Article
Reference
Search for low-mass resonances decaying into two jets and produced in association with a photon using pp collisions at √s = 13 TeV with
the ATLAS detector
ATLAS Collaboration
ADORNI BRACCESI CHIASSI, Sofia (Collab.), et al .
Abstract
A search is performed for localised excesses in dijet mass distributions of low-dijet-mass events produced in association with a high transverse energy photon. The search uses up to 79.8 fb −1 of LHC proton–proton collisions collected by the ATLAS experiment at a centre-of-mass energy of 13 TeV during 2015–2017. Two variants are presented: one which makes no jet flavour requirements and one which requires both jets to be tagged as b -jets.
The observed mass distributions are consistent with multi-jet processes in the Standard Model. The data are used to set upper limits on the production cross-section for a benchmark Z′ model and, separately, on generic Gaussian-shape contributions to the mass distributions, extending the current ATLAS constraints on dijet resonances to the mass range between 225 and 1100 GeV.
ATLAS Collaboration, ADORNI BRACCESI CHIASSI, Sofia (Collab.), et al . Search for
low-mass resonances decaying into two jets and produced in association with a photon using pp collisions at √s = 13 TeV with the ATLAS detector. Physics Letters. B , 2019, vol. 795, p. 56-75
DOI : 10.1016/j.physletb.2019.03.067
Available at:
http://archive-ouverte.unige.ch/unige:120306
Disclaimer: layout of this document may differ from the published version.
1 / 1
Contents lists available atScienceDirect
Physics Letters B
www.elsevier.com/locate/physletb
Search for low-mass resonances decaying into two jets and produced in association with a photon using pp collisions at √
s = 13 TeV with the ATLAS detector
.TheATLAS Collaboration
a r t i c l e i n f o a b s t ra c t
Articlehistory:
Received31January2019
Receivedinrevisedform25March2019 Accepted26March2019
Availableonline30May2019 Editor: M.Doser
Asearchisperformedforlocalisedexcessesindijetmassdistributionsoflow-dijet-masseventsproduced inassociation withahightransverseenergyphoton.The searchusesupto79.8 fb−1 ofLHCproton–
proton collisions collected by the ATLAS experiment at a centre-of-mass energy of 13 TeV during 2015–2017.Two variantsare presented:one whichmakesnojetflavourrequirementsand onewhich requires both jetsto betagged as b-jets.The observed mass distributions are consistentwith multi- jet processesin theStandard Model. Thedata are used toset upper limits onthe productioncross- sectionforabenchmarkZ modeland,separately,ongenericGaussian-shapecontributionstothemass distributions, extendingthe currentATLAS constraintsondijetresonancestothe massrangebetween 225and1100 GeV.
©2019TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.
1. Introduction
Searchesforresonantenhancementsofthedijetinvariantmass distribution (mjj) are an essential part of the LHC physics pro- gramme.Newparticleswithsizeablecouplingstoquarksandglu- ons arepredictedby manymodels,such asthoseincludingreso- nanceswithadditionalcouplingstodark-matterparticles [1,2].
Searchesfordijetresonanceswithmassesofseveralhundreds ofGeVtojustabove1 TeVhavebeencarriedout atlower-energy colliders [3–7] and at the LHC, which has also extended search sensitivities into the multi-TeV mass range [8–22]. Despite us- inghigherintegratedluminositiesthanearliercolliders,theseLHC searches have beenlimited at lower masses by a large multi-jet background.Multi-jet eventsareproduced atsuch highratesthat fullyrecordingevery event wouldsaturatethe onlinedataselec- tion (called trigger) and data acquisition systems. To avoid this, minimumtransversemomentum(pminT )thresholdsareimposedon triggers collecting events with at least one jet (called single-jet triggers).Thesethresholdscreatealowerboundonthesensitivity of searches ata mass of approximatelymjj≈2pminT , where pminT is typically several hundred GeV. Consequently, searches for di- jetresonances attheLHC havepoorsensitivityformassesbelow 1 TeV,andsetlimitsonthecouplingsoftheresonancetoquarks in this light-resonance region which are weaker than limits in heavy-resonance regions [23]. Nevertheless, despite the difficulty
E-mailaddress:atlas.publications@cern.ch.
ofrecordingeventscontaininglightresonances,they remainavi- ablesearchtargetattheLHC,bothfromamodel-agnosticpointof view [24] and,forexample,inmodelsofspin-dependent interac- tionsofquarkswithdarkmatter [1,2].
Recently,ATLASandCMShavepublishedsearchesforlow-mass dijet resonances using severalcomplementary strategies to avoid triggerlimitations.Formjj>450 GeV,themoststringentlimitsare setbysearchesrecordingonlypartialeventinformation [20,21].
Anothersearchavenueisopenedbydatainwhichalightres- onance is boostedin thetransverse directionvia recoil against a high-pT photon [25,26]. Requiring a high-pT photon in the final state reduces signal acceptance but allows efficient recording of events withlowerdijetmasses. At evenlower resonance masses, thedecayproductsoftheresonancewillmergeintoasinglelarge- radius jet. Searches for this event signature have been used to set limits on resonant dijet production at both ATLAS [27] and CMS [28,29].However,thesesearchesbecomelesssensitiveabove 200 GeV–350 GeV,whenthedecayproductsfalloutsidethelarge- radiusjetcone.
This Letter presents a new search for resonances in events containing a dijet and a high-pT photon in the final state, us- ing proton–proton (pp) collisions recorded at a centre-of-mass energy √
s=13 TeV and corresponding to an integrated lumi- nosity up to 79.8 fb−1. The search targets a dijetmass range of 225 GeV–1.1 TeV. This rangecovers massesbelow the range ac- cessible using single-jet triggers or partial-event data and above the mass range where theresonance decay products merge. The search is performedusingsamplesofeventsselected eitherwith https://doi.org/10.1016/j.physletb.2019.03.067
0370-2693/©2019TheAuthor.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP3.
orwithoutcriteriadesignedtoidentifyjetsoriginatingfrombot- tom quarks (b-jets). Searching in a subset of the data selected withb-jetidentificationcriteriaenhancessensitivitytoresonances whichpreferentially decayintobottomquarks.Thissearch probes masses above 225 GeV, obtaining results complementary to the reachofpreviousdijetsearchesatacentre-of-massenergyof√
s= 13 TeV: below approximately 600 GeV, previous ATLAS di-b-jet searcheslosesensitivity [30],whiletherangeoftheCMSboosted di-b-jet search [29] is limited to a mass region up to 350 GeV.
Anothercomplementary CMSsearch for resonances with masses above 325 GeV decaying to b-jets at a centre-of-mass energyof
√s=8 TeVisdescribedinRef. [31].
2. ATLASdetector
The ATLAS experiment [32–35] at the LHC is a multipurpose particle detectorwith a forward–backward symmetric cylindrical geometry1 with layers of tracking, calorimeter, andmuon detec- tors over nearly the entire solid angle around the pp collision point.The directions andenergies ofhightransverse momentum particlesare measured usingtrackingdetectors,finelysegmented hadronicandelectromagneticcalorimeters,andamuonspectrom- eter, within axial and toroidal magnetic fields. The inner tracker consists of silicon pixel, silicon microstrip, and transition radia- tiontrackingdetectors,andreconstructscharged-particletracksin
|η|<2.5. Lead/liquid-argon (LAr) sampling calorimeters provide electromagnetic (EM) energy measurements with high granular- ity.Asteel/scintillator-tilehadroniccalorimetercovers thecentral pseudorapidityrange(|η|<1.7). Theendcapandforwardregions areinstrumented withLArcalorimeters forEM andhadronicen- ergymeasurements up to |η|=4.9.The triggersystem [36] con- sistsofafirst-leveltriggerimplementedinhardware,usingasub- setofthedetectorinformationtoreducetheacceptedrateto100 kHz,followedbyasoftware-basedtriggerthatreducestherateof recordedeventstoabout1kHz.
3. Datasamplesandeventselection
TheresultpresentedinthisLetterisbasedondatacollectedin ppcollisionsat√
s=13 TeV during2015–2017.Thesignalconsists ofeventswithtwo jetsfromthe decayofanewparticle,andan additionalphoton,radiatedoffoneofthecollidingpartons.
Datawerecollectedviaeitherasingle-photontriggeroracom- binedtriggerrequiringadditionaljets,toallowalowerpTrequire- ment on the photon. The data collected with the single-photon trigger are used to search forresonances with masses from225 GeVto450GeV,whilethedatacollectedwiththecombinedtrig- gerare usedtosearch forresonanceswithmassesfrom450GeV to1.1TeV.
The single-photon trigger requires at leastone photon candi- date with Eγ
T,trig>140 GeV, where Eγ
T,trig is the photon trans- verse energyasreconstructed by thesoftware-based trigger. The combined trigger requires a photon and two additional jet can- didates, each with pT>50 GeV. The combined trigger requires Eγ
T,trig>75 GeV for the 2016data, increasing to Eγ
T,trig>85 GeV
1 ATLASusesaright-handedcoordinatesystemwithitsoriginatthenominalin- teractionpoint(IP)inthecentreofthe detectorandthe z-axisalongthebeam pipe.Thex-axispointsfromtheIPtothecentreoftheLHCring,andthe y-axis points upwards. Cylindrical coordinates (r,φ)are usedinthe transverseplane, withφ beingthe azimuthal anglearoundthe z-axis. The pseudorapidityis de- finedinterms ofthe polarangleθ asη= −ln tan(θ/2).It isequivalenttothe rapidityfor masslessparticles.Transversemomentumandenergyaredefinedas pT≡psinθandET≡Esinθ,respectively.Angulardistanceismeasuredinunitsof
R≡
(η)2+(φ)2.
for the 2017 data. This trigger was not active during the 2015 data-taking period. As a consequence, the single-photon trigger recorded 79.8 fb−1 of data and the combined trigger recorded 76.6 fb−1 of data. Both triggers are fully efficient within uncer- taintiesinthekinematicregimesusedforthisanalysis.
Afterrecordingthedata,asubsetofcollisioneventsconsistent withthesignalare selectedtopopulatemjj distributionsforsub- sequentanalysis.Abriefdescriptionofthereconstructionmethods isgivenbelowtogetherwiththeeventselection.
In all of the events selected for analysis, all components of thedetectorarerequiredtobeoperatingcorrectly.Inaddition,all events are required to havea reconstructed primary vertex [37], defined as a vertex withat least two reconstructed tracks, each withpT>500 MeV.
Photoncandidatesarereconstructedfromclustersofenergyde- positsin theelectromagnetic calorimeter [38]. The energyof the candidateis correctedby applying energyscale factorsmeasured withZ→e+e−decays [39].
The trajectory ofthe photon is reconstructed usingthe longi- tudinal segmentation of the calorimeters along the shower axis (shower depth) anda constraintfromthe average collisionpoint oftheprotonbeams.Candidatesarerestrictedtotheregion|η|<
2.37,excludingthetransitionregion1.37<|η|<1.52 betweenthe barrelandendcapcalorimeterstoensurethattheyarisefromwell- calibratedregionsofthecalorimeter.Anadditionalrequirementis appliedonthetransverseenergyofthephotoncandidateafterre- construction,whichisrequiredtohaveEγ
T >95 GeV,where Eγ T is thetransverseenergyofthephotoncandidateafterreconstruction.
Quality requirements are applied to the photon candidates to rejecteventscontainingmisreconstructedphotonsarisingfromin- strumental problems or from non-collision backgrounds. Further tight identification requirements are applied to reduce contami- nation from π0 orother neutralhadronsdecayinginto twopho- tons [38].The photonidentificationisbasedontheprofile ofthe energydeposits inthe firstandsecond layers ofthe electromag- netic calorimeter. In addition to the tight identification require- ment,candidatesmustmeettightisolation criteriausingcalorime- ter and tracking information, requiring that they be separated from nearby event activity [40,41]. Converted photon candidates matched to one track or a pair of tracks passing inner-detector quality requirements [38] and satisfying tight identification and isolation criteriaare alsoconsidered.Any pairofmatching tracks mustformavertexthatisconsistentwithoriginatingfromamass- lessparticle.
Jetsarereconstructed usingtheanti-kt algorithm [42,43] with radius parameter R =0.4 from clusters of energy deposits in thecalorimeters [44].Qualityrequirementsareappliedtoremove events containing spurious jets from detector noise and out-of- timeenergydepositsinthecalorimeterfromcosmicraysorother non-collision sources [45].Jet energies are calibratedtothe scale ofthe constituentparticles ofthejet andcorrected forthepres- enceofmultiplesimultaneous(pile-up)interactions [46,47].
After reconstruction, jets with transverse momentum pjetT >
25 GeV andrapidity |ηjet|<2.8 areconsidered.Tosuppresspile- up contributions, jetswith pjetT <60 GeV and |ηjet|<2.4 are re- quired to originate fromthe primary interaction vertexwith the highestsummedp2Tofassociatedtracks.Ifajetandaphotoncan- didatearewithinR=0.4,thejetcandidateisremoved.
Theserequirementsretainapproximately30%ofatypicalsignal sample.
Jets which likely contain b-hadrons are identified (b-tagged) with the DL1 flavour tagger [48]. Tracks are selected in a cone around the jet axis, using a radius which shrinks with increas- ing pjetT .Theselectedtracksareusedasinputtoalgorithmswhich
Table 1
Eventselectionsusedtoconstructeachofthefoureventcategories,asdescribedin thetext.
Criterion Single-photon trigger Combined trigger
Numberofjets njets≥2
Numberofphotons nγ≥1
Leadingphoton EγT>150 GeV EγT>95 GeV Leading,subleadingjet pjetT >25 GeV pjetT >65 GeV Centrality |y∗| = |y1−y2|/2<0.75 Invariantmass mjj>169 GeV mjj>335 GeV Criterion(appliedtoeach
triggerselection)
Inclusive b-tagged
Jet|η| |ηjet|<2.8 |ηjet|<2.5
b-tagging – nb-tag≥2
attemptto reconstruct a b-hadron decaychain. The resulting in- formation is passed to a neural network which assigns a b-jet probability toeach jet. Toaccount formismodellinginsimulated b-hadrondecays,acomparisonofthediscriminationpowerofthis networkindataandMonteCarlosimulationisperformedandcor- rectionfactorsareappliedtosimulationtoreproducethedata [49].
Jetsareconsideredb-tagged whentheDL1scoreexceedsathresh- old consistent witha 77% b-hadron identification efficiencyon a benchmarkt¯tsample.Atthisthreshold,only0.7%light-flavourjets and25%charm-jetsareretained.
Events which contain at leastone photon candidate and two jetsare selectedusing theabove criteria andseparatedintofour categories for further analysis. Two of the categories are con- structed with flavour-inclusive criteria, for which b-tagging re- sults are ignored. One of these two categories contains events recordedviathesingle-photontrigger,andtheothercategorycon- tainseventsrecordedviathecombinedtrigger.Toensurethetrig- geris fully efficient,events in the single-photon-triggercategory are required to have a photon with Eγ
T >150 GeV and events in the combined-trigger category are required to have a photon withEγ
T >95 GeV andtwojetswithpjetT >65 GeV.Theremaining twocategoriesconsistofeventsselectedasintheflavour-inclusive categories, except that the two highest-pjetT jets must satisfy the b-tagging criteria and have |ηjet|<2.5 to ensure that they fall withintheacceptanceofthetrackingdetectors.
Dijet production at the LHC occurs largely via t-channel pro- cesses, leading to jet pairs with high absolute values of y∗= (y1−y2)/2,where y1 and y2 aretherapiditiesofthehighest-pT (leading) andsecond-highest-pT (subleading)jet, respectively. On the other hand,heavy particles tendto decay moreisotropically, withthe twojetshavinglower|y∗|values.Therefore,|y∗|<0.75 isrequiredforall fourcategories. Thisselectionrejects upto80%
ofthe multi-jet backgroundevents whileaccepting up to80% of thesignaleventsdiscussedbelow.Afurtherselectionisappliedto selecteventsaboveagiveninvariantmassdependingonthetrig- ger,mjj>169 GeV forthesingle-photontriggerandmjj>335 GeV for the combined trigger. This is so that the background can be described by a smoothly falling analytic function satisfying the goodness-of-fitcriteriadescribedin4.
The above selections, summarised in Table 1, yield 2,522,549 and 15,557 events acquired by the single-photon trigger for the flavour-inclusive andb-tagged categories, respectively. Theyyield 1,520,114 and 9,015 events acquiredby the combined trigger in thecorrespondingcategories.
Thedistributionsofmjjforeventsineachofthefourcategories areshowninFig.1.HypotheticalsignalswithmZ=250 GeV and mZ=550 GeV,asfurtherdiscussedinSection6,areoverlaid.
At the largest dijet masses considered, the combined-trigger categories provide greater sensitivity to signals than the single-
photon-trigger categories due to their greater signal acceptance.
ThesensitivityisdefinedasS/√
B,whereSandBarethenumber of signal and background events in the simulation samples de- scribed inSection 6.At thesmallestdijetmassesconsidered, the jet pT thresholds ofthe combined trigger causethose categories to loseefficiency forsignalsandbias themjj distributions ofthe background processes. Therefore,to optimise the search across a wide range of signal masses, the invariant mass spectra selected using the combined-trigger categoriesare used in the search for signalmassesabove450 GeV,whilethespectraobtainedwiththe single-photontriggerareusedforlowermasses.
4. Backgroundestimation
To estimate theStandard Model contributionsto the distribu- tions in Fig. 1, smooth functions are fit to the data. The dijet searches of the CDF, CMS, and ATLAS experiments [6,8,11,15,17, 17,15,7,20] havesuccessfullymodelled dijetmassdistributions in hadroncollidersusingasinglefunctionovertheentiremassrange considered in thosesearches. Thisapproachis not suitable when data constrain the fit too tightly fora single function to reliably modelboth endsofthedistributionsimultaneously.Here,a more flexible techniqueisadopted,similartothatusedinrecentATLAS dijetresonancesearches [22,21].Inthistechnique, asinglefitus- ing a givenfunctionoverthe entiremassdistributionis replaced bymanysuccessivefits.Foreachbinofthemassdistribution,the same function isused to fit a broad mass range centred on the bin,andthebackgroundpredictionforthatbinistakentobethe valueofthefittedfunctioninthecentreoftherange.Theprocess is repeatedfor each binof themass distribution andthe results arecombinedtoformabackgroundpredictioncoveringtheentire distribution. Forinvariant masseshigherthan themjj rangeused for the search (above 1.1TeV), the window is allowed to extend beyondtherangeaslongasdataisavailable.
Asetofparametricfunctionsareconsideredforthesefits:
f(x)=p1x−p2e−p3x−p4x2 (1) or
f(x)=p1(1−x)p2xp3+p4lnx+p5(lnx)2, (2) where x=mjj/√
s and pi are free parameters determined by fit- tingthemjjdistribution.Inadditiontothefive-parameterfunction in Eq. (2), a four-parameter variant with p5=0 and a three- parametervariantwithp5=p4=0 arealsoconsidered.Thewidth ofthemassrangeusedfortheindividualfitswasoptimisedtore- tain the broadestpossible rangewhile maintaininga χ2 p-value above 0.05 inregionsofthedistribution thatdonot containnar- rowexcesses,whereexcessesareidentifiedusingtheBumpHunter algorithmdescribedinthenextsection.Theslidingwindowproce- dure cannotbe extendedbeyondthelower edgeofthemjj range used in each signal selection. Therefore, until the optimal num- berofbinsisreachedoneachsideofagivenbincentre,thestart of the window is fixed to the lower edge of the spectrum and the fitted functional formis evaluated for each bin in turn.This procedure allows for a stable background estimate while main- tainingsensitivitytosignalslocalisedinthemjjdistribution.Tests performed by adding sample signals to smooth pseudo-data dis- tributions confirmedthat thisapproachcanfindsignalsofwidth- to-massratios upto 15%,withsensitivityincreasing fornarrower signals.Therangesoftheindividualfitsvaryfrom750 GeVinthe narrowest case to 1600 GeV in the widest case. A signal with a 15% width-to-mass ratio constrained by the narrowest fit would haveanabsolutewidthof163 GeV,orlessthanonequarterofthe fitrange.
