Jet Algorithms

Transcription

Jet Algorithms
Jet Algorithms
Pierre-Antoine Delsart
LPSC Grenoble
February 27, 2009
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
1 / 13
Introduction
Jets are due to QCD interactions
Several level of jets
partons (QCD perturbative)
hadrons (QCD
non-perturbative)
experimental
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
2 / 13
Jet Algorithms
List of particles :
- partons
- hadrons
- calorimeter clusters
- ...
Groups of particles :
Jet Algorithm
JETS
Theoretical constraints:
infrared & colinear safe (QCD issues, see later)
z-boost invariance
level Independence (parton,hadron, calorimeter)
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
3 / 13
Jets families
Cones
simple geometric motivation
Find stable cones in eta,phi plan (stable : jet 4mom is along cone
axis)
Split/merge procedure for overlapping cones
sa
Clustering (Kt )
Pairwise successive aggregation of proto-jets
Dene a distance between constituents
Aggregate rst closest constituent
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
4 / 13
Cones algorithms
Finding all stable cones : naive approach is extremely slow
=⇒Use
Seeded algorithms
Atlas Cone : a seeded algorithm
search stable cone only around highest Pt constituents
split/merge : if 2 jets overlap, compute fraction of Pt in overlap
region. Split if
P.A. Delsart (LAPP)
f
< 0.75
Jet Algorithms
February 27, 2009
5 / 13
Atlas Cone problems
Atlas cone not Infrared/collinear safe
Dark tower : signicant constituent can be excluded (in black below)
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
6 / 13
Jets (p. 46)
Cone
xC-SM
IRC safety & real-life
Real life does not have infinities, but pert. infinity leaves a real-life trace
α2s + α3s + α4s × ∞ → α2s + α3s + α4s × ln pt /Λ → α2s + α3s + α3s
| {z }
BOTH WASTED
Among consequences of IR unsafety:
Inclusive jets
W /Z + 1 jet
3 jets
W /Z + 2 jets
mjet in 2j + X
Last meaningful order
JetClu, ATLAS MidPoint CMS it. cone
cone [IC-SM]
[ICmp -SM]
[IC-PR]
LO
NLO
NLO
LO
NLO
NLO
none
LO
LO
none
LO
LO
none
none
none
NB: $30 − 50M
Known at
NLO (→ NNLO)
NLO
NLO [nlojet++]
NLO [MCFM]
LO
investment in NLO
Multi-jet contexts much more sensitive: ubiquitous at LHC
And LHC will rely on QCD for background double-checks
extraction of cross sections, extraction of parameters
SISCone
Solution to previous problems : SISCone (Seedless Infrared Safe Cone
algorithm)
Developed by Salam, Soyez et al.
Seedless : no seeds. Absolutely
all
stable cones are considered
Smart split/merge procedure
=⇒No
more theoretical problems.
A smart implementation (using geometrical tricks) makes it relatively fast
(∼
O (N ²))
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
7 / 13
Clustering algorithms
Algorithm description :
Dene a distance
∆Rij2
Dij = min(PTi , PTj ) R 2
Di = PTi2
Compute all {Dij , Di } and d = min({Dij , Di })
2
if
if
2
d = D : combine jet i with jet j
d = D dene jet i as a nal jet
ij
i
Exhaust all proto-jets
Variants : Anti-kt and Cambridge. In distance formula replace
PT2 −→ PT2p
p=1 : standard Kt
p=0 : cambridge
p=-1 : Antit-kt
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
8 / 13
Algorithms in a picture
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
9 / 13
Algorithms in a picture
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
10 / 13
Algorithms performances
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
11 / 13
Efficiency and Purity
Event samples: di­jet samples.
No Pile­up
ΔR Match= 0.2
Jet Size: R=0.4
Input: TopoClusters
AntiKt Topo|η|<0.4
AntiKt Topo|η|<0.4
Cone Topo |η|<0.4
Cone Topo |η|<0.4
ΔR Match= 0.2
P. Francavilla
Efficiency = NTruth
/ NTruth
Matched
Kt Topo |η|<0.4
Kt Topo |η|<0.4
SISCone Topo |η|<0.4
SISCone Topo |η|<0.4
P. Francavilla
Reco
Purity = NMatched
/ NReco
AntiKt Topo |η|<0.4
Cone Topo |η|<0.4
Kt Topo |η|<0.4
SISCone Topo |η|<0.4
Studies done on other samples:
W+Jets, Higgs: ATL­COM­PHYS­2008­021
Campanelli/Geerlings/Huston Top: Sanchez, De Cecco
Studies on different topologies and generators (i.e. AlpGen, Sherpa) are in preparation.
The clustering algorithms have a better efficiency and purity.
Paolo Francavilla Jet Algorithms
29
Calibration based on cell weighting derived for ATLAS Cone
Jet Size: R=0.4
Input: TopoClusters
AntiKt Topo η<0.4
Cone Topo η<0.4
SISCone Topo η<0.4
ΔR Match= 0.2
We will provide soon dedicated calibrations for the new jet algorithms (S. Eckweiler)
P. Francavilla
AntiKt Topo η<0.4
Cone Topo η<0.4
SISCone Topo η<0.4
Paolo Francavilla P. Francavilla
EReco/ETrue
Linearity and Resolution
The jet algorithms show
similar linearities and resolutions Similar results for the spatial resolution (Sanchez, De Cecco)
Jet Algorithms
30
Algorithms and Pileup
Compare same jets in same events with / without pile-up
Low luminosity
Central Calo jets
Study ratio mass ratio :
M(pileup) / M(no-pileup)
Anti-kt R=0.6
Kt R=0.6
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
12 / 13
Conclusions
Choice of Jet Algorithm
matters
!
Theoretical aspects
Inuence of Pile-up
performances
(Largest eect still comes from size parameter)
Atlas will soon give up ATLAS Cone implementation in favor of SISCone or
Anti-Kt
More information in G. Salam lecture
P.A. Delsart (LAPP)
Jet Algorithms
February 27, 2009
13 / 13