Summary of HLT Studies by the JetMET Physics Group

1
Summary of HLT Studies by the Jet/MET Physics
Group

• Shuichi Kunori
• U. of Maryland
• 21-July-2000
• Group Leader Sarah Eno http//home.fnal.gov/scen
  o/jpg/Default.html

2
Milestone and Tasks
(shown on June 6th.)

• July00 Milestone
• Demonstrate that data coming from Level-1 trigger
  can be reduced by a factor of 10 using
  calorimeter information.
• Tasks
• Software tools
• Verify ORCA4
• Create ntuple-files
• Understand offline performance for jet/met/tau at
  10E34.
• Develop algorithms for Level 2 (and implement in
  ORCA).

3
HLT algorithm

• Jets
• Improve Et resolution and Et scale
• improved jet axis with finer granularity (1x1
  instead of 4x4 towers)
• LSB L1(Et0.51.0GeV) -gt L2(Et0.2GeV)
• energy calibration (non-linearity, pile-up
  energy)
• Reduce low Et fake jets due to min-bias pile-up
• jet shape smaller window size
• MET
• Improve resolution
• energy calibration (non-linearity)
• Tau
• improve shape cuts at L2
• use tracker information for isolated charged
  tracks at L3.

4
Software Tools

• The software chain finally established.
• CMSIM- no known bug
• To be updated before next production.
• new eta-phi segmentation - cannot mix with 116
  data.
• ORCA4- big improvement
• Random mixing of min-bias events.
• full pile-up activated - very critical for Jets
  MET studies.
• Re-organization of Calorimeter code
• clean software - good start for future, but
  required large effort to re-establish the
  software chain.
• Ntupel Maker - working!
• Standard ntuple files
• been shared by many members.
• For Jets/MET only
• need to establish common ntuple files for all
  groups.

5
Monte Carlo Samples

• Large data set.
• 100k min-bias events
• 600k QCD events (10-470GeV)
• 10 sets of signal sample
• single particle samples for verification.
• Thank for good production team and system.
• Need planning for next step
• event filter at particle level before cmsim
• an extended list of signal/background samples
• production with different parameters for special
  studies, e.g. study of front end electronics.
• Even larger event sample, e.g. 0.51.0E6 of
  min-bias events for pile-up.

6
Tau Trigger
L1
- New L1 Tau trigger has been introduced since
Nov99. - a big improvement (29-gt64)
for the low mass Higgs (200GeV) () -
High mass Higgs may be triggered by L1 jet
trigger.
() gg-gtbbA/H, A/H-gt2t-gthh-X off-line cuts
Et(t-jet)gt60GeV eta(t-jet)lt2.4, 1 prong
ts.
7
Tau Trigger
L2
L2 Tau trigger operates on the 1st L1 tau.
1) reconstruct a jet 2) calculate e.m.
isolation P(isol)
Etecal(Rlt0.4) - Etecal(Rlt0.13) 3) accept
event if P(isol) lt Pcut.
New isolation cut provides constant efficiency
over large Et region.
(efficiency for ts passed L1)
8
L2 Tau Trigger
L2 reduction 10 at 65 eff. 4kHz -gt
400Hz
L3 Use tracker information
9
Jet Trigger
Problems to solve - large QCD jet rate -
pile-up at E34 - signal pulse spread over
multiple beam crossings. - non-linear
response of calorimeter - Et-eta
dependent - difficult to trigger low Et
jets New 12x12 sliding algorithm has been
introduced in L1 since Nov.99.
Eta-Et dependent correction Et(corr) a0 a1
Et(obs)
10
Jets
Jet Energy resolution before/after correction
- Simple energy correction improves overall
resolution. - Next step - apply this to
L1/L2.
11
Low Et Jet Trigger
12x12 window vs 6x6 window
Problems to solve - large effect by
pile-up. lt17GeVgt in Rlt0.6 due to
min-bias events at E34. -gt Fake
jets - Sensitive to threshold cuts.
(0.5 GeV threshold)
Smaller cone/window pick less pile-up energy,
(I.e. less fake jets)
Rate(12x12)/Rate(6x6) 2.1 for 40GeV for pile-up
events.
12
Low Et Jets ResolutionDependence on window size
and threshold
(35GeV lt Et(genjet) lt 45 GeV)
12x12
6x6
Thres. 0.0 GeV
1.0 GeV
0.5GeV)
Better resolution - larger window - lower
threshold
Next step on -
apply smaller window low Et QCD multijet
sample - apply some shape cut qqH -
energy correction WH
13
MET Trigger
L2/Offline MET
Problems to solve - Long tail due to
pile-up - sensitive to pedestal
pile-up energy subtraction - depend
on threshold - Energy scale - Sensitive to
noise - Sensitive to dead channels
5kHz at MET80 GeV
(L1 rate 810kHz)
14
MET Trigger
Contribution from QCD Jets
How can we reduce and control the rates? -
optimize threshold higher threshold
less pile-up worse
resolution - energy calibration - combine
with jets/leptons 50 reduction with
jet gt 80GeV (???)
Next - Calculate L1/L2 rates - Efficiency
curves for various signal channel. -
Understand offline performance.
L2/offline MET
15
Plans for July Milestone
(as of June 6th)
1) redo L1 rates - done for tau and MET 2)
construct maps for eta-Et dependent jet energy
correction for offline (L2) and L1 - done
(almost) 3) do L1/L2 jet rates with jet energy
correction 4) see effect of jet energy
correction on offline MET 5) do L2 MET rates
with jet energy correction, if 4) works. 6)
calculate L2 tau rejection rate with new L1 tau
trigger. - done 7) support ejets by the egamma
group 8) start developing a common ntuple for
jets/met/tau, electrons/gamma and muons.
16
After July Milestone
(plan as of June 6th)
1) tracker information for L3 tau trigger. 2)
proper HCAL electronics simulation.
(pulse shape, time jitter etc.) 3) optimize
offline MET calculation 4) optimize handling of
low Et jets (Etlt40GeV) 5) improve jet energy
calibration and pile-up energy
subtraction 6) migrate to a common ntuple. 7)
production with new cmsim (updated geometry,
larger HF shower library, etc.)
( pixel for jets ?)
(pedestal subtraction)
17
Conclusion

• Software Tools
• cmsim - orca4 - ntuplemaker finally working!
  (almost)
• L2 Tau
• Isolation cut provided a factor of 10 reduction
  over L1 with efficiency 65.
• Next - use tracker at L3.
• Jets / MET
• Analysis of ORCA4 data just started.
• Follow the plan shown on June 6th.
• Many parameters and algorithms to test.
• Threshold, LSB, pedestalpile-up energy
  subtraction, cone window sizes, energy
  calibration, use of tracker/pixel etc.
• Low Et Trigger and MET trigger are real challenge
  at 10E34, Hope to contribute to triggers on low
  mass scale physics.