High Level Trigger of Muon Spectrometer

Indranil DasSaha Institute of Nuclear

Physics

Description of the problem

Z

y

J/Ψ μ+ μ- Y μ+ μ- B μ+ /μ- + X D μ+ /μ- + X

Signal

Muons from π and K

BackgroundL0 Trigger by Muon Trigger

P2 = (P1 + P2)2 Continue hardware and software trigger…..

(for hadronic background)

Description of the problem• For ALICE Muon Spectrometer L0 (hardware trigger) == Hits in trigger station. + Track pointing to origin + Option for transverse momentum (pT) selection.

• Muons coming from Pions and Kaons decay have low pT .

• Thus to reduce the background two pT cuts of 1 GeV or 2 GeV has been planned to clean invariant mass spectrum of J/Ψ and Y.

px

py

pz

p

pT

Software cut improves the result, so why not online software cut ?

HLT : Thus, on top of the hardware trigger a software trigger is applied, which is called High Level Trigger (HLT). The HLT is an online system that provides the common facilities (i.e. Input/Output Link, PC farm for analysis) and executes the triggering algorithms (full reconstruction) developed for different detectors.

Main motivation of Dimuon HLT : In case of Muon Spectrometer, HLT algorithms are supposed to improve the pT cuts (1 GeV or 2 GeV at most, depending on beam luminosity) as defined by Muon Trigger Station (L0 Trigger), which are important remove the combinatorial background of low momentum particles to clear the J/Ψ and Y signals. A typical processing rate of 1 kHz is the design requirement of muon HLT for heavy ion Pb-Pb collisions.

Description of the problem

How do we deal online data in reality ?

ALICE data flow in online mode

Offline

Detector

Physics Trigger {Level 0 [L0]}

HLT

DAQ

Clusters HLTESD

Tracks

Rawdata (GRID)

Clusters

Tracks

ESD

HLTESD

Event Display

But then it is very straight forward right ?

Description of the problem

• One simulated Pb-Pb event needs few seconds for offline reconstruction (official physics analysis software) and therefore this can not be used for online analysis.

• Thus, a new algorithm has been developed which will perform online analysis of data and satisfies,

– A online processing component has to be robust (24x7)– It is to be fault tolerant against input data stream– It has to provide results of appreciable quality– It has to be fast enough so that the rest of experimental

component does not wait for HLT If any of above conditions are violated dimuon part of HLT is excluded

from the experimental run.Data rate of 500 MB/s for 1 kHz trigger rate with hit multiplicity ~ 300Accuracy as good as offline reconstruction within the time limit of 1 ms

Outline

• Development of the Algorithm– Hit Reconstruction (Thesis work)– Trigger Reconstruction (UCT + Cagliary)– Track Finding (Thesis work)• Cellular Automata• Kalman Filtering

• Validation with the simulated data (Thesis work)• Implementation in HLT PC-Cluster and validation

with cosmic data (Thesis work)• Online display of LHC p-p data (Thesis work) • Ongoing Analysis of p-p data (Thesis work)

At first the Hit Reconstruction algorithm…..

A fast Algorithm The charges are found to spread over 2 or 3 pads along a direction in most of the cases => one cluster.

Each cluster is characterized by one Central Pad (Pad with maximum charge for the given cluster).

Thus, to generate reconstructed hits, it is not essential to make clusters but only identification of Central Pad is sufficient.

Once the central pad is found the reconstructed Y hits in bending plane is calculated using centre of gravity method over three pads around the central pad.

Above method is repeated for non bending plane to find out reconstructed X .

Finally the bending and non bending hits are merged to form reconstructed X and Y.

Comparison with offline…

Comparison of Simulation Results

Fine !!! What about tracking ?

Charged particle track

Z

y

Trigger SeedSt. line track

Kalman χ2 matchCellular Automata

MCS and E.Loss Correction

Trigger Seed

The Trigger chamber issues a Trigger when at least three out of four planes are fired along linear direction

Linear Extrapolation in St4/5

A trigger tracklet can be formed with the trigger seed, which is

extrapolated to find the ROI in St4/5.the hits inside those ROI are

checked for the alignment with the trigger tracks and added to extrapolate

the tracks.

CA in St1 and St2

At First the small tracklets between thetwo chambers of a given station are

formed. Then the tracklets from different stations are collected to make a

track segment in the forward half of the Spectrometer.

Kalman Chi2Test

The tracksegs in the St4/5 are extrapolated through the magnetic fieldusing Kalman filter to meet the trackseg

In St1/2.

Extrapolate

The tracks are then extrapolated to vertex to incorporate energy loss and

MCS (multiple Coulomb Scattering)

in the Muon absorber .

Full Tracker Scheme

What is Cellular Automata ?

Cellular Automaton

Cell Creation

Cell is a small tracklet between thetwo chambers of a given station. The tracklet object contains three

pointers, one corresponds to its rankin the track and the rest two points

towards each other

Connect Tracks

The cells from different stations arecollected into a track from the Muon

Trigger stations towards station1 with certain strict boundary

condition.

How does Kalman Filter Works ?

Kalman Gain

Project to k+1

Update Covariance

Update Estimate

Measurement

Kalman Filtering Process

Projected Estimate

Initial Estimate

Result on simulated data….

Transverse Momentum Reconstruction at a given PT

For other delta PT values…..

Comparison of Different Tracking Approach

0.5 GeV 1.0 GeV 1.5 GeV 2.0 GeV 2.5 GeV

Offline Resolution(MeV)

33.8 +/- 0.7 34.9 +/- 0.2 39.8 +/- 0.2 43.9 +/- 0.2 40.5 +/- 0.1

Efficiency(%) 96.49 +/- 0.912

99.57 +/- 0.533

98.86 +/- 0.473

98.61 +/- 0.458

98.83 +/- 0.457

Full Tracker Resolution(MeV)

44.7 +/- 1.0 47.0 +/- 0.3 61.6 +/- 0.4 74.9 +/- 0.4 78.8 +/- 0.5

Efficiency(%) 95.61 +/- 0.906

97.57 +/- 0.525

97.95 +/- 0.470

98.50 +/- 0.458

97.98 +/- 0.455

How does it fit to the PT efficiency plot ?

PT Cut Efficiency

Offline testing is not same as online implementation….

Pause and discuss….

First Track in Muon Spectrometer of ALICE was detected by SINP group

Run 24841March 3, 2008

And many more…..

Event Snapshots of ALICE Control Room

Run 119842

ALICE Control Room (ACR)First online snapshot of muon track in 7 TeV (highest energy) pp collision beam at record luminosity (~ 6 × 1028 cm-2 s-

1), processing at a rate of 700 Hz

Not Only Display, but Quality Assurance(QA) as well……….

PT Distribution

Cluster D

ist. Attached to

the tra

cks

Real Time Vertexing !!!!!!

Online Inv. Mass Distribution

Bingo!!

Analysis of the quality of the events…..

Number of Clusters per Data Link

Charge Ratio in the Two Planes of the Detectors

Trigger Type as declared by Trigger Chamber

Total ADC Charge per Data Link

Monitoring even deeper…..

Cluster size distribution

Cluster Charge Distribution

Offline Comparison

The performance plots are shown for the Full-Tracker on data from 54 runs from LHC10d:126097 12608 126087 126081 126078 126073 126008 126007 126004 125855 125848 125847 125844 125843 125842 125632 125630 125628 125296 125292 125186 125156 125139 125134 125131 125101 125100 125097 125085 125083 124608 124607 124606 124605 124604124603 124600 124388 124380 124378 124374 124371 124367124364 124362 124360 124359 124358 124355 124187 124186 124183 122375 122374

How does L0, Offline, HLT triggered mass spectrum look like …. ?

Preliminary

Preliminary

Ongoing Analysis of p-p data

Any track passes 1.0 GeV cut

Summary and Future Plan

• Successful development, simulation testing, online implementation, real-time execution and validation by physics analysis of the Hit Reconstruction and Track Reconstruction algorithms

• Histogram incorporation to QA for global run characteristics.

• Continue Validation test with p-p and Pb-Pb collision.