3.4 Grooming Methods
73
The ATLAS experiment has adopted trimming for anti-k T , R = 1.0 jets with f cut
= 0.05 and R sub = 0.3 in analyses of 8 TeV data, and R sub = 0.2 for analyses of
13 TeV data. Extensive studies of different algorithms showed that this choice gives
the best results in terms of pileup stability, jet mass resolution and jet tagging performance [452]. The reclustering of small-R jets [431] has the experimental advantage
of using fully calibrated inputs when clustering jets with a larger distance parameter,
and has already been used in several ATLAS searches for Supersymmetry [456–458].
This method allows for flexibility of optimizing the jet distance parameter depending on the considered phase-space of the analysis without a need to re-derive jet
energy corrections [432]. Note that this method has similarities to trimming with
R sub set to the distance parameter of the small-R jets. The difference is in the value
of p T,sub , which is fixed to the minimum p T of the small-R jets for reclustering, but
dynamically adjusted to the p T of the large-R jet for trimming. The groomed jets
obtained with reclustering are different from trimmed jets also because reclustering
considers all small-R jets in an event, while trimming uses only the ones obtained
from the constituents of large-R jets. In practice these differences lead to small
effects only, resulting in a very similar performance of reclustering and trimming.
A recent ATLAS study [453] on grooming algorithms includes PF reconstruction
and a number of hybrid pileup removal techniques. It is shown that gains in tagging
performance are possible by using soft drop instead of trimming, when used with PF
and a combination of Constituent Subtraction and SoftKiller.
The CMS experiment has employed PF with CHS already early in the data taking,
which resulted in a small impact from pileup on jet substructure during the 8 TeV
data taking. Contributions from the neutral pileup component have been mitigated
using the pruning algorithm for anti-k T , R = 0.8 jets with z prune = 0.1 and R prune
= 0.5 [435]. In analyses of 13 TeV data PUPPI has become the standard pileup
mitigation technique, reinstating the role of grooming algorithms to their original
purpose of removing soft and wide-angle radiation. For this purpose, soft drop with
z cut = 0.1 and β = 0 is used in jet substructure analyses. The combination of PUPPI
and soft drop shows a similar performance as pruning in substructure analyses [436,
459]. The soft drop algorithm has been chosen over pruning due to a better theoretical
control of substructure observables.
The role of grooming algorithms in the performance of W , Z , H and t taggers is
discussed in detail in the respective sections below.
3.5 Jet Substructure Tagging
Particle identification is an experimental challenge that is traditionally met using
charged-particle detectors, pre-shower detectors, high-granularity calorimeters and
muon chambers. Particle identification of electrons, muons, tau leptons, photons and
jets originating from b quark fragmentation processes played an important role in
the design considerations for the ATLAS and CMS detectors and is used extensively in physics analyses. Jet substructure techniques used for the identification of
73
The ATLAS experiment has adopted trimming for anti-k T , R = 1.0 jets with f cut
= 0.05 and R sub = 0.3 in analyses of 8 TeV data, and R sub = 0.2 for analyses of
13 TeV data. Extensive studies of different algorithms showed that this choice gives
the best results in terms of pileup stability, jet mass resolution and jet tagging performance [452]. The reclustering of small-R jets [431] has the experimental advantage
of using fully calibrated inputs when clustering jets with a larger distance parameter,
and has already been used in several ATLAS searches for Supersymmetry [456–458].
This method allows for flexibility of optimizing the jet distance parameter depending on the considered phase-space of the analysis without a need to re-derive jet
energy corrections [432]. Note that this method has similarities to trimming with
R sub set to the distance parameter of the small-R jets. The difference is in the value
of p T,sub , which is fixed to the minimum p T of the small-R jets for reclustering, but
dynamically adjusted to the p T of the large-R jet for trimming. The groomed jets
obtained with reclustering are different from trimmed jets also because reclustering
considers all small-R jets in an event, while trimming uses only the ones obtained
from the constituents of large-R jets. In practice these differences lead to small
effects only, resulting in a very similar performance of reclustering and trimming.
A recent ATLAS study [453] on grooming algorithms includes PF reconstruction
and a number of hybrid pileup removal techniques. It is shown that gains in tagging
performance are possible by using soft drop instead of trimming, when used with PF
and a combination of Constituent Subtraction and SoftKiller.
The CMS experiment has employed PF with CHS already early in the data taking,
which resulted in a small impact from pileup on jet substructure during the 8 TeV
data taking. Contributions from the neutral pileup component have been mitigated
using the pruning algorithm for anti-k T , R = 0.8 jets with z prune = 0.1 and R prune
= 0.5 [435]. In analyses of 13 TeV data PUPPI has become the standard pileup
mitigation technique, reinstating the role of grooming algorithms to their original
purpose of removing soft and wide-angle radiation. For this purpose, soft drop with
z cut = 0.1 and β = 0 is used in jet substructure analyses. The combination of PUPPI
and soft drop shows a similar performance as pruning in substructure analyses [436,
459]. The soft drop algorithm has been chosen over pruning due to a better theoretical
control of substructure observables.
The role of grooming algorithms in the performance of W , Z , H and t taggers is
discussed in detail in the respective sections below.
3.5 Jet Substructure Tagging
Particle identification is an experimental challenge that is traditionally met using
charged-particle detectors, pre-shower detectors, high-granularity calorimeters and
muon chambers. Particle identification of electrons, muons, tau leptons, photons and
jets originating from b quark fragmentation processes played an important role in
the design considerations for the ATLAS and CMS detectors and is used extensively in physics analyses. Jet substructure techniques used for the identification of