5.1 Diboson Resonances
123
5.1.1 WW, WZ and ZZ Resonances
In the all-hadronic decay mode, diboson resonances with masses beyond 1 TeV
appear as an highly-energetic dijet system. The only possibility to suppress the large
background from QCD jet production are jet substructure taggers. Since the jet mass
resolution is not sufficient to resolve W and Z jets, searches in the all-hadronic final
state can not distinguish between W W , W Z and Z Z resonances. An important aspect
of all-hadronic channels are trigger constraints. Because of the high rate of multijet
production, high p T thresholds have to be implemented in order to keep the rate at
which events are recorded and stored, at a manageable level. In ATLAS, the singlejet trigger required a jet to have p T > 360 GeV in 2015, which has been continually
raised to a final threshold of 440 GeV in 2018, due to the increasing instantaneous
luminosity. Besides single-jet triggers, CMS also employed triggers based on the
trimmed jet mass of large-R jets in order to lower the p T thresholds, while maintaining a similar rate. For example, an additional requirement m jet > 30 GeV allowed to
lower the p T threshold from 500 to 360 GeV for the single-jet trigger. Because of
these trigger thresholds, analyses using data with
√
s = 13 TeV usually start at dijet
masses of m jj > 1.1 TeV or higher, such that sensitivity to resonances with masses
above 1.2 TeV is obtained.
In the two most recent all-hadronic V V searches at 13 TeV by ATLAS [494,
705], a W/Z tagger is used based on D 2 , track multiplicity n trk , and trimmed jet
mass. In the earlier search, based on data corresponding to 36.7 fb
−1 [494], the
jet mass is reconstructed using the combined mass, which is a weighted average of
calorimeter and tracking measurements where the weights are inversely proportional
to the square of the jet mass resolution of the corresponding mass terms [427, 525]. In
the more recent analysis, based on the full available Run 2 data with 139 fb
−1 [705],
TCC [410] are used for the reconstruction of jet substructure variables. The resolution
in D 2 for TCC jets is largely improved compared to jets clustered from topoclusters,
where at p T ≈ 2 TeV an improvement by a factor of about two is achieved. For V
tagging, the improvement in D 2 resolution far outweighs the slight degradation in
mass resolution below 1.5 TeV (see Sect. 3.2). The variable n trk is used for additional
discrimination against gluon jets, which have a higher charge multiplicity on average
than quark jets (see Sect. 4.1.4). The V tagging selection based on m jet , D 2 and n trk
is optimised either for a constant signal efficiency of 50% [494] or for the best
overall sensitivity [705]. The latter approach results in an increasing efficiency from
about 20% at p T = 500 GeV to about 60% at p T = 4 TeV. This is achieved by p T -
dependent selection criteria on m jet , D 2 and n trk . The reason for the increase in
efficiency with increasing p T is the steeply falling background, which results in
less background events at high p T , such that a higher signal efficiency with smaller
background suppression is beneficial. The search is performed as a typical bumphunt [706], where a narrow resonance peak is expected to appear on a steeply falling
V -tagged dijet mass spectrum. Since the background originates dominantly from
dijet production, a background parametrisation identical to the ones used in dijet
searches can be used [707]. Both ATLAS searches have found the parametrisation
123
5.1.1 WW, WZ and ZZ Resonances
In the all-hadronic decay mode, diboson resonances with masses beyond 1 TeV
appear as an highly-energetic dijet system. The only possibility to suppress the large
background from QCD jet production are jet substructure taggers. Since the jet mass
resolution is not sufficient to resolve W and Z jets, searches in the all-hadronic final
state can not distinguish between W W , W Z and Z Z resonances. An important aspect
of all-hadronic channels are trigger constraints. Because of the high rate of multijet
production, high p T thresholds have to be implemented in order to keep the rate at
which events are recorded and stored, at a manageable level. In ATLAS, the singlejet trigger required a jet to have p T > 360 GeV in 2015, which has been continually
raised to a final threshold of 440 GeV in 2018, due to the increasing instantaneous
luminosity. Besides single-jet triggers, CMS also employed triggers based on the
trimmed jet mass of large-R jets in order to lower the p T thresholds, while maintaining a similar rate. For example, an additional requirement m jet > 30 GeV allowed to
lower the p T threshold from 500 to 360 GeV for the single-jet trigger. Because of
these trigger thresholds, analyses using data with
√
s = 13 TeV usually start at dijet
masses of m jj > 1.1 TeV or higher, such that sensitivity to resonances with masses
above 1.2 TeV is obtained.
In the two most recent all-hadronic V V searches at 13 TeV by ATLAS [494,
705], a W/Z tagger is used based on D 2 , track multiplicity n trk , and trimmed jet
mass. In the earlier search, based on data corresponding to 36.7 fb
−1 [494], the
jet mass is reconstructed using the combined mass, which is a weighted average of
calorimeter and tracking measurements where the weights are inversely proportional
to the square of the jet mass resolution of the corresponding mass terms [427, 525]. In
the more recent analysis, based on the full available Run 2 data with 139 fb
−1 [705],
TCC [410] are used for the reconstruction of jet substructure variables. The resolution
in D 2 for TCC jets is largely improved compared to jets clustered from topoclusters,
where at p T ≈ 2 TeV an improvement by a factor of about two is achieved. For V
tagging, the improvement in D 2 resolution far outweighs the slight degradation in
mass resolution below 1.5 TeV (see Sect. 3.2). The variable n trk is used for additional
discrimination against gluon jets, which have a higher charge multiplicity on average
than quark jets (see Sect. 4.1.4). The V tagging selection based on m jet , D 2 and n trk
is optimised either for a constant signal efficiency of 50% [494] or for the best
overall sensitivity [705]. The latter approach results in an increasing efficiency from
about 20% at p T = 500 GeV to about 60% at p T = 4 TeV. This is achieved by p T -
dependent selection criteria on m jet , D 2 and n trk . The reason for the increase in
efficiency with increasing p T is the steeply falling background, which results in
less background events at high p T , such that a higher signal efficiency with smaller
background suppression is beneficial. The search is performed as a typical bumphunt [706], where a narrow resonance peak is expected to appear on a steeply falling
V -tagged dijet mass spectrum. Since the background originates dominantly from
dijet production, a background parametrisation identical to the ones used in dijet
searches can be used [707]. Both ATLAS searches have found the parametrisation