3.2 Jet Reconstruction and Calibration
67
the mean of this response distribution,the jet mass resolution (JMR) as one standard
deviation.
5
ATLAS has developed a data-driven approach using forward-folding to extract
the JMS and JMR. In this approach, the jet mass distribution at the particle level is
modified by a response function such that the JMS and JMR are scaled by appropriate
scale parameters, which can be functions of jet mass and p T . The values of these
parameters for which the folded distribution best matches the data are extracted
from a two-dimensional fit [427, 428]. With this approach, the JMS and JMR for
hadronically decaying boosted W bosons with p T 200 GeV are determined with
2–3% and 20% systematic uncertainties, respectively [429]. The results from this
measurement are combined with the so-called R trk method, which constrains the
mass scale by comparing the calorimeter jet mass to the mass calculated from track
jets. The R trk method extends the jet mass calibration up to p T = 3000 GeV [429].
It can be generalised to other variables and is used in ATLAS to constrain the p T
scale of large-R jets as well as to derive systematic uncertainties on jet substructure
variables. ATLAS has also studied combining calorimeter and tracking information
to maintain a stable jet mass reconstruction for highly boosted particles, where the
calorimeter granularity is not sufficient any longer. The track-assisted mass [427]
benefits from the excellent angular resolution of the tracking detector and shows a
better JMR at p T 1000 GeV, while the calorimeter mass performs better below
this value. The combined mass is a weighted average of these two reconstruction
methods, based on the expected resolution.
A similar approach of correcting topoclusters based on tracking information has
been studied as well, the method based on Track-CaloClusters (TCC) improves the
resolution of jet substructure variables at high p T [410]. The TCC method is complementary to the PF algorithm, but is designed to work in a very different energy regime.
While the PF algorithm is especially useful to improve the jet reconstruction at low
p T , the TCC approach is based on a weighting scheme where the track momenta are
used to spatially redistribute the energy measured in the calorimeter. By doing this,
the TCC improves the angular resolution of the calorimeter measurement, without
changing the total energy measured in the calorimeter. This approach improves the
resolution of jet substructure observables at high p T , although the JMR is slightly
worse than for the combined mass below p T < 1.5 TeV. Another approach studied
in ATLAS is a track-assisted reconstruction of substructure observables [430]. This
approach is a generalisation of the track-assisted mass [427], taking into account
local fluctuations inside the jet. There are different flavours of the track-assisted jet
reconstruction, but they all have in common that in a first step tracks are assigned to
calibrated small-R jets.
6 In a second step, the p T of each track is scaled such that the
total p T of the track-jet is equal to the p T of the calibrated calorimeter jet. In this way,
5 ATLAS often uses half of the 68% interquartile range of the response distribution, which is robust
against non-Gaussian tails and is equal to one standard deviation for a Gaussian response distribution.
6 These can be subjets of large-R jets or all small-R jets found in the event, which are re-clustered
into large-R jets in a later stage of the algorithm [431, 432]. This track-assignment is done via ghost
association [433], but in some cases un-matched tracks are added using a R criterion.
67
the mean of this response distribution,the jet mass resolution (JMR) as one standard
deviation.
5
ATLAS has developed a data-driven approach using forward-folding to extract
the JMS and JMR. In this approach, the jet mass distribution at the particle level is
modified by a response function such that the JMS and JMR are scaled by appropriate
scale parameters, which can be functions of jet mass and p T . The values of these
parameters for which the folded distribution best matches the data are extracted
from a two-dimensional fit [427, 428]. With this approach, the JMS and JMR for
hadronically decaying boosted W bosons with p T 200 GeV are determined with
2–3% and 20% systematic uncertainties, respectively [429]. The results from this
measurement are combined with the so-called R trk method, which constrains the
mass scale by comparing the calorimeter jet mass to the mass calculated from track
jets. The R trk method extends the jet mass calibration up to p T = 3000 GeV [429].
It can be generalised to other variables and is used in ATLAS to constrain the p T
scale of large-R jets as well as to derive systematic uncertainties on jet substructure
variables. ATLAS has also studied combining calorimeter and tracking information
to maintain a stable jet mass reconstruction for highly boosted particles, where the
calorimeter granularity is not sufficient any longer. The track-assisted mass [427]
benefits from the excellent angular resolution of the tracking detector and shows a
better JMR at p T 1000 GeV, while the calorimeter mass performs better below
this value. The combined mass is a weighted average of these two reconstruction
methods, based on the expected resolution.
A similar approach of correcting topoclusters based on tracking information has
been studied as well, the method based on Track-CaloClusters (TCC) improves the
resolution of jet substructure variables at high p T [410]. The TCC method is complementary to the PF algorithm, but is designed to work in a very different energy regime.
While the PF algorithm is especially useful to improve the jet reconstruction at low
p T , the TCC approach is based on a weighting scheme where the track momenta are
used to spatially redistribute the energy measured in the calorimeter. By doing this,
the TCC improves the angular resolution of the calorimeter measurement, without
changing the total energy measured in the calorimeter. This approach improves the
resolution of jet substructure observables at high p T , although the JMR is slightly
worse than for the combined mass below p T < 1.5 TeV. Another approach studied
in ATLAS is a track-assisted reconstruction of substructure observables [430]. This
approach is a generalisation of the track-assisted mass [427], taking into account
local fluctuations inside the jet. There are different flavours of the track-assisted jet
reconstruction, but they all have in common that in a first step tracks are assigned to
calibrated small-R jets.
6 In a second step, the p T of each track is scaled such that the
total p T of the track-jet is equal to the p T of the calibrated calorimeter jet. In this way,
5 ATLAS often uses half of the 68% interquartile range of the response distribution, which is robust
against non-Gaussian tails and is equal to one standard deviation for a Gaussian response distribution.
6 These can be subjets of large-R jets or all small-R jets found in the event, which are re-clustered
into large-R jets in a later stage of the algorithm [431, 432]. This track-assignment is done via ghost
association [433], but in some cases un-matched tracks are added using a R criterion.