Transcript Diapositiva 1 - unisalento.it

Muon Reconstruction
with Moore and
MuonIdentification
The Moore/MUID group
Atlas Physics Workshop
Athens, May 2003
Outline








Reconstruction method and architecture
Muon spectrometer inert material treatment
Single m performances
H→4m
Z→2m
Test beam data reconstruction
Moore and Muid as Event Filter algorithms
Conclusions
Tasks

Moore (Muon Object Oriented REconstruction)


reconstruction in the MuonSpectrometer
MuonIdentification


Muon reconstruction and identification
Divided in two parts :
 MuidStandAlone:


Muid Comb:


Back tracking of the MOORE tracks to the interaction point
Combination of the muon and the inner detector tracks
Both work in ATHENA (= the ATLAS reconstruction
framework)
MOORE
rpc
Reconstruction
rpc
Strategy (I)
barrel
F projection
rpc
Search for region
of activity in the
f projection
and
RZ projection
rpc
rpc
rpc
MDT
barrel
RZ projection
MOORE Reconstruction
Strategy(II)
 Pattern recognition in the
MDTs
 the drift distance is calculated
from the drift time, by applying
various corrections on it (TOF,
second coordinate, propagation
along the wire, Lorenz effect).
Among the 4 tangential lines the
best one is found.
 Track segment combination.
MDT pattern
recognition
MDT mutilayer
 Track fit
track parameters (a0, z0, f, cotq, 1./pt ) expressed
at the first measured point
MuonIdentification
Method

MS track parameters are propagated
to beam-axis



multiple scattering parameterised as
scattering planes in calorimeters
energy loss from truth or from Calo
Reconstruction or from
parametrization as function of (h,p)
Muonspectrometer
inner layer
Energy loss
and multiple
scattering
calorimeters
Refit

muon track parameters expressed
at vertex
Beam spot

Muon/ID tracks matching with a c2
cut-off

c2 based on track covariance
matrices and the difference in the
track parameters

Combined track fit
Architecture (I)


C++ and OO technology
High modularisation and flexibility
Easier to develop alternative recostruction approaches
Successfully adapted for the test beam data reconstruction
Successfully integration in the HLT framework
Part of Moore will be used by Calib (MDT calibration)

Track class and Fitter class in common with the inner
detector reconstruction (iPat)
• Separation of the algorithmic classes
from data objects
MooAlgs
MooEvent
Architecture (II)
Pattern recognition is divided
in several steps.
Each step is driven by an
Athena top-algorithm
Algorithms indepent, imply
less dependencies, code more
maintainable, modular, easier
to develop new reconstruction
approaches
Separation of the algorithmic classes
from data objects
Parameterization of the Inert
Material (I)
March 2003
The treatment of the inert material
until March

NO material
No services available yet in Athena for the description of
the inert materials
in the fit
1/PT Pull

The ATHENA geometry service allows to take into
Single m
account in the fit multiple scattering and energy loss in
Pt=20 GeV
the material of the chambers.

Work in progress (waiting for a Material Service from GeoModel):



Parameterization of the material effects:
Data driven approach: define a map of materials; tune
materials with the pull distributions (available)
Geant4 based approach: describe all inert materials;
propagate geantinos; define a fine map of materials in the
Spectrometer (in progress)
Chamber
Material in the fit
Parameterization of the Inert
Material (II)
B C B
Main steps
A
•Define a segmentation of the
Muon Spectrometer:
Binning in h/f/L
(L = trajectory path length)
•Estimate X0 and Energy Loss
in each h/f/L bin
•Refit the track with 2 scattering
centers per station
•Tune X0 and energy loss of the
“scatterers” against the pull
distributions
1.2
1.8
3.1
3.6
5.4
7.8
Overall performances
1st
iteration
Pt = 6 GeV
1/pt pull
distributions
with Parameterization of
Inert Material and with
Detector Material only
|h| < 1
For fixed transverse
momentum single muon
samples
|h| > 1
Pt = 20 GeV
Single m
performances


Efficiency vs pT
Single muons
DC1 data
Moore/MuonID performances
shown here are obtained with
• Release 6.0.3
• A private improved version of
MuonIdentification
•Tracking in the magnetic
fields, bug fixes
• Moore with the full material
description
MooAlgs-00-00-41
MooEvent-00-00-42
PT (GeV)
Rather good agreement with
Physics TDR results
Efficiency vs h
Stepping in the fit procedure needs to be optimized in the region |h|>1
Uniform efficiency vs phi
1/Pt Resolution vs Pt
Rather good agreement
with Physics TDR results
Pt (GeV)
1/Pt resolution vs h
Critical reconstruction in the transition region at low momenta
1/Pt resolution vs f
Worsening of resolution in the MuonSpectrometer
in the feet region at low momenta
s(1/pt pulls) vs h
H4m (I)
(Evelin Meoni)
•DC1 sample (prod. in July 2002):
H4m (with mH=130 GeV)
~ 10 K evt.
• ATHENA 6.0.3 (Moore-00-00-42)
Without Z constraint
s = (3.15±0.09) GeV
Reconstruction with
Muon Spectrometer
Standalone (Moore +
MUID Standalone)
With Z constraint
s = (2.33±0.07) GeV
Phy TDR s = 2.7 GeV
Phy TDR s = 2.1 GeV
H4m (II)
(Evelin Meoni)
Combined Reconstruction
(Moore + MUID + iPat)
Without Z constraint
With Z constraint
s = (1.49±0.05) GeV
s = (1.85±0.06) GeV
Phy TDR s = 1.6 GeV
Phy TDR s = (1.42 ±0.06) GeV
→ mm
0
Z



DC1 Data
Moore/MUID in 6.0.2
Total Number of Events:
~10000
Phy TDR:
Standalone Muon Spectrometer : s = 3.0 GeV
Combined MS + ID : s = 2.5 GeV
Test Beam Reconstruction (I)




MuonTestBeam: ATHENA package designed to read and decode
DAQ data from the Muon Test Beam in H8 and prepare the
reconstruction with Moore algorithms
Status (in ATHENA release 6.2.0):
 Converter to access to the H8-DAQ data in bytestream format 
construct digits in the TDS using the new Muon Event Data Model
 H8 detector description through the ATHENA MuonDetDescrMgr
initialised from H8 Amdb
 Segments and tracks reconstructed via Moore algorithms
 Ntuples for the analysis
 Access to digits from H8 Geant4 simulation also implemented
First tests performed on data collected during 2002 (only MDT)
Also RPC and TGC at this year’s test beam
Test Beam Reconstruction (II)
DAQ
file
Test Beam Converter
MuonDigitContainer
TDC
counts
In the Athena
framework
Moore Algs.
Conditions DB
service
Ntuple
Tracks
angle
First test on
2002 data
• Straight line fit of track segments and full tracks implemented
• Integration with MDT calibration, chambers alignment
• The access to conditions DB through the Athena Interval Of Validity
Service is under development
Moore and MUID in HLT
At the hearth of the philosopy of the
Pesa design is the concept of
seeding.
Algorithms functioning in the context
of the High Level Trigger do not
operate in a general-purpose sense;
rather they must validate or reject
Trigger Element hypotheses formed
at a previous stage in the triggering
process.
Reminder geometry
Strategy
for the Seeding
RecoMuonRoI
(h ±Dh, f±Df)
The RecoMuonRoI gives
DigitsContainer
eta & phiTDS
position of RoI.
seeded
(ZEBRA)
RPCDigitContainer
MDTDigitContainer
From the event store (ZEBRA)
it is possible to access the full
event and to build the RPC
and MDT digitcontainers.
Hash offline IDs
RegionSelector
(RPCs and MDTs)
The
HLT
RegionSelector
translates geometrical regions
within the fiducial volume of
the detector into a set of
corresponding elements of
appropriate granularity in each
sub-detector
Using boths it is possible to build
the seeded DigitsContainers
MooMakePhiSegmentSeeded
MooMakeRZSegmentSeeded
PhiSegmentContainer
RZSegmentContainer
And to follow the
reconstruction chain
Reconstructed Tracks
MooMakeRoads
MooMakeTracks
MuId standalone
Moore/MuId – First time-performance test
300GeV
TDR
200GeV
DC1
H
DC1
142 msec
368 msec
279 msec
572 msec
PT
(t -1)
4m
20GeV TDR 20GeV DC1
155 msec
• t-1
Average execution time per event calculated for the 500 events sample.
The 1st event has not been included in the calculation since in this event several
services are initialized (magnetic field map,… ).
Pt (GeV)
Time (ms)
20
5.1
100
6.3
300
4.9
H->4mu
mH= 130 GeV
25.2
Conclusions
A lot of improvements have been made to MOORE/MUID
in the last two months: it can now be used for Physics Studies
The code has proved to be robust on high statistics DC1 samples
(~106 events processed – No Crash)
A big (and successful!!) effort has been done for having
MOORE/MUID as Event Filter in the HLT framework: results will
appear in the HLT TDR
Alternative tracking methods to be inserted in MOORE (e.g. Kalman
Filter) are under developments
We are aiming at keep going with the developments but always
having a reference version to be used for Physics Studies