Post on 22-Jan-2016
A study on A study on AlpGen and SherpaAlpGen and Sherpa
in Z+jets eventsin Z+jets events
Piergiulio LenziUniversità di Firenze & INFN
CTEQ-MCnet Summer SchoolDebrecen, 8-16 August 2008
IntroductionIntroduction
I’maPhDstudentbasedinFlorence;I’minmythirdyear(thelastinItaly),workingonCMS FlorencegrouphasbeendeeplyinvolvedintheCMSInnerTrackerconstruction,soI’vebeenworkingon: Constructionandcommissioningofthedetector Developmentoftrackreconstructionalgorithms
Inthelastyearwededicatedmoreeffortonthephysics,andwehavechosentheZ+jetschannel AsIwasstartingtostudythenewsubjectthe“MCnet short-term studentship”announcementappearedononeoftheCMSmailinglists!! Iimmediatelydecidedtoapply!
OutlineOutline
SomethingaboutwhatIusedtodobeforemystudentship… BuildingtheCMStracker
TrackReconstruction@CMS
TheMatrixElement–PartonShowermatchingproblem
Twodifferentapproachestothematching: CKKW(Catani,Kuhn,Krauss,Webber),theSHERPAway
MLM(M.Mangano),theAlpGenway
ProductionofZ+jetseventsatLHC(14TeV) Sherpaanditssystematics
AlpGenanditssystematics
Acomparisonbetweenthetwo
CMS Tracker CMS Tracker constructionconstruction
I’vebeeninvolvedinallthestepsofthetrackerconstruction,fromsensorqualityteststothefulldetectorcommissioning
Beam test, Oct 2004
Cosmic rays in a tracker prototype, June 2005
Layer assembly, Feb 2006
Final Inner Barrel assembly, Oct 2007
Final assembly and tests, Mar-Jul 2007
Track reconstructionTrack reconstruction
I’veworkedonthedevelopmentofCMStrackreconstructionalgorithms;inparticular Commissioningofdetectorandreconstructionsoftwarewithcosmicrayrealdata
Track reconstruction efficiencyon cosmic rays as a function of track θ. Black simulation; Red data
DevelopmentoftheKalmanFilterbasedpatternrecognition
Developmentoftrackfittingmethodsbasedonsofthit-to-trackassignment(DAF).Shouldperformbetterindenseornoisyenvironment
Transverse and longitudinal impact parameter resolution; Blue standard tracking Red DAF
preliminary
My studentshipMy studentship
IspentmystudentshipatUniversityCollegeLondon(UCL)fromJantoApr2008,supervisedbyProf.JonButterworth IwasworkinginclosecontactwithdevelopersoftheRIVETpackageandwithMCexpertshttp://projects.hepforge.org/rivet/trac/wiki/ RIVET(Robust Independent Validation of Experiment and
Theory)ispartoftheCEDAR(Combined e-Science Data Analysis Resource)[1]project
RivetmainobjectiveistoprovideageneralframeworkforMCvalidation/tuningandcomparisonwithexperimentalresults
ThemainideaistoreproduceinRivetexperimentalanalyses;resultsobtainedrunningsuchanalysesonMCsamplesaretheneasilycomparedwithexperimentalresultsstoredinHEP-DATADB[2]
[1] J.M. Butterworth et al. “The CEDAR Project”, hep-ph/0412139[2] http://hepdata.cedar.ac.uk
RIVETRIVET
InterfacetomanygeneratorsisprovidedbytheseparateAGILelibraryhttp://projects.hepforge.org/agile/ RIVETisaC++package,withtwomainkindofobjects Projections:theseclassescalculateeventobservables.Projectionsarehandledthroughasmartcachingsystemavoidingthesameobservablebeingcalculatedtwiceforthesameevent
Analyses: RIVETcomeswithmanyanalysesthatreproduceexperimentalpapersfromLEP,HERA,TVT.
UseranalysescanbeeasilypluggedintheRIVETframeworkwithapluginsystem
Comparisonwithexperimentalresultsisstraightforward,thankstotheuseofthelowweightAIDA-xmlplotformat,thesameusedwhenexportingexperimentalresultsfromHEP-DATAdatabase
ME and PSME and PS
MatrixElement(ME)calculationsarebasedonthecomputationoftheFeynmanamplitudesoftheprocessescontributingtothefinalstate
Partonshowers(PS)describethefinalstatestartingfromahard22processasasuccessiveseriesof12splittings.
Splittingprobabilityiscalculatedinthelimitofcollinearemission
Divergencesinthislimitaretamedimposingtheconservationoftotalprobabilityusingformfactors(Sudakov)
The ME-PS matching The ME-PS matching problemproblem
MatrixElementcalculations:✓Describewellseparatedjetsconfigurations✓Areexactatagivenorder✕Runintotroublesinthesoftandcollinearregions✕Can’tdescribetheinternalstructureofjets
- PartonShowercalculations:✓PSisuniversal,giventhebasichardprocessthePSrecipewillproducereasonablepartonconfigurations
✓TheuseofSudakovformfactorsensurescontrolledbehaviorinthesoftandcollinearregion,sojetevolutioniswelldescribed
✕Cannotsteertheshowerevolutiontoomuch,someregionsofthephasespacearenotefficientlyfilled,e.g.wellseparatedpartonconfigurations
- ItwouldbeveryconvenienttousetheMEtopredicthardpartonconfigurationandthePStodescribetheevolutionofjets,butbewaredoublecountingandholesinthephasespace
SHERPA SHERPA approachapproach to the to the matching: CKKWmatching: CKKW
Jetproductionandjetevolutionregionsarewellseparatedusingaktmeasure(y=min(Pt12,Pt22)*ΔR2/D2)cutoff
ycut=Qcut2/Ecm
2
CalculatetheMEcrosssectionsforallthedesiredpartonmultiplicities;ycutisusedtocutoffdivergences;afixeds
MEisused
SelectonepartonmultiplicityaccordingtoMEcrosssectionsandproducejetmomentaaccordingtothecorrespondingMEsquared
a“showerhistory”isreconstructedthroughktclusteringuntila
22processisfound Inotherwordsthisanswerstothequestion:“HowcouldaPShave
producedthispartonconfiguration?”
Arunningcouplingcorrectionweightisapplied
ApplyaSudakovformfactorcorrectionforeachclustering TheeffectofthiscorrectionistoweighttheeventasifitwereproducedbyaPSbutsubstitutingthecollinearsplittingfunctionwiththeexactME
EvolvetheeventwithaPS,vetoingemissionabovethecut
Cluste
r A
bove
y
cut
Bel
ow
ycu
t
AlpGen approach to the AlpGen approach to the matching: MLMmatching: MLM
TheMLMalgorithmproceedsexactlyastheCKKWuptothereweightingofs
EventsemergingfromtheMEarefullyshoweredusingaconventionalshower(PYTHIAorHERWIG) Partonsareclusteredintojets(withaconealgorithmintheAlpGenimplementation) Jetsarematchedtooriginalpartons
Ifnotallthejetsmatchtotheoriginalpartonseventisrejected
ThiseffectivelyreproducesSudakovreweighting EffectivelyvetoesPSemissionabovethemergingscale
Z+jets event Z+jets event generationgeneration
SHERPA SHERPA-MC-1.0.11wasused ppe+e-+upto3jetsat14TeV CTEQ6lpdf Nounderlyingevent 66.GeV<mass(e+e-)<116.GeV MatchingparameterQcut=20GeV
AlpGen V2.13wasused ppe+e-+njets(n=0,1,2,3)at14TeV CTEQ6lpdf 66.GeV<mass(e+e-)<116.GeV -Matchingparameters:
Et(clus)=25GeV R(clus)=0.7 ηmax=5
PythiaandHerwigUEwasswitchedoff
AnalysisAnalysis
Finalstateparticleswith|η|<5wereselected Noptcutonthefinalstateparticles KtjetswerereconstructedusingtheFastJetspackagebyM.Cacciari,G.Salam Ptmin=30GeV,D=0.4
Manyleptonandjetobservableshavebeencalculated Foreachgeneratoracollectionofobservablesisshownandtheeffectofachangeinthematchingparametersandinthescalechoiceisdescribed AcomparisonbetweenAlpGenandSherpaisshown
Sherpa observablesSherpa observables
Pt lepton pairPt e-
Jet multiplicity Pt leading jet
Sherpa, differential jet Sherpa, differential jet ratesrates
Distributionofthektresolutionvariablemakinganjetturnintoan-1jetevent
Criticalregionaroundthecut(log10(20)~1.3)
TheMEkinematicsisalteredbythePS,somismatchesaroundthecutcanoccur
MEPS
Sherpa systematics: Sherpa systematics: change of the matching change of the matching
scalescale ItiredthreevaluesforQcut:20GeV(default),30GeV,50GeV
Theeffectontheoverallcrosssectionissummarizedbelow
Theeffectonleptonobservables,andtherelativedifferencewithrespecttothedefaultisshownintheplotsbelow AstheQcutisincreased,theMEphasespacefractionisreduced
SincetheMEisresponsibleforthehardestpartonkinematicstheincreaseinthemergingscaleresultsinspectrawithsoftertails
Qcut=20GeV
Qcut=30GeV
Qcut=50GeV
1594.1 pb 1411.3 pb 1399.8 pb
Pt lepton pair Pt e-
Sherpa systematics: Sherpa systematics: change of scaleschange of scales
Sherpawasrunwiththedefaultchoiceforrenormalization(R)andfactorization(F)scales,RandFtimes0.5andRandFtimes2
Theeffectontheoverallcrosssecionissummarizedbelow:Default scales
Scales*0.5 Scales*2
1594.1 pb 1292.4 pb 1646.5 pb
Theeffectismoreevidentinthefirstbinswherethe0jetcontributiondominates
Theeffectchangessignforhigherbosonpt
Thisagreeswithpreviousstudies,whereitwasshownthattheLOcrosssectionfor0jetsgrowswiththescale,whiletheLOcrosssectionfor>0jetsdecreasesasthescalegrows
[hep-ph/0308195]
Pt lepton pair
AlpGen systematics, AlpGen systematics, change of the merging change of the merging
scalescale TwovalueswereusedfortheEt(min)oftheconealgorithmusedtosteerthematching:25GeVand40GeV
Theoverallcrosssectioneffectissummarizedbelow
Theeffectonleptonobservablesisalmostnegligible
Et(min)=25GeV Et(min)=40GeV
1534.5 pb 1516.4 pb
AlpGen systematics, AlpGen systematics, Change of scalesChange of scales
AlpGen+Pythiawasrunwiththedefaultscalechoice,Rtimes0.5andRtimes2
Theeffectontheoverallcrosssectionissummarizedbelow: Default
scales Scale*0.5 Scale*2
1534.5pb 1449.1pb 1672.2pb
TheeffectissimilartotheoneobservedinSherpa
Thelowptregion,dominatedbythe0jetcontributionshowsabehaviorsimilartoSherpa
AlpGen-SHERPA AlpGen-SHERPA comparisons:comparisons:
lepton observableslepton observables AlpGenhasbeenshoweredbothwithPythiaandHerwig Theinclusivecrosssectionsare
Sherpashowsthehardestspectrum,Pythiathesoftest DifferencesbetweenPythiaandHerwigareduetothedifferentwaythetwoPSaltertheMEkinematics
ThePtspectrumdifferencestranslateintotheηspectrum,withPYTHIAshowingthelesscentralZandSHERPAthemostcentralone
SHERPA AlpGen+PYTHIA
AlpGen+Herwig
1594.1 pb 1534.5 pb 1523.0 pb
AlpGen-SHERPA AlpGen-SHERPA comparisons:comparisons:
jet observablesjet observables Sherpashowsaharderleadingjetspectrum ThejetmultiplicityismaximumforSherpaandminimumforAlpGen+Pythia
ConclusionConclusion
Aboutmywork AlpGenandSHERPAimplementtwodifferentapproachestothematchingofMEandPSpredictions
Astudyofthe“theoreticaluncertainties”hasbeencarriedoutforbothgenerators,varyingthematchingparametersandthescales
WhencomparingAlpGenwithSHERPA,somenotnegligibledifferenceshavebeenspotted
Andingeneralaboutmystudentship: IworkedinclosecontactwithMCexperts IgotaninsightonthedifferencesamongsomeoftheMCtoolsavailable,andonthesystematicsofsuchtools
….andIspentsomenicetimeinLondon!
Many thanks to Jon Butterworth, James Monk, Andy Buckley, Frank Krauss, Frank Siegert
BackupBackup
Track reconstructionTrack reconstruction Trackreconstructionisacomplexpatternrecognitionproblem AcombinatorialapproachtogetherwithaKalmanfilterestimationoftrackparametersisused
pixelpixel
Inner BarrelInner Barrel
Outer BarrelOuter Barrel
EndcapEndcap
ss
Inn
er
Dis
ksIn
ner
Dis
ks
5.6 m5.6 m
1.2 1.2
mm
Trackcandidatesfoundinthiswayarethenfittedinordertoestimatetrackparametersoneachsurface
The ME-PS matching The ME-PS matching problemproblem
ThePSresponsecoincideswiththeMEinthesoft/collinearregionxi~1
ThedivergenceofthecrosssectioninthisregionistamedinthePSusingareweightingfactorcalledSudakovformfactor
ItwouldbeveryconvenienttousetheMEtopredicthardpartonconfigurationandthePStodescribetheevolutionofjets
BUTbewaredoublecountingandholesinthephasespace!
The PS formulaeThe PS formulae
Divergent in z = 0 and z = 1
Divergencies are tamed with a cutoff .
Evolving the shower a big splitting probability will turn into big probabilities for successive branchings. If we want to conserve total probability the splitting probability abc at scale Q has to be multiplied by the probability that that splitting did not happen before (at > Q).
This “non branching probability” is the
“Sudakov form factor”
PYTHIA approach to the PYTHIA approach to the matching: ME matching: ME correctionscorrections
Let’sconsidertheLOe+e-qqbarandtheNLOrealemissiononee+e-qqbarg
IftheshowerpopulatesthephasespaceaccordingtoWPS,thenafactorWME/WPSneedstobeapplied
WMEissubstitutedtoWPSintheSudakovformfactor
Inthehardregion,wheretheSudavov~1,theMEresultsholds,inthesoftregion,whereWME~WPSthePSresultholds
CKKW (continued)CKKW (continued)
Foreachinternallineinthe“showerhistory”connectingscaleitoscalekreweighteventswithafactor
Foreachexternallinereweighteventswithafactor Δk(qi,qj)istheSudakovformfactorassociatedtoclusteringk,connectingscalesqiandqj
TheneteffectofthisprocedureissimilartowhatisachievedwiththePythia’sMEcorrections:thesplittingfunctionspresentintheSudakovformfactorarereplacedbytheirMEversion
ThePSisthenappliedtotheweightedevents,withaVETOonhard(avovethemergingscale)emission
Sherpa, jet Sherpa, jet observablesobservables
AlpGen+Pythia,AlpGen+Pythia, default settings default settings Leptonobservables
Pt lepton pair
η lepton pair
Pt e- η e-
AlpGen+Pythia,AlpGen+Pythia,default settingsdefault settings
Jetobservables
AlpGen+Pythia,AlpGen+Pythia,default settingsdefault settings
Change of the matching Change of the matching scale (continued)scale (continued)
Jetobservables: Themeanmultiplicitydecreasesasthemergingscalegrows
DiffjetratesaredepletedintheMEregion,andareenhancedinthePSregion
Thesmoothesttransitionseemstobetheoneforthehighestmergingscaleused
Jet observables Jet observables (continued)(continued) Thedifferentialjetratesshow
manydifferences: AlpGen+PythiaandAlpGen+Herwigshowsimilarratesfor21and32transition,whiletheshapeisquitedifferentfor10
SherpaalwaysfillstheMEregionmore
Thisseemstobeinagreementwiththeharderspectraobservedforsherpa,sincetheMEisresponsibleforthehardestpartonkinematics
Change of the merging Change of the merging scale (continued)scale (continued)
LeadingjetPtshowabumparound40GeVforthenon-defaultsample
Smalldifferencesobservedinthejetmultiplicity
Differentialjetrateplotsshowsomemodificationatthemergingscale
Change of scales Change of scales (continued)(continued)
Theeffectonthejetmultiplicity: Thescales*0.5sampleshowsthehighestjetmultiplicity,whilethescales*2samplesshowsthesmallest
Thisagreeswithwhatwasobservedinthebosonptspectrum,wherethetaildueto>0jeteventswasenhancedinthescales*0.5caseanddepletedinthescales*2case
Theeffectofthejetptspectrumisnotveryevident
Lepton observables Lepton observables (continued)(continued)
AsfortheZspectrum,thee-PtspectrumisharderforSherpaandsofterforPythia Thee-ηaresimilar(within8%)
ConclusionsConclusions AlpGenandSHERPAimplementtwodifferentapproachestothematchingofMEandPSpredictions
Astudyofthe“theoreticaluncertainties”hasbeencarriedoutforbothgenerators,varyingthematchingparametersandthescales Thechangeontheshapeofthedistributionsissimilarforthetwogeneratorswhenchangingthescales
WhenchangingthematchingparametersAlpGenshowsminorchangesintheshapeofthedistributionsthanSherpa
WhencomparingAlpGenwithSHERPA,somenotnegligibledifferenceshavebeenspotted: ThebosonleptonandjetspectraareharderforSHERPA TheSherpameanjetmultiplicityishigherthentheAlpGenone
Itmightbeinterestingtoseehowdifferentialjetrateslookondata,andseehowwelltheMCcanreproducetheseplots,sincetheseobservablesareverysensitivetothewayinwhichthephasespaceisfilled