Randall-Sundrum KK Gluon & Energetic Tops at the LHC Joseph Virzi, LBL K. Agashe, A. Belyaev, T. Krupovnickas, G. Perez and JV / hep-ph/612015 Work in Progress with K.Agashe, T.Han, G.Perez
Feb 07, 2016
Randall-Sundrum KK Gluon & Energetic Tops at the LHC
Joseph Virzi, LBL
K. Agashe, A. Belyaev, T. Krupovnickas, G. Perez and JV / hep-ph/612015Work in Progress with K.Agashe, T.Han, G.Perez
May 2007 Joseph Virzi UC Berkeley 2
Outline• Brief Introduction to Randall-Sundrum (RS) Model
• Focus on detection of KKG using top quark pair production
• Top reconstruction @ high PT– discuss associated challenges – propose approaches to address these challenges
• Polarization asymmetry measurement
• Background Analysis & Discussion
• Conclusions
May 2007 Joseph Virzi UC Berkeley 3
Randall-Sundrum Modelwith SM fields propagating in the bulk
• solves the hierarchy problem for
ckrWM ke
22 2 2
& 0
ckrc
PL
ds e dx dx r d
k M
10ckr
• motivation for model is hierarchy problem – vast difference between the weak and Planck scales
4D metric
May 2007 Joseph Virzi UC Berkeley 4
• SM phenomenology constrains profiles in 5th dimension• Yukawa couplings are given by overlap with Higgs on
TeV brane
Warp Factor
ckr
May 2007 Joseph Virzi UC Berkeley 5
Particle Profiles
• Couplings to tops are enhanced and parity violating
• Dominant coupling to tR because of pheno’ constraints
LIGHT HEAVY
ckr
May 2007 Joseph Virzi UC Berkeley 6
Analysis
• The formalism of the RS1 model leads to KK excitations
• We consider here the first excitation of the gluon, G(1)
– Experimental constraints favor masses of G(1) > 2TeV– Case study: 3 TeV KK gluon– Will use 100 fb-1 of data (3 years at high luminosity @
LHC)
May 2007 Joseph Virzi UC Berkeley 7
RS1 KK Gluon
• Prefers decay into heavier quarks, especially to tops.– BR > 0.95
• Heavy quark couplings to G(1) are enhanced relative to the SM. – For tR ~5– For tL & bL ~1.
• Light quarks & bR couplings are suppressed by factor ~5.• SM gluon couplings vanish due to orthogonality conditions
Branching Ratio of KKG vs MKKG
May 2007 Joseph Virzi UC Berkeley 8
Feynman Diagrams• Relevant Tree Level Diagrams for our discussion• The gg→KKG vertex does not exist because of
orthogonality arguments• Primary production mechanism for top quark pairs
+
May 2007 9
Signatures of KK Gluon• The RS1 KK Gluon provides a resonance structure
– Width ~0.2 MKKG ( 600 GeV )
• total cross section 850 pb
• ΔσRS = O(100 fb)
σ vs Invariant Mass
May 2007 10
Signatures of KK Gluon (cont’d)
• The excess production will have more tR than tL
• Strategy– G(1) contribution to PLR is large & opposite sign than SM– Correlate large L/R polarization asymmetry to the mass
peak
L/R Polarization vs Invariant Mass
RS prediction
SM prediction
May 2007 Joseph Virzi UC Berkeley 11
L/R Polarization Asymmetry Introduction to PLR
• Look at the direction of the lepton in the top quark rest frame
N+ & N- are the number of events where the lepton is forward (cos(θ) > 0.0) and where the lepton is backward, respectively in the top rest frame
θ
May 2007 Joseph Virzi UC Berkeley 12
• Dileptonic channel → 2 neutrinos– Difficulty resolving neutrino– 10% BR
• Fully hadronic decay– Background more difficult– 60% BR
semileptonic
hadronic
leptonic
• Semileptonic (ttbar→bbjjℓν) channel most promising for this analysis.– BR(ttbar →{μ,e}) = 30%
Production & Decaytt
May 2007 Joseph Virzi UC Berkeley 13
Monte Carlo Simulation Strategy• Used a customized version of the Sherpa MC
– Full spin correlations in top decays
• 100 fb-1 of signal ( SM/RS ) with MKKG = 3 TeV– Invariant Mass > 1 TeV– σ(M>1TeV) x 0.3 semileptonic BR = 8.8 pb
• 100 fb-1 of W+jets sample– Invariant Mass > 1 TeV & PT > 300 GeV
– σ (M>1TeV) = 6.5 pb
• 100 fb-1 of single top production sample– Invariant Mass > 1 TeV & PT > 50 GeV
– σ (M>1TeV) = 5 pb
May 2007 Joseph Virzi UC Berkeley 14
Signal Reconstruction Overview• Conventional methods of top reconstruction at
the LHC involve reconstruction of whole top decay chain– beats down background– Requires ≥4 jets, of which ≥2 are b-jets
• The approach breaks down at energies ~ TeV– Jets collimate. We will discuss later
• We overhauled the methods to address deficiencies
May 2007 Joseph Virzi UC Berkeley 15
Conventional Signal Reconstruction
• Reconstruction of top pairs– ≥4 jets, 2 are b-tagged– Isolated lepton - ΔR– Missing energy → neutrino– Top mass (174 GeV ) is an
input– 1 b-jet + W reconstructs
leptonic top– 2 light jets reconstruct
hadronic side W– Other b-jet + W reconstructs
hadronic top
May 2007 Joseph Virzi UC Berkeley 16
Problem with Conventional Method
• As the invariant mass of the ttbar event ↑ the jet multiplicity ↓
• Conventional approach works well here
• Reconstruction efficiency is adversely affected @ high invariant mass– Very few 4 jet eventsNumber of Jets
Nu
mb
er
of E
ven
ts
May 2007 Joseph Virzi UC Berkeley 17
TopJet Reconstruction
• Hadronic side – giving up– Use the events where the decay
products of the top are observed as a single jet
– Impose a top-jet hypothesis on the hadronic side jet
– remove b-tagging constraint on hadronic side
– Stiff ( >600 GeV ) PT cut on the leptonic side top decimates background
• Modify leptonic top reconstruction– Lepton isolation difficult (next)
May 2007 Joseph Virzi UC Berkeley 18
Removing B Decay Leptons - MBL
• MBL – the invariant mass between b-jet and lepton– B decay leptons have MBL ~ 5 GeV
– Signal leptons have MBL ~ 50 GeV
• 20% of b-jets contain leptons• descriminate against B decay leptons• Keep leptons from t → bW →bℓν
May 2007 Joseph Virzi UC Berkeley 19
Invariant Mass Plots
• TopJet approach is vastly more statistically significant over the mass window
• The conventional method is more appropriate for lower energies
• Shape of the background
Where’s the peak?
TopJet Method
Conventional Method
May 2007 Joseph Virzi UC Berkeley 20
Efficiency Plot
• Huge increase in reconstruction efficiency• The efficiency & mass curves are shaped by the physics• The mass curve is not shaped by the efficiency curve
Reconstruction Efficiency vs Invariant Mass
May 2007 21
• Boost profile for com is central for large invariant mass
• Primary production is through qqbar
Motivates stiff PT cut
Kinematicstt
May 2007 Joseph Virzi UC Berkeley 22
JETS
May 2007 Joseph Virzi UC Berkeley 23
Jet PT over mass peak
• Distributions are normalized to unit area
• <PT> of b-jets = 555 GeV
• 50% of b-jets have PT > 300 GeV
May 2007 Joseph Virzi UC Berkeley 24
B-tagging @ high PT
• Important Issue but still relatively uncertain– Best estimates at low energies place ε = b-tag efficiency = 60%
• Best estimates are approximately 20% at upper end of the PT spectrum– March, Ros, Salvachua ATL-PHYS-PUB-2006-002
• Remain conservative & use 20% throughout
• Conventional reconstruction methods depend on 2 b-tags. Quadratic dependence on ε
• New approach described here only requires 1 b-tag. Linear dependence on ε
May 2007 Joseph Virzi UC Berkeley 25
Light Jet Rejection
• Ensuring that we do not label jets from lighter partons as b-jets– especially important for W+jets background
• Current estimates– March, Ros & Salvachua ATL-PHYS-PUB-2006-002
– Rc = 30. Ru = 130
• This analysis is performed with a uniform rejection ratio Rq=30
May 2007 Joseph Virzi UC Berkeley 26
L/R Polarization Asymmetry
Challenges• Jet Energy Corrections
– Jet Energy ≠ Parton Energy
– Vital to reconstructing quark cm frame for PLR
– Adds uncertainty to reconstruction of cms kinematics.
Taken from ATL-SOFT-2003-010
Jet Energy Scale for b & light jets
May 2007 Joseph Virzi UC Berkeley 27
L/R Polarization Asymmetry Cont’dLepton PT Distribution
• The L/R polarization asymmetry will manifest itself in the lepton <PT> (A.T.Holloway)
Lepton PT vs Invariant Mass
May 2007 Joseph Virzi UC Berkeley 28
Background Analysis
May 2007 Joseph Virzi UC Berkeley 29
Efficiency of CutsOn Signal & Background
• RED survives all cuts
Signal (RS+SM)
W+JETSSINGLE TOP
May 2007 30
Results of Top Jet Approach
• The peak becomes much more statistically significant
• We correlate the mass peak to the PLR
• Additionally, we can observe the <PT> of the lepton
May 2007 Joseph Virzi UC Berkeley 31
LHC Reach
• Our reconstruction efficiency remains relatively flat to 4 TeV
• Current estimates place the reach of the LHC for our signal to 4 TeV
May 2007 Joseph Virzi UC Berkeley 32
ConclusionsConclusions• With new reconstruction technique, the signature(s) of the RS KK gluon
becomes much more statistically significant– Combination of Topjet and Conventional techniques spans low to high MTT
– The efficiency of reconstruction increases by O(5)– And turns out to stay relatively flat for increasing invariant mass ~4TeV
• The W+jets and single top background is small
• 100 fb-1 of data is a long time.– Depending on the mass of the KK gluon, efficiencies and fake rates, maybe
we can get by with less data– Need to leave some wiggle room ( PDF & other uncertainties )
• Preliminary analysis using more realistic reconstruction techniques shows consistency with the results herein
May 2007 Joseph Virzi UC Berkeley 33
Backup Slides
May 2007 Joseph Virzi UC Berkeley 34
Summary of CutsCUT VALUE
Leptonic Top Mass 174 GeV ± 50 GeV
Hadronic Top Mass 174 GeV ± 50 GeV
Hadronic W Mass 81 GeV ± 50 GeV
Missing ET > 30 GeV
Lepton Isolation ∆R > 0.4 or MBL>40 GeV
Lepton PT > 10 GeV
Lepton η < 2.5
Jet PT > 20 GeV
Jet η < 4.5
Leptonic Top PT > 600 GeV
May 2007 Joseph Virzi UC Berkeley 35
Jet PT Distributionsfrom signal sample
• B-jet spectrum is harder than for light jets
May 2007 36
Single TopBackground
• Sample used is single top production– Representing 100 fb-1
– MCMS > 1 TeV
– PT > 50 GeV
– 5 pb cross section
– PT cut yields high background rejection
– 97% light jet rejection
– t-channel production is dominant
Evolution of cuts for single top production
green is conventional mode
May 2007 37
PT of leptonic top after cuts
May 2007 38
W+JETS background
• Sample is W+jets– representing 100 fb-1
– MCMS > 1.5 TeV– PT > 300 GeV– cross section 6.5 pb– Light jet rejection → 97%
Evolution of cuts for W+jets background
May 2007 39
Efficiencies of Cuts
• Conventional Reconstruction Method
• TopJet Reconstruction Method– Stiff PT cut provides the coup-de-grace
(discuss later)– Has high signal efficiency
RED are events passing all cutsBoth plots are drawn to same scale
GREEN is conventional reconstruction
May 2007 Joseph Virzi UC Berkeley 40
Conventional Reconstruction
RS1 W+Jets Single Top
# of events 880000 650000 478700
# passing preselection 21314 50 36701
# passing lepton cuts 17640 40 31427
# pass leptonic top mass cuts
9483 19 20171
# pass leptonic top PT cut
48 0 137
May 2007 Joseph Virzi UC Berkeley 41
TopJet Cut Statistics
RS1 W+Jets Single Top
# of events 880000 650000 478700
# pass preselection 59081 147 116471
# pass lepton cuts 46348 120 89768
# pass leptonic top mass cuts
18862 22 68097
# pass leptonic top PT cut
2033 0 190
May 2007 Joseph Virzi UC Berkeley 42
Efficiencies of Reconstructionusing Different Modes
Leptonic topPT>600 GeV
May 2007 Joseph Virzi UC Berkeley 43
Efficiencies of Reconstructionusing Different Modes
Leptonic topPT>400 GeV
May 2007 Joseph Virzi UC Berkeley 44
W+JETS background
• I focus here on background most likely to do damage– Invariant mass > 1.5 TeV– PT > 300 GeV– Cross section 6.5 pb
• The background plot looks at all combinations of 2, 3 and 4 jets which pass the indicated cuts on the leptonic side.– Superset of actual background– No b-tagging / light jet rejection
assumptions
Evolution of cuts for W+jets background
May 2007 Joseph Virzi UC Berkeley 45
Single TopBackground
• Sample used is single top production– MCMS > 1.5 TeV.
– PT > 100 GeV
– 5 pb cross section
• The background plot looks at all combinations of 2, 3 and 4 jets which pass the indicated cuts on the leptonic side.– Superset of actual background
– No b-tagging / light jet rejection assumptions
Evolution of cuts for single top production
May 2007 Joseph Virzi UC Berkeley 46
Spectrum of Hadronic SideReconstruction Modes
• 2 light jet + 1 b jet events– b → semileptonic top– 2 light jets summed
• 1 light jet + 2 b jet events– b → semileptonic top– hadronic top = b + j
• 3 light jets + 1 b jet events– b → semileptonic top– hadronic top = j + j + j
• 5+ jet events
• In all cases, the jets on the hadronic side are summed to the top
• Reconstruction modes are separated for different jet multiplicities– The final reconstruction
depends weakly on jet reconstruction algorithm
– Allows for weighing contribution from each mode
May 2007 Joseph Virzi UC Berkeley 47
single tops-channel
PT(leptonic top)
May 2007 Joseph Virzi UC Berkeley 48
May 2007 Joseph Virzi UC Berkeley 49
May 2007 Joseph Virzi UC Berkeley 50
Single top t-channelTruth Level Analysis
May 2007 Joseph Virzi UC Berkeley 51
Single top t-channel
May 2007 Joseph Virzi UC Berkeley 52
May 2007 Joseph Virzi UC Berkeley 53
May 2007 Joseph Virzi UC Berkeley 54
May 2007 Joseph Virzi UC Berkeley 55
May 2007 Joseph Virzi UC Berkeley 56
• jlkajsdkθ
22 2 2sin2BL B L TM p p M
θ
May 2007 Joseph Virzi UC Berkeley 57
W+Jets Jet PT Distribution