5.1 Diboson Resonances
133
a function of m jet to estimate the multijet background. The results are statistically
combined with the fully merged analysis, leading to a significant improvement in the
search sensitivity for resonances with masses between 750 and 2000 GeV. In addition,
signals from the non-resonant production of H H are also accessible, which is not
the case for fully merged final states since such production typically results in an
H H invariant mass that is lower than that of a postulated resonance signal.
Another channel considered is H H → bbτ τ [728], which has been described
above in the context of VH searches. After reconstructing one H candidate through
collimated τ h or τ h τ h decays, the other H candidate is obtained by a requirement
on the soft drop jet mass, 105 < m jet < 135 GeV, and the presence of one or two
b-tagged subjets. Due to the large branching fraction B(H → τ τ ) = 6.3%, in combination with a clean signature, resulting in background events mostly from V +jets
and top quark production, this channel is comparable in sensitivity at high resonance
masses to all-hadronic searches in the H H → bbbb channel. ATLAS has developed
a new reconstruction method for boosted τ h τ h pairs and employed it in a search
for H H → bbτ τ using 139 fb
−1 of 13 TeV data [747]. The constituents of large-R
jets with p T > 300 GeV are clustered into anti-k T subjets with R = 0.2. The two
p T -leading subjets are used to identify the di-τ system, where information from
the calorimeter clusters, tracks and vertices associated with the subjets are used as
input to a BDT. This reconstruction achieves signal efficiencies of about 60% for a
background rejection of 10
4 . The analysis is not as sensitive as the CMS analysis
using 35.9 fb
−1 [728], because leptonic τ decays are not considered. However, the
new boosted τ h τ h reconstruction promises to improve future searches in the H → τ τ
channel.
A new channel in H H searches has been very recently explored by CMS, targeting
the decay H H → bbW W
(∗) , with the subsequent decay W W
(∗)
→ qqν [748].
Both H bosons are reconstructed in a single large-R jet, but in the case of the W W
(∗)
decay, a qq jet with a nearby lepton is required. The lepton is identified using miniisolation, which performs better for boosted qqν decays than the two-dimensional
selection based on p
rel
T and R used in top quark analyses (see Sect. 4.2.3). The qq
jet is required to have a small angular separation to the lepton with R < 1.2, and
two subjets with p T > 20 GeV, obtained from the soft drop algorithm. The qq jet has
to satisfy 0.55 < τ 21 < 0.75 for the LP category and τ 21 < 0.55 for the HP category.
Note that no mass selection is imposed on this jet, as the W initiating this jet may
be off-shell. The bb H jet is classified based on the subjet b tagging discriminator
values. The analysis is conducted via a maximum likelihood fit to the m jet –m H H
mass plane, where m jet denotes the soft drop jet mass of the bb H jet (called m bb in
[748]). The background templates are modelled as conditional probabilities of m jet
as a function of m H H .
Since a parametric function that models the full m H H range for background events
is difficult to obtain, the templates are obtained from simulated events using the kernel
density estimation [749]. The background model is tested in two dedicated control
regions, one for tt and one for W +jets and QCD multijet production. The background
templates are fit to the two-dimensional mass planes in the control regions, with the
133
a function of m jet to estimate the multijet background. The results are statistically
combined with the fully merged analysis, leading to a significant improvement in the
search sensitivity for resonances with masses between 750 and 2000 GeV. In addition,
signals from the non-resonant production of H H are also accessible, which is not
the case for fully merged final states since such production typically results in an
H H invariant mass that is lower than that of a postulated resonance signal.
Another channel considered is H H → bbτ τ [728], which has been described
above in the context of VH searches. After reconstructing one H candidate through
collimated τ h or τ h τ h decays, the other H candidate is obtained by a requirement
on the soft drop jet mass, 105 < m jet < 135 GeV, and the presence of one or two
b-tagged subjets. Due to the large branching fraction B(H → τ τ ) = 6.3%, in combination with a clean signature, resulting in background events mostly from V +jets
and top quark production, this channel is comparable in sensitivity at high resonance
masses to all-hadronic searches in the H H → bbbb channel. ATLAS has developed
a new reconstruction method for boosted τ h τ h pairs and employed it in a search
for H H → bbτ τ using 139 fb
−1 of 13 TeV data [747]. The constituents of large-R
jets with p T > 300 GeV are clustered into anti-k T subjets with R = 0.2. The two
p T -leading subjets are used to identify the di-τ system, where information from
the calorimeter clusters, tracks and vertices associated with the subjets are used as
input to a BDT. This reconstruction achieves signal efficiencies of about 60% for a
background rejection of 10
4 . The analysis is not as sensitive as the CMS analysis
using 35.9 fb
−1 [728], because leptonic τ decays are not considered. However, the
new boosted τ h τ h reconstruction promises to improve future searches in the H → τ τ
channel.
A new channel in H H searches has been very recently explored by CMS, targeting
the decay H H → bbW W
(∗) , with the subsequent decay W W
(∗)
→ qqν [748].
Both H bosons are reconstructed in a single large-R jet, but in the case of the W W
(∗)
decay, a qq jet with a nearby lepton is required. The lepton is identified using miniisolation, which performs better for boosted qqν decays than the two-dimensional
selection based on p
rel
T and R used in top quark analyses (see Sect. 4.2.3). The qq
jet is required to have a small angular separation to the lepton with R < 1.2, and
two subjets with p T > 20 GeV, obtained from the soft drop algorithm. The qq jet has
to satisfy 0.55 < τ 21 < 0.75 for the LP category and τ 21 < 0.55 for the HP category.
Note that no mass selection is imposed on this jet, as the W initiating this jet may
be off-shell. The bb H jet is classified based on the subjet b tagging discriminator
values. The analysis is conducted via a maximum likelihood fit to the m jet –m H H
mass plane, where m jet denotes the soft drop jet mass of the bb H jet (called m bb in
[748]). The background templates are modelled as conditional probabilities of m jet
as a function of m H H .
Since a parametric function that models the full m H H range for background events
is difficult to obtain, the templates are obtained from simulated events using the kernel
density estimation [749]. The background model is tested in two dedicated control
regions, one for tt and one for W +jets and QCD multijet production. The background
templates are fit to the two-dimensional mass planes in the control regions, with the