5.3 Vector-Like Quarks
159
by fully reconstructing the event kinematics, thereby allowing for a reconstruction
of the B mass. The analysis is carried out in three different categories, defined by
the small-R jet multiplicity with four, five or six jets, where two, three or four have
to be b-tagged, respectively. In the categories with four or five jets, large-R jets are
used to reconstruct the boosted H or Z , where one is required to be identified with
the double b tagger. Events are reconstructed and sorted into three categories corresponding to the decay modes bHbH, bHbZ and bZbZ with a χ
2 estimator. The
estimator is built for all possible jet combinations and compares the reconstructed Z
and H masses with the expectations from simulation, as well as the mass difference
between the two VLQs in the event. The multijet background is obtained by a fit
of an exponential function to events in the three categories, before b- and double
b-tagging selections are applied. This sample has a high rate of multijet background
events, with a signal-to-background ratio more than two orders of magnitude smaller
than in the signal regions. The exponential function is observed to describe the shape
of the reconstructed VLQ mass distribution well for masses greater than 1 TeV. The
parametrised shape is then multiplied by the misidentification rate of the b tagging
requirements, obtained in a control region with small reconstructed VLQ mass. The
misidentification rates are checked in validation regions with large values of the χ
2
estimator. The final result of this analysis consists of nine measured VLQ mass distributions for the three jet multiplicity regions and three decay modes. The highest
sensitivity is obtained from the 4-jet, bHbH category. The limits on the B mass
are improved to 1570 GeV in the B(B → Hb) = 1 case, and to 1390 GeV in the
B(B → Zb) = 1 case. The large improvement of about 400 GeV with respect to the
all-hadronic analyses stems mostly from the specialisation to the particular decay
mode, which allows for a reconstruction of m VLQ . The larger data sample, where this
analysis uses about four times the data, plays a role as well, but a smaller one. A
naive extrapolation of the improvement in expected sensitivity of the two all-hadronic
searches results in a higher mass limit by about 100 GeV from the increase in luminosity only. While the previous all-hadronic analyses use advanced machine learning
methods, this analysis accomplishes high sensitivity without multi-dimensional classification algorithms. This shows that in some cases it is advantageous to search for
a particular signature, instead of performing an inclusive analysis aiming at a large
number of channels.
The LHC results with 13 TeV data move the allowed masses of all VLQs considered, B, T , X and Y , to values beyond 1.3 TeV. On the one hand, this is a remarkable
achievement which relies very much on the development and commissioning of jet
substructure methods by the experimental collaborations. On the other hand, this
means that some BSM models where VLQs are an integral component become more
unlikely to be realised in nature. However, non-minimal models, for example models
with an additional scalar [890–892], a new quantum number which is conserved for
VLQs [893], a new gauge symmetry group [894], or models with VLQs embedded
in a two-Higgs-doublet model [895] can lead to significantly lower bounds on VLQ
masses because of the presence of additional decay modes. There is still room left
for these ideas to be explored theoretically and experimentally, which will also lead
to new search strategies for the pair production of VLQs at the LHC.
159
by fully reconstructing the event kinematics, thereby allowing for a reconstruction
of the B mass. The analysis is carried out in three different categories, defined by
the small-R jet multiplicity with four, five or six jets, where two, three or four have
to be b-tagged, respectively. In the categories with four or five jets, large-R jets are
used to reconstruct the boosted H or Z , where one is required to be identified with
the double b tagger. Events are reconstructed and sorted into three categories corresponding to the decay modes bHbH, bHbZ and bZbZ with a χ
2 estimator. The
estimator is built for all possible jet combinations and compares the reconstructed Z
and H masses with the expectations from simulation, as well as the mass difference
between the two VLQs in the event. The multijet background is obtained by a fit
of an exponential function to events in the three categories, before b- and double
b-tagging selections are applied. This sample has a high rate of multijet background
events, with a signal-to-background ratio more than two orders of magnitude smaller
than in the signal regions. The exponential function is observed to describe the shape
of the reconstructed VLQ mass distribution well for masses greater than 1 TeV. The
parametrised shape is then multiplied by the misidentification rate of the b tagging
requirements, obtained in a control region with small reconstructed VLQ mass. The
misidentification rates are checked in validation regions with large values of the χ
2
estimator. The final result of this analysis consists of nine measured VLQ mass distributions for the three jet multiplicity regions and three decay modes. The highest
sensitivity is obtained from the 4-jet, bHbH category. The limits on the B mass
are improved to 1570 GeV in the B(B → Hb) = 1 case, and to 1390 GeV in the
B(B → Zb) = 1 case. The large improvement of about 400 GeV with respect to the
all-hadronic analyses stems mostly from the specialisation to the particular decay
mode, which allows for a reconstruction of m VLQ . The larger data sample, where this
analysis uses about four times the data, plays a role as well, but a smaller one. A
naive extrapolation of the improvement in expected sensitivity of the two all-hadronic
searches results in a higher mass limit by about 100 GeV from the increase in luminosity only. While the previous all-hadronic analyses use advanced machine learning
methods, this analysis accomplishes high sensitivity without multi-dimensional classification algorithms. This shows that in some cases it is advantageous to search for
a particular signature, instead of performing an inclusive analysis aiming at a large
number of channels.
The LHC results with 13 TeV data move the allowed masses of all VLQs considered, B, T , X and Y , to values beyond 1.3 TeV. On the one hand, this is a remarkable
achievement which relies very much on the development and commissioning of jet
substructure methods by the experimental collaborations. On the other hand, this
means that some BSM models where VLQs are an integral component become more
unlikely to be realised in nature. However, non-minimal models, for example models
with an additional scalar [890–892], a new quantum number which is conserved for
VLQs [893], a new gauge symmetry group [894], or models with VLQs embedded
in a two-Higgs-doublet model [895] can lead to significantly lower bounds on VLQ
masses because of the presence of additional decay modes. There is still room left
for these ideas to be explored theoretically and experimentally, which will also lead
to new search strategies for the pair production of VLQs at the LHC.