Contents lists available atScienceDirect
Physics Letters B
www.elsevier.com/locate/physletb
Search for flavour-changing neutral currents in processes with one top quark and a photon using 81 fb
−1of pp collisions at √
s = 13 TeV with the ATLAS experiment
.TheATLASCollaboration
a r t i c l e i n f o a b s t ra c t
Articlehistory:
Received22August2019
Receivedinrevisedform16October2019 Accepted29October2019
Availableonline13November2019 Editor:M.Doser
Asearchforflavour-changingneutralcurrent(FCNC)eventsviathecouplingofatopquark,aphoton,and anuporcharmquarkispresentedusing81 fb−1ofproton–protoncollisiondata takenatacentre-of- massenergyof13 TeV withtheATLASdetectorattheLHC.Eventswithaphoton,anelectronormuon, ab-taggedjet,and missing transversemomentumare selected.Aneuralnetwork basedonkinematic variables differentiatesbetweeneventsfromsignal andbackground processes.The dataare consistent with the background-only hypothesis, and limits are set on the strength of the tqγ coupling in an effectivefieldtheory.Theseare alsointerpretedas 95%CLupperlimits onthecrosssectionforFCNC tγ productionviaaleft-handed(right-handed)tuγ couplingof36 fb(78 fb)andonthebranchingratio fort→γu of2.8×10−5(6.1×10−5).Inaddition,theyareinterpretedas95%CLupperlimitsonthe crosssectionforFCNCtγ productionviaaleft-handed(right-handed)tcγ couplingof40 fb(33 fb)and onthebranchingratiofort→γc of22×10−5(18×10−5).
©2019TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).FundedbySCOAP3.
1. Introduction
Flavour-changingneutralcurrents(FCNCs)areforbiddenattree levelintheStandardModel(SM)andstronglysuppressedathigher ordersvia the GIM mechanism [1].Several extensionsto the SM predict processes involving FCNCs. In particular, some of these modelspredictthebranchingratiosoftop-quarkdecaysvia FCNC tobesignificantlylarger [2] thanthosepredictedbytheSM,which areoftheorder of10−14 [2]. Examplesare R-parity-violatingsu- persymmetricmodels [3] andmodelswithtwoHiggsdoublets [4].
SuchmodelswouldallowtheproductionoftopquarksviaFCNCs atameasurablerate.
ThisLetterpresentsasearchforFCNCsinprocesseswithatop quark(t)andaphoton(γ)basedondatacollectedwiththeATLAS experimentat√
s =13 TeV. Thisanalysisismostsensitive tothe productionofasingletopquarkplusaphoton,butalsoconsiders thedecayofapair-producedtopquarkintoanuporcharmquark (q)plusaphoton.Tree-levelFeynmandiagramsfortheseprocesses are shownin Fig. 1, where inboth cases,exactly one top quark decays via the SM-favoured t W b coupling. Comparedto the SM productionofatopquarkandaphoton,theFCNCprocessesresult inhigherphotontransversemomentaonaverage.
E-mailaddress:atlas.publications@cern.ch.
FCNC contributionsto the decaymode (t→qγ) and thepro- duction mode (q→tγ) can be expressed in terms of effective coupling parameters but also in terms of branching ratios and cross sections [5,6]. In the former case and following the nota- tion in Ref. [7], the corresponding operators are O(uBi j) and O(uWi j), where i=j areindices forthequark generation.Ingeneral, left- handed (LH) andright-handed (RH) couplings could exist, which result in different kinematic properties of the top-quark decay products, such asthe transverse momentum of the charged lep- toninsemileptonictop-quarkdecays.Themoststringentlimitsto datearelimitsonbranchingratiosofB(t→u γ)<1.3·10−4 and B(t→cγ)<1.7·10−3 set by the CMS Collaboration, assuming equalleft- andright-handedcouplings [8].
2. ATLASdetector
The ATLAS experiment [9] atthe LHCis a multi-purposepar- ticledetector witha forward-backwardsymmetric cylindricalge- ometry anda near4π coverage in solidangle.1 It consistsofan inner tracking detector (ID) surrounded by a thin superconduct-
1 ATLASuses aright-handedcoordinatesystemwith itsoriginat thenominal interactionpoint(IP)inthecentreofthedetectorandthez-axisalongthebeam pipe.Thex-axispointsfromtheIPtothecentreoftheLHCring,andthe y-axis pointsupwards.Cylindricalcoordinates(r,φ)areusedinthetransverseplane,φ beingtheazimuthalanglearoundthez-axis.Thepseudorapidityisdefinedinterms https://doi.org/10.1016/j.physletb.2019.135082
0370-2693/©2019TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/).Fundedby SCOAP3.
Fig. 1. Tree-level Feynman diagrams for top-quark production (left) and decay (right) via FCNCs. The tqγvertex, which is not present in the SM, is highlighted.
ingsolenoidprovidinga 2 Taxialmagneticfield, electromagnetic andhadron calorimeters,and amuon spectrometer (MS). The ID covers the pseudorapidity range |η|<2.5. It consists of silicon pixel,siliconmicrostrip,andtransitionradiationtrackingdetectors.
Lead/liquid-argon(LAr)samplingcalorimetersprovideelectromag- netic(EM) energymeasurementswithhighgranularity. Ahadron (steel/scintillator-tile)calorimeter coversthe central pseudorapid- ityrange(|η|<1.7). Theend-capandforwardregions areinstru- mented with LAr calorimeters for both EM and hadronic energy measurementsupto|η|=4.9.TheMSsurroundsthecalorimeters andisbasedonthreelargeair-coretoroidalsuperconductingmag- netswitheightcoilseach.Thefield integralofthetoroidsranges between2.0and6.0 T m acrossmostofthedetector.The MS in- cludesasystemofprecisiontrackingchambersandfastdetectors fortriggering.Atwo-leveltrigger systemis usedtoselectevents.
Thefirst-leveltriggerisimplementedinhardwareandusesasub- setof thedetector informationto reduce the acceptedrateto at mostnearly 100 kHz.Thisisfollowedbyasoftware-basedtrigger thatreducestheacceptedeventrateto1 kHzonaveragedepend- ingonthedata-takingconditions.
3. Analysisstrategy
Thesearchstrategy selectseventswithafinalstate containing onepromptphotonandthedecayproductsofatopquark,namely anelectronoramuon,ab-taggedjet,andmissingtransversemo- mentum,andestimatescontributionsfromFCNCprocessesinthis background-dominated dataset.Asignalregion (SR)isdefinedby looserequirements on thekinematic propertiesof thefinal-state objects,giving rise to alarge acceptance forsignal events inthe production mode. With this selection, the search is most sensi- tivetoFCNCsinthismode,butthedecaymodeisincludedinthe analysis. The mainbackground contributionstems fromelectrons misidentified as photons, primarily in top-quark–anti-top-quark events (tt).¯ These contributions and contributions from hadrons misidentified as photons (both labelled “fakes” in the following) aremodelledbyMonteCarlo(MC)simulationsandscaledtodata- drivenestimates.Photons producedinassociationwitha leptoni- callydecayingW orZ bosonareestimatedincontrolregions(CRs) whichdonotoverlapwiththeSRbutarekinematicallysimilarto it.Thepredictionsforothersmallprompt-photonbackgroundpro- cessesare takenfromMC simulationandincludet¯tγ production, single-topquarkproductioninassociation withaphoton andthe production of two massive gauge bosons with a photon. As the lattertwoprocessesresultinasmallcontributiontothetotalback- groundprediction, it issufficient that in theseprocesses prompt photonsaregenerated bythe parton-showerprogram.Signal and backgroundeventsaredistinguishedusinganeuralnetwork(NN).
ofthepolarangleθasη= −ln tan(θ/2).Angulardistanceismeasuredinunitsof
R≡
(η)2+ (φ)2.
Finally,thesignal contributionisestimatedinaprofile likelihood fittotheNNoutputdistributionsintheSRandtheCRs,inwhich each source of systematicuncertainty is modelled as a nuisance parameter.
4. Dataandsimulation
The proton–proton(pp) collision dataanalysed were recorded with the ATLAS detector from2015 to 2017 ata centre-of-mass energy of √
s=13 TeV. The average number of interactions per bunchcrossingwas13.4,25.1,and37.8in2015,2016and2017,re- spectively.Events wereselectedbysingle-leptontriggers [10] and required to haveat least one reconstructed primary vertexwith at least three assignedtracks that havea transverse momentum greaterthan400 MeV.Aftertheapplicationofdata-qualityrequire- ments,thedatasamplecorrespondstoanintegratedluminosityof 81 fb−1.ItisobtainedusingtheLUCID-2detector [11] forthepri- maryluminositymeasurements.
The data were modelled by MC simulations ofthe signal and backgroundprocesses.Aftereventgeneration,theresponseofthe ATLAS detectorwas simulatedusing Geant 4 [12] with afull de- tectormodel [13] ormodelledbyafastsimulation [14].Toaccount foradditionalpp collisions(pile-up),inelasticpp interactionswere superimposedonthehard-scatteringeventsandweightedaccord- ing to theobserved pile-updistribution.The pile-upevents were simulated using Pythia 8.186 [15],withthe A3 [16] setof tuned parameters(A3tune).
The simulated signal samples were generated using Mad- Graph5_aMC@NLO 2.4.3 [17] with the TopFCNC model [6,18] at next-to-leading order (NLO) in QCD and the NNPDF3.0NLO [19]
set ofpartondistribution functions(PDFs).The partonshowering was done with Pythia 8.212 with the A14 tune set [20]. Simu- latedsamplesofSMtt and¯ single-top-quarkeventsweregenerated using Powheg-Box [21–27] with the NNPDF3.0NLO PDF set. The parton showering, hadronisation, andthe underlying event were modelledusing Pythia 8.230.Thetop-quarkmassmtop wassetto 172.5 GeV inthesesamples,andthehdampparameterthatcontrols thetransversemomentumofthefirstgluonemittedwassetto1.5 timesmtop. Samplesoft¯tγ eventswere generatedas2→7 pro- cess atleading order using MadGraph5_aMC@NLO 2.3.3and the NNPDF2.3LO PDF [28] set andwiththe followingfiducial photon criteria [29]: photon pT>15 GeV and |η|<5.0, charged-lepton pT>15 GeV and|η|<5.0,andR<0.2 betweenthephotonand any chargedfinal-stateparticle. The crosssections forSM t¯t and single-top-quarkproductionarescaled totheNNLO+NNLLpredic- tions [30–33].Theleading-ordercrosssectionfort¯tγ productionof 4.62 pb [29] isscaledtotheNLOpredictions [34] usingak-factor of1.24.TheNNLO+NNLLcrosssectionforSMt¯t productionisalso used tonormalisethesignal inthe decaymode,usingthecorre- spondingFCNCbranchingratio.Thecrosssectionsforthesignalin the productionmode, however,are calculated atNLO with Mad- Graph5_aMC@NLO.
For the study of systematic uncertainties in the modelling of processes involving top quarks, simulated t¯t samples were produced with Powheg-Box + Herwig 7.0.4 [35,36] and Mad- Graph5_aMC@NLO 2.6.0plus Pythia 8.212.TheMMHT2014LO [37]
PDFsetisusedtogetherwiththeH7-UE-MMHT [36] tune.Anad- ditional tt sample¯ was produced with hdamp set to three times mtop andthe factorisation and renormalisation scales set to half their nominal values using the A14tune. For the t W process, a samplewasproducedwithanalternativeschemeforremovingthe overlapwitht¯t production [38]. Moreover, single-top-quark sam- ples were produced with Powheg-Box + Herwig 7.0.4 and Mad- Graph5_aMC@NLO 2.6.2 plus Pythia 8.212.The NNPDF2.3LO PDF setisusedaswellastheA14tune.
Processes with one or two heavy gauge bosons, in particu- larthe processes W +γ+jets and Z +γ+jets, were simulatedusing Sherpa2.2.1 and 2.2.2 [39] with the matrix elements calculated usingComix [40] and OpenLoops [41]. All matrixelements were mergedwith the Sherpa partonshowering [42] accordingto the ME+PS@NLO [43] prescription. The NNPDF3.0NNLO PDF set was used.
An overlap removal scheme was applied to remove double- countingofeventsstemmingfromphotonradiationinsamplesin whichaphotonwasnot explicitlyrequiredinthefinalstate [29].
Thisappliestotheprocessestt,¯ W +jetsand Z +jetsinordertore- movetheoverlapwiththett¯γ,W +γ+jetsandZ +γ+jetssamples.
Mismodellingofthephoton pT distributionis observedinthe W +γ+jetandZ +γ CRs,whicharedefinedinSection5.Thephoton pT spectrum in the W +γ+jets and Z +γ+jets processes was cor- rectedby adjustingthe MCpredictionto thedatainfive pT bins usingalinear functionthat onlychanges theshape ofthedistri- butionandnotitsnormalisation.Thiscorrectiontothephoton pT improvesthemodellingoftheNNoutputdistributionintheCRs.
As discussed in Sections 6 and 7, the contribution of events withelectrons andhadronsmisidentified asphotonsis corrected usingdata.Thecontributionfromprocesseswithhadronsmisiden- tifiedasleptonsisestimatedtobenegligible.
5. Objectandeventselection
Electronsare reconstructedfromclustersofenergydepositsin theelectromagneticcalorimetercellswithamatchedIDtrack [44].
Theyarerequiredtomeetthetight identificationcriteria [44],and their tracksmust point to theprimary vertex. Theymust havea transversemomentum pT largerthan27 GeV and|ηcluster|<2.47, excluding1.37<|ηcluster|<1.52,where|ηcluster|isthepseudora- pidityoftheelectron’senergycluster.Muonsarereconstructedby combining a track in the MS with a track in the ID [45]. They are required to meet the medium identification criteria [45] and must point to the primary vertex. They must have pT>27 GeV and |η|<2.5. Isolated electrons and muons are selected by re- quiring the amount of energy in nearby energy deposits in the calorimeters and the scalar sum of the transverse momenta of nearbytracksintheIDtobesmall.
Photons are reconstructed fromclustersof energydeposits in the electromagnetic calorimeter cells with no matched ID track (unconvertedphotons)orwithoneortwomatchedIDtracksthat arecompatiblewiththetracksfromanelectronorpositronfroma photonconversion(convertedphotons) [44].TheymusthavepT>
20 GeV and|ηcluster|<2.37,excluding1.37<|ηcluster|<1.52.They arerequiredtomeetthetight identificationcriteriafortheshape ofthe shower inthe electromagneticcalorimeter(shower shape) andfortheenergydepositedinthehadroniccalorimeter [44].Pho- tonsmustbeisolatedfromnearbyenergydepositsinthecalorime- terandnearbytracksintheID: thesumoftheenergydeposited (pT of the tracks) within R=0.4 (R=0.2) of the photon
direction is required to be smaller than 0.022×pT+2.45 GeV (0.065×pT), excluding thephoton energydeposition (tracks as- sociatedwiththephoton).
Jets are reconstructed fromtopological clusters [46,47] in the calorimeterswiththeanti-kt algorithm [48] usingFastJet [49] and aradiusparameterof0.4.Theirenergyiscalibrated [50],andthey mustfulfilpT>25 GeV and|η|<2.5.Jetswith pT<120 GeV and
|η|<2.4 are required to pass a requirement on the jet-vertex- tagger (JVT) [51] to suppress pile-upjets. Jetsare b-tagged with the MV2c10 algorithm [52], which uses a boosted decision tree withseveralb-taggingalgorithmsasinput.Theb-taggingefficiency forjetsthatoriginatefromthehadronisationofb-quarksis60%in simulatedt¯t events.Theb-taggingrejection2 forjetsthatoriginate fromthehadronisation ofc-quarks(u-,d-, s-quarksorgluons)is 23(1200).
The magnitude ofthe missing transverse momentum EmissT is reconstructed fromthe vector sumof the pT of leptons,photons, andjets, aswellasIDtracksthatpointtotheprimaryvertexbut arenotassociatedwithareconstructedobject(softterm) [53].
Toavoiddouble-counting,objectsareremovedinthefollowing order: electrons sharing a trackwith a muon; jets within R= 0.2 ofan electron;electronswithin R=0.4 ofa jet;jetswithin
R=0.4 of a muon ifthey haveat mosttwo associatedtracks;
muonswithin R=0.4 ofa jet; photonswithin R=0.4 ofan electronormuon;jetswithinR=0.4 ofaphoton.
Scalefactors(SFs)are usedtocorrecttheefficiencies insimu- lationinordertomatchtheefficienciesmeasured indataforthe electron [44] andmuon [45] trigger,reconstruction,identification, andisolationcriteria, aswell asforthephotonidentification [44]
and isolation requirements. SFs are also applied for the JVT re- quirementandfortheb-taggingefficienciesforjetsthatoriginate fromthehadronisationofb-quarks [54],c-quarks [52],andu-,d-, s-quarksorgluons [55].
The selected events have exactly one electron or muon, ex- actly one photon, exactly one b-tagged jet and no further jets, and EmissT >30 GeV.This selection definesthe SRwithsignal ef- ficiencies3 for the production mode of 3.03% (2.45%) for the LH (RH)tuγ couplingandof3.79%(3.14%)fortheLH(RH)tcγ cou- pling.These efficienciesare definedwithrespect tosignal events that include a leptonic decay of the W boson, i.e. a decay with anelectron,muonortaulepton.Theefficienciesforthecouplings thatinvolveac-quarkarelargerthanthosethatinvolveau-quark becauseofthedifferenceinthekinematicsthatarisesfromthedif- ferencebetweentheu- andc-quarkPDFs.Thedifferencebetween theefficienciesfortheLHandtheRHcouplingsisduetothekine- maticdistributionsoftheW boson’sdecayproducts,whichdiffer dependingonthehandednessofthetopquark.Theefficienciesfor thedecaymodeare0.45%and0.51%forthetuγ andthetcγ cou- pling, respectively, and are lower dueto the requirement ofnot morethanone jetinthefinal state,i.e.thisanalysisisoptimised forthe productionmode. The absolutestatisticaluncertainties in the efficiencies are0.03% orsmaller. Inthe SR, 9557data events are selected. The ratioof production-mode to decay-mode event yieldsis4.2(5.3)fortheLH(RH)tuγ couplinganditis0.86(0.68) fortheLH(RH)tcγ coupling,i.e.inthecaseoftuγ coupling,the dominantsignalprocessisindeedtheproductionmode.However, inthecaseofthetcγ coupling,thedecaymodealsoplaysanim- portant role, because the production mode is suppressed by the c-quarkPDF.
2 Therejectionisdefinedastheinverseoftheefficiency.
3 Here,thesignalefficiencyincludesthesignallossduetothelimitedacceptance ofthedetector.
