Jets and Met Group

1
Jets and Met Group
Many potential sources of new physics have final
states consisting only of jets and MET.
Jets MET is a classic SUSY signature h,H,Att,
ts decay hadronically
In addition, final states containing soft
leptons may not have acceptable trigger rates if
we only use lepton-related trigger requirements
h,H,Att, one t decays leptonically
the other hadronically bbHbbgg wino/zino
3lMET sleptons 2lMET
2
The Challenges

• Jets
• lots of them around
• very slow turn-on curve because the jet ET
  spectra is so steep, causes a very large number
  of jets that are below your offline thresholds to
  cause triggers
• The Jets ARE real. The whole trick here is
  sharpening the turn-on curve by making online as
  similar to offline, and the offline as similar to
  the parton-level as possible!

• MET
• again, slow turn-on curve
• sensitive to tails of the multiple interaction
  distribution
• sensitive to any potential noise source (CDFs
  texas towers, DØs Uranium noise)
• Optimization of MET resolution, to sharpen the
  turn-on curve and to prevent false MET in QCD
  events, especially in the presense of multiple
  interactions, is important. Elimination of noise
  sources is also important.

A look at Level 1clearly shows the challenges...
3
The Challenges
Level 1 uses non-overlapping window algorithm
(very non-offline)
If you set the single jet ET threshold to be 95
efficient for 205.5 GeV jets, you will trigger on
jets with ET down to 100 GeV!
for HLT, want algorithms, calibration as
similar as possible to offline to avoid this
affect!
Sridhara Dasu
4
The Challenges MET
Level 1
Sridhara Dasu
conversely, an 80 GeV MET threshold will only be
95 efficient for parton-level METs of 200 GeV.
Not fully efficient for any kinematic region for
squark and gluinos. Dasu expects similar
results for offline versus parton Met.
experience from CDF/DØ tell us these turn on
curves vary according to event topology, and are
difficult to calculate or to estimate from data.
need strategies for getting the efficiency for
the combined L1/HLT turn-on.
5
Plan
1) Understand how to achieve a factor 10
rejection over L1 using 25 of the data. 2)
while maintaining reasonable efficiency for
difficult channels in the high luminosity physics
menu SUSY jetsMet h,H,Att, ts decay
hadronically h,H,Att, one t decays leptonically
the other hadronically
6
Basic Task List
It will be a long road the current jets and MET
effort is embryonic.
In random order...
1) Level 1 2) Monte Carlo Tuning 3) Monte Carlo
Generation for Backgrounds 4) What
Signals? 5)Design of Trigger Tables 6) Level 2.0
7) ORCA code for GHITS to DIGIS 8) ORCA code
for jet clustering, corrections 9) ORCA code for
MET 10) L2.x specific versions of the ORCA
code 11) rejection, efficiency, data-size, rate
calculations 12) methodology for determining
efficiency

• Manpower situation
• more details on each topic

7
Man Power
Minimum Requirements 8 full-time people.
People on JPG Mailing List (26, besides the
initiators)
Virgil Barnes, Serguei Bitoukov, Volker
Drollinger, Alvise Favara, Vladimir Gavrilov, Dan
Green, Nikolai Krasnikov, Alexandre Nikitenko,
Hans Paar, David Stickland, Hans-Peter Wellisch,
Sedat Ayan, John Hauptmann, Ed McCliment, Eugeny
Doroshkevich, Vivan ODell, Shuichi Kunori, Nural
Akchurin, John Womersley, Sridhara Dasu, Dmitri
Litvintsev, Weimin Wu,Harvey Newman, Lawrence
Sulak
8
People Signed Up to Work
1) Level 1 Sridhara Dasu, Vivan ODell could
use more in this essential area??? 2) Monte Carlo
Tuning Shuichi Kunori, Nural Akchurin, Dmitri
Litvintsev manpower okay? 3) Monte Carlo
Generation for Backgrounds Weimin Wu, Vivian
ODell manpower okay? 4) What Signals? ???
5)Design of Trigger Tables ??? 6) Level 2.0
??? 7) ORCA code for GHITS to DIGIS Shuichi
Kunori., Sarah Eno manpower okay? 8) ORCA code
for jet clustering, corrections Hans-Peter
Wellisch (Volker Drollinger, Sasha Nikitenko,
Lars Sonnenschein, Evgeny Doroshkevich) DØ had
about 9 people working in this area 9) ORCA code
for MET John Hauptmann could use more in this
essential area? 10) L2.x specific versions of the
ORCA code ??? 11) rejection, efficiency,
data-size, rate calculations ???
9 people, all but 2 (?) part-time (some may be
very part time!)
9
Backgrounds
For the November milestone (factor 10 rejection)
we are currently planning on using the
brute-force approach of generating many
background events. For the future, we may need
to re-think this strategy, and develop
parameterized methods, similar to those used in
the past by E. Meschi for HLT electron rates.
Vivian ODell and Weimin Wu of Fermilab are in
charge of this project.
10
Monte Carlo Generation
Fermilab has a 10 Sun farm, and is willing to do
some large-scale Monte Carlo generation. CIT and
Wisconson also have similar facilities, and are
willing to help. Simulation can start once 1)
farm is commissioned (beginning May?) 2) We
understand what HF simulation we will use We are
currently planning on simulating 500,000 QCD
events ( in same mixture as Dasu used for his
Level 1 studies) and 50,000 Min Bias events. We
expect to start in Mid-May, and to be finished by
the end of June. Output will need 100 8mm
tapes. We will keep 50,000 unbiased events, the
min bias events, and all events that pass the L1
trigger (1000) on disk at Fermilab (Dasu expects
his L1 work to be done by mid-May.
11
Level 1
1) get current FORTRAN L1 into ORCA Sridhara Dasu
(mid-April) 2) verify new code Vivan ODell
(mid-May) 3) get HF into L1 simulation (important
for MET) Sridhara Dasu (end-April) 4) check
effect of real digitization simulation on
rates ??? 5) verify that current L1 trigger table
is adequate for all the physics we want to do ???
12
Trigger Tables What Physics?
current high luminosity trigger table
L1 threshold
rate Scalar ET 400 GeV 0.48
khz MET 80 GeV 1.29
khz 1jet 100 GeV 2.06
khz 2jets 60 GeV 2.17
khz 3jets 30 GeV 3.16
khz jete 5012 GeV 1.35
khz (note, jets are 95 efficient at about 2xL1
thres met at 2.5xL1 thres)
benchmarked against SUSY with M(LSP)45 GeV,
M(gluino)300 GeV 83 efficient
but, what about Htt, ts decay
hadronically??? offline 2 jets gt60
METgt40
One issue this group needs to address this at the
same time as the HLT work is this adequate for
all the physics we need to do? What do we need
at L2? Monitor triggers?
13
Monte Carlo Tuning
MET rates may be sensitive to details of detector
simulation, especially near cracks (though,
Dasus studies indicate this is probably not true)
1) Decide on HF simulation strategy for mid-May
MC generation Shuichi Kunori, Nural Akchurin,
Dmitri Litvintsev in time for mid-May MC
generation investigating using latest version of
HF shower library (April 1 version?) 2) re-tune
HCAL simulation after summer test-beam Shuichi
Kunori September
14
Level 2.0
at L2.0, we see what rejection we can get
using the same data as was used in L1, but with
more effective algorithms For jets, we can
probably get our factor of 10 here in L2.0 just
by using a real clustering algorithm, instead of
the L1 fixed window algorithm. Have not yet
thought about MET. pieces needed to do this
have some manpower. need somebody to take
responsibility for putting them together
15
ORCA GHITS to DIGIs
Digitization and pileup effects could have a
substantial effect on the turn-on curves.
Need them to be simulated properly.
Design and implementation of C Paganini Pascal
and Romana Albert done by mid May? Customization
for HCAL Shuichi Kunori and Sarah Eno done by
end-of-May?
manpower okay?
16
ORCA code for jet clustering, corrections
Will want online code as similar as possible to
offline code, to prevent smear-up
effects. especially, want best possible jet
energy calibration (taking into account eta
dependency, etc) in HLT
Library of different clustering
algorithms Hans-Peter Wellisch mid-May Tau-jet
specific algorithms Sasha Nikitenko Using jet
shape and multiplicity to reduce backgrounds Lars
Sonnenschein Hcal calibration using single
pions Evgeny Doroshkevich sharpening
parton/offline turn-on, esp. for Hbb Volker
Drollinger
can use as many people as want to work
17
ORCA code for MET
no thought has gone into this yet, though we
do have 1 volunteer to do this (John Hauptmann)
18
Efficiencies, Rates, Rejections
If we can get L1 into ORCA, HF into L1,
digitization routines into ORCA, finish the Monte
Carlo generation and at least 1 ORCA jet routine
and 1 ORCA MET routine by the end of July, we can
see in August if we can get the factor of 10
simply by using real clustering algorithms
instead of the fixed window algorithms (probably
true) and see that the efficiency for SUSY
squarks/gluinos from the multi-jet trigger
remains high. We will then be in a position to
start to think about the more difficult Higgs to
tau triggers in Sept/Oct. And, well have built
a team capable of dealing with this type of
physics issue in the future. Will be challenging
with current manpower, especially if they are
distracted with other projects (as they are)
manpower needed for this part of the project!
19
Summary
1) We have started to get a team together to
work on this problem. However, more manpower
is certainly needed. 2) We have started to
identify strategies to get the required
information in time for the November
milestone. 3) It is possible to make the
milestone, but only if we get more manpower, and
if our current manpower is not too distracted by
other tasks
20
(No Transcript)
21
Administrative
Mailing List cms-jetsmet_at_cms-lb.cern.ch Web
Page http//hepd01.physics.umd.edu/JPG/Default.ht
m First Meeting March 26 Introduction by
me Overview of L1 jets and MET Dasu Discussion
of Task List all video meetings of this type
are painful. I would like to try to develop a
virtual meeting where people post their talks
on our web-page and the audience makes use of a
netforum-like interface to place questions and
receive answers. However, this will take time to
develop. We would then have real meetings at
CMS collaboration meetings. Second Meeting April
16. update on L1 Dasu MC generation at
Fermilab Odell HCAL digitization Kunori