HCALJETMET Status and Plans

1
HCAL--JET/METStatus and Plans
S.Eno/S.Kunori, U. Maryland For the HCAL-Jet/MET
group
Salavat Abdullin, Sylvia Arcelli, Sunanda
Banerjee, Sudeshna Banerjee, James Branson, Jim
Brooke, Pamela Chumney, Jordan Damgov, Sridhara
Dasu, Isa Dumanoglu, Sarah Eno, Dan Green, Pal
Hidas, R. Kinnunen, V. Konoplianikov, Andrei
Krokhotine,Olga Kodolova,Vladimir Kolosov,Victor
Kolosov, Shuichi Kunori, Kajari Mazumdar,
Alexandre Nikitenko, Alexei Oulianov, Wesley
Smith, Eliane Trepagnier, A. Urkinbaev, Irina
Vardanyan , Hans-Peter Wellisch
2
Outline

• For the DAQ TDR
• develop fast, robust, highly efficient
  algorithms for reconstruction of jets and MET for
  the HLT
• Calculate rates for jet and MET HLT triggers
• develop trigger strategy for SUSY to jetsMET,
  Invisible Higgs, and dijet resonances
• understand the sizes and bandwidth needs of the
  HCAL calibration samples
• verified that the HCAL data size is reasonable
  in view of the DAQ bandwidth
• In preparation for the Physics TDR
• develop algorithms for the reconstruction of
  jets and MET that optimize resolution
• develop and certify a GEANT4-based simulation of
  the HCAL within OSCAR

3
HLT Jet Algorithm
These trigger studies use a simple cone
algorithm, with no splitting/merging ?
appropriate for HLT since fast
Cone size 0.5, seed threshold 2 GeV, minimum jet
ET 10 GeV Optimized at high luminosity for
resolution and fake jet rejection
4
Jet Calibration
e/p goes as E, not ET. Physics goes at ET.
Plus, Different technologies
Calibrate both Level-1 and HLT jets (shown for
low lum)
5
Jet Resolutions
HLT jet resolution at low and high luminosity
6
Jet Rates
Insensitive to details of the simulation
High lum
Low Lum
7
Jet Rates Low Lum
Rates very similar at L1 and HLT
L1
HLT
1 kHz at Level-1 180 GeV (1 jet), 95 GeV (3
jet), 80 GeV (4 jet) 1 Hz at HLT 660 GeV (1
jet), 250 GeV (3 jet), 150 GeV (4 jet)
8
Jet Rates High Lum
Rates very similar at L1 and HLT
L1
HLT
1 kHz at Level-1 250 GeV (1 jet), 125 GeV (3
jet), 110 GeV (4 jet) 1 Hz at HLT 865 GeV (1
jet), 330 GeV (3 jet), 200 GeV (4 jet)
9
HLT MET Algorithm

• Tried 3 algorithms
• simple sum over towers
• using jets with jet energy scale corrections and
  unclustered towers (type-I)
• using jets with jet energy scale corrections and
  unclustered towers corrected using the jet energy
  scale correction for a 30 GeV jet (type-II)
• Simplest works as well as others at trigger level
  (type II works better in some offline analyses)

10
MET calibration and resolution
Mean and RMS of the difference between generator
and HLT MET for 3 different HLT MET algorithms
Low luminosity SUSY events
11
Met Rates
Very sensitive to details of the simulation (and
a great way to find bugs in the calorimeter
simulation ?)

• Sensitive to hardware problem.
• Need good monitoring and quick
• cure.

Raw rates are very difficult to interpret because
of the large and algorithm-dependent difference
between HLT and generator MET
12
Additional Dijet suppression
Only useful for large MET
13
SUSY Intro
Considered MSUGRA with and without R-parity
violation With substantial MET due to stable
Without difficult signature events with many
jets. Some MET due to W, heavy flavor, etc in
decays Points chosen to give a variety of
signatures
Which one has the R-parity violation???
14
SUSY Intro
MSUGRA with A00, tan(b)10, and mgt0
Low lum points near Tevatron reach Used for low
lum optimization High lum points near LHC
reach Used for high lum optimization
Optimized sum of efficiencies for a fixed rate to
tape
15
SUSY Low Lum Level 1
not hard With just two triggers 1 jet gt 130
GeV 3 jets gt 60 GeV We can have efficiencies of
.89,.90,.81,.94,.95,.88 for points 4,5,6,4R,4R,6R
(total efficiency, not with respect to offline
cuts) and a rate to tape of 2 kHz. Add more
triggers and can get a little more.
16
SUSY Low Lum HLT
harder

• METgt170
• 3 jets gt 60 and METgt 110
• 4 jets gt 120
• 1 jet gt 190, METgt90, and Df(j1,j2)gt0.5
• 2 jetsgt40, METgt100, and Df(j1,j2)gt0.5
• 4 jetsgt80, METgt60, and Df(j1,j2)gt0.5
• gives efficiencies of 0.78, 0.74, 0.54, 0.38,
  0.27, 0.17 for points 4,5,6,4R,5R,6R and a rate
  to tape of 3 Hz.
• Some of these are triggers with very steep rate
  curves (4 jet)

17
SUSY High Lum
almost too easy
Level 1 1jetgt90ltMETgt80 3jetsgt90
.90,.98,.92,1.0,1.0,.99 for 2 kHz
HLT METgt 170 4jetsgt150
.95,.94,.83,.90,.89,.65 for 3 Hz to tape
18
Invisible Higgs
Low lum easy to get egt.95 for 0.1 Hz to
tape High lum e0.6 for 0.1 Hz to tape
Level 1 1 jetgt60, METgt65 Low Lum (e0.98) ,
gt70,112 High Lum (e0.79)
Working on other trigger strategies for high lum
(M cut, Df cut with prescaled background trigger
without the Df cut)
19
Dijet Resonances
Limits from Tevatron W 720 920
GeV Z 720 940 GeV E6 diquark
570 780 GeV Axigluon 1160 1300 GeV q
910 1180 GeV
Unprescaled 1 jet trigger is 660 GeV (x2?mass of
1320 GeV)
20
Dijet Resonances
Time to have a statistically significant (5s)
number of Zs in a window around the mass for a
constant 20 Hz to tape
Blue dynamic prescale Red fixed prescale
unprescaled
21
CPU Analysis

• Time is small (budget is 300 ms/event)
• Similar for physics and for backgrounds
• Tower building dominates at low luminosity
• Tower building and jet reconstruction about equal
  at high luminosity

22
Calibration General

• Relative calibration of scintillator tiles
• during construction collimated gamma source
• track changes during running movable gamma
  source tubes
• track changes during running minimum bias data
• Linearity of electronics and timing
• UV laser and blue LEDs
• HCAL Energy scale
• test beam data source data - ADC - GeV
• single particle (QCD events, tau trigger)
• Jet Energy Scale
• g jet data, Zjet data, di-jet data
• W-gtjet data (in ttbar events)

23
HLT for calibration g jet
Several processes will be used for calibration of
jet energy. Among those considered so far (e.g.
Zjet, ttbar etc.), only gj requires a dedicated
HLT.
Statistical error after 3mo. running with 1Hz at
2E33
g jet trigger For
ETgt80GeV - HLT single photon For
ETlt80GeV - L1 e/gamma (ETgt23GeV) -
very tight g isolation with pixel and
ECAL - pre-scale
1
.1
.01
100
1000
ET(g)
24
F-symmetry calibrationwith min-bias events
Process min-bias events in the HLT farm at
1kHz and send histograms to tape.
(1.4 hours)
Accuracy ()
( 5.6 hours)
eta
25
Energy Flow for Jets
Use tracking information to improve jet
reconstruction Used at LEP and CDF
Improves energy scale!
Improves resolution!
26
toward Physics TDR
27
GEANT4
Physics TDR will use OSCAR/GEANT4. Important
that this be ready and well tested

• Done
• HCAL geometry and hits are in OSCAR
• GEANT4 hadron shower generation verified to be
  as good as GEANT3 at describing our 1996 test
  beam data
• To Do
• HCAL needs to be integrated into new OCAR
  framework for geometry,hits, user inferface
• shower library for HF needs to be integrated
• more testing/comparisions for single particles,
  jet, MET between CMSIM/OSCAR
• compare to 2002 test beam

28
Summary

• have
• satisfactory algorithms for reconstructing jets
  and MET in HLT
• calculated rates for jets and MET at low and
  high luminosity
• checked our bandwidth and CPU requirements are
  in budget
• designed satisfactory trigger tables for SUSY,
  dijet resonances, and invisible Higgs
• checked the bandwidth needed for calibration
  samples is reasonable
• started promising work on algorithms to use
  offline (for physics TDR)
• exercised our software and production facilities
• written our portion of the DAQ TDR

THANKS TO ALL THE PARTICIPANTS IN THE
HCAL--JET/MET GROUP!!!