22.6.2010
Jet results and jet reconstruction
techniques in pp and their prospects in
heavy-ion collisions in ATLASMartin Spousta
for the ATLAS Collaboration
Martin Spousta, Charles University in Prague 2
22.6.2010
Outline
• Motivation• Choice of the jet algorithm• Measurement of jet energy spectra in heavy ions
– jet energy scale– jet energy resolution
• Measurement of jet internal structure in heavy ions– jet shapes– fragmentation function and jT distribution– complementary quantities to study the energy loss
• Recent results from jet measurement in pp collisions at 7 TeV– inclusive jet cross section– jet shapes
• Summary
Martin Spousta, Charles University in Prague 3
22.6.2010
Motivation: Why ATLAS?
tracking over 5 units of pseudorapidity –
fragmentation studies
jet reconstruction using calorimeter,full azimuth, 10 units of pseudorapidity
first layer of LAr EM calorimeter excellent for photon isolation,
other layers also well segmented
ZDC to estimate centrality and reaction plane and to
measure UPC
Martin Spousta, Charles University in Prague 4
22.6.2010
Motivation: Expected jet rates
• Assuming 2 weeks of HI running means 1/20 of nominal luminosity (∫L = 25 b-1)
• Full ATLAS acceptance, 100% trigger efficiency
• Reasonable rates
– 4000 jets with ET> 100 GeV
– 50 jets with ET > 200 GeV
by W. Vogelsang
• Pb+Pb (<Ncoll> = 440)
• √sNN=2.75 TeV
• jets with R=0.4
• ∫L = 25 μb-1 (one year)
• σPb+Pb/σp+p=100
• full acceptance |η|<4.5
Martin Spousta, Charles University in Prague 5
22.6.2010
Jet reconstruction strategy
Background / underlying event subtraction
First subtract the background,then reconstruct jets
Cone-type
Jet finding algorithm
Cluster-typekT algorithm
(k=1)
anti-kT algorithm
(k=-1)
Seeded iterative cone algorithm
Infrared & collinear safe SISCone
algorithm
First reconstruct jets,then separate signal jets from background and subtract the background
Cambridge/Aachen(k=0)
Martin Spousta, Charles University in Prague 6
22.6.2010
Comparison among jet finding algorithms
• Which jet algorithm we should use? How the UE influences the jet finding?
• kT algorithm exhibits serious problems with the jet energy scale: for more severe background, kT underestimates the jet area and thus the energy. This is due to the fact that kT preferably clusters the softer part of a jet with the background
• Problematic algorithms: - kT algorithm, - Cambridge/Aachen
• Non-problematic: - anti-kT algorithm, - cone algorithm
mean bkgr Et
size of the fluctuations
in bkgr
jet energy scale
20% underestimation
Toy model: Poisson-like background (“UE”) randomly generated to allow variation of mean and rms of ET
kT algorithm
Martin Spousta, Charles University in Prague 7
22.6.2010
Comparison among jet finding algorithms
• Which jet algorithm we should use? How the UE influences the jet finding?
• kT algorithm exhibits serious problems with the jet energy scale: for more severe background, kT underestimates the jet area and thus the energy. This is due to the fact that kT preferably clusters the softer part of a jet with the background
• Problematic algorithms: - kT algorithm, - Cambridge/Aachen
• Non-problematic: - anti-kT algorithm, - cone algorithm
mean bkgr Et
size of the fluctuations
in bkgr
jet energy scale
no problems with jet energy scale
anti-kT algorithm
Toy model: Poisson-like background (“UE”) randomly generated to allow variation of mean and rms of ET
Martin Spousta, Charles University in Prague 8
22.6.2010
Comparison among jet finding algorithms
Area of jets delivered by different algorithms is very different in the noisy environment
• kT algorithm, R=0.4
• smaller jets with irregular areas• denser UE leads to smaller jet areas
• anti-kT algorithm, R=0.4
• conical jets with the average radius of R
the sameevent
Jet
ID n
um
ber
Jet
ID n
um
ber
jet ET
(GeV)
Martin Spousta, Charles University in Prague 9
22.6.2010
Characteristic properties of anti-kT algorithm
Mach cone? No. This is just an effect of optimization of jet axis with respect to the jet periphery ... characteristic for the anti-kT algorithm.
One needs to be careful about systematic biases coming from usage of a given jet algorithm.
Let’s look at multiplicity of particles as a function of the distance from the jet axis (charged PYTHIA particles, MinBias pp collisions, √s = 900 GeV)
We see a“Mach-cone”!
Martin Spousta, Charles University in Prague 10
22.6.2010
truth jet ET (GeV)
Jet energy resolution
p+p Pb+Pbb=2fm, dN/d~2700
Fitting well known formula: to p+p and Pb+Pb energy resolution
• Jet energy resolution below 25% for 70 GeV jets in the most central collisions (dN/d~2700 b=2 fm, unquenched HIJING, cone algorithm)
• Irreducible background fluctuations from HIJING simulations: ~ 15 GeV / jet ... ... but reality can be very different ...
stochastic term noise term
||<5.0
ATLAS PreliminaryATLAS Preliminaryconstant term
truth jet ET (GeV)
Martin Spousta, Charles University in Prague 11
22.6.2010
Jet energy resolution
... fluctuations in the underlying event are highly model dependent!
Once we know real fluctuations we can easily compute limiting jet energy resolution in PbPb collisions!
=Realistic jet energy resolution
in PbPb
truth jet ET (GeV) calorimeter tower ET (MeV)
<ET>=1.8 GeV, (ET)=0.8 GeV<ET>=1.3 GeV, (ET)=0.6 GeV
— HIJING— HYDJETb = (0-3) fm, ||<5
(ET)=0.8 GeV means p1~15 GeV
(ET)=0.6 GeV means p1~11 GeV!
Real heavy ion underlying event
||<5.0
p+p
ATLAS Preliminary
, PYTHIA
Martin Spousta, Charles University in Prague 12
22.6.2010
Fragmentation function and jT
distribution:unquenched jets
Reconstruction procedure:• tracks are matched to calorimeter towers of a jet• jT and z for tracks with pT>2 GeV is computed• background distributions of jT and z are computed using
tracks that match with HIJING particles, these distributions are subtracted, correction for the jet position resolution is applied
… we can well reproduce jT distribution and fragmentation functionin the presence of heavy-ion UE
z
jT
ATLAS Preliminary
○ - truth● - reconstructed▼ - background(jet ET 70-140 GeV)
Simulation of central PbPb collisions
ATLAS PreliminarySimulation of central PbPb collisions○ - truth● - reconstructed▼ - background(jet ET 70-140 GeV)
Martin Spousta, Charles University in Prague 13
22.6.2010
Fragmentation function and jT
distribution:
quenched jets
Large jT suppressed gluons radiated from large angles
Low z enhanced, higher z suppressed leading particle suppressed, redistribution of
energy out of a jet core
Effect of jet quenching simulated by PYQUEN (without heavy ion UE) ...... if the quenching is of the order of PYQUEN prediction we should be
able to measure it
Reconstructed○ - unquenched● - quenched(jet ET=70-140 GeV)
ATLAS Preliminary
Reconstructed○ - unquenched● - quenched(jet ET=70-140 GeV)
ATLAS Preliminary
Martin Spousta, Charles University in Prague 14
22.6.2010
• black = reference, computed using simulation of pp jets, <ET>=60 GeV
• open = smearing of the jet axis by convoluting with gaussian (R)=0.06 - leads to: characteristic shift in jT distribution; fragmentation function remains almost unaffected
• blue = simulation of jet energy scale shift by +10% - leads to:characteristic shift in fragmentation function; jT distribution remains unaffected
UE or detector effects could mimic the jet quenching – careful cross-checks are needed.
Simulation of pp jets(jet ET 70-140 GeV)
ATLAS Preliminary
Simulation of pp jets(jet ET 70-140 GeV)
ATLAS Preliminary
Fragmentation function and jT
distribution: detector / UE effects
Martin Spousta, Charles University in Prague 15
22.6.2010
Jet shapes
• Measure the energy flow inside the jet at the calorimeter level
• First jet shapes already measured in 7 TeV pp collisions at ATLAS! More details later
jet axis
... integral jet shape
... differential jet shape
Martin Spousta, Charles University in Prague 16
22.6.2010
• black = reference jet shape computed using simulated pp jets
• red = adding gaussian noise with <ET>=0, (ET)=1.4 GeV leads to:
– jet energy resolution deterioration, (Et/Et) = 20%
– jet position resolution deterioration (R) = 0.06
– characteristic change in the jet shape, and pT spectra
• green = we can recover the jet shape e.g. by re-computing the jet axis using simple cone algorithm with R randomly chosen in <0.1,0.5>
Jet shape measurements: UE / detector effects
jet shape jet energy flowSimulation of pp jets(jet ET 70-140 GeV)
Simulation of pp jets(jet ET 70-140 GeV)
ATLAS Preliminary ATLAS Preliminary
Martin Spousta, Charles University in Prague 17
22.6.2010
Effect of the calorimeter can be corrected by unfolding (bin-by-bin as good as bayesian unfolding)
Jet shape measurements: UE / detector effects
○ - particle level● - detector level
○ - particle level● - detector level□□ - bayesian□□ - bin-by-bin
Simulation of pp jetsjet ET 70-140 GeV
Simulation of pp jetsjet ET 70-140 GeV
ATLAS Preliminary ATLAS Preliminary
Martin Spousta, Charles University in Prague 18
22.6.2010
Modification of jet shapes
• Jet quenching effect predicted by PYQUEN is well measurable
PbPb – no quenching(jet ET = 70-140 GeV)
(jet ET = 70-140 GeV)
ATLAS Preliminary ATLAS Preliminary
• Jet shape can be well measured – it is almost centrality independent (after applying corrections on jet position resolution)
Martin Spousta, Charles University in Prague 19
22.6.2010
jet shape
jet shape
jet energy spectrum
jet energy spectrum
PYQUEN with only radiative
energy loss
PYQUEN with only collisional
energy loss
○ - non-quenched● - quenched
○ - non-quenched● - quenched - non-quenched
- quenched
- non-quenched - quenched
Complementary quantities to measure different energy loss
mechanisms
We need to use different quantities to learn about energy loss mechanism
ATLAS Preliminary ATLAS Preliminary
ATLAS Preliminary ATLAS Preliminary
Martin Spousta, Charles University in Prague 20
22.6.2010
Inclusive jet cross section measured
in pp collisions at 7 TeV• Inclusive jet differential cross section
measured in pp collisions at 7 TeV using L ~ 17 nb-1 (uncertainty of 11%).
• Jets defined using R=0.6 anti-kT algorithm. Topological calorimeter clusters used in the input for the jet reconstruction.
• Data corrected for detector effects using bin-by-bin correction and compared to NLOJET++ with CTEQ6.6 PDFs.
• Theory uncertainties due to the choice of PDFs, renormalization and factorization scheme, s, and non-perturbative effects estimated and summed in quadrature (orange band)
• jet energy scale < 9%
• perturbative effects estimated and summed in quadrature (orange band).
• Main contribution to the systematic errors (grey band) is due to the jet energy scale uncertainty which is below 9% in the entire pT and rapidity range (below 7% for the central jets with ||<0.8).
Martin Spousta, Charles University in Prague 21
22.6.2010
Inclusive jet cross section measured
in pp collisions at 7 TeV
• Double differential single inclusive jet cross section for different rapidity bins as a function of pT (left) and for different pT bins as a function of rapidity (right).
• Data matches theory within systematic and statistical errors in all pT and rapidity bins.
Martin Spousta, Charles University in Prague 22
22.6.2010
Jet shapes in pp collisions at 7 TeV
• Jet shapes measured at the calorimeter level (left) and using charged tracks (right).
• Differential jet shape measured using anti-kT jets with R=0.6. Data compared with detector level MC simulation.
• Data agrees with Monte Carlo up to 10% in the whole pT,jet range of 40 – 210 GeV and rapidity range of |y| < 2.8.
• ATLAS tune of Herwig + Jimmy provide the best description of the jet shapes.
Martin Spousta, Charles University in Prague 23
22.6.2010
Summary
• ATLAS detector is well suited for jet studies in heavy-ion collisions (it is also well suited for soft heavy ion physics, quarkonia, ultra-peripheral collisions, and other important measurements).
• We have stressed some caveats of the measurement of jets and their properties that can influence our understanding of the jet energy loss mechanism.
• We have shown first results on inclusive single jet cross section measurement and jet shape measurement in pp collisions at √sNN = 7 TeV.
Martin Spousta, Charles University in Prague 24
22.6.2010
Backup
Backup
Martin Spousta, Charles University in Prague 25
22.6.2010
Jet reconstruction strategy
Cone jet reconstruction:
• regions of interest found (seed regions) – fast sliding window algorithm used (size of the sliding window 0.3x0.3)
• background computed • excluding the seed-regions
(R=0.8 outside of a seed)• vs. (binning of 0.1) • vs. layer (24 calorimeter layers)
• background subtracted
• standard p+p jet finding algorithm used (seeded iterative R=0.4 cone algorithm)
An alternative: (anti)kT-algorithm based reconstruction strategy
PYTHIA dijets embedded into the unquenched HIJING events
one event before the background subtraction
one event after the background subtraction
Martin Spousta, Charles University in Prague 26
22.6.2010
Fast (anti)kT jet reconstruction strategy
(anti)kT jet reconstruction:
• run fast (anti)kT algorithm• separate “signal” jets from the
“background” jets • use regions outside of signal jets to
estimate the background• subtract the background from jets ... motivated by Cacciari and Salam
An alternative: Use the same procedure as for the cone algorithm (previous slide) but subtract the background after the jet finding.
Martin Spousta, Charles University in Prague 27
22.6.2010
Jet energy scale and efficiency
dN/d~2700
• Jet energy scale within 5% above 60 GeV in the most central collisions simulated by “unquenched” HIJING (dN/d~2700 b=2 fm). Cone algorithm R=0.4.
• Below 60 GeV the efficiency is steeply decreasing jet energy scale shift (it is more probable to find jets with higher energy)
To trust any jet observation we need to have efficiency under the control.
PbPbb=2 fm, dN/d~2700
ATLAS Preliminary ATLAS Preliminary
Martin Spousta, Charles University in Prague 28
22.6.2010
Jet position resolution
• Jet position resolution in similar to that in (full field simulated)
• It improves with increasing jet energy, in the whole energy range jet position resolution is better than half of the tower size
• Jet position resolution can be improved using smaller cones: jet axis of a reconstructed jet is substituted by the jet axis from jet reconstructed with R<Rorig
Pb+PbdN/d~2700
Pb+PbdN/d~2700
Towersize
ATLAS Preliminary ATLAS Preliminary
Cone algorithm Cone algorithm
Martin Spousta, Charles University in Prague 29
22.6.2010
Efficiency and fake-rate
• Efficiency is almost centrality independent – easier interpretation of jet properties vs. centrality
• Above 70 GeV the efficiency is above 90%
• Above 70 GeV very low fake rate < 5% (without any fake rejection)
ATLAS Preliminary
ATLAS Preliminary
Cone algorithm Cone algorithm
Martin Spousta, Charles University in Prague 30
22.6.2010
• associated and leading jets with ET>100 GeV
Dijet correlations
ATLAS Preliminary ATLAS Preliminary
Cone algorithm Cone algorithm
Martin Spousta, Charles University in Prague 31
22.6.2010
Tracking – resolution, efficiency
70%
ATLAS Preliminary
ATLAS Preliminary
ATLAS Preliminary
ATLAS Preliminary
Martin Spousta, Charles University in Prague 32
22.6.2010Backup slides
Fake jet reduction
ATLAS Preliminary
ATLAS Preliminary
Martin Spousta, Charles University in Prague 33
22.6.2010
Statistics
only the most central collisions
x jet quenching effect simulated with 5000 PYQUEN events
Martin Spousta, Charles University in Prague 34
22.6.2010
pp vs PbPb – important parameters