Tau reconstruction at the ATLAS experiment– Electric field: 5000 V/mm ... Björn Gosdzik • DESY HH • [email protected] Student Seminar 22.01.09 11 Motivation Tau reconstruction

Björn Gosdzik • DESY HH • [email protected] 1Student Seminar 22.01.09

Tau reconstruction at Tau reconstruction atthe ATLAS experimentthe ATLAS experiment

Björn GosdzikBjörn GosdzikStudent Seminar 2009Student Seminar 2009

37 Countries37 Countries169 Institutes169 Institutes2500 Authors2500 Authors


OverviewOverview

I. The Large Hadron Collider (LHC)

II. A Toroidal LHC AparatuS (ATLAS)

III.Tau reconstruction and identification

IV.“Safe Variables” for first data

V. Summary

Tau reconstruction atTau reconstruction atthe ATLAS experimentthe ATLAS experiment

OverviewOverviewThe Large Hadron Collider (LHC)The Large Hadron Collider (LHC)

ATLASATLASTau reconstruction and identificationTau reconstruction and identification

““Safe Variables” for first dataSafe Variables” for first dataSummarySummary

3

Circumference:Circumference: 27 km27 kmTotal energy:Total energy: 14.0 TeV14.0 TeVLuminosity:Luminosity: 10^34 cm^-2 s^-110^34 cm^-2 s^-1Number of bunches:Number of bunches: 28082808Protons per bunch:Protons per bunch: 1.15x10^111.15x10^11Number of Dipoles:Number of Dipoles: 12321232Beam energy:Beam energy: 2808 x 1.15x10^11 x 7x10^12 x 1.602x10^-19 J = 362 MJ2808 x 1.15x10^11 x 7x10^12 x 1.602x10^-19 J = 362 MJ

The The LLarge arge HHadron adron CCollider (LHC)ollider (LHC)

4

First beam on Sep. 10th at 10:28 a.m.


First events in ATLASFirst events in ATLAS


● Collimater hit by beam 140 m in front of ATLAS

● Mainly muons from pi0 production




6

● Diameter: Diameter: 25 m25 m

● Length: Length: 46 m46 m

● Barrel toroid length:Barrel toroid length: 26 m26 m

● Overall weight:Overall weight: 7000 t7000 t

● ~100 million electronic channels~100 million electronic channels

● ~3000 km of cables~3000 km of cables


The ATLAS detector (Inner detector)The ATLAS detector (Inner detector)


● Pixel detector:

– 80 million pixels

– Pixel size: 50 x 400 µm²

● Semiconductor Tracker (SCT)

– 6 million channels

– 60 m² silicon over 4 cylindrical barrel layers and 18 planar endcap discs

● Transition Radiation Tracker (TRT)

– 400 000 channels

– Volume: 16 m³

– Precision measurement of 0.17 mm





The ATLAS detector (Calorimeters)The ATLAS detector (Calorimeters)


● Liquid Argon (LAr) Calorimeter

– Barrel 6.4 m long, 53 cm thick


– LAr endcap consist of forward calorimeter, electromagnetic (EM) and hadronic endcaps

● Tile Calorimeter (TileCal)

– Barrel made of 64 wedges, each 5.6 m long and 20 tonnes

– 500 000 plastic scintillator tiles





The ATLAS detector (Muon and magnet system)The ATLAS detector (Muon and magnet system)


● Thin Gap Chambers


● Resistive Plate Chambers


– Electric field: 5000 V/mm

● Monitored Drift Chambers

– 1171 chambers with 354 240 tubes

– Tube resolution: 80 µm

● Cathode Strip Chambers:

– 70 000 channels

● Central Solonoid Magnet

– 2 Tesla magnet field

● Barrel Toroid


● End-cap Toroid






MotivationMotivation


● Standard model: large number of taus from decay of Z and W bosons with 100

● Important for search of Higgs boson and MSSM

● Polarisation sensitive to SUSY parameters

Tau properties: Tau properties: ● m = 1.78 GeVm = 1.78 GeV● c = 87 µm c = 87 µm ● BR ( ->hadrons) = 64.8%BR ( ->hadrons) = 64.8%

pb−1







qq

20

10

● mSugra: Neutralino1 LSP => possible candidate for dark matter

● Missing Et gives mass of neutralino 1







Transverse energy of tau-leptons for different processes:

● top quark decay

● W->tau nu

● Z->tautau

● SM Higgs (120 GeV)->

● VBF SM Higgs->

● SU1 sample







● Difficult to distinguish leptonic modes from primary electrons or muons

→ Reconstruction only hadronic decay modes

● Hadronic mode divided in 1-prong (one charged pion) and 3-prong (three charged pions)

● Five prong decay hard to distinguish from QCD jets





Tau reconstructionTau reconstruction


● Main background for taus in Main background for taus in ATLAS: QCD jets => require us to ATLAS: QCD jets => require us to distinguish taus from jets with distinguish taus from jets with high rejection powerhigh rejection power

● Tau leptons at ATLAS:Tau leptons at ATLAS:

– Collimated calorimeter clusterCollimated calorimeter cluster

– Low charged tracks multiplicityLow charged tracks multiplicity

– Displaced secondary vertexDisplaced secondary vertex

– Decay hadronically to 1 or 3 Decay hadronically to 1 or 3 charged daughterscharged daughters

● Veto against:Veto against:

– QCD jetsQCD jets

– ElectronsElectrons

– MuonsMuons

¿

−¿

¿

0




15


Tau reconstructionTau reconstruction

● Track seed: good quality tracks ( > 6 GeV)

● Associates other > 1 GeV tracks (1-7) to candidate in < 0.2

● , using weighting of tracks

● determination using energy flow method

● Reconstruction of subcluster

● Calorimeter seed: Cone4TopoJets with > 10 GeV not matched with track-seed

● Associates loose quality tracks with > 1 GeV in < 0.3

● (calorimetric) using H1-style calibration on cells from calo-seed

● Define the , using calo-seed

Calo Only candidate

Find matching Cone4TopoJets ( > 10 GeV < 0.2) as calo-seed?

Track Only candidateCalo+Track candidate

NONOYESYES

Two reconstruction algorithms merged now

pT

pT

pTpT

R

R

ET

ET

ET

ET

0

R






Tau identificationTau identification


● Distribution from calorimeter and track based variables are used to identify taus

● Many ID methods available:– Cut based (later more to this topic)

– Logarithmic Likelihood (Llh)– Neural Network (NN)

– Boosted Decision Tree (BDT)– Probability Density Estimators with Range search

(PDRS)





Tau identificationTau identification


● In early data taking taus in W/Z decay and tt production

● Main background: QCD jets

Overall reconstruction and identification efficiency





Motivation for cut-based approachMotivation for cut-based approach


● Two approaches:Two approaches:

– Calo based approach:Calo based approach:● Safe approach, calorimeter has to (and Safe approach, calorimeter has to (and

will be) understood at very early stage of will be) understood at very early stage of data takingdata taking

● Requirement: variable safe according to Requirement: variable safe according to calorimeter experts and not highly calorimeter experts and not highly correlatedcorrelated

– Calo+Track based approach:Calo+Track based approach:● More “aggressive” approachMore “aggressive” approach● Same calo variables as calo based Same calo variables as calo based

approach plus additional track variablesapproach plus additional track variables

Calo based approach Calo based approach

● Radius in EM calorimeterRadius in EM calorimeter● Isolation FractionIsolation Fraction● Width in strip layerWidth in strip layer● Et(EM)/EtEt(EM)/Et

Calo + Track based approachCalo + Track based approach

● 4 calo based variables4 calo based variables● Width of track momentaWidth of track momenta● Et/pT (Lead track)Et/pT (Lead track)● Et(Had)/sum_pTEt(Had)/sum_pT● Et(EM)/sum_pTEt(EM)/sum_pT● Sum_pT/EtSum_pT/Et





IntermezzoIntermezzo


Nomenclature:

Interaction point: x = y = z = 0x is pointing upwardsy is pointing to the center of the LHCz is pointing in anticlockwise direction

azimuthal anglepolar anglepseudorapidity is defined as:

distance is defined as:

transverse momentum and transverse energy are defined in the x-y plane

=−ln tan

2

R R=2

2

pT ET





Calo-based variablesCalo-based variables


Radius in EM CalorimeterRadius in EM Calorimeterruns over all cells inruns over all cells inthe EM in a clusterthe EM in a clusterwith dR<0.4with dR<0.4

Isolation FractionIsolation Fractionring of 0.1<dR<0.2 as isolation regionring of 0.1<dR<0.2 as isolation regioni and j run over all EM cells in a conei and j run over all EM cells in a conewith 0.1<dR<0.2 and dR<0.4 respectivelywith 0.1<dR<0.2 and dR<0.4 respectively

Width in strip layerWidth in strip layertransverse energy width in striptransverse energy width in striplayer (first layer of EM barrel)layer (first layer of EM barrel)

Et(EM)/EtEt(EM)/EtFraction of transverse energy in theFraction of transverse energy in theelectromagnetic calorimeter and the totalelectromagnetic calorimeter and the totaltransverse energytransverse energy

Rem=

∑i=1

n

ET , ii−cluster 2i−cluster

2

∑i=1

n

ET ,i

ET12=

∑i

ET ,i

∑j

ET , j

=∑i=1

n

ET ,istrip

i−cluster 2

∑i=1

n

ET ,istrip

ETEM=

ETEM

ETtotal





Track-based variablesTrack-based variables


Width of track momentaWidth of track momentawidth of tracks, weighted with width of tracks, weighted with their transverse momentatheir transverse momenta

Et over pT of the leading trackEt over pT of the leading tracklarge fraction of the tau energy is expected to belarge fraction of the tau energy is expected to becarried by the leading track => ratio of Et and pTcarried by the leading track => ratio of Et and pTof leading track small, close to 1of leading track small, close to 1

Et(EM/Had)/sum_pTEt(EM/Had)/sum_pTfraction of transverse energy in the electromagneticfraction of transverse energy in the electromagnetic(hadronic) calorimeter and total transverse(hadronic) calorimeter and total transversemomentummomentum

sum_pT/Etsum_pT/Etfraction of total transverse momentum and totalfraction of total transverse momentum and totaltransverse energytransverse energy

W tracks =

∑ ,track 2⋅pTtrack

∑ pTtrack

−∑ ,track⋅pT

track2

∑ pTtrack

2

ETpT , 1=

ETtotal

pT , 1

ETEM /Had

=ET

EM /Had

pTtotal

pT=pT

total

ETtotal





Optimization of Safe VariablesOptimization of Safe Variables


● Signal: Z->tautau and A->tautau with 198250 events

● Background: several di-jet samples with 459650 events

● Rectangular cuts (Genetic Algorithm with TMVA)

– Classifier only returns a binary response

– Maximizes background rejection for a given signal efficiency

– Provides optimal cut

● TMVA output used only as proposal, final cut values estimated by human

● Optimizing in different bins and selections for both approaches:

– 3 cut selections: loose, medium, tight (Eff.: 0.7, 0.5, 0.3)

– 1-prong and 3-prong (multi-prong for background)

– 5 pT-bins: 10-25 GeV, 25-45 GeV, 45-70 GeV, 70-100 GeV, >100 GeV





Safe Variables for calo approachSafe Variables for calo approach


EMRadius:taus have a smaller transverse shower profile than QCD-jetsThis is the strongest variable

Clusters from had. decaying taus are well collimated => tighter isolation criteria are used

Tight cutsMedium cutsLoose cuts





Safe Variables for calo+track approachSafe Variables for calo+track approach


Et over pT of leading track:large fraction of the energy is expected to be carried by leading track => ratio expected to be large, close to 1. QCD-jets are more uniform distributed





Track distributionTrack distribution


Z->tautau andZ->tautau andA->tautau signalA->tautau signal

di-jetdi-jetbackgroundbackground





Optimal cut values, Calo approachOptimal cut values, Calo approach


Efficiency StripWdith2 <= Efficiency StripWdith2 <=

0,7 0,11 0,047 0,75 1,02 0,7 0,21 0,057 0,96 1,02

0,5 0,072 0,045 0,65 1,02 0,5 0,15 0,056 0,90 1,02

0,3 0,055 0,033 0,32 1,01 0,3 0,096 0,048 0,62 1,01


0,7 0,087 0,048 0,89 0,0010 0,7 0,15 0,045 0,79 1,01

0,5 0,058 0,043 0,24 0,0011 0,5 0,088 0,044 0,68 1,01

0,3 0,048 0,0030 0,20 0,023 0,3 0,068 0,044 0,42 1,01


0,7 0,081 0,050 0,31 0,00040 0,7 0,25 0,031 0,25 1,08

0,5 0,050 0,035 0,26 0,00040 0,5 0,071 0,031 0,25 1,03

0,3 0,037 0,035 0,26 0,0028 0,3 0,053 0,026 0,25 1,01


0,7 0,15 0,040 0,70 0,0010 0,7 0,23 0,036 0,50 1,02

0,5 0,045 0,036 0,14 0,029 0,5 0,061 0,035 0,20 1,01

0,3 0,045 0,015 0,058 0,029 0,3 0,048 0,035 0,18 1,00


0,7 0,50 0,030 0,60 0,00033 0,7 0,071 0,038 0,47 1,00

0,5 0,034 0,030 0,60 0,00033 0,5 0,061 0,037 0,20 1,00

0,3 0,034 0,00069 0,049 0,0015 0,3 0,036 0,030 0,18 0,99

CaloOnly, 1prong, 1025 GeV CaloOnly, 3prong, 1025 GeV

EMRadius <= IsoFrac <= EtEM/Et <= EMRadius <= IsoFrac <= EtEM/Et <=


EMRadius <= IsoFrac <= EtEM/Et > EMRadius <= IsoFrac <= EtEM/Et <=





CaloOnly, 1prong, >100 GeV CaloOnly, 3prong, >100 GeV


Background: 3-prong cuts applied on multi-prong (track multiplicity >= 2)





Performance calo approachPerformance calo approach


● Good rejection power for 1-prong taus (left)

● 1-2 magnitudes worse for 3-prong taus (right)

● Likelihood is official logarithmic likelihood for tau reconstruction (about 20 variables)

● Compared with cutting only on EMRadius (strongest variable)





Performance calo+track approachPerformance calo+track approach






● Taus play an important role in SM physics and search for the Higgs boson and MSSM

● Large number of taus expected in early data

● Reconstruction of hadronically decaying taus (1-prong and 3-prong)

● Collimated CaloCluster, low track multiplicity

● Fakes: electrons, muons, QCD-jets

● Two algorithm:

– Calorimeter as seed

– Tracks as seed

● Tau identification of early data with two cutbased approaches:

– Calobased uses only calorimeter information

– Calo+Trackbased uses calorimeter and tracking information

SummarySummary





Documents

Tau reconstruction at the ATLAS experiment– Electric field: 5000 V/mm ... Björn Gosdzik • DESY HH • [email protected] Student Seminar 22.01.09 11 Motivation Tau reconstruction