38
2 Phenomenology of Jet Substructure
Trimming
Trimming [230] is similar to filtering. In its commonly used form, trimming reclusters all particles from a jet with distance parameter R into subjets with R sub . All
subjets that fulfil p T,sub > f cut p T,jet are kept and merged to form the trimmed jet,
where p T,sub and p T,jet are the transverse momenta of the subjet and the original jet,
respectively. All subjets not fulfilling this requirement are discarded.
9 Experimentally, an advantage of trimming in relation to filtering is that the number of maximum
subjets does not have to be specified, instead the parameter f cut controls the number of subjets dynamically. Trimming became standard in ATLAS, where it is used
to reduce the effects of pileup and the underlying event on anti-k T , R = 1.0 jets.
Typical parameter settings are R sub = 0.2 and f cut = 0.05.
Pruning
Pruning [236] reclusters all jet constituents, and in each step checks the angular
separation R i j < R prune and for a soft splitting min( p T,i , p T,j ) ≥ z prune p T,i+j . If
both criteria are met, pseudojets i and j are combined, otherwise only the hardest
of i and j are kept and the other is rejected. Pruning takes a different approach than
filtering and trimming. While in the latter cases a jet is reclustered with R sub < R
and soft subjets are discarded, pruning rejects pseudojets during the jet clustering. Another difference is that for pruning the radius R prune is adjusted dynamically
through R prune = m jet / p T,jet , while R sub is kept fixed for filtering and trimming. Here,
m jet and p T,jet are the mass and p T of the original jet. A particularity of pruning are
cases where soft, wide-angle radiation leads to large values of m jet and thus sets the
scale for R prune , while being rejected by the pruning conditions when reclustering
the jet. The mass of the pruned jet is then determined by radiation at a much smaller
angle and the jet shows a 1-prong structure, which is referred to as I-pruning [237].
Conversely, jets which feature at least one clustering step passing the pruning criteria
show a two-prong structure, referred to as Y-pruning. For practical applications the
difference between I- and Y-pruning has been shown to be mostly irrelevant, but it
is important for the analytical control of calculations for pruned jets.
Mass Drop Tagger
The mass drop tagger (MDT) [40] relies on the CA clustering, where it steps backwards through the clustering history. At each step, the mass drop condition
max(m i , m j ) < μ cut m i+ j
(2.39)
and the symmetry of the splitting,
min( p T,i
2
, p T,j
2
))R
2
i j > y cut m
2
i+ j ,
(2.40)
9 In its original version, trimming was formulated using a hard scale hard , instead of the original
jet’s transverse momentum p T,jet .
2 Phenomenology of Jet Substructure
Trimming
Trimming [230] is similar to filtering. In its commonly used form, trimming reclusters all particles from a jet with distance parameter R into subjets with R sub . All
subjets that fulfil p T,sub > f cut p T,jet are kept and merged to form the trimmed jet,
where p T,sub and p T,jet are the transverse momenta of the subjet and the original jet,
respectively. All subjets not fulfilling this requirement are discarded.
9 Experimentally, an advantage of trimming in relation to filtering is that the number of maximum
subjets does not have to be specified, instead the parameter f cut controls the number of subjets dynamically. Trimming became standard in ATLAS, where it is used
to reduce the effects of pileup and the underlying event on anti-k T , R = 1.0 jets.
Typical parameter settings are R sub = 0.2 and f cut = 0.05.
Pruning
Pruning [236] reclusters all jet constituents, and in each step checks the angular
separation R i j < R prune and for a soft splitting min( p T,i , p T,j ) ≥ z prune p T,i+j . If
both criteria are met, pseudojets i and j are combined, otherwise only the hardest
of i and j are kept and the other is rejected. Pruning takes a different approach than
filtering and trimming. While in the latter cases a jet is reclustered with R sub < R
and soft subjets are discarded, pruning rejects pseudojets during the jet clustering. Another difference is that for pruning the radius R prune is adjusted dynamically
through R prune = m jet / p T,jet , while R sub is kept fixed for filtering and trimming. Here,
m jet and p T,jet are the mass and p T of the original jet. A particularity of pruning are
cases where soft, wide-angle radiation leads to large values of m jet and thus sets the
scale for R prune , while being rejected by the pruning conditions when reclustering
the jet. The mass of the pruned jet is then determined by radiation at a much smaller
angle and the jet shows a 1-prong structure, which is referred to as I-pruning [237].
Conversely, jets which feature at least one clustering step passing the pruning criteria
show a two-prong structure, referred to as Y-pruning. For practical applications the
difference between I- and Y-pruning has been shown to be mostly irrelevant, but it
is important for the analytical control of calculations for pruned jets.
Mass Drop Tagger
The mass drop tagger (MDT) [40] relies on the CA clustering, where it steps backwards through the clustering history. At each step, the mass drop condition
max(m i , m j ) < μ cut m i+ j
(2.39)
and the symmetry of the splitting,
min( p T,i
2
, p T,j
2
))R
2
i j > y cut m
2
i+ j ,
(2.40)
9 In its original version, trimming was formulated using a hard scale hard , instead of the original
jet’s transverse momentum p T,jet .