Muon PRS Report
Tasks & Manpower Needs
DC04 Preparations
High Momentum Muon Reconstruction
H → 4µ Studies
Darin Acosta
PRS Task List and Manpower
As part of the CPT Annual Review and the LHCC
review, we were asked to provide a task list for the
next couple years and the estimated manpower
required
Tasks:
Î
Simulation
Î Reconstruction
Î Calibration & Alignment
Î Data Handling & Monitoring
Î L1 Trigger Simulation
Î HLT
Î Testbeam Software
EMU Meeting, 19 Oct 2003
2
Annual Review Recommendations
Recommendations (in Order of Priority):
1. A renewed effort is needed in detector calibration. This
includes all the PRS detector groups as well as
calibration infrastructure from CCS. Since an exercise
of calibration was already planned for DC04 as input to
the computing TDR, this is clearly an urgent problem.
2. The detector groups should supply manpower to
perform time critical activities such as calibration.
3. PRS should participate in the specification of a
minimal (Distributed) Analysis Model which can be
implemented in time for the Physics TDR. In fact, it
would be good if we could have an implementation to
study in time for the Computing TDR.
4. PRS should rapidly move to produce persistent
physics objects in the new framework.
EMU Meeting, 19 Oct 2003
3
PRS Tasks & Estimated Manpower Needs
Muons
(FTE)
Simulation
2003
2003
(now)
(need)
1.5
(Geometry, simHits, digitization, Validation,
CommonDet and Tracker domains, fast simulation)
8
Reconstruction
2
(pulse reconstruction; maintenance and development
of clustering algorithms)
5
Calibration
(Simulation studies, application of constants)
Data handling & Monitoring
0.5
2004
2005
2006
2007
2.5
2.5
2.5
2.5
2.5
2.5
4
4
3
4
2
4
4
4
4
0.5
2
3
2
3
2
0.3
(Any off-line and quasi-online software needed to
cope with real data, Tracker and DAQ Proto.)
0
Level-1 Trigger simulation
2
2
1.5
1
1
2
HLT
1
1.5
4
4
3
5
4
4
4
4
4
15
22
22.5
19.5
24.5
muon object reconstruction, and full analysis like for
Physics TDR
10
Test-beam software and analysis
4
(Normally included in Data handling and
Monitoring)
12
TOTAL
EMU Meeting, 19 Oct 2003
11.3
4
Effort needs to double in the next year
Simulation Tasks
√ Final adjustments & tuning to OSCAR
(Geant-4 detector simulation)
√ Refinement of simulation based on testbeam results
Implementation of muon performance
parameterizations into FAMOS
(fast detector simulation)
Development of a long-term supportable XML
geometry description for DDD (and short term
maintenance/update of existing XML)
√ Support for the production of necessary MC samples
for muon analyses
( √ = task started, partially covered)
EMU Meeting, 19 Oct 2003
5
Reconstruction Tasks
√ Further improve/optimize local reconstruction
(DT, CSC, RPC)
Î
In presence of backgrounds
Î From results of testbeam studies
Î Develop tools for MC/testbeam comparison
√ Optimize global reconstruction for efficiency, speed,
rate reduction:
Î
Define/implement “reconstruction geometry” (different level
of details than “simulation geom.”; same DDD source )
Î Develop improved navigation & extrapolation tools
Î Develop improved B-field tools (data retrieval, interpolation)
EMU Meeting, 19 Oct 2003
6
Calibration & Alignment Tasks
For DT, CSC, RPC separately:
Î
Establish calibration procedures, frequency, and data
volume
Î Define format of constants stored in DB
Î Possibly transfer exisiting data from chamber/electronics
production DB to CMS-wide DB (once latter is defined)
Î Define interface between DB and HLT/reconstruction code
Î Develop software to apply DB constants to
HLT/reconstruction
Same for alignment (⇒ constants from alignment
hardware)
Additionally:
√ Î Interface “misalignment tools” to “reconstruction geometry”
√
(from DDD) in ORCA
Î Develop calibration strategies & software (e.g. vs magnetic
field knowledge)
Î Develop alignment strategies with tracks & software
(“internal” barrel/endcap, overlap region, muon-Tracker )
EMU Meeting, 19 Oct 2003
7
Data Handling & Monitoring Tasks
Define monitored quantities and possible alarms for DCS
Define monitored quantities and tolerance criteria for
online farm
Î
Probably a lot of experience at FAST sites…
Develop software to produce histograms and alarms
from online data, HLT monitoring (once CMS-wide
framework developed)
Define muon data sets (streams) and data volume
√ Î Single muon and di-muon streams
Î
Other physics streams?
Î Prescaled loose triggers (e.g. HLT pass-through triggers)
Define data handling procedure for reconstruction
(application of calibration and alignment corrections)
√ Introduce realistic data formats in HLT processing in
ORCA
Î
Started in “slice test” software development
EMU Meeting, 19 Oct 2003
8
L1 Trigger Simulation Tasks
Update L1 simulation to exactly match electronics
algorithms:
√ Î RPC trigger with 4/6 logic and HO added
√ Î DT BTI+TRACO+TSS updates
Î
RPC → CSC interface (TMB)
Refine L1 algorithms to further improve efficiency, rate
reduction and functionality:
√ Î CSC single µ trigger for |η|>1.6 without RPC and ME4/2
√ Î CSC di-muon trigger with increased η coverage
(See Alexei’s talk)
Î
Cosmic ray trigger
‡ We’ll need a cosmic trigger for slice tests on the surface
Î Accelerator muon trigger
‡ Needed for beam running (alignment studies)
EMU Meeting, 19 Oct 2003
9
HLT and Full Analyses
√ Minimize HLT execution time
Adapt HLT code to any specific online filter farm
requirements
√ Refine HLT algorithms to further improve efficiency,
rate reduction and functionality (especially based on
analyses)
Muon reconstruction at very high pT
Î
Study/optimize performance
√ Muon identification
Î
Isolated & inside jets (⇒ b-physics)
√ Perform full analysis of H → 4µ
Î
No U.S. group?
√ Perform full analysis of H → 2µ
√ Perform full analysis of Z' → 2µ
EMU Meeting, 19 Oct 2003
10
For Physics
TDR
Test Beam Software & Analysis
√ Integrate testbeam DAQ software into ORCA framework
Log data into native ORCA data format
‡ Work of R.Wilkinson for EMU, with P.Kreuzer (UCLA) joining
Î Develop testbeam geometries
(prototype exists for Barrel with simple “one-station” setup)
‡ M.Case (UCD) assists for EMU (?)
Î
√ Perform analysis of local reconstruction quantities (e.g.
residuals) for comparison and tuning of detector simulation
Î
Could benefit from an analysis of the latest CSC beam test data with
production electronics
Perform dedicated mini-experiments (e.g. scattering in iron) to
check HLT and validate Geant4 physics simulation
Î
Develop software for 2 stations setup (geometry, L1 TrackFinder
LUTs, reconstruction...)
testbeam software into system software for slice tests of
√ Evolve
many chambers and for CMS commissioning
Î
Starting for CSCs
EMU Meeting, 19 Oct 2003
11
DC04 Status
The 2004 Data Challenge is a 50M event production
Î
In fact 70M Pythia events were requested by the PRS groups
Î 35M events have already been pushed through CMSIM
OSCAR (G4) is ready for detector simulation
Î
Will be used for all PRS/µ samples except W/Z
(i.e. we’ve been waiting…)
ORCA is about ready for digitization
Î
First 4M events will be ready within a month
‡ 0.7M requested by PRS/µ: single muons, high-mass DY,
t-tbar events
Î Remaining 46M available after Feb’04 challenge
Data will be available at CERN or FNAL
Î
But perhaps not at both
EMU Meeting, 19 Oct 2003
12
High Momentum Muon Reconstruction
Improvement to muon reconstruction algorithm driven
by physics study of Z’ search
Î
Work of R. Cousins and J. Mumford (UCLA)
Î Significant tails in mass resolution of high-mass di-muon
pairs are a major source of background for the Z´ signal
Î Would like to reduce tails while maintaining or improving Z´
FWHM
Detailed report given at Muon PRS Meeting during
Sept’03 CMS Week
EMU Meeting, 19 Oct 2003
13
Some Possible Track Fitting Methods
GMR
•
•
5 Parameter measurement at inner surface of tracker from backward part of Kalman filter fit to tracker and
muon hits (default L3 fitting method).
http://agenda.cern.ch/askArchive.php?base=agenda&categ=a03423&id=a03423s1t0/transparencies for more
information.
Tracker System
Tracker Only
•
•
Measure here
Use scaled final
forward values
to initialize
backward filter
Muon System
Forward filter
Backward filter
Iron CSC, DT
Use only tracker hits for Kalman filter. Take measurement from innermost tracker surface.
Available as reconstruction option in ORCA 6.3.0 (implementation by Norbert Neumeister).
Tracker System
Muon System
Tracker + 1 Muon-Station
•
Use tracker and hits from first muon station which contains hits.
Tracker System
µ-PRS Meeting: September 16 , 2003
Muon System
R. Cousins, J. Mumford UCLA
4
1/pT Resolution : Fit Comparison I
Entries
GMR
eta < 1.04
Entries
160
140
"Barrel"
0.01098
RMS
0.1577
Underflow
120
In barrel, tracker-only has
smaller tails, (truncated) RMS
1068
Mean
0
Overflow
18
100
Tracker Only
Entries
80
60
eta < 1.04
Entries
160
1068
Mean
0.0005788
RMS
140
40
120
0.1184
Underflow
0
Overflow
6
20
Tracker + 1 mu-station
-0.8
-0.6
-0.4
-0.2
-0
0.2
0.4
0.6
0.8
1
((1/pT)rec - (1/pT)gen)/(1/pT)gen
Entries
80
60
GMR
Entries
Mean
"Endcap"
RMS
120
40
670
-0.009759
0
20
Overflow
7
0
-1
160
Mean
1068
0.005602
RMS
120
0.133
Underflow
0
Overflow
13
100
0.1689
Underflow
eta < 1.04
Entries
140
eta > 1.04
160
140
Entries
100
0
-1
-0.8
-0.6
-0.4
-0.2
-0
80
0.2
0.4
0.6
0.8
1
((1/pT)rec - (1/pT)gen)/(1/pT)gen
60
100
Tracker Only
Entries
80
60
eta > 1.04
Entries
160
40
120
20
-0.01675
RMS
140
40
670
Mean
0.1849
Underflow
0
Overflow
5
0
-1
-0.8
-0.6
-0.4
-0.2
-0
0.2
0.4
0.6
0.8
1
((1/pT)rec - (1/pT)gen)/(1/pT)gen
20
Tracker + 1 mu-station
-0.8
-0.6
-0.4
-0.2
-0
0.2
0.4
0.6
0.8
1
((1/pT)rec - (1/pT)gen)/(1/pT)gen
Entries
100
0
-1
80
60
RMS
120
20
ORCA 6.3.0 Digis
ORCA 6.3.0 MuonAnalysis package
Drell Yan Mass > 1.5 TeV Sample
pT > 100 GeV
Mean
140
40
0
-1
eta > 1.04
Entries
160
671
-0.01558
0.1467
Underflow
0
Overflow
4
100
-0.8
-0.6
-0.4
-0.2
-0
0.2
0.4
0.6
0.8
1
((1/pT)rec - (1/pT)gen)/(1/pT)gen
80
60
Differences between fits
in endcap less clear
µ-PRS Meeting: September 16 , 2003
40
20
0
-1
-0.8
-0.6
-0.4
-0.2
R. Cousins, J. Mumford UCLA
-0
0.2
0.4
0.6
0.8
1
((1/pT)rec - (1/pT)gen)/(1/pT)gen
5
1/pT Resolution : Fit Comparison II
eta < 1.04
50
40
30
1068
0.0007453
0.08179
23
58
108.7 / 86
0.0494
33.86 ± 1.62
0.002935 ± 0.00211
0.06216 ± 0.00213
In barrel and endcap, tracker + 1
mu-station has best fitted-sigma
0.062
Tracker Only
20
50
10
40
30
-0.2
-0.1
0
1068
-0.0005495
0.09854
12
18
105.2 / 89
0.1153
24.76 ± 1.107
-0.0003817 ± 0.003048
0.09069 ± 0.002867
0.1
0.2
0.3
((1/pT)rec - (1/pT)gen)/(1/pT)gen
20
eta > 1.04
Entries
Mean
RMS
Underflow
Overflow
χ2 / ndf
Prob
Constant
Mean
Sigma
60
50
40
30
670
-0.000949
0.09816
26
26
98.92 / 82
0.09837
16.79 ± 1.171
-0.002843 ± 0.00338
0.07455 ± 0.004219
0.075
20
0
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
((1/pT)rec - (1/pT)gen)/(1/pT)gen
eta < 1.04
Entries
Mean
RMS
Underflow
Overflow
χ2 / ndf
Prob
Constant
Mean
Sigma
60
50
40
0
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
((1/pT)rec - (1/pT)gen)/(1/pT)gen
Tracker Only
30
eta > 1.04
Entries
Mean
RMS
Underflow
Overflow
χ2 / ndf
Prob
Constant
Mean
Sigma
60
50
40
10
Tracker + 1 mu-station
10
Entries
Entries
GMR
0.091
Entries
0
-0.3
eta < 1.04
Entries
Mean
RMS
Underflow
Overflow
χ2 / ndf
Prob
Constant
Mean
Sigma
60
30
670
-0.00588
0.1041
36
27
90.03 / 88
0.42
12.97 ± 0.879
-0.007093 ± 0.004398
0.09662 ± 0.005328
0.097
20
0
-0.3
-0.2
-0.1
0
-0.1
0
0.1
0.2
0.3
((1/pT)rec - (1/pT)gen)/(1/pT)gen
ORCA 6.3.0 Digis
ORCA 6.3.0 MuonAnalysis package
Drell Yan Mass > 1.5 TeV Sample
pT > 100 GeV
0.1
0.2
0.3
((1/pT)rec - (1/pT)gen)/(1/pT)gen
Tracker + 1 mu-station
eta > 1.04
Entries
Mean
RMS
Underflow
Overflow
χ2 / ndf
Prob
Constant
Mean
Sigma
60
50
40
-0.2
30
671
-0.001058
0.08231
24
11
83.81 / 75
0.2275
19.49 ± 1.213
-0.000199 ± 0.002979
0.06841 ± 0.003234
0.068
20
10
0
-0.3
µ-PRS Meeting: September 16 , 2003
0.051
10
10
0
-0.3
1068
0.004161
0.07089
20
36
115.7 / 75
0.001778
42.53 ± 1.869
0.005539 ± 0.00174
0.05062 ± 0.001485
20
Entries
Entries
Mean
RMS
Underflow
Overflow
χ2 / ndf
Prob
Constant
Mean
Sigma
60
Entries
Entries
GMR
-0.2
-0.1
R. Cousins, J. Mumford UCLA
0
0.1
0.2
0.3
((1/pT)rec - (1/pT)gen)/(1/pT)gen
6
A Prototype Selection Criterion Using “Prob” in
Tail of F2 Distribution > Observed F2
4.
If difference is
greater than 50
(optimized value),
replace with
tracker-only fit.
Track is now
referred to as
“optimized”.
Replace with
tracker-only
160
140
30
-ln(Prob) of
replacement muon
from tracker-only.
25
120
20
Entries
119
Mean
3.676
RMS
7.749
Underflow
0
Overflow
0
100
15
80
60
~11% of tracks in "barrel"
10
40
0
0
5
No change
20
5
10
15
20
25
30
35
eta > 1.04
Entries
671
Mean x
2.227
Mean y
3.39
RMS x
5.07
RMS y
6.618
44
7
0
0
620
0
0
0
0
200
180
Replace with
tracker-only
160
140
0
0
40
45
50
-ln(Prob) Tracker
10
15
20
25
30
35
40
45
50
-ln(Prob) Replacement
Entries
44
Mean
3.657
RMS
8.02
Underflow
0
Overflow
0
18
16
14
12
100
10
80
8
60
6
~7% of tracks in "endcap"
4
No change
20
0
0
5
eta > 1.04
120
40
ORCA 6.3.0 Digis
ORCA 6.3.0 MuonAnalysis package
Drell Yan Mass > 1.5 TeV Sample
Muon pT > 100 GeV
eta < 1.04
Events
180
1068
2.512
4.423
4.419
7.05
2
0
0
Events
3.
Check difference in
-ln(Prob) between
tracker + 1 mustation and trackeronly fits
eta < 1.04
Entries
Mean x
Mean y
RMS x
RMS y
0
119
0
947
0
0
200
-ln(Prob) Tracker + Mu-Station
2.
Start with tracker +
1 mu-station fit.
-ln(Prob) Tracker + Mu-Station
1.
5
10
15
20
25
30
35
40
45
50
-ln(Prob) Tracker
µ-PRS Meeting: September 16 , 2003
2
0
0
5
10
15
20
25
R. Cousins, J. Mumford UCLA
30
35
40
45
50
-ln(Prob) Replacement
11
Zssm (3 TeV) vs Drell-Yan background
Optimized Opp-sign dimuon mass
45
Events/100 GeV/100 fb-1
Events/100 GeV/100 fb-1
Default GMR Opp-sign dimuon mass
40
35
30
45
40
35
30
25
25
20
20
15
15
10
10
5
5
0
1500
2000
2500
3000
3500
ORCA 6.3.0 Digis
MuonAnalysis package
Note: Optimization algorithm was
tuned on different sample.
4000 4500 5000
mu+ mu- mass
0
1500
2000
2500
3000
3500
4000 4500 5000
mu+ mu- mass
Optimized has narrower signal,
perhaps reduced background tails.
µ-PRS Meeting: September 16 , 2003
R. Cousins, J. Mumford UCLA
15
H→4µ Full Analyses
Two groups have started investigations using
CMSIM & ORCA
Î
Karlsruhe
‡ Concentrating on low mass region MH < 2MZ mostly
Î INFN Firenze
‡ Concentrating on high mass region MH > 2MZ mostly
‡ CMS-IN 2003-046
Work is under review by PRS Higgs and Muon groups
Î
Only results approved during a PRS plenary meeting can be
released publicly for conferences (Nov. CPT Week in this
case)
EMU Meeting, 19 Oct 2003
14
Preselection
Signal and background after preselection (L = 20 fb-1):
Selection optimized for MH > 2MZ and MH < 2MZ .
Matteo Sani - PRS mu, 14/10/2003
7
Selection (MH > 2MZ)
di-muon mass: both muon pairs with an
invariant mass
mµ + µ − − M Z < 8.0 GeV/c2
transverse momenta
– thresholds at 20, 15, 15, 10 GeV/c
pT of the Higgs candidate
Matteo Sani - PRS mu, 14/10/2003
8
Efficiencies (MH > 2MZ)
reco
Z mass + pT
pT4µ
tt
42.4 %
0.05 %
0.05 %
Zbb
40.2 %
0.3 %
0.3 %
ZZ
68.6 %
35.8 %
15.2 %
H (200 GeV/c2)
71.4 %
49.4 %
37.4 %
H (300 GeV/c2)
76.7 %
51.5 %
42.1 %
H (500 GeV/c2)
76.6 %
50.6 %
43.8 %
Signal and background
after the selection
described before
(L = 20 fb-1)
Matteo Sani - PRS mu, 14/10/2003
10
1
fb )
Results (L = 20
(MH < 2MZ)
(MH > 2MZ)
|mµµ- MZ|< 6 GeV/c2
|mµµ- MZ|< 8 GeV/c2
pT thres. (20, 15, 10 , 7 GeV/c)
pT thres. (20, 15, 15, 10 GeV/c)
iso.: ∆R=0.2, pTmax = 4.0 GeV/c
four µ system pT > 15 GeV/c
-0.9 < cosθD < 0.8
MH (GeV/c2)
Higgs
ZZ(*)
tt
Zbb
130
5.0
0.65
< 0.5
1.2
150
10.9
0.71
< 0.5
< 1.7
200
19.4
2.5
0.35
< 1.7
300
13.9
2.0
< 0.5
< 1.7
500
7.0
1.5
< 0.5
< 1.7
Matteo Sani - PRS mu, 14/10/2003
Events falling
inside ±2σH
When no events are
found the upper
limit at 68 % C.L. is
given
16
after all cuts (20 fb-1)
after isolation cut with default value
after all cuts (20 fb-1)
after optimized isolation cut
isolation cut improves signal visibility significantly
Valeria Bartsch, CPT week, May 2003
9
mH [GeV]
Valeria Bartsch, CPT week, May 2003
12
Conclusions
With CSC production nearly over, FAST sites nearly
finished, and installation well in hand:
Î
Need to transfer experience to PRS groups!
‡ Especially in the areas of local reconstruction, test beam
analysis, calibration & alignment schemes, data handling
& monitoring
Physics studies are starting to drive reconstruction
improvements
Î
Muon reconstruction in Z’ search
Î Isolation techniques for H→4µ studies
EMU Meeting, 19 Oct 2003
15
Recent Agenda
EMU Meeting, 19 Oct 2003
16