roberto crupi infn lecce & physics department, university of salento - italy on behalf of the

Como, ICATPP 2009 R.Crupi 1

Roberto CrupiRoberto CrupiINFN Lecce & Physics Department, INFN Lecce & Physics Department,

University of Salento - ItalyUniversity of Salento - Italyon behalf of theon behalf of the

ATLAS CollaborationATLAS Collaboration

11th ICATPP Conference on Astroparticle, Particle, Space Physics, Detectors and Medical Physics Applications Villa Olmo, Como 5-9 October 2009

Muon reconstruction and selection at the last trigger level of the ATLAS

experiment


OutlineOutline

• Overview on the ATLAS Trigger SystemOverview on the ATLAS Trigger System

• The Muon Event Filter The Muon Event Filter

• Performance StudiesPerformance Studies


ATLAS (A Toroidal LHC ApparatuS)ATLAS (A Toroidal LHC ApparatuS)


MDT RPC

TGC

CSC

The ATLAS Muon SpectrometerThe ATLAS Muon Spectrometer

Trigger ChambersTrigger Chambers:• RPC RPC (Resistive Plate Chambers) ||<1.05 • TGC TGC (Thin Gap Chambers ) 1.05<||<2.4

Tracking ChambersTracking Chambers: • MDTMDT (Monitored Drift Tubes) ||<2• CSC CSC (Cathode Strip Chambers) 2<||<2.7

Technologies deployed in the muon system:Technologies deployed in the muon system:

A magnetic field allows muon momentum measurementsA magnetic field allows muon momentum measurements

The Muon Spectrometer has the standalone capability of measuring muon momenta ppTT/p/pTT < 10% up to 1 TeV < 10% up to 1 TeV

ΔΔppTT/p/pTT of few % up to 100 GeV of few % up to 100 GeV


To obtain the required event rate reduction, the trigger is structured in three levels:

• Level-1Level-1 : hardware based, with coarse granularity from calorimeters and muon systems. It reduces the ~40MHz initial 40MHz initial rate to 75 kHz75 kHz in a maximum latency of 2.5 2.5 ss. It looks for regions of activity in the detector (Regions of Interest, RoIRoIs)

• Event FilterEvent Filter : software based. It uses the same algorithms as in the offline and runs after event building, then it can access the complete event data.It reduces Level-2 output rate to ~200 Hz200 Hz with a 4s4s mean processing time. It accesses the full granularity data inside a Level-2 RoI.Level-2 & EF are together referred as High Level Trigger (HLT)

• Level-2Level-2 : software based, algorithms optimized for fast rejection. It reduces Level-1 output rate to ~3 kHz3 kHz with a 40 ms mean processing time. It accesses the full granularity data inside a Level-1 RoI.

The ATLAS TriggerThe ATLAS Trigger

~75 kHz

~3 kHz

~200 Hz

2.5 s

40 ms

4 s

40 MHz

BUNCH CROSSING RATE


• The Muon Spectrometer is a powerful tool for precise measurements of high pT muons and identification of low pT muons

• The longer lever arm improves charge determination for high energy muons

• Extrapolation to the Interaction Point (IP) can better discriminate from secondary muons and the ones coming from the in-flight decay of π/K mesons.

• The Inner DetectorInner Detector (ID) performs precise track reconstruction with high efficiency for low pT muons

• Muon Identification is achieved by combining ID measurements with the ones provided by MS

• The final momentum resolution takes advantage of both ID and MS

• Isolation criteria applied to muons before reaching the MS allows a good muons discrimination from the ones deriving from jets

Reconstruction and identificationReconstruction and identificationin the muon systemin the muon system


The Muon TriggerThe Muon Trigger

Level-1 Level-1 resultresult

Region of Interest selected Region of Interest selected By By Level-1Level-1

Event Filter (EF)Event Filter (EF) performs muon Identification and Reconstruction thanks to dedicated offline packages developed in the ATHENA framework • Refines Level-2 estimates • Confirms/Discards Level-2 Hypothesis• For debug purposes EF could directly receive a seed from Level-1 RoIs

Level-2:Level-2:• muFastmuFast: standalone MS muon reconstruction, fast muon pT estimation via LUT

• muCombmuComb: combined ID/MS muon reconstruction; improvement of pT resolution wrt μFast

• muIsomuIso: Calorimeter and ID isolation

• muTilemuTile: Muon Tagging in Tile Calorimeter

Level-2 tasks:

• rejection of low pT muons• rejection of secondary muons deriving from the decay in flight of π/k• rejection of fake muon tracks generated by hits due to cavern background


Two algorithms have been implemented in muon EF:Two algorithms have been implemented in muon EF:

TrigMuonEFTrigMuonEF andand TrigMuGirlTrigMuGirl.

Both are wrappers of muon offline reconstruction tools.

They have anThey have an opposite strategy in opposite strategy in muonmuon reconstruction/identificati reconstruction/identification:on:

TrigMuonEFTrigMuonEF starts from the reconstruction in the MS MS and performs a backward extrapolation to the IPextrapolation to the IP and track combination in the inner track combination in the inner detectordetector.

TrigMuGirlTrigMuGirl starts muonmuon reconstruction reconstruction from the ID ID and extrapolates extrapolates tracks tracks to MS MS..

Muon Event FilterMuon Event Filter


TrigMuonEF and TrigMuGirl are structured inTrigMuGirl are structured in :

• Feature EXtraction (FEX)FEX) algorithms: normally RoI seeded, retrieve detector data inside RoI and look for featurelook for feature in this data.

• HYPOthesis (HYPOHYPO) algorithms: compare featurescompare features provided by by the FEXs against some hypothesisFEXs against some hypothesis and validate or reject the TE output according to the success or failure of the hypothesis

Trigger selection strategy is based on the definition of:Trigger selection strategy is based on the definition of:

• Trigger Elements (TEs)TEs) : abstract objects representing the : abstract objects representing the state of the reconstructionsstate of the reconstructions.

• Chains: Chains: sequences of several TEssequences of several TEs

• For each chain,For each chain, the trigger steering runs the trigger steering runs the sequence of the sequence of algorithms configuredalgorithms configured to produce all the output TEs at to produce all the output TEs at each step of the chaineach step of the chain.

TE

FEX

HYPO

UpdatedTE

Muon Event FilterMuon Event Filter


TrigMuonEF ImplementationTrigMuonEF Implementation• TrigMuonEFTrigMuonEF is ais a wrapper of offline packages developed for the purpose of muon reconstruction and identification:

TrigMuonEFTrigMuonEF is composed of: is composed of:

• A sequence of fourfour FEXFEX algorithms which compute physics quantities.

• Four HYPOHYPO algorithmsalgorithms after each FEX.

• Possible configurations:– Seed from Level 2

(standard)– Seed from Level 1 (for

debug)– Unseeded (full event

reconstruction)


• Segment FinderSegment Finder: selects MDT precision hits to found a segment.

• Track BuilderTrack Builder: Reconstructs muon tracks inside the MS taking advantage of its high precision tracking system and provides a precise measurement of the track parameter outside the calorimeter.

• TrackExtrapolatorTrackExtrapolator: performs a backward extrapolation to the interaction region through the calorimeters.

• It takes into account the magnetic field and corrections for energy loss and multiple scattering effects from all crossed material

• TrackCombinerTrackCombiner: Extrapolated tracks are combined with the corresponding matching tracks, if they exist, in the ID.

• MS and ID tracks are matched by forming a 2 built with the parameters of both tracks

• use a global fit of all the hits collected in both ID and MS

TrigMuonEF ImplementationTrigMuonEF Implementation


• Muon candidate starts from an ID trackMuon candidate starts from an ID track inside a Level-2 RoI, provided by EF ID algorithms.

• Tracks are extrapolated to the MS chambers.Tracks are extrapolated to the MS chambers.

• A roadA road in eta and phi is created around the around the extrapolated coordinates.extrapolated coordinates.

• The algorithm looks for hitslooks for hits around the extrapolated trackextrapolated track to fit segments.

• The best fittedThe best fitted segmentsegment is used to refineto refine the estimated track coordinatesestimated track coordinates and road in the current station.

TrigMuGirl ImplementationTrigMuGirl Implementation

TrigMuGirlTrigMuGirl

HypoAlgHypoAlg

CandidateToolCandidateTool

GlobalFitToolGlobalFitTool

ANNSelectionToolANNSelectionTool

EF ID TracksEF ID TracksIn In

Muon RoIMuon RoI

• Hits and segmentsHits and segments collected in the various stations of the MS chambers allow TrigMuGirl to improve extrapolationimprove extrapolation and the to identify muon-like candidate.

• Muon-like candidatesMuon-like candidates are selected using precalibrated ANN functions.precalibrated ANN functions.

• A global fitglobal fit, icluding ID and MS hits is applied to tracks belonging to identified muons for a further improvement of the momentum estimatefurther improvement of the momentum estimate.


• TrigMuGirl includes a dedicated tool to trigger Stable Massive Particles (SMP, e.g. in Susy R-parity violating scenario).

• Reconstruction efficiency takes advantage of hits collected from both the previous and the following Bunch Crossing.

• and mass reconstruction using RPC and MDT technologies.

• Specific R-Hadrons selection for candidates that don’t have an ID track.

• Seed search in the MS from a Level-2 trigger element.

Source 1034 rate (Hz) 1031 rate(Hz)

L2_mu40 Slow Slow

c-cbar 125 2.5 0.0024

b-bbar 123 1.65 0.0016

Wmunu 19 0.4 0.0004

t-tbar 0.2 small 0.0000

zmumu small 0.0000

SMP trigger runs in a separate

chain

(Slow) in the trigger Menu for

1031

L1: mu10L1: mu10

L2: beta/mass reco in the L2: beta/mass reco in the

barrelbarrel

EF: TrigMuGirl beta/mass recoEF: TrigMuGirl beta/mass reco

Estimated beta resolution and

reconstructed mass for single

stau

Background Rates

M=110 GeV Mtruth=110 GeV

M = 113.4 ± 9.7

GeV

TrigMuGirl: Slow Particle ModeTrigMuGirl: Slow Particle Mode


• Simulated t-tbar sample

• Association to true muon in a cone of DR<0.002 for muon EF combined algorithms.

• No pT thresholds have been applied for TrigMuonEF while the requirement of a 10GeV pT muon in the event is applied for TrigMuGirl.

• The pT resolution as a function of transverse momentum is shown for

ExtrapolatorExtrapolator

CombinerCombiner

TrigMuGirlTrigMuGirl

with respect to Level-2 muComb

ppTT resolution vs p resolution vs pTT

Muon EF PerformanceEF efficiencies wrt Level-2EF efficiencies wrt Level-2

ATLAS Preliminary


TrigMuGirlTrigMuonEF

ZZμμμμ MC sampleMC sample

Eff

icie

ncy

wrt

MC

Tru

th

eta resolution vs peta resolution vs pTT

phi resolution vs pphi resolution vs pTT

TrigMuonEFTrigMuonEF and and TrigMuGirlTrigMuGirl::

• improve significantlyimprove significantly trigger trigger rejectionrejection power due to better ppower due to better pTT resolution wrt resolution wrt

Level-2 Level-2

• have have similar performancesimilar performance in in efficiency and resolutionsefficiency and resolutions

Muon EF Performance Level-2: muFastmuFast

muCombmuComb


Results on 2008 cosmic dataResults on 2008 cosmic data

• TrigMuonEF TrackBuilder algorithm: comparison of eta () and phi (Φ) wrt to offline algorithm

=0.007, =17mrad

• Solenoidal and toroidal field on


ConclusionsConclusions• The Muon Event Filter has been designed and The Muon Event Filter has been designed and implemented to cope with the demanding requirements implemented to cope with the demanding requirements of the ATLAS trigger system in the high luminosity and of the ATLAS trigger system in the high luminosity and background enviroment at LHCbackground enviroment at LHC

• TrigMuonEF and TrigMuGirl have been successfully TrigMuonEF and TrigMuGirl have been successfully integrated in the Muon EF and constantly tested and integrated in the Muon EF and constantly tested and validatedvalidated

• TrigMuonEF and TrigMuGirl show very good and TrigMuonEF and TrigMuGirl show very good and similar performancesimilar performance

• At the startup muon selection will take advantage of At the startup muon selection will take advantage of both algorithms runningboth algorithms running


BACK UP SLIDESBACK UP SLIDES


LHCLHC• Center of Mass Energy (p-p) – 14 TeV (design)– 10 TeV (start-up)

• Instantaneous Luminosity:– L = 1031cm-2s-1

– L = 1034cm-2s-1

• Integrated Luminosity / year (@ 1034) Ldt 100 fb-1

• The cross sections of interesting physics processes are highly suppressed w.r.t. tot

– Calibrations and precision physics

• Leptonic W decay

– New Physics

• Higgs boson production

10-6

10-9

• Event rate at LHC (@ design lumi) ~1 GHz– 40 MHz (BC frequency) x 23 interactions per BC

• Max allowed acquisition rate ~200 Hz– Bandwidth ~ 300 MByte/s– Event Size ~ 1.5 MByte


10103131 Muon menu and rates Muon menu and rates

Total of muons: 1.7k -220 ~40Total of muons: 1.7k -220 ~40


Slow Particle TriggerSlow Particle Trigger


At Level-2; StauHypo: pStauHypo: pTT> 40GeV, > 40GeV, < 0.097 and m>40 GeV < 0.097 and m>40 GeV

TrigMuGirl for Slow ParticlesTrigMuGirl for Slow Particles

• Enables to select candidates when segment reconstruction is imperfect

• Efficiency for low is improved using TGC hits form next BC

• estimation using RPC- as in Level-2

• estimation using MDT

• Loop over possible

• Change TOF according to –MDT radii change accordingly

• Create MDT segments from the re-time radii

• Minimize segment 2 wrt → estimate

• Determine and Mass


New mu20_slow is in trigger menu for 1031

L1 mu10 L2 stau (B) | mu20 (EC) TrigMuGirlSlow

Estimated background rates

roberto crupi infn lecce & physics department, university of salento - italy on behalf of the

Documents

muon ef

trigger level

muon systems

muon event filter trigmuonef

high level trigger hlt

detector data

rois event filter

complete event data