3.2 Jet Reconstruction and Calibration
69
initiated by quarks and gluons. The resolution improves with increasing m u and is
around 9–13% for the most probable value of m u ≈ 100–150 GeV. For a given value
of m u < 200 GeV, the resolution worsens with increasing jet p T due to a higher
degree of collimation. Remarkably, the resolution obtained in CMS is comparable
to the combined mass in ATLAS, even though quark/gluon jets are compared with
W /Z jets and very different technologies are used to reconstruct the jet mass.
3.3 Pileup Mitigation
Pileup originates from simultaneous pp collisions that occur in addition to a hard
scattering, triggering the readout and reconstruction of an event. The interaction vertex of the hard scattering is referred to as leading vertex (LV), and is usually taken
to be the vertex with the highest scalar sum of p T , calculated from reconstructed
objects associated to it. Pileup interactions are uncorrelated to the primary interaction (unlike MPIs) and typically consist of an admixture of inelastic, elastic and
diffractive pp processes. Vertices from pileup interactions are distributed along the
longitudinal direction in the sensitive detector region. As the detector response is not
instantaneous, pileup events from both the same (in-time) and neighbouring (outof-time) bunch crossings can contribute. This section focusses on the mitigation of
in-time pileup, though out-of-time pileup is also mitigated by dedicated reconstruction algorithms developed by the ATLAS and CMS Collaborations. During the LHC
data taking between 2010 and 2012 the mean number of pileup interactions reached
μ = 21, and μ values up to 60 were attained in certain runs of 2017 as shown in
Fig. 3.4 (left). For the data taking in 2022 even higher values are expected, culminating at μ = 140−200 for the high-luminosity LHC (HL-LHC). On average, pileup
leaves approximately 0.5 GeV of energy in the detector per unit area in (η, φ), per
pileup vertex. The effects of this are present in all aspects of LHC physics, from
detector design and software performance to the final sensitivity of measurements
and searches.
3.3.1 Mitigation Methods
Properties of pileup interactions can be exploited to remove energy contributions
from pileup to individual jets. Pileup can be approximated as a spatially uniform
deposition of energy. Using this approximation, the so-called area subtraction [185]
corrects the jet four-momentum through p
corr
T
= p
orig
T − ρ A, where A is the jet area.
The estimator ρ for the contribution from pileup per unit area is obtained from
reconstructed jets in the event, associated to pileup. An example of ρ is shown in
Fig. 3.4, with a slope of approximately 0.5 GeV per pileup vertex. There are many
subtleties in defining both ρ and A, which are discussed in e.g. [185, 441, 442]. An
69
initiated by quarks and gluons. The resolution improves with increasing m u and is
around 9–13% for the most probable value of m u ≈ 100–150 GeV. For a given value
of m u < 200 GeV, the resolution worsens with increasing jet p T due to a higher
degree of collimation. Remarkably, the resolution obtained in CMS is comparable
to the combined mass in ATLAS, even though quark/gluon jets are compared with
W /Z jets and very different technologies are used to reconstruct the jet mass.
3.3 Pileup Mitigation
Pileup originates from simultaneous pp collisions that occur in addition to a hard
scattering, triggering the readout and reconstruction of an event. The interaction vertex of the hard scattering is referred to as leading vertex (LV), and is usually taken
to be the vertex with the highest scalar sum of p T , calculated from reconstructed
objects associated to it. Pileup interactions are uncorrelated to the primary interaction (unlike MPIs) and typically consist of an admixture of inelastic, elastic and
diffractive pp processes. Vertices from pileup interactions are distributed along the
longitudinal direction in the sensitive detector region. As the detector response is not
instantaneous, pileup events from both the same (in-time) and neighbouring (outof-time) bunch crossings can contribute. This section focusses on the mitigation of
in-time pileup, though out-of-time pileup is also mitigated by dedicated reconstruction algorithms developed by the ATLAS and CMS Collaborations. During the LHC
data taking between 2010 and 2012 the mean number of pileup interactions reached
μ = 21, and μ values up to 60 were attained in certain runs of 2017 as shown in
Fig. 3.4 (left). For the data taking in 2022 even higher values are expected, culminating at μ = 140−200 for the high-luminosity LHC (HL-LHC). On average, pileup
leaves approximately 0.5 GeV of energy in the detector per unit area in (η, φ), per
pileup vertex. The effects of this are present in all aspects of LHC physics, from
detector design and software performance to the final sensitivity of measurements
and searches.
3.3.1 Mitigation Methods
Properties of pileup interactions can be exploited to remove energy contributions
from pileup to individual jets. Pileup can be approximated as a spatially uniform
deposition of energy. Using this approximation, the so-called area subtraction [185]
corrects the jet four-momentum through p
corr
T
= p
orig
T − ρ A, where A is the jet area.
The estimator ρ for the contribution from pileup per unit area is obtained from
reconstructed jets in the event, associated to pileup. An example of ρ is shown in
Fig. 3.4, with a slope of approximately 0.5 GeV per pileup vertex. There are many
subtleties in defining both ρ and A, which are discussed in e.g. [185, 441, 442]. An