra- diation,finelysegmentedhadronicandelectromagneticcalorime- ters,andamuon spectrometer.Asteel/scintillator-tile calorimeter provides hadronic energy measurements for the pseudorapidity range |η_|<1.7. A lead/liquid-argon (LAr) calorimeter provides electromagnetic(EM)energymeasurementswithhighergranular- itywithin theregion |η|<3.2.The end-capandforward regions areinstrumented withLArcalorimeters forEM andhadronicen- ergymeasurementsupto|η_|=4.9.Thefirst-leveltriggerisimple- mentedinhardwareandusesasubsetofthedetectorinformation to reduce the accepted rate to 100 kHz. This is followed by a software-basedtriggerthatreducestherateofeventsrecordedto 1 kHz.

3. Dataselection

Collisioneventsarerecordedusingatriggerrequiringthepres- enceofatleastonejetreconstructedinthesoftware-basedtrigger witha pT ofatleast 360 GeV.Groups ofcontiguous calorimeter cells(topologicalclusters)areformedbasedonthesignificanceof

ofthepolarangleθasη= −^{ln tan}(θ/2).Itisequivalenttotherapidityformassless particles.

the energy deposit over calorimeter noise [22]. Topological clus- tersaregroupedintojetsusingtheanti-kt algorithm[23,24]with radiusparameterR=⁰.4.Jetfour-momentaarecomputedbysum- mingoverthetopologicalclustersthatconstituteeachjet,treating the energyof each cluster asa four-momentum withzero mass.

The reconstruction efficiency for jets with pT above 20 GeV is 100%. Jet calibrations derived from √

s=13 TeV simulation, and collision datatakenat√

s=^{8 TeV and}√

s=^{13 TeV, areused to} correct the jet energies and directions to those of the particles from the hard-scatter interaction. This calibration procedure, de- scribedinRefs. [25–27],isimprovedby adata-derivedcorrection totherelativecalibrationofjetsinthecentralandtheforwardre- gions. The dijet mass resolution is2.4% and 2%, for dijetmasses of2and5 TeVrespectively.Thejetenergyscaleuncertaintyfrom 8 TeV data iscomplementedby systematicuncertaintiescovering thedifferencesbetween8 TeV and 13 TeV data. Thetotal jeten- ergy scale uncertaintyis 1% forcentral jetswith pT of 500 GeV, and3%forjetsof 2 TeV. Analysisofjet dataat13 TeV using the insitu techniques described in Ref. [28] confirms the jet calibra- tionanduncertaintyestimates.Beyondthe p_T rangeoftheinsitu techniques,forthequantitiesusedtocalibratejetsaswellasother kinematicquantities,thedataagreewithsimulationwithinquoted uncertainties.

Eventscontainingatleasttwojetsareselectedforofflineanal- ysis if the pT of the leading andsubleading jets is greater than 440 GeV and 50 GeV respectively. This requirement ensures a trigger efficiency of at least 99.5% for collisions with |^y∗|<1.7 andremoves a negligible number ofevents fromunbalanced di- jeteventsoriginatingfromadditionalinteractionswithinthesame bunchcrossingorjetresolutiontails.Eventsarediscardedfromthe search ifanyofthe threeleading jetswith pT>50 GeV is com- patiblewithnon-collisionbackgroundorcalorimeternoise[29].

4. Simulatedcollisions

Forthissearch,eventsfromQCDprocessesaresimulatedwith Pythia 8[30] usingthe A14[31] setoftuned parameters forthe underlyingeventandtheleading-orderNNPDF2.3[32] partondis- tribution functions (PDFs). The renormalization and factorization scales are setto theaverage pT of thetwo leading jets.Detector effects are simulated using Geant4 [33] within the ATLAS soft- ware infrastructure [34]. The same software used to reconstruct datawasalsousedtoreconstructsimulatedevents.Thesimulated eventsareusedtopredicttheangulardistributionfromQCDpro- cesses and forqualitative comparisons to kinematic distributions indata.

Pythia 8calculationsusematrixelementsthatareatleadingor- derintheQCDcouplingconstantwithsimulationofhigher-order contributions partially covered by the parton shower (PS) mod- elling. They also include modelling of hadronization effects. The distributions of events predicted by Pythia 8 are reweighted to the next-to-leading-order (NLO) predictions of NLOJET++ [35–37]

using mass- and χ-dependent correction factors defined as in Ref. [19].The correction factors modify the shape ofthe angular distributions atthelevelof15% atlowvaluesof χ andhighval- ues ofm_{j j}. The correction is 5% orless at the highestvalues of

χ. The Pythia 8 predictions also omit electroweak effects. These areincludedasadditionalmass- and χ-dependentcorrectionfac- tors[38] thatareunityatlowm_{j j} anddifferfromunitybyup to 3%inthemj j>3.4 TeV region.

BSMsignalsamplesofexcitedquarks[39,40],newheavyvector bosons [41–43], quantum black holes [44–46] andcontact inter- actions [47–49] are simulated andreconstructed using the same procedure asfor QCD processes.The models andthe parameters chosenforgenerationaredescribedinSection7.

Fig. 1. The reconstructed dijetmass distribution(filled points) for eventswith

|y^∗|<0.6 and pT>440(50)GeV fortheleading (subleading)jets.Thesolidline depictsthefittoEq.(1),asdiscussedinthetext.Predictionsforanexcitedquark andaquantumblackholesignalpredictedbythe BlackMax generator(QBHBM) areshownabovethefit,normalizedtothepredictedcross-section.Theverticallines indicatethemostdiscrepantintervalidentifiedbythe BumpHunter algorithm,for whichthe p-valueisstatedinthefigure.Themiddlepanelshowsthebin-by-bin significancesofthe data-fitdifferences,considering onlystatistical uncertainties.

Thelowerpanelshowstherelativedifferencesbetweenthedataandthepredic- tionof Pythia 8 simulationofQCDprocesses,correctedforNLOandelectroweak effects,andisshownpurelyforcomparison.Theshadedbanddenotestheexperi- mentaluncertaintyinthejetenergyscalecalibration.

5. Selectionforthemassdistributionanalysis

Them_{j j} distributionofeventswith|^y∗|<0.6 (χ<3.3)isanal- ysedforevidenceofcontributionsfromresonantBSMphenomena.

The requirementon |^y∗| ^{reducesthe backgroundfrom QCD pro-} cesses. To avoid kinematic bias from the y^∗ and p_T selections describedabove,theanalysisisconfinedtomj j>1.1 TeV.

Fig. 1 shows the observed mj j distribution for the resonance selection,overlaid withexamplesof thesignalsdescribed inSec- tion 7. The bin widths are chosen to approximate the m_{j j} res- olution as derived from the simulation of QCD processes, and therefore widen as the mass increases. The largest value of m_{j j} measuredis6.9 TeV.

ToestimatetheSMbackground,theansatz,

f(z)=^p1(1−^z)^p2z^p3, (1) wherez≡^mj j/√

s,isfittothemj j distributioninFig. 1toobtain theparameters pi.Thefitrangeis1.1–7.1 TeV.CDF,CMS,andAT- LAS dijetsearches such as those described in Refs. [6,8,13,14,17]

havefoundthat expressionssimilarto Eq.(1)describe dijetmass distributionsobserved atlower collisionenergies.The ansatzalso describes leading-order and next-to-leading order simulations of QCDdijetproductionat√

s=^{13 TeV.A}log-likelihood-ratiostatis- ticemploying Wilks’stheorem[50] was used todetermineifthe backgroundestimationwouldbesignificantly improvedby anad- ditionaldegreeof freedom.Withthe currentdataset,Eq.(1)was foundtobesufficient.

Fig. 1alsoshowstheresultofthefit.Thefitdescribestheob- serveddatawitha p-valueof0.87,usingaPoissonlikelihoodtest

statistic.Themiddlepanelofthefigureshowsthesignificancesof bin-by-bin differences betweenthe dataandthe fit.These Gaus- siansignificancesarecalculatedfromthePoissonprobability,con- sidering only statistical uncertainties. The lower panel compares the data to the prediction of Pythia 8 simulation of QCD pro- cesses, correctedforNLO andelectroweakeffects.Eventhough it is notusedinthe analysisofthem_{j j} distribution,thesimulation isshowntobeingoodagreementwiththedata.

TheuncertaintyinvaluesoftheparametersinEq.(1)isevalu- atedbyfittingthemtopseudo-datadrawnviaPoissonfluctuations around the fittedbackground model.The uncertainty inthe pre- dictionineachmj j binistakentobetherootmeansquare ofthe functionvalue forall pseudo-experimentsinthatbin.Toestimate an uncertainty due to the choice of the background parameteri- zation,aparameterizationwithoneadditionaldegreeoffreedom, z^{p4log z}, iscompared tothe nominalansatz, andthedifference is takenasanuncertainty.Thepredictionofthem_{j j}distributiondoes not involvesimulatedcollisions andthusisnot affectedby theo- reticalorexperimentaluncertainties.

The statistical significance of any localized excess in the m_{j j} distributionisquantifiedusingthe BumpHunter algorithm[51,52].

Thealgorithmcomparesthebinnedmj j distributionofthedatato thefittedbackgroundestimate,consideringcontiguousmassinter- valsinallpossiblelocations,fromawidthoftwobinstoawidth of half ofthe distribution. Foreach interval inthe scan, it com- putesthesignificanceofanyexcessfound.Thealgorithmidentifies the interval 1.53–1.61 TeV,indicated by thetwo vertical linesin Fig. 1,asthe mostdiscrepantinterval. The statisticalsignificance of thisoutcomeis evaluated usingtheensemble ofpossibleout- comes across all intervalsscanned, by applying the algorithm to manypseudo-datasamplesdrawnrandomlyfromthebackground fit.Withoutincludingsystematicuncertainties,theprobabilitythat fluctuationsofthebackgroundmodelwouldproduceanexcessat least assignificant astheone observed inthe data,anywhere in the distribution,is 0.67.Thus, there isnoevidence ofa localized contributiontothemassdistributionfromBSMphenomena.

6. Selectionfortheangulardistributionsanalysis

The dN/dχ (angular) distributions of events with |^y∗|<1.7 (i.e. χ <30.0) and|^yB|<1.1 are alsoanalysed for contributions from BSM signals. Fig. 2 shows the angular distributions of the data in different m_{j j} ranges, the SM prediction for the shape of theangulardistributions,andexamplesofthesignalsdescribedin Section 7.The data with mj j<2.5 TeV are discarded to remove bias from the kinematic selections described earlier. The highest mj j measuredis7.9 TeV.TheSMpredictionisobtainedfromsim- ulation,asdescribedinSection4.Intheanalysis,thepredictionin eachm_{j j} rangeisnormalizedtomatchtheintegralofthedatain thatrange.

Theoreticaluncertaintiesinsimulationsoftheangulardistribu- tions fromQCDprocessesare estimatedasdescribed inRef. [19].

The effect on the QCD prediction of varying the PDFs is esti- mated using NLOJET++ withthree differentPDF sets: CT10 [53], MSTW2008[54] andNNPDF23 [32]. Asthe choice ofPDF largely affectsthetotalcross-sectionratherthantheshapeofthe χdistri- butions, theseuncertainties arenegligible(<1%).Theuncertainty dueto thechoice ofrenormalizationandfactorizationscaleswas estimatedusing NLOJET++ by varyingeach independentlyup and downbyafactortwo,excludingoppositevariations.Theresulting uncertainty,takenastheenvelopeofthevariationsinthenormal- ized χ distributions,dependsonbothm_{j j} and χ,risingto20%at thesmallest χvaluesathighmj jvalues.Thestatisticaluncertainty ofthesimulatedNLOcorrectionsislessthan1%.Thedominantex- perimentaluncertaintyinthepredictionsofthe χ distributionsis