198
5 Direct Searches for New Physics
are excluded for a light ˜
χ
0
1 . For a ˜
t mass of 1 TeV, ˜
χ
0
1 masses below 600 GeV are
excluded. An ATLAS search in the all-hadronic final state, using 139 fb
−1 [1165] and
targeting the t ¯
t+ p
miss
T
signature, uses the mass of reclustered R = 0.8 and R = 1.2
jets to select events with fully-merged and partially-merged t decays. The sensitivity
of the analysis is comparable to the CMS search in the +jets channel [1164], where
˜
t masses below 1.25 TeV are excluded for ˜
χ
0
1 masses below 200 GeV.
Top squark production from the decays of gluinos is another example of SUSY
searches where jet substructure techniques help to improve the sensitivity. One of the
first searches in this channel in the all-hadronic final state considering W tagging has
been performed by CMS using 19.7 fb
−1 of 8 TeV data [1166]. This search has been
optimised for ˜
g pair production for a compressed spectrum with m ≈ m t or smaller,
in the decay channel ˜
g → ˜
tt. At
√
s = 8 TeV, the ˜
g mass reach is below 1 TeV, such
that semi-merged t decays are considered and identified with a large-R W jet and an
accompanying small-R b jet. Tagged W jets serve as a means to select events, where
small-R jets are used to build razor variables to discriminate signal from background.
The SM backgrounds are obtained from control regions, defined by inverting the
τ 21 selection used for W tagging and other kinematic requirements. The analysis
excludes ˜
g masses below 700 GeV and ˜
t masses below 300 GeV for m = 175 GeV.
A CMS search using 2.3 fb
−1 of 13 TeV data [1167] considers triplets of smallR jets to construct resolved t decays [1168], as well as doublets and mono-jet
signatures to account for semi-merged and fully-merged t decays. While the tagging
efficiency of the algorithm is high, increasing from 30% at p T = 200 GeV to 80%
at p T = 1 TeV, the misidentification rate is also high, between 30–40% on average.
The reason for the large misidentification rate lies in the use of ungroomed small-R
jets, where the mass increases for background jets with increasing p T following the
behaviour of the Sudakov peak. Nevertheless, at high p
miss
T
it is beneficial to have
high signal efficiency, where the SM backgrounds are small. For models with gluinomediated ˜
t production pp → ˜
g ˜
g, with ˜
g → tt ˜
χ
0
1 , gluino masses up to 1550 GeV are
excluded. The deficiencies of the t identification have been addressed in a later CMS
analysis based on 35.9 fb
−1 [529]. The fully-merged and semi-merged t decays are
identified using large-R jets with soft drop grooming and N -subjettiness ratios. These
changes result in a reduction of a factor of about two in the average misidentification
rate, while retaining the signal efficiency. Consequently, the ˜
g mass limits could be
improved by nearly 500 GeV, masses below 2040 GeV are excluded for a light ˜
χ
0
1 .
This also corresponds to an improvement by about 100 GeV relative to searches in
the all-hadronic final state based on the same amount of data, but without top tagging
or jet substructure [1169, 1170].
In most SUSY decay cascades, a large number of jets is produced. These may arise
from gluino-mediated ˜
t production with a t ¯
tt ¯
t+2 ˜
χ
0
1 final state as discussed above,
from chargino and neutralino decays or gauge-mediated SUSY symmetry breaking
scenarios [1171–1173] with W , Z , H bosons in the final state, or from R-parity violating (RPV) [1174, 1175] decays with SUSY particle decays to SM quarks. Boosted
heavy SM particles are produced in scenarios with large mass splittings between the
SUSY particles produced in the primary interaction, the LSP, and the SM particles.
A way to discriminate between these signals and SM background processes is the
5 Direct Searches for New Physics
are excluded for a light ˜
χ
0
1 . For a ˜
t mass of 1 TeV, ˜
χ
0
1 masses below 600 GeV are
excluded. An ATLAS search in the all-hadronic final state, using 139 fb
−1 [1165] and
targeting the t ¯
t+ p
miss
T
signature, uses the mass of reclustered R = 0.8 and R = 1.2
jets to select events with fully-merged and partially-merged t decays. The sensitivity
of the analysis is comparable to the CMS search in the +jets channel [1164], where
˜
t masses below 1.25 TeV are excluded for ˜
χ
0
1 masses below 200 GeV.
Top squark production from the decays of gluinos is another example of SUSY
searches where jet substructure techniques help to improve the sensitivity. One of the
first searches in this channel in the all-hadronic final state considering W tagging has
been performed by CMS using 19.7 fb
−1 of 8 TeV data [1166]. This search has been
optimised for ˜
g pair production for a compressed spectrum with m ≈ m t or smaller,
in the decay channel ˜
g → ˜
tt. At
√
s = 8 TeV, the ˜
g mass reach is below 1 TeV, such
that semi-merged t decays are considered and identified with a large-R W jet and an
accompanying small-R b jet. Tagged W jets serve as a means to select events, where
small-R jets are used to build razor variables to discriminate signal from background.
The SM backgrounds are obtained from control regions, defined by inverting the
τ 21 selection used for W tagging and other kinematic requirements. The analysis
excludes ˜
g masses below 700 GeV and ˜
t masses below 300 GeV for m = 175 GeV.
A CMS search using 2.3 fb
−1 of 13 TeV data [1167] considers triplets of smallR jets to construct resolved t decays [1168], as well as doublets and mono-jet
signatures to account for semi-merged and fully-merged t decays. While the tagging
efficiency of the algorithm is high, increasing from 30% at p T = 200 GeV to 80%
at p T = 1 TeV, the misidentification rate is also high, between 30–40% on average.
The reason for the large misidentification rate lies in the use of ungroomed small-R
jets, where the mass increases for background jets with increasing p T following the
behaviour of the Sudakov peak. Nevertheless, at high p
miss
T
it is beneficial to have
high signal efficiency, where the SM backgrounds are small. For models with gluinomediated ˜
t production pp → ˜
g ˜
g, with ˜
g → tt ˜
χ
0
1 , gluino masses up to 1550 GeV are
excluded. The deficiencies of the t identification have been addressed in a later CMS
analysis based on 35.9 fb
−1 [529]. The fully-merged and semi-merged t decays are
identified using large-R jets with soft drop grooming and N -subjettiness ratios. These
changes result in a reduction of a factor of about two in the average misidentification
rate, while retaining the signal efficiency. Consequently, the ˜
g mass limits could be
improved by nearly 500 GeV, masses below 2040 GeV are excluded for a light ˜
χ
0
1 .
This also corresponds to an improvement by about 100 GeV relative to searches in
the all-hadronic final state based on the same amount of data, but without top tagging
or jet substructure [1169, 1170].
In most SUSY decay cascades, a large number of jets is produced. These may arise
from gluino-mediated ˜
t production with a t ¯
tt ¯
t+2 ˜
χ
0
1 final state as discussed above,
from chargino and neutralino decays or gauge-mediated SUSY symmetry breaking
scenarios [1171–1173] with W , Z , H bosons in the final state, or from R-parity violating (RPV) [1174, 1175] decays with SUSY particle decays to SM quarks. Boosted
heavy SM particles are produced in scenarios with large mass splittings between the
SUSY particles produced in the primary interaction, the LSP, and the SM particles.
A way to discriminate between these signals and SM background processes is the