5.7 Supersymmetry
203
signal purity up to a factor of four for events with an average pileup of 35. Searches
including a → bb decays have been reported by ATLAS [1207] and CMS [1208] in
2μ2b final states. Resolved a decays are considered in these analyses, which have
been optimised for 20 < m a < 62.5 GeV. Smaller masses are not accessible in these
searches, which would need boosted reconstruction techniques. The first search in
the 2τ 2b channel has been reported by CMS [1209], considering eτ h , μτ h and eμ
final states and resolved b decays. Final states with τ h τ h are not considered because
of high τ h trigger thresholds. The large branching fractions for a → bb and a → τ τ
result in upper limits as low as 6% on B(H → aa) in certain BSM scenarios from this
analysis alone. The 4b final state is not accessible in gluon-gluon fusion production at
the LHC, because of the overwhelming background from multijet production. However, it can be analysed in VH production, with leptonic V decays. The first search
of this kind has been performed by ATLAS with 36.1 fb
−1 of 13 TeV data [1210].
Resolved a decays are targeted, reconstructed using small-R jets with p T > 20 GeV.
The signal is enriched using BDTs in signal regions categorised by the lepton, jet and
b-jet multiplicity. The analysis is sensitive to m a in the range 20–60 GeV, but the best
sensitivity is achieved for m a ≈ 30 GeV. Below this value, the acceptance decreases
because of overlapping b jets. This is remedied in a recent analysis by ATLAS, using
the same dataset, but considering collimated a → bb decays, optimised for m a in
the range 15–30 GeV [1211]. This signature needs a new identification method for
boosted a → bb jets, because of their much smaller p T compared to typical boosted
signatures. Large-R jets with R = 0.8 are obtained from a reclustering of small-R
jets. The corresponding ghost-associated tracks are clustered using the exclusive
k T algorithm, returning either two or three track-jets. These serve as proxies for the
flight directions of the two b quarks, where three track-jets are considered to capture
cases with significant additional radiation. The use of exclusive k T clustering instead
of inclusive k T clustering, as used for the identification of H → bb jets, results in
an improved matching of b quarks to track-jets. While the inclusive clustering finds
correct matches only in 46% of the cases for m a = 20 GeV, the exclusive clustering
has a success rate of nearly 100%. A BDT is trained to identify a → bb jets, using
b-tagging information calculated from the track-jets, their angular separation R,
and their p T asymmetry. The BDT achieves signal efficiencies of 25% and 35%
for background rejections of 1% and 2.6%, respectively, where the latter is derived
for b jets from tt decays. Efficiencies and corrections for simulated events are measured in a multijet sample enriched in g → bb splittings. Since the distributions of
the BDT input variables can change due to the non-zero a mass, the measurement
is repeated using input distributions as expected from the decay a → bb instead
of distributions from a massless gluon. Within the statistical uncertainties of the
measurement, the correction factors are found to be independent of m a . Two signal
regions are constructed to test for H → aa decays, defined by high purity and low
purity a → bb jets. The sensitivity achieved by this analysis improves the sensitivity
at m a = 20 GeV by a factor of 2.5 compared to the resolved 4b analysis [1210].
This highlights the potential gain from jet substructure methods in future analyses
in a → bb decays.
203
signal purity up to a factor of four for events with an average pileup of 35. Searches
including a → bb decays have been reported by ATLAS [1207] and CMS [1208] in
2μ2b final states. Resolved a decays are considered in these analyses, which have
been optimised for 20 < m a < 62.5 GeV. Smaller masses are not accessible in these
searches, which would need boosted reconstruction techniques. The first search in
the 2τ 2b channel has been reported by CMS [1209], considering eτ h , μτ h and eμ
final states and resolved b decays. Final states with τ h τ h are not considered because
of high τ h trigger thresholds. The large branching fractions for a → bb and a → τ τ
result in upper limits as low as 6% on B(H → aa) in certain BSM scenarios from this
analysis alone. The 4b final state is not accessible in gluon-gluon fusion production at
the LHC, because of the overwhelming background from multijet production. However, it can be analysed in VH production, with leptonic V decays. The first search
of this kind has been performed by ATLAS with 36.1 fb
−1 of 13 TeV data [1210].
Resolved a decays are targeted, reconstructed using small-R jets with p T > 20 GeV.
The signal is enriched using BDTs in signal regions categorised by the lepton, jet and
b-jet multiplicity. The analysis is sensitive to m a in the range 20–60 GeV, but the best
sensitivity is achieved for m a ≈ 30 GeV. Below this value, the acceptance decreases
because of overlapping b jets. This is remedied in a recent analysis by ATLAS, using
the same dataset, but considering collimated a → bb decays, optimised for m a in
the range 15–30 GeV [1211]. This signature needs a new identification method for
boosted a → bb jets, because of their much smaller p T compared to typical boosted
signatures. Large-R jets with R = 0.8 are obtained from a reclustering of small-R
jets. The corresponding ghost-associated tracks are clustered using the exclusive
k T algorithm, returning either two or three track-jets. These serve as proxies for the
flight directions of the two b quarks, where three track-jets are considered to capture
cases with significant additional radiation. The use of exclusive k T clustering instead
of inclusive k T clustering, as used for the identification of H → bb jets, results in
an improved matching of b quarks to track-jets. While the inclusive clustering finds
correct matches only in 46% of the cases for m a = 20 GeV, the exclusive clustering
has a success rate of nearly 100%. A BDT is trained to identify a → bb jets, using
b-tagging information calculated from the track-jets, their angular separation R,
and their p T asymmetry. The BDT achieves signal efficiencies of 25% and 35%
for background rejections of 1% and 2.6%, respectively, where the latter is derived
for b jets from tt decays. Efficiencies and corrections for simulated events are measured in a multijet sample enriched in g → bb splittings. Since the distributions of
the BDT input variables can change due to the non-zero a mass, the measurement
is repeated using input distributions as expected from the decay a → bb instead
of distributions from a massless gluon. Within the statistical uncertainties of the
measurement, the correction factors are found to be independent of m a . Two signal
regions are constructed to test for H → aa decays, defined by high purity and low
purity a → bb jets. The sensitivity achieved by this analysis improves the sensitivity
at m a = 20 GeV by a factor of 2.5 compared to the resolved 4b analysis [1210].
This highlights the potential gain from jet substructure methods in future analyses
in a → bb decays.