116
4 Standard Model Measurements
p
rel
T because of the small hadron masses involved, while leptons from W decays
are kept. This 2D lepton selection shows better performance in boosted tt decays
against QCD multijet production compared to mini-isolation. With the 2D selection,
a QCD rejection rate of more than 99% is achieved for a selection efficiency of 91%
for muons, compared to a rejection rate of 60% for mini-isolation at the same efficiency [650]. In this measurement, boosted hadronic t decays are identified using the
CMSTT for large-R jets with p T > 400 GeV. The dominant sources of background
are W +jets, single top quark in the t W channel and multijet production. The contribution from the latter is derived from a control region, obtained by inverting the 2D
lepton selection. The normalisations of the other backgrounds are derived through a
maximum-likelihood fit, where their shapes are modelled using simulation. In this
fit, also the tt signal yield and t tagging efficiency are obtained simultaneously. The
cross sections, unfolded at the particle level, show a similar trend as in the ATLAS
measurement, where the NLO+PS prediction from Powheg and the multi-leg prediction from Madgraph have a harder p T spectrum. A better description of the data
up to the highest p T of 1200 GeV is obtained with MC@NLO interfaced to Herwig
for the PS. In addition, the data at the parton level are compared to fixed-order predictions in perturbative QCD at NNLO [651] and approximate next-to-next-to-next-toleading-order (aNNNLO) [620] accuracy. While the NNLO calculation shows very
good agreement with the data, the aNNNLO calculation predicts a significantly too
hard p T spectrum.
A particular challenge is the measurement of tt production at high top quark
p T in the all-hadronic channel, which resembles a dijet topology. The overwhelming amount of background from QCD multijet production is difficult to model and
technically unfeasible to simulate with accurate statistical precision. Hence, jet substructure methods enabling a background estimation from data are the only possibility
to carry out measurements in this channel at the LHC. In a CMS analysis using 8 TeV
data [652], the tt signal is identified by requiring two subjet b-tagged large-R jets
with p T > 400 GeV, where the leading jet in p T has to have a three prong structure
with τ 32 < 0.55. The distribution in the pruned jet mass of the leading jet is used to
perform a maximum likelihood fit to estimate the tt cross section in bins of p T . The τ 32
selection, in conjunction with p T > 400 GeV, leads to a sculpting of the m jet distribution for light quark and gluon jets, resulting in a peak approximately at the same
position as the signal peak. In order to reliably estimate the multijet background,
sidebands in the jet mass with 100 < m jet < 140 GeV and 250 < m jet < 400 GeV
are used, as well as events failing the τ 32 requirement. With the help of these sideband regions, the multijet background can be interpolated into the signal region with
140 < m jet < 250 GeV, taking correlations between m jet and τ 32 fully into account.
In order to achieve higher accuracy, also events with zero and one subjet b tag are
considered in this analysis. The final differential cross section unfolded at the parton
level agrees well with the results obtained in the +jets channel, albeit with larger
uncertainties by factors of about two to three.
The increase of
√
s from 8 to 13 TeV in 2015 has offered a unique possibility
for tt measurements in the highly boosted regime. While the total cross section for
tt production increased by a factor of three, the cross cross section for top quark
4 Standard Model Measurements
p
rel
T because of the small hadron masses involved, while leptons from W decays
are kept. This 2D lepton selection shows better performance in boosted tt decays
against QCD multijet production compared to mini-isolation. With the 2D selection,
a QCD rejection rate of more than 99% is achieved for a selection efficiency of 91%
for muons, compared to a rejection rate of 60% for mini-isolation at the same efficiency [650]. In this measurement, boosted hadronic t decays are identified using the
CMSTT for large-R jets with p T > 400 GeV. The dominant sources of background
are W +jets, single top quark in the t W channel and multijet production. The contribution from the latter is derived from a control region, obtained by inverting the 2D
lepton selection. The normalisations of the other backgrounds are derived through a
maximum-likelihood fit, where their shapes are modelled using simulation. In this
fit, also the tt signal yield and t tagging efficiency are obtained simultaneously. The
cross sections, unfolded at the particle level, show a similar trend as in the ATLAS
measurement, where the NLO+PS prediction from Powheg and the multi-leg prediction from Madgraph have a harder p T spectrum. A better description of the data
up to the highest p T of 1200 GeV is obtained with MC@NLO interfaced to Herwig
for the PS. In addition, the data at the parton level are compared to fixed-order predictions in perturbative QCD at NNLO [651] and approximate next-to-next-to-next-toleading-order (aNNNLO) [620] accuracy. While the NNLO calculation shows very
good agreement with the data, the aNNNLO calculation predicts a significantly too
hard p T spectrum.
A particular challenge is the measurement of tt production at high top quark
p T in the all-hadronic channel, which resembles a dijet topology. The overwhelming amount of background from QCD multijet production is difficult to model and
technically unfeasible to simulate with accurate statistical precision. Hence, jet substructure methods enabling a background estimation from data are the only possibility
to carry out measurements in this channel at the LHC. In a CMS analysis using 8 TeV
data [652], the tt signal is identified by requiring two subjet b-tagged large-R jets
with p T > 400 GeV, where the leading jet in p T has to have a three prong structure
with τ 32 < 0.55. The distribution in the pruned jet mass of the leading jet is used to
perform a maximum likelihood fit to estimate the tt cross section in bins of p T . The τ 32
selection, in conjunction with p T > 400 GeV, leads to a sculpting of the m jet distribution for light quark and gluon jets, resulting in a peak approximately at the same
position as the signal peak. In order to reliably estimate the multijet background,
sidebands in the jet mass with 100 < m jet < 140 GeV and 250 < m jet < 400 GeV
are used, as well as events failing the τ 32 requirement. With the help of these sideband regions, the multijet background can be interpolated into the signal region with
140 < m jet < 250 GeV, taking correlations between m jet and τ 32 fully into account.
In order to achieve higher accuracy, also events with zero and one subjet b tag are
considered in this analysis. The final differential cross section unfolded at the parton
level agrees well with the results obtained in the +jets channel, albeit with larger
uncertainties by factors of about two to three.
The increase of
√
s from 8 to 13 TeV in 2015 has offered a unique possibility
for tt measurements in the highly boosted regime. While the total cross section for
tt production increased by a factor of three, the cross cross section for top quark