106
4 Standard Model Measurements
objects using the full detector simulation [243, 435, 436, 459, 483, 522, 525, 525,
526, 530, 531]. Substructure measurements unfolded at the particle level are rare
for boosted W and t jets. So far, only a single measurement has been performed by
ATLAS using 13 TeV data [601]. In this measurement a number of jet substructure
distributions has been measured and compared to simulations for light quark and
gluon jets, W jets and t jets. Trimmed large-R jets are used for a selection of events,
before the original jet constituents are groomed with the soft drop algorithm. There
are three distinct selections, all using anti-k T R = 1.0 jets. Inclusive jets from a
dijet sample are selected with p T > 450 GeV and |η| < 2.5. Events from tt production in the lepton+jets final state are used to select W and t jets with |η| < 1.5
and p T > 200 and 350 GeV, respectively. The soft drop jet mass is required to be
60 < m jet < 100 GeV for W jets and m jet > 140 GeV for t jets. In addition, a condition of angular separation to an identified small-R b jet is required for W jets. In
the selection of t jets, the b jet has to overlap with the t jet. After this selection, the
kinematic phase space is sufficiently different for the inclusive, W and t jets such that
a direct comparison of substructure distributions between these is not meaningful.
However, comparisons with simulation are very useful to study the modelling of substructure observables. The observables studied are subjet multiplicity, Les Houches
angularity, ECF(2, 1), ECF(3, 1), e
(1)
2 , e
(1)
3 , C
(1,1)
2
, D
(1,1)
2
, τ 21 and τ 32 , where the N -
subjettiness ratios have been calculated with the winner-takes-all (WTA) axis [488].
The largest uncertainties in the unfolded distributions originate from modelling differences between different simulations used for the correction of detector effects.
These uncertainties are typically around 15% for W and t jets, highlighting the need
for a better understanding of differences in the simulations. The largest experimental uncertainties can be attributed to uncertainties in the cluster energy, derived with
the R trk method (see Sect. 3.2). These amount to about 8% for W and t jets and
are smaller for inclusive jets. Overall, the simulation describes the data within the
uncertainties, but some trends are visible. In Fig. 4.7 the measured distributions of
τ 21 for W jets and τ 32 for t jets are shown. These are important observables for W
and t tagging, used in several tagger studies and analyses. A potential mismodelling
of these distributions results in efficiency differences between data and simulation
for signal jets. It should be noted that soft drop groomed N -subjettiness ratios have
been measured, whereas ungroomed versions are usually used for tagging. However,
the data provide valuable input for improving the simulation of W and t jets, which
have not been considered so far in the tuning of event generators. The combination of Powheg+Pythia describes the distribution in τ 21 adequately, but discrepancies are observed for τ 32 . A similar disagreement between data and simulation is
observed in CMS [531]. The τ 32 distribution has been shown to be sensitive to the
tuning of MPI and the underlying event [602]. In fact, N -subjettiness ratios can help
to constrain uncertainties in the simulation of the final state radiation. Initial studies
suggest that the value of α S used in final state shower is preferred to be smaller for t
jets than the value used in the CMS default simulation with Powheg+Pythia [603],
similar to the observation for light quark jets [596]. However, a complete analysis of
4 Standard Model Measurements
objects using the full detector simulation [243, 435, 436, 459, 483, 522, 525, 525,
526, 530, 531]. Substructure measurements unfolded at the particle level are rare
for boosted W and t jets. So far, only a single measurement has been performed by
ATLAS using 13 TeV data [601]. In this measurement a number of jet substructure
distributions has been measured and compared to simulations for light quark and
gluon jets, W jets and t jets. Trimmed large-R jets are used for a selection of events,
before the original jet constituents are groomed with the soft drop algorithm. There
are three distinct selections, all using anti-k T R = 1.0 jets. Inclusive jets from a
dijet sample are selected with p T > 450 GeV and |η| < 2.5. Events from tt production in the lepton+jets final state are used to select W and t jets with |η| < 1.5
and p T > 200 and 350 GeV, respectively. The soft drop jet mass is required to be
60 < m jet < 100 GeV for W jets and m jet > 140 GeV for t jets. In addition, a condition of angular separation to an identified small-R b jet is required for W jets. In
the selection of t jets, the b jet has to overlap with the t jet. After this selection, the
kinematic phase space is sufficiently different for the inclusive, W and t jets such that
a direct comparison of substructure distributions between these is not meaningful.
However, comparisons with simulation are very useful to study the modelling of substructure observables. The observables studied are subjet multiplicity, Les Houches
angularity, ECF(2, 1), ECF(3, 1), e
(1)
2 , e
(1)
3 , C
(1,1)
2
, D
(1,1)
2
, τ 21 and τ 32 , where the N -
subjettiness ratios have been calculated with the winner-takes-all (WTA) axis [488].
The largest uncertainties in the unfolded distributions originate from modelling differences between different simulations used for the correction of detector effects.
These uncertainties are typically around 15% for W and t jets, highlighting the need
for a better understanding of differences in the simulations. The largest experimental uncertainties can be attributed to uncertainties in the cluster energy, derived with
the R trk method (see Sect. 3.2). These amount to about 8% for W and t jets and
are smaller for inclusive jets. Overall, the simulation describes the data within the
uncertainties, but some trends are visible. In Fig. 4.7 the measured distributions of
τ 21 for W jets and τ 32 for t jets are shown. These are important observables for W
and t tagging, used in several tagger studies and analyses. A potential mismodelling
of these distributions results in efficiency differences between data and simulation
for signal jets. It should be noted that soft drop groomed N -subjettiness ratios have
been measured, whereas ungroomed versions are usually used for tagging. However,
the data provide valuable input for improving the simulation of W and t jets, which
have not been considered so far in the tuning of event generators. The combination of Powheg+Pythia describes the distribution in τ 21 adequately, but discrepancies are observed for τ 32 . A similar disagreement between data and simulation is
observed in CMS [531]. The τ 32 distribution has been shown to be sensitive to the
tuning of MPI and the underlying event [602]. In fact, N -subjettiness ratios can help
to constrain uncertainties in the simulation of the final state radiation. Initial studies
suggest that the value of α S used in final state shower is preferred to be smaller for t
jets than the value used in the CMS default simulation with Powheg+Pythia [603],
similar to the observation for light quark jets [596]. However, a complete analysis of