Search for t q
FermilabMarch 28-29, 2006
Sarah Demers
Outline
Motivation
CDF and the Tevatron
Event Selection
Tau Identification
Result
The Standard Model
Encompasses 3 of 4 fundamental forces
Fundamental particles:
6 quarks6 leptons
Interactions mediated by force carriers
The Standard Model: What’s Missing?
Why are there three generations?
Why the large variation in quark masses?
Why is there so little antimatter in the universe?
What about gravity?
Top quarks and tau leptons
The Higgs BosonExtensions to the standard model may be neededMSSM Charged Higgs: t -> H+b, H+ ->gives identical final state
Result from CDF Run 1 (mid-1990s)Expected 3.2 events (0.7 signal events) with theoretical cross section
Found four events with three b tagged events
Need to improve signal to noise
Top quarks and tau leptons
Experience with tau leptons in the environment of a hadron collider
Room for new physicsThe “recently” discovered top quarkThe heavy third generation
The Tevatron
The Tevatron (and the LHC)
Tevatron LHC
proton-antiproton proton-proton
Beam Energy 1 TeV 7 TeV
Radius 1 km 4.24 km
Interactions/Crossing
3 20-30
Instantaneous Lum.
1032 cm-2s-1 1034 cm-2s-1
Time between bunch crossings
396 ns 25 ns
Fermilab’s Accelerators
Cockroft-WaltonHydrogen gas ionizedIons accelerated to 750 keV
Linac500 ft longOscillating electric fields accelerate protons to 400 MeV
BoosterCircular acceleratorProtons make 20,000 lapsAccelerated to 8 GeV
Fermilab’s Accelerators
Main InjectorAccelerates protons and anti-protons to 150 GeVInjects the particles into the Tevatron
Anti-protonsFrom 120 GeV proton beam extracted from the Main Injector
TevatronAccelerates to almost 1 TeVParticles move only 200 mph slower than speed of light
Collisions
Main Injector
Tevatron
DØCDF
Chicago
p source
Booster
Detectors
Event Selection: Decay Chain
top, anti-top events needed for statistics
t t
W+
b bW-
jet jet e e
Final State
t t
W+
b bW-
jet jet e e
hadrons(1 or 3 prong)
Strategy for this analysis
First CDF Run II top -specific analysisClosely follow {e+} dilepton analysis (but)
admit only the lowest background categories with tight, central electron and muon requirementsplace a premium on ensuring non-tau top final states are excluded
t t
W+
b bW-
jet jet e e
hadrons(1 or 3 prong)
Event Selection
Reconstructed tau passing all ID cuts, Et > 15 GeV
t t
W+
b bW-
jet jet e e
hadrons(1 or 3 prong)
CMS CR 2005/018
elemuon1 prong (had)3 prong (had)5 prong (had)
Tau decay modes
Event Selection
Tau identification requirements
No impact parameter
t t
W+
b bW-
jet jet e e
hadrons(1 or 3 prong)
Event Selection
Corrected Missing Et
greater than 20 GeV
Opposite sign tau and electron (muon)
t t
W+
b bW-
jet jet e e
hadrons(1 or 3 prong)
Event Selection
>= 2 jets, < 2.0
1st jet > 25 GeV
2nd jet > 15 GeV
Ht > 205 GeV
Z Mass veto
t t
W+
b bW-
jet jet e e
hadrons(1 or 3 prong)
Optimization
The HT and lead jet ET cuts are chosen by a formal optimization procedure
2D optimization with MC signal & data+MC bkgndMinimize S/sqrt(B), the stat. uncertainty in Gaussian limit in “no signal observed” caseMaximize likelihood ratio: LS+B/LB
Acceptance
Pythia Monte Carlo
Before scale factors:
eh : 50%
h : 42%
eh : 4%
h : 4%
50%
42%
4% 4%
Acceptance
35% ID efficiency from Monte Carlo
With W->, compare data to Monte Carlo
W+
Acceptance Summary
0.076 ± 0.005 (stat) ± 0.013 (sys) %
BR ~ 3% with ~2.5% efficiency
Expected signal:
1.00 ± 0.06 ± 0.16 events
Expected background:
1.29 ± 0.14 ± 0.21 events
Background Summary
jet fakes
We measure the jet to tau fake rate in 4 datasets:
20 GeV jet 50 GeV jet 70 GeV jet Large total event
Energy
Rates from 0.1% to 10%
e fakes
Measure e tau fake rate in data with Z ee
Electron veto variable (HadE/SumP) shown for loose leg Zee
We calculate a (1.2±0.3)% etau fake rate at our 0.15 cut
fakes
Z MC predicts background
Cross-check in data
Fakes are extremely rare, so data statistics only allow a cross-check
Agreement at level of 30%
Z
First require of missing energy, “taus” consistent with Z Then reconstruct mass by assigning MET to “taus”
Z
65 GeV < Mass < 115 GeV88% reduction of BG, 4% reduction of signal
The Result
We could report this result as a cross-section, as is done with other rate analyses
However, clearly this analysis has little to contribute to a cross-section average
The motivation for the analysis is a universality testWe quote:
Conclusions
We predict 2.3 events and see 2 events.
We set a limit on:
r < 5.2 at the 95% confidence levelFrequentist Method: profile likelihood (Rolke et al)
Acknowledgements
Thank you to the Fermilab group for inviting me!www-cdf.fnal.govwww.fnal.govwww.particleadventure.orglhc.web.cern.ch/lhc/
backup slides
CDF (and CMS)
CMS Tracker25,000 Silicon Strip CensorsTotal Area of 210 m2
9600000 readout channels
Crystal Electromagnetic CalorimeterSampling Hadronic Calorimeter with
copper Absorbing PlatesTrigger
Store ~100 events per second (out of 40 million +)
CDF Tracker405,504 silicon readout channelsOpen cell drift chamber (30,240 readout channels)
Lead/Scintillator Sampling Electromagnetic CalorimeterIron/Scintillator Sampling Hadronic CalorimeterTrigger
Three level system~8 s decision time at Level 1
Systematics
Data
• Fake background from jets and electrons.
• Electrons pass isolation cuts but fail electron veto (lower left corner)
• Jets fill plot but tend to fail the track and 0 isolation
Candidate events
• Two events survive all cuts.
• Jet 1 of Candidate 1 is tagged as a b quark jet.
Jet Multiplicity
We have background predictions in hand
As an a priori test:
predict rates in 0 and 1 jet multiplicity bins(no HT or Z Mass cut)
did not look at 2+ jet bin until satisfied
result more likely than 41% of pseudoexperiments
identification cuts
Our ID cuts are similar to other tau analysis
We have tighter a tighter calorimeter isolation cut
Our electron veto is bracketed by cuts in other analyses
Our W+jets background is reduced at the expense of reduced tau ID efficiency
1D version, fixing ET(1) cut…
Optimization (cont’d)
HT
HT
Signal/SQRT(Bkgnd)
Likelihood Ratio
Signal/BkgndHT
HT
Optimization (cont’d)
HT and lead jet ET cuts can distinguish signal from background
Integral distributions above cuts shown
HT
ET(1)
HT
Signal
Background
identification cuts
Tightening the calorimeter isolation cut is a concern because it is the worst modeled
Using the “tight” sample fromCDF 6010, thisis a 5%(relative) scalefactor effect
W analysis
CDF 6010 cut
Our cut
MC
Data
Acceptance
Pythia ttopei MCNcand includes jet fakesBefore scale factors:
eh : 50%
h : 42%
eh : 4%
h : 4%
Efficiencies and Scale Factors
Zvertex 0.948 +/- 0.003 6917
Ele trigger 0.966 +/- 0.001 6234
CMUP trigger
0.904 +/- 0.012 6293
CMX trigger 0.901 +/- 0.016 6293
Ele ID SF 0.965 +/- 0.006 6590
CMUP ID SF 0.94 +/- 0.01 6825
CMX ID SF 1.015 +/- 0.007 6825
value CDF Note
Systematics: Techniques
•Jet Energy Corrections:
•Level 5, half of difference between +1 and –1
•Monte Carlo Generator Dependence (half of difference):
• ttop2e (pythia) with no QED FSR
• weighted for BR ttopli (herwig)
•ISR: ttopei (ISR on) compared to ttop0e (ISR off)
•FSR: ttopei (tune A) compared to ttop5e (tune B)
•Statistical uncertainty dominates
Systematics
•PDFs: Compare # expected events in ttopei with:
• ttop3e (MRST PDFs)
• ttop4e (MRST PDFs, lower ISR)
• ttop6e (MRST PDFs, lower FSR)
•For ttop2e and ttop4e comparisons our systematics are limited by statistics
jet fakes
• Cross-checking samples yeilds a maximum difference of 26%, which we take as our systematic error
• jet50 is closest in Et to spectrum in data fakes so we use jet50 to determine our backgrounds
Candidate Event
• Run 167229
• Event 2376337
Candidate Event
Plot optimization variable in 2D vs cuts
Choose cuts at lower left corner of “mouth”
highest acceptance for same optimization
Optimization
HT
ET(1)
HT
Signal/SQRT(Bkgnd)
ET(1)
Likelihood Ratio
e fakes
Form an electron veto sample in the data with tau candidates that pass all ID cuts but fail electron veto
Apply 1.2% fake rate
Plotted on the right is our veto sample before Ht cut
Peak at zero gives us confidence that we have electrons in this sample
WW and WZ
WW and WZ are sources of real taus in our background
Cross Section * Branching ratio combined with low (relative to e and mu) tau reconstruction reduces BG
These two backgrounds combined are <15% total BG
WW: atop4x (Herwig + Alpgen +0 parton) atop5x (Herwig + Alpgen +1 parton)
WZ: atop0y (Herwig + Alpgen + 0 parton) Statistical error is 100%
Jet Multiplicity Study
Summary of all bins with probability of our data,and distribution of pseudoexperiment results
predicted
seen
0j, e, OS 23±3 18
1j, e OS 4.4±0.6 5
0j, , OS 23±3 14
1j, , OS 2.7±0.6 4
0j, e, SS 6.5±1.7 5
1j, e, SS 1.8±0.5 1
0j, , SS 4.3±1.3 3
1j, , SS 0.6±0.2 0
Jet Multiplicity Study
If you believe there are patterns of discrepancy, can study subsets
worst is muon 0j OS bin. 7% probable