Hadron Collider Physics symposium 2006 Duke 26 May 2006 Tracking and vertexing at ATLAS

Hadron Collider Physics symposium 2006Pamela Ferrari 1Hadron Collider Physics symposium 2006Pamela Ferrari 1

Hadron Collider Physics symposium 2006Hadron Collider Physics symposium 2006

Duke 26 May 2006

Tracking and vertexing at Tracking and vertexing at ATLASATLAS

P.Ferrari (CERN)


Magnet system ( 2Tesla)

The ATLAS DetectorMuon Spectrometer

CalorimeterInner Detector


Forward SCT

Transition Radiation straw tube Tracker

Barrel SCT

Pixel Detectors

(3 layers + 6 disks)

The ATLAS Inner Detector


pixel:pixel:Barrel: 3 layers of Silicon detectors (average R=5.05,8.85,12.25

cm).endcap:3 disks of Silicon pixel detectors. For some simulation the intermediate layer is not present for

historical reasons ( 2-layer layout)

SCT:SCT:8 layers of semiconductor tracker SCT in the barrel and 9 disks

per side in the EndCaps ( stereo strip detectors)

TRT:TRT:Several layers of 4 mm straws in the barrel region ( arranged in

3 layers of modules). 14 Transition Radiation Tracker wheels in the

endcap ~30 hits per track

The Inner DetectorThe Inner Detector

points

(R) (m) (Rz) (m)

pixel 3 12 60

SCT 4 17 580

TRT 36 170 -


Magnetic field and material Magnetic field and material knowledgeknowledge

The magnetic field ~2T in IDThe magnetic field ~2T in ID B= 1T at the end of solenoidB= 1T at the end of solenoid B= 0.8 T at end of IDB= 0.8 T at end of ID

more complicated tracking algsmore complicated tracking algs track resolution degradation on track resolution degradation on

mainly at high pmainly at high pT T

on 1/pon 1/pTT for | for || >1.5| >1.5 on don d00 small effect since B field small effect since B field

is uniform around Int. Pointis uniform around Int. Point

Knowledge of material in the Knowledge of material in the detectordetector To reconstruct tracks one needs To reconstruct tracks one needs toto

take into account:take into account:• multiple scattering multiple scattering

• energy loss in materialenergy loss in material


Number of tracks per Number of tracks per eventevent

At high luminosity per each At high luminosity per each

bunch crossing ( 25 ns)bunch crossing ( 25 ns)

more than 200 tracksmore than 200 tracks about 15-20 vertex about 15-20 vertex

candidatescandidates

Complex task for tracking andComplex task for tracking and

Vertexing because of pile-up.Vertexing because of pile-up.

Triggering algorithms have to Triggering algorithms have to be be

fast and robust to avoid to miss fast and robust to avoid to miss

rare eventsrare events

ℒℒintint/y (fb/y (fb-1-1)) ℒ ℒ (cm(cm22/s)/s) s (TeV)s (TeV) Minimum Minimum bias/bco bias/bco

LHC ( low ℒ) LHC ( low ℒ) 1010 2x102x103333 1414 55

LHC (high LHC (high ℒ) ℒ)

100100 10103434 1414 2525


ATLAS triggerATLAS trigger

LVL1 trigger:LVL1 trigger:

-hardware trigger (2.5-hardware trigger (2.5s latency)s latency)

-calorimeter + muon chambers.-calorimeter + muon chambers.

-Defines Regions Of Interest (ROI)-Defines Regions Of Interest (ROI)

LVL2:LVL2: processing in parallel info from processing in parallel info from

ROI, uses ID information (latency ROI, uses ID information (latency 10ms)10ms)

Event Filter:Event Filter:

uses tools similar to “offline” code uses tools similar to “offline” code (thanks to longer latency ~1s)(thanks to longer latency ~1s)

Challenge:Challenge: have traking and b-tagging at have traking and b-tagging at

trigger level trigger level speed! speed!

HLT:software triggersHLT:software triggers


Tracks seeds formed by fitting with a straight line pairs of Tracks seeds formed by fitting with a straight line pairs of space points in pixel B layer and in second logical layer (in a space points in pixel B layer and in second logical layer (in a given RoI).given RoI).

- Tracks extrapolated back to beam line - Tracks extrapolated back to beam line impact parameter impact parameter (IP)(IP)

- Track retained if IP is small in transverse plane.- Track retained if IP is small in transverse plane.

The Z coordinate of the primary vertex = maximum of the The Z coordinate of the primary vertex = maximum of the histogram filled with the z intersection of the seeds with the histogram filled with the z intersection of the seeds with the beam line.beam line.

Third space point is extracted in modules situated in positions Third space point is extracted in modules situated in positions where the hits may lay ( LookUpTables). Space points where the hits may lay ( LookUpTables). Space points compatible with linear extrapolation of track extend the seed.compatible with linear extrapolation of track extend the seed.

30 30 m at high pm at high pTT

Tracking algorithm at Tracking algorithm at LVL2LVL2

remove ambiguities due to overlapping remove ambiguities due to overlapping space space points in triplets with extrapolation points in triplets with extrapolation qualityquality

Triplets fitted and identified with tracksTriplets fitted and identified with tracks

- Efficiency for tracks in jets 80-90% - Efficiency for tracks in jets 80-90% depending depending on luminosity and event topology on luminosity and event topology - 95% for single electrons.- 95% for single electrons.


B-tagging at LVL2B-tagging at LVL2

b-jet selection is performed by using impact parameters b-jet selection is performed by using impact parameters significance (S=dsignificance (S=d00//(d(d00), where ), where (d(d00) dependence from p) dependence from pTT is is obtained from simulation) obtained from simulation)

Secondary vertex algorithm similar to offline but fasterSecondary vertex algorithm similar to offline but faster b-jet estimator uses likelihood ratiob-jet estimator uses likelihood ratio

WH (mWH (mHH=120 GeV/c=120 GeV/c22), low ), low luminosityluminosity

Timing: Timing: 3 ms per RoI (track 3 ms per RoI (track

rec.) +rec.) + < 2 ms Sec. Vtx rec< 2 ms Sec. Vtx rec..

R ~ 25 (15) for εR ~ 25 (15) for εbb=50%(60%)=50%(60%)

Impact parameters d0 & z0

Impact parameters + Sec. Vertex

trN

1i i

ijet )u(S

)b(SlnW


Track extrapolation Track extrapolation ingredientsingredients

1.1.The first step is the The first step is the geometrical transport geometrical transport of the track of the track parameters and their parameters and their covariance matrices to covariance matrices to a given detector a given detector surfacesurface

2.2. The second procedure The second procedure is the update of the is the update of the propagated propagated parameters and errors, parameters and errors, taking multiple taking multiple Coulomb scattering Coulomb scattering and energy loss effects and energy loss effects during the propagation during the propagation process into account.process into account.


ATLAS offline tracking ATLAS offline tracking algorithmsalgorithms

The xKalman algorithm:The xKalman algorithm: finding space points defining primary finding space points defining primary trajectories in SCT & pixel. Kalman filter associates clusters to trajectories in SCT & pixel. Kalman filter associates clusters to tracks.In TRT reconstruct track in a narrow region around the tracks.In TRT reconstruct track in a narrow region around the extrapolated trajectory retaining all hits in that region.extrapolated trajectory retaining all hits in that region. Kalman fitter= track fitting with Gaussian noise and all Kalman fitter= track fitting with Gaussian noise and all

measurements and material effects are approximately measurements and material effects are approximately GaussianGaussian

The iPatRec algorithm:The iPatRec algorithm: form track-candidates using space-point form track-candidates using space-point combinatorials subject to criteria on maximum curvature and combinatorials subject to criteria on maximum curvature and crude vertex region projectivity. crude vertex region projectivity.

Global χ2 fitter used to fit tracks and associate clusters.Global χ2 fitter used to fit tracks and associate clusters.

Only good tracks are retained for extrapolation in TRT, where Only good tracks are retained for extrapolation in TRT, where TRT hits are added.TRT hits are added. Using global χUsing global χ22 fitter = minimises track χ fitter = minimises track χ22 by considering all by considering all

measurements simultaneouslymeasurements simultaneously


New TrackingNew Tracking NewTracking algorithm: Logical reorganization of

tracking code. Largely based on xKalman, but in the future it will combine as well some tools from iPatRec optimised tracking algorithm.

Can be used at event filter level, offline and for the Combined testbeam and Cosmics runs.

Uses better geometry description builtfrom full geomodel easy development of new code


Comparison of resultsComparison of results Using tt events

xkalmanxkalman iPatReciPatRec newTrackinnewTrackingg

Multiplicity (P>1 GeV)Multiplicity (P>1 GeV) 16.6916.69 17.0617.06 16.8816.88

Barrel Track eff/Barrel Track eff/ fake ratefake rate 99%/99%/0.6%0.6% 99%/99%/0.70.7 96%/96%/2.5%2.5%

Transition eff/Transition eff/ fake ratefake rate 98%/98%/0.6%0.6% 98%/98%/0.5%0.5% 96%/96%/3.6%3.6%

Forward eff/Forward eff/ fake ratefake rate 98%/98%/0.3%0.3% 99%/99%/1.3%1.3% 95%/95%/2.7%2.7%

iPatrec

σ(z0

)σ(d0

)

WH(400 GeV/cWH(400 GeV/c22) W() W(μνμν))H(H(uu)uu)

iPatrec


Averaged over all η

η~0

Momentum Resolution vs PMomentum Resolution vs PTT

PT(GeV)

Resolutions obtained here using iPatRec (xKalman gives same results)

Single μ± ( DC1 / 2 pixel layer layout pT = 1, 5, 20,100, 1000 GeV/c)


Cosmics with SCT & TRTCosmics with SCT & TRT

the SCT & TRT barrel are integrated on the surfacethe SCT & TRT barrel are integrated on the surface We are having cosmics data taking since the 9We are having cosmics data taking since the 9thth of May of May

We expect to We expect to collect 300K of collect 300K of cosmics until mid cosmics until mid of Juneof June

remember that we remember that we have still a non-have still a non-aligned detector.aligned detector.

The alignment The alignment precision precision

is given by the is given by the module module

placement precision placement precision on the barrelon the barrel


View from outside towards

Side A

Cosmics data takingCosmics data taking

SCT:SCT:• Read 504 modules grouped Read 504 modules grouped

in a sector at the top and in a sector at the top and another one at the bottomanother one at the bottom

• The bottom sector is not The bottom sector is not fully cabled up will be fully cabled up will be ready the the 22ready the the 22ndnd of May of May

TRT:TRT: • Read 2 sectors in top +2 Read 2 sectors in top +2

sectors in bottomsectors in bottom


Cosmics with SCT & TRTCosmics with SCT & TRT

First Cosmics tracks in top sectorWe aim to use the cosmics to do a first exercise on the

alignment

Present alignment == module positioning precision


e/e/ separation using the TRT separation using the TRT

Typical TR photon energy depositions in TRT ~ 8-10 keV pions deposit ~ 2 keV

Electron identification using large energy depositions due to Electron identification using large energy depositions due to transition radiation (X-rays) when they traverse radiators between transition radiation (X-rays) when they traverse radiators between TRT strawsTRT straws

Results from TB 2002Results from TB 2002 @20 @20 GeVGeV

Results from CTB2004


Primary vertex Primary vertex reconstructionreconstruction

Large multiplicity of tracksLarge multiplicity of tracks ( several hundreds as we ( several hundreds as we have seen) have seen) vertex reconstruction must be fast vertex reconstruction must be fast

Input needed consists of 3D trajectory & error matrix Input needed consists of 3D trajectory & error matrix of tracks. Quality requirements on track are appliedof tracks. Quality requirements on track are applied

Approximate primary vertex position in Z:Approximate primary vertex position in Z: sliding sliding window of 0.7 cm is moved along all interaction region. window of 0.7 cm is moved along all interaction region. The window with largest number of tracks weighted The window with largest number of tracks weighted with pwith pTT is chosen. is chosen.

The <z> is the mean obtained by all the tracks in that The <z> is the mean obtained by all the tracks in that windowwindow

Tracks belonging to primary vertex are taken away and Tracks belonging to primary vertex are taken away and the procedure is iterated to get other (pile-up) vertices.the procedure is iterated to get other (pile-up) vertices.


Primary vertex reconstruction Primary vertex reconstruction cont’dcont’d

All tracks at ± 5mm in z and ± 1mm in transverse All tracks at ± 5mm in z and ± 1mm in transverse plane are accepted as coming from primary vertexplane are accepted as coming from primary vertex

At this point the vertex fitting is performed using a At this point the vertex fitting is performed using a Billoir method: Billoir method: if the if the 22 is too high, the tracks that give is too high, the tracks that give too high too high 22 are rejected and everything is recalculated ( are rejected and everything is recalculated ( outliers are removed)outliers are removed)

There are two different implementation of this method There are two different implementation of this method which are basically using the same strategy: which are basically using the same strategy: • VxPrimaryVxPrimary

• VKalVrtVKalVrt


Primary vertex with Primary vertex with IPatRecIPatRec

WHWHbbbbevents mevents mHH=120 GeV/c=120 GeV/c22

x[x[m]m] z[z[m]m]

VxPrimary VxPrimary 12.6 12.6 ± 0.1± 0.1 50.0 50.0 ± ± 0.50.5

VKalVrtVKalVrt 10.8 10.8 ± 0.1± 0.1 42.7 42.7 ± ± 0.40.4


New Adaptive Vertex FitterNew Adaptive Vertex Fitter

The “Adaptive Vertex Fitter” solves the problem The “Adaptive Vertex Fitter” solves the problem

of outlier tracks that spoil the fit, not by of outlier tracks that spoil the fit, not by

discarding them, but by down-weighting them.discarding them, but by down-weighting them.

Minimises instead than residuals, weighted sum Minimises instead than residuals, weighted sum

of squared residuals (weight depending on of squared residuals (weight depending on 22))

(10000 events of WH(120) with H->bb)(10000 events of WH(120) with H->bb)

x ( x ( m)m) z ( z ( m)m)

AVFAVF 11.07±0.0911.07±0.09 46.76±0.0546.76±0.05

VKalVrtVKalVrt 11.07±0.0911.07±0.09 45.43±0.0545.43±0.05


B-Tagging methods

BBa0<0

a0>0

x

y

Secondary Vertex

Primary vertex

Jet axis

Soft lepton

1.1. Based on lifetime of b-hadrons jets, high multiplicity of b-jetsBased on lifetime of b-hadrons jets, high multiplicity of b-jets• Impact parameter of tracksImpact parameter of tracks• Secondary vertexSecondary vertex

2.2. Soft-lepton tagging: development ongoingSoft-lepton tagging: development ongoing• Low pT electron from B (D)Low pT electron from B (D)• Low pT muon from B (D)Low pT muon from B (D)

Key ingredients:Key ingredients:• tracking (IP resolution, tracking (IP resolution, PrimaryVertex)PrimaryVertex)• jets (axis)jets (axis)


Use normalised SUse normalised S= = dd00//dd00

for each trackfor each track

compare it to predefined compare it to predefined calibration p.d.f. for the calibration p.d.f. for the b and light q hypothesis: b and light q hypothesis: get probabilities b(S) get probabilities b(S) and u(S)and u(S)

IP in transverse planeIP in transverse plane

sum over all tracks jet btag weight

trN

1i i

ijet )u(S

)b(SlnW


3D Impact Parameter3D Impact Parameter

Improvement can be obtained by combining theImprovement can be obtained by combining the longitudinal longitudinal and the transverse significance. and the transverse significance.

W=PW=Pbb(S(Sdd00

,S,Szz00)/P)/Puu(S(Sdd00

,S,Szz00))

b jet u jet


Secondary vertex searchSecondary vertex search

1. Track selection with quality cut:1. Track selection with quality cut:(Typically pT > 1 GeV/c, |(Typically pT > 1 GeV/c, |ηη| < 2.5; |d0| < 1 mm, |z0| < 1.5 mm; NPixB | < 2.5; |d0| < 1 mm, |z0| < 1.5 mm; NPixB > 0, NPix > 1, NSi > 6)> 0, NPix > 1, NSi > 6)

2. Search for good 2-track vertices in jet2. Search for good 2-track vertices in jet

3. At this point one can remove V0s, identified interaction 3. At this point one can remove V0s, identified interaction with material,…with material,…

4. Common (inclusive) vertex for remaining tracks4. Common (inclusive) vertex for remaining tracks

Ks Λbeam-pipe,pixel layers


Final jet tagging weightFinal jet tagging weight

Input variables have to be independent from flight

distance

E(SVX)/E(jet)

SVX mass

# 2-track vertices

+ IP3D

1 2D variable

1D variable

b jet u jet

b jet u jet

b jet u jet


Taggers availableTaggers available

1st stream 2nd stream

IP (long.

impact)Lifetime1D

(trans. impact)

IP2D Lifetime2D

IP3D Lifetime3D

Inclusive Secondary

Vertex

SV1 SecVtxBU

SV2 SecVtxTD

Pre-definedcombinatio

n

VKalVrt=VKalVrt=“weight”: IP3D + SV1

lhSig=lhSig= Lifetime1D

+Lifetime2D+SecVtxBU

Different taggers are used as Different taggers are used as cross-check since they are cross-check since they are almost identical wrt almost identical wrt discriminating variables:discriminating variables:

Lifetime2D ~ IP2D Lifetime2D ~ IP2D Lifetime3D ~ IP3DLifetime3D ~ IP3D

Slight differences:Slight differences:

• refined track selection in refined track selection in IPxD, IPxD, • one 2D vs one 1D pdf for IP3D one 2D vs one 1D pdf for IP3D vs Lifetime3Dvs Lifetime3D


PerformancesPerformances

Labelling of jets:Labelling of jets: label a jet as a b-jet if there is a b-quark within label a jet as a b-jet if there is a b-quark within ΔΔR<0.3R<0.3..efficiency efficiency εεbb : (# b jets )/(#jets labelled as b with p: (# b jets )/(#jets labelled as b with pTT>15 GeV/c, |>15 GeV/c, |ηη||<2.5 )<2.5 )light-jet rejection:light-jet rejection: Ru= 1 / Ru= 1 / εεuu

Overlapping jets and purification:Overlapping jets and purification: Overlaps in jets Overlaps in jets mislabelling mislabellingJet isolation very dependent on the type of events and physics Jet isolation very dependent on the type of events and physics processes processes (gluon jets) + jet (gluon jets) + jet algorithmalgorithm

Purification to factorize it from Purification to factorize it from pure b-tagging issuespure b-tagging issues

do not consider lights jets do not consider lights jets

whenwhenthere is a b/c/quark/hadron there is a b/c/quark/hadron

within within ΔΔR<0.8R<0.8

Using WH events mUsing WH events mHH = 120 = 120 GeV/cGeV/c22 2 layer-layout, xKalman 2 layer-layout, xKalman tracks, Cone 0.4 jetstracks, Cone 0.4 jets

IP2DIP2D

IP3DIP3D

SV1+IP3DSV1+IP3D

SV2+IP3DSV2+IP3D


Shared hits and bad tracksShared hits and bad tracks

b-tagging is obviously very demanding for track quality.b-tagging is obviously very demanding for track quality.One might try to ‘clean them up’One might try to ‘clean them up’

1)1) Tracks in jet may share some hits, resulting in lower Tracks in jet may share some hits, resulting in lower quality tracks: special treatment, by either rejecting quality tracks: special treatment, by either rejecting them, or using dedicated calibrations. Fraction in b-jets:them, or using dedicated calibrations. Fraction in b-jets:• tt events tt events 3.5%3.5%• WH events (400 GeV) 8.5%WH events (400 GeV) 8.5%

2)2) Tracks may originate from V0 Tracks may originate from V0 or interaction with material.or interaction with material.They usually have “more They usually have “more lifetime” lifetime” reject them reject them ((bad bad trackstracks))Fraction of V0 tracks in b-jets: Fraction of V0 tracks in b-jets: • tt tt 1.2% 1.2%• WH(400 GeV)WH(400 GeV) 3.6% 3.6%

in light jets


B-tagging performanceB-tagging performance

IP2DIP2D IP3DIP3D IP3D+SV1IP3D+SV1

efficiencyefficiency 50%50% 60%60% 50%50% 60%60% 50%50% 60%60%

Rej:just taggerRej:just tagger

VKalVrtVKalVrt135±9135±9 55 ±255 ±2 214 ±18214 ±18 75 ±475 ±4 609 ±86609 ±86 157 ±11157 ±11

Rej:just taggerRej:just tagger

AVFAVF 130 ±9 130 ±9 52 ±252 ±2 205 ±17205 ±17 73 ±4 73 ±4 612 ±87612 ±87 147 ±10147 ±10

Rej:bad tracks +Rej:bad tracks +

VkalVrtVkalVrt206±17206±17 69 ±369 ±3 339 ±35339 ±35 101 ±6101 ±6 815 ±134815 ±134 192 ±15192 ±15

Rej:bad tracks +Rej:bad tracks +

AVFAVF 199 ±16 199 ±16 66 ±366 ±3 327 ±34327 ±34 98 ±6 98 ±6 794 ±129794 ±129 164 ±12164 ±12

B-tagging performance using different primary vertex B-tagging performance using different primary vertex finders.finders.

• WH>uuWH>uu, m, mH H = 120 GeV/c xKalman= 120 GeV/c xKalman

• Geometry for this study: Final Layout for pixels (3 layers/disks) Geometry for this study: Final Layout for pixels (3 layers/disks)

Physics performance limited by gluon splittingPhysics performance limited by gluon splitting


b-tagging in ttbar eventsb-tagging in ttbar events

190 K tt ttbar, cone ΔR=0.4, iPatrec tracks 60k ttH ( m60k ttH ( mHH=120 GeV)=120 GeV) cone ΔR=0.4, iPatrec tracks (2-layer

layout)

RRuu ( (εεbb=60%)=60%) RRuu ( (εεbb=50%)=50%)

ttbar SV1+IP3D ttbar SV1+IP3D 259 259 ±7.8±7.8 858858±42.9±42.9

ttbarttbar

SV1+IP3D + SV1+IP3D + shared hits shared hits

326 326 ±9.8±9.8 1133 1133 ±56.6±56.6

ttbar eventsttbar events

SV1+IP3D + SV1+IP3D + shared hits +shared hits +

bad tracks bad tracks

375 375 ±11.2±11.2 1326 1326 ±66.3±66.3

ttH events lhSigttH events lhSig 313 313 ±9.4±9.4 1392 1392 ±56.6±56.6


ConclusionsConclusions

There has been a lot of work/improvement on There has been a lot of work/improvement on tracking and vertexing in the past year(s).tracking and vertexing in the past year(s).

Tracking algorithms are available at LVL2, Event Tracking algorithms are available at LVL2, Event filter, offline, for cosmic running, combined testbeam filter, offline, for cosmic running, combined testbeam etc..etc..

Different parallel software developments for the Different parallel software developments for the tracking and vertexing algorithms have been tracking and vertexing algorithms have been produced, giving comparable resultsproduced, giving comparable results

We are already reconstructing cosmic events with We are already reconstructing cosmic events with the SCT and TRT barrelsthe SCT and TRT barrels

We are looking forward to the commissioning of all We are looking forward to the commissioning of all those tools with the final detector.those tools with the final detector.


Back-up


Tracking the basics:Tracking the basics:

1.1. Pattern recognition:Pattern recognition: finding hits in SCT and pixel and then finding hits in SCT and pixel and then make a fast fit to extrapolate tracks to TRT to find TRT hits. make a fast fit to extrapolate tracks to TRT to find TRT hits. At any stage the effect of the magnetic field is taken into At any stage the effect of the magnetic field is taken into accountaccount

2.2.Track fitting:Track fitting: uses the list of hits that the pattern recognition uses the list of hits that the pattern recognition associates to a track, and fits the track. It needs as input the associates to a track, and fits the track. It needs as input the track parameters at the perigee ( point of closest approach track parameters at the perigee ( point of closest approach to the z axis for the track). There are 5 parameters: f0,q0,d0, to the z axis for the track). There are 5 parameters: f0,q0,d0, z0 and q/p. The track fit can correct for energy loss and z0 and q/p. The track fit can correct for energy loss and multiple scattering for each scattering plane.multiple scattering for each scattering plane.

3.3.Residuals:Residuals: difference between the track prediction and the difference between the track prediction and the hit hit

4.4.Track parameters pulls:Track parameters pulls: a measure of the reliability of the a measure of the reliability of the track fit are the pull distributions (rec-tru)/(error on rec) for track fit are the pull distributions (rec-tru)/(error on rec) for the 5 track parameters.the 5 track parameters.


Comparison of resultsComparison of results

Using 100 tt events

xkalmanxkalman iPatReciPatRec newTrackingnewTracking

Multiplicity (P>1 GeV)Multiplicity (P>1 GeV) 16.6916.69 17.0617.06 16.8816.88

Barrel Track eff/Barrel Track eff/ fake fake raterate

99%/99%/0.6%0.6% 99%/99%/0.7%0.7% 96%/96%/2.5%2.5%

Transition TrackTransition Track eff/eff/ fake ratefake rate

98%/98%/0.6%0.6% 98%/98%/0.5%0.5% 96%/96%/3.6%3.6%

ForwardTrack eff/ForwardTrack eff/ fake fake raterate

98%/98%/0.3%0.3% 99%/99%/1.3%1.3% 95%/95%/2.7%2.7%

BarrelBarrel

# hits# hits Pixel/ Pixel/SCTSCT/TRT/TRT2.9/2.9/8.1/8.1/28.128.1 2.9/2.9/8.0/8.0/27.527.5 2.9/2.9/7.9/7.9/28.928.9

TransitionTransition


ForwardForward



Impact parameter resolutions

WH(400 GeV) W(μν)H(uu)iPatrec

iPatrec xKalman

xKalman

σ(z0

)

σ(d0

)

WH(400 GeV/cWH(400 GeV/c22) W() W(μνμν))H(H(uu)uu)


Primary vertex with Primary vertex with xKalmanxKalman

x[x[m]m] z[z[m]m]

VxPrimary VxPrimary 13.0 13.0 ± 0.2± 0.2 51.4 51.4 ± ± 0.60.6

VKalVrtVKalVrt 11.4 11.4 ± 0.1± 0.1 47.4 47.4 ± ± 0.50.5


Shared hits: IP distributions

Typical criteria:• ‘’Good’’ track: no shared pixel AND < 2 shared SCT hits• ‘’Shared’’ track: the rest

b-jets light jets

Rome samples Fraction of tracks with shared hits

b-jets light jets

WH(400 GeV) 8.5% 4.4%

ttH, ttbb, tt(jj) 3.5% 1.4%


Performances: ttH vs ttjjPerformances: ttH vs ttjj

Complicated/busy events due to overlaps and Complicated/busy events due to overlaps and mislabellings.mislabellings.

Purification done by factorising these effects to disentagle them from b-Purification done by factorising these effects to disentagle them from b-tagging tagging b-jets: ttH Pythia (samples 4867, 4868) b-jets: ttH Pythia (samples 4867, 4868) u-jets: tt(jj) MC@NLOu-jets: tt(jj) MC@NLO cone cone ΔΔR=0.4, iPatrec tracksR=0.4, iPatrec tracks Statistics: 75k b-jets, 1.2M u-jetsStatistics: 75k b-jets, 1.2M u-jets

as a function ofas a function of ||| and p| and pTT

@ 60%@ 60% efficiencyefficiency

R @ R @ b b 50%50% R@ R@ b b 60%60% R@ R@ b b 70%70%

IP2DIP2D 218 218 ± 3± 3 66 66 ± 1± 1 2323

SV1+IP3DSV1+IP3D 882 882 ± 24± 24 297 297 ± 5± 5 5959


Soft Electron Tagging

use Soft Electron identification variables to build a probability for each track in a jet the track with the highest probability is the “electron candidate”

light jet rejection vs algorithm efficiency :

@ 60% algorithm efficiency (i.e. 0.6*BR(beX) ~7.8% b-jet efficiency)

Ru = 134 (WH mH=120 GeV events)

Pions

Electrons


Impact of (mis)alignment

Random misalignments:

IP3D tagger , ttH events, realistic conditions

will redo the exercise with misaligned detectors from simulation (more realistic than random misalignments)


Use CDF experience : Use CDF experience :

Map CDF commissioning misalignments from CDF run II to ATLASMap CDF commissioning misalignments from CDF run II to ATLAS and propagate to b-tagging performancesand propagate to b-tagging performances

~ 10% loss

Influence of alignment (II)


Performances versus jet pT, η

IP2DIP3D+SV1εb= 60%

IP2DIP3D+SV1εb= 60%

Non-uniform performances: tagging b-jets can bias kinematics

How to improve bad regions ?:• large pseudo-rapidity (|η|>2): z-anolog clusters, matter descrip, interaction in disks• low pT (<50 GeV) [bbh,…]: better matter description• high pT (>200 GeV) [Susy, little Higgs,…]: tracks w/o hit in b-layer, ambiguities

A bit more subtle, being investigated now


Track ClassificationDefine track quality categories to:• reject/use dedicated calib for tracks w/ shared hits (≥1 Pix || ≥2 Sct)• reject tracks from V0 or interactions with material

Tracks w/ shared hits:

Samples Fraction of shared tracks in b-jets

tt 3.5%

WH(400 GeV) 8.5%

Tracks from V0, interactions:

Samples Fraction of tracks from V0 in b-jets

tt 1.2%

WH(400 GeV) 3.6%

LV, JBdV (CPPM)

Date post:	16-Jan-2016
Category:	Documents
Upload:	nhu
View:	23 times
Download:	0 times

Hadron Collider Physics symposium 2006 Duke 26 May 2006 Tracking and vertexing at ATLAS

Documents