ATLAS Paper Draft [621530]

ATLAS Paper Draft
EXOT-2016-38
Version 3.0
Comments are due by: 17 April 2018
Supporting internal notes
A model independent general search for new phenomena with the ATLAS detector atps=13TeV:
https://cds.cern.ch/record/2243631
A strategy for a general search for new phenomena
using data-derived signal regions and
its application within the ATLAS experiment
This paper describes a strategy for a general search used by the ATLAS Collaboration to find potential
indications of new physics. Events are classified according to their final state into many event classes.
For each event class an automated search algorithm tests whether the data are compatible with the
Monte Carlo simulated expectation in various distributions sensitive to the effects of new physics. The
significance of a deviation is quantified using pseudo-experiments. A data selection with a signific-
ant deviation defines a signal region for a dedicated follow-up analysis with an improved background
expectation. The analysis of the data-derived signal regions on a new dataset allows a statistical in-
terpretation without the large look-elsewhere effect. The sensitivity of the approach is discussed using
Standard Model processes and benchmark signals of new physics. As an example, results are shown
for 3.2 fb1of proton–proton collision data at a centre-of-mass energy of 13 TeV collected with the AT-
LASdetectorattheLHCin2015,inwhichmorethan700eventclassesandmorethan 105regionshave
been analysed. No significant deviations are found and consequently no data-derived signal regions
for a follow-up analysis have been defined.
To be submitted to: Eur. Phys. J. C
Document created on 18th July 2018 using ATLAS L ATEX Version 05-06-00.

Analysis Team
[email:[anonimizat]]
Sara Alderweireldt, Simone Amoroso, Fabio Cardillo, Sascha Caron, Adam Jinaru,
Alan Litke, Marina Rotaru, Jeroen Schouwenberg, Alexandra Tudorache
Editorial Board
[email:[anonimizat]]
Kevin Black (chair), Christopher Lester, Zach Marshall

ATLAS Paper
EXOT-2016-38
18th July 2018
Draft version 3.0
1
A strategy for a general search for new phenomena2
using data-derived signal regions and 3
its application within the ATLAS experiment 4
The ATLAS Collaboration 5
This paper describes a strategy for a general search used by the ATLAS Collaboration to
find potential indications of new physics. Events are classified according to their final state
into many event classes. For each event class an automated search algorithm tests whether
the data are compatible with the Monte Carlo simulated expectation in various distributions
sensitive to the effects of new physics. The significance of a deviation is quantified using
pseudo-experiments. A data selection with a significant deviation defines a signal region
for a dedicated follow-up analysis with an improved background expectation. The analysis
of the data-derived signal regions on a new dataset allows a statistical interpretation without
the large look-elsewhere effect. The sensitivity of the approach is discussed using Standard
Model processes and benchmark signals of new physics. As an example, results are shown
for 3.2 fb1of proton–proton collision data at a centre-of-mass energy of 13 TeV collected
withtheATLASdetectorattheLHCin2015,inwhichmorethan700eventclassesandmore
than 105regions have been analysed. No significant deviations are found and consequently
no data-derived signal regions for a follow-up analysis have been defined.6
7
8
9
10
11
12
13
14
15
16
17
18
19
To be submitted to: Eur. Phys. J. C
©2018 CERN for the benefit of the ATLAS Collaboration.
Reproduction of this article or parts of it is allowed as specified in the CC-BY-4.0 license.20

ATLAS DRAFT
1 Introduction21
Direct searches for unknown particles and interactions are one of the primary objectives of the physics22
programme at the Large Hadron Collider (LHC). The ATLAS experiment at the LHC has thoroughly23
analysed the Run 1 ppcollision dataset (recorded in 2010–2012) and roughly a quarter of the expected 24
Run 2 dataset (2015–2018). No evidence of physics beyond the Standard Model (SM) has been found in25
any of the searches performed so far.26
Searches that have been performed to date do not fully cover the enormous parameter space of masses,27
cross-sectionsanddecaychannelsofpossiblenewparticles. Signalsmightbehiddeninkinematicregimes28
and final states that have remained unexplored. This motivates a model-independent 1analysis to search 29
forphysicsbeyondtheStandardModel(BSM)inastructured,globalandautomatedway,wheremanyof30
the final states not yet covered can be probed.31
GeneralsearcheswithoutanexplicitBSMsignalassumptionhavebeenbeenperformedbytheDØCollab-32
oration [1–4] at theTevatron, by the H1 Collaboration [5, 6]at HERA, and by the CDFCollaboration [7,33
8]attheTevatron. AttheLHC,preliminaryversionsofsuchsearcheshavebeenperformedbytheATLAS34
Collaboration atps=7,8and13TeV, and by the CMS Collaboration atps=7TeV. 35
This paper outlines a strategy employed by the ATLAS Collaboration to search in a systematic and36
(quasi-)model-independentwayfordeviationsofthedatafromtheSMprediction. Thisapproachassumes37
only generic features of the potential BSM signals. Signal events are expected to have reconstructed38
objects with relatively large momentum transverse to the beam axis. In contrast to previous generic39
searches, the main objective of this strategy is not to finally access the exact level of significance of a40
deviationwithallavailabledata,butrathertoidentifywithafirstdatasetthosephase-spaceregionswhere41
significant deviations of the data from SM prediction are present for a further dedicated analysis. The42
observation of one or more significant deviations in some phase-space region(s) serves as a trigger to43
perform dedicated and model-dependent analyses where these ‘data-derived’ phase-space region(s) can44
be used as signal regions. Such an analysis can then determine the level of significance using a second45
dataset. Themainadvantageofthisprocedureisthatitallowsalargenumberofphase-spaceregionstobe46
testedwiththeavailableresources,therebyminimizingthepossibilityofmissingasignalfornewphysics,47
while simultaneously maintaining a low false discovery rate by testing the data-derived signal region(s)48
onanindependentdatasetinadedicatedanalysis. Thededicatedanalysiswithdata-derivedsignalregions49
also allows an improved background prediction.50
Inthisapproach,eventsarefirstclassifiedintodifferent(exclusive)categories,labelledwiththemultiplicity51
of final-state objects (e.g. muons, electrons, jets, missing transverse momentum, etc.) in an event. These52
final-statecategoriesarethenautomaticallyanalysedfordeviationsofthedatafromtheSMpredictionin53
several BSM-sensitive distributions using an algorithm that locates the region of largest excess or deficit.54
Sensitivitytestsforspecificsignalmodelsareperformedtodemonstratetheeffectivenessofthisapproach.55
Themethodologyhasbeenappliedtoasubsetoftheps=13TeVproton–protoncollisiondataasreported 56
in this paper. The data were collected with the ATLAS detector in 2015, and correspond to an integrated57
luminosity of 3:2fb1. 58
1‘Model-independent’ refers to the absence of a beyond the Standard Model signal assumption. The analysis depends on the
Standard Model prediction.
18th July 2018 – 12:25 4

ATLAS DRAFT
The paper is organized as follows: the general analysis strategy is outlined in Section 2, while Section 359
provides specific details about its application to the ATLAS 2015 ppcollision dataset. Conclusions are 60
given in Section 4.61
2 Strategy62
Theanalysisstrategyassumesthatasignalofunknownorigincanberevealedasastatisticallysignificant63
deviationoftheeventcountsinthedatafromtheexpectationinaspecificdataselection. Adataselection64
canbeanysetofrequirementsonobjectsorvariablesneededtodefineasignalregion(e.g.aneventclass65
or a specific range in one or multiple observables). In order to search for these signals a large variety of66
data selections need to be tested. This requires a high degree of automation and a categorization of the67
data events according to their main features. The main objective of this analysis is to identify selections68
for which the data deviates significantly from the SM expectation. These selections can then be applied69
as data-derived signal regions in a dedicated analysis to determine the level of significance using a new70
dataset. Thishastheadvantageofamorereliablebackgroundexpectation,whichshouldallowanincrease71
in signal sensitivity compared to a strategy that only relies on Monte Carlo expectations with a typically72
conservative evaluation of uncertainties. The strategy is divided into the seven steps described below.73
2.1 Step 1: Data selection and Monte Carlo simulation74
The recorded data are reconstructed via the ATLAS software chain. Events are selected by applying75
event-qualityandtriggercriteria,andareclassifiedaccordingtothetypeandmultiplicityofreconstructed76
objects with high transverse momentum ( pT). Objects that can be considered in the classification are 77
those typically used to characterize hadron collisions such as electrons, muons, -leptons, photons, jets, 78
b-tagged jets and missing transverse momentum. More complex objects, which were not implemented 79
in the example described in Section 3, could also be considered. Examples are resonances reconstructed80
by a specific decay (e.g. Zor Higgs bosons decaying into two or four isolated leptons respectively, or 81
decayinghadronicallyandgivingrisetolargeradiusjetswithsubstructure)anddisplacedvertices. Event82
classes(orchannels)arethendefinedasthesetofeventswithagivennumberofreconstructedobjectsfor83
each type, e.g. two muons and a jet.84
Monte Carlo (MC) simulations are used to estimate the expected event counts from SM processes. To85
allow the investigation of signal regions with a low number of expected events it is important that the86
equivalent integrated luminosity of the MC samples significantly exceeds that of the data, and that all87
relevant background processes are included, in particular rare processes which might dominate certain88
multi-object event classes.89
2.2 Step 2: Systematic uncertainties and validation90
The particular nature of this analysis, in which a large number of final states are explored, makes the91
definition of control and validation regions difficult. In searches for BSM physics at the LHC, control92
regionsareusedtoconstrainMC-basedbackgroundpredictionswithauxiliarymeasurements. Validation93
regions are used to test the validity of the background model prediction with data.94
18th July 2018 – 12:25 5

ATLAS DRAFT
The simplest way to construct a background model is to obtain the background expectation from the MC95
prediction including the corresponding theoretical and experimental uncertainties. This approach, which96
is applied in the example in Section 3, has the advantage that it prevents the absorption of BSM signal97
contributionsintoa rescalingoftheSMprocesses. Anotherpossibleapproach istoautomaticallydefine,98
for each data selection and algorithmic hypothesis test, statistically independent control selections. The99
data in the control selections can be used to rescale the MC background predictions and to constrain the100
systematic uncertainties. This comes at the price of reduced sensitivity for the case in which a BSM101
model predicts a simultaneous effect in the signal region and control region, which would be absorbed in102
the rescaling.103
To verify the proper modelling of the SM background processes, several validation distributions are104
defined using inclusive selections for which observable signals for new physics are excluded. If these105
validation distributions show problems in the MC modelling, either corrections to the MC backgrounds106
are applied or the affected event class is excluded.107
Uncertainties in the background estimate arise from experimental effects, and the theoretical accuracy108
of the prediction of the (differential) cross-section and acceptance of the MC simulation. Their effect is109
evaluated for all contributing background processes as well as for benchmark signals.110
2.3 Step 3: Sensitive variables and search algorithm111
Distributions of observables in the form of histograms are investigated for all event classes considered in112
the analysis. Observables are included if they have a high sensitivity to a wide range of BSM signals.113
Thetotalnumberofobservablesconsideredis,however,restrictedtoafewtoavoidalargeincreaseinthe114
numberofhypothesistests,asthelatteralsoincreasestherateofdeviationsfrombackgroundfluctuations.115
In high-energy physics this effect is commonly known as the ‘trial factor’ or ‘look-elsewhere effect’.116
Examples of such observables are the effective mass meff(defined as the sum of the scalar transverse 117
momenta of all objects plus the scalar missing transverse momentum), the total invariant mass minv118
(defined as the invariant mass of all visible objects), the invariant mass of any combination of objects119
(suchasthedielectroninvariantmassineventswithtwoelectronsandtwomuons),eventshapevariables120
such as thrust [9, 10] or even more complicated variables such as the output of a machine-learning121
algorithm.122
Astatisticalalgorithmisusedtoscanthesedistributionsforeacheventclassandquantifythedeviationsof123
thedatafromtheSMexpectation. Thealgorithmidentifiesthedataselectionthathasthelargestdeviation124
inthedistributionoftheinvestigatedobservablebytestingmanydataselectionstominimizeateststatistic.125
Apossibleteststatisticisalocal p0-value,whichgivestheexpectedprobabilityofobservingafluctuation 126
that is at least as far from the SM expectation as the observed number of data events in a given region, if127
the experiment were to be repeated. A possible definition of the local p0-value is: 128
p0=2
minh
P¹nNobsș;P¹nNobsși
(1)
P¹nNobsș=¹1
0dx G¹x;NSM;NSMș
NobsÕ
n=0exxn
n!+¹0
1dx G¹x;NSM;NSMș (2)
P¹nNobsș=¹1
0dx G¹x;NSM;NSMș
1Õ
n=Nobsexxn
n!(3)
18th July 2018 – 12:25 6

ATLAS DRAFT
where nis the independent variable of the Poisson probability mass function (pmf), Nobsis the observed 129
number of data events for a given selection, P¹nNobsșis the probability of observing no more than 130
the number of events observed in the data and P¹nNobsșis the probability of observing at least the 131
numberofeventsobservedinthedata. Thequantity NSMistheexpectationforthenumberofeventswith 132
its total uncertainty NSMfor a given selection. The convolution of the Poisson pmf (with mean x) with 133
a Gaussian probability density function (pdf), G¹x;NSM;NSMșwith mean NSMand widthNSM, takes 134
the effect of both non-negligible systematic uncertainties and statistical uncertainties into account. 2If 135
the Gaussian pdf Gis replaced by a Dirac delta function ¹xNSMșthe estimator p0results in the usual 136
Poisson probability. The selection with the largest deviation identified by the algorithm is defined as the137
selection giving the smallest p0-value. The smallest p0for a given channel is defined as pchannel, which 138
therefore corresponds to the local p0-value of the largest deviation in that channel. 139
Data selections are not considered in the scan if large uncertainties in the expectation arise due to a lack140
of MC events, or from large systematic uncertainties. To avoid overlooking potential excesses in these141
selections the p0-values of selections with more than three data events are monitored separately. 142
The result of scanning the distributions for all event classes is a list of data selections, one per event143
class containing the largest deviation in that class, and their local statistical significance. Details of the144
procedure and the statistical algorithm used for the 2015 dataset are explained in Section 3.3.145
2.4 Step 4: Generation of pseudo-experiments146
The probability that for a given observable one or more deviations of a certain size occur somewhere in147
the event classes considered is modelled by pseudo-experiments. Each pseudo-experiment consists of148
exactlythesameeventclassesasthoseconsideredwhenapplyingthesearchalgorithmtodata. However,149
thedatacountsarereplacedbypseudo-datacountswhicharegeneratedfromtheSMexpectationusingan150
MCtechnique. Pseudo-datadistributionsareproducedtakingintoaccountbothstatisticalandsystematic151
uncertaintiesbydrawingpseudo-randomdatacountsforeachbinfromtheconvolvedpmfusedinEqs.(1)–152
(3) to compute a p0-value. 153
Correlations in the uncertainties of the SM expectation affect the chance of observing one or more154
deviations of a given size. The effect of correlations between bins of the same distribution or between155
distributions of different event classes are therefore taken into account when generating pseudo-data156
for pseudo-experiments. Correlations between distributions of different observables are not taken into157
account, since the results obtained for different observables are not combined in the interpretation.158
The search algorithm is then applied to each of the distributions, resulting in a pchannel-value for each 159
event class. The pchanneldistributions of many pseudo-experiments and their statistical properties can be 160
compared with the pchanneldistribution obtained from data to interpret the test statistics in a frequentist 161
manner. The fraction of pseudo-experiments having one of the pchannel-values smaller than a given value 162
2ThesecondterminEq.(2)givestheprobabilityofobservingnoeventsgivenanegativeexpectationfromdownwardvariations
of the systematic uncertainties. It can be derived as follows:
¹0
1dx G¹x;NSM;NSMș
NobsÕ
n=0lim
!0en
n!
=¹0
1dx G¹x;NSM;NSMș
NobsÕ
n=0n0=¹0
1dx G¹x;NSM;NSMș
whereisthemeanofthePoissonpmfand n0=f1ifn=0,0ifn,0gistheKroneckerdelta. InEq.(3)thistermvanishes
forNobs>0.
18th July 2018 – 12:25 7

ATLAS DRAFT
pminindicatestheprobabilityofobservingsuchadeviationbychance,takingintoaccountthenumberof 163
selections and event classes tested.164
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
2
−
10
1
−
10
1
Simulation
ATLAS
-1
= 13 TeV, 3.2 fb
s
min
< p
channel
p
no. of channels with
inv
variable: m
1
≥
2
≥
3
≥
σ
1
σ
2
σ
3
σ
4
σ
5
5%
Figure 1: The fractions of pseudo-experiments ( Pexp;i¹pminș) in the minvscan, which have at least one, two or
three pchannel-valuessmallerthanagiventhreshold( pmin). Pseudo-datasetsaregeneratedfromtheSMexpectation.
Dotted lines are drawn at Pexp;i=5%and at the corresponding log10¹pminș-values.
Toillustratethis,Figure1showsthreecumulativedistributionsof pchannel-valuesfrompseudo-experiments. 165
The number of event classes (686) and the minvdistributions used to generate these pseudo-experiments 166
coincide with the example application in Section 3. The distribution in Figure 1 with circular markers is167
the fraction of pseudo-experiments with at least one pchannel-value smaller than pmin. For example, about 168
15% of the pseudo-experiments have at least one pchannel-value smaller than pmin=104. Therefore, the 169
estimated probability ( Pexp;i) of obtaining a single pchannel-value ( i=1) smaller than 104from data in 170
the absence of a signal is about 15%, or Pexp;1¹104ș=0:15. To estimate the probability of observing 171
deviations of a given size in at least two or three different event classes, the second or third smallest172
pchannel-value of a pseudo-experiment is compared with a given pminthreshold. From Figure 1 it follows 173
for instance that 2% of the pseudo-experiments have at least three pchannel-values smaller than 104. 174
Consequently, the probability of obtaining a third smallest pchannel-value smaller than 104from data in 175
the absence of a signal is about 2%, or Pexp;3¹104ș=0:02 176
InFigure1ahorizontaldottedlineisdrawnatafractionofpseudo-experimentsof5%andcorresponding177
vertical dotted lines are drawn at the three pminthresholds. The observation of one, two or three pchannel- 178
values in data below the corresponding pminthreshold, i.e. an observation with a Pexp;i<0:05, promotes 179
the selections that yielded these deviations to signal regions that can be tested in a new dataset.180
2.5 Step 5: Evaluation of the sensitivity181
The sensitivity of the procedure to a priori unspecified BSM signals can be evaluated with two different182
methods that either use a modified background estimation through the removal of SM processes or in183
which signal contributions are added to the pseudo-data sample.184
In the first method, a rare SM process (with either a low cross-section or a low reconstruction efficiency)185
is removed from the background model. The search algorithm is applied again to test the data or ‘signal’186
pseudo-experiments generated from the unmodified SM expectation, against the modified background187
18th July 2018 – 12:25 8

ATLAS DRAFT
expectation. The data samples would be expected to reveal excesses relative to the modified background188
prediction.189
In the second method, pseudo-experiments are used to test the sensitivity of the analysis to benchmark190
signalmodelsofnewphysics. ThepredictionofamodelisaddedtotheSMprediction,andthismodified191
expectation is used to generate ‘signal’ pseudo-experiments. The search algorithm is applied to the192
pseudo-experiments and the distribution of pchannel-values is derived. 193
Toprovideafigureofmeritforthesensitivityoftheanalysis,thefractionof‘signal’pseudo-experiments194
with Pexp;i<5%fori=1;2;3is computed. 195
2.6 Step 6: Results196
Finding one or more deviations in the data with Pexp;i<5%triggers a dedicated analysis that uses the 197
data selection in which the deviation is observed as a signal region (step 7). If no significant deviations198
are found, the outcome of the analysis technique includes information such as: the number of events199
and expectation per event class, a comparison of the data with the SM expectation in the distributions of200
observables considered, the scan results (i.e. the location and the local p0-value of the largest deviation 201
per event class) and the comparison with the expectation from pseudo-experiments.202
2.7 Step 7 (only in the case of Pexp;i<5%): Dedicated analysis of deviation 203
Dedicated analysis on original dataset Deviations are investigated using methods similar to those of 204
a conventional analysis. In particular, the background prediction is determined using control selections205
to control and validate the background modelling. Such a procedure further constrains the background206
expectation and uncertainty, and reduces the dependence on simulation.207
Dedicatedanalysisonanindependentdataset Ifadeviationpersistsinadedicatedanalysisusingthe 208
originaldataset,thedataselectioninwhichthedeviationisobserveddefinesadata-derivedsignalregion209
that is tested in an independent new dataset with a similar or larger integrated luminosity. At this point,210
a particular model of new physics can be used to interpret the result of testing the data-derived signal211
region. Since the signal region is known, the corresponding data can be excluded (‘blinded’) from the212
analysis until the very end to minimize any possible bias in the analysis. Additionally, since only a few213
optimized hypothesis tests are performed on the independent dataset, the large look-elsewhere effect due214
to the large number of hypothesis tests performed in step 3 is not present in the dedicated analysis of the215
signal region(s).216
2.8 Advantages and disadvantages217
The features of this strategy lead to several advantages and disadvantages that are outlined below.218
Advantages:219
•It can find unexpected signals for new physics due to the large number of event classes and phase- 220
space regions probed, which may otherwise remain uninvestigated. 221
18th July 2018 – 12:25 9

ATLAS DRAFT
•A relatively small excess in two or three independent data selections, each of which is not big 222
enough to trigger a dedicated analysis by itself ( Pexp;1>5%), can trigger one in combination 223
(Pexp;2;3<5%). 224
•The approach is broad, and the resulting distributions can be used to probe the overall description 225
of the data by the event generators for many SM processes. 226
•The probability of a deviation occuring in anyof the many different event classes under study can 227
bedeterminedwithpseudo-experiments,resultinginatrulyglobalinterpretationoftheprobability 228
of finding a deviation within an experiment such as ATLAS. 229
Disadvantages:230
•The outcome depends on the MC-based description of physics processes and simulations of the 231
detector response. Event classes in which the majority of the events contain misreconstructed 232
objects are typically poorly modelled by MC simulation and might need to be excluded from the 233
analysis. Although step 2 validates the description of the data by the MC simulation, there is still 234
a possibility of triggering false positives due to an MC mismodelling in a corner of phase space. 235
Step7aimstominimizethisbyreducingthedependenceonMCsimulationsinadedicatedanalysis 236
performed for each significant deviation. In future implementations a better background model 237
could be constructed with the help of control regions or data-derived fitting functions. This might 238
allowthedetectionofexcessessmallcomparedtotheuncertaintiesintheMC-baseddescriptionof 239
the SM processes. 240
•SincethisanalysisisnotoptimizedforaspecificclassofBSMsignals,adedicatedanalysisoptimized 241
for a given BSM signal achieves a larger sensitivity to that signal. The enormous parameter space 242
of possible signals makes an optimized search for each of them impossible. 243
•The large number of data selections introduce a large look-elsewhere effect, which reduces the 244
significance of a real signal. Step 7 circumvents this problem since the final discovery significance 245
is determined with a dedicated analysis of one or a few data selection(s) and a statistically inde- 246
pendentdataset. Thiscanyieldanimprovedsignalsensitivityifthebackgrounduncertaintycanbe 247
constrained in the dedicated analysis. 248
•Despitebeingbroad,theproceduremightmissacertainsignalbecauseitdoesnotshowalocalized 249
excessinoneofthestudieddistributions. AsanexamplethisstrategyislargelyinsensitivetoBSM 250
processes resulting in small overall normalization differences in large cross-section processes such 251
as multijet production. This might be overcome with better observables or modified algorithms, 252
which may then be sensitive to such signals. 253
3 Application of the strategy to ATLAS data254
This section describes the application of the strategy outlined in the previous section to the 13 TeV pp 255
collision data recorded by the ATLAS experiment in 2015.256
18th July 2018 – 12:25 10

ATLAS DRAFT
3.1 Step 1: Data selection and Monte Carlo simulation257
3.1.1 ATLAS detector and dataset258
TheATLASdetector[11]isamultipurposeparticlephysicsdetectorwithaforward-backwardsymmetric259
cylindricalgeometryandacoverageofnearly 4insolidangle. 3Theinnertrackingdetector(ID)consists 260
of silicon pixel and microstrip detectors covering the pseudorapidity region jj<2:5, surrounded by 261
a straw-tube transition radiation tracker which enhances electron identification in the region jj<2:0. 262
BetweenRun1andRun2,anewinnerpixellayer,theinsertableB-layer[12],wasinsertedatameansensor263
radius of 3.3 cm. The inner detector is surrounded by a thin superconducting solenoid providing an axial264
2Tmagneticfieldandbyafine-granularitylead/liquid-argon(LAr)electromagneticcalorimetercovering265
jj<3:2. A steel/scintillator-tile calorimeter provides hadronic coverage in the central pseudorapidity 266
range (jj<1:7). The endcap and forward calorimeter coverage ( 1:5<jj<4:9) is completed by LAr 267
active layers with either copper or tungsten as the absorber material. An extensive muon spectrometer268
withanair-coretoroidmagnetsystemsurroundsthecalorimeters. Threelayersofhigh-precisiontracking269
chambers provide coverage in the range jj<2:7, while dedicated fast chambers provide a muon trigger 270
in the regionjj<2:4. The ATLAS trigger system consists of a hardware-based level-1 trigger followed 271
by a software-based high-level trigger [13].272
The data used in this analysis were collected by the ATLAS detector during 2015 in ppcollisions at the 273
LHC with a centre-of-mass energy of 13 TeV and a 25 ns bunch crossing interval. After applying quality274
criteria for the beam, data and detector, the available dataset corresponds to an integrated luminosity275
of 3.2 fb1. In this dataset, each event includes an average of approximately 14 additional inelastic pp 276
collisions in the same bunch crossing (pile-up).277
Candidateeventsarerequiredtohaveareconstructedvertex[14],withatleasttwoassociatedtrackswith278
pT>400MeV. Thevertexwiththehighestsumofsquaredtransversemomentaofthetracksisconsidered 279
to be the primary vertex.280
3.1.2 Monte Carlo samples281
Monte Carlo simulated event samples [15] are used to describe SM background processes and to model282
possible signals. The ATLAS detector is simulated either by a software system based on G/e.sc/a.sc/n.sc/t.sc4[16] or 283
by a faster simulation based on a parameterization of the calorimeter response and G/e.sc/a.sc/n.sc/t.sc4for the other 284
detector systems. To account for additional ppinteractions from the same or nearby bunch crossings, 285
a set of minimum-bias interactions generated using P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 [17], the MSTW2008LO [18] parton 286
distribution function (PDF) set and the A2 set of tuned parameters (tune) [19] was superimposed onto287
the hard-scattering events to reproduce the observed distribution of the average number of interactions288
per bunch crossing. In all MC samples, except those produced by S/h.sc/e.sc/r.sc/p.sc/a.sc[20], the E/v.sc/t.scG/e.sc/n.sc v1.2.0 289
program [21] was used to model the properties of the bottom and charm hadron decays. The SM MC290
programs are listed in Table 1 and a detailed explanation can be found in Appendix A.1.291
3ATLAS uses a right-handed coordinate system with its origin at the nominal interaction point in the centre of the detector.
Thepositive x-axisisdefinedbythedirectionfromtheinteractionpointtothecentreoftheLHCring,withthepositive y-axis
pointingupwards,whilethebeamdirectiondefinesthe z-axis. Cylindricalcoordinates( r,)areusedinthetransverseplane, 
beingtheazimuthalanglearoundthe z-axis. Thepseudorapidity isdefinedintermsofthepolarangle by=ln tan¹2ș.
The angular distance is defined as
R=p
¹
ș2+¹
ș2. Rapidity is defined as y=0:5
ln»¹E+pzș¹Epzș¼where E
denotes the energy and pzis the component of the momentum along the beam direction.
18th July 2018 – 12:25 11

ATLAS DRAFTTable 1: A summary of the MC samples used in the analysis to model SM background processes. For each sample the corresponding generator, matrix element
(ME)accuracy,partonshower,cross-sectionnormalizationaccuracy,PDFsetandtuneareindicated. DetailsaregiveninAppendixA.1. Sampleswith‘data’in
the ‘cross-section normalization’ column are scaled to data as described in Section 3.2.3. Zrefers to
/Z.
Physics process Generator ME accuracy Parton shower Cross-section PDF set Tune
normalization
W(!`) + jets S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2j@NLO + 3,4j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NNLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Z(!`+`) + jets S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2j@NLO + 3,4j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NNLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Z/W(!q¯q) + jets S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 1,2,3,4j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NNLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Z/W+
S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2,3j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Z/W+

S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2,3j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault

+ jets S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2,3,4j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 data NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault

+ jets S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 data NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault

+ jets MG5_aMC@NLO 2.3.3 0,1j@LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.212 LO NNPDF23LO A14
t¯t P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v2 NLO P/y.sc/t.sc/h.sc/i.sc/a.sc6.428 NNLO+NNLL NLO CT10 Perugia 2012
t¯t+W MG5_aMC@NLO 2.2.2 0,1,2j@LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 NLO NNPDF2.3LO A14
t¯t+Z MG5_aMC@NLO 2.2.2 0,1j@LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 NLO NNPDF2.3LO A14
t¯t+WW MG5_aMC@NLO 2.2.2 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 NLO NNPDF2.3LO A14
t¯t+
MG5_aMC@NLO 2.2.2 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 LO NNPDF2.3LO A14
t¯t+b¯b S/h.sc/e.sc/r.sc/p.sc/a.sc2.2.0 NLO S/h.sc/e.sc/r.sc/p.sc/a.sc2.2.0 NLO NLO CT10f4 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Single-top (t-channel) P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v1 NLO P/y.sc/t.sc/h.sc/i.sc/a.sc6.428 app. NNLO NLO CT10f4 Perugia 2012
Single-top (s- and Wt-channel) P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v2 NLO P/y.sc/t.sc/h.sc/i.sc/a.sc6.428 app. NNLO NLO CT10 Perugia 2012
tZ MG5_aMC@NLO 2.2.2 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 LO NNPDF2.3LO A14
3-top MG5_aMC@NLO 2.2.2 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 LO NNPDF2.3LO A14
4-top MG5_aMC@NLO 2.2.2 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 NLO NNPDF2.3LO A14
WW S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0j@NLO + 1,2,3j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
W Z S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0j@NLO + 1,2,3j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Z Z S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1j@NLO + 2,3j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
Multijets P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 data NNPDF2.3LO A14
Higgs (ggF/VBF) P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v2 NLO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 NNLO NLO CT10 AZNLO
Higgs ( t¯tH) MG5_aMC@NLO 2.2.2 NLO Herwig++ NNLO NLO CT10 UEEE5
Higgs ( WZH) P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 LO P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 NNLO NNPDF2.3LO A14
Tribosons S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 0,1,2j@LO S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 NLO NLO CT10 S/h.sc/e.sc/r.sc/p.sc/a.scdefault
18th July 2018 – 12:25 12

ATLAS DRAFT
In addition to the SM background processes, two possible signals are considered as benchmarks. The292
first benchmark model considered is the production of a new heavy neutral gauge boson of spin 1 ( Z0), 293
as predicted by many extensions of the SM. Here, the specific case of the sequential extension of the SM294
gauge group (SSM) [22, 23] is considered, for which the couplings are the same as for the SM Zboson. 295
This process was generated at leading order (LO) using P/y.sc/t.sc/h.sc/i.sc/a.sc8.212 with the NNPDF23LO [24] PDF 296
set and the A14 tune [25], as a Drell–Yan process, for five different resonant masses, covering the range297
from 2 TeV to 4 TeV, in steps of 0.5 TeV. The considered decays of Z0bosons are inclusive, covering the 298
fullrangeofleptonandquarkpairs. InterferenceeffectswithSMDrell–Yanproductionarenotincluded,299
and the Z0boson is required to decay into fermions only. 300
The second signal considered is the supersymmetric [26–31] production of gluino pairs through strong301
interactions. Thegluinosareassumedtodecaypromptlyintoapairoftopquarksandanalmostmassless302
neutralinoviaanoff-shelltopsquark ˜g!tt˜0
1. SamplesforthisprocessweregeneratedatLOwithupto 303
twoadditionalpartonsusingMG5_aMC@NLO2.2.2[32]withtheCTEQ6L1[33]PDFset,interfacedto304
P/y.sc/t.sc/h.sc/i.sc/a.sc8.186withtheA14tune. ThematchingwiththepartonshowerwasdoneusingtheCKKW-L[34] 305
prescription, with a matching scale set to one quarter of the pair-produced resonance mass. The signal306
cross-sectionswerecalculatedatnext-to-leadingorder(NLO)inthestrongcouplingconstant,addingthe307
resummation of soft gluon emission at next-to-leading-logarithm (NLL) accuracy [35–37].308
3.1.3 Object reconstruction309
Reconstructed physics objects considered in the analysis are: prompt and isolated electrons ( e), muons 310
() and photons (
), as well as b-jets ( b) and light (non- b-tagged) jets ( j) reconstructed with the anti- kt311
algorithm [38] with radius parameter R=0:4, and large missing transverse momentum ( Emiss
T). Table 2 312
lists the reconstructed physics objects along with their pTand pseudorapidity requirements. Jets and 313
electrons misidentified as hadronically decaying -leptons are difficult to model with the MC-based 314
approach used in this analysis. Therefore, the identification of hadronically decaying -leptons is not 315
considered; they are mostly reconstructed as light jets. Details of the object reconstruction can be found316
in Appendix B.317
Table 2: The physics objects used for classifying the events, with their corresponding label, minimum pTrequire-
ment, and pseudorapidity requirement.
Object Label pT(min) [GeV] Pseudorapidity
Isolated electron e 25jj<1:37or1:52<jj<2:47
Isolated muon  25 jj<2:7
Isolated photon
50jj<1:37or1:52<jj<2:37
b-tagged jet b 60 jj<2:5
Light (non- b-tagged) jet j 60 jj<2:8
Missing transverse momentum Emiss
T200
After object identification, overlaps between object candidates are resolved using the distance variable318

Ry=p
¹
yș2+¹
ș2. IfanelectronandamuonsharethesameIDtrack,theelectronisremoved. Any 319
jet within a distance
Ry=0:2of an electron candidate is discarded, unless the jet has a value of the 320
b-tagging MV2/c.sc20 discriminant [39, 40] larger than that corresponding to approximately 85% b-tagging 321
efficiency, in which case the electron is discarded since it probably originated from a semileptonic b- 322
hadrondecay. Anyremainingelectronwithin
Ry=0:4ofajetisdiscarded. Muonswithin
Ry=0:4of 323
18th July 2018 – 12:25 13

ATLAS DRAFT
ajetarealsoremoved. However,ifthejethasfewerthanthreeassociatedtracks,themuoniskeptandthe324
jet is discarded instead to avoid inefficiencies for high-energy muons undergoing significant energy loss325
in the calorimeter. If a photon candidate is found within
Ry=0:4of a jet, the jet is discarded. Photons 326
within a cone of size
Ry=0:4around an electron or muon candidate are discarded. 327
The missing transverse momentum (with magnitude Emiss
T) is defined as the negative vector sum of the 328
transverse momenta of all selected and calibrated physics objects (electrons, photons, muons and jets)329
in the event, with an additional soft-term [41]. The soft-term is constructed from all tracks that are not330
associated with any physics object, but are associated with the primary vertex.331
3.1.4 Event selection and classification332
The events are divided into mutually exclusive classes that are labelled with the number and type of333
reconstructed objects listed in Table 2. The division can be regarded as a classification according to the334
mostimportantfeaturesoftheevent. Theclassificationincludesallpossiblefinal-stateconfigurationsand335
object multiplicities, e.g. if a data event with seven reconstructed muons and no other objects is found, it336
is classified in a ‘ 7-muon’ event class ( 7). Similarly an event with missing transverse momentum, two 337
muons, one photon and four jets is classified and considered in the corresponding event class denoted338
Emiss
T21
4j. 339
All events contributing to a particular event class are also required to be selected by a trigger from a340
corresponding class of triggers by imposing a hierarchy in the event selection. The flow diagram in341
Figure2givesagraphicalrepresentationofthetriggerandofflineeventselection,basedontheclassofthe342
event. Sincethethresholdsforthesingle-photonandsingle-jettriggersarehigherthanthe pTrequirements 343
inthephotonandjetobjectselection,anadditionalreconstruction-level pTcutisimposedtoavoidtrigger 344
inefficiencies. For the other triggers, the pTrequirements in the object definitions exceed the trigger 345
thresholds by a sufficient margin to avoid additional trigger inefficiencies. Events with Emiss
T>200GeV 346
are required to pass the Emiss
Ttrigger, otherwise they are rejected and not considered for further event 347
selection. If the event has Emiss
T<200GeV but contains an electron with pT>25GeV it is required to 348
pass the single-electron trigger. However, events with more than one electron with pT>25GeV or with 349
an additional muon with pT>25GeV can be selected by the dielectron trigger or electron-muon trigger 350
respectively if the event fails to pass the single-electron trigger. Events with a muon with pT>25GeV 351
but no reconstructed electrons or large Emiss
Tare required to pass the single-muon trigger. If the event 352
has more than one muon with pT>25GeV and fails to pass the single-muon trigger, it can additionally 353
be selected by the dimuon trigger. Remaining events with a photon with pT>140GeV or two photons 354
with pT>50GeV are required to pass the single-photon or diphoton trigger, respectively. Finally, any 355
remaining event with no large Emiss
T, leptons, or photons, but containing a jet with pT>500GeV is 356
required to pass the single-jet trigger.357
In addition to the thresholds imposed by the trigger, a further selection is applied to event classes with358
Emiss
T<200GeVcontainingoneleptonoroneelectronandonemuonandpossiblyadditionalphotonsor 359
jets( 1+X,1e+Xand11e+X),toreducetheoveralldatavolume. Intheseeventclasses,onelepton 360
is required to have pT>100GeV if the event has less than three jets with pT>60GeV. 361
To suppress sources of fake Emiss
T, additional requirements are imposed on events to be classified in 362
Emiss
Tcategories. Theratioof Emiss
Ttomeffisrequiredtobegreaterthan0.2,andtheminimumazimuthal 363
separationbetweenthe Emiss
Tdirectionandthethreeleadingreconstructedjets(ifpresent)hastobegreater 364
than 0.4, otherwise the event is rejected.365
18th July 2018 – 12:25 14

ATLAS DRAFT
Event
Emiss
T>
200GeVMET
triggerAccepted
Rejected
has
electron(s)Single
electron
triggers
N¹eș2Dielectron
triggers
N¹ș1Electron-
muon
triggersAccepted
Rejected
has
muon(s)Single
muon
triggers
N¹ș2Dimuon
triggersAccepted
Rejected
has
photon(s)pT¹
ș>
140GeVSingle
photon
triggers
N¹
ș2Diphoton
triggersAccepted
Rejected
has
jet(s)pT¹jetș>
500GeVSingle jet
triggerAccepted
Rejectedyes
nopassed
failed
yes
nopassedfailed
yes
nopassed
failed
yes
nopassed
failed
yes
nopassed
failed
yes
nofailedpassed
yes
noyes
no
yes
nopassed
passedfailed
failed
yes
noyes
nopassed
failed
Figure 2: Flow diagram for the trigger and offline event selection strategy. The offline requirements are shown on
the left of the dashed line and the trigger requirements are shown on the right of the dashed line.
18th July 2018 – 12:25 15

ATLAS DRAFT
3.2 Step 2: Systematic uncertainties and validation366
3.2.1 Systematic uncertainties367
Experimentaluncertainties ThedominantexperimentalsystematicuncertaintiesintheSMexpectation 368
for the different event classes are the jet energy scale (JES) and resolution (JER) [42] and the scale and369
resolution of the Emiss
Tsoft-term. The uncertainty related to the modelling of Emiss
Tin the simulation 370
is estimated by propagating the uncertainties in the energy and momentum scale of each of the objects371
entering the calculation, with an additional uncertainty in the resolution and scale of the soft-term [41].372
The uncertainties in the b-tagging scale factors are determined in data samples enriched in top quark 373
decays,andinsimulatedevents[39]. Leptonicdecaysof J mesonsand Zbosonsindataandsimulation 374
areexploitedtoestimatetheuncertaintiesinleptonreconstruction,identification,momentum/energyscale375
and resolution, and isolation criteria [43–45]. Photon reconstruction and identification efficiencies are376
evaluatedfromsamplesof Z!eeandZ+
events[45,46]. Theluminositymeasurementwascalibrated 377
duringdedicatedbeam-separationscans,usingthesamemethodologyasthatdescribedinRef.[47]. The378
uncertainty of this measurement is found to be 2.1%.379
In total, 35 sources of experimental uncertainties are identified pertaining to one or more physics objects380
considered. For each source the one-standard-deviation ¹1șconfidence interval (CI) is propagated to 381
a1CI around the nominal SM expectation. The total experimental uncertainty of the SM expecta- 382
tion is obtained from the sum in quadrature of these 35 1CIs and the uncertainty of the luminosity 383
measurement.384
Theoreticalmodelling uncertainties Twodifferent sourcesof uncertaintyin thetheoreticalmodelling 385
oftheSMproductionprocessesareconsidered. Afirstuncertaintyisassignedtoaccountforourknowledge386
of the cross-sections for the inclusive processes. A second uncertainty is used to cover the modelling of387
theshapeofthedifferentialcross-sections. Inordertoderivethemodellinguncertainties,eithervariations388
of the QCD factorization, renormalization, resummation and merging scales are used or comparisons of389
the nominal MC samples with alternative ones are used. For some SM processes additional modelling390
uncertaintiesareincluded. AppendixA.2describesalltheoreticaluncertaintiesconsideredforthevarious391
SM processes. The total uncertainty is taken as the sum in quadrature of the two components and the392
statistical uncertainty of the MC prediction.393
3.2.2 Validation procedures394
The evaluated SM processes, together with their standard selection cuts and the studied validation dis-395
tributions, are detailed in Table 3. These validation distributions rely on inclusive selections to probe396
the general agreement between data and simulation and are evaluated in restricted ranges where large397
new-physics contributions have been excluded by previous direct searches.398
There are some cases in which the validation procedure finds modelling problems and MC background399
correctionsareneeded(multijets,
¹
ș+jets). Inothercases,theaffectedeventclassesareexcludedfrom 400
theanalysisastheirSMexpectationmostlyarisesfromobjectmisidentification(e.g.jetsreconstructedas401
electrons)whichispoorlymodelledinMCsimulation. Theexcludedclassesare: 1e1j,1e2j,1e3j,1e4j, 402
1e1b,1e1b1j,1e1b2j,1e1b3j. Eventclassescontainingasingleobject,aswellasthosecontainingonly 403
Emiss
Tand a lepton are also discarded from the analysis. 404
18th July 2018 – 12:25 16

ATLAS DRAFT
Table 3: A summary of the SM processes and their inclusive selections used to validate the background modelling.
For each selection the pT,, anddistributions of the objects used in the selection and of additional jets are
included as validation distributions by default. Additional validation distributions are listed per selection. In all
cases,‘jet(s)’referstoboth b-taggedandnon- b-taggedjets,exceptwhere‘ b-jet’ismentionedexplicitly. HT¹jetsșis
defined as the scalar pTsum of all the jets in the event. Some selections rely on the transverse mass ( mT¹`;Emiss
Tș)
which is defined as »2pT¹`șEmiss
T¹1cos
¹`;Emiss
Tșș¼12.N¹¹b-șjetsșis the number of ( b-)jets in an event. For
the distance variables
Rand
, the two instances of the objects with the minimum distance between them are
used. The HT¹jetsș,mT¹`;Emiss
Tș,pT¹„șandminvvalidation distributions are evaluated in restricted ranges where
largenew-physicscontributionshavebeenexcludedbypreviousdirectsearches. Same-flavouropposite-chargesign
lepton pairs are referred to as SFOS pairs.
Physics process Event selection Additional validation distributions
W¹!`ș+jets 1 lepton, Emiss
T>25GeV N¹jetsșN¹b-jetsșmT¹`;Emiss
TșHT¹jetsș
andmT¹`;Emiss
Tș>50GeV
R¹`;jetș
¹`;Emiss
Tș
¹jet;Emiss
Tș
& N¹jetsș3 HT¹jetsș
& N¹b-jetsș1 mT¹`,Emiss
Tș
& N¹b-jetsș2 mT¹`,Emiss
Tș
Z¹!„ș+jets 1 SFOS pair N¹jetsșN¹b-jetsșminv¹„șpT¹„ș
66<minv¹„ș<116GeV HT¹jetsș
R¹`;`ș
R¹`;jetș
& N¹jetsș2 HT¹jetsș
& N¹b-jetsș1 minv¹„șpT¹„ș
& N¹b-jetsș2 minv¹„șpT¹„ș
W+
¹
ș same selection as W¹!`ș+jets same distributions as W¹!`ș+jets and
and 1(2) additional photon(s)
R¹`;
ș
Z+
¹
ș same selection as Z¹!„ș+ jets same distributions as Z¹!„ș+jets and
and 1(2) additional photon(s)
R¹`;
ș

¹
ș+jets 1(2) photon(s), no leptons N¹jetsșN¹b-jetsșHT¹jetsșminv¹

ș
and at least 1(0) jet(s)
R¹
;
ș
R¹
;jetș
t¯t! 1 lepton, at least 2 b-jets and at least 2 light jets N¹jetsșN¹b-jetsșminv¹j jș
W¹!j jș+ 50<minv¹j jș<110GeV mT¹`;Emiss
Tș
R¹jet;jetș
R¹`;jetș
W¹!`ș+electron channel: mT¹e,Emiss
Tș>50GeV
bb muon channel: mT¹,Emiss
Tș>60GeV
Emiss
T>40GeV
Diboson
WW 1 electron and 1 muon of opposite charge and no jets
Emiss
T>50GeV
W Z 1 SFOS pair¹„ș minv¹„ș(SFOS pair(s))
and 1 lepton of different flavour ¹`0ș pT¹„ș(SFOS pair(s))
66<minv¹„ș<116GeV mT¹`0;Emiss
Tș
Emiss
T>50GeV and mT¹`0;Emiss
Tș>50GeV
R¹`;`ș
¹`;Emiss
Tș
Z Z 2 SFOS pairs
66<minv¹„ș<116GeV (both SFOS pairs)
Multijets at least 2 jets N¹jetsșN¹b-jetsș
R¹jet;jetș
no leptons or photons Emiss
Tmeff
18th July 2018 – 12:25 17

ATLAS DRAFT
3.2.3 Corrections to the MC background405
The MC samples for multijet and
+jets production, while giving a good description of kinematic 406
variables, predict an overall cross-section and a jet multiplicity distribution that disagrees with data.407
Following step 2, correction procedures were applied.408
In classes containing only jandbthe multijet MC samples are scaled to data with normalization factors 409
ranging between approximately 0.8 and 1.2. The normalization factors are derived separately in each410
exclusivejetmultiplicityclassbyequatingtheexpectedtotalnumberofeventstotheobservednumberof411
events. If a channel contains less than five data events, no modifications are made.412
For
+jets event classes the same rescaling procedure is applied to classes with exactly one photon, no 413
leptons or Emiss
T, and any number of jets. 414
TheS/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1MCgeneratorhasaknowndeficiencyinthemodellingof Emiss
Tduetotoolargeforward 415
jet activity. This results in a visible mismodelling of the Emiss
Tdistribution in event classes with two 416
photons, which also affects the meffdistribution. To correct for this mismodelling a reweighting [48] is 417
applied to the background events containing two real photons (

+jets). The diphoton MC events are 418
reweighted as a function of Emiss
Tand of the number of selected jets to match the respective distributions 419
inthedatafortheinclusivediphotonsampleintherange Emiss
T<100GeV. Innoothereventclasseswas 420
the mismodeling large enough to warrant such a procedure.421
3.2.4 Comparison of the event yields with the MC prediction422
After classification, 704 event classes are found with at least one data event or an SM expectation greater423
than0.1events. ThedataandthebackgroundpredictionsfromMCsimulationfortheseclassesareshown424
inFigure3andAppendixC.Agreementisobservedbetweendataandthepredictioninmostoftheevent425
classes. In events classes having more than two b-jets and where the SM expectation is dominated by t¯t 426
production, the nominal SM expectation is systematically slightly below the data. Data events are found427
in 528 out of 704 event classes. These include events with up to four leptons (muons and/or electrons),428
threephotons,twelvejetsandeight b-jets. Thereare18eventclasseswithanSMexpectationoflessthan 429
0.1events;nomorethantwodataeventsareobservedinanyofthese,andtheyarenotconsideredfurther430
in the analysis. The remaining 686 classes are retained for statistical analysis.431
3.3 Step 3: Sensitive variables and search algorithm432
In order to quantitatively determine the level of agreement between thedata and the SM expectation, and433
to identify regions of possible deviations, this analysis uses an algorithm for multiple hypothesis testing.434
The algorithm locates a single region of largest deviation for specific observables in each event class.435
Inthefollowing,analgorithmderivedfromthealgorithmusedinRef.[5]isappliedtothe2015dataset.436
18th July 2018 – 12:25 18

ATLAS DRAFT
1e1j
missT
E1e2j
missT
E1e3j
missT
E1e4j
missT
E1e5j
missT
E1e6j
missT
E1e7j
missT
E1e8j
missT
E1e1b
missT
E1e1b1j
missT
E1e1b2j
missT
E1e1b3j
missT
E1e1b4j
missT
E1e1b5j
missT
E1e1b6j
missT
E1e1b7j
missT
E1e2b
missT
E1e2b1j
missT
E1e2b2j
missT
E1e2b3j
missT
E1e2b4j
missT
E1e2b5j
missT
E1e2b6j
missT
E1e3b
missT
E1e3b1j
missT
E1e3b2j
missT
E1e3b3j
missT
E1e3b4j
missT
E1e4b
missT
E1e4b1j
missT
E1e4b2j
missT
E1j
µ1missT
E2j
µ1missT
E3j
µ1missT
E4j
µ1missT
E5j
µ1missT
E6j
µ1missT
E7j
µ1missT
E8j
µ1missT
E1b
µ1missT
E1b1j
µ1missT
E1b2j
µ1missT
E1b3j
µ1missT
E1b4j
µ1missT
E1b5j
µ1missT
E1b6j
µ1missT
E1b7j
µ1missT
E2b
µ1missT
E2b1j
µ1missT
E2b2j
µ1missT
E2b3j
µ1missT
E2b4j
µ1missT
E2b5j
µ1missT
E2b6j
µ1missT
E3b
µ1missT
E3b1j
µ1missT
E3b2j
µ1missT
E3b3j
µ1missT
E3b4j
µ1missT
E4b
µ1missT
E4b1j
µ1missT
E4b2j
µ1missT
EEvents / class1
−10110210310410510610Data 2015
3-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbs + single lepton
TmissE
Figure 3: The number of events in data, and for the different SM background predictions considered, for classes
with large Emiss
T, one lepton and ( b-)jets (no photons). The classes are labelled according to the multiplicity and
type( e,,
,j,b,Emiss
T)ofthereconstructedobjectsforthegiveneventclass. Thehatchedbandsindicatethetotal
uncertaintyoftheSMprediction. Thisfigureshows60outof704eventclasses,theremainingeventclassescanbe
found in Figures 11–23 of Appendix C.
3.3.1 Choice of variables437
Foreacheventclass,the meffandminvdistributionsareconsideredintheformofhistograms. Theinvariant 438
massiscomputedfromallvisibleobjectsintheevent,withnoattempttousethe Emiss
Tinformation. These 439
variableshavebeenwidelyusedinsearchesfornewphysics,andaresensitivetoalargerangeofpossible440
signals, manifesting either as bumps, deficits or wide excesses. Several other commonly used kinematic441
variables have also been studied for various models, but were not found to significantly increase the442
sensitivity. The approach is however not limited to these variables, as discussed in Section 2.443
For each histogram, the bin widths h¹xșas a function of the abscissa xare determined using: 444
h¹xș=vutNobjectsÕ
i=1k22
i¹x2ș
where Nobjectsis the number of objects in the event class, i¹x2șis the expected detector resolution 445
for the pTof object ievaluated at pT=x2and=0, and kis the width of the bin in standard 446
deviations. An exception to this is the missing transverse momentum resolution ( Emiss
T), which is a 447
function ofÍET, whereÍETis approximated by the effective mass minus the Emiss
Tobject requirement: 448
Emiss
T¹ÍET=x200GeVș. The Emiss
Tobject is only considered in the binning of the effective mass 449
histograms. A1interval is used for the bin width ( k=2) for all objects except for photons and 450
electrons,forwhicha 3intervalisused( k=6)toavoidhavingtoofinelybinnedhistogramswithfew 451
MC events. This results in variable bin widths with values ranging from 20 GeV to about 2000 GeV. For452
a given event class, the scan starts at a value of the scanned observable larger than two times the sum of453
the minimum pTrequirement of each contributing object considered (e.g. 100 GeV for a 2class). This 454
minimises spurious deviations which might arise from insufficiently well modelled threshold regions.455
18th July 2018 – 12:25 19

ATLAS DRAFT
3.3.2 Algorithm to search for deviations of the data from the expectation456
Thealgorithmidentifiesthesingleregionwiththelargestupwardordownwarddeviationinadistribution,457
providedintheformofahistogram,astheregionofinterest(ROI).Thetotalnumberofindependentbins458
is36936, leading to 518320combinations of contiguous bins (regions 4) with an SM expectation larger 459
than 0.01 events. For each region with an SM expectation larger than 0.01, the statistical estimator p0460
is calculated as defined in Eqs. (1)–(3). Here, p0is to be interpreted as a local p0-value. The region of 461
largestdeviationfoundbythealgorithmistheregionwiththesmallest p0-value. Suchamethodisableto 462
find narrow resonances and single outstanding bins, as well as signals spread over large regions of phase463
space in distributions of any shape.464
Toillustratetheoperationofthealgorithm,sixexampledistributionsarepresented. Figure4(a)showsthe465
invariantmassdistributionoftheeventclasswithonephoton,threelightjetsandlargemissingtransverse466
momentum ( Emiss
T1
3j), which has the smallest pchannel-value in the minvscan. Figure 4(b) shows the 467
effective mass distribution of the event class with one muon, one electron, four b-jets and two light jets 468
(11e4b2j), which has the smallest pchannel-value in the meffscan. Figure 4(c) shows the invariant mass 469
distribution of the event class with one electron, one photon, two b-jets and two light jets ( 1e1
2b2j). 470
Figure 4(d) shows the effective mass distribution of the event class with six light jets ( 6j). Figure 4(e) 471
shows the invariant mass distribution of the event class with two muons, a light jet and large missing472
transversemomentum( Emiss
T21j)andFigure4(f)showstheeffectivemassdistributionoftheeventclass 473
with three light jets and large missing transverse momentum ( Emiss
T3j). The regions with the largest 474
deviation found by the search algorithm in these distributions, an excess in Figures 4(a), 4(b), 4(c) and475
4(f), and a deficit in Figures 4(d) and 4(e), are indicated by vertical dashed lines.476
To minimize the impact of few MC events, 213992regions where the background prediction has a total 477
relativeuncertaintyofover100%arediscardedbythealgorithm. Discardingaregionforcesthealgorithm478
to consider a different or larger region in the event class, or if no region in the event class satisfies479
the condition, to discard the entire event class. 5For all discarded regions with Nobs>3ap0-value is 480
calculated. If the p0-value is smaller than the pchannel-value (or if there is no ROI and hence no pchannel- 481
value),itisevaluatedmanuallybycomparingitwiththedistributionof pchannel-valuesfromthescan. This 482
is done for 27 event classes among which the smallest p0-value observed in a discarded region is 0:01. 483
To model the analysis of discarded regions in pseudo-experiments, regions are allowed to have larger484
uncertainties if they fulfil the Nobs>3criterion. 485
Inadditiontomonitoringregionsdiscardedduetoatotaluncertaintyinexcessof100%,regionsdiscarded486
due to NSM<0:01but with Nobs>3would also be monitored individually; however, no such region has 487
been observed.488
Tables4and5listthethreeeventclasseswiththelargestdeviationsinthe minvandmeffscansrespectively. 489
The largest deviation reported by a dedicated search using the same dataset was observed in an inclusive490
diphoton data selection at a diphoton mass of around 750 GeV with a local significance of 3:9[49]. 491
Due to the different event selections and background estimates the excess has a lower significance in this492
analysis. The excess was not confirmed in a dedicated analysis with 2016 data [50].493
4A histogram of nbins has 1 region of ncontiguous bins, 2 regions of n1contiguous bins, etc. down to nregions of single
bins. Therefore,ithasÍn
i=1i=n¹n+1ș2regions. Whencombiningbins,backgrounduncertaintiesareconservativelytreated
as correlated among the bins.
5In the minvandmeffscan respectively, 72 and 87 event classes are discarded since they have no ROI.
18th July 2018 – 12:25 20

ATLAS DRAFT
GeV]
[ invm50010001500 2000 2500 3000 Events / bin1
−10110210ROI:
DATA 9
0.66
±BKG 2.15 -VALUE 0.0028
0PData 2015Total SM MC stat.
MC stat.⊕Exp. syst. SM processes:
)
γ(γZ/W+Z/W+jets γ
+tt+jets ttdi-/triboson
single top other
-1 = 13 TeV, 3.2 fbs3j
γ1missT
Event class: E ATLASGeV]
[ invm50010001500 2000 2500 3000 Data / MC01234
(a)
GeV]
[ effm100012001400160018002000Events / bin1
−10110210ROI:
DATA 2
0.04
±BKG 0.04 -VALUE 0.0027
0PData 2015Total SM MC stat.
MC stat.⊕Exp. syst. SM processes:
+jets
tt3-/4-top +Z/W/WW
ttother -1 = 13 TeV, 3.2 fbs1e4b2j
µEvent class: 1 ATLASGeV]
[ effm100012001400160018002000Data / MC01234 (b)
GeV]
[ invm10001500 2000 2500 3000 Events / bin1
−10110210ROI:
DATA 3
0.13
±BKG 0.27 -VALUE 0.0078
0PData 2015Total SM MC stat.
MC stat.⊕Exp. syst. SM processes:
γ
+tt+jets tt)
γ(γZ/W++Z/W/WW ttsingle top
Z/W+jets other
-1 = 13 TeV, 3.2 fbs2b2j
γEvent class: 1e1 ATLASGeV]
[ invm10001500 2000 2500 3000 Data / MC01234
(c)
GeV]
[ effm20003000 4000 5000 6000 7000 Events / bin1
−10110210310410510610ROI:DATA 3
3.05
±BKG 10.01 -VALUE 0.11
0PData 2015Total SM MC stat.
MC stat.⊕Exp. syst. SM processes:
multijets
Z/W+jets other
-1 = 13 TeV, 3.2 fbsEvent class: 6j
ATLASGeV]
[ effm20003000 4000 5000 6000 7000 Data / MC01234 (d)
GeV]
[ invm5001000 1500 2000 Events / bin1
−10110210ROI:
DATA 0
1.97
±BKG 5.52 -VALUE 0.049
0PData 2015Total SM MC stat.
MC stat.⊕Exp. syst. SM processes:
+jets
ttdi-/triboson single top
Z/W+jets other
-1 = 13 TeV, 3.2 fbs1j
µ2missT
Event class: E ATLASGeV]
[ invm5001000 1500 2000 Data / MC01234
(e)
GeV]
[ effm10001500200025003000350040004500Events / bin1
−10110210310410510ROI:
DATA 2
0.14
±BKG 0.24 -VALUE 0.059
0PData 2015Total SM MC stat.
MC stat.⊕Exp. syst. SM processes:
Z/W+jets
+jets ttdi-/triboson
) γ(γZ/W+other
-1 = 13 TeV, 3.2 fbs3j
missT
Event class: E ATLASGeV]
[ effm10001500200025003000350040004500Data / MC01234 (f)
Figure 4: Example distributions showing the region of interest (ROI), i.e. the region with the smallest p0-value,
between the vertical dashed lines. (a) Emiss
T1
3jchannel, which has the largest deviation in the minvscan. (b)
11e4b2jchannel,whichhasthelargestdeviationinthe meffscan. (c)Anupwardfluctuationinthe minvdistribution
of the 1e1
2b2jchannel. (d) A downward fluctuation in the meffdistribution of the 6jchannel. (e) A downward
fluctuationinthe minvdistributionofthe Emiss
T21jchannel. (f)Anupwardfluctuationinthe meffdistributionofthe
Emiss
T3jchannel. ThehatchedbandincludesallsystematicandstatisticaluncertaintiesfromMCsimulations. Inthe
ratio plots the inner solid uncertainty band shows the statistical uncertainty from MC simulations, the middle solid
band includes the experimental systematic uncertainty, and the hatched band includes the theoretical systematic
uncertainty.
18th July 2018 – 12:25 21

ATLAS DRAFT
Table 4: List of the three channels with the smallest pchannel-values in the scan of the minvdistributions.
Largest deviations in minvscan
Channel pchannel¹
103șNobs NSMNSMRegion»GeV¼
Emiss
T1
3j 2:81 9 2 :150:66 670–732
11e4b2j 2:91 2 0 :0420:0371227–1569
1e1b4j 3:44 160 105 14 726–809
Table 5: List of the three channels with the smallest pchannel-values in the scan of the meffdistributions.
Largest deviations in meffscan
Channel pchannel¹
103șNobs NSMNSMRegion»GeV¼
11e4b2j 2:66 2 0 :0400:036992–1227
11
5j 3:98 4 0 :450:18 750–895
3b1j 4:87 4 0 :420:243401–3923
3.4 Step 4: Generation of pseudo-experiments494
As described in Section 2.4, pseudo-experiments are generated to derive the probability of finding a495
p0-value of a given size, for a given observable and algorithm. The pchannel-value distributions of the 496
pseudo-experiments and their statistical properties can be compared with the pchannel-value distribution 497
obtained from data. Correlations in the uncertainties of the SM expectation affect this probability and498
their effect is taken into account in the generation of pseudo-data as outlined in the following.499
For the experimental uncertainties, each of the 35 sources of uncertainty is varied independently by500
drawing a value at random from a Gaussian pdf. This value is assumed to be 100% correlated across501
all bins and event classes. The uncertainty in the normalization of the various backgrounds is also502
consideredas100%correlated. Likewise,theoreticalshapeuncertainties,includingthoseestimatedfrom503
scale variations or the differences with alternative generators, are assumed to be 100% correlated, with504
the exception of the uncertainties which are used for some SM processes with small cross-sections. The505
latter uncertainties are assumed to be uncorrelated, both between event classes and between bins of the506
same event class. Scale variations are applied in the generation of pseudo-experiments by varying the507
renormalization,factorization,resummationandmergingscalesindependently. Thevaluesforeachscale508
of a given pseudo-experiment are 100% correlated between all bins and event classes. The scales are509
correlated between processes of the same type which are generated with a similar generator set-up, i.e.510
scales are correlated among the WZ
+jets processes, among all the diboson processes, among the 511
t¯t+WZprocesses, and among the single-top processes. 512
Changing the size of the theoretical uncertainties by a factor of two leads to a change of less than 5% in513
thelog10¹pminșthresholdsatwhichadedicatedanalysisistriggered. Thecorrelationassumptionsinthe 514
theoreticaluncertaintieswerealsotested. Figure5showstheeffectofchangingthecorrelationassumption515
for all theoretical shape uncertainties that are nominally taken as 100% correlated. This test decorrelates516
the bin-by-bin variations due to the theoretical shape uncertainties in the pseudo-data while retaining517
the correlation when summing over selected bins in the scan, thus testing the impact of an incorrect518
assumption in the correlation model. By comparing the nominal assumption of 100% correlation with519
18th July 2018 – 12:25 22

ATLAS DRAFT
a 50% correlated component, and a fully uncorrelated assumption, the threshold at which a dedicated520
analysis is triggered is changed by a negligible amount.521
)
min
(p
10
-log
0
1
2
3
4
5
6
Fraction of pseudo-experiments
2
−
10
1
−
10
1
Simulation
σ
1
σ
2
σ
3
σ
4
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
1 channel
≥
in
min
< p
channel
p
inv
variable: m
100% correlated
50% correlated
0% correlated
(a)
)
min
(p
10
-log
0
1
2
3
4
5
6
Fraction of pseudo-experiments
2
−
10
1
−
10
1
Simulation
σ
1
σ
2
σ
3
σ
4
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
3 channels
≥
in
min
< p
channel
p
eff
variable: m
100% correlated
50% correlated
0% correlated (b)
Figure 5: A comparison of different correlation assumptions for scale variations: 100% correlated; 50% correlated
and 50% uncorrelated; and 100% uncorrelated. The fractions of pseudo-experiments in the scan of the minv
distribution having at least one pchannel-value smaller than pminare shown on the left (a), while the fractions in the
scan of the meffdistribution having at least three pchannel-values smaller than pminare shown on the right (b).
3.5 Step 5: Evaluation of the sensitivity of the strategy522
3.5.1 Sensitivity to Standard Model processes523
The sensitivity of the procedure is evaluated with two different methods that either use a modified524
backgroundestimationthroughtheremovalofSMprocessesorinwhichsignalcontributionsareaddedto525
thepseudo-datasample. Asafigureofmerit,thefractionof‘signal’pseudo-experimentswith Pexp;i<5% 526
fori=1;2;3is computed. 527
Figure 6 shows how removing the W Zprocess from the background prediction affects the three smallest 528
expected pchannel-values. In Figures 6(a) and 6(b), the dashed curves show the nominal expected pchannel529
distribution obtained from pseudo-experiments. These define the pminthresholds for which Pexp;i<5% 530
andverticaldashedlinesaredrawnatthethresholdvalues. Thesolidlinesshowthe pchanneldistributions 531
obtainedbytestingpseudo-experimentsgeneratedfromtheSMpredictionagainstthemodifiedbackground532
prediction which has the W Zdiboson process removed. It can be observed that in this case the meffscan 533
ismoresensitive;thefractionof‘signal’pseudo-experimentswith Pexp;i<5%isatleast80%inallthree 534
cases i=1;2;3. 535
Additionally, in Figures 6(a) and 6(b), the three smallest pchannel-values observed in the data are shown 536
byarrows,bothwhentestedagainstthefullSMprediction(dashed)andwhentestedagainstthemodified537
prediction(solid). Forallthreecases( i=1;2;3),Pexp;i<5%isfoundagain. Thismeansthatadedicated 538
analysis would be performed for the three event classes in which the pchannel-values are observed, i.e. 3, 539
12e1j, and 21e1j, likely resulting in the discovery of an unexpected signal due to W Zproduction. 540
Figures6(c)and6(d)showsthe meffdistributionsofthedatawiththefullSMpredictionandthemodified 541
prediction respectively. This test uses the conclusion from Section 3.6 and is performed in retrospect. In542
18th July 2018 – 12:25 23

ATLAS DRAFT
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
2
−
10
1
−
10
1
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
min
< p
channel
p
no. of channels with
inv
variable: m
1
≥
2
≥
3
≥
SM (w.r.t. SM)
Data (w.r.t. SM)
SM
(w.r.t. SM, WZ removed)
Data
(w.r.t. SM, WZ removed)
(a)
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
2
−
10
1
−
10
1
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
min
< p
channel
p
no. of channels with
eff
variable: m
1
≥
2
≥
3
≥
SM (w.r.t. SM)
Data (w.r.t. SM)
SM
(w.r.t. SM, WZ removed)
Data
(w.r.t. SM, WZ removed) (b)
GeV]
[
eff
m
200
400
600
800
Events / bin
1
−
10
1
10
2
10
ROI:
DATA 41
5.81
±
BKG 31.70
-VALUE 0.28
0
P
Data 2015
Total SM
WZ
ZZ
+jets
t
t
Higgs
Other
-1
= 13 TeV, 3.2 fb
s
µ
Event class: 3
ATLAS
GeV]
[
eff
m
200
400
600
800
Data / MC
0
1
2
3
4
(c)
GeV]
[
eff
m
200
400
600
800
Events / bin
1
−
10
1
10
2
10
ROI:
DATA 41
3.80
±
BKG 6.09
-VALUE 4.4e-08
0
P
Data 2015
Total SM
ZZ
+jets
t
t
Higgs
Other
-1
= 13 TeV, 3.2 fb
s
µ
Event class: 3
ATLAS
GeV]
[
eff
m
200
400
600
800
Data / MC
0
1
2
3
4 (d)
Figure 6: The fraction of pseudo-experiments which have at least one, two and three pchannel-values below a given
pmin, given for both the pseudo-experiments generated from the nominal SM expectation and tested against the
nominal expectation (dashed) and for those tested against the modified expectation (‘SM, W Zremoved’) in which
theW Zdibosonprocessisremoved(solid). The minvscanisshownin(a)andthe meffscanin(b). Thescanresults
ofthedatatestedagainstthemodifiedbackgroundpredictionareindicatedwithsolidarrows. Forreferencethescan
resultsundertheSMhypothesisareplottedasdashedarrows. Thelargestdeviationafterremovingthe W Zprocess
fromthebackgroundexpectationisfoundinthe meffdistributionofthe 3eventclass. Thedistributionsofthedata
and the expectation with both W Zincluded and W Zremoved are shown in (c) and (d) respectively.
18th July 2018 – 12:25 24

ATLAS DRAFT
thecaseofasignificantdeviation,thistestwouldbeperformedwithpseudo-datatoassessthesensitivity543
of the search to a missing background.544
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
2
−
10
1
−
10
1
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
min
< p
channel
p
no. of channels with
inv
variable: m
1
≥
2
≥
3
≥
SM (w.r.t. SM)
Data (w.r.t. SM)
SM
removed)
γ
+
t
(w.r.t. SM, t
Data
removed)
γ
+
t
(w.r.t. SM, t
(a)
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
2
−
10
1
−
10
1
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
min
< p
channel
p
no. of channels with
eff
variable: m
1
≥
2
≥
3
≥
SM (w.r.t. SM)
Data (w.r.t. SM)
SM
removed)
γ
+
t
(w.r.t. SM, t
Data
removed)
γ
+
t
(w.r.t. SM, t (b)
GeV]
[
eff
m
600
800
1000
1200
1400
1600
1800
Events / bin
1
−
10
1
10
2
10
ROI:
DATA 14
1.45
±
BKG 5.88
-VALUE 0.019
0
P
Data 2015
Total SM
γ
+
t
t
)
γ
(
γ
Z/W+
+jets
t
t
Z/W+jets
)+jets
γ
(
γ
Other
-1
= 13 TeV, 3.2 fb
s
1b2j
γ
Event class: 1e1
ATLAS
GeV]
[
eff
m
600
800
1000
1200
1400
1600
1800
Data / MC
0
1
2
3
4
(c)
GeV]
[
eff
m
600
800
1000
1200
1400
1600
1800
Events / bin
1
−
10
1
10
2
10
ROI:
DATA 22
1.50
±
BKG 3.98
-VALUE 1.4e-06
0
P
Data 2015
Total SM
)
γ
(
γ
Z/W+
+jets
t
t
Z/W+jets
)+jets
γ
(
γ
Other
-1
= 13 TeV, 3.2 fb
s
1b2j
γ
Event class: 1e1
ATLAS
GeV]
[
eff
m
600
800
1000
1200
1400
1600
1800
Data / MC
0
1
2
3
4 (d)
Figure 7: The fraction of pseudo-experiments which have at least one, two and three pchannel-values below a given
pmin, given for both the pseudo-experiments generated from the nominal SM expectation and tested against the
nominal expectation (dashed) and for those tested against the modified expectation (‘SM, t¯t
removed’) in which
thet¯t
processisremoved(solid). The minvscanisshownin(a)andthe meffscanin(b). Thescanresultsofthedata
tested against the modified background prediction are indicated with solid arrows. For reference the scan results
under the SM hypothesis are plotted as dashed arrows. The largest deviation after removing the t¯t
process from
the background expectation is found in the meffdistribution of the 1e1
1b2jevent class. The distributions of the
data and the expectation with both t¯t
included and t¯t
removed are shown in (c) and (d) respectively.
Figure 7 shows the effect of removing the t¯t+
process. Again the meffscan is slightly more sensitive, 545
andatleast70%of‘signal’pseudo-experimentshave Pexp;i<5%inallthreecases i=1;2;3. Inthedata, 546
Pexp;i<5%is found again for all three cases ( i=1;2;3). A dedicated analysis would be performed for 547
thethreeclasses 11
2b1j,1e1
2b2j,and 11
1b3j,likelyresultinginthediscoveryofan unexpected 548
signal due to t¯t+
production. 549
It is interesting to note that these discoveries would have been made without a priori knowledge of the550
existence of these processes.551
18th July 2018 – 12:25 25

ATLAS DRAFT
3.5.2 Sensitivity to new-physics signals552
)
min
(p
10
-log
0
1
2
3
4
5
6
7
Fraction of pseudo-experiments
0.0
0.2
0.4
0.6
0.8
1.0
Simulation
σ
1
σ
2
σ
3
σ
4
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
1 channel
≥
in
min
< p
channel
p
SM + Z' (SSM)
inv
variable: m
SM-only
= 2.0 TeV
Z'
m
= 2.5 TeV
Z'
m
= 3.0 TeV
Z'
m
= 3.5 TeV
Z'
m
= 4.0 TeV
Z'
m
(a)
)
min
(p
10
-log
0
1
2
3
4
5
6
7
Fraction of pseudo-experiments
0.0
0.2
0.4
0.6
0.8
1.0
Simulation
σ
1
σ
2
σ
3
σ
4
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
1 channel
≥
in
min
< p
channel
p
1
0
χ
∼
+
t
t
→
g
~
SM +
eff
variable: m
SM-only
) = (1000,1) GeV
1
0
χ
∼
,m
g
~
(m
) = (1200,1) GeV
1
0
χ
∼
,m
g
~
(m
) = (1400,1) GeV
1
0
χ
∼
,m
g
~
(m
) = (1600,1) GeV
1
0
χ
∼
,m
g
~
(m (b)
Figure 8: The fraction of pseudo-experiments in which a deviation is found with a pchannel-value smaller than a
given pmin. Distributionsareshownforpseudo-experimentsgeneratedfromtheSMexpectation(circularmarkers),
and after injecting signals of (a) inclusive Z0decays or (b) gluino pairs with ˜g!t¯t˜0
1decays and various masses.
The line corresponding to the injection of a Z0boson with a mass of 4 TeV and a gluino with a mass of 1600 GeV
overlap with the line obtained from the SM-only pseudo-experiments due to the small signal cross-section.
Figure 8 shows the sensitivity for the two benchmark signals considered as a function of the mass of the553
producedparticle. Forthe Z0model,wherethemassoftheresonancecanbereconstructedfromitsdecay 554
products,thesensitivitytothesignalisfoundtobethelargestinthescanofthe minvdistribution. Gluinos 555
undergo a cascade decay process to the lightest neutralino, which is undetected and leads to missing556
transverse momentum. It is not possible to fully reconstruct an event from gluino pair production due to557
the presence of neutralinos in the final state. The sensitivity to the gluino signal is therefore found to be558
the largest in the meffscan, where a broad excess at large values of this quantity is expected. 559
Exclusion and discovery sensitivity have to be carefully distinguished when the results of this search are560
comparedwithmodel-basedsearches. Anexclusionsensitivityatthe95%CLinadedicatedsearchroughly561
corresponds to a single class having a p0-value for a discovery test smaller than 0.05. Consequently, the 562
sensitivity to a benchmark signal corresponding to a given particle mass should be compared with the563
discovery sensitivity of other searches for a Z0boson or gluino. 564
As previously described, a deviation for which Pexp;i<5%promotes the selection to a signal region for 565
a dedicated analysis. By applying this sensitivity criterion, it can be seen in Figure 8 that this search is566
sensitive to a Z0boson with a mass of about 2.5 TeV as more than 90% of the signal-injected pseudo- 567
experiments show a deviation for which Pexp;1<5%. Similar sensitivity is expected for a gluino with a 568
massofabout1TeV. Theprobabilityofdiscoveringanew-physicssignalinanewdatasetwithadedicated569
search in the selected event classes is estimated in the next section.570
3.5.3 Sensitivity of a second independent dataset571
In step 7 a dedicated analysis of a deviation is performed on an independent dataset. The sensitivity of572
step 7 is evaluated with pseudo-experiments.573
18th July 2018 – 12:25 26

ATLAS DRAFT
A first pseudo-experiment emulates the original dataset on which this analysis is performed. The scan574
algorithm is applied after which eight different cases can be distinguished where Pexp;iis either larger or 575
smaller than 5% for i=1;2;3. In seven cases at least one Pexp;i<5%and a new independent pseudo- 576
experiment is generated to emulate a new independent dataset with the same integrated luminosity. The577
one, two or three data selections for which Pexp;i<5%are applied to the second pseudo-experiment 578
to obtain the p0-values for these selections. Although the systematic uncertainties may be reduced by 579
applyingdata-drivenestimatesofthebackground,theywereassumedtohavethesamesizeinthesecond580
pseudo-experimenttomakeaconservativeestimate. Thesystematicuncertaintiesarealsoexpectedtobe581
partially correlated between two datasets but here they were assumed to be uncorrelated.582
In four of the seven cases Pexp;1<5%and these cases are grouped together into a ‘one signal region’ 583
class. This class shows the sensitivity when there would be only a single data-derived signal region. The584
case where only Pexp;3<5%is called the ‘three signal region’ class. The two remaining cases where 585
Pexp;2<5%define the ‘two signal region’ class. These classes show cases when a data-derived signal 586
region is found only by a combination of two and three regions.587
)
min,k
(p
10
log
n
k=0
Σ
–
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
7
−
10
6
−
10
5
−
10
4
−
10
3
−
10
2
−
10
1
−
10
1
ATLAS
Simulation
-1
= 13 TeV, 3.2 fb
s
inv
variable: m
SM + Z' (SSM)
1
2
signal regions
SM-only
= 2.5 TeV
Z'
m
σ
5
σ
5
σ
5
3
(a)
)
min,k
(p
10
log
n
k=0
Σ
–
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
7
−
10
6
−
10
5
−
10
4
−
10
3
−
10
2
−
10
1
−
10
1
ATLAS
Simulation
-1
= 13 TeV, 3.2 fb
s
eff
variable: m
1
0
χ
+
t
t
→
g
~
SM +
1
2 signal regions
SM-only
= 1.0 TeV
g
~
m
σ
5
σ
5
σ
53 (b)
Figure 9: Thefractionofcasesinwhichthefirstpseudo-experimenthas Pexp;i<5%andtriggersasecondpseudo-
experiment which yields a value for Ín
k=1log10¹p0kșsmaller than a given value ( Ín
k=1log10¹pmin;kș). Here n
denotes the minimum number of event classes (1, 2 or 3) for which Pexp;i<0:05. The distribution is shown for
pseudo-experiments generated from the SM expectation, and from an SM-plus-signal expectation of (a) inclusive
decays of a Z0boson with mass mZ0=2:5TeV or (b) gluino pairs with mass m˜g=1:0TeV and g!t¯t˜0
1decays.
The signal region tested in the follow-up pseudo-experiment is defined by the preceding pseudo-experiment. The
5thresholds are obtained by extrapolating the SM-only fractions to 5:7
107and are indicated at the top of the
figure for n=1(left), n=2(middle) or n=3(right) event classes.
For each of the three classes ( n=1;2;3) the statistical estimatorÎn
k=1p0kis computed. Figure 9 shows, 588
asafunctionoftheestimatorinthelogarithmicform Ín
k=1log10¹p0kșthefractionofcasesinwhichthe 589
firstpseudo-experimenthas Pexp;i<5%andtriggersasecondpseudo-experimentwhichyieldsavaluefor 590
Ín
k=1log10¹p0kșabove a threshold given by the value on the horizontal axis. This is done for pseudo- 591
experiments generated from the SM expectation (SM-only) and for pseudo-experiments generated from592
the SM expectation plus Z0or gluino signal contributions. The 5lines are derived from the fractions 593
givenbytheSM-onlylines,asthesecorrespondtotheprobabilityoffalsepositiveswhichdefinesthelevel594
of significance. It should be noted that the SM-only lines with circular markers start at a fraction of 0:05 595
byconstructionofthe Pexp;i<5%definition. The n=2andn=3linesshowthegaininsensitivitywhen 596
18th July 2018 – 12:25 27

ATLAS DRAFT
adeviationinoneortwochannels,respectively,isnotlargeenoughtodefineadata-derivedsignalregion.597
Itdoesnotshowthegaininsensitivityfromconsideringmultiplechannelswhenasinglechanneldefines598
a signal region. Signals which produce one or more large deviations therefore lower the number of cases599
in the n=2and 3 categories, while signals producing deviations close to the Pexp;i<5%threshold (e.g. 600
for higher Z0masses) would raise the number of cases in the n=2and 3 categories. A Z0boson with a 601
mass of 2.5 TeV would yield a discovery in almost all cases.602
In the case of a 1:0TeV gluino the sensitivity is about 5. The sensitivity increases to about 1:1TeV 603
if the integrated luminosity of the two datasets combined is increased to about 10 fb1by doubling the 604
size of the second dataset to 6.4 fb1. ATLAS has determined the discovery sensitivity of the dedicated 605
searches for gluinos decaying to quarks, a Wboson and a neutralino. This dedicated search estimates a 606
localsignificance(i.e. notcorrectedfortrialfactorsofthededicatedsearches)of 5withaluminosityof 607
10 fb1for gluinos with a mass of 1:35TeV assuming a systematic uncertainty of 25% [51]. 608
It should be noted that, with this strategy, these signals are found without any a priori assumptions about609
the model, including the mass and the decay chain of the gluinos or the Z0boson. It can therefore be 610
concluded that this procedure could also be sensitive to possible unexpected signals for new physics.611
3.6 Step 6: Results612
In step 6 the pchannel-values found in the analysis of the 2015 ATLAS data are interpreted by comparing 613
them with the pchannel-values found in the pseudo-experiments. 614
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
0.0
0.2
0.4
0.6
0.8
1.0
σ
1
σ
2
σ
3
σ
4
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
in
min
< p
channel
p
no. of channels with
inv
variable: m
1
≥
2
≥
3
≥
Data
(a)
)
min
(p
10
-log
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
0.0
0.2
0.4
0.6
0.8
1.0
σ
1
σ
2
σ
3
σ
4
σ
5
5%
ATLAS
-1
= 13 TeV, 3.2 fb
s
in
min
< p
channel
p
no. of channels with
eff
variable: m
1
≥
2
≥
3
≥
Data (b)
Figure 10: The fractions of pseudo-experiments ( Pexp;i¹pminș) which have at least one, two or three pchannel-values
(circular, square, and rhombic markers respectively) smaller than a given threshold ( pmin) in the scans of (a) the
minvdistributions,andof(b)the meffdistributions. Thecolouredarrowsrepresentthethreesmallest pchannel-values
observed in data. Dashed lines are drawn at Pexp;i=5%and at the pmin-values corresponding to 1,2,3,4
and5local significances.
Figure 10 shows the fractions of pseudo-experiments that have at least one, two or three pchannel-values 615
below a given threshold ( pmin) in the scans of the minvand the meffdistributions. The statistical tests in 616
both distributions for the three leading pchannel-values are all consistent at the Pexp;i>50%level with the 617
SMexpectationof pchannel-valuesobtainedfrompseudo-experiments. Changingthesizeofthetheoretical 618
18th July 2018 – 12:25 28

ATLAS DRAFT
shape uncertainties by a factor of two leads to a change in the three smallest pchannel-values of a factor of 619
two. It therefore does not lead to an appreciable change in the result.620
In conclusion, no significant deviations are found in the 2015 dataset and consequently no dedicated621
analysis using data-derived signal regions (step 7) is initiated.622
4 Conclusions623
Astrategyforamodel-independentgeneralsearchtofindpotentialindicationsofnewphysicsispresented.624
Eventsareclassifiedaccordingtotheirfinalstateintomanyeventclasses. Foreacheventclassanautomated625
search algorithm tests whether the data is compatible with the Monte Carlo simulated Standard Model626
expectation in various distributions sensitive to the effects of new physics. For each distribution the627
searchalgorithmisrepeatedonmanypseudo-experimentstomakeafrequentistestimateofthestatistical628
significance of the three largest deviations. A data selection in which a significant deviation is observed629
defines a data-derived signal region which will be tested on a new dataset in a dedicated analysis with an630
improved background model.631
The strategy has been applied to the data collected by the ATLAS experiment at the LHC during 2015,632
corresponding to a total of 3.2 fb1of 13 TeV ppcollisions. In this dataset, exclusive event classes 633
containingelectrons,muons,photons, b-taggedjets,non- b-taggedjetsandmissingtransversemomentum 634
have been scanned for deviations from the MC-based SM prediction in the distributions of the effective635
mass and the invariant mass. Sensitivity studies with various toy signals ( t¯t+
,W Z, gluino, and Z0636
production) have shown that the strategy could discover signals for new physics without an a priori637
knowledge of the existence of the processes.638
No significant deviations are found in the 2015 dataset and consequently no dedicated analysis using639
data-derivedsignalregionsisperformed. Thestrategydiscussedinthispaperwillbeusefultosearchfor640
signals of unknown particles and interactions in the subsequent Run 2 datasets.641
18th July 2018 – 12:25 29

ATLAS DRAFT
Acknowledgements642
We thank CERN for the very successful operation of the LHC, as well as the support staff from our643
institutions without whom ATLAS could not be operated efficiently.644
We acknowledge the support of ANPCyT, Argentina; YerPhI, Armenia; ARC, Australia; BMWFW645
and FWF, Austria; ANAS, Azerbaijan; SSTC, Belarus; CNPq and FAPESP, Brazil; NSERC, NRC and646
CFI, Canada; CERN; CONICYT, Chile; CAS, MOST and NSFC, China; COLCIENCIAS, Colombia;647
MSMT CR, MPO CR and VSC CR, Czech Republic; DNRF and DNSRC, Denmark; IN2P3-CNRS,648
CEA-DRF/IRFU, France; SRNSFG, Georgia; BMBF, HGF, and MPG, Germany; GSRT, Greece; RGC,649
Hong Kong SAR, China; ISF, I-CORE and Benoziyo Center, Israel; INFN, Italy; MEXT and JSPS,650
Japan;CNRST,Morocco;NWO,Netherlands;RCN,Norway;MNiSWandNCN,Poland;FCT,Portugal;651
MNE/IFA, Romania; MES of Russia and NRC KI, Russian Federation; JINR; MESTD, Serbia; MSSR,652
Slovakia; ARRS and MIZŠ, Slovenia; DST/NRF, South Africa; MINECO, Spain; SRC and Wallenberg653
Foundation,Sweden;SERI,SNSFandCantonsofBernandGeneva,Switzerland;MOST,Taiwan;TAEK,654
Turkey;STFC,UnitedKingdom;DOEandNSF,UnitedStatesofAmerica. Inaddition,individualgroups655
and members have received support from BCKDF, the Canada Council, CANARIE, CRC, Compute656
Canada, FQRNT, and the Ontario Innovation Trust, Canada; EPLANET, ERC, ERDF, FP7, Horizon657
2020 and Marie Skłodowska-Curie Actions, European Union; Investissements d’Avenir Labex and Idex,658
ANR, Région Auvergne and Fondation Partager le Savoir, France; DFG and AvH Foundation, Germany;659
Herakleitos,ThalesandAristeiaprogrammesco-financedbyEU-ESFandtheGreekNSRF;BSF,GIFand660
Minerva, Israel; BRF, Norway; CERCA Programme Generalitat de Catalunya, Generalitat Valenciana,661
Spain; the Royal Society and Leverhulme Trust, United Kingdom.662
The crucial computing support from all WLCG partners is acknowledged gratefully, in particular from663
CERN, the ATLAS Tier-1 facilities at TRIUMF (Canada), NDGF (Denmark, Norway, Sweden), CC-664
IN2P3(France),KIT/GridKA(Germany),INFN-CNAF(Italy),NL-T1(Netherlands),PIC(Spain),ASGC665
(Taiwan), RAL (UK) and BNL (USA), the Tier-2 facilities worldwide and large non-WLCG resource666
providers. Major contributors of computing resources are listed in Ref. [52].667
18th July 2018 – 12:25 30

ATLAS DRAFT
Appendix668
A Details of the Monte Carlo samples669
A.1 Monte Carlo programs and settings670
Samples of multijet production were simulated with 2!2matrix elements (ME) at leading order (LO) 671
using the P/y.sc/t.sc/h.sc/i.sc/a.sc8.186 generator [17]. The A14 [25] set of shower and multiple parton interactions 672
parameters (tune) was used together with the NNPDF2.3LO PDF set [24]. Alternative multijet samples673
with 2!2ME at LO were generated with H/e.sc/r.sc/w.sc/i.sc/g.sc++ 2.7.1 [53] with the UEEE5 underlying-event tune 674
and the CTEQ6L1 [33] PDF set, and with S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 [20] with ME for 2!2and2!3partons at 675
LO merged using the ME+PSLO prescription. All S/h.sc/e.sc/r.sc/p.sc/a.scsamples use the CT10 [54] PDF set and the 676
S/h.sc/e.sc/r.sc/p.sc/a.scparton shower [55] with a dedicated shower tuning developed by the S/h.sc/e.sc/r.sc/p.sc/a.scauthors. 677
Events containing leptonic decays of a WorZbosons with associated jets ( W/Z+jets) were simulated 678
using the S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 generator [56]. Matrix elements were calculated using the Comix [57] and 679
Open-Loops [58] generators. They include up to two partons at NLO and four partons at leading order680
(LO), merged using the ME+PS@NLO prescription [59]. Samples with WandZdecaying hadronically 681
are also generated with S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 including up to four partons at LO. The W/Z+jets events were 682
normalized to their inclusive next-to-next-to-leading-order (NNLO) cross-sections [60, 61]. Simulated683
samplesofmassivevectorbosonsproducedinassociationwithoneortworealphotonswerealsogenerated684
with S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 with a ME calculated at LO for up to three partons. They are scaled to their NLO 685
cross-sections computed with MCFM [62, 63].686
Samples of prompt photon production in association with jets (
+jets) were generated using S/h.sc/e.sc/r.sc/p.sc/a.sc 687
2.1.1. For these samples up to four real parton emissions are included at LO. Events containing two688
promptphotons(

+jets)werealsogeneratedwith S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1. Matrixelementswerecalculatedwith 689
up to two partons at LO. The gluon-induced box process is also included. These samples were scaled to690
data following the procedure described in Section 3.2.3.691
Top-quark pair production events, and single top quarks in the Wt- and s-channels, were simulated using 692
theP/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v2 [64] generator with the CT10 PDF set, as detailed in Ref. [65]. The top quark mass 693
was set to 172.5 GeV. The hdampparameter, which regulates the transverse momentum of the first extra 694
emission beyond the Born configuration and thus controls the pTof the t¯tsystem, was set to the mass of 695
the top quark. Electroweak t-channel single-top-quark events were generated using the P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v1 696
generator. This generator uses the four-flavour scheme for the NLO matrix-element calculations together697
with the fixed four-flavour PDF set CT10f4. For all top-quark processes, top-quark spin correlations are698
preserved (for the single-top t-channel, top quarks were decayed using MadSpin [66]). An alternative699
sampleof t¯twasgeneratedwiththe S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1generator,includinguptooneadditionalpartonatNLO 700
anduptofouradditionalpartonsatLOaccuracy,interfacedtothepartonshowerusingtheME+PS@NLO701
prescription. The parton shower (PS), fragmentation, and the underlying event of the P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc 702
samples were simulated using P/y.sc/t.sc/h.sc/i.sc/a.sc6.428 [67] with the CTEQ6L1 PDF set and the corresponding 703
Perugia2012tune(P2012)[68]. The t¯tandsingle-top-quarkeventswerenormalizedtotheNNLOcross- 704
section including the resummation of soft gluon emission at next-to-next-to-logarithm accuracy [69–71]705
using Top++2.0 [72]. Both the default and the alternative t¯tsamples were corrected to reproduce the 706
NNLOprediction[73,74]ofthetopquark pTandthe pTofthe t¯tsystem. Thecontributionof t¯t+b¯bwas 707
18th July 2018 – 12:25 31

ATLAS DRAFT
generatedseparatelywith S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1atNLO;thecalculationwasperformedinthefour-flavourscheme 708
and with the CT10f4 PDF set.709
Diboson samples were generated with the S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1 generator, and are described in Ref. [75]. The 710
matrixelementscontainthe WW,W ZandZ Zprocessesandallotherdiagramswithfourorsixelectroweak 711
vertices (such as same-electric-charge Wboson production in association with two jets, WWj j). Fully 712
leptonic triboson processes ( WWW,WW Z,W Z ZandZ Z Z) with on-shell bosons and up to six charged 713
leptons were also simulated using S/h.sc/e.sc/r.sc/p.sc/a.sc2.1.1. The ME for the Z Zprocesses were calculated at NLO 714
for up to one additional parton; final states with two and three additional partons were calculated at LO.715
TheW ZandWWprocesses were calculated at NLO with up to three extra partons at LO using the 716
ME+PS@NLO prescription. The WWfinal states were generated without bottom quarks in the hard- 717
scattering process, to avoid contributions from top-quark-mediated processes. The triboson processes718
were calculated with the same configuration and with up to two extra partons at LO. The generator719
cross-sections were used for the normalization of these backgrounds.720
Samples of top quark production in association with vector bosons [76] ( W,Z,
andWW, including the 721
non-resonant
Zcontributions) were generated at LO with MG5_aMC@NLO 2.2.2 [32] interfaced 722
toP/y.sc/t.sc/h.sc/i.sc/a.sc8.186, with up to two ( t¯tW), one ( t¯tZ) or no ( t¯tWW,t¯t
) extra partons included in the 723
matrix element. The A14 tune was used together with the NNPDF2.3LO PDF set. The t¯t
sample 724
uses a fixed QCD renormalization and factorization scale of 2mt, and the top decay was performed in 725
MG5_aMC@NLOtoaccountforhardphotonradiationfromthetopdecayproducts. Thesamegenerator726
wasalsousedtosimulatethe tZ,3-topand4-topquarksprocesses. The t¯tW,t¯tZ,t¯tWWand4-topsamples 727
werenormalizedtotheirNLOcross-sections[32]whiletheLOcross-sectionfromthegeneratorwasused728
fortZand 3-top quarks. 729
The Higgs boson mass was set to 125 GeV and all SM Higgs boson decay modes were considered.730
The production of the SM Higgs boson in the gluon–gluon fusion (ggF) and vector-boson fusion (VBF)731
channels was modelled using the P/o.sc/w.sc/h.sc/e.sc/g.sc-B/o.sc/x.sc v2 generator with the CT10 PDF set. It was interfaced 732
toP/y.sc/t.sc/h.sc/i.sc/a.sc8.186 with the CTEQ6L1 PDF set and the AZNLO tune [77]. Production of a Higgs boson 733
in association with a pair of top quarks was simulated using MG5_aMC@NLO 2.2.2 interfaced to734
H/e.sc/r.sc/w.sc/i.sc/g.sc++ 2.7.1 [78] for showering and hadronization. The UEEE5 underlying-event tune was used 735
togetherwiththeCT10(matrixelement)andCTEQ6L1(partonshower)PDFsets. Simulatedsamplesof736
SMHiggsbosonproductioninassociationwitha WorZbosonwereproducedwith P/y.sc/t.sc/h.sc/i.sc/a.sc8.186,using 737
theA14tuneandtheNNPDF2.3LOPDFset. Eventswerenormalizedtotheirmostaccuratecross-sections738
calculations (typically NNLO) [79].739
To avoid double counting, events with a hard photon from final-state radiation were removed from the740
multijet, t¯tandW/Z+jets samples. 741
A.2 Theoretical uncertainties742
The inclusive Wand Zcross-sections are known at NNLO, with an uncertainty of about 5% [60, 743
61]. Modelling uncertainties for W+jets and Z+jets are determined by varying the renormalization, 744
factorizationandresummationscalesintheMEbyfactors 0:5and2,togetherwithachangeofthemerging 745
scale from 20 GeV to 15 GeV or 30 GeV.746
For top quark pair or single-top production, processes known to NNLO+NNLL [72] or approximate747
NNLO [69–71], respectively, the cross-section uncertainty is 7%. The modelling uncertainty for t¯tis 748
18th July 2018 – 12:25 32

ATLAS DRAFT
determined by comparing the nominal P/o.sc/w.sc/h.sc/e.sc/g.sc+P/y.sc/t.sc/h.sc/i.sc/a.sc NLO+PS sample with an alternative sample 749
generated with S/h.sc/e.sc/r.sc/p.sc/a.scincluding up to two partons at NLO and four at LO accuracy in the ME. The 750
single-top quark uncertainty is estimated by varying the renormalization and factorization scales, and by751
changingthe hdampparameterandtheshowertuneofthe P/o.sc/w.sc/h.sc/e.sc/g.sc+P/y.sc/t.sc/h.sc/i.sc/a.sc sample. Anuncertaintyinthe 752
interference between the Wtandt¯tproduction is estimated by comparing the nominal Wtsample, where 753
all doubly resonant NLO Wtdiagrams are removed, with a sample where the cross-section contribution 754
from Feynman diagrams containing two top quarks is subtracted [80].755
Diboson cross-sections ( WW,W ZandZ Z) are calculated at NLO, and a 6% uncertainty, evaluated with 756
the MCFM program [62, 63], is applied to their cross-sections. Their modelling uncertainty is evaluated757
analogously to V+jets by varying the scales used to perform the calculation. For W+
andZ+
758
samples, which are computed at LO, a 20% uncertainty in the cross-section is assumed, with a further759
20% modelling uncertainty assigned in accord with the measurement in Ref. [81].760
Thecross-sectionsfortop-quarkpairproductioninassociationwithoneortwovectorbosonsarecalculated761
atNLOandanuncertaintyof15%isused[32]. Theirmodellinguncertaintyisevaluatedfromvariations762
oftherenormalizationandfactorizationscales,togetherwithachangeinthemergingscale. For t¯t+
,an 763
additional12%uncertaintyinthenormalizationisassumed,whileauniform30%uncertaintyisassigned764
to the modelling.765
Multijet and
+jets processes are scaled to data following the procedure described in Section 3.2.2. 766
Therefore, no uncertainty is applied to their normalization. For multijets the maximum bin-by-bin767
difference between the P/y.sc/t.sc/h.sc/i.sc/a.sc8 nominal sample and alternative samples generated with S/h.sc/e.sc/r.sc/p.sc/a.scand 768
H/e.sc/r.sc/w.sc/i.sc/g.sc++ isconsideredasashapeuncertainty. Inadditionthestandarddeviationofthe100replicasets 769
oftheNNPDF2.3LOPDFisused. Themodellinguncertaintyfor
+jetsisestimatedfromscalevariations 770
with the same methodology as for the V+jets samples. The uncertainty in the

+jets modelling is 771
instead taken to be 30% from parton-level comparisons of samples with varied scales.772
Aconservativeuncertaintyof20%[79]isassumedforHiggsproductionintheggF,VBFand VHchannels. 773
A further uncertainty of 20% is assigned as a shape uncertainty.774
For the subdominant triboson (including V+

),ttH, 3-top, 4-top and tZproduction processes a 50% 775
uncertainty is assigned to the event yields, similarly to Ref. [82]. Uncertainties associated with the PDFs776
arefoundtobenegligibleinallbutthemultijetsamples,forwhichtheyhavebeenexplicitlyestimated.777
B Details of the object reconstruction778
Electroncandidatesarereconstructedfromanisolatedelectromagneticcalorimeterenergydepositmatched779
to an ID track and are required to have jj<2:47, a transverse momentum pT>10GeV, and to pass a 780
loose likelihood-based identification requirement [44, 83]. The likelihood input variables include mea-781
surementsofcalorimetershowershapesandmeasurementsoftrackpropertiesfromtheID.Thecandidate782
electrons are selected if the matched tracks have a transverse impact parameter significance relative to783
the reconstructed primary vertex of jd0j¹d0ș<5. Candidates within the transition region between the 784
barrel and endcap electromagnetic calorimeters, 1:37<jj<1:52, are removed. 785
Muon candidates are reconstructed in the region jj<2:7from muon spectrometer tracks matched to ID 786
tracks. Themuoncandidatesareselectediftheyhaveatransversemomentumabove10GeVandpassthe787
medium identification requirements defined in Ref. [43], based on selections on the number of hits in the788
18th July 2018 – 12:25 33

ATLAS DRAFT
different ID and muon spectrometer subsystems, and the significance of the charge to momentum ratio789
qp. 790
All candidate leptons (electrons and muons) are used for the object overlap removal, as discussed in791
Section 3.1.3. Tighter requirements on the lepton candidates are imposed, which are then referred to792
as ‘signal’ electrons or muons and are used further in the analysis, i.e. to establish the accuracy of the793
backgroundmodellingprocessesortoclassifytheevents. Signalelectronsmustsatisfyatightlikelihood-794
based identification requirement [44, 83]. Signal muons must fulfil the requirement of jd0j¹d0ș<3. 795
The track associated with a signal lepton must have a longitudinal impact parameter relative to the796
reconstructed primary vertex, z0, satisfyingjz0
sinj<0:5mm. Isolation requirements are applied to 797
both the signal electrons and muons. The calorimeter isolation is computed as the sum of the energies of798
calorimeterenergyclustersinaconeofsize
R=0:2aroundthelepton. Trackisolationisdefinedasthe 799
scalar sum of the pTof tracks within a variable-size cone around the lepton, in a cone of size
R=0:2 800
(0.3) for electron (muon) transverse momenta pT<50GeV ( pT<33GeV) and of size
R=10GeVpT801
forpT>50GeV ( pT>33GeV). The efficiency of these criteria increases with the lepton transverse 802
momentum, reaching 95% at 25 GeV and 99% at 60 GeV, as determined in a control sample of Zdecays 803
intoleptonsselectedwithatag-and-probetechnique[43,44]. CorrectionsareappliedtotheMCsamples804
to match the leptons’ trigger, reconstruction and isolation efficiencies in data.805
Photon candidates are reconstructed from an isolated electromagnetic calorimeter energy deposit and806
are required to satisfy the tight identification criteria described in Refs. [46, 84]. Furthermore, photons807
are required to have pT>25GeV andjj<2:37, excluding the barrel–endcap calorimeter transition 808
in the range 1:37<jj<1:52. Photons must further satisfy isolation criteria based on both track and 809
calorimeterinformation[46]. Aftercorrectingforcontributionsfrompile-up,theenergywithinaconeof810

R=0:4aroundtheclusterbarycentreisrequiredtobelessthan 2:45GeV+0:022p

T,where p

Tisthe 811
transversemomentumofthephotoncandidate. Theenergyoftracksinaconeof
R=0:2shouldbeless 812
than 0:05p

T. 813
Jet candidates are reconstructed with the anti- ktalgorithm [38] implemented in the FastJet package [85] 814
witharadiusparameterof R=0:4,usingasinputthree-dimensionalenergyclustersinthecalorimeter[86] 815
calibrated to the electromagnetic scale. The reconstructed jets are then calibrated to the jet energy scale816
(JES)derivedfromsimulationandinsitucorrectionsbasedon13TeVdata[42,87]. Foralljetstheexpected817
averageenergycontributionfrompile-upclustersissubtractedaccordingtothejetareaprescription[88].818
Quality criteria are imposed to identify jets arising from non-collision sources or detector noise and any819
event containing such a jet is removed [89]. All jet candidates are required to have pT>20GeV and 820
jj<2:8. 821
Identification of jets containing b-hadrons ( b-tagging) is performed with a multivariate discriminant, 822
MV2/c.sc20, making use of track impact parameters, the b- and c-hadron flight paths inside the jet and 823
reconstructed secondary vertices [39, 40]. The algorithm working point used corresponds to a 77%824
average efficiency obtained for b-jets in simulated t¯tevents. The rejection factors for light-quark jets, 825
c-quark jets and hadronically decaying -leptons in simulated t¯tevents are approximately 140, 4.5 and 826
10, respectively [40]. Jets with jj<2:5which satisfy this b-tagging requirement are identified as b-jet 827
candidates. To compensate for differences between data and MC simulation in the b-tagging efficiencies 828
and mistag rates, correction factors are applied to the simulated samples [40].829
18th July 2018 – 12:25 34

ATLAS DRAFT
C Event yields for all event classes830
1j
γ12j
γ13j
γ14j
γ15j
γ16j
γ17j
γ18j
γ19j
γ110j
γ11b
γ11b1j
γ11b2j
γ11b3j
γ11b4j
γ11b5j
γ11b6j
γ11b7j
γ11b8j
γ12b
γ12b1j
γ12b2j
γ12b3j
γ12b4j
γ12b5j
γ12b6j
γ12b7j
γ13b
γ13b1j
γ13b2j
γ13b3j
γ13b4j
γ13b5j
γ14b
γ14b1j
γ14b2j
γ1γ
21j
γ22j
γ23j
γ24j
γ25j
γ26j
γ27j
γ21b
γ21b1j
γ21b2j
γ21b3j
γ21b4j
γ21b5j
γ21b6j
γ22b
γ22b1j
γ22b2j
γ22b3j
γ23b2j
γ2γ
31j
γ3Events / class1
−10110210310410510610710810Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbsphoton(s) (normalized to data)
Figure 11: The number of events in data, and for the different SM background predictions considered, for classes
withatleastonephotonand( b-)jets(noleptonsor Emiss
T). Theclassesarelabelledaccordingtothemultiplicityand
type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. In event classes with four or more
dataevents,the
+jetsand

+jetsMCsamplesarescaledtodatainthesingle-photonanddiphotoneventclasses
respectively. The hatched bands indicate the total uncertainty of the SM prediction.
18th July 2018 – 12:25 35

ATLAS DRAFT
1e5j
1e6j
1e7j
1e8j
1e9j
1e1b4j
1e1b5j
1e1b6j
1e1b7j
1e1b8j
1e1b9j
1e2b
1e2b1j
1e2b2j
1e2b3j
1e2b4j
1e2b5j
1e2b6j
1e2b7j
1e2b8j
1e3b
1e3b1j
1e3b2j
1e3b3j
1e3b4j
1e3b5j
1e3b6j
1e3b7j
1e4b
1e4b1j
1e4b2j
1e4b3j
1e4b4j
1e4b5j
1e5b
1e5b1j
1e5b2j
Events / class1
−10110210310410510610Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbssingle electron
Figure 12: The number of events in data, and for the different SM background predictions considered, for classes
with one electron and ( b-)jets (no muons, photons or Emiss
T). The classes are labelled according to the multiplicity
and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the
total uncertainty of the SM prediction.
1j
µ12j
µ13j
µ14j
µ15j
µ16j
µ17j
µ18j
µ19j
µ11b
µ11b1j
µ11b2j
µ11b3j
µ11b4j
µ11b5j
µ11b6j
µ11b7j
µ11b8j
µ11b9j
µ12b
µ12b1j
µ12b2j
µ12b3j
µ12b4j
µ12b5j
µ12b6j
µ12b7j
µ12b8j
µ13b
µ13b1j
µ13b2j
µ13b3j
µ13b4j
µ13b5j
µ13b6j
µ13b7j
µ14b
µ14b1j
µ14b2j
µ14b3j
µ14b4j
µ14b5j
µ15b
µ15b1j
µ15b2j
µ1Events / class1
−10110210310410510610710Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbssingle muon
Figure 13: The number of events in data, and for the different SM background predictions considered, for classes
with one muon and ( b-)jets (no electrons, photons or Emiss
T). The classes are labelled according to the multiplicity
and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the
total uncertainty of the SM prediction.
18th July 2018 – 12:25 36

ATLAS DRAFT
1e
µ11e1j
µ11e2j
µ11e3j
µ11e4j
µ11e5j
µ11e6j
µ11e7j
µ11e1b
µ11e1b1j
µ11e1b2j
µ11e1b3j
µ11e1b4j
µ11e1b5j
µ11e1b6j
µ11e1b7j
µ11e2b
µ11e2b1j
µ11e2b2j
µ11e2b3j
µ11e2b4j
µ11e2b5j
µ11e2b6j
µ11e3b
µ11e3b1j
µ11e3b2j
µ11e3b3j
µ11e3b4j
µ11e4b
µ11e4b1j
µ11e4b2j
µ11e4b5j
µ1Events / class1
−10110210310410510Data 2015
3-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbsmuon + electron
Figure 14: The number of events in data, and for the different SM background predictions considered, for classes
withonemuon,oneelectronand( b-)jets(nophotonsor Emiss
T). Theclassesarelabelledaccordingtothemultiplicity
and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the
total uncertainty of the SM prediction.
2e
2e1j
2e2j
2e3j
2e4j
2e5j
2e6j
2e7j
2e8j
2e1b
2e1b1j
2e1b2j
2e1b3j
2e1b4j
2e1b5j
2e1b6j
2e1b7j
2e2b
2e2b1j
2e2b2j
2e2b3j
2e2b4j
2e2b5j
2e3b
2e3b1j
2e3b2j
2e3b3j
2e3b4j
2e4b
2e4b1j
µ
21j
µ22j
µ23j
µ24j
µ25j
µ26j
µ27j
µ28j
µ21b
µ21b1j
µ21b2j
µ21b3j
µ21b4j
µ21b5j
µ21b6j
µ21b7j
µ22b
µ22b1j
µ22b2j
µ22b3j
µ22b4j
µ22b5j
µ22b6j
µ23b
µ23b1j
µ23b2j
µ23b3j
µ24b
µ24b1j
µ24b2j
µ2Events / class1
−10110210310410510610710810910Data 2015
3-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbsdilepton
Figure 15: Thenumberofeventsindata,andforthedifferentSMbackgroundpredictionsconsidered,forclasseswith
two same-flavour leptons and ( b-)jets (no photons or Emiss
T). The classes are labelled according to the multiplicity
and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the
total uncertainty of the SM prediction.
18th July 2018 – 12:25 37

ATLAS DRAFT
3e
3e1j
3e2j
3e3j
3e4j
3e1b
3e1b1j
3e1b2j
3e1b3j
3e2b
3e2b1j
3e2b2j
2e
µ12e1j
µ12e2j
µ12e3j
µ12e4j
µ12e1b
µ12e1b1j
µ12e1b2j
µ12e1b3j
µ12e2b
µ12e2b1j
µ12e2b2j
µ12e2b5j
µ11e
µ21e1j
µ21e2j
µ21e3j
µ21e4j
µ21e1b
µ21e1b1j
µ21e1b2j
µ21e1b3j
µ21e2b
µ21e2b1j
µ21e2b2j
µ2µ
31j
µ32j
µ33j
µ34j
µ35j
µ31b
µ31b1j
µ31b2j
µ31b3j
µ31b4j
µ32b
µ32b1j
µ32b2j
µ31e2b
µ34e
4e1j
4e1b
3e
µ12e
µ22e1j
µ22e2j
µ22e1b
µ22e1b1j
µ2µ
41j
µ42j
µ4Events / class1
10210310Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbs3- and 4-lepton
Figure 16: The number of events in data, and for the different SM background predictions considered, for classes
with three or four leptons and ( b-)jets (no photons or Emiss
T). The classes are labelled according to the multiplicity
and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the
total uncertainty of the SM prediction.
γ
1e11j
γ1e12j
γ1e13j
γ1e14j
γ1e15j
γ1e16j
γ1e17j
γ1e11b
γ1e11b1j
γ1e11b2j
γ1e11b3j
γ1e11b4j
γ1e11b5j
γ1e12b
γ1e12b1j
γ1e12b2j
γ1e12b3j
γ1e12b4j
γ1e12b5j
γ1e13b
γ1e13b1j
γ1e13b2j
γ1e13b3j
γ1e1γ
1e21j
γ1e22j
γ1e23j
γ1e21b1j
γ1e21b2j
γ1e2γ
1µ11j
γ1µ12j
γ1µ13j
γ1µ14j
γ1µ15j
γ1µ16j
γ1µ11b
γ1µ11b1j
γ1µ11b2j
γ1µ11b3j
γ1µ11b4j
γ1µ11b5j
γ1µ12b
γ1µ12b1j
γ1µ12b2j
γ1µ12b3j
γ1µ12b4j
γ1µ13b
γ1µ13b1j
γ1µ13b2j
γ1µ1γ
2µ11j
γ2µ12j
γ2µ13j
γ2µ11b2j
γ2µ1Events / class1
−10110210310410Data 2015
3-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbssingle lepton + photon(s)
Figure 17: The number of events in data, and for the different SM background predictions considered, for classes
with one lepton, at least one photon and ( b-)jets (no Emiss
T). The classes are labelled according to the multiplicity
and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the
total uncertainty of the SM prediction.
18th July 2018 – 12:25 38

ATLAS DRAFT
γ
2e11j
γ2e12j
γ2e13j
γ2e14j
γ2e15j
γ2e11b
γ2e11b1j
γ2e11b2j
γ2e11b3j
γ2e12b
γ2e12b1j
γ2e12b2j
γ2e13b
γ2e1γ
2e2γ
3e1γ
1e1µ11j
γ1e1µ12j
γ1e1µ13j
γ1e1µ11b
γ1e1µ11b1j
γ1e1µ11b2j
γ1e1µ11b3j
γ1e1µ11b4j
γ1e1µ12b
γ1e1µ12b1j
γ1e1µ12b2j
γ1e1µ1γ
1µ21j
γ1µ22j
γ1µ23j
γ1µ24j
γ1µ21b
γ1µ21b1j
γ1µ21b2j
γ1µ21b3j
γ1µ22b
γ1µ22b1j
γ1µ22b2j
γ1µ2γ
2µ21j
γ2µ22j
γ2µ2γ
1e1µ21j
γ1e1µ2γ
1µ31e
µ3Events / class1
10210310Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbsmultilepton + photon(s)
Figure 18: The number of events in data, and for the different SM background predictions considered, for classes
with at least two leptons, at least one photon and ( b-)jets (no Emiss
T). The classes are labelled according to the
multiplicity and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands
indicate the total uncertainty of the SM prediction.
1j
missT
E2j
missT
E3j
missT
E4j
missT
E5j
missT
E6j
missT
E7j
missT
E8j
missT
E9j
missT
E1b
missT
E1b1j
missT
E1b2j
missT
E1b3j
missT
E1b4j
missT
E1b5j
missT
E1b6j
missT
E1b7j
missT
E1b8j
missT
E2b
missT
E2b1j
missT
E2b2j
missT
E2b3j
missT
E2b4j
missT
E2b5j
missT
E2b6j
missT
E2b7j
missT
E2b8j
missT
E3b
missT
E3b1j
missT
E3b2j
missT
E3b3j
missT
E3b4j
missT
E3b5j
missT
E4b
missT
E4b1j
missT
E4b2j
missT
E4b3j
missT
E4b4j
missT
E5b1j
missT
EEvents / class1
−10110210310410510610710Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbs + jets
TmissE
Figure 19: The number of events in data, and for the different SM background predictions considered, for classes
withlarge Emiss
Tand( b-)jets(noleptonsorphotons). Theclassesarelabelledaccordingtothemultiplicityandtype
(e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands indicate the total
uncertainty of the SM prediction.
18th July 2018 – 12:25 39

ATLAS DRAFT
γ
1missT
E1j
γ1missT
E2j
γ1missT
E3j
γ1missT
E4j
γ1missT
E5j
γ1missT
E6j
γ1missT
E1b
γ1missT
E1b1j
γ1missT
E1b2j
γ1missT
E1b3j
γ1missT
E1b4j
γ1missT
E1b5j
γ1missT
E2b
γ1missT
E2b1j
γ1missT
E2b2j
γ1missT
E2b3j
γ1missT
E2b4j
γ1missT
E3b
γ1missT
E3b1j
γ1missT
E3b2j
γ1missT
Eγ
2missT
E1j
γ2missT
E2j
γ2missT
E1b1j
γ2missT
E1b2j
γ2missT
Eγ
1e1missT
E1j
γ1e1missT
E2j
γ1e1missT
E3j
γ1e1missT
E4j
γ1e1missT
E1b
γ1e1missT
E1b1j
γ1e1missT
E1b2j
γ1e1missT
E1b3j
γ1e1missT
E1b4j
γ1e1missT
E2b
γ1e1missT
E2b1j
γ1e1missT
E2b2j
γ1e1missT
E2b3j
γ1e1missT
E1j
γ1e2missT
Eγ
2e1missT
E1b
γ2e1missT
Eγ
1µ1missT
E1j
γ1µ1missT
E2j
γ1µ1missT
E3j
γ1µ1missT
E4j
γ1µ1missT
E5j
γ1µ1missT
E1b
γ1µ1missT
E1b1j
γ1µ1missT
E1b2j
γ1µ1missT
E1b3j
γ1µ1missT
E1b4j
γ1µ1missT
E2b
γ1µ1missT
E2b1j
γ1µ1missT
E2b2j
γ1µ1missT
E2b3j
γ1µ1missT
Eγ
1e1µ1missT
E1j
γ1e1µ1missT
E2j
γ1e1µ1missT
E1b
γ1e1µ1missT
E1b1j
γ1e1µ1missT
E1j
γ1µ2missT
E2j
γ1µ2missT
EEvents / class1
10210310Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbs + photon(s) (+ lepton)
TmissE
Figure 20: The number of events in data, and for the different SM background predictions considered, for classes
withlarge Emiss
T,atleastonephoton,leptonsand( b-)jets. Theclassesarelabelledaccordingtothemultiplicityand
type( e,,
,j,b,Emiss
T)ofthereconstructedobjectsforthegiveneventclass. Thehatchedbandsindicatethetotal
uncertainty of the SM prediction.
1e
µ1missT
E1e1j
µ1missT
E1e2j
µ1missT
E1e3j
µ1missT
E1e4j
µ1missT
E1e5j
µ1missT
E1e6j
µ1missT
E1e1b
µ1missT
E1e1b1j
µ1missT
E1e1b2j
µ1missT
E1e1b3j
µ1missT
E1e1b4j
µ1missT
E1e1b5j
µ1missT
E1e2b
µ1missT
E1e2b1j
µ1missT
E1e2b2j
µ1missT
E1e2b3j
µ1missT
E1e2b4j
µ1missT
E1e3b
µ1missT
E1e3b1j
µ1missT
E1e3b2j
µ1missT
E1e4b
µ1missT
EEvents / class1
10210310Data 2015
3-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbs + muon + electron
TmissE
Figure 21: The number of events in data, and for the different SM background predictions considered, for classes
with large Emiss
T, one muon, one electron and ( b-)jets (no photons). The classes are labelled according to the
multiplicity and type ( e,,
,j,b,Emiss
T) of the reconstructed objects for the given event class. The hatched bands
indicate the total uncertainty of the SM prediction.
18th July 2018 – 12:25 40

ATLAS DRAFT
2e
missT
E2e1j
missT
E2e2j
missT
E2e3j
missT
E2e4j
missT
E2e5j
missT
E2e1b
missT
E2e1b1j
missT
E2e1b2j
missT
E2e1b3j
missT
E2e1b4j
missT
E2e2b
missT
E2e2b1j
missT
E2e2b2j
missT
E2e2b3j
missT
E2e3b
missT
E2e3b1j
missT
E3e
missT
E3e1j
missT
E3e2j
missT
E2e
µ1missT
E2e1j
µ1missT
E2e2j
µ1missT
E2e3j
µ1missT
E2e1b
µ1missT
Eµ
2missT
E1j
µ2missT
E2j
µ2missT
E3j
µ2missT
E4j
µ2missT
E5j
µ2missT
E1b
µ2missT
E1b1j
µ2missT
E1b2j
µ2missT
E1b3j
µ2missT
E1b4j
µ2missT
E1b5j
µ2missT
E2b
µ2missT
E2b1j
µ2missT
E2b2j
µ2missT
E2b3j
µ2missT
E2b4j
µ2missT
E3b
µ2missT
E3b1j
µ2missT
E3b2j
µ2missT
E1e
µ2missT
E1e1j
µ2missT
E1e2j
µ2missT
E1e3j
µ2missT
E1j
µ3missT
E2j
µ3missT
E1b1j
µ3missT
EEvents / class1
10210Data 20153-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbs + multilepton
TmissE
Figure 22: The number of events in data, and for the different SM background predictions considered, for classes
with large Emiss
T, at least one pair of same flavour leptons and ( b-)jets (no photons). The classes are labelled
accordingtothemultiplicityandtype( e,,
,j,b,Emiss
T)ofthereconstructedobjectsforthegiveneventclass. The
hatched bands indicate the total uncertainty of the SM prediction.
2j
3j
4j
5j
6j
7j
8j
9j
10j
11j
12j
13j
14j
1b1j
1b2j
1b3j
1b4j
1b5j
1b6j
1b7j
1b8j
1b9j
1b10j
1b11j
2b
2b1j
2b2j
2b3j
2b4j
2b5j
2b6j
2b7j
2b8j
2b9j
2b10j
3b
3b1j
3b2j
3b3j
3b4j
3b5j
3b6j
3b7j
3b8j
3b9j
4b
4b1j
4b2j
4b3j
4b4j
4b5j
4b6j
5b
5b1j
5b2j
5b3j
5b4j
5b7j
6b
6b1j
6b2j
6b3j
7b
8b1j
Events / class1
−10110210310410510610710810910Data 2015
3-/4-top Higgs +Z/W/WW ttγ +ttsingle top di-/triboson
) γ(γZ/W++jets tt)+jets γ(γZ/W+jets multijets ATLAS-1
= 13 TeV, 3.2 fbsdijet and multijet (normalized to data)
Figure 23: The number of events in data, and for the different SM background predictions considered, for classes
with( b-)jets(no Emiss
T,leptonsorphotons). Theclassesarelabelledaccordingtothemultiplicityandtype( e,,
,
j,b,Emiss
T)ofthereconstructedobjectsforthegiveneventclass. Ineventclasseswithfourormoredataevents,the
multijet MC sample is scaled to data. The hatched bands indicate the total uncertainty of the SM prediction.
18th July 2018 – 12:25 41

ATLAS DRAFT
References831
[1] D0 Collaboration, Search for new physics in eXdata at DØ using SLEUTH: A 832
quasi-model-independent search strategy for new physics , Phys. Rev. D 62(2000) 092004, 833
arXiv: hep-ex/0006011 . 834
[2] D0 Collaboration, 835
Quasi-model-independent search for new physics at large transverse momentum , 836
Phys. Rev. D 64(2001) 012004, arXiv: hep-ex/0011067 . 837
[3] D0 Collaboration, Quasi-Model-Independent Search for New High pTPhysics at DO , 838
Phys. Rev. Lett. 86(2001) 3712, arXiv: hep-ex/0011071 . 839
[4] D0 Collaboration, 840
Model independent search for new phenomena in p¯pcollisions atps=1:96TeV, 841
Phys. Rev. D 85(2012) 092015, arXiv: 1108.5362 [hep-ex] . 842
[5] H1 Collaboration, A general search for new phenomena in epscattering at HERA , 843
Phys. Lett. B 602(2004) 14, arXiv: hep-ex/0408044 . 844
[6] H1 Collaboration, A general search for new phenomena at HERA , Phys. Lett. B 674(2009) 257, 845
arXiv: 0901.0507 [hep-ex] . 846
[7] CDF Collaboration, 847
Model-independent and quasi-model-independent search for new physics at CDF , 848
Phys. Rev. D 78(2008) 012002, arXiv: 0712.1311 [hep-ex] . 849
[8] CDF Collaboration, Global search for new physics with 2.0 fb1at CDF, 850
Phys. Rev. D 79(2009) 011101, arXiv: 0809.3781 [hep-ex] . 851
[9] S. Brandt, C. Peyrou, R. Sosnowski and A. Wroblewski, 852
The principal axis of jets – an attempt to analyse high-energy collisions as two-body processes , 853
Phys. Lett. 12(1964) 57. 854
[10] E. Farhi, Quantum Chromodynamics Test for Jets , Phys. Rev. Lett. 39(1977) 1587. 855
[11] ATLAS Collaboration, The ATLAS Experiment at the CERN Large Hadron Collider , 856
JINST3(2008) S08003. 857
[12] ATLAS Collaboration, ATLAS Insertable B-Layer Technical Design Report , (2010), 858
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/1291633 , 859
ATLAS Insertable B-Layer Technical Design Report Addendum , ATLAS-TDR-19-ADD-1, 2012, 860
URL:https://cds.cern.ch/record/1451888 . 861
[13] ATLAS Collaboration, Performance of the ATLAS Trigger System in 2015 , 862
Eur. Phys. J. C 77(2017) 317, arXiv: 1611.09661 [hep-ex] . 863
[14] ATLASCollaboration, VertexReconstructionPerformanceoftheATLASDetectoratps=13TeV, 864
ATL-PHYS-PUB-2015-026, 2015, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2037717 . 865
[15] ATLAS Collaboration, The ATLAS Simulation Infrastructure , Eur. Phys. J. C 70(2010) 823, 866
arXiv: 1005.4568 [physics.ins-det] . 867
[16] S. Agostinelli et al., G/e.sc/a.sc/n.sc/t.sc4a simulation toolkit , Nucl. Instrum. Meth. A 506(2003) 250. 868
[17] T. Sjöstrand, S. Mrenna and P. Z. Skands, A brief introduction to PYTHIA 8.1 , 869
Comput. Phys. Commun. 178(2008) 852, arXiv: 0710.3820 [hep-ph] . 870
18th July 2018 – 12:25 42

ATLAS DRAFT
[18] A. Martin, W. Stirling, R. Thorne and G. Watt, Parton distributions for the LHC , 871
Eur. Phys. J. C 63(2009) 189, arXiv: 0901.0002 [hep-ph] . 872
[19] ATLAS Collaboration, Summary of ATLAS Pythia 8 tunes , ATL-PHYS-PUB-2012-003, 2012, 873
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/1474107 . 874
[20] T. Gleisberg et al., Event generation with SHERPA 1.1 , JHEP02(2009) 007, 875
arXiv: 0811.4622 [hep-ph] . 876
[21] D. J. Lange, The EvtGen particle decay simulation package , 877
Nucl. Instrum. Meth. A 462(2001) 152. 878
[22] D. London and J. L. Rosner, Extra gauge bosons in E6, Phys. Rev. D 34(1986) 1530. 879
[23] P. Langacker, The physics of heavy Z0gauge bosons , Rev. Mod. Phys. 81(2009) 1199, 880
arXiv: 0801.1345 [hep-ph] . 881
[24] R. D. Ball et al., Parton distributions with LHC data , Nucl. Phys. B 867(2013) 244, 882
arXiv: 1207.1303 [hep-ph] . 883
[25] ATLAS Collaboration, ATLAS Pythia 8 tunes to 7TeV data, ATL-PHYS-PUB-2014-021, 2014, 884
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/1966419 . 885
[26] Yu. A. Golfand and E. P. Likhtman,886
Extension of the Algebra of Poincare Group Generators and Violation of p Invariance , 887
JETP Lett. 13(1971) 323, [Pisma Zh. Eksp. Teor. Fiz. 13(1971) 452]. 888
[27] D.V.VolkovandV.P.Akulov, IstheNeutrinoaGoldstoneParticle? ,Phys.Lett.B 46(1973)109. 889
[28] J. Wess and B. Zumino, Supergauge Transformations in Four-Dimensions , 890
Nucl. Phys. B 70(1974) 39. 891
[29] J. Wess and B. Zumino, Supergauge Invariant Extension of Quantum Electrodynamics , 892
Nucl. Phys. B 78(1974) 1. 893
[30] S. Ferrara and B. Zumino, Supergauge Invariant Yang-Mills Theories , 894
Nucl. Phys. B 79(1974) 413. 895
[31] A. Salam and J. Strathdee, Super-symmetry and non-Abelian gauges , Phys. Lett. B 51(1974) 353. 896
[32] J. Alwall et al., The automated computation of tree-level and next-to-leading order differential 897
cross sections, and their matching to parton shower simulations , JHEP07(2014) 079, 898
arXiv: 1405.0301 [hep-ph] . 899
[33] J. Pumplin et al.,900
New Generation of Parton Distributions with Uncertainties from Global QCD Analysis , 901
JHEP07(2002) 012, arXiv: hep-ph/0201195 . 902
[34] L. Lonnblad and S. Prestel, Matching tree-level matrix elements with interleaved showers , 903
JHEP03(2012) 019, arXiv: 1109.4829 [hep-ph] . 904
[35] W. Beenakker, M. Kramer, T. Plehn, M. Spira and P. M. Zerwas,905
Stop production at hadron colliders , Nucl. Phys. B 515(1998) 3, arXiv: hep-ph/9710451 . 906
[36] W. Beenakker et al., Supersymmetric top and bottom squark production at hadron colliders , 907
JHEP08(2010) 098, arXiv: 1006.4771 [hep-ph] . 908
[37] W. Beenakker et al., Squark and gluino hadroproduction , Int. J. Mod. Phys. A 26(2011) 2637, 909
arXiv: 1105.1110 [hep-ph] . 910
18th July 2018 – 12:25 43

ATLAS DRAFT
[38] M. Cacciari, G. P. Salam and G. Soyez, The anti- ktjet clustering algorithm , JHEP04(2008) 063, 911
arXiv: 0802.1189 [hep-ph] . 912
[39] ATLAS Collaboration, Performance of b-jet identification in the ATLAS experiment , 913
JINST11(2016) P04008, arXiv: 1512.01094 [hep-ex] . 914
[40] ATLAS Collaboration, Expected performance of the ATLAS b-tagging algorithms in Run-2 , 915
ATL-PHYS-PUB-2015-022, 2015, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2037697 . 916
[41] ATLAS Collaboration, Performance of missing transverse momentum reconstruction with the 917
ATLAS detector using proton–proton collisions atps=13TeV, (2018), 918
arXiv: 1802.08168 [hep-ex] . 919
[42] ATLAS Collaboration, Jet energy scale measurements and their systematic uncertainties in 920
proton–proton collisions atps=13TeV with the ATLAS detector , 921
Phys. Rev. D 96(2017) 072002, arXiv: 1703.09665 [hep-ex] . 922
[43] ATLASCollaboration, MuonreconstructionperformanceoftheATLASdetectorinproton–proton 923
collision data atps=13TeV, Eur. Phys. J. C 76(2016) 292, arXiv: 1603.05598 [hep-ex] . 924
[44] ATLAS Collaboration, Electron efficiency measurements with the ATLAS detector using the 2015 925
LHC proton–proton collision data , ATLAS-CONF-2016-024, 2016, 926
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/2157687 . 927
[45] ATLAS Collaboration, Electron and photon energy calibration with the ATLAS detector using 928
data collected in 2015 atps=13TeV, ATL-PHYS-PUB-2016-015, 2016, 929
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/2203514 . 930
[46] ATLAS Collaboration, Photon identification in 2015 ATLAS data , ATL-PHYS-PUB-2016-014, 931
2016, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2203125 . 932
[47] ATLAS Collaboration,933
Luminosity determination in ppcollisions atps=8TeV using the ATLAS detector at the LHC , 934
Eur. Phys. J. C 76(2016) 653, arXiv: 1608.03953 [hep-ex] . 935
[48] ATLAS Collaboration, Search for supersymmetry in a final state containing two photons and 936
missingtransversemomentuminps=13TeVppcollisionsattheLHCusingtheATLASdetector , 937
Eur. Phys. J. C 76(2016) 517, arXiv: 1606.09150 [hep-ex] . 938
[49] ATLAS Collaboration,939
Search for resonances in diphoton events atps=13TeV with the ATLAS detector , 940
JHEP09(2016) 001, arXiv: 1606.03833 [hep-ex] . 941
[50] ATLAS Collaboration, Search for new phenomena in high-mass diphoton final states using 942
37fb1of proton–proton collisions collected atps=13TeV with the ATLAS detector , 943
Phys. Lett. B 775(2017) 105, arXiv: 1707.04147 [hep-ex] . 944
[51] ATLAS Collaboration, Expected sensitivity studies for gluino and squark searches using the early 945
LHC 13TeV Run-2 dataset with the ATLAS experiment , ATL-PHYS-PUB-2015-005, 2015, 946
/u.sc/r.sc/l.sc:https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PUBNOTES/ATL-PHYS-PUB- 947
2015-005 . 948
[52] ATLAS Collaboration, ATLAS Computing Acknowledgements , ATL-GEN-PUB-2016-002, 949
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/2202407 . 950
[53] G. Corcella et al., HERWIG 6: an event generator for hadron emission reactions with interfering 951
gluons (including supersymmetric processes) , JHEP01(2001) 010, arXiv: hep-ph/0011363 . 952
18th July 2018 – 12:25 44

ATLAS DRAFT
[54] H.-L. Lai et al., New parton distributions for collider physics , Phys. Rev. D 82(2010) 074024, 953
arXiv: 1007.2241 [hep-ph] . 954
[55] S. Schumann and F. Krauss,955
A parton shower algorithm based on Catani-Seymour dipole factorisation , JHEP03(2008) 038, 956
arXiv: 0709.1027 [hep-ph] . 957
[56] ATLAS Collaboration, Monte Carlo Generators for the Production of a WorZ
Boson in 958
Association with Jets at ATLAS in Run 2 , ATL-PHYS-PUB-2016-003, 2016, 959
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/2120133 . 960
[57] T. Gleisberg and S. Höche, Comix, a new matrix element generator , JHEP12(2008) 039, 961
arXiv: 0808.3674 [hep-ph] . 962
[58] F. Cascioli, P. Maierhöfer and S. Pozzorini, Scattering Amplitudes with Open Loops , 963
Phys. Rev. Lett. 108(2012) 111601, arXiv: 1111.5206 [hep-ph] . 964
[59] S. Höche, F. Krauss, M. Schönherr and F. Siegert,965
QCD matrix elements + parton showers. The NLO case , JHEP04(2013) 027, 966
arXiv: 1207.5030 [hep-ph] . 967
[60] R. Gavin, Y. Li, F. Petriello and S. Quackenbush,968
FEWZ 2.0: A code for hadronic Zproduction at next-to-next-to-leading order , 969
Comput. Phys. Commun. 182(2011) 2388, arXiv: 1011.3540 [hep-ph] . 970
[61] R. Gavin, Y. Li, F. Petriello and S. Quackenbush, W physics at the LHC with FEWZ 2.1 , 971
Comput. Phys. Commun. 184(2013) 208, arXiv: 1201.5896 [hep-ph] . 972
[62] J. M. Campbell and R. K. Ellis, Update on vector boson pair production at hadron colliders , 973
Phys. Rev. D 60(1999) 113006, arXiv: hep-ph/9905386 . 974
[63] J. M. Campbell, R. K. Ellis and C. Williams, Vector boson pair production at the LHC , 975
JHEP07(2011) 018, arXiv: 1105.0020 [hep-ph] . 976
[64] S. Alioli, P. Nason, C. Oleari and E. Re, A general framework for implementing NLO calculations 977
in shower Monte Carlo programs: the POWHEG BOX , JHEP10(2010) 043, 978
arXiv: 1002.2581 [hep-ph] . 979
[65] ATLAS Collaboration,980
Simulation of top-quark production for the ATLAS experiment atps=13TeV, 981
ATL-PHYS-PUB-2016-004, 2016, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2120417 . 982
[66] P. Artoisenet, R. Frederix, O. Mattelaer and R. Rietkerk,983
Automatic spin-entangled decays of heavy resonances in Monte Carlo simulations , 984
JHEP03(2013) 015, arXiv: 1212.3460 [hep-ph] . 985
[67] T. Sjöstrand, S. Mrenna and P. Skands, PYTHIA 6.4 physics and manual , JHEP05(2006) 026, 986
arXiv: hep-ph/0603175 . 987
[68] P. Z. Skands, Tuning Monte Carlo generators: The Perugia tunes , Phys. Rev. D 82(2010) 074018, 988
arXiv: 1005.3457 [hep-ph] . 989
[69] N. Kidonakis,990
Two-loopsoftanomalousdimensionsforsingletopquarkassociatedproductionwitha WorH, 991
Phys. Rev. D 82(2010) 054018, arXiv: 1005.4451 [hep-ph] . 992
18th July 2018 – 12:25 45

ATLAS DRAFT
[70] N. Kidonakis,993
Next-to-next-to-leading logarithm resummation for s-channel single top quark production , 994
Phys. Rev. D 81(2010) 054028, arXiv: 1001.5034 [hep-ph] . 995
[71] N. Kidonakis, Next-to-next-to-leading-order collinear and soft gluon corrections for t-channel 996
single top quark production , Phys. Rev. D 83(2011) 091503, arXiv: 1103.2792 [hep-ph] . 997
[72] M. Czakon and A. Mitov,998
Top++: A program for the calculation of the top-pair cross-section at hadron colliders , 999
Comput. Phys. Commun. 185(2014) 2930, arXiv: 1112.5675 [hep-ph] . 1000
[73] M. Czakon, D. Heymes and A. Mitov, 1001
High-Precision Differential Predictions for Top-Quark Pairs at the LHC , 1002
Phys. Rev. Lett. 116(2016) 082003, arXiv: 1511.00549 [hep-ph] . 1003
[74] M. Czakon, D. Heymes and A. Mitov, 1004
Dynamical scales for multi-TeV top-pair production at the LHC , JHEP04(2017) 071, 1005
arXiv: 1606.03350 [hep-ph] . 1006
[75] ATLAS Collaboration, Multi-boson simulation for 13TeV ATLAS analyses , 1007
ATL-PHYS-PUB-2016-002, 2016, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2119986 . 1008
[76] ATLAS Collaboration, 1009
Modelling of the t¯tHandt¯tV¹V=W;Zșprocesses forps=13TeV ATLAS analyses , 1010
ATL-PHYS-PUB-2016-005, 2016, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2120826 . 1011
[77] ATLAS Collaboration, Measurement of the Z
boson transverse momentum distribution in pp 1012
collisions atps=7TeV with the ATLAS detector , JHEP09(2014) 145, 1013
arXiv: 1406.3660 [hep-ex] . 1014
[78] M. Bahr et al., Herwig++ Physics and Manual , Eur. Phys. J. C 58(2008) 639, 1015
arXiv: 0803.0883 [hep-ph] . 1016
[79] LHC Higgs Cross Section Working Group, 1017
Handbook of LHC Higgs Cross Sections: 2. Differential Distributions , CERN-2012-002, 1018
arXiv: 1201.3084 [hep-ph] . 1019
[80] S. Frixione, E. Laenen, P. Motylinski, B. R. Webber and C. D. White, 1020
Single-top hadroproduction in association with a Wboson, JHEP07(2008) 029, 1021
arXiv: 0805.3067 [hep-ph] . 1022
[81] ATLAS Collaboration, Measurements of W
andZ
production in ppcollisions atps=7TeV 1023
with the ATLAS detector at the LHC , Phys. Rev. D 87(2013) 112003, 1024
arXiv: 1302.1283 [hep-ex] , Erratum: Phys. Rev. D 91(2015) 119901. 1025
[82] ATLAS Collaboration, Search for supersymmetry atps=13TeV in final states with jets and two 1026
same-sign leptons or three leptons with the ATLAS detector , Eur. Phys. J. C 76(2016) 259, 1027
arXiv: 1602.09058 [hep-ex] . 1028
[83] ATLAS Collaboration, Electron efficiency measurements with the ATLAS detector using 2012 1029
LHC proton–proton collision data , Eur. Phys. J. C 77(2017) 195, arXiv: 1612.01456 [hep-ex] . 1030
[84] ATLAS Collaboration, Measurement of the photon identification efficiencies with the ATLAS 1031
detector using LHC Run-1 data , (2016), arXiv: 1606.01813 [hep-ex] . 1032
[85] M. Cacciari, G. P. Salam and G. Soyez, FastJet user manual , Eur. Phys. J. C 72(2012) 1896, 1033
arXiv: 1111.6097 [hep-ph] . 1034
18th July 2018 – 12:25 46

ATLAS DRAFT
[86] ATLAS Collaboration, 1035
Topological cell clustering in the ATLAS calorimeters and its performance in LHC Run 1 , 1036
Eur. Phys. J. C 77(2017) 490, arXiv: 1603.02934 [hep-ex] . 1037
[87] ATLAS Collaboration, Monte Carlo Calibration and Combination of In-situ Measurements of Jet 1038
Energy Scale, Jet Energy Resolution and Jet Mass in ATLAS , ATLAS-CONF-2015-037, 2015, 1039
/u.sc/r.sc/l.sc:https://cds.cern.ch/record/2044941 . 1040
[88] ATLAS Collaboration, Performance of pile-up mitigation techniques for jets in ppcollisions at 1041ps=8TeV using the ATLAS detector , Eur. Phys. J. C 76(2016) 581, 1042
arXiv: 1510.03823 [hep-ex] . 1043
[89] ATLAS Collaboration, 1044
Selection of jets produced in 13TeV proton–proton collisions with the ATLAS detector , 1045
ATLAS-CONF-2015-029, 2015, /u.sc/r.sc/l.sc:https://cds.cern.ch/record/2037702 . 1046
18th July 2018 – 12:25 47

ATLAS DRAFT
Auxiliary material 1047
)
min
(p
10
log
–
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
7
−
10
6
−
10
5
−
10
4
−
10
3
−
10
2
−
10
1
−
10
1
ATLAS
Simulation
-1
= 13 TeV, 3.2 fb
s
inv
variable: m
σ
5
SM + Z' (SSM)
SM-only
= 2.5 TeV
Z'
mPexp,1 < 5%
(a)
)
min
(p
10
log
–
0
1
2
3
4
5
6
7
8
Fraction of pseudo-experiments
7
−
10
6
−
10
5
−
10
4
−
10
3
−
10
2
−
10
1
−
10
1
ATLAS
Simulation
-1
= 13 TeV, 3.2 fb
s
eff
variable: m
1
0
χ
+
t
t
→
g
~
SM + Pexp,1 < 5%
SM-only
= 1.0 TeV
g
~
m
σ
5 (b)
Figure 24: The fraction of cases in which the first pseudo-experiment has Pexp;1<0:05and triggers a second
pseudo-experimentwhichyieldsavaluefor log10¹p0șsmallerthanagivenvalue( log10¹pminș). Thedistribution
isshownforpseudo-experimentsgeneratedfromtheSMexpectation,andfromanSM-plus-signalexpectationof(a)
inclusivedecaysofa Z0bosonwithmass mZ0=2:5TeVor(b)gluinopairswithmass m˜g=1:0TeVand g!t¯t˜0
1
decays. Thesignalregiontestedinthefollow-uppseudo-experimentisdefinedbytheprecedingpseudo-experiment.
The5threshold is indicated at the top of the figure.
18th July 2018 – 12:25 48