Jets and jet substructure 4 - Physics Learning LaboratoriesGavin Salam (CERN) Jets and jet...
Transcript of Jets and jet substructure 4 - Physics Learning LaboratoriesGavin Salam (CERN) Jets and jet...
Jets and jet substructure 4:
substructure
1
Gavin Salam (CERN)with extensive use of material by Matteo Cacciari
and Gregory Soyez
TASIJune 2013
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Two things that make jets@LHC special
The large hierarchy of scales√s ⨠ MEW
The huge pileupnpileup ~ 20 – 40
[These involve two opposite extremes: low pt and high pt,which nevertheless talk to each other]
2
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 3
E.g. X ! tt̄ resonances of varying di!culty[1 jet ! 2 partons]
RS KK resonances ! tt̄, from Frederix & Maltoni, 0712.2355
NB: QCD dijet spectrum is " 103 times tt̄Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 10 / 29
e.g. ttbar resonances
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Boosted massive particles ! fat jets
Normal analyses: two quarks fromX ! qq̄ reconstructed as two jets
jet 1
jet 2
X at restX
High-pt regime: EW object Xis boosted, decay is collimated,
qq̄ both in same jet
singlefat jet
z
(1−z)
boosted X
Happens for pt ! 2m/R
pt ! 320 GeV for m = mW , R = 0.5
Gavin Salam (CERN/LPTHE/Princeton) Jets in Higgs Searches HC2012 2012-11-18 19 / 29
4
Boosted EW scale objects
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Very active research field
5
Some taggers and jet-substructure observables
Jet Declustering
Jet Shapes
Matrix−Element
Seymour93
YSplitter
Mass−Drop+Filter
JHTopTagger TW
CMSTopTagger
N−subjettiness (TvT)
CoM N−subjettiness (Kim)
N−jettiness
HEPTopTagger(+ dipolarity)
Trimming
Pruning
Planar Flow
Twist
ATLASTopTagger
Templates
Shower Deconstruction
Qjets
Multi−variate tagger
ACF
apologies for omitted taggers, arguable links, etc.
Gavin Salam (CERN/Princeton/CNRS) Boost Theory Summary Boost 2012-07-27 6 / 33
Some of the tools developedfor boosted W/Z/H/top
reconstruction
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Papers on jet substructure
6
More than 100 papers since 2008
(+ some background noise)
Number of papers containing the words ‘jet substructure’
Mike Se
ymou
r
Butte
rwort
h, Cox
, Fors
haw
Pioneered by M. Seymour in the early ‘90s
0
1
2
3
4
5
6
7
1990 1995 2000 2005 2010 2015
pape
rs /
mon
th
year
Papers containing "jet substructure"+ pioneering papers by Mike Seymour in 1991 and 1994(Source: INSPIRE)
Exploded around 2008
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Extensive experimental work
7
From a list compiledfor a recent workshopat Perimeter Institute
Many more analyses in the pipeline
ATLAS Public Results• Large-R, groomed jets with pile-up• Large-R jets with substructure• Quark/gluon jets (see also this link)• Jet substructure at LHC7• Jet properties for boosted searches
Resonance searches• Boosted top (hadronic)• Boosted top (semileptonic)• Three-jet resonance (gluino RPV)• Two-jet resonance (sgluon)
CMS Public Results• Jet substructure in CMS• Subjet multiplicity• Jet mass and grooming
Resonance searches:• Boosted top (hadronic)• Boosted top (semileptonic)• Boosted W/Z
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Jet masses
8
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0 50 100 150 200
1/N
dN
/dm
jet [
GeV
-1]
mjet [GeV]
Wj events
pt,jets > 700 GeV
anti-kt, R = 0.7
pp
7 T
eV
, Pyth
ia 6
.4, n
oU
E
Pythia, underlying event switched off
Look at jet mass distribution for 2 leading jets in
‣ pp → W+jet events
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Problem #1a: QCD jets have masses too
9
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0 50 100 150 200
1/N
dN
/dm
jet [
GeV
-1]
mjet [GeV]
qq ! qq events
pt,jets > 700 GeV
anti-kt, R = 0.7
pp
7 T
eV
, Pyth
ia 6
.4, n
oU
E
Look at jet mass distribution for 2 leading jets in
‣ pp → W+jet events‣ qq → qq events
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Problem #1b: there are lots of QCD jets
10
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0 50 100 150 200
1/N
dN
/dm
jet [
GeV
-1]
mjet [GeV]
qq ! qq + Wj mixture
pt,jets > 700 GeV
anti-kt, R = 0.7
pp
7 T
eV
, Pyth
ia 6
.4, n
oU
E
Look at jet mass distribution for 2 leading jets in
‣ pp → W+jet events‣ qq → qq events‣ mixture of the two,
in rough proportion
Jet mass gives clear sign of massive particles inside the jet; but QCD jets are massive too —
must learn to reject them
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Problem #2: jet mass v. sensitive to PU
11
m/N
dN
/dm
jet
mjet [GeV]
Wj eventspt,jets > 700 GeV
anti-kt, R = 0.7
pp
7 T
eV
, Pyth
ia 6
.4, n
oU
E
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0 50 100 150 200
+ !10" pileup
Jet mass is extremely sensitive to pileup
→ loss of mass resolution
→ loss of ability to find signals
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Tagging & Grooming
12
Tagging• reduces the background, leaves much of signal
Grooming• improves signal mass resolution (removing pileup,
etc.), without significantly changing background & signal event numbers
Two widely used termsthough there’s not a
consensus aboutwhat they mean
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 13
One core idea for tagging
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Inside the jet mass
14
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq!qq, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
Inside the jet mass[1 jet ! 2 partons]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq→qq, no UE
anti-kt, R=0.7LHC, 7 TeV
pt,jets > 700 GeV
QCD jet mass distribution has theapproximate
dN
d lnm! !s ln
ptR
m" Sudakov
Work from ’80s and ’90s
+ Almeida et al ’08
The logarithm comes from integralover soft divergence of QCD:
! 12
m2
p2t R2
dz
z
A hard cut on z reduces QCD back-ground & simplifies its shape
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 15 / 29
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Inside the jet mass
14
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq!qq, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
10 100
mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
Inside the jet mass[1 jet ! 2 partons]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq→qq, no UE
anti-kt, R=0.7LHC, 7 TeV
pt,jets > 700 GeV
QCD jet mass distribution has theapproximate
dN
d lnm! !s ln
ptR
m" Sudakov
Work from ’80s and ’90s
+ Almeida et al ’08
The logarithm comes from integralover soft divergence of QCD:
! 12
m2
p2t R2
dz
z
A hard cut on z reduces QCD back-ground & simplifies its shape
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 15 / 29
Inside the jet mass[1 jet ! 2 partons]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq→qq, no UE
anti-kt, R=0.7LHC, 7 TeV
pt,jets > 700 GeV
10 100mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
QCD jet mass distribution has theapproximate
dN
d lnm! !s ln
ptR
m" Sudakov
Work from ’80s and ’90s
+ Almeida et al ’08
The logarithm comes from integralover soft divergence of QCD:
! 12
m2
p2t R2
dz
z
A hard cut on z reduces QCD back-ground & simplifies its shape
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 15 / 29
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Inside the jet mass
15
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq!qq, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
after cut on z > 0.25
(a la BERS)
10 100
mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
cut on z
keep
reject
Inside the jet mass[1 jet ! 2 partons]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq→qq, no UE
anti-kt, R=0.7LHC, 7 TeV
pt,jets > 700 GeV
QCD jet mass distribution has theapproximate
dN
d lnm! !s ln
ptR
m" Sudakov
Work from ’80s and ’90s
+ Almeida et al ’08
The logarithm comes from integralover soft divergence of QCD:
! 12
m2
p2t R2
dz
z
A hard cut on z reduces QCD back-ground & simplifies its shape
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 15 / 29
Inside the jet mass[1 jet ! 2 partons]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq→qq, no UE
anti-kt, R=0.7LHC, 7 TeV
pt,jets > 700 GeV
10 100mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
QCD jet mass distribution has theapproximate
dN
d lnm! !s ln
ptR
m" Sudakov
Work from ’80s and ’90s
+ Almeida et al ’08
The logarithm comes from integralover soft divergence of QCD:
! 12
m2
p2t R2
dz
z
A hard cut on z reduces QCD back-ground & simplifies its shape
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 15 / 29
Inside the jet mass[1 jet ! 2 partons]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq→qq, no UE
anti-kt, R=0.7LHC, 7 TeV
pt,jets > 700 GeV
after cut on z > 0.25
(a la BERS)
10 100mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
cut on z
keep
reject
QCD jet mass distribution has theapproximate
dN
d lnm! !s ln
ptR
m" Sudakov
Work from ’80s and ’90s
+ Almeida et al ’08
The logarithm comes from integralover soft divergence of QCD:
! 12
m2
p2t R2
dz
z
A hard cut on z reduces QCD back-ground & simplifies its shape
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 15 / 29
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Inside the jet mass
16
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq!qq, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
after cut on z > 0.25
(a la BERS)
10 100
mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
cut on z
keep
reject
0
0.5
1
1.5
2
2.5
3
3.5
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
W+jet Jet Mass distributionPythia 6.4, pp!Wj, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
10 100
mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Inside the jet mass
17
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
QCD Jet Mass distributionPythia 6.4, qq!qq, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
after cut on z > 0.25
(a la BERS)
10 100
mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
cut on z
keep
reject
0
0.5
1
1.5
2
2.5
3
3.5
10 100
1/N
dN
/dlo
g(m
jet)
mjet [GeV]
W+jet Jet Mass distributionPythia 6.4, pp!Wj, no UE
anti-kt, R=0.7
LHC, 7 TeV
pt,jets > 700 GeV
after cut on z > 0.25
10 100
mjet [GeV]
0.01
0.1
1
z
2-bodyphasespace
cut on z
keep
reject
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 18
m/N
dN
/dm
jet
mjet [GeV]
qq ! qq + Wj mixture
pt,jets > 700 GeV
anti-kt, R = 0.7
0
0.05
0.1
0.15
0.2
0.25
0 50 100 150 200
+ CA subjet (z > 0.25)
Signal + bkgd after cut on z
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 19
One core idea for grooming
[see blackboard]
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
“Grooming”
20
Noise removal from jets[1 jet ! 2 partons]
[Some of the other ideas]
[Boost 2010
writeup]
! Filtering Butterworth et al ’08
! Pruning Ellis, Vermillion and Walsh ’09
! Trimming Krohn, Thaler & Wang ’09
[With earlier methods by Seymour ’93 and Kodolova et al ’07;
Rubin ’10 for filtering optimisation; also Soper & Spannowsky ’10, ’11]
Jets 2 (M. Cacciari and G. Salam) GGI September 2011 24 / 29
Plain jet mass(anti-kt)
Groomed jetmasses
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 21
How do the tools workin practice?
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 3.57137e−05
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.000496598
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.000688842
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.000805103
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.000773759
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.0014577
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is diB = 0.00147749
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is diB = 1.96
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out anti-kt
anti-kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Anti-kt gradually makes its way
through the secondary blob ! no
clear identification of substructure
associated with 2nd parton.
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.318802
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 0.977453
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 1.48276
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 2.34277
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 13.5981
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 30.8068
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is dij = 717.825
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Its last step is to merge two hard
pieces. Easily undone to identify un-
derlying kinematics
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Its last step is to merge two hard
pieces. Easily undone to identify un-
derlying kinematics
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
dmin is diB = 11432
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Its last step is to merge two hard
pieces. Easily undone to identify un-
derlying kinematics
Identifying jet substructure: try out kt
kt algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
This is crucial for identifying thekinematic variables of the partons inthe jet (e.g. z).
kt clusters soft “junk” early on in theclustering
Its last step is to merge two hard
pieces. Easily undone to identify un-
derlying kinematics
This meant it was the first algorithmto be used for jet substructure.
Seymour ’93
Butterworth, Cox & Forshaw ’02
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} = 0.142857
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} = 0.214286
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} = 0.415037
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} = 0.686928
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} = 1.20645
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} = 1.93202
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem, and then adds in remainingsoft junk
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem, and then adds in remainingsoft junk
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} > 2
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem, and then adds in remainingsoft junk
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem, and then adds in remainingsoft junk
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
DeltaR_{ij} > 2
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem, and then adds in remainingsoft junk
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Identifying jet substructure: Cam/Aachen
Cambridge/Aachen algorithm
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
How well can an algorithm identifythe “blobs” of energy inside a jet thatcome from di!erent partons?
C/A identifies two hard blobs withlimited soft contamination, joinsthem, and then adds in remainingsoft junk
The interesting substructure is buriedinside the clustering sequence — it’sless contamined by soft junk, butneeds to be pulled out with specialtechniques
Butterworth, Davison, Rubin & GPS ’08Kaplan, Schwartz, Reherman & Tweedie ’08
Butterworth, Ellis, Rubin & GPS ’09Ellis, Vermilion & Walsh ’09
G. Salam (CERN/Princeton/Paris) test-bed., G. Salam June 13, 2013 1 / 1
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
73
anti-kt algorithmpt/GeV
50
40
20
00 1 2 3 4 y
30
10
kt algorithmpt/GeV
50
40
20
00 1 2 3 4 y
30
10
Cambridge/Aachen
pt/GeV
50
40
20
00 1 2 3 4 y
30
10
Jets lecture 3 (Gavin Salam) CERN Academic Training March/April 2011 19 / 29
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
Cluster event, C/A, R=1.2
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
Zbb BACKGROUND
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
Fill it in, ! show jets more clearly
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
Zbb BACKGROUND
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
Consider hardest jet, m = 150 GeV
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
0
0.05
0.1
0.15
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
Zbb BACKGROUND
0
0.002
0.004
0.006
0.008
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
split: m = 150 GeV, max(m1,m2)m
= 0.92 ! repeat
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
0
0.05
0.1
0.15
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
Zbb BACKGROUND
0
0.002
0.004
0.006
0.008
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
split: m = 139 GeV, max(m1,m2)m
= 0.37 ! mass drop
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
0
0.05
0.1
0.15
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
Zbb BACKGROUND
0
0.002
0.004
0.006
0.008
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
check: y12 "pt2pt1
" 0.7 ! OK + 2 b-tags (anti-QCD)
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
0
0.05
0.1
0.15
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
Zbb BACKGROUND
0
0.002
0.004
0.006
0.008
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
Rfilt = 0.3
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
0
0.05
0.1
0.15
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
Zbb BACKGROUND
0
0.002
0.004
0.006
0.008
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
arbitrary norm.
pp ! ZH ! !!̄bb̄, @14TeV, mH=115GeV
Herwig 6.510 + Jimmy 4.31 + FastJet 2.3
Rfilt = 0.3: take 3 hardest, m = 117 GeV
Butterworth, Davison, Rubin & GPS ’08
SIGNAL
0
0.05
0.1
0.15
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
Zbb BACKGROUND
0
0.002
0.004
0.006
0.008
80 100 120 140 160mH [GeV]
200 < ptZ < 250 GeV
arbitrary norm.
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Boosted Higgs analysis
82
Cluster with a large RUndo the clustering into subjets,
until a large mass drop is observed
Re-cluster with smaller R, and keep only 3
hardest jets
pp →ZH → ννbb--
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 83
#include “fastjet/tools/MassDropTagger.hh”#include “fastjet/tools/Filter.hh”
JetDefinition jet_def(cambridge_algorithm, 1.2);ClusterSequence cs(input_particles, jet_def);jets = sorted_by_pt(cs.inclusive_jets());
// define the tagger and use itdouble mu = 0.667, ycut = 0.09;MassDropTagger md_tagger(mu, ycut);PseudoJet tagged = md_tagger(jets[0]);// check it was tagged OK by verifying (tagged != 0)
// define the filter and use itFilter filter(0.3,SelectorNHardest(3));PseudoJet filtered = filter(tagged); // this is the Higgs!!
Mass-Drop Tagger + Filtering in FastJet
The real analysis is slightly more refined (b-tagging, dynamical filter radius, etc) but the main features are already present here
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Pruning
84
0
0.05
0.1
0.15
0.2
0.25
0.3
0.001 0.01 0.1 1
10 100 1000
m/σ
dσ
/ dm
m/pt
Monte Carlo
m [GeV], for pt = 4 TeV
mMDT, ycut = 0.03mMDT, ycut = 0.13
0
0.05
0.1
0.15
0.2
0.25
0.3
0.001 0.01 0.1 1
10 100 1000
m/σ
dσ
/ dm
m/pt
Analytic Calculation
m [GeV], for pt = 4 TeV
mMDT, ycut = 0.03mMDT, ycut = 0.13mMDT, ycut = 0.13 some finite ycut
Figure 4: Comparison of Monte Carlo and analytic results for the mMDT tagged mass spectrumfor two values of ycut. In all cases we have used µ = 0.67.
We compare our analytical predictions with the Monte Carlo simulation results in
Fig. 4 for several ycut values. There is acceptable agreement — remaining di!erences
are attributable to finite ycut e!ects and di!erences beyond single-log accuracy between
our treatment here and that in the Monte Carlo (e.g. subleading terms in the running
coupling). Furthermore, the pattern of ycut dependence, with smaller values of ycut leading
to an increase in the mass spectrum at moderate ! values and a steeper fall-o! towards
lower ! values, in line with the pattern visible in the second-order expansion of Eq. (5.4).
6. Pruning
Pruning [7, 8] takes an initial jet, and from its mass deduces a pruning radius Rprune =
Rfact · 2mpt
, where Rfact is a parameter of the tagger. It then reclusters the jet and for
every clustering step, involving objects a and b, it checks whether "ab > Rprune and
min(pta, ptb) < zcutpt,(a+b), where zcut is a second parameter of the tagger. If so, then the
softer of the a and b is discarded. Otherwise a and b are recombined as usual. Clustering
then proceeds with the remaining objects, applying the pruning check at each stage.
In analysing pruning, we will take Rfact =12 , i.e. its default suggested value. In analogy
with our approach for the ycut parameter in the (m)MDT, we will work in the limit of small
zcut (but ln zcut not too large). And we will assume that the reclustering is performed with
the Cambridge/Aachen algorithm, the most common choice.
At leading order, i.e. a jet involving a single 1 ! 2 splitting, then Rprune = mpt
=
"ab
!
z(1" z), which guarantees that "ab is always larger than Rprune. To establish the
pruned jet mass, one then needs to examine the second part of the pruning condition: if
min(z, 1 " z) > zcut then the clustering is accepted and the pruned jet has a finite mass.
Otherwise the pruned jet mass is zero. This pattern is true independently of the angle
between the two prongs. Thus, as for the MD tagger, pruning removes the divergent low-z
region from the jet-mass phasespace. As a result, in the small zcut limit, one obtains the
same LO result for the pruned mass distribution as for the MD tagger case (replacing
– 13 –
#include “fastjet/tools/Pruner.hh”
// define prunerdouble zcut = 0.1, Rfact = 0.5;Pruner pruner(cambridge_algorithm, zcut, Rfact);
PseudoJet pruned_jet = pruner(jet);
Ellis, Vermilion and Walsh ’09
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Trimming
85
Different condition for retaining jets (pT-cut rather than nfilt hardest)
with respect to filtering, but otherwise identical
#include “fastjet/tools/Filter.hh”
// define trimmerFilter trimmer(0.3,SelectorPtFractionMin(0.03));
Krohn, Thaler & Wang ’09
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
different (2-body) substructure tools
86
Gro
omin
g
Tagging
jet mass
N-subjettiness
filteringtrimming
pruning
MDT
Detailed relative positions depend on physics context(and are possibly contentious!)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
different (2-body) substructure tools
86
Gro
omin
g
Tagging
jet mass
N-subjettiness
filteringtrimming
pruning
MDT
combined
Detailed relative positions depend on physics context(and are possibly contentious!)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
different (2-body) substructure tools
86
Gro
omin
g
Tagging
jet mass
N-subjettiness
filteringtrimming
pruning
MDT
combined
multivariatetaggers
Detailed relative positions depend on physics context(and are possibly contentious!)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
different (2-body) substructure tools
86
Gro
omin
g
Tagging
jet mass
N-subjettiness
filteringtrimming
pruning
MDT
combined
multivariatetaggers
Detailed relative positions depend on physics context(and are possibly contentious!)
template tagger
jet deconstruction
Q-jets
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
3rd party tools for FastJet
87
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Comparing top taggers
88
Comparing top taggers: QCD fakes rate v. signal e!.[Methods]
Herwig, 500 < pt < 600 GeV Herwig++, 200 < pt < 800 GeV
From the extensive “Boost 2011” report, which reviewed taggersdiscussed software, determined performance on MC, etc.
Bottom line: some taggers clearly better than others.But many taggers behave similarly & details depend on analysis
(+ MC choice)
Gavin Salam (CERN/Princeton/CNRS) Theory of Fat Jets Higgs Hunting 2012-07-19 16 / 28
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Boosted Ws and tops in single jets: data!
W’s in a single jet
with Pruning + Mass Drop requirement
NB: combined in IR unsafe way. . .
tops in a single jet
with HEPTopTaggerGavin Salam (CERN) Perturbative QCD in hadron collisions SILAFAE 2012-12-10 32 / 35
Seeing W’s and tops in a single jet
89
CMS single-jet W mass peakin events with a lepton andseparate b-tagged jet.
Uses pruning (+ mass-dropcondition on split jet)
Gavin Salam (CERN/Princeton/CNRS) Theory of Fat Jets Higgs Hunting 2012-07-19 19 / 28
Leading Fat Jet Mass [GeV]100 200 300 400 500 600
Even
ts /
20 G
eV
0
20
40
60
80
100
120
140
160
180 Data 2011
= 1.3 pbσZ' (1 TeV)
ttMultijet
ATLAS-1 L dt = 4.7 fb∫
= 7 TeVs
(a)
Leading Top-Quark Candidate Mass [GeV]140 150 160 170 180 190 200 210
Even
ts /
2 G
eV0
20
40
60
80
100
120Data 2011
= 1.3 pbσZ' (1 TeV)
ttMultijet
ATLAS-1 L dt = 4.7 fb∫
= 7 TeVs
(b)
Figure 7. Signal region distributions of (a) the mass of the leading pT fat jet and (b) the massof the leading pT top-quark candidate. Also shown are the prediction for SM tt̄ production, themultijet background contribution as estimated from data, and a hypothetical Z 0 boson signal.
– 16 –
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Searches with substructure tools
90
Some BSM searches with jet-substructure techniques
A range of techniques being used for varied BSM scenarios
Gavin Salam (CERN/LPTHE/Princeton) Jets in Higgs Searches HC2012 2012-11-18 24 / 29
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 92
EXTRA MATERIAL
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
What do we know currently?
At the time:• No clear picture of why the taggers might be similar or
different• No clear picture of how the parameter choices affect
the taggers
93
Boost 2010 proceedings:
1. Introduction
The Large Hadron Collider (LHC) at CERN is increasingly exploring phenomena at ener-
gies far above the electroweak scale. One of the features of this exploration is that analysis
techniques developed for earlier colliders, in which electroweak-scale particles could be con-
sidered “heavy”, have to be fundamentally reconsidered at the LHC. In particular, in the
context of jet-related studies, the large boost of electroweak bosons and top quarks causes
their hadronic decays to become collimated inside a single jet. Consequently a vibrant
research field has emerged in recent years, investigating how best to tag the characteristic
substructure that appears inside the single “fat” jets from electroweak scale objects, as
reviewed in Refs. [?,?,26]. In parallel, the methods that have been developed have started
to be tested and applied in numerous experimental analyses (e.g. [23–25] for studies on
QCD jets and [some searches]).
The taggers’ action is twofold: they aim to suppress or reshape backgrounds, while re-
taining signal jets and enhancing their characteristic jet-mass peak at the W/Z/H/top/etc.
mass. Nearly all the discussion of these aspects has taken place in the context of Monte
Carlo simulation studies [Some list], with tools such as Herwig [?, ?], Pythia [?, ?] and
Sherpa [?]. While Monte Carlo simulation is an extremely powerful tool, its intrinsic nu-
merical nature can make it di!cult to extract the key characteristics of individual taggers
and the relations between taggers (examining appropriate variables, as in [4], can be helpful
in this respect). As an example of the kind of statements that exist about them in the
literature, we quote from the Boost 2010 proceedings:
The [Monte Carlo] findings discussed above indicate that while [pruning,
trimming and filtering] have qualitatively similar e"ects, there are important
di"erences. For our choice of parameters, pruning acts most aggressively on the
signal and background followed by trimming and filtering.
While true, this brings no insight about whether the di"erences are due to intrinsic proper-
ties of the taggers or instead due to the particular parameters that were chosen; nor does it
allow one to understand whether any di"erences are generic, or restricted to some specific
kinematic range, e.g. in jet transverse momentum. Furthermore there can be significant
di"erences between Monte Carlo simulation tools (see e.g. [22]), which may be hard to diag-
nose experimentally, because of the many kinds of physics e"ect that contribute to the jet
structure (final-state showering, initial-state showering, underlying event, hadronisation,
etc.). Overall, this points to a need to carry out analytical calculations to understand the
interplay between the taggers and the quantum chromodynamical (QCD) showering that
occurs in both signal and background jets.
So far there have been three investigations into the analytical features that emerge from
substructure taggers. Ref. [19, 20] investigated the mass resolution that can be obtained
on signal jets and how to optimize the parameters of a method known as filtering [1].
Ref. [13] discussed constraints that might arise if one is to apply Soft Collinear E"ective
Theory (SCET) to jet substructure calculations. Ref. [14] observed that for narrow jets the
distribution of the N -subjettiness shape variable for 2-body signal decays can be resummed
– 2 –
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 201394
0
0.1
0.2
0.3
0.4
0.5
0.6
0.001 0.01 0.1 1
10 100 1000
m/m
dm
/ dm
m/pt
m [GeV], for pt = 4 TeV
Pythia 6 DW
, parton-shower level, no U
E, pp 14 TeV, pt,gen > 4 TeV, qq A qq, R
= 1
plain jet mass
The “right” MC study can already be instructive(testing on background [quark] jets)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 201395
0
0.1
0.2
0.3
0.4
0.5
0.6
0.001 0.01 0.1 1
10 100 1000
m/m
dm
/ dm
m/pt
m [GeV], for pt = 4 TeV
Pythia 6 DW
, parton-shower level, no U
E, pp 14 TeV, pt,gen > 4 TeV, qq A qq, R
= 1
plain jet massMass-drop tagger (ycut=0.09, µ=0.67)
Pruner (zcut=0.1)
Trimmer (zcut=0.1, Rtrim=0.2) Different taggersare apparentlyquite similar
The “right” MC study can already be instructive(testing on background [quark] jets)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 201396
0
0.1
0.2
0.3
0.4
0.5
0.6
0.001 0.01 0.1 1
10 100 1000
m/m
dm
/ dm
m/pt
m [GeV], for pt = 4 TeV
Pythia 6 DW
, parton-shower level, no U
E, pp 14 TeV, pt,gen > 4 TeV, qq A qq, R
= 1
plain jet massMass-drop tagger (ycut=0.09, µ=0.67)
Pruner (zcut=0.1)
Trimmer (zcut=0.1, Rtrim=0.2) But only for a limited range
of masses
The “right” MC study can already be instructive(testing on background [quark] jets)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Pileup
97
Jet
perturbative radiation
non-perturbative phase (hadronisation)
background radiation (underlying event)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Pileup
97
Jet
perturbative radiation
non-perturbative phase (hadronisation)
background radiation (underlying event)
background radiation (pileup)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Pileup
97
Jet
perturbative radiation
non-perturbative phase (hadronisation)
background radiation (underlying event)
background radiation (pileup)
Out of time pileup(especially ATLAS)
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Effect of pileup on 2 TeV Z’
98
1/N
dn/
dbin
/ 2
dijet mass [GeV]
qq, M = 2000 GeV
arXiv:0810.1304 0
0.02
0.04
0.06
0.08
1900 2000 2100
anti-kt, R=0.7Qw
f=0.12 = 26.1 GeV
1/N
dn/
dbin
/ 2
dijet mass [GeV]
qq, M = 2000 GeV
arXiv:0810.1304
0
0.02
0.04
0.06
0.08
1900 2000 2100
anti-kt, R=0.7Qw
f=0.12 = 51.9 GeVPU = 0.25 mb-1
no pileup ~25 pileup
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 99
ATLAS
Pileup for real
a few cm
~ 20
m
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
What goes into the jets?
100Figure: There are also magnets and things.
I Current grooming techniques only use calorimeter information.
I Particles tracks automatically remove pileup.
I Can tracking information be used to improve grooming?
Matthew Low (UChicago) Supercharged Jet Grooming June 12, 2013 6 / 11
ATLASCalorimeter towers, afterpre-clustering them into
“topoclusters”
Figure
: Ther
e are a
lsoma
gnets a
ndthin
gs.
I Curren
t groom
ingtec
hnique
s only
usecalo
rimete
r inform
ation.
I Particl
es trac
ksaut
omatic
allyrem
ovepile
up.
I Cantrac
king in
formatio
n beuse
d to imp
rove g
roomin
g?
M
a
t
t
h
e
w
L
o
w
(
U
C
h
i
c
a
g
o
)
S
u
p
e
r
c
h
a
r
g
e
d
J
e
t
G
r
o
o
m
i
n
g
J
u
n
e
1
2
,
2
0
1
3
6
/
1
1
CMS“Particle flow” objects ~
1) charged tracks
2) neutrals: calorimeter towers not associated with
charged tracks(or leftover bits of calo if
Ecalo – Etrack ≫ √Ecalo )
Matthew Low’s ideal detector
A CMS particle-flowexpert would shudder
at this description
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
What goes into the jets?
100Figure: There are also magnets and things.
I Current grooming techniques only use calorimeter information.
I Particles tracks automatically remove pileup.
I Can tracking information be used to improve grooming?
Matthew Low (UChicago) Supercharged Jet Grooming June 12, 2013 6 / 11
ATLASCalorimeter towers, afterpre-clustering them into
“topoclusters”
Figure
: Ther
e are a
lsoma
gnets a
ndthin
gs.
I Curren
t groom
ingtec
hnique
s only
usecalo
rimete
r inform
ation.
I Particl
es trac
ksaut
omatic
allyrem
ovepile
up.
I Cantrac
king in
formatio
n beuse
d to imp
rove g
roomin
g?
M
a
t
t
h
e
w
L
o
w
(
U
C
h
i
c
a
g
o
)
S
u
p
e
r
c
h
a
r
g
e
d
J
e
t
G
r
o
o
m
i
n
g
J
u
n
e
1
2
,
2
0
1
3
6
/
1
1
CMS“Particle flow” objects ~
1) charged tracks
2) neutrals: calorimeter towers not associated with
charged tracks(or leftover bits of calo if
Ecalo – Etrack ≫ √Ecalo )
Matthew Low’s ideal detector
A CMS particle-flowexpert would shudder
at this description
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
How do you remove pileup (PU)?
101
2. CMS
Throw out charged tracks not from the main primary
vertex
Use area/median(FastJet)method for neutral PU
1. Offset method
Count # of pileup vertices (npileup):
c ~ 0.5 GeV
3. Area/median (FastJet) method
Determine density of pileup pt / unit area ≡ ρ
psubtractedt,jet = pt,jet � ⇢⇥Ajet
psubtractedt,jet = pt,jet � c⇥ npileup
[Tevatron / old ATLAS]
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 102
Add “ghosts”, infinitesimally soft particles, to track “area”
of jet in y–φ plane
yφ
pt
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 102
Add “ghosts”, infinitesimally soft particles, to track “area”
of jet in y–φ plane
yφ
pt
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013 102
Add “ghosts”, infinitesimally soft particles, to track “area”
of jet in y–φ plane
yφ
pt
0
50
100
150
200
250
-4 -3 -2 -1 0 1 2 3 4
ptj / A
j [G
eV
]
yj
kt algorithm, R=0.5
Most jets are from pileup; all have similar pt / area
Estimate ρ for a given event as
⇢ = medianjets
⇢pt,jetAjet
�
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
pileup subtraction performance
103
I Can subtract pT based ondensity of event [Cacciari,Salam; arxiv:0707.1378]
p
(sub)
T = pT � ⇢A
⇢ = median
⇢pT ,j
Aj
�� 0
0.005
0.01
0.015
1900 2000 2100
1/N
dN
/dm
[G
eV
-1]
m [GeV]
kt, R=0.7
LHC, high lumiZ! at 2 TeV
no pileup
no pileup, sub
pileup
pileup, sub
I Jet shapes can also be subtracted [Soyez, Salam, Kim, Duta, Cacciari;arxiv:1211.280110]
V
jet,sub
= V
jet
� ⇢V [1]
jet
+1
2⇢2V [2]
jet
+ . . .
Matthew Low (UChicago) Supercharged Jet Grooming June 12, 2013 5 / 11
Used in CMS for neutral part of PU, and (I think) for recent ATLAS results
Gavin Salam (CERN) Jets and jet substructure (4) TASI, June 2013
Some observables particularly sensitive to PU
104
Beyond simply providing a pile-up-independent average jet mass, the optimal grooming configura-tions render the full jet mass spectrum insensitive to high instantaneous luminosity. Figure 3 demon-strates this by comparing the jet mass spectrum for leading ungroomed and trimmed anti-kt jets. Thecomparison is performed both in data and using the Z0 ! tt̄ MC sample. The result using the inclusivejet sample obtained from data shows that a nearly identical trimmed mjet spectrum is obtained regardlessof the level of pile-up. The peak of the leading jet mass distribution for events with NPV � 12 is shiftedcomparatively more due to trimming: from mjet ⇡ 125 GeV to mjet ⇡ 45 GeV as compared to an initialpeak position of mjet ⇡ 90 GeV for events with 1 NPV 4. Nonetheless, the resulting trimmed jetmass spectra exhibit no dependence on NPV.
1jetLeading jet mass, m
0 50 100 150 200 250 300
Arbi
trary
uni
ts
0
0.02
0.04
0.06
0.08
0.1
ATLAS Preliminary-1 Ldt = 4.7 fb∫Data 2011,
with R=1.0 LCW, No jet groomingtanti-k| < 0.8η < 800 GeV, |
Tjet p≤600
4≤PVN≤17≤PVN≤511≤PVN≤8
12≥PVN
(a) Data: anti-kt, R = 1.0: Ungroomed1jetLeading jet mass, m
0 50 100 150 200 250 300
Arbi
trary
uni
ts
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14 ATLAS Preliminary-1 Ldt = 4.7 fb∫Data 2011,
=0.3sub=0.05, Rcut with R=1.0 LCW, ftanti-k| < 0.8η < 800 GeV, |
T
jet p≤600 4≤PVN≤17≤PVN≤511≤PVN≤8
12≥PVN
(b) Data: anti-kt, R = 1.0: Trimmed
1jetLeading jet mass, m
120 140 160 180 200 220 240 260
Arbi
trary
uni
ts
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18 ATLAS Preliminary - Simulation= 1.6 TeV)
Z' (mt t→Z' with R=1.0 LCW, No jet groomingtanti-k
| < 0.8η < 800 GeV, |Tjet p≤600
4≤PVN≤17≤PVN≤511≤PVN≤8
12≥PVN
(c) Z0: anti-kt, R = 1.0: Ungroomed1jetLeading jet mass, m
120 140 160 180 200 220 240 260
Arbi
trary
uni
ts
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18 ATLAS Preliminary - Simulation= 1.6 TeV)
Z' (mt t→Z'
=0.3sub=0.05, Rcut with R=1.0 LCW, ftanti-k| < 0.8η < 800 GeV, |
Tjet p≤600
4≤PVN≤17≤PVN≤511≤PVN≤8
12≥PVN
(d) Z0: anti-kt, R = 1.0: Trimmed
Figure 3: Jet mass spectra for four primary vertex multiplicity ranges for anti-kt jets with R = 1.0 inthe range 600 pjet
T < 800 GeV. Both untrimmed anti-kt jets (left) and trimmed anti-kt jets (right) arecompared for the various NPV ranges in data (top) and for a Z0 ! tt̄ sample (bottom).
8
E.g. jet mass
For these, plain pileup subtraction
is only OKish
But why are weinterested in the
jet mass?