4.2 Measurements Using Jet Substructure
117
p T > 400 GeV increased by more than a factor of 10. The accumulated data in 2015,
corresponding only to 3.2 fb
−1 , lead already to an improved statistical precision for
these measurements. ATLAS has performed a differential measurement in the +jets
channel [653] using these data. The measurement uses a very loose top tagging
selection based on τ 32 and trimmed jet mass, with varying selection criteria as a
function of p T , resulting in an approximately flat top tagging efficiency of 80%
for p T > 400 GeV. The obtained cross sections at the particle level are overall
well modelled by the SM predictions. The least-well described distribution is the
p T spectrum of the leading- p T top quark, which is observed to be softer than the
predictions, similar to the observations in 8 TeV data.
The large dataset collected in the years 2015–2018 offers unprecedented precision
for studies at high top quark p T . Recently, first analyses using about 36 fb
−1 of data
have been completed, corresponding to an increase in the number of events at high
p T of more than a factor of 15 in relation to the 8 TeV measurements. Both, ATLAS
and CMS have performed measurements in the +jets and all-hadronic final states.
In ATLAS, the all-hadronic analysis [654] uses the same top tagging algorithm as the
+jets analysis based on 3.2 fb
−1 [653], but with tighter selection criteria, because
of the higher background levels. The resulting top tagging efficiency is 50% for
p T > 500 GeV, with a misidentification rate for light quark and gluon jets of 6%
at p T = 500 GeV, increasing to 10% at p T = 1000 GeV. To further suppress nontop SM backgrounds, b tagging on small-R jets, matched to the large-R jets, is
used. In the +jets channel [655], background suppression is not as crucial, such
that a compromise between background rejection and modelling uncertainties is
made in order to achieve the best total precision. The analysis uses reclustered largeR jets [431], improving the precision of the jet reconstruction by using calibrated
inputs to the clustering. The trimmed mass of the reclustered jet is used for t tagging,
where the selection of 120 < m jet < 220 GeV results in an efficiency of about 60%.
This results in a total non-tt background of about 15%, which is estimated mostly
from simulation. Only a small contribution from multijet production is estimated
from a control region with a looser lepton selection. The unfolded results in the
all-hadronic and +jets channels are consistent; the shapes of distributions are well
described by NLO+PS calculations, but the measured total cross section is lower than
the prediction by about 30% for p T > 500 GeV. The large dataset also allows for
the first time to perform double-differential cross section measurements in boosted
final states. State-of-the-art fixed-order NNLO calculations [651, 656] describe the
data very well, as shown in Fig. 4.12, where the ratio of data to predictions at NNLO
accuracy are shown for measurements in resolved and boosted final states in the +jets
channel [655], also highlighting the gain from the boosted measurement. The NNLO
prediction provides a very good description of the data from top quarks produced at
rest up to p T = 2 TeV. This provides a very stringent test of the SM description of
tt production. Measurements of differential cross sections for boosted tt production
have also been reported by CMS in the +jets and all-hadronic channels using 36 fb
−1
of data [657]. In the all-hadronic channel, a selection on the soft drop jet mass is used
of 120 < m jet < 220 GeV, identical to the trimmed mass selection in the ATLAS
analysis. In addition, a NN has been trained on tt and QCD jets with τ 1 , τ 2 and τ 3 as
117
p T > 400 GeV increased by more than a factor of 10. The accumulated data in 2015,
corresponding only to 3.2 fb
−1 , lead already to an improved statistical precision for
these measurements. ATLAS has performed a differential measurement in the +jets
channel [653] using these data. The measurement uses a very loose top tagging
selection based on τ 32 and trimmed jet mass, with varying selection criteria as a
function of p T , resulting in an approximately flat top tagging efficiency of 80%
for p T > 400 GeV. The obtained cross sections at the particle level are overall
well modelled by the SM predictions. The least-well described distribution is the
p T spectrum of the leading- p T top quark, which is observed to be softer than the
predictions, similar to the observations in 8 TeV data.
The large dataset collected in the years 2015–2018 offers unprecedented precision
for studies at high top quark p T . Recently, first analyses using about 36 fb
−1 of data
have been completed, corresponding to an increase in the number of events at high
p T of more than a factor of 15 in relation to the 8 TeV measurements. Both, ATLAS
and CMS have performed measurements in the +jets and all-hadronic final states.
In ATLAS, the all-hadronic analysis [654] uses the same top tagging algorithm as the
+jets analysis based on 3.2 fb
−1 [653], but with tighter selection criteria, because
of the higher background levels. The resulting top tagging efficiency is 50% for
p T > 500 GeV, with a misidentification rate for light quark and gluon jets of 6%
at p T = 500 GeV, increasing to 10% at p T = 1000 GeV. To further suppress nontop SM backgrounds, b tagging on small-R jets, matched to the large-R jets, is
used. In the +jets channel [655], background suppression is not as crucial, such
that a compromise between background rejection and modelling uncertainties is
made in order to achieve the best total precision. The analysis uses reclustered largeR jets [431], improving the precision of the jet reconstruction by using calibrated
inputs to the clustering. The trimmed mass of the reclustered jet is used for t tagging,
where the selection of 120 < m jet < 220 GeV results in an efficiency of about 60%.
This results in a total non-tt background of about 15%, which is estimated mostly
from simulation. Only a small contribution from multijet production is estimated
from a control region with a looser lepton selection. The unfolded results in the
all-hadronic and +jets channels are consistent; the shapes of distributions are well
described by NLO+PS calculations, but the measured total cross section is lower than
the prediction by about 30% for p T > 500 GeV. The large dataset also allows for
the first time to perform double-differential cross section measurements in boosted
final states. State-of-the-art fixed-order NNLO calculations [651, 656] describe the
data very well, as shown in Fig. 4.12, where the ratio of data to predictions at NNLO
accuracy are shown for measurements in resolved and boosted final states in the +jets
channel [655], also highlighting the gain from the boosted measurement. The NNLO
prediction provides a very good description of the data from top quarks produced at
rest up to p T = 2 TeV. This provides a very stringent test of the SM description of
tt production. Measurements of differential cross sections for boosted tt production
have also been reported by CMS in the +jets and all-hadronic channels using 36 fb
−1
of data [657]. In the all-hadronic channel, a selection on the soft drop jet mass is used
of 120 < m jet < 220 GeV, identical to the trimmed mass selection in the ATLAS
analysis. In addition, a NN has been trained on tt and QCD jets with τ 1 , τ 2 and τ 3 as