ALICE Collaboration / Physics Letters B 746 (2015) 385–395 387
samples (see Section 2) only jets with p
ch+ne
T
,jet
> 40 GeV/c for the
low energy trigger and p
ch+ne
T
,jet
> 60 GeV/c for the higher energy
trigger are used. In these kinematic regimes the triggers are fully
efficient and no fragmentation bias is observed with respect to
the minimum-bias jet sample. The |k
Ty
| distributions are reported
at particle level involving a correction for detector effects and
p
T
-smearing due to the underlying event, see Section 3.3.
The
dijet sample is biased due to the requirement that the full
jet has a larger transverse momentum than the associated charged
jet, while the full jet momentum is used to estimate k
Ty
. In an
unbiased measurement, the full jet would correspond to the lead-
ing
jet of the event in only 50% of the cases. A PYTHIA [29,30]
study
was performed in which the particle-level jet was defined as
the jet containing all final state particles (no kinematic selection
on constituents and full azimuthal acceptance). Applying the selec-
tion
of this analysis to detector-level reconstructed jets, results in
a correct tagging of the leading jet in 70% of the dijet events. This
results in a slightly harder |k
Ty
| distribution with a 10% smaller
yield at low |k
Ty
| and 20% higher yield at large |k
Ty
|. The results of
the p–Pb data analysis will be compared to particle-level PYTHIA
with the same dijet selection incorporating the mentioned bias.
The
dijet acoplanarity is measured as a function of the trans-
verse
momentum of the full jet while the kinematic interval of the
associated charged jet is also varied to explore k
Ty
for more or less
balanced dijets in transverse momentum. In addition k
Ty
distribu-
tions
are also presented for two event multiplicity classes.
3.3. Corrections and systematic uncertainties
The measured dijet |k
Ty
| distributions are corrected to the par-
ticle
level, defined as the dijet |k
Ty
| from jets clustered from all
prompt particles produced in the collision including all decay
products, except those from weak decays of light flavor hadrons
and muons. Both full and charged jets are accepted at particle
level in the full azimuthal acceptance and in the pseudorapid-
ity
range of |η
jet
| < 0.5. The correction to particle level is based
on a data-driven method to correct for the influence of the un-
derlying
event fluctuations and on simulated PYTHIA events (tune
Perugia-2011 [31]) transported through the ALICE detectors layout
with GEANT3 [32]. The correction procedure takes into account the
p
T
and angular resolution of the measured dijets.
Detector-level
jets are defined as jets reconstructed from recon-
structed
tracks and EMCal clusters after subtraction of the charged
energy deposits. The jet energy scale and resolution are affected
by unmeasured particles (predominantly K
0
L
and neutrons), fluc-
tuations
of the energy deposit by charged tracks in the EMCal, the
EMCal energy scale and the charged particle tracking efficiency and
p
T
resolution. A response matrix as a function of p
T,jet
of the full
and associated charged jet, ϕ
dijet
and k
Ty
is created after match-
ing
the detector-level to the particle-level jets as described in [33].
The
p
T
-smearing due to fluctuations of the underlying event
is estimated with the random-cones technique which is also ap-
plied
in the analysis of Pb–Pb data [24]. Cones with a radius equal
to the resolution parameter R are placed in the measured p–Pb
events at random positions in the η–ϕ plane ensuring the cone
is fully contained in the detector acceptance. The fluctuations of
the background are characterized by the difference between the
summed p
T
of all the tracks and clusters in the random cone (RC)
and the estimated background: δ p
T
=
RC
i
p
T,i
− A · ρ, where A is
the area of the random cone (A = π R
2
) and the subscript i indi-
cates
a cluster or track pointing inside the random cone. A random
cone can overlap with a jet but to avoid oversampling in small sys-
tems
like p–Pb, a partial exclusion of overlap with the leading jet
in the event is applied. This is achieved by excluding random cones
overlapping with a leading jet with a given probability, p = 1/N
coll
where N
coll
is the number of binary collisions. N
coll
is taken from
estimates applying a Glauber fit to the multiplicity measured in
the V0A detector resulting in values between 14.7 and 1.52 de-
pending
on the event activity measured in the V0A detector. The
width of the background fluctuations for full (charged) jets varies
between 2.12 (1.59) and 0.73 (0.56) GeV/c depending on the mul-
tiplicity
of the event.
The
influence of background fluctuations is added to the re-
sponse
extracted from detector simulation through a Monte Carlo
model assuming that the background fluctuation for the full and
associated charged jet are uncorrelated within 20% wide bins of
V0A multiplicity classes. Within these selected multiplicity classes
the variation of the background fluctuations is negligible. Since the
Monte Carlo model does not generate full events and only accounts
for the δ p
T
smearing on a jet-by-jet basis, additional jet finding
inefficiencies and worsening of angular resolution due to the back-
ground
fluctuation are not taken into account. These effects are
negligible since the contribution of the underlying event to the jets
in p–Pb collisions is small. No correction for the angular resolution
of the charged jet due to missing neutral fragments is applied. This
effect increases the width of ϕ
dijet
by ∼ 0.03 and is present, and
of the same magnitude, in the p–Pb data and the PYTHIA refer-
ence.
The
most probable correction to the jet energy, taking into ac-
count
detector effects and background fluctuations, for fully recon-
structed
jets is 28% at p
ch+ne
T
,jet
= 20 GeV/c and decreases to 20%
for jets with p
ch+ne
T
,jet
> 40 GeV/c. The uncertainty on the jet energy
scale is evaluated by changing the tracking efficiency in data and
full detector simulation [34], varying the double counting correc-
tion
for the hadronic energy deposit in the EMCal and by using
different estimates of the underlying-event fluctuations. The final
uncertainty on the jet energy scale is 4%. The jet energy resolution
for full jets is 22% at p
ch+ne
T
,jet
= 20 GeV/c and decreases gradually to
18% at p
ch+ne
T
,jet
= 120 GeV/c. The influence of the uncertainties on
the jet energy scale and resolution on the dijet |k
Ty
| measurement
are discussed in the following.
The
measured |k
Ty
| distributions are corrected to the parti-
cle
level by applying bin-by-bin correction factors, which are
parametrized by a linear fit to the ratio between the particle- and
detector-level |k
Ty
| distributions for a given dijet selection. The cor-
rection
factors take into account the effects of feed-in and feed-out
of the selected kinematic and angular intervals of the full and as-
sociated
charged jets. These effects slightly change the shape of
the |k
Ty
| distributions resulting in correction factors which vary
between 0.9for small |k
Ty
| to 1.2at large |k
Ty
|. The correction is
relatively small, because while feed-in from lower p
ch+ne
T
,jet
narrows
the |k
Ty
| distribution, feed-in from higher p
ch+ne
T
,jet
broadens the dis-
tribution
resulting in a cancellation. Similarly the feed-out to high
and low p
ch+ne
T
,jet
has a small effect on the observable. By using a
linear fit to the correction factors the statistical fluctuations of the
detector simulation are not propagated to the measurement. The
95% confidence limit of the parametrization using the linear fit is
included in the systematic uncertainty of the measurement. Cor-
rection
factors are extracted as a function of V0A event multiplicity
class and kinematic intervals of the full and associated charged jet.
The
dominant systematic uncertainty on the measurement orig-
inates
from the extraction of bin-by-bin correction factors. The un-
certainty
of the parametrization of the correction factors results in
10–20% correlated systematic uncertainty on the dijet |k
Ty
| yields.
An additional 2.5% uncertainty arises from the uncertainty on the
tracking efficiency which is 4%. Systematic uncertainties originat-
ing
from the charged hadron energy deposit in EMCal towers,