176 The ATLAS Collaboration / Physics Letters B 784 (2018) 173–191
Table 1
Observed
number of events in the different bins of the H → γγ and H → ZZ
∗
→ 4 searches, using 13 TeV data corresponding to an integrated luminosity of 79.8 fb
−1
.
The observed yields are compared with the sum of expected t
¯
tH signal, normalised to the SM prediction, background from non-t
¯
tH Higgs boson production and other
background sources, with the systematic uncertainties assigned to the observed result in the H → γγ analysis, and expected systematic uncertainties in the H → ZZ
∗
→ 4
analysis. The numbers for H → γγ are counted in the smallest m
γγ
window containing 90% of the expected signal. The numbers for H → ZZ
∗
→ 4 are derived in a
four-lepton mass window of 115 GeV < m
4
< 130 GeV. In the H → γγ analysis, the background yield is extracted from the fit with freely floating signal. The BDT bins are
in descending order of signal purity.
Bin Expected Observed
t
¯
tH (signal) Non-t
¯
tH Higgs Non-Higgs Total Total
H → γγ
Had 1 4.2 ±1.10.49 ±0.33 1.8 ±0.56.4 ±1.310
Had 2 3
.4 ±0.70.7 ±0.67.5 ±1.111.6 ±1.514
Had 3 4
.7 ±0.92.0 ±1.732.9 ± 2.239.6 ± 3.247
Had 4 3
.0 ±0.53.2 ±3.155.0 ±2.861±567
Lep 1 4
.5 ±1.00.24 ±0.09 2.2 ± 0.66.9 ±1.27
Lep 2 2
.2 ±0.40.27 ±0.10 4.6 ±0.97.1 ±1.07
Lep 3 0
.82 ±0.18 0.30 ±0.13 4.6 ±0.95.7 ±0.95
H
→ ZZ
∗
→ 4
Had 1 0.169 ±0.031 0.021 ±0.007 0.008 ± 0.008 0.198 ±0.033 0
Had 2 0
.216 ±0.032 0.20 ±0.09 0.22 ±0.12 0.63 ±0.16 0
Lep 0
.212 ±0.031 0.0256 ±0.0023 0.015 ±0.013 0.253 ±0.034 0
put to a likelihood fit that extracts the t
¯
tH yield. The expected
dominant uncertainties in the cross section are due to the parton-
shower
modelling affecting the acceptance of the selection, and
to the cross-section uncertainty in the Higgs boson plus heavy-
flavour
background (about 10% each). The leading experimental
uncertainty arises from the calibration of the jet energy scale (6%).
The expected and observed numbers of events are presented in
Table 1. No event is observed. The expected significance is 1.2 stan-
dard
deviations.
4. Combination
The t
¯
tH searches in the H → γγ and H → ZZ
∗
→ 4 decay
channels are combined with the H →b
¯
b and multilepton searches
from Refs. [10,15]. These analyses use a dataset corresponding to
an integrated luminosity of 36.1 fb
−1
at
√
s = 13 TeV, and find
observed (expected) significances of 1.4 (1.6) standard deviations
for H → b
¯
b and 4.1 (2.8) for the multilepton search. The combina-
tion
is performed using the profile likelihood method described in
Ref. [54], based on simultaneous fits to the signal regions and con-
trol
regions of the individual analyses. The overlap between the
selected events in the different analyses is found to be negligi-
ble.
The asymptotic approximation used in the fit is verified with
pseudo-experiments, and the results are corrected if necessary. The
effect of systematic uncertainties in the predicted yields and distri-
butions
is incorporated into the statistical model through nuisance
parameters. The correlation scheme of all systematic uncertainties
between the H →b
¯
b and multilepton analyses, as well as the cor-
relation
scheme of the theory uncertainties between all channels
are the same as in Ref. [10]. Since the H → γγ and H → ZZ
∗
→
4 analyses employ improved reconstruction software compared
with the H → b
¯
b and multilepton analyses, the correlations be-
tween
the experimental systematic uncertainties are evaluated for
each source individually. Some components of the systematic un-
certainties
in the luminosity, the jet energy scale, the electron/pho-
ton
resolution and energy scale, and in the electron reconstruction
and identification efficiencies are correlated between the channels.
All Higgs boson production processes other than t
¯
tH, including
Higgs boson production in association with a single top quark, are
considered as background and their cross sections are fixed to the
SM predictions [37]. The respective cross-section uncertainties are
considered as systematic uncertainties. The total t
¯
tH cross section
is extracted assuming SM branching fractions and using the detec-
tor
acceptance and efficiencies predicted from the t
¯
tH simulation
Table 2
Summary
of the systematic uncertainties affecting the combined t
¯
tH cross-section
measurement at 13 TeV. Only systematic uncertainty sources with at least 1% im-
pact
are listed. The fake-lepton uncertainty is due to the estimate of leptons from
heavy-flavour decay, conversions or misidentified hadronic jets. The jet, electron,
and photon uncertainties, as well as the uncertainties associated with hadronically
decaying τ -leptons, include those in reconstruction and identification efficiencies,
as well as in the energy scale and resolution. The Monte Carlo (MC) statistical un-
certainty
is due to limited numbers of simulated events. More detailed descriptions
of the sources of the systematic uncertainties are given in Refs. [10,15].
Uncertainty source σ
t
¯
tH
/σ
t
¯
tH
[%]
Theory uncertainties (modelling) 11.9
t
¯
t + heavy flavour 9.9
t
¯
tH 6.0
Non-t
¯
tH Higgs boson production 1.5
Other background processes 2
.2
Experimental uncertainties 9
.3
Fake leptons 5
.2
Jets, E
miss
T
4.9
Electrons, photons 3
.2
Luminosity 3
.0
τ -leptons 2.5
Flavour tagging 1
.8
MC statistical uncertainties 4
.4
discussed above. The respective uncertainties are included in the
fit.
A combination is also performed with the t
¯
tH searches based
on datasets corresponding to integrated luminosities of 4.5 fb
−1
at
√
s = 7 TeV and 20.3 fb
−1
at
√
s = 8TeV[16]. The com-
bined
observable is the signal strength μ = σ /σ
SM
. The SM cross-
section
expectations σ
SM
and branching ratios used in the 7 and
8 TeV analyses are updated with the values in Ref. [37], while
their uncertainties are not changed. Theoretical uncertainties in
the SM cross-section prediction for t
¯
tH are included in the signal-
strength
extraction. The branching-fraction uncertainties and the
uncertainties due to missing higher-order corrections in the t
¯
tH
cross-section
prediction are correlated between the 7 and 8TeV
and
13 TeV analyses. Furthermore, the relevant uncertainties in the
electron/photon energy scale and resolution are correlated.
5. Results
Table 2 shows a summary of the systematic uncertainties in the
13 TeV t
¯
tH production cross-section measurement. The dominant
uncertainties arise from the modelling of the t
¯
t+ heavy-flavour
processes in the H → b
¯
b analysis [15] and the modelling of the
t
¯
tH process, which affects the acceptance of the selection in all