5.1 Diboson Resonances
125
An improvement of the background estimation has been developed by CMS in
an all-hadronic search based on 77.3 fb
−1 of data [254]. Instead of searching for a
peak in the m jj distribution only, advantage is taken of the fact that the signal peaks
in three observables simultaneously: in m jj and in m jet of the two leading jets. The
full spectrum in m jet is fitted for the two leading jets, such that no selection in m jet is
needed. The V tag is based solely on the decorrelated τ
DDT
21 . This improves the signal
efficiency by about 20% and helps to better constrain the background shape because
of the higher statistical precision, especially at high m jj . In this approach, templates
for non-resonant and resonant backgrounds in m jet have to be built, where nonresonant contributions entail smoothly falling distributions in m jet , while resonant
contributions from W/Z +jets production and tt production will result in a peak in
one or both distributions of m jet . Since all backgrounds result in falling distributions
in m jj , this approach allows to constrain the main backgrounds by a fit of these
templates to data, while obtaining high sensitivity to resonant signals. This method
yields an improvement in sensitivity of up to 30% relative to the one-dimensional
fit in m jj . Even though this search is based on a dataset only half as large as the one
used by the most recent ATLAS search in this channel [705], the limits are better by
up to a factor of three at high masses, highlighting the power of this method.
Searches for diboson resonances in +jets final states can be performed in three
different channels, W V → ν J , Z V → J and Z V → νν J , where J stands for
the V -tagged large-R jet. Each of these comes with its particular advantages and
challenges. The ν J final state has the second-highest branching fraction for W W
decays with 17.0%, when including τ
−
→ e
−
¯
ν e ν τ and τ
−
→ μ
−
¯
ν μ ν τ decays and
their charge conjugates. Only the all-hadronic channel has a higher branching fraction
of 45.4%, but much higher background levels. The same is true for W Z resonances,
where the single-lepton final state has a branching fraction of 17.7% compared to
48.9% for the all-hadronic final state. The presence of an isolated lepton allows for
triggering these events with lepton triggers with thresholds as low as 25–30 GeV. This
results in lowest reconstructed diboson masses of about 500 GeV, extending the lower
reach of the all-hadronic analyses. For Z V resonances, the νν J final state has the
highest branching fractions after the all-hadronic final state, with B(Z Z → νν J ) =
14.0% and B(W Z → νν J ) = 13.5%. Recording these events is usually achieved
with triggers requiring significant missing transverse momentum, p
miss
T , which have
much higher thresholds than lepton triggers.
1 Dilepton signatures from Z V → J
decays have the smallest branching fractions with 4.7% for Z W and 4.9% for Z Z.
While dilepton triggers could be used for these events, usually single-lepton triggers
are employed. For these, trigger paths without lepton isolation requirements are
available, with the advantage of a stable efficiency if the two leptons from the
highly boosted Z decay are within each other’s isolation cone. The higher lepton p T
thresholds of these triggers compared to dilepton triggers with isolation requirements
is inconsequential for heavy resonance searches.
1 The p miss
T
denotes the magnitude of the two-component vector p miss
T , which is calculated as the
negative vectorial sum of the transverse momenta of electrons, muons, small-R jets, and unassociated
tracks in ATLAS [708], or of all PF candidates in CMS [709].
125
An improvement of the background estimation has been developed by CMS in
an all-hadronic search based on 77.3 fb
−1 of data [254]. Instead of searching for a
peak in the m jj distribution only, advantage is taken of the fact that the signal peaks
in three observables simultaneously: in m jj and in m jet of the two leading jets. The
full spectrum in m jet is fitted for the two leading jets, such that no selection in m jet is
needed. The V tag is based solely on the decorrelated τ
DDT
21 . This improves the signal
efficiency by about 20% and helps to better constrain the background shape because
of the higher statistical precision, especially at high m jj . In this approach, templates
for non-resonant and resonant backgrounds in m jet have to be built, where nonresonant contributions entail smoothly falling distributions in m jet , while resonant
contributions from W/Z +jets production and tt production will result in a peak in
one or both distributions of m jet . Since all backgrounds result in falling distributions
in m jj , this approach allows to constrain the main backgrounds by a fit of these
templates to data, while obtaining high sensitivity to resonant signals. This method
yields an improvement in sensitivity of up to 30% relative to the one-dimensional
fit in m jj . Even though this search is based on a dataset only half as large as the one
used by the most recent ATLAS search in this channel [705], the limits are better by
up to a factor of three at high masses, highlighting the power of this method.
Searches for diboson resonances in +jets final states can be performed in three
different channels, W V → ν J , Z V → J and Z V → νν J , where J stands for
the V -tagged large-R jet. Each of these comes with its particular advantages and
challenges. The ν J final state has the second-highest branching fraction for W W
decays with 17.0%, when including τ
−
→ e
−
¯
ν e ν τ and τ
−
→ μ
−
¯
ν μ ν τ decays and
their charge conjugates. Only the all-hadronic channel has a higher branching fraction
of 45.4%, but much higher background levels. The same is true for W Z resonances,
where the single-lepton final state has a branching fraction of 17.7% compared to
48.9% for the all-hadronic final state. The presence of an isolated lepton allows for
triggering these events with lepton triggers with thresholds as low as 25–30 GeV. This
results in lowest reconstructed diboson masses of about 500 GeV, extending the lower
reach of the all-hadronic analyses. For Z V resonances, the νν J final state has the
highest branching fractions after the all-hadronic final state, with B(Z Z → νν J ) =
14.0% and B(W Z → νν J ) = 13.5%. Recording these events is usually achieved
with triggers requiring significant missing transverse momentum, p
miss
T , which have
much higher thresholds than lepton triggers.
1 Dilepton signatures from Z V → J
decays have the smallest branching fractions with 4.7% for Z W and 4.9% for Z Z.
While dilepton triggers could be used for these events, usually single-lepton triggers
are employed. For these, trigger paths without lepton isolation requirements are
available, with the advantage of a stable efficiency if the two leptons from the
highly boosted Z decay are within each other’s isolation cone. The higher lepton p T
thresholds of these triggers compared to dilepton triggers with isolation requirements
is inconsequential for heavy resonance searches.
1 The p miss
T
denotes the magnitude of the two-component vector p miss
T , which is calculated as the
negative vectorial sum of the transverse momenta of electrons, muons, small-R jets, and unassociated
tracks in ATLAS [708], or of all PF candidates in CMS [709].