cms and a light higgs with t t
DESCRIPTION
CMS and a Light Higgs with t t. J. R. Incandela University of California Santa Barbara. Ht t Group: The goal was to perform a realistic study of the feasibility of detecting SM Higgs in this channel. CMS Ht t study. Recently completed study for the CMS Physics TDR vol. 2 - PowerPoint PPT PresentationTRANSCRIPT
CMS and a Light Higgs with ttCMS and a Light Higgs with tt
J. R. IncandelaUniversity of California Santa Barbara
J. R. IncandelaUniversity of California Santa Barbara
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 2
Htt Group: The goal was to perform a realistic study of the feasibility of
detecting SM Higgs in this channel
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 3
CMS Htt study
Recently completed study for the CMS Physics TDR vol. 2
This is a publicly available note: (49 pages).
UCSB (JI) Led the group through the completion of this effort and note.
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 4
Overview
• Analysis involved 4 subgroups • Standard top channels: Dilepton, All-hadronic, and e/m + jets
• Largely independent over past two years
• Goal: As realistic as possible• Include backgrounds with higher order processes (Alpgen)
• Include multiple interactions (aka pile-up)
• Fully simulate the detectors
• Develop and use realistic algorithms:• Electron and muon identification
• Jet and missing Et reconstruction
• b and c jet tagging, and light quark/gluon mis-tagging
• A huge amount of work!!
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 5
Light H in conjunction with tt
Production Modes
Decay Mode
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 6
ISR
H→bb
MPI
t→W
- b
W+b←t
ET > 3 GeV
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 7
Backgrounds
•Main backgrounds are ttjj, ttbb, ttZ• ttjj dominates numerically even though a mis-tagged light quark
or gluon jet is required• The ttjj xsec is nearly 3 orders of magnitude higher than signal
• Should one get beyond ttjj, one must still confront ttbb, and ttZ with Zbb, which are irreducible
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 8
Generators and Cross sections
What’s new:
Used a more sophisticated generation scheme for the ttNj background
•PYTHIA alone under-estimates the hard radiation
•CompHEP overestimates
•Alpgen+MLM matching is “just right”
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 9
ALPGEN v.2 & MLM
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 10
Triggering
• After full CMS detector simulation and digitization, trigger simulations were run.
• Typically a bit lower efficiency than we ultimately expect• More complex and efficient triggers will likely be available. .
Trigger
offline
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 11
Leptons
•Likelihoods • Developed explicitly for Htt
•Categorized in Monte Carlo (MC)
• SIGNAL• Matched to lepton form W
• use an cone of radius 0.1 (0.01) for electrons (muons)
• BACKGROUND• All others:
• leptons from b or c hadron decays, fakes
)()(
)(
ibkg
iisig
i
isig
ii xPxP
xPL
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 12
Muon Reconstruction
• Muon likelihood: 4 “obvious” variables:
• Muon pT
• Track Isolation
• Calorimeter Isolation
• 2-D Impact Parameter significance
SIPPTCalo Iso
Track Iso
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 13
Muon Reconstruction
• 90% for signal and 1.0% for background • calculated from the semi-leptonic Htt sample
cut is –Log(L)<1.4
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 14
Electron Reconstruction
• Electron Likelihood: 5 variables:
• pT
• E/p
• Had/EM
• pT tracks inside R=0.3 cone and
outside veto cone (R=0.015)
• R between electron candidate and closest track outside veto cone
R ≡ √( ∆η2 + ∆φ2 )
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 15
Electron Reconstruction
• 84% for signal 1.5% for background
cut is –Log(L) < 1.3
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 16
•JETS• Iterative Cone Algorithm
R=0.5 (0.4 for All-had)
• ET > 20 (25 All-had)
• |η| < 2.5 (2.7 All-had)
• Use MC calibrated jets
• Remove electrons that match within R < 0.4
R ≡ √( ∆η2 + ∆φ2 )
Signal (ttH, MH = 115 GeV)
Raw (IC 0.5)
MonteCarlo calibrated
γ-jet calibrated
Jets
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 17
Jet definition: All Hadronic Case
• Different Cone Sizes Tested
• Signal and 3 most dangerous Bkg
used for testing (ttbb, tt2j, qcd170)
• 8 most energetic jets in ||<2.7 and
ET>25GeV
• Jet-Parton pairing
2 for masses of 2 W and 2 top
within 3 sigma
• Jets paired to b-parton have to be
b-tagged
2222
2
)()()()(
t
jjt
t
jjt
W
jjW
W
jjWmass m
mm
m
mm
m
mm
m
mm
Cone R=0.4 chosen
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 18
ET(ecal + hcal)
− ∑ [ET(calib)−ET(raw)]
− ∑ pT(μ)
Missing ET
Calorimeter tower measurements
Jet corrections
Muon momenta
In semi-leptonic Htt channel/
Missing Momentum
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 19
bTagging
• Combined Secondary Vertex Algorithm
• Mistagging rate as a function of b-jet efficiency for signal (left) and ttjj
(center) are shown for various types of jets
• Gluon jets include splitting to bb or cc in center plot...
• Tag Rates for b-, c- and uds-jets vs discriminator cut for ttjj sample (right)
c
udsuds
c
g
cb
uds
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 20
Semi-Leptonic Selection
• Preselection• HLT + Isolated Lepton + 6 or 7 Jets• ET > 20GeV • 4 bTagged jets
• D>0.5 (70% bTag efficiency)
• Veto events with two leptons, or wrong lepton.
• Jet Pairing• Likelihood method: Levent=LmassxLbTagxLkinematics
• Mass refers to likelihoods for the obtained masses for hadronic W and tops
• Kinematics • Takes into account b jets from top quarks slightly more
energetic than those from H or jj (rather complicated formula…)
• LbtagLbsele=Dh1xDh2xDbTopHadxDbTopLep
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 21
Semi-Leptonic Selection
Example of performance for muon selection
Chosen working points are: 0.55 and 0.72
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 22
Final results 60 fb-1
Muon Channel
A bit less S/ √B cause mass constraintbut betterS/N
Electron Channel
No discrepancies at Just less efficiency in e channel for HLT and isolation.
S
√B
= 2.35
Muon + electron, no systematics
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 23
di-Lepton Selection
• Signal• H forced to bb
• One W forced to (e,,); the other free • Find large contribution from single lepton events (1 real + 1 fake
lepton)
• Preselection• 3 b-jets and D>0.7
• Selection• 2 leptons (e,) passing Likelihood criteria
• –Log(Lmu)<1.8 and –Log(Lele)<1.3 with pT>20 GeV
• At least 4 jets with ET> 20 GeV• No additional tagging requirement
• Corrected ETmiss > 40 GeV
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 24
Full Hadron Selection• Same analysis as Jet Algo choice and same mass for jet pairing
• 8 most energetic jets in ||<2.7
• Centrality Cuts
mass for 2W and 2tops within 3 sigma from expected values
• 2 working points
• Low S/N Higher Significance
• ET(7th)>30GeV – ET(8th)>20GeV ordered ET jet
• 3 out of Dh1 Dh2 DbTopHad1 DTopHad2 > 0.80
• CentH>0.55
• High S/N Lower Significance
• ET(7th)>30GeV – ET(8th)>20GeV ordered ET jet
• Dh1 Dh2 DbTopHad1 DTopHad2 orderd in D; D(3th)>0.85 and D(4th)>0.70
• CentH>0.55 – CentAll>0.80
jet
jetT
E
E
8
8All Centrality
jet
jetT
E
E
2
2Higgs Centrality
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 25
Full Hadron Selection: Results 60 fb-1
Analyzed Ev. Sel.Eff . Opt.1(%) Nexp 60fb- 1 Sel.Eff . Opt.2(%) Nexp 60fb- 1
ttH (115) 49636 0.294 44 2.317 347
ttH (120) 163494 0.366 45 2.550 314
ttH (130) 43254 0.358 27 2.797 214
ttbb 203135 0.0645 109 0.7015 1187
tt1j 1031551 0.0005 49 0.0084 860
tt2j 559111 0.0009 54 0.0333 1996
tt3j 68015 0.0015 35 0.0794 1905
tt4j 97334 0.0021 75 0.1818 6656
ttZ 80226 0.0312 11 0.3577 121
q170 264310 0.0004 76 0.0238 4805
q120 55128 0.0000 0 0.0018 83
Total 410 17614
M(H)=115GeV S/sqrt(N) 2,17 S/sqrt(N) 2.61
S/N 10.7% S/N 2.0%
M(H)=120GeV S/sqrt(N) 2,23 S/sqrt(N) 2,37
S/N 11.0% S/N 1.8%
M(H)=130GeV S/sqrt(N) 1.33 S/sqrt(N) 1.61
S/N 6.5% S/N 1.2%
An example: tables of this type are in the note for all channels
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 26
Speculation on Mature Experiments
• The mature CMS working point is taken to have the following systematic uncertainties:
• Flat 3% JES
• 10% Jet Resolution
• 4% in bc-Jet tagging efficiencies
• 10% in uds-Jet tagging efficiencies
• 3% in luminosity
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 27
Single Lepton Table
• Uncertainties from JES and uds-Tagging efficiencies
• Most dangerous BKGD are tt1j and tt2j
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 28
Single Leptons 60 fb-1
•Significances are drastically reduced once reasonable systematic uncertainties are included
•Led to re-optimization with a looser selection
• Still pretty grim
Muon
Electron
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 29
Di-Lepton Table
• Again - big uncertainties from JES and uds-tagging
• Most dangerous BKGDs are ttNj
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 30
Hadron Table
• Again big uncertainties from JES and uds-Tagging efficiencies but…Surprise! The most dangerous background is QCD (look at JES) and then tt4j, tt3j and tt2j…
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 31
The message!
Jet Energy Scale
and
light quark jet mis-tagging
are major systematics
Furthermore, when signal and background production
systematics are included, it gets worse.
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 32
Cross Section uncertainties
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 33
CMS-CDF comparison
• Exercise performed with the diLepton channel• CMS, CDF tag uncertainties taken to be the same (4% bc and 10% uds)
• Some indication we’re not too far off the mark.
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 34
Is there any hope?
• A couple of possible mitigating factors:1. Backgrounds from Data: 60fb-1 of integrated luminosity will
provide plenty of data for which detailed studies can be performed to understand the detector and algorithms
2. The availability of large control samples of top events will enable b tagging of high energy jets to be very well understood. This will probably enable some further suppression of light quark and charm jet tagging relative to b tagging. Similarly, experience with real data will likely improve jet reconstruction and energy measurements.
• Will they be enough?
• How much work and how much data will it take?
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 35
Summary
• Statistical Significance for the 4 sub-channels lower than previous studies but still combine to greater than 2.0 in 60 fb-1.
• But the systematic uncertainties: JES and uds-tagging are major problems
• They eliminate all sensitivity! (Combined significance of less than 0.2 for all channels)
• Real data will give the final answer on the feasibility of this channel, but all indications are that it will be very difficult.
More Information
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 37
Muon Selection: Results 60 fb-1
Analyzed Ev. PreSel. Eff .(%) Nexp 60fb- 1 Sel.Eff . Opt.1(%) Nexp 60fb- 1 Sel.Eff . Opt.2(%) Nexp 60fb- 1
ttH (115) 27768 5.66 271 2.3 111 1.13 54
ttH (120) 91447 5.66 223 2.3 91 1.1 43
ttH (130) 19467 6.3 145 2.0 65 1.23 30
ttbb 372737 0.92 1573 0.30 503 0.13 221
tt1j 393000 0.08 8616 0.007 675 0.0013 130
tt2j 569000 0.15 8878 0.014 843 0.0021 126
tt3j 110000 0.14 3493 0.009 218 0.0009 22
tt4j 86697 0.05 1857 0.005 169 0.0000 0
ttZ 50000 0.27 90 0.078 26 0.034 12
Total 24507 2433 511
M(H)=115GeV S/sqrt(N) 2.25 S/sqrt(N) 2.4
S/N 4.5% S/N 10.3%
M(H)=120GeV S/sqrt(N) 1.85 S/sqrt(N) 1.9
S/N 3.7% S/N 8.5%
M(H)=130GeV S/sqrt(N) 1.3 S/sqrt(N) 1.3
S/N 2.7% S/N 5.9%
P.K.’s visit to UCSB, September 14, 2006 J. Incandela 38
Electron Selection: Results 60 fb-1
Analyzed Ev. PreSel. Eff .(%) Nexp 60fb- 1 Sel.Eff . (%) Nexp 60fb- 1
ttH (115) 27500 4.25 204 1.26 60
ttH (120) 100000 4.28 170 1.26 50
ttH (130) 24500 4.50 111 1.25 31
ttbb 420000 0.7 1188 0.17 286
tt1j 528000 0.08 8500 0.0038 386
tt2j 286000 0.14 8266 0.0056 336
tt3j 110000 0.12 2924 0.0045 109
tt4j 102000 0.10 3732 0.0039 143
ttZ 50000 0.20 75 0.05 16
Total 24685 1276
M(H)=115GeV S/sqrt(N) 1.7
S/N 4.7%
M(H)=120GeV S/sqrt(N) 1.4
S/N 3.9%
M(H)=130GeV S/sqrt(N) 0.86
S/N 2.4%