A search for diboson resonances at ATLAS using
boson-tagged jets
Using the jet substructure thresher on the QCD haystack
Alex Martyniuk, UCL
October 15, 2015
1 / 48
Alex Martyniuk
X → VV → JJ
Introduction
Resonance searches are the classic
methodology to search for new
particles and their excitations
In essence they boil down to,
‘Look for a peak on a smooth
background’
Used in searches ranging from
quarkonia to the Higgs
2 / 48
Alex Martyniuk
X → VV → JJ
Introduction
Resonance searches are the classic
methodology to search for new
particles and their excitations
In essence they boil down to,
‘Look for a peak on a smooth
background’
Used in searches ranging from
quarkonia to the Higgs
Rediscovering the SM
2 / 48
Alex Martyniuk
X → VV → JJ
Resonance searches are the classic
methodology to search for new
particles and their excitations
In essence they boil down to,
‘Look for a peak on a smooth
background’
Events / 20 MeV
Introduction
20
18
ATLAS
16
fb
∫ Ldts == 4.9
7 TeV
14
12
10
QB ππ = 288 ± 5 MeV
c
σBcππ = 18 ± 4 MeV
NBcππ = 22 ± 6
-1
Data
Wrong-charge
combinations
8
6
4
Used in searches ranging from
quarkonia to the Higgs
2
0
0
Alex Martyniuk
200
300
400
500
600
700
m(Bcππ)-m(Bc)-2m(π) [MeV]
Rediscovering the SM
New meson states
2 / 48
100
X → VV → JJ
∑ weights / GeV
Introduction
∫ L dt = 4.5 fb , s = 7 TeV
-1
∫ L dt = 20.3 fb , s = 8 TeV
180
-1
160
ATLAS
Data
S/B weighted sum
Signal+background
Signal strength categories
140
Background
120
Used in searches ranging from
quarkonia to the Higgs
Rediscovering the SM
New meson states
New bosons!
2 / 48
60
20
40
18
ATLAS
16
fb
∫ Ldts == 4.9
7 TeV
14
12
10
8
QB ππ = 288 ± 5 MeV
c
σBcππ = 18 ± 4 MeV
NBcππ = 22 ± 6
20
-1
∑ weights - fitted bkg
In essence they boil down to,
‘Look for a peak on a smooth
background’
80
Events / 20 MeV
Resonance searches are the classic
methodology to search for new
particles and their excitations
Signal
m H = 125.4 GeV
100
0
10
5
Data
Wrong-charge
0
combinations
-5
110
120
130
140
150
6
2
0
0
100
200
300
400
500
600
700
m(Bcππ)-m(Bc)-2m(π) [MeV]
Alex Martyniuk
X → VV → JJ
160
m γ γ [GeV]
4
Introduction
Searches for exotic models at ATLAS use resonances to probe the very highest
mass ranges, in very early data
Examples from Run-1/2 are the ATLAS dijet resonance searches,
[arXiv:1407.1376] and ATLAS-CONF-2015-042
10
9
10
8
ATLAS
∫
6
10
5
3
10
[data-fit]/fit
W’
Observed 95% CL upper limit
2
10
Expected 95% CL upper limit
10
Data
Fit
q*, m = 0.6 TeV
q*, m = 2.0 TeV
q*, m = 3.5 TeV
102
Signif.
103
68% and 95% bands
104
10
-1
s=8 TeV, L dt=20.3 fb
107
10
σ × A [pb]
Prescale-weighted events
Uses pairs of high pT anti-kt 0.6 jets
Searches for narrow mass resonances, alongside broader signals
Data driven background model
∫L dt = 20.3 fb
-1
1
1
s=8 TeV
10-1
1
0
-1
10-2
2
0
-2
ATLAS
1
0.3
0.4 0.5
1
2
3
4
5
Alex Martyniuk
3
mW ’ [TeV]
Reconstructed mjj [TeV]
3 / 48
2
X → VV → JJ
Introduction
Searches for exotic models at ATLAS use resonances to probe the very highest
mass ranges, in very early data
Examples from Run-1/2 are the ATLAS dijet resonance searches,
[arXiv:1407.1376] and ATLAS-CONF-2015-042
Uses pairs of high pT anti-kt 0.6 jets
Searches for narrow mass resonances, alongside broader signals
Data driven background model
Events
104
ATLAS Preliminary
s=13 TeV, 80 pb-1
Data
Background fit
BumpHunter interval
BlackMax, m = 4.0 TeV
BlackMax, m = 5.0 TeV
3
10
102
10
1
Signif.
3
2
1
0
−1
−2
−3
3 / 48
p -value = 0.79
Fit Range: 1.1 - 5.3 TeV
|y*| < 0.6
2
3
2
3
4
5
6 7
Reconstructed mjj [TeV]
4
5
6 7
mjj [TeV]
Alex Martyniuk
X → VV → JJ
Introduction
Today I will present a complementary analysis to the dijet, a search for narrow
diboson resonances using jet-substructure performed on the full 8 TeV
dataset from ATLAS.
Which can be found here [arXiv:1506.00962], for those with no patience
Diboson resonances appear in many
extensions to the standard model
The following analysis concentrates on
two benchmark models
Extended gauge sector models
(W 0 → WZ )
Extra dimensions models
(GRS → WW /ZZ )
Low branching ratios hinder the
leptonic searches at the highest masses
Obviously, a fully hadronic search has
access to these lost events
The problem, is controlling the
enormous QCD background that the
leptonic searches were avoiding
4 / 48
Alex Martyniuk
W ⇓
Diboson branching ratios
qq (67%)
23%
47%
7%
13%
3%
7%
Z ⇒
qq (70%)
νν (20%)
ll (10%)
lν (33%)
X → VV → JJ
Boosted Bosons
Vector bosons have mass
O(0.1 TeV)
Low Mass <500GeV Large ΔR Small ΔR 5 / 48
Z
Z
WI W
Large ΔR We are interested in particles
of mass ≥ O(1 TeV)
Increasing Mass Therefore the decays of the
form, X → VV with large mX ,
lead to vector bosons with
very high pT
High Mass >1TeV Therefore, boosted decay
products become more
collimated
WI W
Alex Martyniuk
Small ΔR X → VV → JJ
Rule of thumb for angular
separation of decay products:
p
∆R = ∆φ2 + ∆η 2 ≈ 2m
pT
Boosted Bosons
Can roughly separate hadronic boson
decays into two regimes
1
Resolved
Lower momentum W , pT < 160 GeV
W decay resolved in two distinct anti-kt 0.4
jets
2
Boosted
Higher momentum W , pT 160 GeV
W decay products can be captured within a
single large-R jets (R ≥ 1.0)
Some overlap in between for partially
resolved systems
So, how can we use this information to our
advantage?
6 / 48
Alex Martyniuk
X → VV → JJ
A quick aside: ATLAS calorimetry
Jets in ATLAS are formed from
topoclusters
Logical combinations of
adjacent energy deposits in
the calorimeter cells
The calorimeters in ATLAS
have a fine granularity
Tile: ∆R ≈ 0.1
EM LAr: ∆R ≈ 0.025
We have the resolution to pick
apart large-R jets and look at
the substructure
Therefore we can use the guts
of boosted jets to our
advantage
7 / 48
Alex Martyniuk
X → VV → JJ
Bosonic vs QCD Jets
QCD jets
Bosonic jets
Form two narrow regions with high
energy density corresponding to each
quark
Each quark carries a roughly equal
fraction of the boson momentum in
the lab frame
Jet mass originates from the boson
mass, i.e. peaked
8 / 48
Alex Martyniuk
Narrow region with high energy
density corresponding to a single
quark/gluon
Majority of the jet momentum is
concentrated in this single region
Jet mass originates from the spread of
the energy deposition by the single
parton/any final state radiation, i.e.
essentially random
X → VV → JJ
Fat Jet→Grooming→Tagging
1
Reconstruct decay as fat-jet
2
Groom the jet
Use large-R parameter jet to collect radiation from the original decay
Signal: Remove unwanted jet constituents not from the signal, e.g. pile-up
Background: Preserve the background characteristics
3
Tag as boson jet
Use differences between signal and background jet characteristics to reject
background jets
9 / 48
Alex Martyniuk
X → VV → JJ
Cambridge-Aachen Jets
Cambridge-Aachen jets (CA jets)
[arXiv:9707323] or [arXiv:0802.2470]
Part of the sequential recombination
family of jet reconstruction algorithms
Calculate the ∆Rij between all jet
constituents
Combine closest constituents first
Merge while R ≤ 1.2 (in this analysis)
If there are no components within 1.2,
redefine as a jet and remove from the
collection of constituents
Merge until there are no components
left
NO pT dependence!
Therefore can look into the history and
use the pT splitting information
10 / 48
Alex Martyniuk
X → VV → JJ
BDRS Split-Filter
The BDRS split filtering algorithm, [arXiv:0802.2470], decomposes CA jets
sequential clustering to find hard substructure within
Originally defined to find boosted H → bb decays
The decomposition follows some simple steps
For jet j, undo the last step of clustering forming jets j1 and j2 (mj1 > mj2 )
If there was a large mass drop, mj1 < µmax mj and the pT balance is not too
asymmetric,
min(pT2 ,pT2 )
j1
mj2
0
j2
∆Rj2 ,j ≥ ymin , define j as from a hard splitting and stop
1 2
Otherwise redefine j as j1 , discard j2 , and continue
Filter the resulting jet by re-clustering as nr × Rr sized subjets
11 / 48
Alex Martyniuk
X → VV → JJ
BDRS-A Split-Filter
In this analysis a modified BDRS-A
split filtering algorithm is used
Starts from R = 1.2 CA jets seeded
from local cluster weighted (LCW)
topological clusters
Loose BDRS tagger, with no mass
drop requirement
12 / 48
Alex Martyniuk
Iterative parameter
√
ymin
µmax
Filtering parameter
nr
Rr
X → VV → JJ
Value
0.20
1.00
Value
3
0.3
BDRS-A CA R = 1.2 Jet Calibration
Derive the mean
reconstructed jet energy,
jet
< Ereco >
jet
RE
Fit the <
> vs
jet
< Ereco > distribution to
gain a calibration
function
Repeat process for
mass calibration using
the LCW+JES jets
13 / 48
1.1
Jet mass response before calibration
1.2
1.15
ATLAS Simulation
E = 450 GeV
E = 800 GeV
E = 1100 GeV
E = 1400 GeV
E = 1800 GeV
Pythia 8 dijets, s = 8 TeV
BDRS-A jets, C/A R = 1.2
Tagging selection applied
1.05
1
0.95
0.9
0.85
-2
-1.5
-1
-0.5
0
0.5
1
1.5
1.2
1.15
1.1
ATLAS Simulation
E = 450 GeV
E = 800 GeV
E = 1100 GeV
E = 1400 GeV
E = 1800 GeV
Pythia 8 dijets, s = 8 TeV
BDRS-A jets, C/A R = 1.2
Tagging selection applied
1.05
1
0.95
0.9
0.85
-2
-1.5
-1
-0.5
Alex Martyniuk
0
0.5
1
1.5
2
Jet η
X → VV → JJ
1.2
1.15
1.1
ATLAS Simulation
E = 450 GeV
E = 800 GeV
E = 1100 GeV
E = 1400 GeV
E = 1800 GeV
Pythia 8 dijets, s = 8 TeV
BDRS-A jets, C/A R = 1.2
Tagging selection applied
1.05
1
0.95
0.9
0.85
-2
2
Jet η
Jet mass response after calibration
Fit the responses with a
Gaussian fit, to gain
mean response in each
jet
bin, < RE >
Jet energy response after calibration
Calculate the jet energy
response in bins of ηdet
and Etruth
Jet energy response before calibration
Particle level jet energy and mass calibrations were derived and applied to the
BDRS-A CA R = 1.2 jets used in the analysis
Effectively restores jet energy/mass response over the full jet E and η range
-1.5
-1
-0.5
0
0.5
1
1.5
2
Jet η
1.2
1.15
1.1
ATLAS Simulation
E = 450 GeV
E = 800 GeV
E = 1100 GeV
E = 1400 GeV
E = 1800 GeV
Pythia 8 dijets, s = 8 TeV
BDRS-A jets, C/A R = 1.2
Tagging selection applied
1.05
1
0.95
0.9
0.85
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Jet η
Killing the Background
Let me briefly try to quantify the level of the dominant QCD background the
analysis will encounter
Other backgrounds contribute, at a significantly lower rates
All modelled to be smoothly falling
Events / 100 GeV
It is a lot.... an awful lot
Leading jet pT
[TeV]
0.5
1.0
8
10
ATLAS
107
Data
s = 8 TeV, 20.3 fb-1
Background model
6
10
Significance (stat)
Significance (stat + syst)
5
10
No boson tagging
104
10
102
Significance
10
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
14 / 48
dσ
W 0 dp
T
[fb/GeV]
10−1
10−3
S/B
[−]
10−4
10−4
Rough order of magnitude differential cross
sections taken from MC show the extent of
the problem, 1 signal in 10k background
events
3
4
2
0
−2
dσ
QCD dp
T
[fb/GeV]
103
10
Alex Martyniuk
Obvious problem when you look at the raw
events selected by the jet trigger used in the
analysis
Our jet substructure thresher has quite the
haystack against it
X → VV → JJ
Tools at our disposal
What do we have to remove the
QCD background?
0.2
0.18
bulk GRS → WW (m = 1.8 TeV)
ATLAS Simulation
s = 8 TeV
G
bulk GRS → ZZ (m = 1.8 TeV)
G
0.16
Pythia QCD dijet
0.14
0.12
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
y > 0.45
0.1
0.08
0.06
0.3
ATLAS Simulation
s = 8 TeV
EGM W' → WZ (m = 1.8 TeV)
W'
Pythia QCD dijet
0.25
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
60 ≤ mj < 110 GeV
0.2
0.15
0.25
0.2
ATLAS Simulation
s = 8 TeV
0.15
0.05
0.1
0.15
0.2
0.25
Filtered jet mass
Separates peaked
boson mass from
falling QCD spectrum
15 / 48
0
0
0.3
0.35
mj [TeV]
3
W'
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
60 ≤ mj < 110 GeV
y > 0.45
0.1
0.05
0.05
0
0
EGM W' → WZ (m = 1.8 TeV)
Pythia QCD dijet
0.1
0.04
0.02
2
0.35
Fraction of jets / 3
Fraction of jets / 5.0 GeV
0.22
Fraction of jets / 0.05
The BDRS-A filtered CA R = 1.2 jets
Selects two (three) pronged
decays within jets
1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Subjet momentum balance
Boson jets
symmetric, QCD
unbalanced
Alex Martyniuk
0
0
0.9
1
Jet y
X → VV → JJ
4
20
40
60
80
100
120
Jet ungroomed ntrk
Number of tracks ghost
matched to the unfiltered jet
More hadronic
activity in QCD jets
Tools: Jet mass
2
Filtered jet mass
Apply ±13 GeV window cuts
around boson mass from MC
simulation peak (mW = 82.4,
mZ = 92.8)
For example, in the WZ cut;
Leading mass jet
79.8 GeV < mjet < 105.8 GeV
Subleading mass jet
69.4 GeV < mjet < 95.4 GeV
Fraction of jets / 5.0 GeV
Separates peaked boson mass from falling QCD spectrum
Very powerful cut!
signal ≈ 80%
background ≈ 10 − 15%
0.22
0.2
0.18
G
bulk GRS → ZZ (m = 1.8 TeV)
G
0.16
Pythia QCD dijet
0.14
0.12
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
y > 0.45
0.1
0.08
0.06
0.04
0.02
Cuts optimised using a data CR
0
0
0.05
Dijet formed from two
tagged/un-tagged regions
N.B. Windows overlap!!!
16 / 48
bulk GRS → WW (m = 1.8 TeV)
ATLAS Simulation
s = 8 TeV
Alex Martyniuk
X → VV → JJ
0.1
0.15
0.2
0.25
0.3
0.35
mj [TeV]
Tools: Subjet momentum balance
3
Subjet momentum balance
Soft gluon radiation leads to
asymmetric splittings
W /Z → q q̄ decays tend to share
momentum more equally between
decay products
√
Apply a more stringent y ≥ 0.45
cut on the subjet momentum
balance
Fraction of jets / 0.05
Boson jets symmetric, QCD unbalanced
0.3
ATLAS Simulation
s = 8 TeV
EGM W' → WZ (m = 1.8 TeV)
W'
Pythia QCD dijet
0.25
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
60 ≤ mj < 110 GeV
0.2
0.15
0.1
Another powerful cut!
signal ≈ 70%
background ≈ 30%
0.05
Cuts optimised using MC, using a
wide mass window,
60 GeV < mjet < 110 GeV
17 / 48
0.35
Alex Martyniuk
0
0
0.1
0.2
X → VV → JJ
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Jet y
Tools: Ghost matched jets
4
Number of tracks ghost matched to the unfiltered jet
More hadronic activity in QCD jets
Expect increased hadronic activity
from gluon
Use the number of ghost
associated ungroomed tracks, ntrk ,
as a proxy for hadronic activity,
[arXiv:0802.1188]
Fraction of jets / 3
Emission of hard gluon dominates
after mass/asymmetry cuts
0.25
0.2
ATLAS Simulation
s = 8 TeV
Efficiency after mass/asymmetry
signal = 83 ± 7%
background ≈ 65%
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
60 ≤ mj < 110 GeV
y > 0.45
0.05
0
0
Very hard to model in MC
20
Cuts optimised using V +jets
enriched data CR
Efficiency calibrated in this CR
18 / 48
Alex Martyniuk
W'
Pythia QCD dijet
0.15
0.1
Apply ntrk ≤ 30 cut
EGM W' → WZ (m = 1.8 TeV)
X → VV → JJ
40
60
80
100
120
Jet ungroomed ntrk
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
3
Require mJJ > 1.05 TeV
Ensures on trigger plateau
4
Rapidity gap between leading jets,
|∆y12 | < 1.2
Events / 100 GeV
Event selection: Putting it all together
s-channel signal more central than
t-channel QCD
5
ATLAS
107
Leading jets pT asymmetry ApT < 0.15
Background model
6
10
Significance (stat)
Significance (stat + syst)
5
10
No boson tagging
3
10
102
Leading jets |η| < 2.0
7
Correction for jets on calorimetry holes
8
Boson tagging cuts
Significance
10
Ensures a good overlap with tracker
4
2
0
−2
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
Alex Martyniuk
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
Background efficiencies
19 / 48
Data
s = 8 TeV, 20.3 fb-1
104
Used as proxy for large-R jet cleaning
6
8
10
X → VV → JJ
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
3
Require mJJ > 1.05 TeV
Events / 0.1
Event selection: Putting it all together
×103
160
W' (1400 GeV) x 3000 ATLAS
s = 8 TeV , 20.3fb-1
QCD (Pythia)
QCD (Herwig)
QCD (PowPy)
Data
140
120
100
Ensures on trigger plateau
1.26 ≤ mjj < 1.54 TeV
80
4
Rapidity gap between leading jets,
|∆y12 | < 1.2
60
40
s-channel signal more central than
t-channel QCD
Used as proxy for large-R jet cleaning
6
0
0
Leading jets pT asymmetry ApT < 0.15
0.5
Ensures a good overlap with tracker
12000
10000
7
Correction for jets on calorimetry holes
8
Boson tagging cuts
2
2.5
3
|∆y |
2.16 ≤ mjj < 2.64 TeV
8000
6000
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
Background efficiencies
4000
2000
0
0
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
19 / 48
1.5
W' (2400 GeV) x 15000 ATLAS
s = 8 TeV , 20.3fb-1
QCD (Pythia)
QCD (Herwig)
QCD (PowPy)
Data
14000
Leading jets |η| < 2.0
1
jj
Events / 0.1
5
20
0.5
1
1.5
2
2.5
3
|∆y |
jj
Alex Martyniuk
X → VV → JJ
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
3
Require mJJ > 1.05 TeV
40000
Rapidity gap between leading jets,
|∆y12 | < 1.2
s-channel signal more central than
t-channel QCD
6
20000
10000
0
0
Leading jets pT asymmetry ApT < 0.15
Used as proxy for large-R jet cleaning
Leading jets |η| < 2.0
Ensures a good overlap with tracker
1000
8
Boson tagging cuts
600
19 / 48
0.15
0.2
0.25
0.3
0.35
(p -p )/(p +p )
T1
T2
|η| < 2
2.16 ≤ mjj < 2.64 TeV
400
200
0
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
(p -p )/(p +p )
T1
Alex Martyniuk
T2
W' (2400 GeV) x 2500 ATLAS
QCD (Pythia)
s = 8 TeV , 20.3fb-1
QCD (Herwig)
QCD (PowPy)
|∆ y | < 1.2
Data
jj
1200
800
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
0.1
T1
Correction for jets on calorimetry holes
Background efficiencies
0.05
1400
7
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
|η| < 2
1.26 ≤ mjj < 1.54 TeV
30000
Events / 0.01
5
W' (1400 GeV) x 2500 ATLAS
QCD (Pythia)
s = 8 TeV , 20.3fb-1
QCD (Herwig)
QCD (PowPy)
|∆ y | < 1.2
Data
jj
50000
Ensures on trigger plateau
4
Events / 0.01
Event selection: Putting it all together
X → VV → JJ
T2
T1
T2
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
Require mJJ > 1.05 TeV
Rapidity gap between leading jets,
|∆y12 | < 1.2
s-channel signal more central than
t-channel QCD
5
W' (1400 GeV) x 2500
QCD (Pythia)
QCD (Herwig)
QCD (PowPy)
Data
40000
Ensures on trigger plateau
4
50000
30000
6
Leading jets |η| < 2.0
7
Correction for jets on calorimetry holes
8
Boson tagging cuts
Ensures a good overlap with tracker
0
-3
Background efficiencies
1400
1200
1000
Alex Martyniuk
-1
0
W' (2400 GeV) x 2500
QCD (Pythia)
QCD (Herwig)
QCD (PowPy)
Data
1
2
3
jet η
ATLAS
s = 8 TeV , 20.3fb-1
|∆ y | < 1.2
jj
2.16 ≤ mjj < 2.64 TeV
600
400
200
0
-3
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
19 / 48
-2
1600
800
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
|∆ y | < 1.2
jj
1.26 ≤ mjj < 1.54 TeV
10000
Leading jets pT asymmetry ApT < 0.15
Used as proxy for large-R jet cleaning
ATLAS
s = 8 TeV , 20.3fb-1
20000
Jets / 0.1
3
Jets / 0.1
Event selection: Putting it all together
X → VV → JJ
-2
-1
0
1
2
3
jet η
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
Require mJJ > 1.05 TeV
Rapidity gap between leading jets,
|∆y12 | < 1.2
s-channel signal more central than
t-channel QCD
5
W' (1400 GeV) x 2500
QCD (Pythia)
QCD (Herwig)
QCD (PowPy)
Data
40000
Ensures on trigger plateau
4
50000
30000
6
Leading jets |η| < 2.0
7
Correction for jets on calorimetry holes
8
Boson tagging cuts
Ensures a good overlap with tracker
0
-3
Background efficiencies
1400
1200
1000
Alex Martyniuk
-1
0
W' (2400 GeV) x 2500
QCD (Pythia)
QCD (Herwig)
QCD (PowPy)
Data
1
2
3
jet η
ATLAS
s = 8 TeV , 20.3fb-1
|∆ y | < 1.2
jj
2.16 ≤ mjj < 2.64 TeV
600
400
200
0
-3
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
19 / 48
-2
1600
800
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
|∆ y | < 1.2
jj
1.26 ≤ mjj < 1.54 TeV
10000
Leading jets pT asymmetry ApT < 0.15
Used as proxy for large-R jet cleaning
ATLAS
s = 8 TeV , 20.3fb-1
20000
Jets / 0.1
3
Jets / 0.1
Event selection: Putting it all together
X → VV → JJ
-2
-1
0
1
2
3
jet η
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
3
Require mJJ > 1.05 TeV
Fraction of jets / 5.0 GeV
Event selection: Putting it all together
Leading jets pT asymmetry ApT < 0.15
G
bulk GRS → ZZ (m = 1.8 TeV)
G
Pythia QCD dijet
0.14
0.12
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
y > 0.45
0.1
0
0
Fraction of jets / 0.05
5
bulk GRS → WW (m = 1.8 TeV)
ATLAS Simulation
s = 8 TeV
0.16
0.04
0.02
Rapidity gap between leading jets,
|∆y12 | < 1.2
s-channel signal more central than
t-channel QCD
0.2
0.18
0.08
0.06
Ensures on trigger plateau
4
0.22
0.05
0.1
0.15
0.3
ATLAS Simulation
s = 8 TeV
0.25
0.3
0.35
mj [TeV]
EGM W' → WZ (m = 1.8 TeV)
W'
Pythia QCD dijet
0.25
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
60 ≤ mj < 110 GeV
0.2
0.15
Used as proxy for large-R jet cleaning
0.2
0.35
0.1
6
Leading jets |η| < 2.0
0.05
0
0
7
8
Correction for jets on calorimetry holes
Boson tagging cuts
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
Background efficiencies
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Jet y
0.25
0.2
ATLAS Simulation
s = 8 TeV
EGM W' → WZ (m = 1.8 TeV)
W'
Pythia QCD dijet
0.15
|∆ y | < 1.2
jj
|η | < 2
j
1.62 ≤ mjj < 1.98 TeV
60 ≤ mj < 110 GeV
y > 0.45
0.1
0.05
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
19 / 48
Fraction of jets / 3
Ensures a good overlap with tracker
0
0
Alex Martyniuk
X → VV → JJ
20
40
60
80
100
120
Jet ungroomed ntrk
1
Trigger: pT > 360 GeV anti-kt 1.0
2
Apply BDRS-A split-filter
3
Require mJJ > 1.05 TeV
4
Rapidity gap between leading jets,
|∆y12 | < 1.2
Ensures on trigger plateau
Selection efficiency
Event selection: Putting it all together
Ensures a good overlap with tracker
7
Correction for jets on calorimetry holes
8
Boson tagging cuts
Jet mass (WZ , WW , ZZ ), momentum
balance, ntrk
EGM W'→ WZ
bulk GRS → WW
bulk GRS → ZZ
0.8
0.6
Event topology requirements
0.5
Selection efficiency
Leading jets |η| < 2.0
ATLAS Simulation
s = 8 TeV
0.9
Leading jets pT asymmetry ApT < 0.15
Used as proxy for large-R jet cleaning
6
1
0.7
s-channel signal more central than
t-channel QCD
5
1.1
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
Resonance Mass [TeV]
0.3
0.25
ATLAS Simulation
s = 8 TeV
EGM W'→ WZ
bulk GRS → WW
bulk GRS → ZZ
0.2
0.15
0.1
0.05
Background efficiencies
0
Topological ≈ 48%
Tagger ≈ 1.2 − 0.6%
19 / 48
Alex Martyniuk
1.4
X → VV → JJ
1.6
1.8
2
2.2
2.4
2.6
2.8
3
Resonance Mass [TeV]
Modelling the Background
Events / 100 GeV
After trying to kill the background we now arrive at the point of modelling it
MC statistics needed to properly model the high mJJ tail are prohibitively large
Assume a steeply and smoothly falling
distribution models the background
8
10
ATLAS
107
Data
s = 8 TeV, 20.3 fb-1
Any resonance should be narrow, thus only
affect a few bins
Background model
6
10
Significance (stat)
Use a parametric function to model the
background from the data
dn
= p1 (1 − x)p2 −ξp3 x p3
dx
Significance (stat + syst)
5
10
No boson tagging
4
10
3
10
102
Where,
Significance
10
4
2
0
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
20 / 48
Alex Martyniuk
√
x = mJJ / s
mJJ is the dijet invariant mass,
p1 is a normalisation factor,
p2 and p3 are dimensionless shape
parameters
ξ is a dimensionless constant chosen after
fitting to minimise the correlations between
p2 and p3
X → VV → JJ
Modelling the Background
Events / 100 GeV
After trying to kill the background we now arrive at the point of modelling it
MC statistics needed to properly model the high mJJ tail are prohibitively large
Assume a steeply and smoothly falling
distribution models the background
8
10
ATLAS
107
Data
s = 8 TeV, 20.3 fb-1
Any resonance should be narrow, thus only
affect a few bins
Background model
6
10
Significance (stat)
Use a parametric function to model the
background from the data
dn
= p1 (1 − x)p2 −ξp3 x p3
dx
Significance (stat + syst)
5
10
No boson tagging
4
10
3
10
102
Where,
Significance
10
4
2
0
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
20 / 48
Alex Martyniuk
√
x = mJJ / s
mJJ is the dijet invariant mass,
p1 is a normalisation factor,
p2 and p3 are dimensionless shape
parameters
ξ is a dimensionless constant chosen after
fitting to minimise the correlations between
p2 and p3
X → VV → JJ
Modelling the Background
Events / 100 GeV
After trying to kill the background we now arrive at the point of modelling it
MC statistics needed to properly model the high mJJ tail are prohibitively large
8
10
ATLAS
107
Data
s = 8 TeV, 20.3 fb-1
Assume a steeply and smoothly falling
distribution models the background
Background model
6
10
Significance (stat)
Any resonance should be narrow, thus only
affect a few bins
Significance (stat + syst)
5
10
No boson tagging
104
Use a parametric function to model the
background from the data
dn
= p1 (1 − x)p2 −ξp3 x p3
dx
3
10
102
Significance
10
4
2
0
−2
1.5
2
2.5
3
3.5
Error taken from errors on functional
parameters
1.5
2
2.5
3
3.5
mjj [TeV]
Sample
Pythia dijet events
Herwig++ dijet events
Data with 110 < mj1 ≤ 140 GeV and 40 < mj2 ≤ 60 GeV
Data with 40 < mj ≤ 60 GeV for both jets
Data with 110 < mj ≤ 140 GeV for both jets
2
χ /nDOF
24.6/22
15.9/22
12.1/11
19.8/13
5.0/6
Probability
0.31
0.82
0.79
0.56
0.91
Fit tested on,
Raw data
P YTHIA/H ERWIG MC
Mass sideband data CRs
Alternate fit functions give similar results
20 / 48
Alex Martyniuk
X → VV → JJ
Modelling the Background
Events / 100 GeV
After trying to kill the background we now arrive at the point of modelling it
MC statistics needed to properly model the high mJJ tail are prohibitively large
104
ATLAS
3
10
Data
s = 8 TeV, 20.3 fb-1
Assume a steeply and smoothly falling
distribution models the background
Background model
Significance (stat)
102
Any resonance should be narrow, thus only
affect a few bins
Significance (stat + syst)
40 < mj < 60 GeV
10
Use a parametric function to model the
background from the data
dn
= p1 (1 − x)p2 −ξp3 x p3
dx
1
10−1
Significance
10−2
4
2
0
−2
1.5
2
2.5
3
3.5
Error taken from errors on functional
parameters
1.5
2
2.5
3
3.5
mjj [TeV]
Sample
Pythia dijet events
Herwig++ dijet events
Data with 110 < mj1 ≤ 140 GeV and 40 < mj2 ≤ 60 GeV
Data with 40 < mj ≤ 60 GeV for both jets
Data with 110 < mj ≤ 140 GeV for both jets
χ2 /nDOF
24.6/22
15.9/22
12.1/11
19.8/13
5.0/6
Probability
0.31
0.82
0.79
0.56
0.91
Fit tested on,
Raw data
P YTHIA/H ERWIG MC
Mass sideband data CRs
Alternate fit functions give similar results
20 / 48
Alex Martyniuk
X → VV → JJ
Modelling the Background
Events / 100 GeV
After trying to kill the background we now arrive at the point of modelling it
MC statistics needed to properly model the high mJJ tail are prohibitively large
5
10
ATLAS
104
Assume a steeply and smoothly falling
distribution models the background
Data
s = 8 TeV, 20.3 fb-1
Background model
3
Significance (stat)
10
Any resonance should be narrow, thus only
affect a few bins
Significance (stat + syst)
102
40 < mj1,j2 < 60 GeV
Use a parametric function to model the
background from the data
dn
= p1 (1 − x)p2 −ξp3 x p3
dx
110 < mj2,j1 < 140 GeV
10
1
10−1
Significance
10−2
4
2
0
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
Sample
Pythia dijet events
Herwig++ dijet events
Data with 110 < mj1 ≤ 140 GeV and 40 < mj2 ≤ 60 GeV
Data with 40 < mj ≤ 60 GeV for both jets
Data with 110 < mj ≤ 140 GeV for both jets
20 / 48
χ2 /nDOF
24.6/22
15.9/22
12.1/11
19.8/13
5.0/6
Probability
0.31
0.82
0.79
0.56
0.91
Alex Martyniuk
Error taken from errors on functional
parameters
Fit tested on,
Raw data
P YTHIA/H ERWIG MC
Mass sideband data CRs
Alternate fit functions give similar results
X → VV → JJ
Modelling the Background
Events / 100 GeV
After trying to kill the background we now arrive at the point of modelling it
MC statistics needed to properly model the high mJJ tail are prohibitively large
104
ATLAS
3
10
Assume a steeply and smoothly falling
distribution models the background
Data
s = 8 TeV, 20.3 fb-1
Background model
Significance (stat)
102
Any resonance should be narrow, thus only
affect a few bins
Significance (stat + syst)
110 < mj < 140 GeV
10
Use a parametric function to model the
background from the data
dn
= p1 (1 − x)p2 −ξp3 x p3
dx
1
10−1
Significance
10−2
4
2
0
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
Sample
Pythia dijet events
Herwig++ dijet events
Data with 110 < mj1 ≤ 140 GeV and 40 < mj2 ≤ 60 GeV
Data with 40 < mj ≤ 60 GeV for both jets
Data with 110 < mj ≤ 140 GeV for both jets
20 / 48
χ2 /nDOF
24.6/22
15.9/22
12.1/11
19.8/13
5.0/6
Probability
0.31
0.82
0.79
0.56
0.91
Alex Martyniuk
Error taken from errors on functional
parameters
Fit tested on,
Raw data
P YTHIA/H ERWIG MC
Mass sideband data CRs
Alternate fit functions give similar results
X → VV → JJ
Systematic uncertainties: Shape
Background: Taken from the uncertainties on the fit parameters
Signal: Various systematics affect the signal reconstruction and selection
efficiency
The jet pT and jet mass scale
uncertainties determined by the
track/calo double ratio technique
For example for a variable x,
data /x data
xtrack
calo
MC
MC
xtrack /xcalo
18
16
ATLAS Simulation
s = 8 TeV , 20.3fb-1
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
y >= 0.45
f
ntrk < 30
14
12
W'(1.4 TeV) → WZ → qqqq
10
8
6
4
Applied as a Gaussian with µ = 1 and σ
equal to the observed uncertainty
jet pT scale: 2%
jet mass scale: 3%
2
0
0
Nominal
JMS up
JMS down
JMR
20
40
60
80
100
120
140
160
max(m , m2) [GeV]
1
An uncertainty on the jet pT resolution of
20% is applied as an additional
smearing on top of the nominal 5%
21 / 48
Jets / 5 GeV
Shape systematics:
Alex Martyniuk
Source
Jet pT scale
Jet pT resolution
Jet mass scale
X → VV → JJ
Uncertainty
2%
20%
3%
Constraining pdf
G(αPT |1, √
0.02)
G(σrE | 0, 0.05 × 1.22 − 12 )
G(αm |1, 0.03)
Systematic uncertainties: Normalisation
Large uncertainty on the ntrk cut
evaluated in the data driven V +jets
study used to define the efficiency of the
cut
Jet mass scale affects both shape and
normalisation strongly
√
y scale evaluated using the double
ratio method
Jets / 5 GeV
Background: Taken from the uncertainties on the fit parameters
Signal: Various systematics affect the signal reconstruction and selection
20
efficiency
BDRS-A
Normalisation systematics:
18 ATLAS Simulation -1
16
14
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
s = 8 TeV , 20.3fb
W'(1.4 TeV) → WZ → qqqq
12
10
8
6
4
2
0
0
Nominal
JMS up
JMS down
JMR
20
40
60
80
100
120
140
160
min(m , m2) [GeV]
1
Resolutions taken as 20% smearings
Shower model evaluated by comparing
MC showered by P YTHIA or H ERWIG
PDF4LHC method used to evaluate
PDF uncertainties
ATLAS luminosity uncertainty
assumed
22 / 48
Alex Martyniuk
Source
Efficiency of the track-multiplicity cut
Jet mass scale
Jet mass resolution
Subjet momentum-balance scale
Subjet momentum-balance resolution
Parton shower model
Parton distribution functions
Luminosity
X → VV → JJ
Uncertainty
20.0%
5.0%
5.5%
3.5%
2.0%
5.0%
3.5%
2.8%
Events / 100 GeV
The long and winding road....
8
10
ATLAS
107
Data
s = 8 TeV, 20.3 fb-1
Background model
6
10
Significance (stat)
Significance (stat + syst)
5
10
No boson tagging
104
OK, enough with the
build-up.....
3
10
What does the triggered
data look like after
applying our selection???
102
Significance
10
4
2
0
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
23 / 48
Alex Martyniuk
X → VV → JJ
Events / 100 GeV
X → WZ selection
104
Full WZ selection applied
to the data
Data
Background model
1.5 TeV EGM W', c = 1
2.0 TeV EGM W', c = 1
2.5 TeV EGM W', c = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
s = 8 TeV, 20.3 fb-1
10
102
Z mass window applied
to leading mass jet
W mass window
applied to sub-leading
mass jet
WZ Selection
10
Good agreement seen
with steeply, smoothly
falling background model
in the low/high mass
regions
1
Significance
10−1
3
2
1
0
−1
−2
1.5
1.5
2
2
2.5
2.5
3
3
3.5
3.5
mjj [TeV]
24 / 48
Alex Martyniuk
X → VV → JJ
Deviation from the
background observed at
around 2 TeV
Benchmark extended
gauge model W 0 signal
MC shown for
comparison purposes
X → WW selection
Events / 100 GeV
104
Full WW selection applied
to the data
Data
Background model
1.5 TeV Bulk GRS, k/MPI = 1
2.0 TeV Bulk GRS, k/MPI = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
s = 8 TeV, 20.3 fb-1
102
10
W mass window
applied to both jets
Good agreement again
seen with steeply,
smoothly falling
background model
WW Selection
1
10−1
Deviation from the
background still observed
at around 2 TeV
10−2
−3
Significance
10
3
2
1
0
−1
−2
1.5
1.5
2
2
2.5
2.5
3
3
3.5
3.5
mjj [TeV]
25 / 48
Alex Martyniuk
X → VV → JJ
Remember: There is an
overlap between the W /Z
mass windows
Benchmark Bulk
Randall-Sundrum graviton
signal MC shown for
comparison purposes
X → ZZ selection
Events / 100 GeV
104
Full ZZ selection applied
to the data
Data
Background model
1.5 TeV Bulk GRS, k/MPI = 1
2.0 TeV Bulk GRS, k/MPI = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
s = 8 TeV, 20.3 fb-1
102
10
Z mass window
applied to both jets
Good agreement again
seen with steeply,
smoothly falling
background model
ZZ Selection
1
10−1
Deviation from the
background still observed
at around 2 TeV
10−2
−3
Significance
10
3
2
1
0
−1
−2
1.5
1.5
2
2
2.5
2.5
3
3
3.5
3.5
mjj [TeV]
26 / 48
Alex Martyniuk
X → VV → JJ
Remember: There is an
overlap between the W /Z
mass windows
Benchmark Bulk
Randall-Sundrum graviton
signal MC shown for
comparison purposes
What do these events look like? Dramatic!
27 / 48
Alex Martyniuk
X → VV → JJ
What do these events look like? Energetic!
28 / 48
Alex Martyniuk
X → VV → JJ
Digging deeper into the jets....
The ATLAS detector volume is
shown unfolded in y and φ
φ
These jet event displays take a bit more explanation, but offer a powerful
insight into the analysis jets
Inner detector track positions are
shown as crosses
Black Tracks: From primary
vertex
Blue Tracks: From secondary
vertices
6
5
ATLAS
s = 8 TeV
Run: 203027 Event: 5303984
4
1
3
1
Calorimeter deposits are displayed
on the rainbow scale
The outlines of the CA 1.2 jets are
shown
2
1
1
1
1
1
0
-4
Black: Leading pT jet
Mauve: Sub-leading jet
-3
-2
Grey area: Sub-jets after filtering
29 / 48
Alex Martyniuk
X → VV → JJ
-1
0
1
2
3
4 1
y
φ
φ
Digging deeper into the jets....
6
5
6
ATLAS
s = 8 TeV
5
Run: 201556 Event: 7295269
4
ATLAS
s = 8 TeV
Run: 201489 Event: 75855145
4
3
10
3
1
3
102
10
2
1
2
1
1
1
1
1
10-1
1
10-2
0
-4
-3
-3
-2
-1
0
1
2
3
4
y
10
1
0
-4
-3
-2
-1
What do we see here?
Subjets are highly collimated
PV tracks are highly correlated with the selected sub-jets
Energy deposits concentrated in the sub-jet
Pile-up tracks/deposits sparsely distributed over the events
Successfully picked the boson out of the pileup?
30 / 48
Alex Martyniuk
X → VV → JJ
0
1
2
3
4 1
y
Time to cross-check.....
Deviations from the expected background, especially ones at the tail of the data
distribution, mean one thing for physicists....
What on Earth did we do wrong?
Try to evaluate any possible issues with the analysis
Operation Cross-check begins
Data
Background model
1.5 TeV EGM W', c = 1
2.0 TeV EGM W', c = 1
2.5 TeV EGM W', c = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
s = 8 TeV, 20.3 fb-1
10
102
104
Data
Background model
1.5 TeV Bulk GRS, k/MPI = 1
2.0 TeV Bulk GRS, k/MPI = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
Events / 100 GeV
104
Events / 100 GeV
Events / 100 GeV
Look for mistakes, bugs or shaping effects in:
Detector/data taking effects
Jet reconstruction effects
Event selection effects
s = 8 TeV, 20.3 fb-1
2
10
10
WZ Selection
104
Data
Background model
1.5 TeV Bulk GRS, k/MPI = 1
2.0 TeV Bulk GRS, k/MPI = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
s = 8 TeV, 20.3 fb-1
2
10
10
WW Selection
10
ZZ Selection
1
1
−1
1
10−1
10−1
10−2
10−2
10
−3
−3
2
2.5
3
1.5
2
2.5
3
3.5
3.5
3
2
1
0
−1
−2
10
1.5
2
2.5
3
1.5
2
2.5
3
mjj [TeV]
31 / 48
3.5
3.5
mjj [TeV]
Alex Martyniuk
X → VV → JJ
Significance
1.5
Significance
Significance
10
3
2
1
0
−1
−2
3
2
1
0
−1
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
Time to cross-check.....
From E. Kajomovitz
32 / 48
Alex Martyniuk
X → VV → JJ
1
0.5
15
0
10
Is one cut driving all of the
deviation?
Many, many, many..... more,
you can only get so much
approved...
33 / 48
s = 8 TeV, 20.3 fb-1
25
0.5
20
15
10
-1
5
40 < mj1 < 60 GeV
1.5
2
2.5
3
67.5 < mj1 < 105.8 GeV
-1.5
40 < mj2 < 60 GeV
3.5
-2
1
0
1.5
2
2.5
3
25
1
20
0.5
15
0
Sub-leading jet η
ATLAS
s = 8 TeV, 20.3 fb-1
3.5
0
mjj [TeV]
2
1.5
5
40 < mj2 < 60 GeV
mjj [TeV]
Sub-leading jet η
30
1
-0.5
-1
-2
1
35
ATLAS
45
2
ATLAS
1.5
40
s = 8 TeV, 20.3 fb-1
35
1
30
0.5
25
0
20
-0.5
10
-1
5
40 < mj1 < 60 GeV
67.5 < mj2 < 105.8 GeV
1.5
2
2.5
3
3.5
mjj [TeV]
Alex Martyniuk
15
-1
-1.5
-2
1
-0.5
X → VV → JJ
0
10
67.5 < mj1 < 105.8 GeV
-1.5
-2
1
67.5 < mj2 < 105.8 GeV
1.5
2
2.5
3
3.5
mjj [TeV]
5
0
Number of events
-0.5
Look at the effect of single cuts
on the distribution
2
1.5
0
-1.5
Look at the effect of N − 1 cuts
on the distribution
20
Number of events
s = 8 TeV, 20.3 fb-1
Sub-leading jet η
ATLAS
Number of events
25
2
1.5
Number of events
Started trawling through
SR/CR kinematic distributions,
looking for unusual features in
the signal regions
Sub-leading jet η
Time to cross-check.....
ATLAS
Data
s = 8 TeV, 20.3 fb-1
6
Background model
5
2.0 TeV EGM W', c = 1
4
Significance (stat + syst)
3
only jet ungroomed n
10
10
10
8
10
107
Background model
5
2.0 TeV EGM W', c = 1
4
Significance (stat + syst)
3
only jet
10
10
10
1
10−1
−2
−2
10
1.5
2
1.5
2
2.5
3
2.5
3
3.5
Significance
10
3
2
1
0
−1
−2
3.5
1.5
2
2.5
3
1.5
2
2.5
3
mjj [TeV]
107
ATLAS
Data
s = 8 TeV, 20.3 fb-1
6
Background model
10
5
2.0 TeV EGM W', c = 1
104
Significance (stat + syst)
10
only m
3
10
j
10
1
Significance
10−1
10−2
3
2
1
0
−1
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
33 / 48
Alex Martyniuk
X → VV → JJ
3.5
3.5
mjj [TeV]
8
10
102
Many, many, many..... more,
you can only get so much
approved...
f
10
1
Is one cut driving all of the
deviation?
y
102
10−1
Look at the effect of N − 1 cuts
on the distribution
Data
s = 8 TeV, 20.3 fb-1
10
102
3
2
1
0
−1
−2
ATLAS
6
10
trk
Events / 100 GeV
Look at the effect of single cuts
on the distribution
107
10
Significance
Started trawling through
SR/CR kinematic distributions,
looking for unusual features in
the signal regions
8
10
Events / 100 GeV
Events / 100 GeV
Time to cross-check.....
Is one cut driving all of the
deviation?
Background model
5
2.0 TeV EGM W', c = 1
4
Significance (stat + syst)
3
w/o mj
10
10
107
ATLAS
Data
s = 8 TeV, 20.3 fb-1
6
Background model
5
2.0 TeV EGM W', c = 1
4
Significance (stat + syst)
3
w/o jet
10
10
10
y
f
102
10
1
1
10−1
10−1
−2
−2
10
3
2
1
0
−1
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
Significance
10
3
2
1
0
−1
−2
3.5
1.5
2
2.5
3
1.5
2
2.5
3
mjj [TeV]
8
10
107
ATLAS
Data
s = 8 TeV, 20.3 fb-1
6
Background model
10
5
2.0 TeV EGM W', c = 1
104
Significance (stat + syst)
10
10
3.5
8
10
107
ATLAS
Data
s = 8 TeV, 20.3 fb-1
6
Background model
10
5
2.0 TeV EGM W', c = 1
104
Significance (stat + syst)
10
3
WZ Selection
10
trk
102
102
10
10
1
1
10−1
10−2
3
2
1
0
−1
−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
X → VV → JJ
Significance
10−1
10−2
3
2
1
0
−1
−2
Alex Martyniuk
3.5
mjj [TeV]
w/o jet ungroomed n
3
mjj [TeV]
33 / 48
8
10
10
10
Significance
Many, many, many..... more,
you can only get so much
approved...
Data
s = 8 TeV, 20.3 fb-1
Events / 100 GeV
Look at the effect of N − 1 cuts
on the distribution
ATLAS
6
10
102
Significance
Look at the effect of single cuts
on the distribution
107
10
Events / 100 GeV
Started trawling through
SR/CR kinematic distributions,
looking for unusual features in
the signal regions
8
10
Events / 100 GeV
Events / 100 GeV
Time to cross-check.....
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
Did we find any errors?
In so many cross-checks, yes,
bugs were found and fixed
Always attempted to shield the
data from any possible biases
by using control regions to test
After almost a year and a half
of scrutiny by the
analysers/Exotics
group/ATLAS we found no
major issues
Therefore we continue to
publish the final results
Run-2 was looming!
34 / 48
Alex Martyniuk
X → VV → JJ
Local p0
Observed p0 value
10
ATLAS
s=8TeV, 20.3 fb-1
WZ Selection
WW Selection
ZZ Selection
1
0σ
1σ
−1
10
2σ
10−2
3σ
10−3
10−4
A discrepancy was seen
with respect to the
expected background
distribution
1400 1600 1800 2000 2200 2400 2600 2800 3000
mjj [GeV]
Therefore, no statistically significant deviation from
the background has been observed
Alex Martyniuk
In the WZ channel:
Local p0 = 3.4σ
3.5 σ
4σ
35 / 48
Once suitably confident it
is not an error, its
significance should be
quantified
X → VV → JJ
Global p0 = 2.5σ
Global σ takes into
account the look
elsewhere effect
LEE includes weighted
contribution from
WW /ZZ channels due
to the overlap
X → WZ limits
As no significant deviation was observed, we continue to set limits on the
observed distributions
σ(pp → W') × BR(W' → WZ) [fb]
95% confidence limits set on σ × B using the CLS prescription taking into
account the systematic uncertainties and background fit
104
ATLAS
Observed 95% CL
s = 8 TeV, 20.3 fb-1
Expected 95% CL
103
± 1σ uncertainty
± 2σ unceirtainty
2
10
EGM W', c = 1
1
36 / 48
Exclusion of EGM W 0
from 1.3 − 1.5 TeV
Broad deviation from the
background observable at
around 2 TeV
10
10−1 1.4
Expected limits broadly
agree with the observed
limits
1.6
1.8
2
2.2
2.4
Alex Martyniuk
2.6
2.8
3
mW' [TeV]
X → VV → JJ
Benchmark extended
gauge model W 0 σ × B
shown for comparison
purposes
X → WW limits
As no significant deviation was observed, we continue to set limits on the
observed distributions
104
ATLAS
Observed 95% CL
s = 8 TeV, 20.3 fb-1
Expected 95% CL
103
± 1σ uncertainty
± 2σ uncertainty
102
RS
σ(pp → G ) × BR(G
RS
→ WW) [fb]
95% confidence limits set on σ × B using the CLS prescription taking into
account the systematic uncertainties and background fit
Bulk GRS k/ MPI = 1
1
1.6
1.8
2
2.2
2.4
2.6
2.8
3
mG [TeV]
RS
37 / 48
Exclusion of graviton
production at no masses
Deviation from the
background observable at
around 2.1 TeV
10
10−1 1.4
Expected limits broadly
agree with the observed
limits
Alex Martyniuk
X → VV → JJ
Benchmark Bulk
Randall-Sundrum graviton
σ × B shown for
comparison purposes
X → ZZ limits
As no significant deviation was observed, we continue to set limits on the
observed distributions
104
ATLAS
Observed 95% CL
s = 8 TeV, 20.3 fb-1
Expected 95% CL
103
± 1σ uncertainty
± 2σ uncertainty
102
Bulk GRS k/ MPI = 1
10
1
10−1 1.4
1.6
1.8
2
2.2
2.4
2.6
2.8
3
mG [TeV]
RS
38 / 48
Expected limits broadly
agree with the observed
limits
Exclusion of graviton
production at no masses
Broad deviation from the
background observable at
around 2 TeV
RS
σ(pp → G ) × BR(G
RS
→ ZZ) [fb]
95% confidence limits set on σ × B using the CLS prescription taking into
account the systematic uncertainties and background fit
Alex Martyniuk
X → VV → JJ
Benchmark Bulk
Randall-Sundrum graviton
σ × B shown for
comparison purposes
Prescale-weighted events
Are we alone? ATLAS searches
10
9
10
8
ATLAS
∫
s=8 TeV, L dt=20.3 fb
107
10
6
10
5
104
10
10
[data-fit]/fit
ATLAS resolved Run-2
dijet search
ATLAS-CONF-2015-042,
nothing seen 4
Data
Fit
q*, m = 0.6 TeV
q*, m = 2.0 TeV
q*, m = 3.5 TeV
3
102
Signif.
ATLAS resolved dijet
search [arXiv:1407.1376],
nothing seen 4
-1
ATLAS semi-leptonic
search W (lν)Z (jj)
[arXiv:http://1503.04677],
using similar BDRS-A CA
1.2 reconstruction, in
tail/nothing seen 4
1
1
0
-1
2
0
-2
0.3
0.4 0.5
1
2
3
4
5
Reconstructed mjj [TeV]
39 / 48
Alex Martyniuk
X → VV → JJ
ATLAS semi-leptonic
search W (jj)Z (ll)
[arXiv:1409.6190], using
similar BDRS-A CA 1.2
reconstruction, in
tail/nothing seen 4
Are we alone? ATLAS searches
Events
104
ATLAS Preliminary
s=13 TeV, 80 pb-1
Data
Background fit
BumpHunter interval
BlackMax, m = 4.0 TeV
BlackMax, m = 5.0 TeV
3
10
102
10
1
Signif.
3
2
1
0
−1
−2
−3
39 / 48
p -value = 0.79
Fit Range: 1.1 - 5.3 TeV
|y*| < 0.6
2
3
2
3
4
5
6 7
Reconstructed mjj [TeV]
4
Alex Martyniuk
5
6 7
mjj [TeV]
X → VV → JJ
ATLAS resolved dijet
search [arXiv:1407.1376],
nothing seen 4
ATLAS resolved Run-2
dijet search
ATLAS-CONF-2015-042,
nothing seen 4
ATLAS semi-leptonic
search W (lν)Z (jj)
[arXiv:http://1503.04677],
using similar BDRS-A CA
1.2 reconstruction, in
tail/nothing seen 4
ATLAS semi-leptonic
search W (jj)Z (ll)
[arXiv:1409.6190], using
similar BDRS-A CA 1.2
reconstruction, in
tail/nothing seen 4
mlvJ [GeV]
Alex Martyniuk
39 / 48
Diboson
Uncertainty
G*(1200 GeV)
W'(1200 GeV)
102
10
X → VV → JJ
ATLAS semi-leptonic
search W (jj)Z (ll)
[arXiv:1409.6190], using
similar BDRS-A CA 1.2
reconstruction, in
tail/nothing seen 4
800 1000 1200 1400 1600 1800 2000 2200 2400
2
1.5
1
0.5
0
∫
10
Events / 100 GeV
W → lν + ≥ 1 large-R jet
tt+single top
Multijet
ATLAS resolved Run-2
dijet search
ATLAS-CONF-2015-042,
nothing seen 4
xxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxx
xxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxx
xxxxxxxxxxx
xxxxxxxxx
xxxxxxxxxxx
xxxxxxxxx
xxxxxxx
xxxxxxxxxxx
xxxxxxxxx
xxxxxxx
xxxx
xxxxxxx
xxxxxxx
xxxx
xxxxxxx
xxxxxxx
xxxx
xxxxxx
xxxxxxx
xxxxxxx
xxxx
xxxxxx
xxxxxxx
xxxx
xxxxxx
xxxxxxx
xxxxxx
xxxxxxx
xxxx
xxxxxx
xxxxxxx
xxxxx
xxxx
xxxxx
xxxx
xxxxx
xxxx
xxxx
xxxxx
xxxx
xxxxx
xxxx
xxxxx
xxxx
xxxx
xxxxx
xxxx
xxxx
xxxx
xxxx
xxxx
xxxx
xxxx
xxx
xxxx
xxxx
xxx
xxx
xxx
xxx
xxx
xx
x
xx
xxx
xx
xxx
xx
xxx
xx
xxx
xxx
x
x
xxx
xxx
xxx
xxx
xxx
xx
x
xx
xxx
xx
xxx
xx
xxx
xx
x
x
xxx
xx
xxx
xx
xxx
xx
xxx
xx
x
x
xxx
xx
xxx
xx
xxx
xx
xxx
xxx
xxx
xxx
xxx
xxx
xxx
xxxxxxxxxxxxxxx
xxxxxxxxxxx
xxx
xxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxx
xxxxxxxxxxx
1
10-1
ATLAS resolved dijet
search [arXiv:1407.1376],
nothing seen 4
ATLAS
Data
W/Z+jets
s = 8 TeV, Ldt = 20.3 fb-1
3
ATLAS semi-leptonic
search W (lν)Z (jj)
[arXiv:http://1503.04677],
using similar BDRS-A CA
1.2 reconstruction, in
tail/nothing seen 4
Data / Bkg
Are we alone? ATLAS searches
Events / GeV
Are we alone? ATLAS searches
103
102
10
ATLAS
s = 8 TeV
∫ L dt = 20.3 fb-1
High-p Res. Region
T
Z → ee, µµ Channel
Data
Z+jets
ZZ/ZW/WW
tt+Single Top
Sys+Stat Uncertainty
G*, m=800 GeV
σNominal × 10.0
1
10-1
10-3
Data / BG
10-4
39 / 48
ATLAS resolved Run-2
dijet search
ATLAS-CONF-2015-042,
nothing seen 4
ATLAS semi-leptonic
search W (lν)Z (jj)
[arXiv:http://1503.04677],
using similar BDRS-A CA
1.2 reconstruction, in
tail/nothing seen 4
10-2
1.4
1.2
1
0.8
0.6
0
ATLAS resolved dijet
search [arXiv:1407.1376],
nothing seen 4
500
1000
1500
2000
2500
mlljj [GeV]
Alex Martyniuk
X → VV → JJ
ATLAS semi-leptonic
search W (jj)Z (ll)
[arXiv:1409.6190], using
similar BDRS-A CA 1.2
reconstruction, in
tail/nothing seen 4
Are we alone? ATLAS combination
ATLAS diboson combination ATLAS-CONF-2015-045
In the full combination discrepancy is reduced to a 2.5σ local excess
Limits increased to 1.81TeV in the W 0 search channels
Fully hadronic channel in (significant) tension with the leptonic channels
Pointing to statistical fluctuation in fully hadronic channel?
Run-2 will sort this out!
Full paper following up CONF note in the ATLAS pipeline
40 / 48
Alex Martyniuk
X → VV → JJ
Are we alone? Across the ring....CMS
CMS dijet search
[arXiv:1501.04198], blip in
2-btag at 2 TeV? Trick of
the eye? 4
CMS run-2 dijet search
[CMS-PAS-EXO-15-001],
nothing 4
CMS WR search
[arXiv:1407.3683], excess
at 2 TeV in eejj channel
only 42
CMS WW /WZ /ZZ
semi-leptonic search
[arXiv:1405.3447], in
tails/nothing seen 4
41 / 48
Alex Martyniuk
X → VV → JJ
CMS WW /WZ /ZZ fully
hadronic search
[arXiv:1405.3447], uses
n-subjettiness, broad blip
at ≈ 1.8TeV 42
Are we alone? Across the ring....CMS
dσ / dm jj [pb / GeV]
42 pb-1(13 TeV)
1
CMS
Preliminary
data
10
-1
background fit to data
QCD MC
q* (4.5 TeV)
10 -2
10
-4
10
-5
(Data-Fit)/ σ
10
-3
3
2
1
0
-1
-2
-3
CMS run-2 dijet search
[CMS-PAS-EXO-15-001],
nothing 4
CMS WR search
[arXiv:1407.3683], excess
at 2 TeV in eejj channel
only 42
|η | < 2.5, |Δη | < 1.3
Mjj > 1.1 TeV
Wide Jets
1500 2000 2500 3000 3500 4000 4500 5000 5500
Dijet Mass (GeV)
1500 2000 2500 3000 3500 4000 4500 5000 5500
Dijet Mass [GeV]
41 / 48
CMS dijet search
[arXiv:1501.04198], blip in
2-btag at 2 TeV? Trick of
the eye? 4
Alex Martyniuk
X → VV → JJ
CMS WW /WZ /ZZ
semi-leptonic search
[arXiv:1405.3447], in
tails/nothing seen 4
CMS WW /WZ /ZZ fully
hadronic search
[arXiv:1405.3447], uses
n-subjettiness, broad blip
at ≈ 1.8TeV 42
Are we alone? Across the ring....CMS
CMS dijet search
[arXiv:1501.04198], blip in
2-btag at 2 TeV? Trick of
the eye? 4
CMS run-2 dijet search
[CMS-PAS-EXO-15-001],
nothing 4
CMS WR search
[arXiv:1407.3683], excess
at 2 TeV in eejj channel
only 42
CMS WW /WZ /ZZ
semi-leptonic search
[arXiv:1405.3447], in
tails/nothing seen 4
41 / 48
Alex Martyniuk
X → VV → JJ
CMS WW /WZ /ZZ fully
hadronic search
[arXiv:1405.3447], uses
n-subjettiness, broad blip
at ≈ 1.8TeV 42
Are we alone? Across the ring....CMS
CMS dijet search
[arXiv:1501.04198], blip in
2-btag at 2 TeV? Trick of
the eye? 4
CMS run-2 dijet search
[CMS-PAS-EXO-15-001],
nothing 4
CMS WR search
[arXiv:1407.3683], excess
at 2 TeV in eejj channel
only 42
CMS WW /WZ /ZZ
semi-leptonic search
[arXiv:1405.3447], in
tails/nothing seen 4
41 / 48
Alex Martyniuk
X → VV → JJ
CMS WW /WZ /ZZ fully
hadronic search
[arXiv:1405.3447], uses
n-subjettiness, broad blip
at ≈ 1.8TeV 42
Are we alone? Across the ring....CMS
CMS dijet search
[arXiv:1501.04198], blip in
2-btag at 2 TeV? Trick of
the eye? 4
CMS run-2 dijet search
[CMS-PAS-EXO-15-001],
nothing 4
CMS WR search
[arXiv:1407.3683], excess
at 2 TeV in eejj channel
only 42
CMS WW /WZ /ZZ
semi-leptonic search
[arXiv:1405.3447], in
tails/nothing seen 4
41 / 48
Alex Martyniuk
X → VV → JJ
CMS WW /WZ /ZZ fully
hadronic search
[arXiv:1405.3447], uses
n-subjettiness, broad blip
at ≈ 1.8TeV 42
The Future: Run-2
42 / 48
Alex Martyniuk
X → VV → JJ
The Future Now: Data is here!
Fun fact: The Run-1 paper was submitted around 13 hours before the first collisions of Run-2
43 / 48
Alex Martyniuk
X → VV → JJ
The Future Now: Data is here!
44 / 48
Alex Martyniuk
X → VV → JJ
Events / 100 GeV
How much data is needed?
104
3
So question one is how
much Run-2 13 TeV data
do we need to surpass
the Run-1 result?
Data
Background model
1.5 TeV EGM W', c = 1
2.0 TeV EGM W', c = 1
2.5 TeV EGM W', c = 1
Significance (stat)
Significance (stat + syst)
ATLAS
s = 8 TeV, 20.3 fb-1
10
102
At 2 TeV production
cross-sections grow
considerably
WZ Selection
10
1
Significance
10−1
3
2
1
0
−1
−2
1.5
1.5
2
2
2.5
2.5
3
3
3.5
3.5
mjj [TeV]
45 / 48
Alex Martyniuk
X → VV → JJ
For gluon-gluon fusion
×15, i.e. for GRS
production
For q q̄ initiated ×8, i.e.
for W0 production
Unfortunately QCD
background also
increases
So a smaller amount of
Run-2 data (1 − 3fb−1 )
should be roughly
equivalent to the 8 TeV
dataset
How much data is needed?
WJS2013
100
luminosity ratio
ratios of LHC parton luminosities: 13 TeV / 8 TeV
At 2 TeV production
cross-sections grow
considerably
gg _
Σqq
qg
10
MSTW2008NLO
1
100
1000
Alex Martyniuk
For gluon-gluon fusion
×15, i.e. for GRS
production
For q q̄ initiated ×8, i.e.
for W0 production
Unfortunately QCD
background also
increases
So a smaller amount of
Run-2 data (1 − 3fb−1 )
should be roughly
equivalent to the 8 TeV
dataset
MX (GeV)
45 / 48
So question one is how
much Run-2 13 TeV data
do we need to surpass
the Run-1 result?
X → VV → JJ
How much data is needed?
WJS2013
100
luminosity ratio
ratios of LHC parton luminosities: 13 TeV / 8 TeV
At 2 TeV production
cross-sections grow
considerably
gg _
Σqq
qg
10
MSTW2008NLO
1
100
1000
Alex Martyniuk
For gluon-gluon fusion
×15, i.e. for GRS
production
For q q̄ initiated ×8, i.e.
for W0 production
Unfortunately QCD
background also
increases
So a smaller amount of
Run-2 data (1 − 3fb−1 )
should be roughly
equivalent to the 8 TeV
dataset
MX (GeV)
45 / 48
So question one is how
much Run-2 13 TeV data
do we need to surpass
the Run-1 result?
X → VV → JJ
How much data is needed?
What local σ values can
we see assuming
Different 13 TeV
recorded luminosity
points
Systematics size w.r.t.
Run-1
Injecting a signal with a
cross-section the size
of the discrepancy
seen in Run-1
If realised in nature,
observation should be
possible even with the
reduced 2015 data
expectations
Semi-leptonic channels can control backgrounds more easily (mostly V +jets)
Have made additional cuts and should have an equivalent reach at 2 TeV to the
fully hadronic channel in Run-2!
Hadronic channel will still be competitive at the highest masses in early data
46 / 48
Alex Martyniuk
X → VV → JJ
Ready for data?
New ATLAS data format, xAOD working 2
Physics derivations, DxAOD, running 2
On MC 2
On data 2
CxAOD analysis framework in place,
producing plots 2
ResonanceFinder statistical framework in
place 2
New R2D2 boosted boson tagger defined 2
No really... R = 0.2 subjet, D2β
substructure variable
[arXiv:1409.6298]
Ready, to improve the analysis with new
techniques and data YES! 2
Steady, to push search to higher masses
YES!! 2
47 / 48
Alex Martyniuk
Go, to confirm/disprove excess YES!!! 2
X → VV → JJ
Conclusions
Deviation from expected steeply, smoothly falling
background seen at 2 TeV
Cross-checks performed, no major issues
discovered
Excess p0 = 3.4σ local, 2.5σ global
0
Limits exclude EGM W models with
1.3 < mW 0 < 1.5 TeV
Preparations underway to repeat search in
13TeV data
Events / 100 GeV
s = 8 TeV, 20.3 fb-1
102
WZ Selection
10
1
10−1
3
2
1
0
−1
−2
Alex Martyniuk
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
mjj [TeV]
104
ATLAS
Observed 95% CL
s = 8 TeV, 20.3 fb-1
Expected 95% CL
103
± 1σ uncertainty
± 2σ unceirtainty
2
10
EGM W', c = 1
10
1
10−1 1.4
48 / 48
Data
Background model
1.5 TeV EGM W', c = 1
2.0 TeV EGM W', c = 1
2.5 TeV EGM W', c = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
Significance
Used 20.3fb−1 8TeV ATLAS data
Jet substructure (BDRS-A CA 1.2 jets) used to
separate signal from background
QCD dominated background modelled by
parametric function
σ(pp → W') × BR(W' → WZ) [fb]
Presented a search for a high mass diboson
resonance, [arXiv:1506.00962]
104
X → VV → JJ
1.6
1.8
2
2.2
2.4
2.6
2.8
3
mW' [TeV]
Backup
Backup
48 / 48
Alex Martyniuk
X → VV → JJ
Mass window optimisation
A data CR was formed from two tagged/untagged samples
A wide mass window (40 < mJ < 400 GeV) used
CR A: Leading jet tagged, sub-leading fails tag
CR B: Leading jet fails tag, sub-leading passes tag
1
2
Forms dijet sample by taking one jet from CR A and one from CR B
Use this as a high stats QCD region for comparison with signal MC
ATLAS Simulation
s = 8 TeV , 20.3fb-1
For W'(1.4 TeV)
Zllr
εs
εb
2
2.5
1
1
0.5
0.5
48 / 48
For W'(1.8 TeV)
Zllr
εs
εb
1.5
1.5
0
ATLAS Simulation
s = 8 TeV , 20.3fb-1
2
0
5
10
15
20
half window [GeV]
0
0
5
Alex Martyniuk
10
15
20
half window [GeV]
X → VV → JJ
Zllr, εs, εb
2.5
Zllr, εs, εb
Zllr, εs, εb
Compute mass window efficiencies as function of window width for different W 0
masses
2.2
2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
ATLAS Simulation
s = 8 TeV , 20.3fb-1
0
5
For W'(2.2 TeV)
Zllr
εs
εb
10
15
20
half window [GeV]
Jets / 5 GeV
ntrk optimisation
2000
ATLAS
s = 8 TeV, 20.3 fb-1
1800
1600
1400
1200
ntrk is poorly modelled in MC
1000
800
Cannot trust it to optimise cut or measure its
efficiency
600
400
200
Select V +jets enriched data sample
2
Optimise the cut by fitting QCD background vs
selected signal peak
Measure the efficiency of selected cut in data
sample
Two models used to fit background in this region
Both fit well, efficiency error taken from a
combined PDF of both fits
48 / 48
Alex Martyniuk
Jets / 5 GeV
1
0
2200
2000
1800
1600
1400
1200
1000
ntrk < 30, Fourth deg. polynomial
60
80
100
120
140
160
180
200
m [GeV]
ATLAS
s = 8 TeV, 20.3 fb-1
800
600 n < 30, Third deg. polynomial
trk
400
200
0
40 50 60 70 80 90 100 110 120 130 140
m [GeV]
X → VV → JJ
ATLAS Simulation
s = 8 TeV , 20.3fb-1
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
14
12
W'(1.4 TeV) → WZ → qqqq
10
0
0
16
14
ATLAS Simulation
s = 8 TeV , 20.3fb-1
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
W'(1.4 TeV) → WZ → qqqq
2.5
1.5
6
4
Nominal
JMS up
JMS down
JMR
20
2
40
60
80
100
ATLAS Simulation
s = 8 TeV , 20.3fb-1
0
0
120
140
160
max(m , m2) [GeV]
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
W'(1.8 TeV) → WZ → qqqq
2
1
Nominal
JMS up
JMS down
JMR
20
0.5
40
60
80
100
0
0
0.9
0.8
0.7
0.6
ATLAS Simulation
s = 8 TeV , 20.3fb-1
W'(2.2 TeV) → WZ → qqqq
20
0.2
0.1
40
60
80
100
120
140
160
min(m , m2) [GeV]
20
40
60
80
100
0
0
120
140
160
max(m , m2) [GeV]
1
0.8
0.7
0.6
ATLAS Simulation
s = 8 TeV , 20.3fb-1
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
W'(2.2 TeV) → WZ → qqqq
0.5
0.4
0.3
0.2
Nominal
JMS up
JMS down
JMR
20
0.1
40
1
48 / 48
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
0.3
Nominal
JMS up
JMS down
JMR
Nominal
JMS up
JMS down
JMR
1
0.4
1
0
0
120
140
160
min(m , m2) [GeV]
0.5
1.5
0.5
BDRS-A
|∆yjj | < 1.2
|η| < 2.0
A < 0.15
yf >= 0.45
ntrk < 30
W'(1.8 TeV) → WZ → qqqq
8
Jets / 5 GeV
Jets / 5 GeV
3
ATLAS Simulation
s = 8 TeV , 20.3fb-1
3
2
1
3.5
4
3.5
2.5
10
6
2
18
12
8
4
20
Jets / 5 GeV
16
Jets / 5 GeV
18
Jets / 5 GeV
Jets / 5 GeV
Systematic variations
60
80
100
120
140
160
max(m , m2) [GeV]
1
Alex Martyniuk
X → VV → JJ
0
0
Nominal
JMS up
JMS down
JMR
20
40
60
80
100
120
140
160
min(m , m2) [GeV]
1
Selection overlaps
ATLAS
3
10
Data
s = 8 TeV, 20.3 fb-1
Background model
Significance (stat)
102
Significance (stat + syst)
WW+ZZ+WZ Selection
10
104
ATLAS
3
10
Events / 100 GeV
104
Events / 100 GeV
Events / 100 GeV
Comparing the effect of applying all signal selections at once, and the same
flavour selections at once
Data
s = 8 TeV, 20.3 fb-1
Background model
Significance (stat)
102
Significance (stat + syst)
104
Data
Background model
1.5 TeV EGM W', c = 1
2.0 TeV EGM W', c = 1
2.5 TeV EGM W', c = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
s = 8 TeV, 20.3 fb-1
10
102
WW+ZZ Selection
10
WZ Selection
10
1
1
10−1
10−1
1
10−1
10−2
1.5
2
2.5
3
1.5
2
2.5
3
3.5
4
2
0
−2
3.5
1.5
2
2.5
3
1.5
2
2.5
3
Data
Background model
1.5 TeV Bulk GRS, k/MPI = 1
2.0 TeV Bulk GRS, k/MPI = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
s = 8 TeV, 20.3 fb-1
102
10
3.5
2.5
3
1.5
2
2.5
3
Data
Background model
1.5 TeV Bulk GRS, k/MPI = 1
2.0 TeV Bulk GRS, k/MPI = 1
Significance (stat)
Significance (stat + syst)
ATLAS
3
10
s = 8 TeV, 20.3 fb-1
102
10
ZZ Selection
1
10−1
10−1
10−2
10−2
−3
−3
10
1.5
2
2.5
3
1.5
2
2.5
3
3.5
3.5
Significance
10
Significance
2
3
2
1
0
−1
−2
1.5
2
2.5
3
1.5
2
2.5
3
mjj [TeV]
Alex Martyniuk
3.5
3.5
mjj [TeV]
WW Selection
48 / 48
1.5
104
1
3
2
1
0
−1
−2
3
2
1
0
−1
−2
mjj [TeV]
104
Events / 100 GeV
Events / 100 GeV
mjj [TeV]
3.5
Significance
4
2
0
−2
Significance
Significance
10−2
3.5
3.5
mjj [TeV]
X → VV → JJ
Signal properties
The resonance width (Γ) and the product of cross sections and branching ratios
(BR) to four-quark final states used in modelling W ? → WZ , GRS → WW , and
GRS → ZZ , for several values of resonance pole masses (m). The fraction of
events in which the invariant mass of the W ? or GRS decay products lies within
10% of the nominal resonance mass (f10% ) is also displayed.
m
[TeV]
1.3
1.6
2.0
2.5
3.0
48 / 48
ΓW 0
[GeV]
47
58
72
91
109
ΓGRS
[GeV]
76
96
123
155
187
W0 → WZ
σ ×BR f10%
[fb]
19.1
0.83
6.04
0.79
1.50
0.72
0.31
0.54
0.088 0.31
Alex Martyniuk
GRS → W W
σ ×BR f10%
[fb]
0.73
0.85
0.14
0.83
0.022
0.83
0.0025
0.78
0.00034 0.72
X → VV → JJ
GRS → ZZ
σ ×BR f10%
[fb]
0.37
0.84
0.071
0.84
0.010
0.82
0.0011
0.78
0.00017 0.71
Observed and expected numbers
Number of events observed in the WZ, WW, and ZZ selected samples in each
dijet mass bin used in the analysis, compared to the prediction of the
background-only fit.
mjj bin [GeV]
1050–1150
1150–1250
1250–1350
1350–1450
1450–1550
1550–1650
1650–1750
1750–1850
1850–1950
1950–2050
2050–2150
2150–2250
2250–2350
2350–2450
2450–2550
2550–2650
2650–2750
2750–2850
2850–2950
2950–3050
3050–3150
3150–3250
3250–3350
3350–3450
3450–3550
48 / 48
obs
299
149
60
38
16
15
6
5
5
8
2
0
1
0
0
0
0
0
0
0
0
0
0
0
0
W Z selection (Events)
exp
+1σ
−1σ
296.65
312.60
280.90
140.82
146.83
134.68
71.37
75.30
67.21
38.24
40.98
35.40
21.50
23.36
19.64
12.61
13.88
11.39
7.67
8.56
6.86
4.82
5.47
4.25
3.12
3.62
2.70
2.08
2.47
1.75
1.41
1.74
1.15
0.98
1.26
0.77
0.70
0.94
0.52
0.50
0.71
0.35
0.37
0.55
0.24
0.28
0.44
0.17
0.21
0.36
0.12
0.16
0.29
0.08
0.13
0.25
0.06
0.10
0.21
0.04
0.08
0.18
0.03
0.06
0.16
0.02
0.05
0.15
0.02
0.04
0.13
0.01
0.04
0.12
0.01
obs
203
97
47
37
13
8
4
2
4
7
2
0
0
0
0
0
0
0
0
0
0
0
0
1
0
Alex Martyniuk
W W selection (Events)
exp
+1σ
−1σ
202.39
215.30
189.33
97.98
102.93
92.97
50.79
54.10
47.42
27.91
30.28
25.52
16.13
17.78
14.49
9.75
10.87
8.62
6.12
6.90
5.35
3.98
4.54
3.43
2.67
3.09
2.27
1.84
2.18
1.53
1.30
1.58
1.06
0.95
1.18
0.74
0.70
0.91
0.52
0.53
0.71
0.37
0.41
0.58
0.27
0.32
0.48
0.20
0.25
0.40
0.15
0.21
0.35
0.11
0.17
0.30
0.08
0.14
0.27
0.06
0.12
0.25
0.05
0.10
0.23
0.03
0.09
0.22
0.03
0.08
0.21
0.02
0.07
0.20
0.01
X → VV → JJ
obs
169
86
30
20
10
8
0
1
5
3
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
ZZ selection (Events)
exp
+1σ
−1σ
180.83
194.82
166.70
78.04
82.65
73.32
36.01
39.53
32.51
17.59
20.02
15.26
9.02
10.55
7.58
4.83
5.76
3.96
2.68
3.25
2.16
1.54
1.90
1.22
0.91
1.14
0.71
0.55
0.71
0.42
0.34
0.46
0.25
0.22
0.30
0.15
0.14
0.20
0.09
0.09
0.14
0.06
0.06
0.10
0.03
0.04
0.07
0.02
0.03
0.06
0.01
0.02
0.04
0.01
0.01
0.03
0.00
0.01
0.03
0.00
0.01
0.02
0.00
0.01
0.02
0.00
0.00
0.01
0.00
0.00
0.01
0.00
0.00
0.01
0.00
Observed and expected limits
Observed and expected limits on the EGM W ? models in the WZ selection
m [GeV]
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
2600
2700
2800
2900
3000
48 / 48
W0 → WZ
σBRW Z [fb]
40.5
27.3
18.7
12.8
8.93
6.23
4.46
3.14
2.28
1.64
1.21
0.889
0.666
0.502
0.383
0.297
0.236
0.186
Γ [GeV]
46.66
50.30
53.99
57.60
61.32
65.00
68.66
72.30
76.00
79.60
83.34
87.00
90.68
94.30
98.03
101.70
105.40
109.00
obs
23.47
22.08
15.79
18.82
14.40
30.47
34.96
36.04
37.12
20.92
8.87
7.49
6.75
6.40
7.28
8.08
8.91
10.58
Alex Martyniuk
exp
42.43
35.01
26.28
19.84
16.19
15.22
13.35
12.18
10.74
9.88
9.72
9.61
9.37
9.69
10.88
11.26
12.32
13.54
95% CL Limits [fb]
−2σ
−1σ
+1σ
20.51
27.61
69.49
16.96
23.41
58.15
12.82
16.81
41.96
10.97
13.99
30.62
8.55
11.18
25.36
7.95
10.67
23.64
6.69
8.96
20.40
6.12
8.39
18.71
6.02
7.31
16.77
5.36
6.73
15.40
5.24
6.70
14.89
5.40
6.68
15.05
5.47
6.76
14.83
5.25
6.80
15.22
6.24
7.49
16.53
6.96
8.24
17.88
7.27
8.86
18.61
8.86
10.12
20.71
X → VV → JJ
+2σ
120.84
96.95
70.12
50.61
42.78
38.51
33.37
30.84
27.06
25.36
23.30
23.37
23.97
23.60
26.11
27.75
29.37
32.11
Observed and expected limits
Observed and expected limits on the bulk GRS models in the WW selection
m [GeV]
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
2600
2700
2800
2900
3000
48 / 48
GRS → W W
σBRW W [fb]
1.59
0.908
0.532
0.317
0.192
0.119
0.0744
0.0470
0.0300
0.0194
0.0136
0.0083
0.0055
0.0036
0.0024
0.0016
0.0011
0.0008
Γ [GeV]
76.0
82.8
89.5
96.2
103
109
116
123
129
136
142
149
155
161
168
174
181
187
obs
59.16
59.00
27.57
16.53
12.47
18.18
29.01
30.23
47.39
13.70
8.48
6.76
6.39
6.21
6.41
6.62
6.87
8.24
Alex Martyniuk
exp
53.46
40.90
32.60
26.04
19.92
17.83
15.99
14.68
13.83
12.34
12.63
12.06
11.32
10.87
11.07
10.84
10.93
10.76
95% CL Limits [fb]
−2σ
−1σ
+1σ
29.46
37.36
80.36
23.03
28.67
62.92
15.39
21.85
49.69
12.63
17.18
41.93
9.44
13.56
31.84
9.15
12.96
27.57
7.68
10.97
24.93
7.87
10.18
23.25
7.72
9.46
21.67
7.02
8.53
19.19
6.67
9.01
19.53
5.78
7.86
18.78
5.68
7.36
17.92
5.64
7.19
16.99
5.58
7.10
16.94
5.41
7.41
16.72
6.02
7.50
17.07
6.61
8.25
16.39
X → VV → JJ
+2σ
139.90
109.16
81.17
64.89
50.82
44.35
40.50
37.90
35.14
29.46
31.31
30.54
28.18
26.76
26.91
26.60
26.76
25.85
Observed and expected limits
Observed and expected limits on the bulk GRS models in the ZZ selection
m [GeV]
1300
1400
1500
1600
1700
1800
1900
2000
2100
2200
2300
2400
2500
2600
2700
2800
2900
3000
48 / 48
GRS → ZZ
σBRZZ [fb]
0.753
0.415
0.245
0.144
0.0888
0.0557
0.0338
0.0213
0.0137
0.0088
0.0058
0.0037
0.0025
0.0019
0.0011
0.0008
0.0005
0.0003
Γ [GeV]
76.0
82.8
89.5
96.2
103
109
116
123
129
136
142
149
155
161
168
174
181
187
obs
29.55
27.71
24.30
18.39
10.36
29.71
29.24
29.73
26.23
13.27
7.50
6.37
6.61
7.27
6.24
6.34
6.62
7.31
Alex Martyniuk
exp
50.91
34.78
25.93
21.13
17.07
12.98
10.98
10.26
9.25
7.84
8.55
8.02
8.30
8.85
7.53
7.41
7.66
8.05
95% CL Limits [fb]
−2σ
−1σ
+1σ
25.19
33.72
81.26
18.59
24.56
56.97
14.29
18.15
41.45
11.35
14.58
33.67
8.46
11.42
26.44
7.74
8.78
20.04
6.28
7.83
16.94
6.07
7.72
15.68
5.93
6.87
13.70
5.41
6.09
11.76
5.51
6.30
13.15
4.91
6.01
12.16
4.96
6.15
12.57
5.64
6.73
13.23
5.00
5.88
11.26
4.96
5.75
11.00
5.28
6.22
11.27
5.69
6.42
11.52
X → VV → JJ
+2σ
152.30
93.29
67.91
53.01
44.35
34.98
28.79
24.91
21.35
18.20
20.57
18.93
19.61
20.72
17.05
16.88
17.50
18.21