Muon Reconstruction
in CMS
27th September - 1st October, 2004
Norbert Neumeister
CERN and Institute for High Energy Physics, Vienna
Outline
• Introduction
– CMS Muon System
– High-Level Trigger
• Muon Reconstruction
–
–
–
–
Software Design
Local Pattern Recognition
Standalone Reconstruction
Global Reconstruction
• Performance
• Summary
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
2
Compact Muon Solenoid Detector
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
3
CMS Muon System
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
4
Muon Detectors
Three types of gaseous particle detectors for muon identification:
• Drift Tubes (DT) in the central barrel region
• Cathode Strip Chambers (CSC) in the endcap region
• Resistive Plate Chambers (RPC) in both the barrel and endcaps
The DT and CSC detectors are used to obtain a precise
measurement of the position and thus the momentum of the muons,
whereas the RPC chambers are dedicated to providing fast
information for the Level-1 trigger
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
5
Muon System
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
6
Muon Reconstruction
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
7
CMS Trigger Overview
• Level-1: hardware trigger, 40 MHz  100 kHz (75 kHz)
–
–
–
–
–
Only calorimeter and muon information used
Electron/photon triggers
Jet and missing ET triggers
Muon triggers
Level-1 decision based on trigger objects
with / information
– Custom-built electronics
– Latency: < 3.2 s (128 bx)
40 MHz
100 kHz
• HLT: software trigger, 100 kHz  O(102)Hz
– Beyond Level-1 there is a High-Level Trigger running
on a single processor farm (no dedicated L2 hardware)
– DAQ designed to accept Level-1 rate of 100 kHz
– Access to full event data (full granularity and resolution)
– Rejection: 1:1000
100 Hz
– ~1000 processor units
– Once trigger rate is low enough (1 kHz) apply full reconstruction
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
8
Software Design
• Offline and High-Level Trigger
– Reconstruction software is designed to work for both, offline and HLT
– Level-1 Trigger provides “seeds” (Regions of Interest) for HLT
– Offline reconstruction makes use of complete calibration, alignment, etc.
• Robust, high quality reconstruction software
– Object-Oriented design
– Use of a common framework
• Basic concepts
– Reconstruction on demand
• Access data as needed
• Avoid unnecessary calculations; reject events as soon as possible
– Regional/Partial reconstruction
• Using data in a region around a “seed”
• Need seeds: objects with ,  information
• Reconstruction/selection applied to regions only
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
9
Local Pattern Recognition (I)
DT’s and CSC’s are multi-layer detectors:
 First step of muon reconstruction is local pattern recognition
 Reconstruct track segments in the DT and CSC detectors
• Barrel (DT):
– Reconstruct position of each channel above threshold
using effective drift velocity
– Reconstruct  super-layer hits (time-space conversion)
• Drift velocity depends on B field and impact angle
– Cluster hits (linear fit): 2D segment
– Fit 2D lines separately in r- and r-z through the 8+4 layers
of chamber
• L/R ambiguities solved by best 2
– Combine into 3D segments, use segment position and
direction for tracking
• Resolution: 100 m in  view, direction ~1mrad
– Apply impact angle correction on time-to-distance relation
and refit
– Calculate position (center of gravity) of the track-segment
and its angle in the super-layer (+error matrix)
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
up to 12 hits/station
10
Local Pattern Recognition (II)
• Endcaps (CSC):
– Reconstruct 3D hits
– Associate hits with linear fit (only one hit per
layer)
– Fit “Gatti” function to the spatial shape of
3-strip charge distribution to determine
centroid of cluster in layer
up to 6 hits/station
– Associate two projections by time coincidence
– Fit 3D segments through the collection of wire
and strip clusters in chamber; linear fit
– Resolution: 120–250 m form bending
coordinate, depending on chamber
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
11
Residuals and Pulls in r-
residual
pull
 = 109 m
600
2
 / ndf
49.27/17
1400
Constant
1200
Mean
0.0436
Sigma
1000
129
0.9774
400
800
DT
600
200
400
0
-1000
0
-500
X0sim
X0 fit m
200
500
1000
/ ndf
1231/23
0
2

-5
-4
-3
-2
Constant 1.272e+04
1400
Mean
1200
Sigma
0.0002398
0.02302
-1 0
rec
1
7000
5
210.4/57
Constant
6000
Mean
6310
0.01169
Sigma
1.238
4000
600
3000
4000
2000
2000
1000
0
4
2
5000
CSC
3
 / ndf
1000
800
2
(x -xsim )/ x
0
-0.2 -0.15 -0.1 -0.05 0
rec

sim

x -x
Norbert Neumeister
CERN PH / HEPHY Vienna
-5
0.05 0.1 0.15 0.2
(cm)
-4
-3
-2
-1 0
rec
1
(x -xsim )/  x
2
3
4
5

CHEP 2004
12
Standalone Muon Reconstruction
• All muon detectors (DTBX, CSC and RPC) are used
• Seed generation:
– external: Level-1 trigger (vector at 2nd station)  Level-2 reconstruction
– internal: track segments from local pattern recognition
• Fit:
–
–
–
–
–
–
Kalman filter technique applied to DT/CSC/RPC track segments
Use segments in barrel and 3D hits in endcaps
Trajectory building works from inside out
Apply 2 cut to reject bad hits
Track fitting works from outside in
Fit track with beam constraint
• Propagation:
– Non constant magnetic field
– Iron between stations, propagation through iron (more difficult than in tracker!)
– GEANE used for propagation through iron
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
13
Global Muon Reconstruction
Inclusion of Tracker Hits
Start from standalone reconstructed muons:
•
Seed generation
– Get muon trajectory at innermost muon station
– Propagate to outer tracker surface and to interaction point
– Open window for track reconstruction
• define region of interest through tracker based on Level-2 track with
parameters at vertex
• fixed/dynamic region
– Create one or more seeds for each Level-2 muon
•
Construction of trajectories for a given seed
– Propagate from innermost layers out, including hits in muon chambers
– Resolve ambiguities
– Final fit of trajectories
tremendous gain in resolution
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
14
Tracking Strategy
Rely on few measurement layers, each able to provide
robust (clean) and precise coordinate determination
2 to 3 Silicon Pixel, and 10 to 14 Silicon Strip Measurement Layers
At high luminosity (~ 20 min. bias events every 25 ns):
R
= 10 cm
25 cm
60 cm
Nch/(cm2*25ns) = 1.0
0.10
0.01
Radius ~ 110 cm, Length ~ 270 cm
R-phi (Z-phi) only
measurement layers
6 layers
TOB
R-phi (Z-phi) & Stereo
measurement layers
~2.4
4 layers
TIB
3 disks TID
Norbert Neumeister
CERN PH / HEPHY Vienna
~1.7
9 disks TEC
CHEP 2004
15
Muon Reconstruction
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
16
Muon Reconstruction @ high L
W  
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
17
Muon Reconstruction @ high L
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
18
Muon Reconstruction @ high L
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
19
Muon Reconstruction @ high L
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
20
Muon pT Resolution
barrel
overlap
 = 0.12
 = 0.14
endcaps
 = 0.17
Level-2:
 = 0.013
 = 0.015
Level-3:
Norbert Neumeister
CERN PH / HEPHY Vienna
 = 0.018Order of
magnitude
improvement
CHEP 2004
21
Muon pT Resolution
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
22
Status and Plans
• Propagation
– Extrapolate track parameters and errors
– Reconstruction geometry
• Currently, includes only chambers (active material)
• Include geometry of non-sensitive materials with optimized navigation
• Fast location of homogeneous volumes
• Volumes know: magnetic field, material properties and neighbors
– Propagation in a single volume through homogeneous material and
continuous (non-constant) magnetic field
– Large amount of material  Multiple scattering, energy loss
– Replace GEANE
• Magnetic Field
– Need accurate field map
– Fast Look up
– Non constant magnetic field (Runge-Kutta Integration)
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
23
Summary
• Muon reconstruction
–
–
–
–
Local: pattern recognition
Regional: standalone muon reconstruction
Global: Muon system + tracker
Same base software for offline and HLT
• HLT
– HLT offers high level of flexibility
• Status and Plans
– Propagation
– Magnetic Field
Norbert Neumeister
CERN PH / HEPHY Vienna
CHEP 2004
24