Search for heavy long-lived charged R-hadrons with the ATLAS detector in 3.2 fb$^{-1}$ of proton–proton collision data at $\sqrt{s}=13$ TeV

Contents lists available atScienceDirect

Physics Letters B

www.elsevier.com/locate/physletb

Search for heavy long-lived charged R-hadrons with the ATLAS detector in 3.2 fb

of proton–proton collision data at √

s = ^{13 TeV}

.TheATLASCollaboration^

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

Articlehistory:

Received17June2016

Receivedinrevisedform11July2016 Accepted15July2016

Availableonline19July2016 Editor:W.-D.Schlatter

A search for heavy long-lived charged R-hadrons is reported using adata sample corresponding to 3.2 fb⁻¹ ofproton–proton collisions at√

s=13 TeV collected bythe ATLASexperiment atthe Large HadronCollideratCERN.Thesearchisbasedonobservablesrelatedtolargeionisationlossesandslow propagationvelocities,whicharesignaturesofheavychargedparticlestravellingsignificantlyslowerthan thespeedoflight.Nosignificantdeviationsfromtheexpectedbackgroundareobserved.Upperlimitsat 95%confidencelevelareprovidedontheproductioncrosssectionoflong-lived R-hadrons inthemass rangefrom600 GeV to2000 GeV andgluino,bottomandtopsquarkmassesareexcludedupto1580 GeV, 805 GeV and890 GeV,respectively.

©^{2016TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense} (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP³.

1. Introduction

Heavylong-livedparticles(LLP)arepredictedinarangeofthe- oriesextendingtheStandardModel(SM)inanattempttoaddress the hierarchy problem [1]. These theories include supersymme- try(SUSY)[2–7],whichallowsforlong-livedchargedsleptons(˜), squarks(q),˜ gluinos(g)˜ andcharginos (χ˜₁^±) inmodelsthateither violate[8–10]orconserve[11–17]R-parity.

Heavylong-livedchargedparticlescanbeproducedattheLarge Hadron Collider (LHC). A search for composite colourless states ofsquarks or gluinos together withSM quarks or gluons, called R-hadrons[11],ispresentedinthisLetter.Thesearchexploitsthe fact that theseparticles are expected to propagate witha veloc- ity,β=^v/c,substantiallylowerthanoneandtoexhibit aspecific ionisationenergyloss,dE/dx,largerthanthatforanychargedSM particle.Similar searches havebeen performedpreviously by the ATLAS and CMS Collaborations [18,19] using data samples from Run 1attheLHC.Noexcessesofeventsabovetheexpectedback- grounds were observed, and lower mass limits were set at 95%

confidencelevel(CL)around1300 GeV forgluinoR-hadrons.

R-hadrons can be produced in pp collision as eithercharged or neutral states, and can be modified to a state with different charge by interactions with the detector material [20,21], arriv- ing asneutral, chargedor doubly charged particles in the muon spectrometer(MS)oftheATLASdetector.Thissearchdoesnotuse informationfromtheMS andfollowsthe“MS-agnostic” R-hadron search approach in Ref. [18]. This strategy avoids assumptions

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

about R-hadron interactions with the detector, especially in the calorimeters,andissensitivetoscenarios inwhich R-hadrons de- cayorbecomeneutral(viapartonexchangewiththedetectorma- terialinhadronicinteractions)beforereachingtheMS.

2. ATLASdetector

TheATLASdetector[22]isamulti-purposeparticle-physicsde- tector consisting of an inner detector (ID) immersed in an ax- ial magnetic field to reconstructtrajectories of chargedparticles, calorimeterstomeasuretheenergyofparticlesthatinteractelec- tromagneticallyor hadronically anda MS within a toroidal mag- netic fieldto providetrackingformuons. Withnear4π coverage in solid angle,¹ the ATLAS detector is able to deduce the miss- ing transversemomentum, p^miss_T ,associated witheach event.The componentsofparticular importanceto thissearchare described inmoredetailbelow.

The ID consists of two distinct silicon detectors and a straw tracker,whichjointlyprovidegoodmomentummeasurements for charged tracks. The innermost partof the ID, a silicon pixel de- tector,typicallyprovidesfourormoreprecisionmeasurementsfor each track in the region |η_|<2.5 at radial distances 3.4<r<

13 cm fromtheLHCbeamline.Allpixellayers aresimilar,except

1 ATLASusesaright-handedcoordinatesystemwithitsoriginatthenominalin- teractionpoint(IP)inthecentreofthedetectorandthez-axiscoincidingwiththe axisofthebeampipe.Thex-axispointsfromtheIPtothecentreoftheLHCring, andthe y-axispointsupward.Cylindricalcoordinates(r,φ)areusedinthetrans- verseplane,φbeingtheazimuthalanglearoundthebeampipe.Thepseudorapidity isdefinedintermsofthepolarangleθasη= −ln tan(θ/2).

http://dx.doi.org/10.1016/j.physletb.2016.07.042

0370-2693/©^{2016TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense}(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP³.

Fig. 1. ResolutionofβfordifferentcellsintheATLAStilecalorimeterobtainedfromaZ→μμselectionindata.Thefinalβmeasurementisaweightedaverageoftheβ measurementsinthecellstraversedbythecandidate.

the innermost Insertable B-Layer(IBL) [23],which has a smaller pixelsizeanda reducedthickness,butalso4-bitinsteadof 8-bit encodingandhencepoorerchargeresolutionthantheotherpixel layers. The charge released by the passage of a charged particle israrelycontainedwithin asingle siliconpixel,andaneuralnet- workalgorithm[24]isusedtoformclustersfromthesinglepixel charges. Foreach cluster in the pixel detectora dE/dx estimate can be provided, from which an overall dE/dx measurement is calculated as a truncated mean to reduce the effect of the tail oftheLandaudistribution,bydisregardingtheoneortwolargest measurements[25].RadiationsensitivityoftheIBLelectronicsre- sults in the measured dE/dx drifting over time. This effect is correctedby applyinga dedicatedtime-dependent ionisationcor- rectionof1.2%onaverage.ThemeanandRMSofthedE/dx mea- surement fora minimum-ionisingparticle are 1.12 MeV g⁻¹cm² and 0.13 MeV g⁻¹cm², respectively, while the distribution ex- tendstohigherdE/dx values,duetotheremnants ofthe Landau tail.

TheATLAScalorimeterinthecentraldetectorregionconsistsof anelectromagneticliquid-argoncalorimeterfollowedbyahadronic tile calorimeter. The estimation of β from time-of-flight mea- surements relies on timing and distance information from tile- calorimeter cells crossed by the extrapolated candidate track in threeradiallayersinthecentralbarrelaswellasanextendedbar- reloneachside,asillustratedinFig. 1.Toreduceeffectsofdetec- tornoise,onlycellsinwhichtheassociatedparticlehasdeposited aminimumenergyE_min,cell=500 MeV aretakenintoaccount.The time resolutiondepends onthe energydeposited inthe cell and alsothelayertypeandthicknessofthecell.

Aseriesofcalibrationtechniquesisappliedtoachieve optimal performance,usinga Z→μμcontrol sample.Muonsonaverage depositslightlylessenergythan expectedfromsignal, butvaria- tions sufficiently cover the relevant range. First, a common time shiftisappliedforeachshortperiodofdatataking(run)followed byfiveadditionalcell-by-cellβ corrections.A geometry-basedcell correctionisintroducedtominimisethe ηdependenceofβwithin eachindividualcell.Thisisdonebytakingintoaccounttheactual trajectory (η and path length) of the extrapolated track in each calorimetercell,torecalculatethedistance-of-flight,insteadofus- ingthecentreofthecell,asdoneinpreviousATLASsearches(e.g.

in[18]).The effectismostprominentattheedges ofthelargest cells at high|η_|with shiftsof up to 0.05 in β,andalmost neg- ligiblefor thecells at low |η|^{.An additional}correction, linearin

|η|^{andonlyappliedin}simulation,isaddedtoaccount foratim- ingmismodellingduetoanimperfectsimulation.Thiscorrectionis againmostprominentforthecellsathigh|η_|withshiftsupto0.1 inβ.TheOptimalFilteringAlgorithm(OFA)[26]usedfortheread- outofthetilecalorimetercellsisoptimisedforin-timesignalsand introduces abiastowards lowervaluesof β inthe measuredcell

Fig. 2. Distributionsofβ fordataandsimulationaftera Z→μμselection.The valuesgivenforthemeanandwidtharetakenfromGaussianfunctionsmatchedto dataandsimulation.

timeoflate-arrivingparticles.Tocompensateforthisbiasforlate- arrivingparticles,a correctionisestimatedfromafittosimulated latesignals.Celltimeslargerthan25 nsarediscarded,tolimitthe sizeoftherequiredcorrection.Thesizeofthecorrectionisup to 0.05 in β. A cell-timesmearingis appliedto adjustthecell-time resolutioninsimulationtothat observedindata.Theuncertainty inthesingleβ measurementsisscaledupbyabout12%,basedon therequirementthatthepulldistribution(β− βtrue)/σ_β beaunit Gaussian.Finallytheβ associatedwiththeparticleisestimatedas a weighted average, usingthe β measurement in each traversed cellanditsuncertainty, σβ.

Afterallcalibrations,thesinglecell-timeresolutionrangesfrom 1.3 ns incells atlarge radii to 2.5 nsin cells atsmall radii. The distancesfromthenominalinteractionpoint(IP)tothecellcentres are2.4 mto3.6 m(4.2 mto5.7 m)at|η_|∼0 (|η_|∼1.25).Thisin turnresultsinaresolutionof0.06to0.23inβ,asshowninFig. 1.

The larger cells atlargeradii havea better resolutiondueto the higherenergydepositsandtheirincreaseddistancefromthe IP.

As described in Section 5,theexpected β distribution forthe backgroundis determinedfromdata.However, theβ distribution forthe R-hadron signalisobtainedfromsimulation.Fig. 2shows the β distributions obtained for both data and simulation for a control sample of Z→μμ events that isused to validate theβ measurement.Goodagreementbetweendataandsimulationsup- portstheuseofthesimulationtopredictthebehaviourexpected forthe R-hadron signal.

Table 1

FinalselectionrequirementsasafunctionofthesimulatedR-hadron mass.

Simulated R-hadron mass [GeV]

600 800 1000 1200 1400 1600 1800 2000

βγ^max 1.35 1.35 1.35 1.35 1.35 1.15 1.15 1.15

β^max 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75

m^min_βγ 350 450 500 575 650 675 750 775

m^min_β 350 450 500 575 650 675 750 775

3. Dataandsimulatedevents

The work presented in this Letter is based on 3.2 fb⁻¹ of pp collision data collected in 2015 at a centre-of-mass energy

√s=^{13 TeV.} Reconstructed Z →μμ eventsin dataandsimula- tionareusedfortimingresolutionstudies.Simulatedsignalevents areusedtostudytheexpectedsignalbehaviour.

R-hadron signal events are generated with gluino (bottom- squark and top-squark) masses from 600 GeV to 2000 GeV (600 GeV to1400 GeV).Pairproductionofgluinosandsquarksis simulatedin Pythia 6.427[27]withtheAUET2B[28] setoftuned parametersfortheunderlyingeventandtheCTEQ6L1[29] parton distributionfunction (PDF) set, incorporating Pythia-specificspe- cialised hadronisation routines [20,30,31] to produce final states containing R-hadrons. ThemassesoftheotherSUSY particlesare settoveryhighvaluestoensurethattheircontributiontothepro- ductioncross sectionis negligible.Fora givensparticle massthe productioncross section forgluino R-hadrons is typically an or- der of magnitude higherthan forbottom-squark and top-squark R-hadrons. The probability for a gluino to form a gluon-gluino bound state is assumed, based on a colour-octet model, to be 10%[12].Theassociatedhadronicactivityproducedbythecolour field of the sparticle typically only possesses a small fraction of theinitialenergyofthesparticle [12],whichshould thereforebe reasonablyisolated.

To achieve a more accurate description of QCD radiative ef- fects, the Pythia eventsare reweighted to matchthe transverse- momentumdistributionofthegluino–gluinoorsquark–squarksys- temto that obtainedin dedicated MG5_aMC@NLOv2.2.3.p0 [32]

events,asMG5_aMC@NLOcanproduceadditionalQCDinitial-state radiation(ISR)jetsaspartofthehardprocess, while Pythia only includesshoweringtoaddjetstotheevent.

Allevents passthrough a full detectorsimulation[33],where interactions with matter are handled by dedicated Geant4 [34]

routines based on different scattering models: the model used to describe gluino (squark) R-hadron interactions is referred to as the generic (Regge) model [21]. The R-hadrons interact only moderatelywith the detector material, asmost of the R-hadron momentum is carried by the heavy gluino orsquark, which has littleinteractioncrosssection.Typically,theenergydepositinthe calorimetersislessthan10 GeV.

Allsimulatedeventsincludeamodellingofcontributionsfrom pile-up by overlaying minimum-bias pp interactions from the same (in-time pile-up) and nearby (out-of-time pile-up) bunch crossings,andarereconstructedusingthesamesoftwareusedfor collisiondata.Simulatedeventsarereweightedso thatthedistri- bution of the expected number of collisions per bunch crossing matchesthatofthedata.

4. Eventselection

Eventsareselectedonlineviaatriggerbasedonthemagnitude of the missing transverse momentum, E^miss_T . Large E^miss_T values areproducedmainlywhenQCD initial-stateradiation(ISR)boosts

the R-hadron system, resultinginan imbalancebetweenISR and R-hadrons whosemomentaarenotfullyaccountedforintheE^miss_T calculation.In particular,theadoptedtrigger imposesa threshold of70 GeV on E^miss_T calculatedsolely fromenergydepositsinthe calorimeters [35].The signal efficiencyof the E^miss_T triggervaries between 32% and 50%, depending on the mass and type of the R-hadron.

The offline eventselection requires all relevant detectorcom- ponents to befullyoperational; a primary vertex(PV)builtfrom atleasttwo well-reconstructedcharged-particle tracks,each with a transverse momentum, pT, above 400 MeV; and at least one R-hadron candidatetrackthatmeetsthecriteriaspecifiedbelow.

R-hadron candidatesare basedonIDtrackswithpT>50 GeV and |η_| < 1.65. Candidates must not be within R =

(η)²+ (φ)²=⁰.3 ofanyjetwithpT>50 GeV,reconstructed using theanti-kt jet algorithm [36] withradius parameter set to 0.4. Furthermore, the candidates must not have any additional nearby(R<0.2) trackswith pT>10 GeV.Tracks reconstructed with p>6.5 TeV are rejected as unphysical. To ensure a well reconstructedtrack, a minimum numberofsevenhitsinthe sili- con detectors isrequired. Ofthese, atleast two clustersused to measure dE/dx in the pixel detector are required, to ensure a good dE/dx measurement.Candidateswith|^zPV0 sin(θ )|>0.5 mm or |^d0|>2.0 mm are removed, where d₀ is the transverse im- pactparameter atthecandidate track’spointof closestapproach to the IP and z^PV₀ is the z coordinate of this point relative to the PV. To suppress background muons stemming from cosmic- rayinteractions,candidateswithdirection(η,φ)arerejectedifan oppositely-charged trackwithalmost specular direction, i.e.with

|η_|<0.005 and |φ|<0.005 with respect to (−η, π_{− φ)}, is identifiedontheoppositesideofthedetector.Inordertominimise the background from Z→μμ decays, candidates are rejected if they resultin an invariant masscloser than10 GeV to the mass ofthe Z bosonwhencombinedwiththehighest-p_T muoncandi- dateintheevent.Inadditiontotheabovementionedtrack-quality criteria,candidatesmustalsosatisfyobservable-qualitycriteria,de- finedbyanunambiguousβγ determinationfromthedE/dx value, estimatedusingan empirical relation(more details canbe found in Ref. [37]), determined fromlow-momentum pions, kaons and protons[37],andaβ measurement,withanuncertainty σβofless than 0.12. In the following, βγ refers to quantities derived from thedE/dx measurementinthesiliconpixeldetectorandβ refers tothetime-of-flight-basedmeasurementinthetilecalorimeter.

Afterthisinitialselection,226 107oftheapproximately36 mil- lion initially triggered dataeventsas well as10% to15% of sim- ulated signal events (the percentage increaseswith hypothesised mass) remain. Only thecandidate withthe largest p_T is used in eventswithmultiple R-hadron candidates.The final signalselec- tion, requiring a momentum above 200 GeV as well as criteria summarisedinTable 1,isbasedonβγ andβ,requiringβγ<1.35 (<1.15) for R-hadron masses up to (greater than) 1.4 TeV and β <0.75 inall cases.Thesignal regionisdefinedinthem_βγ –mβ plane foreach R-hadron masspoint,wheremβγ andmβ are ex- tracted independently from the measurement of the momentum

Fig. 3. Data(blackdots)andbackgroundestimates(redsolidline)formβ(left)andmβγ (right)forthegluinoR-hadron search(1000 GeV).Thegreenshadedbandillustrates thestatisticaluncertaintyofthebackgroundestimate.Thebluedashedlinesillustratetheexpectedsignal(ontopofbackground)forthegivenR-hadron masshypothesis.

Theblackdashedverticallinesat500 GeV showthemassselectionandthelastbinincludesallentries/massesabove.(Forinterpretationofthereferencestocolourinthis figurelegend,thereaderisreferredtothewebversionofthisarticle.)

aswell as βγ andβ,respectively, via m=^p/βγ. The minimum mass requirements, m^min_βγ and m^min_β , are set to correspond to a value about 2σ below the nominal R-hadron mass value, given themassresolutionexpectedforthesignal.

Thetotalselectionefficiencydependsonthesparticlemassand variesbetween9% and15% forgluino andtop-squark R-hadrons and6% to 8% forbottom-squark R-hadrons. The lower efficiency forbottomsquarksisexpected, asR-hadrons aremostlikelypro- duced in mesonic states, where those with down-type squarks tend to be neutral more often than those withup-type squarks, due to light-quark production ratios of u:^d:^s≈¹:¹:⁰.3 [12]

duringhadronisation.The expectedsignal yieldandefficiency,es- timated background and observed number of events in data for the full mass range after the final selection are summarised in Table 3.

5. Backgroundestimation

The background is evaluated in a data-driven manner. First, probabilitydistributionfunctions(pdf)inthemomentum,andalso in the β and βγ values, are determined from data. These pdfs areproducedfromcandidatesindata,whichhavepassedtheini- tialselectionmentionedearlier,butfallinsidebandsofthesignal region, as described below. Background distributions in mβ and m_βγ are obtainedby randomly samplingthe pdfs derived above andthenusingtheequationm=^p/βγ.Thesemassdistributions, whicharenormalisedtothedataeventsoutsidethesignalregion (i.e.notpassingbothmassrequirementsofthehypothesisinques- tion),areshowninFig. 3alongwiththedataandexpectedsignal forthe1000 GeV gluino R-hadron masshypothesis.

Each R-hadron mass hypothesis hasa different selection, and thereforecorresponding individual backgroundestimatesare pro- duced accordingly. The momentum pdf is produced fromevents thatpassthemomentumcut,butfailtheβ andβγ requirements inTable 1forthechosen R-hadron masshypothesis,butnonethe- lesshaveβ <1 andβγ <2.5. Theβ andβγ pdfs are produced by selecting events which pass the respective β and βγ selec- tion and have momentum in the range 50 GeV<p<200 GeV.

Since momentum iscorrelated with|η_|,any correlation between

|η|^andβ (βγ)willleadtoacorrelationbetweenmomentumand

Table 2

Summaryofallstudiedsystematicuncertainties.Rangesindicateadependencyon theR-hadron masshypothesis(fromlowtohighmasses).

Source Relativeuncertainty

[±^%]

Theoretical uncertainty on signal 14–57

Uncertainty on signal efficiency 20–16

Trigger efficiency 2

QCD uncertainty (ISR, FSR) 14

Pile-up 7–1

Pixelβγmeasurement 1–3

Calorimeterβmeasurement 10–2

Luminosity 5

Uncertainties on background estimate 30–43

β (βγ),invalidatingthebackgroundestimate.Thesizeandimpact ofsuchcorrelationsarereducedbydeterminingthethreepdfs in fiveequal-widthbinsof|η|^{.Thisprocedurealsoensures thatdif-} ferentdetectorregionsaretreatedseparately.

6. Systematicuncertainties

Thesystematicuncertaintiesareobtainedfromdata,whenever possible.Thetwomajoruncertaintiesforwhichthisisnotthecase arecrosssectionsandISR,thelatterbeingfoldedwiththetrigger efficiency curve obtainedfrom data to produce the overall E^miss_T trigger efficiency.Theindividual contributions areoutlined below andsummarisedinTable 2.

6.1. Theoreticalcrosssections

Signal crosssectionsarecalculated tonext-to-leadingorderin the strong coupling constant, including the resummation of soft gluon emission at next-to-leading-logarithmic accuracy (NLO + NLL) [38–40].The nominal crosssection andthe uncertainty are taken froman envelopeof cross-section predictions using differ- ent PDF sets andfactorisation andrenormalisation scales, asde- scribed inRef. [41]. Thisprescriptionresults inan uncertaintyof 14% (at 600 GeV) rising to 24% (at 1600 GeV) and to 32% (at 2000 GeV) for gluino R-hadrons andmarginally larger valuesfor squark R-hadrons.