Post on 30-Sep-2020
A Top Quark Mass in The Dilepton Channel
Marco Trovato,
SNS & INFN Pisa
CDF Italia meeting
Marco Trovato SNS & INFN Pisa 22 September 2009 – CDF Italia
Top mass measurement: Motivation
Enormous mass Same scale as EWSB Yukawa coupling of ~ 1
Within the SM By generating large radiative corrections
helps constraining the Higgs mass
Beyond the SM discrepancies in the results across different decay
channels could provide hints of new physics
∝ log(mH)∝ mtop2
W
W W
WH0
b
t
Marco Trovato SNS & INFN Pisa 32 September 2009 – CDF Italia
Top Production and Signature
l+jets 30% → 2 bjets, 2 qjets → high PT e/ μ → large MET
all jets 44%→ 2 bjets→ 4 qjets
dilepton 5% → 2 bjets → 2 high PT e/ μ → large missing transverse energy (MET)
Produced in pairs mainly via strong interactions ~ 6.7 pb (Mtop =175 GeV)→
Decaying in Wb ~ 100 % of the times 3 possible signatures depending on W’s products→
q
q
q
q
q
q
e/μ
e/μ
e/μ
ν
νν
Marco Trovato SNS & INFN Pisa 42 September 2009 – CDF Italia
State of the art @ Tevatron
GeV .)( 1.1 .)(6.0 1.173 syststatM top ±±=
Tevatron world average (March ’09)
%)75.0/(
.)( 1.1.)( 6.01.173
=
±±=
toptop
top
MM
GeVsyststatM
δ
arXiv:0903.2503
Marco Trovato SNS & INFN Pisa 52 September 2009 – CDF Italia
Dilepton channel
Features → Very clean sample (S:B ~ 3:1) → Less affected from JES systematic (2 jets)
→ Low statistics → Unconstrained kinematics (2 neutrinos)
Importance• As uncorrelated sample:
→ contribute to improve overall resolution on Mtop
• Due its unique bg: → helps including/excluding other nonSM signal processes (stop) → sensitive to nonSM background processes
Marco Trovato SNS & INFN Pisa 62 September 2009 – CDF Italia
A particular top mass measurement
Documentation for this work: 340/pb: PRD 73 112006, CDF note 7641 2.9/fb: CDF notes 9048, 9433, 9505, arXiv:0901.3773
Authors: I. Suslov, M. Trovato, G. Velev, V. Glagolev, Oleg Pukhov, G. Bellettini, G. Chlachidze, J. Budagov, Alexei Sissakian
Marco Trovato SNS & INFN Pisa 72 September 2009 – CDF Italia
Selection of the events: LTRK algorithm
A well identified isolated lepton (“tl”) Cluster ET > 20 GeV
PT >20 GeV
Two or more jetsET
@L5 > 20 GeV
A well isolated track lepton (trkl)PT > 20 GeV
Missing energy
GeV 25>/ TE
General cuts
• MET < 40 GeV
• Cosmic• Same sign dilepton ltrk • Δφ(tl, MET) < 5°;> 175°• Δφ(trkl, MET) < 5°• Δφ(MET, jet) < 25°
(if MET<50GeV)
Vetoeselectronm
uon
)2/ 1062GeV/ 76 if( cGeVtrkltlMc <−<
Marco Trovato SNS & INFN Pisa 82 September 2009 – CDF Italia
Why LTRK?
Compared to the other competive selection (DIL)
1) Statistics is increased at the price of higher (but
well modeled) background rates
6) We investigate a sample otherwise unexplored
PRD 73 112006
Percentaged normalized to LTRK U DIL
Marco Trovato SNS & INFN Pisa 92 September 2009 – CDF Italia
)~( tt MΓ
The method: event by event
22222
22
)( MMmM
BWinv Γ+−
Γ=
222constrreso χχχ +=
∑∑∑=
−+℘−+−
=
yxi UE
iUEiUE
l
lTPj
TPtf
l P
lTPl
TPreso
ilT ,
2
2)(2
))~|(ln(2
2
2
2)~(2
σσχ
))~,~|(ln(2))~,
~|(ln(2
)),|(ln(2)),|(ln(22
222111
2211
tt
WW
MtMjlinvmBWMtMjl
invmBW
MWMlinvmBWMWMl
invmBWconstr
Γ−Γ−
Γ−Γ−=
νν
ννχ
Kinematical event reconstruction: → for each (φ1,φ2) we minimize χ2 with respect to
Constrain the kinematics – PHI method: Assume φ1, φ2 of the neutrinos as known
→ optimal 144 point grid created in (0,π)x(0,π)
jetTPjet
TPpartTP /)( −
Best event top mass (Mtreco):
→ a weighted average of the 144 masses
tM~
Marco Trovato SNS & INFN Pisa 102 September 2009 – CDF Italia
The method: signal templates
Signal templates are built from events generated within a top mass range 155195 GeV/c2
Templates are parametrized with analytic functions (fs) depending on Mt
A signal mass template defines the probability density function to renconstruct Mtreco from
a ttbar event generated with Mt
Marco Trovato SNS & INFN Pisa 112 September 2009 – CDF Italia
Bg contamination
LTRK sample is affected by three main bg processes: DY, Diboson, W+jets (“fakes”)
→ DY & Diboson are simulated and weighted according to their crosssection → Fakes are obtained from data
Reconstructed events are normalized to the expected rates
145.0 ± 17.3Total background
80.2 ± 15.7Fakes
49.6 ± 7.2DrellYan
15.2 ± 1.0Diboson
162.6 ± 5.1Signal (ttbar)
Expected rateProcess) /9.2 ( fbLdt =∫
Bg control region
Signal region
Marco Trovato SNS & INFN Pisa 122 September 2009 – CDF Italia
The method: bg templates
A bg mass template defines the probability density function to renconstruct Mtreco from a
nonttbar event
Templates are parametrized with analytic functions (fb) Mt undependent
Marco Trovato SNS & INFN Pisa 132 September 2009 – CDF Italia
The method: Likelihood
• The likelihood expresses the probability that a distribution be obtained from a sample of N events composed of ns signal events and nb background events
Fit parameters
Mtop corresponds to –ln[L]min
Expected background rate and its uncertainty
recotM
Template parametrizations and covariance matrices
Marco Trovato SNS & INFN Pisa 142 September 2009 – CDF Italia
Calibrations
ationunderestimerror for correct to009.1*)()()(
bias for thecorrect to/ 13.0 2
toptoptop
toptoptop
MMM
cGeVMMM
σσσ =→
−=→
Marco Trovato SNS & INFN Pisa 152 September 2009 – CDF Italia
Systematic uncertainty
JES is ~94% of the overall systematic uncertainty
Marco Trovato SNS & INFN Pisa 162 September 2009 – CDF Italia
Results in Data: Constrained Fit
fbLdt /9.2=∫3.170.145exp ±=→ bgN
Fit result
235.322.3 / (stat.)35.165 cGeVM top ±=
1.150.151.146 ±=bgN
18% PE’s with larger error
Marco Trovato SNS & INFN Pisa 172 September 2009 – CDF Italia
Previous Result and Improvement
With respect to the previous published result
we note:
A ~ 20% improvement in mass sensitivity due to the introduction in the χ2 of a BW instead of a Gaussian distribution
Error now
Error before
2006 73 PRD ,)/ )(1.4 )(7.169( 29.80.9 cGeVsyststatM top ±= +
−
1340 −≈∫ pbdtL
Marco Trovato SNS & INFN Pisa 182 September 2009 – CDF Italia
Conclusions
First dilepton mass measurement with ~ 3/fb data → first to investigate the LTRK sample at high luminosities (50% overlap with DIL sample) → first to use relativistic BreitWigner distribution function and topmass dependent top
width in the kinematical event reconstruction
Plans at nex step: Analysing separately btagged and untagged events Exploiting event by event probability Increasing statistics
24.33.3 / 3.1(Syst) )( 165.5 cGeVstatM top +±=
Marco Trovato SNS & INFN Pisa 192 September 2009 – CDF Italia
BACKUP
Marco Trovato SNS & INFN Pisa 202 September 2009 – CDF Italia
Based on mtdependence of σtt
: Orthogonal measurement
l+jets, dilepton samples → τ included
→ ≥3 jets (l+jets), ≥2 jets (dilepton)→ 1 btag
pb .).( 18.8 GeV) 170( 98.087.0 syststatM top
measuredtt +±==σ
Depending on the input from theory we obtain
QCD) (NLO GeV 5.1 65 1.69.5±=topM
QCD) NLL(NLO GeV 5.167 8.56.5 +±=topM
QCD) NNLO te(approxima GeV 1.169 9.54.5±=topM
Less accurate than direct easurements Does not heavily rely on (LO) simulated
ttbar → less sensitive to difference between
MC and pole masses