Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

ATLAS NOTE

ATLAS-CONF-2016-030
14th June 2016

Search for light dijet resonances with the ATLAS detector


√ using a
Trigger-object Level Analysis in LHC p p collisions at s = 13 TeV

The ATLAS Collaboration

Abstract
ATLAS-CONF-2016-030

Searches for dijet resonances with sub-TeV masses using the ATLAS detector are statistic-
ally limited by the bandwidth available to inclusive single-jet triggers. Due to large Stand-
ard Model multijet backgrounds, these triggers must be prescaled to record full events at a
manageable rate. However, one can avoid this limitation by recording only the subset of
information in each event needed for such a search, allowing much higher trigger rates. This
15 June 2016

note describes a search for new physics in the dijet final state using this strategy, targeting
low-mass dijet resonances between 450 GeV and 950 GeV. The analysed dataset has an in-
tegrated luminosity of 3.4 fb−1 and was recorded at a center-of-mass energy of 13 TeV. No
excesses are found, and limits are set on Dark Matter mediators with axial vector couplings
to quarks and Dark Matter particles, and on generic Gaussian resonances.

© 2016 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.
1 Introduction

Searches for new particles decaying to pairs of jets have historically been among the first analyses per-
formed at new collision energies [1–16]. Such searches usually focus on the production of heavy particles
inaccessible at lower collision energies, as in the case of recent results from ATLAS and CMS focusing
on the mass region above 1 TeV [15, 16]. Nevertheless, the region below 1 TeV remains an important
target of searches for resonances with lower masses and lower cross-sections, and is challenging to access
at the LHC. Despite the unprecedented number of hadron collisions recorded at the LHC, earlier SPS and
Tevatron searches are more sensitive to light dijet resonances in this regime than previous ATLAS and
CMS searches.
The bandwidth allocated to inclusive single-jet triggers limits the statistical power of dijet resonance
searches for particles lighter than approximately 1 TeV at ATLAS, due to the large Standard Model
(SM) multijet rate. If full event information is to be recorded, single-jet triggers must be prescaled, a
procedure wherein only a fraction of events corresponding to the inverse of the prescale fraction and
passing the trigger requirements are recorded. This translates into a large fraction of discarded events
with a jet pT below roughly 400 GeV. This limitation can be avoided by recording only a summary of
the jet information needed for performing a resonance search in the dijet mass (m j j ) spectrum. These
partially-built events are used for the search in this note, performed at ATLAS for the first time. This note
compares the performance of trigger-level jet reconstruction to the performance of the reconstruction
using full detector readout performed after the trigger decision, and describes the search for narrow,
light dijet resonances using this strategy, called Trigger-object Level Analysis (TLA) in ATLAS and Data
Scouting in CMS [17, 18]. A complementary strategy is described in [19]: there, the threshold limitations
are avoided by triggering on a high-pT photon balanced against the light dijet resonance of interest, which
is boosted in the transverse direction. Another complementary search for resonances with masses of
600 GeV and above, decaying to heavy-flavour jets, is described in Ref. [20].
The analysis follows the standard resonance search approach [15]. Section 2 describes the ATLAS ex-
periment and its jet trigger system, together with the TLA technique. The search uses 3.4 fb−1 of 13 TeV
collision data recorded in 2015 with the TLA stream. Though some calibration is applied online to trigger
jets, additional corrections are needed to bring them closer to the performance of jets from full events that
have been calibrated offline. We apply event- and jet-level cleaning criteria and jet calibrations derived
with the methods employed offline, insofar as possible with the partial event data. These techniques are
described in Section 3. Additional jet performance studies were conducted with full events collected in
parallel, to validate the trigger-level jet calibration and cleaning. The event selection is described in Sec-
tion 4. After requiring at least two high-pT jets, we search for excesses in the mass distribution of the
two leading jets. The background estimate for the search is obtained from fits of smooth shapes to the
data, described in Section 5. Since no significant excesses are identified in the "search phase" detailed in
Section 6 and comprising a background-only parameterisation and statistical assessment, Bayesian limits
are set on benchmark signals, and the results are shown in Section 7.

2
2 ATLAS detector and trigger system

2.1 ATLAS detector

The ATLAS experiment [21] at the LHC is a multi-purpose particle detector with a forward-backward
symmetric cylindrical geometry with layers of tracking, calorimeter, and muon detectors over nearly the
entire solid angle around the collision point 1. The directions and energies of high-pT hadronic jets are
measured using silicon tracking detectors and straw tubes detecting transition radiation, finely segmented
hadronic and electromagnetic calorimeters, and a muon spectrometer. A steel/scintillator-tile calorimeter
provides hadronic energy measurements for the pseudorapidity range |η| < 1.7. A lead/liquid-argon
(LAr) calorimeter provides electromagnetic (EM) energy measurements with higher granularity within
the region |η| < 3.2. The end-cap and forward regions are instrumented with LAr calorimeters for EM
and hadronic energy measurements up to |η| = 4.9.

2.2 Jet triggers

A first-level (L1) trigger in ATLAS is implemented in hardware and uses a subset of the detector inform-
ation to record events at a rate of 100 kHz. The L1 trigger identifies jet regions of interest (jet ROIs) from
∆η × ∆φ = 0.2 × 0.2 calorimeter segments using a sliding window algorithm. Events selected by the
hardware trigger are processed by a software-based high-level trigger (HLT). The HLT reconstructs and
calibrates jets (trigger jets) with similar techniques and inputs as in the jet reconstruction performed offline
on fully recorded events [22]. Groups of contiguous calorimeter cells (topological clusters) are formed
based on the significance of the energy deposit over calorimeter noise [23]. These are then clustered into
jets using the anti-k t algorithm [24, 25] with distance parameter R = 0.4. Jet four-momenta are computed
summing over the massless four-momenta of the topological clusters that comprise the jet, pointing to the
center of the ATLAS detector. Using these trigger jets and other reconstructed physics quantities, the HLT
then selects full events to be recorded at a total rate across all triggers of 1 kHz for offline reconstruction
and analysis.

2.3 Trigger-object Level Analysis in ATLAS

Partially-built events are collected by means of an additional TLA data stream. In this stream, all events
containing at least one L1 jet ROI with ET > 75 GeV at the electromagnetic scale 2 are recorded. For
each trigger jet recorded at the HLT with pT > 4 GeV, the stream records the jet four-momentum and
a set of variables characterizing the detector information within the jet (such as variables needed for
jet identification [26], and limited information about the structure of the jet). The data format does not
include any readout of individual calorimeter elements, nor does it record information from the tracking
or muon detectors. The size of partially-built TLA events in 2015 data is less than 5% of the size of full
1 ATLAS uses a right-handed coordinate system with its origin at the nominal interaction point (IP) in the centre of the detector
and the z-axis along the beam pipe. The x-axis points from the IP to the centre of the LHC ring, and the y-axis points
upwards. Cylindrical coordinates (r, φ) are used in the transverse plane, φ being the azimuthal angle around the z-axis. The
pseudorapidity is defined in terms of the polar angle θ asqη = − ln tan(θ/2). It is equivalent to the rapidity, y, for massless
particles. Angular distance is measured in units of ∆R ≡ (∆η) 2 + (∆φ) 2 .
2 The electromagnetic scale correctly measures the energy deposited by electromagnetic showers in the calorimeter, and needs
to be corrected to the jet energy scale.

3
events, allowing for higher event rates to be recorded. TLA events are recorded with a rate of 2 kHz, in
addition to the fully-recorded higher pT events.

3 Trigger jet calibration and performance

Offline jet four-momenta are calibrated for their response to incident hadrons using the procedures de-
scribed in Refs. [27, 28]. A dedicated calibration procedure is used for trigger jets in partially built events,
closely matching the calibration scheme used offline. Pile-up is corrected for using an event-by-event jet
area-based calibration as in Ref. [29] based on HLT calorimeter quantities; additional pile-up residual
corrections requiring information from the inner detector are not applied. A Monte-Carlo based calib-
ration derived specifically for trigger jets restores the jet energy scale to that of jets built from stable
particles with a lifetime longer than 10 ps, excluding muons and neutrinos. The Global Sequential Calib-
ration corrections [28] present in the offline calibration scheme are not applied in the case of trigger jets,
as they require information from the inner detector and the muon spectrometers. The agreement in the
response of trigger with respect to offline jets is improved applying a dedicated correction to trigger jets.
This correction is derived from data, in bins of jet η and pT , and is based on the inverse of the response
ratio of trigger to offline jets matched within ∆R < 0.4. This additional calibration step also allows for a
cross-check of the jet energy resolution modelling.
After this correction, the energy scale of trigger jets is equivalent to that of offline jets. The same jet-wise
calibration constants obtained for offline jets with in situ techniques [30] are then applied to both trigger
and offline jets, to restore agreement between the data and the simulation used to derive the Monte Carlo
(MC) calibration.
The dijet mass and pT response for trigger jets with respect to offline jets is shown in Figures 1 and 2.
Events for the dijet invariant mass response have at least one trigger jet of ET > 110 GeV 3, selected using
the HLT_j110 single-jet trigger. In order to avoid trigger biases in the jet pT response at the transverse
momenta relevant for this search, events are selected using the HLT_j60 trigger in order to have at least
one trigger jet with ET > 60 GeV. In both distributions, the triggers chosen for the event selection are
fully efficient with the specified pT cuts. In these events, y ∗ = (y1 − y2 )/2 is defined as half the difference
in rapidities of the two jets and its absolute value is required to be less than 0.6.
After the trigger jet calibration procedure, the jet energy scale uncertainty for trigger jets is estimated
starting from the offline jet energy scale uncertainty, with the same components as in [31]. The trigger jet
energy scale uncertainty includes also the full trigger-to-offline correction, and an additional uncertainty
of less than 1% as a function of the number of primary vertices to account for differences in the pile-up
subtraction method between trigger and offline jets. The total jet energy scale uncertainty for trigger jets
ranges from 3.5% in the central region (|η| <0.8) to a maximum of 5% in the transition region between
barrel and endcaps. The jet energy scale uncertainty for jets with η = 0 is shown in Figure 3. The main
difference between the offline and the trigger jet energy scale is due to the flavor uncertainties 4, which
are significantly reduced for offline jets by the application of the Global Sequential Correction. The jet
energy scale uncertainty is only applied to simulated trigger jets.

3 The ET threshold for selecting jets for full reconstruction at the HLT is specified in the name of the trigger after the HLT_
prefix.
4 A flavor composition of 50 (± 50)% quark- and 50 (± 50)% gluon-initiated jets uncertainty is used to estimate the flavor
uncertainties, for both offline and trigger jets.

4
1.3 103

/ moffline
jj
1.2

mHLT
jj
1.1
102

0.9
ATLAS Preliminary 10
0.8 Events passing HLT_j110 trigger
subleading jet
pleading jet > 185 GeV, pT > 85 GeV, |y*| < 0.6
T
0.7
600 800 1000 1200 1400
mjj [GeV]

Figure 1: Dijet invariant mass built from leading and subleading trigger (HLT) jets compared to m j j for offline jets
in data, as a function of randomly chosen HLT or offline m j j to reduce resolution biases. Events are selected using
the HLT_j110 single jet trigger. Selected events have leading jet pT > 185 GeV, subleading jet pT > 85 GeV and
|y ∗ | < 0.6. The average m j j response of trigger jets relative to offline jets is shown in black and it is within 1% of
unity, independent of jet m j j .

1.3 103
pHLT / poffline
T

1.2
T

1.1
102

0.9
ATLAS Preliminary 10
0.8 Events passing HLT_j60 trigger
pT > 85 GeV, |η| < 2.8, |y*| < 0.6
0.7
100 150 200 250 300 350 400 450 500
p [GeV]
T

Figure 2: Transverse momentum of trigger (HLT) jets compared to offline jets in data, as a function of randomly
chosen HLT or offline jet pT to reduce resolution biases. Events are selected using the HLT_j60 single jet trigger.
HLT and offline jets are matched using ∆R < 0.4. Selected events have leading jet pT > 185 GeV, subleading jet
pT > 85 GeV and |y ∗ | < 0.6. The average pT response of trigger jets relative to offline jets is shown in black and it
is within 1% of unity, independent of jet pT .

5
0.1

Fractional JES uncertainty


Data 2015, s = 13 TeV ATLAS Preliminary
anti-kt R = 0.4, EM+JES + in situ correction
0.08 η = 0.0
Total Uncertainty, Trigger
Total Uncertainty, Offline
Absolute in situ JES
0.06 Relative in situ JES
Trigger Flav. composition
Trigger Flav. response
Trigger Pileup, average 2015 conditions
0.04 Trigger data-derived correction

0.02

0
100 200 300 400 500 600 700 800 900 1000
p jet
T
[GeV]

Figure 3: Jet energy scale uncertainties for trigger jets in the central region (η=0), compared to the total offline jet
energy scale uncertainties (solid green line). Jet energy scale uncertainty subcomponents are only shown for trigger
jets. Further information on the relative magnitude of the components of the trigger and offline jet energy scale
uncertainties are given in the text.

4 Event selection

Dijet events are selected for this analysis with criteria similar to those used for heavy dijet resonance
searches [15]. Events must contain at least two trigger jets within |η| < 2.8 and with transverse momenta
above 85 GeV. The analysis is performed using two signal regions, one consisting of events with |y ∗ | <
0.6 and one consisting of events with |y ∗ | < 0.3. The leading trigger jet must have pT > 185 GeV, chosen
to ensure that the L1 trigger is fully efficient.
Restrictions to lower |y ∗ | values reduce the background to resonant new physics phenomena from QCD
processes, which contribute mainly at high y ∗ values. The additional signal region with |y ∗ | < 0.3 lowers
the mass range affected by the turn-on of the L1 trigger efficiency for a given pT cut, allowing extension
of the search to m j j values below 550 GeV, since higher pT jets are selected at a given invariant mass with
respect to the wider y ∗ cut.
Events which contain jets induced by minor noise bursts, beam induced background or cosmic rays are
rejected. Events are discarded from the search if either of the two highest-pT (leading) jets, or a third jet
with pT > 50 GeV, are poorly measured, or are compatible with non-collision background or calorimeter
noise. The criteria used are the same as in Ref. [26], except for the track-based charged fraction, which is
not computed for trigger jets. The charged fraction requirement has a negligible effect on events satisfying
the dijet kinematic criteria.
The efficiency and purity of jets passing the above criteria are checked with a tag and probe method in
data containing full events. The trigger jet reconstruction efficiency is 100% for jets with pT > 85 GeV.
The rate of trigger jets not also reconstructed offline is below 0.3%.

6
Events
10
8 TLA jets
Offline jets selected by any single-jet trigger
Offline jets selected by j110 single-jet trigger

6
10

104

102
ATLAS Preliminary
s=13 TeV, 3.4 fb-1
|y*| < 0.6
1
TLA/Offline (j110)

1.5
1
0.5
500 600 700 800 900 1000
mjj [TeV]

Figure 4: Comparison between the data used by the Trigger-object Level Analysis and the data selected by using
an OR of any single jet trigger. The ratio plot compares the shape of the dijet mass distribution of TLA trigger jets
and offline jets collected using the HLT_j110 trigger, normalised to the same integral as the trigger-level spectrum.
The average prescale factor for the HLT_j110 trigger is 2300.

Figure 4 shows the m j j distribution of events built from trigger jets and selected for the TLA compared
to the m j j distribution of fully-recorded events passing any single-jet trigger. The bin width chosen
varies with mass to approximate the m j j resolution for trigger jets. The m j j resolution is derived from
simulation of QCD dijet processes and is shown in Figure 5 for events with |y ∗ | < 0.6.
The m j j distribution from full events passing the event selection and at least one trigger jet with ET >
110 GeV is also given in Figure 4. All triggers selecting full events with a trigger jet ET threshold lower
than 360 GeV apply a prescale factor, meaning that only the fraction of events corresponding to the
inverse of the prescale factor is recorded. The figure highlights the increase in statistics achieved by the
TLA strategy as well as the agreement between the shapes of the m j j distributions of jets reconstructed
in the trigger and jets reconstructed offline.

0.075
σ (m ) / <m >
jj

0.07
ATLAS Simulation Preliminary
0.065
jj

Pythia 8 QCD
0.06 |y*| < 0.6
0.055
0.05
0.045
0.04
0.035
0.03
400 600 800 1000 1200
mjj [GeV]

Figure 5: The ratio of the dijet mass resolution σ m j j to average dijet mass < m j j > as a function of m j j as
determined in Pythia 8 simulation of QCD processes.

7
5 Background estimate

Distributions of the dijet invariant mass, m j j , for events passing the criteria described in Section 4 are
analyzed for evidence of localized excesses. Prior dijet searches at various collision energies [9, 12, 32–
35] have found that a variety of simple functional forms describe the shapes of dijet mass distributions.
The Standard Model backgrounds to the present search are estimated by fitting the binned m j j distribution

to a functional form, dependent on the variable z ≡ m j j / s and on a number of free parameters pi . A
number of functions are evaluated for their capability to describe the mass distributions of high-statistics
simulation of QCD dijet processes using an iterative procedure. A fit of each of the candidate functions to
the mass distribution in data is performed in a mass range initially chosen to avoid kinematic bias at low
masses from the L1 trigger threshold and to extend the search below the range studied by the high-mass
dijet search [15]. If a χ2 p-value larger than 0.05 cannot be obtained for this range, even after excluding
a single contiguous subrange of bins compatible with signal, the lower end of the mass range is increased
one bin at a time until this criterion is satisfied. The procedure stops when at least two functions are
selected for a given starting point in dijet mass.
Among the seven functions, two are selected as viable candidates to model the distributions in data,
starting from 443 GeV and 396 GeV for the |y ∗ | < 0.6 and |y ∗ | < 0.3 selections, respectively. The
function with the largest χ2 probability is found to be the one given by Equation (1):
p1 −p3 z−p4 z 2
f (z) = e , (1)
z p2
and it is selected to provide the background estimate. The function with the second-largest probability,

f (z) = p1 (1 − z) p2 z p3 +p4 log z (2)


is used to define a shape uncertainty due to the choice of the function.
The statistical significance of any localized excess in this distribution is evaluated using the BumpHunter
algorithm [36, 37]. The algorithm operates on the binned m j j distribution, comparing the data with the
fitted background estimate from Equation (1) in all possible mass intervals, up to half the mass range
spanned by the data. For each interval considered, it computes the significance of any excess found. A p-
value is then computed using the ensemble of possible outcomes across all intervals considered, repeating
the algorithm on pseudo-data drawn from the background fit.

6 Search for localized excesses in the dijet mass distribution

Figure 6(a) shows the mass distribution for events with |y ∗ | < 0.6. The function in Equation (1) is fit to
this distribution in the interval 443–1236 GeV with a χ2 p-value of 0.86. The bottom panel of the figure
shows the significances of the bin by bin differences between the data and the fit [38]. These Gaussian
significances are calculated from the Poisson probability. The significance takes statistical uncertainties
but no systematic uncertainties into account. Figure 6(b) shows the mass distribution for events with
|y ∗ | < 0.3. The function in Equation (1) is fit in the interval 396–1236 GeV with a χ2 p-value of 0.80.
The BumpHunter algorithm identifies the interval 574–685 GeV, indicated by the two vertical lines in
the figures, as the single most discrepant interval for both mass distributions (|y ∗ | < 0.3 and |y ∗ | <

8
107 107
Events

Events
ATLAS Preliminary ATLAS Preliminary
s=13 TeV, 3.4 fb-1 s=13 TeV, 3.4 fb-1
Data Data
Background fit Background fit
6 BumpHunter interval 6 BumpHunter interval
10 10

5 5
10 10
p-value = 0.44 p-value = 0.19
Fit Range: 443 - 1236 GeV Fit Range: 394 - 1236 GeV
|y*| < 0.6 |y*| < 0.3
Significance

Significance
2 2
0 0
−2 −2

500 600 700 800 900 1000 400 500 600 700 800 9001000
mjj [GeV] mjj [GeV]
(a) (b)

Figure 6: The reconstructed dijet mass distribution (filled points) for events with pT > 185 (85) GeV for the
leading (subleading) jets and with (a) |y ∗ | < 0.6 and (b) |y ∗ | < 0.3. The solid line depicts the fit to Equation (1),
as discussed in the text. The vertical lines indicate the most discrepant interval identified by the BumpHunter
algorithm, for which the p-value is stated in the figure. The lower panels show the bin-by-bin significances of the
data–fit differences, considering only statistical uncertainties.

0.6). The probability of observing a background fluctuation at least as significant as that observed in
data, anywhere in the distribution, is 0.44 for the mass distribution with |y ∗ | < 0.6 and 0.19 for the
mass distribution with |y ∗ | < 0.3, taking only statistical uncertainties into account, corresponding to 0.15
and 0.88 σ, respectively. Thus, there is no evidence for an anomalous localized contribution in either
selection.

7 Limit setting

Limits are set on the cross-section, σ, times acceptance, A, times branching ratio, BR of a leptophobic
Z’ simplified model [39]. The model’s matrix elements are calculated in MadGraph 5 [40] and parton
showering is performed in Pythia 8. Consistently with the model studied in [15], the Z 0 model assumes
axial-vector couplings to SM quarks and to a Dirac fermion dark matter candidate. No interference with
the SM is simulated for this model. The Z 0 model considered follows a scenario [41] where its decays to
dark matter particles are negligible, hence the dijet production rate and resonance width depend only on
the coupling to quarks, gq , and the mass of the resonance m Z 0 . The acceptance A for a mass of 550 GeV
is 20% for the Z 0 model with gq = 0.15 for the analysis selection with |y ∗ | < 0.3, and 40% for a signal of
mass equal to 650 GeV for the analysis selection with |y ∗ | < 0.6.
Starting from the m j j distribution obtained with the resonance selection, a Bayesian method [9] is applied
to the data and simulation of signals at a series of discrete masses to set 95% credibility-level upper
limits on the cross-section times acceptance for the signals described above. The method uses a constant

9
prior for signal cross-section and Gaussian priors for nuisance parameters corresponding to systematic
uncertainties.
Limits are also provided for a possible signal that produces a Gaussian contribution to the observed
m j j distribution. For sufficiently narrow resonances, these results may be used to set limits on models
of new phenomena beyond those considered explicitly in this result. Such limits should be used when
parton distribution function and nonperturbative effects can be safely truncated or neglected. These limits
are applicable if the m j j distribution for a signal approaches a Gaussian distribution after applying the
kinematic selection criteria of the resonance analysis. Models of new resonances with an intrinsic width
much smaller than 7% should be compared to the results with a width equal to the experimental resolution.
For models with a larger width after detector effects, the limit that best matches their width should be used.
More detailed instructions can be found in Appendix A of Ref. [12].
Systematic uncertainties on the estimate of the background to the search arise from imperfect knowledge
of the shape of the m j j distribution. To account for the statistical uncertainty on the fitted parameters in
Equation (1), a large number of pseudo-experiments are performed, fitting pseudo-data drawn from Pois-
son fluctuations around the nominal background model. The uncertainty on the background prediction
in each m j j bin is taken to be the root mean square of the function value in that bin for all pseudo-
experiments. To account for the choice of the function used in the fit, the alternative parameterization of
Equation (2) is compared to the nominal fit. The mean difference between the two results as calculated
over a range of pseudoexperiments is taken as the uncertainty. Statistical uncertainties and uncertainties
on the jet energy scale of trigger jets are included when setting limits on the signal models considered
while the impact of the uncertainty on the jet energy resolution is negligible. The jet energy scale un-
certainty is treated using a set of nuisance parameters and accounting for the effect of each on the signal
Monte Carlo. For model-independent limits on Gaussian shapes, a flat jet energy scale uncertainty of 3%
is applied in accordance with the measured impact of this uncertainty on the Z’ samples. The uncertainty
on the integrated luminosity is ±5%. It is derived, following a methodology similar to that detailed in
Ref. [42], from a preliminary calibration of the luminosity scale using a pair of x-y beam-separation scans
performed in June 2015.
The expected limits are calculated using pseudo-experiments generated from the maximum-likelihood
values for parameters of the background-only model in Equation (1) and using the full systematic uncer-
tainties in both the signal and background models.
Figure 7 shows limits on the coupling to quarks gq as a function of the mass m Z 0 for the Z’ model.
Coupling values above the solid curves are excluded.
Figure 8, Table 1 and Table 2 show limits on a possible Gaussian contribution to the observed m j j dis-
tribution obtained for a mean mass mG and three different widths; a width equal to the detector mass
resolution as well as equal to 7% and 10% of mG are chosen. Limits for a given signal region are set for
mG values above the starting point of the fit and separated from the endpoints of this range by at least the
width of the Gaussian. As the width increases, the expected signal contribution is distributed across more
bins. Wider signals are thus less affected by statistical fluctuations of the data in a single bin.

10
gq

0.16

0.14 obs. limit


exp. limit

0.12

0.1

0.08
ATLAS Preliminary
0.06 |y*| < 0.3 |y*| < 0.6 s = 13 TeV; 3.4 fb-1

400 500 600 700 800 900 1000


mZ’ [GeV]

Figure 7: The 95% credibility-level observed and expected upper limits for the Z 0 model described in the text. The
limits are obtained from the m j j distribution on the coupling to quarks, gq , as a function of the mass, m Z 0 . Coup-
lings above the solid lines are excluded. Markers indicate the mass and coupling points that have been simulated.
The solid and dashed curves represent the observed and expected limit, respectively. They are obtained from the
simulated points, correctly accounting for the scaling of the signal cross-section with gq2 . Limits for masses includ-
ing and above 550 GeV are derived from the m j j distribution with |y ∗ | < 0.6 while those at and below 550 GeV are
derived using the distribution with |y ∗ | < 0.3.

11
102 102
σ × A × BR [pb]

σ × A × BR [pb]
|y*| < 0.3 s=13 TeV, 3.4 fb-1
|y*| < 0.6 s=13 TeV, 3.4 fb-1
σG/mG = 0.10 σG/mG = 0.10
σG/mG = 0.07 σG/mG = 0.07
σG/mG = Res. σG/mG = Res.
10 10

1 1

ATLAS Preliminary
ATLAS Preliminary
−1 −1
10 0.4 10 0.6 0.6 0.8 0.8 1 1
mG [TeV] mG [TeV]

Figure 8: The 95% credibility-level observed upper limits on cross-section times acceptance times branching ratio
to two jets, σ × A × BR, for a hypothetical signal producing a Gaussian contribution to the observed m j j distribu-
tion. The limits are shown as a function of the mean mass of the Gaussian distribution, mG , for different relative
widths σG /mG . The smallest relative width shown (“Res”) corresponds to a width σG equal to the detector mass
resolution. This relative width is between 4% and 6% for high and low m j j values, respectively. Results are also
shown for larger relative widths of 7% and 10%. For all mass points to which the nominal |y ∗ | < 0.6 distribution is
sensitive, it is selected to define the limit shown. Lower mass points accessible only to the |y ∗ | < 0.3 spectrum are
shown on the left. A range of mass points spanning [425 GeV, 1100 GeV] are thus covered using the combination
of the two spectra. While the y axis is shared between the two selections, the signal acceptance varies, thus the two
sets of limit points relate to two different interpretations of σ × A × BR. As an example, the acceptance of the mass
points from the |y ∗ | < 0.3 signal region ranges from 14% to 16% for the Z’ signal considered, while the acceptance
of the mass points from the |y ∗ | < 0.6 signal region ranges from 32% to 46%.

12
Limits on σ× A × BR
Mass [GeV] Detector resolution width 7% width/mass 10% width/mass
500 4.1 - -
550 2.0 2.5 -
600 4.3 5.4 9.3
650 3.9 5.2 7.2
700 2.5 3.4 4.7
750 1.3 1.9 2.4
800 1.1 1.2 1.5
850 0.73 0.81 0.95
900 0.45 0.58 0.82
950 0.56 0.73 0.92
1000 0.81 1.0 1.4
1050 1.0 1.6 -
1100 1.1 - -

Table 1: Limits on cross section in pb times acceptance times branching ratio for generic Gaussian signals using the
|y ∗ | < 0.6 cut.

Limits on σ× A × BR
Mass [GeV] Detector resolution width 7% width/mass 10% width/mass
425 3.8 4.1 -
450 2.4 2.7 3.5
500 - 1.8 2.2
550 - - 4.4

Table 2: Limits on cross section in pb times acceptance times branching ratio for generic Gaussian signals using the
|y ∗ | < 0.3 cut. Mass points listed are those for which no limit can be set using the |y ∗ | < 0.6 selection.

13
8 Conclusions

The analysis examines dijets events using the Trigger-object Level Analysis technique in 3.4 fb−1 of

proton-proton collisions with a centre-of-mass energy of s = 13 TeV at the LHC, which was performed
in ATLAS for the first time. The dijet invariant mass distribution presents no significant local excesses
atop a data-driven estimate of the smoothly-falling distribution predicted by the Standard Model. This
analysis excludes at 95% credibility-level Z’ signals and cross-sections for new processes that would
produce a Gaussian contribution to the dijet mass distribution. Gaussian contributions with effective
cross-sections ranging from approximately 3 pb at 450 GeV, to 9 pb at 600 GeV, to 0.7 pb at 850 GeV are
excluded.

References

[1] UA1 Collaboration, G. Arnison et al., Angular distributions and structure functions from two jet
events at the CERN SPS p anti-p collider, Phys. Lett. B136 (1984) 294.
[2] UA1 Collaboration, C. Albajar et al.,
Two jet mass distributions at the CERN Proton–Anti-Proton Collider,
Phys. Lett. B209 (1988) 127.
[3] UA2 Collaboration, P. Bagnaia et al.,
Measurement of jet production properties at the CERN pp Collider, Phys. Lett. B 144 (1984) 283.
[4] CDF Collaboration, T. Aaltonen et al.,

Search for new particles decaying into dijets in proton-antiproton collisions at s = 1.96 TeV,
Phys. Rev. D 79 (2009) 112002, arXiv: 0812.4036 [hep-ex].
[5] ATLAS Collaboration, Search for new particles in two-jet final states in 7 TeV proton-proton
collisions with the ATLAS detector at the LHC, Phys. Rev. Lett. 105 (2010) 161801,
arXiv: 1008.2461 [hep-ex].
[6] CMS Collaboration, Search for dijet resonances in 7 TeV pp collisions at CMS,
Phys. Rev. Lett. 105 (2010) 211801, arXiv: 1010.0203 [hep-ex].
[7] CMS Collaboration, Measurement of dijet angular distributions and search for quark

compositeness in pp Collisions at s = 7 TeV, Phys. Rev. Lett. 106 (2011) 201804,
arXiv: 1102.2020 [hep-ex].
[8] CMS Collaboration,
Search for resonances in the dijet mass spectrum from 7 TeV pp collisions at CMS,
Phys. Lett. B 704 (2011) 123, arXiv: 1107.4771 [hep-ex].
[9] ATLAS Collaboration, Search for new physics in dijet mass and angular distributions in pp

collisions at s = 7 TeV measured with the ATLAS detector, New J. Phys. 13 (2011) 053044,
arXiv: 1103.3864 [hep-ex].
[10] ATLAS Collaboration, Search for new physics in the dijet mass distribution using 1 fb−1 of pp

collision data at s = 7 TeV collected by the ATLAS detector, Phys. Lett. B708 (2012) 37,
arXiv: 1108.6311 [hep-ex].
[11] ATLAS Collaboration, ATLAS search for new phenomena in dijet mass and angular distributions

using pp collisions at s = 7 TeV, JHEP 01 (2013) 029, arXiv: 1210.1718 [hep-ex].

14
[12] ATLAS Collaboration, Search for new phenomena in the dijet mass distribution using pp

collision data at s = 8 TeV with the ATLAS detector, Phys. Rev. D 91 (2015) 052007,
arXiv: 1407.1376 [hep-ex].
[13] CMS Collaboration,

Search for narrow resonances using the dijet mass spectrum in pp collisions at s = 8 TeV,
Phys. Rev. D 87 (2013) 114015, arXiv: 1302.4794 [hep-ex].
[14] R. M. Harris and K. Kousouris, Searches for dijet resonances at hadron colliders,
Int. J. Mod. Phys. A26 (2011) 5005, arXiv: 1110.5302 [hep-ex].
[15] ATLAS Collaboration, Search for new phenomena in dijet mass and angular distributions from

pp collisions at s = 13 TeV with the ATLAS detector, Phys. Lett. B754 (2016) 302,
arXiv: 1512.01530 [hep-ex].
[16] CMS Collaboration,

Search for narrow resonances decaying to dijets in proton-proton collisions at s = 13 TeV,
Phys. Rev. Lett. 116 (2016) 071801, arXiv: 1512.01224 [hep-ex].
[17] CMS Collaboration,

Search for Resonances Decaying to Dijet Final States at s = 8 TeV with Scouting Data,
CMS-PAS-EXO-14-005, 2015, url: https://cds.cern.ch/record/2063491.

[18] CMS Collaboration, Search for narrow resonances in dijet final states at s=8 TeV with the
novel CMS technique of data scouting, 2016, arXiv: 1604.08907 [hep-ex].
[19] ATLAS Collaboration, Search for new low-mass resonances decaying to jet pairs in association

with a photon in proton-proton collisions at s = 13 TeV with the ATLAS detector,
ATLAS-CONF-2016-029, 2016, url: https:
//atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/CONFNOTES/ATLAS-CONF-2016-029/.
[20] ATLAS Collaboration, Search for resonances below 1.2 TeV from the mass distribution of b-jet

pairs in proton-proton collisions at s=13 TeV with the ATLAS detector, ATL-CONF-2016-031,
2016, url: https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/CONFNOTES/ATLAS-
CONF-2016-031/.
[21] ATLAS Collaboration, The ATLAS Experiment at the CERN Large Hadron Collider,
JINST 3 (2008) S08003.
[22] ATLAS Collaboration, 2015 start-up trigger menu and initial performance assessment of the
ATLAS trigger using Run-2 data, ATL-DAQ-PUB-2016-001, 2016,
url: https://cds.cern.ch/record/2136007.
[23] ATLAS Collaboration,
Topological cell clustering in the ATLAS calorimeters and its performance in LHC Run 1, (2016),
arXiv: 1603.02934 [hep-ex].
[24] M. Cacciari, G. P. Salam and G. Soyez, The anti-k t jet clustering algorithm, JHEP 04 (2008) 063,
arXiv: 0802.1189 [hep-ph].
[25] M. Cacciari and G. Salam, Dispelling the N 3 myth for the k t jet-finder,
Phys. Lett. B 641 (2006) 57, arXiv: hep-ph/0512210.
[26] ATLAS Collaboration,
Selection of jets produced in 13 TeV proton-proton collisions with the ATLAS detector,
ATLAS-CONF-2015-029, 2015, url: https://cds.cern.ch/record/2037702.

15
[27] ATLAS Collaboration, Pile-up subtraction and suppression for jets in ATLAS,
ATLAS-CONF-2013-083, 2013, url: https://cds.cern.ch/record/1570994.
[28] ATLAS Collaboration, Jet global sequential corrections with the ATLAS detector in

proton-proton collisions at s = 8 TeV, ATLAS-CONF-2015-002, 2015,
url: http://cds.cern.ch/record/2001682.
[29] ATLAS Collaboration, Performance of pile-up mitigation techniques for jets in pp collisions at

s = 8 TeV using the ATLAS detector, 2015, arXiv: 1510.03823 [hep-ex].
[30] ATLAS Collaboration, Data-driven determination of the energy scale and resolution of jets

reconstructed in the ATLAS calorimeters using dijet and multijet events at s = 8 TeV,
ATLAS-CONF-2015-017, 2015, url: http://cds.cern.ch/record/2008678.
[31] ATLAS Collaboration, Jet calibration and systematic uncertainties for jets reconstructed in the

ATLAS detector at s = 13 TeV, ATL-PHYS-PUB-2015-015, 2015,
url: http://cds.cern.ch/record/2037613.
[32] CDF Collaboration, T. Aaltonen et al.,

Search for new particles decaying into dijets in proton-antiproton collisions at s = 1.96-TeV,
Phys.Rev. D79 (2009) 112002, arXiv: 0812.4036 [hep-ex].
[33] ATLAS Collaboration, Search for new particles decaying into dijets in proton–proton collisions

at s = 7 TeV with the ATLAS detector, ATLAS-CONF-2010-080, 2010,
url: http://cdsweb.cern.ch/record/1298844.
[34] CMS Collaboration,
Search for Resonances in the Dijet Mass Spectrum from 7 TeV pp Collisions at CMS,
Phys. Lett. B 704 (2011) 123, arXiv: 1107.4771 [hep-ex].
[35] UA2 Collaboration, J. Alitti et al.,
A Measurement of two jet decays of the W and Z bosons at the CERN p̄p collider,
Z.Phys. C49 (1991) 17.
[36] CDF Collaboration, T. Aaltonen et al., Global search for new physics with 2.0 fb−1 at CDF,
Phys. Rev. D79 (2009) 011101, arXiv: 0809.3781 [hep-ex].
[37] G. Choudalakis, On hypothesis testing, trials factor, hypertests and the BumpHunter, 2011,
arXiv: 1101.0390 [physics.data-an].
[38] G. Choudalakis and D. Casadei, Plotting the Differences Between Data and Expectation,
Eur. Phys. J. Plus (2012) 25, arXiv: 1111.2062v3.
[39] D. Abercrombie et al., Dark Matter Benchmark Models for Early LHC Run-2 Searches: Report of
the ATLAS/CMS Dark Matter Forum, 2015, arXiv: 1507.00966 [hep-ex].
[40] J. Alwall et al., The automated computation of tree-level and next-to-leading order differential
cross sections, and their matching to parton shower simulations, JHEP 07 (2014) 079,
arXiv: 1405.0301 [hep-ph].
[41] M. Chala et al., Constraining Dark Sectors with Monojets and Dijets, JHEP 07 (2015) 089,
arXiv: 1503.05916 [hep-ph].
[42] ATLAS Collaboration, Improved luminosity determination in pp collisions at sqrt(s) = 7 TeV
using the ATLAS detector at the LHC, Eur. Phys. J. C73 (2013) 2518,
arXiv: 1302.4393 [hep-ex].

16
Appendix

gSM Acceptance
0.10 0.148
0.15 0.155
0.20 0.145
0.30 0.135

Table 3: Acceptances for the m Z 0 = 0.45 TeVsignal with a cut of |y ∗ | < 0.3.

Mass [TeV] gSM Acceptance


0.55 0.10 0.34
0.55 0.15 0.33
0.55 0.20 0.34
0.55 0.30 0.32
0.65 0.10 0.40
0.65 0.15 0.40
0.65 0.20 0.40
0.65 0.30 0.39
0.65 0.40 0.38
0.75 0.10 0.45
0.75 0.15 0.44
0.75 0.20 0.43
0.75 0.30 0.42
0.75 0.40 0.40
0.85 0.10 0.46
0.85 0.15 0.45
0.85 0.20 0.46
0.85 0.30 0.45
0.95 0.10 0.46
0.95 0.15 0.46
0.95 0.20 0.46
0.95 0.30 0.45

Table 4: Acceptances for all Z’ mass points used in this analysis with a cut of |y ∗ | < 0.6.

17
m j j [GeV] N(obs) N(exp), nominal N(exp), alternate
394–410 5390567 5390128 5388798
410–426 4411654 4409364 4409554
426–443 3833944 3835360 3836149
443–460 3138721 3141048 3141892
460–478 2727544 2726298 2726990
478–496 2236952 2236153 2236566
496–515 1937582 1938115 1938272
515–534 1589876 1592565 1592500
534–554 1378060 1379134 1378909
554–574 1135360 1135567 1135255
574–595 984389 983013 982655
595–617 847477 846157 845799
617–639 697622 696122 695813
639–662 600096 599683 599431
662–685 495318 494785 494610
685–709 426353 426675 426571
709–733 352999 353078 353045
733–758 305134 304844 304872
758–783 252372 253015 253092
783–809 218529 218754 218870
809–836 188321 188480 188625
836–863 156612 156629 156784
863–891 135126 135227 135388
891–919 112157 112759 112910
919–948 97520 97559 97697
948–978 84071 84195 84313
978–1008 70331 70364 70454
1008–1039 61105 60885 60945
1039–1070 51074 51074 51101
1070–1102 44439 44312 44307
1102–1135 38530 38379 38338
1135–1168 32266 32294 32221
1168–1202 28161 28057 27951
1202–1236 23637 23705 23571

Table 5: Data and integrated background counts for the search result, for |y ∗ | < 0.3, for nominal and alternate fit
function.

18
m j j [GeV] N(obs) N(exp), nominal N(exp), alternate
443–460 7540437 7541523 7539877
460–478 6560043 6556391 6556677
478–496 5386035 5386046 5387050
496–515 4675610 4675111 4676209
515–534 3843828 3847001 3847826
534–554 3333458 3335887 3336373
554–574 2747607 2750212 2750347
574–595 2386661 2383569 2383430
595–617 2054198 2054065 2053726
617–639 1692756 1691649 1691216
639–662 1459964 1458728 1458261
662–685 1205615 1204659 1204235
685–709 1039450 1039695 1039336
709–733 861039 861014 860756
733–758 743941 743902 743743
758–783 617642 617805 617747
783–809 534906 534438 534468
809–836 460493 460698 460805
836–863 382721 383003 383164
863–891 330497 330782 330983
891–919 274712 275898 276112
919–948 238675 238755 238974
948–978 205878 206080 206286
978–1008 172126 172239 172413
1008–1039 149573 149038 149174
1039–1070 125153 125015 125101
1070–1102 108738 108453 108486
1102–1135 94153 93914 93888
1135–1168 79014 79007 78922
1168–1202 68625 68620 68474
1202–1236 57748 57955 57753

Table 6: Data and integrated background counts for the search result, for |y ∗ | < 0.6, for nominal and alternate fit
function.

Selection criteria Number of events passing cut


Generated events 20000
At least two jets with pT > 10 GeV 16999
pT,lead > 185 GeV && pT,sublead > 85 GeV 14599
|y ∗ | < 0.6 9055

Table 7: Events passing each of the selection cuts for the |y ∗ | < 0.6 selection, for the Z’ signal point with 750 GeV
mass and gq =0.15.

19

You might also like