Top pair resonance searches with the ATLAS detector

Report
Top pair resonance searches
with the ATLAS detector
钟家杭
University of Oxford
[email protected]
Frontier Physics Working Month
Outline

Background information

Top reconstruction

Top pair resonance searches

Boosted tops
1
[email protected]
31 Aug 2012
Top quark

Lifetime ~ 5x10-25 s


Decay before hadronization
Almost exclusively via t -> W + b
Spin=1/2, charge=2/3
 The heaviest known quark


2
m(t)=173.2±0.9 GeV (Tevatron)
[email protected]
hadron
67%
e
11%
μ
11%
τ
11%
31 Aug 2012
The energy frontier at TeV
3
[email protected]
31 Aug 2012
Beyond the Standard Model

Two benchmark BSM models used in experiments

Z’ in a leptophobic topcolor model
Proxy to narrow resonance: Γ/m=1.2%

Kaluza-Klein gluon (KKG) in Randall-Sundrum extra dimension models
KKG branching ratio
Proxy to broad resonance: Γ/m=15.3%
Phys. Rev. D 77 (2008) 015003

Generic search, applicable
to other BSM models



4
Spin-0 Lee-Wick Higgs
Spin-2 KK graviton
…
[email protected]
31 Aug 2012
The ATLAS detector
5
[email protected]
31 Aug 2012
Leptons in ATLAS

Only prompt leptons are considered signal



Fixed-cone isolation to suppress QCD contribution



Electron:
Energy cluster of high EM fraction, matching to a track
Muons:
Combined tracking in both Inner Tracker and Muon Chambers
Mostly real leptons from heavy-flavor quark
Both calo-based and track-based
Hadronic tau channel not included
6
[email protected]
31 Aug 2012
Jets in ATLAS
Sequential clustering algorithms : Kt, C/A, anti-Kt
 AntiKt as the mainstream jet algorithm





R=0.4 as the standard jet
R=1.0 known as the fat jet (boosted hadronic top jet)
C/A algorithm with R=1.5 used for HEPTopTagger
B-tagging



7
For antiKt4 jets
Using tracks associated with the jet

Secondary vertices

Impact parameter
Multivariate algorithms, 70% efficiency
[email protected]
31 Aug 2012
Leptonic top reconstruction

t -> W + b -> l+v+b





One Lepton
High missing transverse energy (MET)
High transverse mass MT between lepton and MET (due to W mass)
mT  2 pTl ETmiss (1  cos  )
One b-tagged antiKt4 jet.
Neutrino reconstruction


8
Assuming MET fully from neutrino, solve pz(v) using W-mass
Under-constrained in di-lepton channel
[email protected]
31 Aug 2012
Hadronic top reconstruction

t -> W + b -> q+q+b

Resolved:


3 antiKt4 jets
2 antiKt4 jets, if one has high mass.
Boost

Boosted:



One energetic antiKt10 jet
with substructure cuts
One energetic C/A1.5 jet
using HEPTopTagger
Discrimination against QCD
R ~ m / p
9
[email protected]
31 Aug 2012
Hadronic top reconstruction

Jet substructure

Jet mass> 100 GeV


2
2
2 = (  ) −(  )
First splitting scale 12 >40 GeV

Re-clustering jet constitutes with Kt algorithm.
The splitting scale of the last step.  =min(pTi, PTj) x ΔRij
mt
10
mt/2
[email protected]
31 Aug 2012
Top pair resonance search

Select ttbar-like events






Di-lepton
1 lepton + 4(3) jets (resolved)
1 lepton + 1 jet + 1 fat jet (boosted)
Fully hadronic (HEPTopTagger)
2 fb-1, EPJC72 (2012) 2083
2 fb-1, arXiv:1207.2409
5 fb-1, ATLAS-CONF-2012-102
Reconstruct  or equivalent
Look for peaks in  spectrum
11
τ (had)
14%
Di-lepton
6%
[email protected]
Fully
hadronic
46%
1-lepton
(e, µ)
34%
31 Aug 2012
Single Lepton Boosted ttbar
Single lepton trigger
 Exactly one offline lepton






ETmiss>35GeV, MT>25GeV
 Solve neutrino pz with W
mass constraint
 Closest antiKt4 jet as from
the leptonic top

One antiKt10 fat jet

Electron pT > 25 GeV
Muon pT > 20 GeV



pT > 250 GeV
m > 100 GeV
12 > 40 GeV
dR(akt4, akt10)>1.5
pT > 30 GeV
0.4 < ΔR(lepton, jet) <1.5
Signal selection efficiency
12
[email protected]
31 Aug 2012
Single Lepton Boosted ttbar
M=2.5 TeV
13
[email protected]
31 Aug 2012
Single Lepton Boosted ttbar

tt= l + v + akt4 + akt10 (4-vector sum)
Leptonic top mass
(l + v + akt4)
Hadronic top mass
(fat jet)
14
[email protected]
31 Aug 2012
Single Lepton Boosted ttbar

W+jets background


Data-driven normalization
Multijets

Fully data-driven
Can be further
improved by b-tagging
15
[email protected]
31 Aug 2012
Single Lepton Boosted ttbar
16
[email protected]
31 Aug 2012
Single Lepton Boosted ttbar

Search for local data excess
with BumpHunter

Set 95% CL upper limits on xsec
Replace the theoretical line with your favorite model
17
[email protected]
31 Aug 2012
Top pair resonance search
Di-lepton
One-lepton
(Resolved)
One-lepton
(Boosted)
Fully hadronic
Integrated
luminosity
2 fb-1
2 fb-1
2 fb-1
4.7 fb-1
Z’ limits
-
0.5 – 0.88 TeV
0.6 – 1.15 TeV
0.7 – 1.3 TeV
KKG limits
0.5 – 1.08 TeV
0.5 – 1.13 TeV
0.6 – 1.5 TeV
0.7 – 1.5 TeV
More results are coming…
18
[email protected]
31 Aug 2012
Boosted Top

New challenge: TeV frontier

Top decay products are more collimated
ΔR ~ m/P
19
[email protected]
31 Aug 2012
Boosted Top: Leptonic

Lepton collinear with the b-quark

Signal acceptance suffers from the fixed-cone isolation cuts
Signal selection efficiency
20
[email protected]
31 Aug 2012
Boosted Top: Leptonic

Mini-isolation


JHEP 1103:059 (2011)
Variable-cone size ΔR=KT/pT

Parameter KT, e.g. 15 GeV

Lepton pT (easier than top pT)
Sum up tracks pt within the cone

Sufficient angular resolution
Fixed-cone
isolation
b-jet
Isolation cut
Boost, dR=mtop/Etop
Mini-isolation
21
[email protected]
31 Aug 2012
lepton

Three jets tend to overlap.

Use single jet with large radius

Need rejection against QCD
=> Substructure variable

Need to get rid of soft component from
underlying event and pileup
=> Jet Grooming

Not limited to top decay
22
[email protected]
Boost
Boosted Top: Hadronic
31 Aug 2012
Boosted Top: Jet grooming

Algorithms to reduce soft components from UE and PU


I.
II.
III.
23
Jet kinematics more close to the constituents of hard scattering
Better resolution/discrimination of the substructure variables
Mass drop/filtering
Trimming
Pruning
[email protected]
31 Aug 2012
Boosted Top: Jet grooming
Phys.Rev.Lett.100:242001 (2008)
Mass drop/filtering
(J. Butterworth, A. Davidson, M. Rubin, G. Salam)
 Works on C/A jet
 More optimized for two-body hadronic decay

W/Z -> qq, H -> bb
Mass drop
Filtering
24
[email protected]
31 Aug 2012
Boosted Top: Jet grooming
JHEP 1002:084 (2010)
Trimming
(D. Krohn, J. Thaler, L. Wang)
 Use jet constituents to build Kt subjets (e.g. R=0.2)
 Remove soft subjets
 Applicable to any jet, any physics scenario
25
[email protected]
31 Aug 2012
Boosted Top: Jet grooming
arXiv:0912.0033 (2009)
Pruning
(S. Ellis, C. Vermilion, J. Walsh)
 Recluster jet constituents with C/A or Kt algorithm
(no need of subjets)
 Veto wide angle and soft constituents during jet formation
26
[email protected]
31 Aug 2012
Boosted Top: Jet grooming

Reduce unnecessary catchment area
antiKt R=1.0 (ungroomed)
27
antiKt R=1.0 (trimmed)
[email protected]
31 Aug 2012
Boosted Top: Substructure

Jet mass are more discriminating after trimming
28
[email protected]
31 Aug 2012
Boosted Top: Substructure

Splitting scale

Re-clustering jet constitutes with Kt algorithm.
The splitting scale of the last step.  =min(pTi, PTj) x ΔRij
12 ≈  /2
29
23 ≈  /2
[email protected]
31 Aug 2012
Boosted Top: Substructure

N-subjettiness (τN)

Re-clustering with Kt algorithm until exactly N subjets are formed


30
Smaller τN+1 /τN => Structure described better with additional sujet
[email protected]
31 Aug 2012
Boosted Top: HEPTopTagger




A multi-step algorithm starting from a large-R
C/A jet
Grooming: filter out soft component
Form up subjets
Impose Top and W mass constraints
JHEP 1010:078 (2010)
ATLAS-CONF-2012-065
31
[email protected]
31 Aug 2012
Summary

ttbar resonance are searched in all channels at ATLAS


Systematics still have large impact on the sensitivity




Unfortunately, we don’t have the luck yet…
Uncertainty of performance at high pt
Understanding realistic performance of new techniques
Rooms to improve…
New techniques for new challenges


32
Boosted top/object
Increased luminosity
[email protected]
31 Aug 2012

similar documents