Electron Identification with the ALICE TRD

Clemens Adler

Physikalisches Institut Heidelberg

For the TRD collaboration

HCP2005, Les Diablerets, July, 6 2005

ALICETRD:Identification of electrons (p>1GeV) -0.9<η<0.9

ITS

TPC

TRD

ALICE TRD principle

TRD in numbersPurpose: Electron ID in the central barrel

at p > 1 GeV/c Fast (6 μs) trigger for high-pt

Particles (pt > 3 GeV/c) +PID

Parameters: 540 modules → 767 m2 area18 “supermodules” 6 layers, 5 longitudinal stacks Length: 7 m 28 m3 Xe/CO2 (85:15)

1.2 million read out channels 15 TB/s on-detector bandwidth

Physics with the TRDTogether with TPC and ITS (dE/dx, good

momentum resolution), the TRD provides electron identification sufficient to study:

Di-electron channel: production of J/Psi, Upsilon and continuum (complementary to muon arm measurement).

+ Displaced vertex from ITS: E.g. Identify J/Psi from B decays

Single electron channel: semi-leptonic decays of open charm and beauty:

Handle on c+b production x-section

TRD alone:

L1 trigger on high-Pt particles+electron identification: Factor 100 Enhancement of potentially interesting events (PbPb).

Upsilon enrichment Jets: Study “jet quenching” under LHC

conditions

TPC dE/dx:~7% resolution

TRD pion efficiency

Test beam data:90% electron efficiency

Goal

Quarkonia performance

Phd. thesis Tariq Mahmoud, Heidelberg

pt/pt < 2% up to 10 GeV/c < 9% up to 100 GeV/c

B = 0.5 T

Central Barrel Pt-resolution

Signal/Background Significance

What is new at LHC

Plenty of c+b to start with

central AA

Upsilon suppression should be observable at LHC

RHICLHC

hard gluon induced quarkonium

breakup hep-ph/0311048

Complete primary J/Psi suppression expected

Strong (centrality dependant) secondary J/Psi production (statistical hadronization)

->strong QGP Signal

Read Out Chambers Large area chambers (1-1,7 m²)

-> need high rigidity

Low rad. length (15%Xo) -> low Z, low mass material

-> Carbon reinforced sandwich construction

Read out chambers II

• 5 chamber production sites:– Bucharest (NIPNE)

– Dubna (JINR)

– GSI (Darmstadt)

– Heidelberg (University)

– Frankfurt (University)

Dubna

Bukarest

• QA:– Standardized chamber

building prescription

– Chambers have to pass well defined set of Quality control steps

2d gain uniformity

Electronics

• 1.2 million channels• 18 channels in 1 MCM• 16(+1) MCMs per readout board (4104 pc.)• 260 000 CPUs working in parallel during readout

Electronics Status

• PASA and TRAP chips ready• PASA: have full quantity• TRAP: several Wafers

• Readout boards: last design changes

• Integration of electronics on chambers ongoing

PASATRAP

Electron ID

LQ Method:

Likelihood with total charge

Typical signal of single particle

LQX Method:

2d-Likelihood: Total charge + position of maximum cluster

()(

)(

PeP

ePL

Likelihood distributionExtract probabilities

Integrated Charge

Total charge spectra

Depos. Energy (keV)

Co

un

ts

Max. cluster position

Distribution of maximum cluster position

PID with Neural Network I

Each neuron of one Layer is connected to every neuron of the following Layer.

Input Layer: Charge per timebin

One hidden Layer: 22 neurons

Output layer per chamber: Probability to be Electron/Pion

Connect 6 Chambers by NN, or multiplication of Probabilities.

Submitted to NIM A, arXiv:physics/0506202v1

PID with Neural Network IISo far analysis done for Testbeam data with 4 small prototype chambers->extrapolation to 6 Chambers

Momentum dependence of Pion efficiency

To do:• Test with higher statistics and on generalized dataset (new Testbeam data)•Try to understand this significant improvement analytically

Testbeam Oct. 2004

• 4 small size prototype chambers (Transition radiation spectra measurement).

• 6 real size production chambers (2 different size types)• (Almost) final electronics

Signal in production chambers

Online Event displayElectrons

Pions

Position/Angle Resolution

Large chambers Prototype

Position resolution (y): 200-300 micron

Angle Resolution: <0.5°

Pion efficiency

2004 Test beam data compared to 2002 Test beam data:Somewhat worse separation

Pions

Electrons

Points: 2002 dataLines: 2004 data

Pion efficiency slightly worse than in previous test beam

Transition radiation

Transition radiationEnergy spectrum

data

simulation

Number of produced TR photons with different RadiatorsRegular: foil stacksSandwich: ALICE TRD radiator

Online Tracking

256

512

768

1024

AD

C V

alu

e

ADC Channel

Tim

e B

in [

0.1

us]

0 5 10 15 20

0

5

10

15

20

Comparison: Online tracking ↔ Offline tracking

Very Good Agreement!Outliers on per mille level due toCalculation precisionO

ffli

lne

Online

Summary

• TRD enhances ALICE Heavy flavour physics capabilities

• Detector mass production under way.

• Electronics finalized• Electronics Integration in

final iteration• First Supermodule to be

assembled end of the year

• Testbeam:– Detector performance is

well understood and satisfies design considerations

• Neural network approach:– New test beam data (6

real size chambers, different angles, higher statistics)

– Can information used by NN be extracted analytically?

TRD CollaborationMain Contributions:Germany:• Frankfurt University (IKF)• Gesellschaft für Schwerionenforschung (GSI) Darmstadt• Heidelberg University (Physikalisches Institut, Kirchhoff Institut)• Münster University (IKP)Russia:• JINR Dubna Romania:• NIPNE Bukarest

Additional Subsystems:Japan: Tokyo University, Nagasaki UniversityGreece: Athens University Germany: FH Köln, University Kaiserslautern, FH Worms, TU Darmstadt