HEP seminar Josu Cantero (OSU) Heavy resonances 1 / 36 - - PowerPoint PPT Presentation

Search for heavy resonances decaying into third generation quarks with the ATLAS detector

Josu Cantero (Oklahoma State University) July 15, 2020

HEP seminar

Josu Cantero (OSU) Heavy resonances 1 / 36

Introduction

Theories beyond the Standard Model (SM) involve enhanced symmetries that predict new gauge bosons, usually called W’ or Z’ bosons.

→ Some models favor couplings of these new gauge bosons to third generation quarks. → Good signal/background ratio thanks to b-tagging and top-tagging techniques. → Complement searches using final states with first and second generation quarks.

→ This motivate searches for new heavy resonances: 1) W’→ tb (fully hadronic channel)

• Phys. Lett. B 781 (2018) 327 (pdf)

2) Z’→ bb JHEP 03 (2020) 145 (pdf) 3) Z’→ tt (fully hadronic channel) EXOT-2018-48 (pdf) Outline: → Jet reconstruction and calibration (arXiv:2007.02645) → Jet b-tagging (arXiv:1907.05120) → Jet substructure and top tagging (arXiv:1808.07858) → Analysis results

Josu Cantero (OSU) Heavy resonances 2 / 36

Jet reconstruction and calibration

Jets are reconstructed using the anti-kt algorithm with radius parameters R = 0.4 (small-R) and 1.0 (large-R). For use in jet reconstruction, calorimeter cells are first clustered into three-dimensional, massless, topological clusters using a nearest-neighbour algorithm.

→ An event-by event correction to account for the position of the primary vertex in each event is applied to every topo-cluster.

Jets reconstructed using only calorimeter-based energy information are referred to as EMtopo jets. Hadronic final-state measurements can be improved by making more complete use of the information from both the tracking and calorimeter systems.

→ Particle flow algorithm used. It combines information from the tracker and the calorimeter. Specifically, energy deposited in the calorimeter by charged particles is subtracted from the observed topo-clusters and replaced by the momenta of tracks that are matched to those topo-clusters → this improves energy and angular resolution, reconstruction efficiency, and pile-up stability compared to calorimeter jets. → Jets reconstructed with PFlow objects are referred to as PFLow jets. → Only available for jets with R = 0.4.

Josu Cantero (OSU) Heavy resonances 3 / 36

Jet reconstruction and calibration

Jets need to be calibrated to restore the energy to that of jets reconstructed at particle level. This calibration is applied in different steps:

→ pile-up corrections remove the excess energy due to additional proton–proton interactions. → The absolute JES calibration to correct the jet so that it agrees in energy and direction with truth jets from the MC. → Global sequential corrections to improve jet resolution and to remove the dependence on the flavour of the jet. → In situ calibration to remove the remaining differences between data and MC simulation. It is derived using well-measured reference objects, including γ, Z bosons, and calibrated jets.

Applied as a function of event pile-up pT density and jet area. Removes residual pile-up dependence, as a function of μ and NPV. Reconstructed jets Jet finding applied to tracking- and/or calorimeter-based inputs. Corrects jet 4-momentum to the particle-level energy

• scale. Both the energy and

direction are calibrated. Reduces flavour dependence and energy leakage effects using calorimeter, track, and muon-segment variables. A residual calibration is applied only to data to correct for data/MC differences. pT-density-based pile-up correction Residual pile-up correction Absolute MC-based calibration Global sequential calibration Residual in situ calibration

Josu Cantero (OSU) Heavy resonances 4 / 36

Jet reconstruction and calibration

0.5 1 1.5 2 2.5 3 3.5 4 4.5 |

det

η | 0.8 − 0.6 − 0.4 − 0.2 − 0.2 0.4 0.6 0.8 [GeV]

PV

N ∂ /

T

p ∂

Before any correction After area-based correction After residual corrections

Simulation ATLAS = 13 TeV, Pythia8 dijet s = 0.4 (PFlow) R

t

Anti-k 0.5 1 1.5 2 2.5 3 3.5 4 4.5 |

det

η | 0.8 − 0.6 − 0.4 − 0.2 − 0.2 0.4 0.6 0.8 [GeV] µ ∂ /

T

p ∂

Before any correction After area-based correction After residual corrections

Simulation ATLAS = 13 TeV, Pythia8 dijet s = 0.4 (PFlow) R

t

Anti-k

The jet-area method uses to estimate the energy density (ρ) due to pile-up.

→ pcorr

T

= pT - ρ × A - α× (NPV - 1) - β × µ

The negative dependence on µ for out-of-time pile-up is a result of the liquid-argon calorimeter’s pulse shape. Good stability of the pT of the jet after all corrections.

Josu Cantero (OSU) Heavy resonances 5 / 36

Jet reconstruction and calibration

3040 50

2

10

2

10 × 2

3

10

3

10 × 2 [GeV]

reco

E 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 Jet energy response = 0

det

η = 1

det

η = 1.4

det

η = 2.5

det

η = 4

det

η

ATLAS Simulation = 13 TeV, Pythia8 dijet s = 0.4 (PFlow) R

t

k Anti-

0.05 0.1 0.15 0.2 0.25

0.95 1 1.05 1.1 1.15 1.2 Response

T

p Jet

0.05 0.1 0.15 0.2 0.25

trk

Track width, w

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 Normalized entries < 25 GeV

true T

p 20 < < 100 GeV

true T

p 80 < < 250 GeV

true T

p 200 < < 1200 GeV

true T

p 1000 < ATLAS Simulation = 13 TeV, Pythia8 dijet s = 0.4 (PFlow+JES) R

t

k Anti- | < 0.3

det

η 0.2 < |

The absolute JES correction corrects the reconstructed jet four-momentum accounting for non-compensating calorimeter response, energy losses in dead material and out-of-cone effects. (R = Ereco/Etrue) The calibration is derived using a Pythia MC simulation of dijet events after the application of the pile-up corrections. After the JES correction, the response can vary from jet to jet depending

• n the flavour and energy distribution of the constituent particles.

→ A quark-initiated jet includes hadrons with a higher fraction of the jet pT that penetrate further into the calorimeter, while a gluon-initiated jet contains more particles of softer pT, leading to a lower calorimeter response and a wider transverse profile.

Josu Cantero (OSU) Heavy resonances 6 / 36

Jet reconstruction and calibration

20 30

2

10

2

10 × 2

3

10

3

10 × 2 [GeV]

jet T

p 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15

MC

R /

data

R ATLAS

• 1

= 13 TeV, 80 fb s = 0.4 (PFlow+JES) R

t

k Anti- +jet γ + jet ee → Z + jet µ µ → Z Multijet Total uncertainty Statistical component 20 30

2

10

2

10 × 2

3

10

3

10 × 2 [GeV]

jet T

p 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15

MC

R /

data

R ATLAS

• 1

= 13 TeV, 80 fb s = 0.4 R

t

k Anti- Total uncertainty, PFlow+JES Total uncertainty, EM+JES

One final calibration step to account for differences between the jet response in data and simulation causes by imperfect simulation of both the detector materials and the physics processes involved.

→ Final in situ calibration measures the jet response in data and MC and uses the ratio as an additional correction in data: c = Rdata

in situ

RMC

in situ

η intercalibration corrects the energy scale of forward (0.8 < ∣η∣ < 4.5) jets to match those of central (∣η∣ < 0.8) jets using the pT balance in dijet events. Z+jet and γ +jet analysis balance the hadronic recoil in an event against the pT of a calibrated Z boson or γ.

Josu Cantero (OSU) Heavy resonances 7 / 36

Jet reconstruction and calibration

Component Description η intercalibration Systematic mis-modelling Envelope of the generator, pile-up, and event topology variations Statistical component Statistical uncertainty (single component) Non-closure Three components describing non-closure at high energy and at η ∼ ±2.4 Non-closure, 2018 only Single component describing non-closure at η ∼ ±1.5 due to Tile calibration Z + jet Electron scale Uncertainty in the electron energy scale Electron resolution Uncertainty in the electron energy resolution Muon scale Uncertainty in the muon momentum scale Muon resolution (ID) Uncertainty in muon momentum resolution in the ID Muon resolution (MS) Uncertainty in muon momentum resolution in the MS MC generator Difference between MC event generators JVT cut Jet vertex tagger uncertainty ∆φ cut Variation of ∆φ between the jet and Z boson Subleading jet veto Radiation suppression through second-jet veto Showering & topology Modelling energy flow and distribution in and around a jet Statistical Statistical uncertainty in 28 discrete pT terms γ + jet Photon scale Uncertainty in the photon energy scale Photon resolution Uncertainty in the photon energy resolution MC generator Difference between MC event generators JVT cut Jet vertex tagger uncertainty ∆φ cut Variation of ∆φ between the jet and photon Subleading jet veto Radiation suppression through second-jet veto Showering & topology Modelling energy flow and distribution in and around a jet Photon purity Purity of sample used for γ + jet balance Statistical Statistical uncertainty in 16 discrete pT terms Multijet balance ∆φ (lead, recoil system) Angle between leading jet and recoil system ∆φ (lead, any sublead) Angle between leading jet and closest subleading jet MC generator Difference between MC event generators pasym

T

selection Second jet’s pT contribution to the recoil system Jet pT Jet pT threshold Statistical Statistical uncertainty in 28 discrete pT terms Pile-up µ offset Uncertainty in the µ modelling in MC simulation NPV offset Uncertainty in the NPV modelling in MC simulation ρ topology Uncertainty in the per-event pT density modelling in MC simulation pT dependence Uncertainty in the residual pT dependence Jet flavour Flavour composition Uncertainty in the proportional sample composition of quarks and gluons Flavour response Uncertainty in the response of gluon-initiated jets b-jets Uncertainty in the response of b-quark-initiated jets Punch-through Uncertainty in GSC punch-through correction Single-particle response High-pT jet uncertainty from single-particle and test-beam measurements AFII non-closure Difference in the absolute JES calibration for simulations in AFII

Josu Cantero (OSU) Heavy resonances 8 / 36

Jet reconstruction and calibration

20 30

2

10

2

10 × 2

3

10

3

10 × 2 [GeV]

jet T

p 0.02 0.04 0.06 0.08 Fractional JES uncertainty

ATLAS = 0.4 (PFlow+JES) R

t

k Anti- = 13 TeV s Data 2015-2017, = 0.0 η Inclusive jets Total uncertainty JES in situ Absolute JES in situ Relative

• Flav. composition
• Flav. response

Pile-up Punch-through

4 − 3 − 2 − 1 − 1 2 3 4 η 0.02 0.04 0.06 0.08 Fractional JES uncertainty

ATLAS = 0.4 (PFlow+JES) R

t

k Anti- = 13 TeV s Data 2015-2017, = 60 GeV

jet T

p Inclusive jets Total uncertainty JES in situ Absolute JES in situ Relative

• Flav. composition
• Flav. response

Pile-up Punch-through

5% of uncertainty for pT ≈ 20 GeV. It decreases to 1% for pT ≈ 200 GeV and < 1% for 200 GeV < pT < 2 TeV

→ the high-pT ‘single particle’ uncertainty is derived from studies of the response to individual hadrons and is used to cover the region beyond 2.4 TeV, where in-situ measurements no longer have statistical power.

Uncertainty due to pile-up and jet flavor response dominates at low pT.

Josu Cantero (OSU) Heavy resonances 9 / 36

Jet b-tagging

The identification of jets containing b-hadrons (b-jets) against the large jet background containing c-hadrons but no b-hadron (c-jets) or containing neither b- or c-hadrons (light-flavour jets) is of major importance in many areas of the ATLAS physics programme.

→ ATLAS uses various b-tagging algorithms. These algorithms exploit the long lifetime, high mass and high decay multiplicity of b-hadrons as well as the properties of the b-quark fragmentation.

Performance of a b-tagging algorithm is characterised by the probability of tagging a b-jet and the probability of mistakenly identifying a c-jet or a light-flavour jet as a b-jet. Identification of b-jets based on:

→ Track reconstructed in the ID with pT > 500 MeV and ∣η∣ < 2.5. → Primary vertex reconstruction: displaced tracks from b-hadron decays selected using d0 and z0 (transverse and longitudinal impact parameters): low-level b-tagging algorithm IP3D. → Secondary vertex consistent to b-hadron decay: low-level b-tagging algorithm SV1. → Topological structure of weak b- and c-hadron decays inside the jet: low-level b-tagging algorithm JetFitter .

High level b-tagging algorithms such as MV2 (DL1) uses boosted decision trees (deep neural networks) combining the information previously listed.

→ Mixed of tt and Z’ samples used for the training

Josu Cantero (OSU) Heavy resonances 10 / 36

Jet b-tagging

Input Variable Description Kinematics pT Jet pT η Jet |η| IP2D/IP3D log(Pb/Plight) Likelihood ratio between the b-jet and light- flavour jet hypotheses log(Pb/Pc) Likelihood ratio between the b- and c-jet hypo- theses log(Pc/Plight) Likelihood ratio between the c-jet and light- flavour jet hypotheses SV1 m(SV) Invariant mass of tracks at the secondary vertex assuming pion mass fE(SV) Energy fraction of the tracks associated with the secondary vertex NTrkAtVtx(SV) Number of tracks used in the secondary vertex N2TrkVtx(SV) Number of two-track vertex candidates Lxy(SV) Transverse distance between the primary and secondary vertex Lxyz(SV) Distance between the primary and the second- ary vertex Sxyz(SV) Distance between the primary and the second- ary vertex divided by its uncertainty ∆R( pjet, pvtx)(SV) ∆R between the jet axis and the direction of the secondary vertex relative to the primary vertex. JetFitter m(JF) Invariant mass of tracks from displaced vertices fE(JF) Energy fraction of the tracks associated with the displaced vertices ∆R( pjet, pvtx)(JF) ∆R between jet axis and vectorial sum of mo- menta of all tracks attached to displaced vertices Sxyz(JF) Significance of average distance between PV and displaced vertices NTrkAtVtx(JF) Number of tracks from multi-prong displaced vertices N2TrkVtx(JF) Number of two-track vertex candidates (prior to decay chain fit) N1-trk vertices(JF) Number of single-prong displaced vertices N≥2-trk vertices(JF) Number of multi-prong displaced vertices JetFitter c-tagging Lxyz(2nd/3rdvtx)(JF) Distance of 2nd or 3rd vertex from PV Lxy(2nd/3rdvtx)(JF) Transverse displacement of the 2nd or 3rd vertex mTrk(2nd/3rdvtx)(JF) Invariant mass of tracks associated with 2nd or 3rd vertex ETrk(2nd/3rdvtx)(JF) Energy fraction of the tracks associated with 2nd or 3rd vertex fE(2nd/3rdvtx)(JF) Fraction of charged jet energy in 2nd or 3rd vertex NTrkAtVtx(2nd/3rdvtx)(JF) Number of tracks associated with 2nd or 3rd vertex Y min

trk ,Y max trk ,Y avg trk (2nd/3rdvtx)(JF)

Min., max. and avg. track rapidity of tracks at 2nd or 3rd vertex

Josu Cantero (OSU) Heavy resonances 11 / 36

Jet b-tagging

0.5 0.6 0.7 0.8 0.9 1

b-jet tagging efficiency

0.5 1 1.5 2

Ratio to MV2 0.5 0.6 0.7 0.8 0.9 1 1 10

2

10

3

10

4

10

5

10 Light-flavour jet rejection ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p

MV2 DL1 IP3D SV1 JetFitter 0.5 0.6 0.7 0.8 0.9 1

b-jet tagging efficiency

0.5 1 1.5 2

Ratio to MV2 0.5 0.6 0.7 0.8 0.9 1 1 10

2

10 c-jet rejection ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p ATLAS Simulation t = 13 TeV, t s 2.5 ≤ | η 20 GeV, | ≥

T

Jet p

MV2 DL1 IP3D SV1 JetFitter

Four WPs based on the efficiency of b-flavoured jets are derived: 60%, 70%, 77%, 85% WPs. Improvements in the light-flavour jet and c-jet rejections by factors of around 10 and 2.5 for high-level b-tagging algorithms at ǫb = 70%.

Josu Cantero (OSU) Heavy resonances 12 / 36

Jet b-tagging

The performance of each b-tagging WP in the MC is corrected to the one

• bserved in data.

→ This is done by means of scale factors, SF (pT, η) = ǫdata(pT, η)/ǫMC(pT, η)

tt events in the di-lepton channel are selected in data and MC.

→ High purity of b-flavoured jets. → Events classified to extract flavour fractions: bb, bl, ll. → bb flavour fraction used to extract ǫb in data and MC.

30 40

2

10

2

10 × 2 [GeV]

T

Jet p 0.5 0.6 0.7 0.8 0.9 b-jet tagging efficiency ATLAS

• 1

= 13 TeV, 80.5 fb s = 70% single-cut OP

b

ε MV2,

Data (stat. unc.) Data (total unc.) MC t t

30 40

2

10

2

10 × 2 [GeV]

T

Jet p 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 b-jet tagging efficiency SF ATLAS

• 1

= 13 TeV, 80.5 fb s DL1, various single-cut OP

= 85% scale factor (total unc.)

b

ε = 77% scale factor (total unc.)

b

ε = 70% scale factor (total unc.)

b

ε = 60% scale factor (total unc.)

b

ε

Adequate description of ǫb by the MC. Similar SFs derived for each b-tagging WPs.

Josu Cantero (OSU) Heavy resonances 13 / 36

Jet b-tagging

Several different uncertainty sources considered.

→ High pT extrapolation uncertainties derived from MC to cover the high pT region where data is not available.

MV2 70% WP:

Source of uncertainty Relative uncertainty on εb [%] per jet pT bin [GeV] 20–30 30–40 40–60 60–85 85–110 110–140 140–175 175–250 250–600 Data statistics 3.7 1.7 0.7 0.6 0.6 0.6 0.8 1.1 2.8 MC statistics 2.2 1.0 0.4 0.2 0.2 0.2 0.2 0.2 0.5 Jet energy scale 4.5 0.8 0.3 0.1 0.1 0.1 0.1 0.2 0.4 t¯ t modelling 3.2 1.5 1.0 0.7 0.7 0.8 1.0 0.8 0.5 Single top modelling 2.5 0.5 0.6 0.6 0.4 0.3 0.3 0.4 1.1 Fake leptons modelling 1.8 1.1 0.1 0.2 < 0.1 < 0.1 0.2 < 0.1 0.2 Other sources 1.4 0.9 0.2 0.3 0.2 0.1 0.1 0.1 0.3 Total 7.7 3.0 1.4 1.1 1.0 1.1 1.3 1.5 3.1 [GeV]

T

Jet p

2

10

3

10 b-jet tagging efficiency SF 0.7 0.8 0.9 1 1.1 1.2 1.3 Scale factor (data-based, total unc.) Scale factor (smoothed, extrapolated) Uncertainty (data-based, smoothed) Uncertainty (extrapolation)

• 1

= 13 TeV, 80.5 fb s ATLAS = 70% single-cut OP

b

ε MV2, [GeV]

T

Jet p

2

10

3

10 b-jet tagging efficiency SF 0.7 0.8 0.9 1 1.1 1.2 1.3 Scale factor (data-based, total unc.) Scale factor (smoothed, extrapolated) Uncertainty (data-based, smoothed) Uncertainty (extrapolation)

• 1

= 13 TeV, 80.5 fb s = 70% single-cut OP

b

ε DL1, ATLAS Josu Cantero (OSU) Heavy resonances 14 / 36

Jet substructure and top tagging

Large-R jets (R = 1.0) using EMTopo objects as inputs.

→ Since mass of Z, W and top larger than light quarks, a large radious jet is needed to collect all the decay products.

Grooming technique used to remove the effects of pile-up and the underlaying event.

→ Difference with respect to small-R jets. Larger effects expected since R is large. → Trimming procedure in which original constituents of the jets are reclustered using the kt algorithm with a radius parameter Rsub = 0.2 to produce a collection of subjets. These subjets are then discarded if the pT is less than 5% of the pT of the original jet. → Jet mass calibration (JMS) step included in the calibration chain of large-R

• jets. The rest similar to what is done for small-R jets.

[GeV]

T

Truth jet p 1000 2000 3000 〉

truth

/ m

calo

m 〈 1 1.5 2 Simulation Preliminary ATLAS

| < 0.4

det

η = 13 TeV, QCD dijets, | s Energy-calibration only < 60 GeV

truth

40 GeV < m < 100 GeV

truth

80 GeV < m < 200 GeV

truth

160 GeV < m

[GeV]

T

Truth jet p 1000 2000 3000 〉

truth

/ m

calo

m 〈 1 1.5 2 Simulation Preliminary ATLAS

| < 0.4

det

η = 13 TeV, QCD dijets, | s Energy and mass calibration < 60 GeV

truth

40 GeV < m < 100 GeV

truth

80 GeV < m < 200 GeV

truth

160 GeV < m

Josu Cantero (OSU) Heavy resonances 15 / 36

Jet substructure and top tagging

From large-R jet constituents several observables can be defined to quantify a particular feature of the jet in an analytic way:

→ jet mass. → Splitting scales: d12, d23 ... → Energy correlation functions: C2, D2 ... → N-subjettiness: τ2, τ3, τ32 ...

23

d [GeV] 20 40 60 80 100 120 Normalized amplitude 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- = [500, 1000] GeV

true T

p | < 2

true

η | > 60 GeV

comb

m Jets W multijets Top Jets

2

C 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 Normalized amplitude 0.02 0.04 0.06 0.08 0.1 0.12 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- = [500, 1000] GeV

true T

p | < 2

true

η | > 60 GeV

comb

m Jets W multijets Top Jets

wta 32

τ 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Normalized amplitude 0.01 0.02 0.03 0.04 0.05 0.06 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- = [500, 1000] GeV

true T

p | < 2

true

η | > 60 GeV

comb

m Jets W multijets Top Jets

These variables can be used to derive “low-level” W/top taggers or combined using multivariate classifiers (BDT, DNN ...) to derive “high-level” top taggers. Shower deconstruction: top-tagger based on the reconstruction of subjets to determine whether the subjet pattern is compatible with a parton shower profile typical of a top-quark decay.

Josu Cantero (OSU) Heavy resonances 16 / 36

Jet substructure and top tagging

From large-R jet constituents several observables can be defined to quantify a particular feature of the jet in an analytic way:

→ jet mass. → Splitting scales: d12, d23 ... → Energy correlation functions: C2, D2 ... → N-subjettiness: τ2, τ3, τ32 ...

23

d [GeV] 20 40 60 80 100 120 Normalized amplitude 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- = [500, 1000] GeV

true T

p | < 2

true

η | > 60 GeV

comb

m Jets W multijets Top Jets

2

C 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 Normalized amplitude 0.02 0.04 0.06 0.08 0.1 0.12 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- = [500, 1000] GeV

true T

p | < 2

true

η | > 60 GeV

comb

m Jets W multijets Top Jets

SD

χ log 15 − 10 − 5 − 5 10 15 Normalized amplitude 0.02 0.04 0.06 0.08 0.1 0.12 0.14 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- = [500, 1000] GeV

true T

p | < 2

true

η | > 60 GeV

comb

m Jets W multijets Top Jets

These variables can be used to derive “low-level” W/top taggers or combined using multivariate classifiers (BDT, DNN ...) to derive “high-level” top taggers. Shower deconstruction: top-tagger based on the reconstruction of subjets to determine whether the subjet pattern is compatible with a parton shower profile typical of a top-quark decay.

Josu Cantero (OSU) Heavy resonances 17 / 36

Jet substructure and top tagging

W Boson Tagging Top Quark Tagging DNN Test Groups Chosen Inputs DNN Test Groups Chosen Inputs Observable 1 2 3 4 5 6 7 8 9 BDT DNN 1 2 3 4 5 6 7 8 9 BDT DNN mcomb

• pT
• e3
• C2
• D2
• τ1
• τ2
• τ3
• τ21
• τ32
• RFW

2

• P
• a3
• A
• zcut
• pd12
• pd23
• KtDR
• Qw
• Josu Cantero (OSU)

Heavy resonances 18 / 36

Jet substructure and top tagging

Training input groups

Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9

)

bkg rel

∈ Relative background rejection (1/ 10 20 30 40 50 60

ATLAS Simulation Tagging W = 13 TeV, DNN s = 1.0 jets R

t

k Trimmed anti- = 50%

rel sig

∈ = [200,2000] GeV

true T

p | < 2.0

true

η > 40 GeV, |

comb

m

Training input groups

Group 1 Group 2 Group 3 Group 4 Group 5 Group 6 Group 7 Group 8 Group 9

)

bkg rel

∈ Relative background rejection (1/ 1 2 3 4 5 6 7 8 9

ATLAS Simulation = 13 TeV, DNN Top Tagging s = 1.0 jets R

t

k Trimmed anti- = 80%

rel sig

∈ = [350,2000] GeV

true T

p | < 2.0

true

η > 40 GeV, |

comb

m

Different scenarios have been tested by grouping different set of variables. The performance of the DNN tagger depends on both the number of variables and the information content in the group. Found to be 12 variables for W -boson tagging (Group 8) and 13 variables for top-quark tagging (Group 9).

Josu Cantero (OSU) Heavy resonances 19 / 36

Jet substructure and top tagging

)

sig

∈ Signal efficiency (

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

)

bkg

∈ Background rejection (1 /

1 10

2

10

3

10

4

10 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- | < 2.0

true

η | = [500, 1000] GeV

true T

p Top tagging DNN top BDT top Shower Deconstruction 2-var optimised tagger HEPTopTagger v1 > 60 GeV

comb

m ,

32

τ

)

sig

∈ Signal efficiency (

0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

)

bkg

∈ Background rejection (1 /

1 10

2

10

3

10

4

10 ATLAS Simulation = 13 TeV s = 1.0 jets R

t

k Trimmed anti- | < 2.0

true

η | = [1500, 2000] GeV

true T

p Top tagging DNN top BDT top Shower Deconstruction 2-var optimised tagger HEPTopTagger v1 > 60 GeV

comb

m ,

32

τ TopoDNN

Similar performance for BDT and DNN multivariate classifiers. Large improvement on top-tagging performance by using multivariate classifiers with respect low-level taggers. Shower deconstruction (SD) top-tagger better than low-level taggers. Worse performance at high pT due to granularity of the calorimeter.

Josu Cantero (OSU) Heavy resonances 20 / 36

Jet substructure and top tagging

Performance of the top-tagging studied in data using tt events.

→ One top quark decays hadronically and the other semileptonically in both the electron and the muon decay channels. → b-tagged jet required within the top-candidate large-R jet to ensure t/t boosted topologies.

60 80 100 120 140 160 180 200 220 240 Events / 5 GeV 500 1000 1500 2000 2500

Data 2015+2016 (top) t t ) W ( t t (other) t t ) W Single Top ( Single Top (other) + jets W + jets, multijet Z , VV Total uncert.

• Stat. uncert.

modelling uncert. t t

ATLAS

• 1

= 13 TeV, 36.1 fb s =1.0 jets R

t

k Trimmed anti-

• jet) < 1.0

b jet, R (large- R ∆ > 350 GeV

T

p

[GeV]

comb

m jet R Leading large- 60 80 100 120 140 160 180 200 220 240 Data/Pred. 0.5 1 1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Events / 0.025 1000 2000 3000 4000 5000 6000

Data 2015+2016 (top) t t ) W ( t t (other) t t ) W Single Top ( Single Top (other) + jets W + jets, multijet Z , VV Total uncert. (excl. tagger)

• Stat. uncert.

modelling uncert. t t

ATLAS

• 1

= 13 TeV, 36.1 fb s =1.0 jets R

t

k Trimmed anti-

• jet) < 1.0

b jet, R (large- R ∆ > 350 GeV

T

p > 40 GeV

comb

m

jet DNN top discriminant R Leading large- 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Data/Pred. 0.5 1 1.5

→ Adequate description of the jet mass and DNN score distributions. → tt modelling uncertainties dominates.

Josu Cantero (OSU) Heavy resonances 21 / 36

Jet substructure and top tagging

The performance of top-tagging in the MC is corrected to the one

• bserved in data.

→ As in the b-tagging, this is done using scale factors. → Uncertainties on SD estimated by propagating the uncertainties on the subjet pT to the SD score. → Overall, good agreement on top tagging efficiencies between data and MC across the studied pT range.

400 600 800 1000 )

sig

∈ Signal efficiency ( 0.5 1 1.5

Data 2015+2016 PowhegPythia6 Total uncert. ATLAS

• 1

= 13 TeV, 36.1 fb s lepton+jets selection =1.0 jets R

t

k Trimmed anti- = 80%): DNN

sig

∈ Top tagger (

[GeV]

T

p jet R Leading large- 400 600 800 1000 Data/Pred. 0.5 1 1.5

400 600 800 1000 )

sig

∈ Signal efficiency ( 0.5 1 1.5

Data 2015+2016 PowhegPythia6 Total uncert. ATLAS

• 1

= 13 TeV, 36.1 fb s lepton+jets selection =1.0 jets R

t

k Trimmed anti- = 80%): SD

sig

∈ Top tagger (

[GeV]

T

p jet R Leading large- 400 600 800 1000 Data/Pred. 0.5 1 1.5

Josu Cantero (OSU) Heavy resonances 22 / 36

Analysis results: W ′ → tb

Theories beyond the Standard Model (SM) involve enhanced symmetries that predict new gauge bosons, usually called W ′ or Z ′ bosons. Many models such as those with extra dimensions, strong dynamics, composite Higgs, or the Little Higgs predict new vector charged-current interactions, some with preferential couplings to quarks or third-generation particles.

→ Sequential Standard Model (SSM) used to capture main phenomenology.

For large W’ masses, decay products of top quark decay become more collimated, such that, the top quark is reconstructed in a single large-R jet.

→ SD top tagging to identify jets from boosted top-quark decays, whereas b-tagging used to identify jets coming from b-quark.

Signal bump expected in the top (large-R jet) and b (small-R jet) candidates invariant mass mtb. L = 36.1 fb−1 of data used to perform this search.

→ Work in progress to include all Run 2 data, L = 139 fb−1.

q ¯ q′ W ′ ¯ b q ¯ q′ b t W +

Event reconstruction and selection Large-R jet (J) pJ

T > 420 GeV, |η| < 2.0

Small-R jet (j) pj

T > 25 GeV, |η| < 2.5

Top-quark jet candidate (Jcand

top )

jet J with highest mj + 0.15 × mJ b-quark jet candidate (jcand

b

) highest-pT jet j with pj

T > 420 GeV,

∆R(Jcand

top , j) > 2.0

Lepton veto zero leptons with pT > 25 GeV, |η| < 2.5 b-quark jet candidate η zero jcand

b

with |η| > 1.2 0 b-tag in zero b-tagged jets j with ∆R(Jcand

top , j) < 1.0

1 b-tag in exactly one b-tagged jet j with ∆R(Jcand

top , j) < 1.0

Josu Cantero (OSU) Heavy resonances 23 / 36

Analysis results: W ′ → tb

The dominant background from multi-jet production is estimated directly from data using a six-region “2D sideband” method that predicts both the shape and normalisation of mtb distribution.

→ Nbkg

A

= Rcorr

A

⋅

(Ndata

C

−Ntt

C )⋅(Ndata D

−Ntt

D)

Ndata

F

−Ntt

F

→ Nbkg

B

= Rcorr

B

⋅

(Ndata

C

−Ntt

C )⋅(Ndata E

−Ntt

E )

Ndata

F

−Ntt

F

→ Rcorr

A

and Rcorr

B

estimated from MC samples.

Three ortoghonal signal regions are defined based on top-tagging and b-tagging information. E B A D

Tight top-tagged Loose but not tight top-tagged Not b-tagged b-tagged

Large-R jet top candidate Small-R jet b-candidate

Not loose top-tagged

C F SR1 VR

0 b-tag in category

E B A D

Tight top-tagged Loose but not tight top-tagged Not b-tagged b-tagged

Large-R jet top candidate Small-R jet b-candidate

Not loose top-tagged

C F SR3 SR2

1 b-tag in category

Josu Cantero (OSU) Heavy resonances 24 / 36

Analysis results: W ′ → tb

Several systematic sources taken into account related to jet calibration, b-tagging SFs, top-tagging SFs, multijet background estimation, pile-up and tt modelling.

1 −

10 1 10

2

10

3

10

4

10

Events / 100 GeV

ATLAS

• 1

= 13 TeV, 36.1 fb s SR1

Data multi-jet + W/Z+jets t all-had t t non all-had t uncertainty pre-fit 3 TeV W'

1000 2000 3000 4000 5000 6000

[GeV]

tb

m

0.6 0.8 1 1.2 1.4

Data / Pred

1 −

10 1 10

2

10

3

10

4

10

5

10

Events / 100 GeV

ATLAS

• 1

= 13 TeV, 36.1 fb s VR

Data multi-jet + W/Z+jets t all-had t t non all-had t uncertainty pre-fit 3 TeV W'

1000 2000 3000 4000 5000 6000

[GeV]

tb

m

0.6 0.8 1 1.2 1.4

Data / Pred

To test for the presence of a massive resonance, mtb obtained from signal MC and backgrounds are fit to data using a binned maximum-likelihood approach. Systematic uncertainties incorporated into the fit as nuisance parameters with log-normal constraints. The p0-value estimated using the log-likelihood ratio (LLR) test statistic.

→ If no significant excess, upper limits at the 95% CL on the signal production cross-section times branching ratio are derived using the CLs method.

Josu Cantero (OSU) Heavy resonances 25 / 36

Analysis results: W ′ → tb

Several systematic sources taken into account related to jet calibration, b-tagging SFs, top-tagging SFs, multijet background estimation, pile-up and tt modelling.

1 −

10 1 10

2

10

3

10

4

10

Events / 100 GeV

ATLAS

• 1

= 13 TeV, 36.1 fb s SR3

Data multi-jet + W/Z+jets t all-had t t non all-had t uncertainty pre-fit 3 TeV W'

1000 2000 3000 4000 5000 6000

[GeV]

tb

m

0.6 0.8 1 1.2 1.4

Data / Pred

2 −

10

1 −

10 1 10

2

10

3

10

4

10

5

10

Events / 100 GeV

ATLAS

• 1

= 13 TeV, 36.1 fb s SR2

Data multi-jet + W/Z+jets t all-had t t non all-had t uncertainty pre-fit 3 TeV W'

1000 2000 3000 4000 5000 6000

[GeV]

tb

m

0.6 0.8 1 1.2 1.4

Data / Pred

To test for the presence of a massive resonance, mtb obtained from signal MC and backgrounds are fit to data using a binned maximum-likelihood approach. Systematic uncertainties incorporated into the fit as nuisance parameters with log-normal constraints. The p0-value estimated using the log-likelihood ratio (LLR) test statistic.

→ If no significant excess, upper limits at the 95% CL on the signal production cross-section times branching ratio are derived using the CLs method.

Josu Cantero (OSU) Heavy resonances 26 / 36

Analysis results: W ′ → tb

1000 1500 2000 2500 3000 3500 4000 4500 5000 ) [GeV]

R

m(W'

2 −

10

1 −

10 1 10 tb) [pb] →

R

B(W' × )

R

W' → (pp σ ATLAS

• 1

= 13 TeV, 36.1 fb s

Observed 95% CL limit Expected 95% CL limit σ 1 ± Expected 95% CL limit σ 2 ± Expected 95% CL limit NLO W' cross-section (ZTOP)

1000 1500 2000 2500 3000 3500 4000 4500 5000 ) [GeV]

L

m(W'

2 −

10

1 −

10 1 10 tb) [pb] →

L

B(W' × )

L

W' → (pp σ ATLAS

• 1

= 13 TeV, 36.1 fb s

Observed 95% CL limit Expected 95% CL limit σ 1 ± Expected 95% CL limit σ 2 ± Expected 95% CL limit NLO W' cross-section (ZTOP)

Exclusion limits derived for right- and left-handed couplings. NLO theoretical prediction for W ′ production computed using Ztop program. For m(W ′) ≳ 2.0 TeV, σ × B > 0.1 pb excluded. Assuming Ztop parameters (SM couplings), m(W ′

R) (m(W ′ L)) < 3.0

(2.85) TeV excluded.

Josu Cantero (OSU) Heavy resonances 27 / 36

Analysis results: Z ′ → bb

Models favouring couplings fo gauge bosons to third generation quarks in general.

→ Different models tested in this search: SSM, DM models with Z ′ mediator, KK resonances.

L = 139 fb−1 of data used to perform this search. New b-tagging algorithm used for this search: DL1r.

→ Better performance for high pT jets.

Signal bump expected in the invariant mass of two leading small-R jets.

→ Both small-R jets fulfilling 77% b-tagging WP.

[GeV]

T

p

2

10

3

10

b-tagging efficiency SF

0.7 0.8 0.9 1 1.1 1.2

Scale factor Smoothed and extrapolated scale factor Data-based uncertainty Extrapolation uncertainty ATLAS

• 1

= 13 TeV, 80.5 fb s = 77% Fixed Cut

b

ε DL1r,

Event display with two high-pT jets; pT = 3.0 and 2.9 TeV respectively

Josu Cantero (OSU) Heavy resonances 28 / 36

Analysis results: Z ′ → bb

SR defined by requiring ∣y ∗∣ < 0.8 → contribution from s-channel enhanced. Multijet events main background.

→ Estimated using sliding-window fitting method using a parametric function: f(x) = p1(1 − x)p2xp3+p4 log x. → x = mjj/√s. → Fit validated in a CR with no b-tagging requirement multiplied by the appropiate b-tagging efficiencies. → Signal injection and spurius signal tests performed to evaluate the robustness of the background fitting strategy.

Bumphunter tool to look for local excesses in the mjj distribution.

→ No (significant) local excess was found.

Jet-related and b-tagging uncertainties propagated to signal templates.

1.5 2 2.5 3 3.5 4 4.5

1 −

10 1 10

2

10

3

10

4

10

5

10

6

10

7

10

Events 1.5 2 2.5 3 3.5 4 4.5

[TeV]

jj

m

2 − 2 Significance

ATLAS

• 1

=13 TeV, 139 fb s 2 b-tag Data Background fit BumpHunter interval = 2 TeV

Z'

DM Z', m = 3 TeV

Z'

DM Z', m

• value = 0.83

p 10 × σ =0.25,

q

DM Z' g

1.5 2 2.5 3 3.5 4 [TeV]

SSM Z'

m

4 −

10

3 −

10

2 −

10

1 −

10 BR [pb] × ∈ × A × σ

Theory Observed 95% CL Expected 95% CL σ 1 ± σ 2 ±

ATLAS

• 1

= 13 TeV, 139 fb s ), 2 b-tag b SSM Z'(b Josu Cantero (OSU) Heavy resonances 29 / 36

Analysis results: Z ′ → bb

SR defined by requiring ∣y ∗∣ < 0.8 → contribution from s-channel enhanced. Multijet events main background.

→ Estimated using sliding-window fitting method using a parametric function: f(x) = p1(1 − x)p2xp3+p4 log x. → x = mjj/√s. → Fit validated in a CR with no b-tagging requirement multiplied by the appropiate b-tagging efficiencies. → Signal injection and spurius signal tests performed to evaluate the robustness of the background fitting strategy.

Bumphunter tool to look for local excesses in the mjj distribution.

→ No (significant) local excess was found.

SSM Z ′ with mZ′ ≲ 2.8 TeV excluded.

[TeV]

DM mediator Z'

m 1 1.5 2 2.5 3 3.5 4 4.5 5 BR [pb] × σ

3 −

10

2 −

10

1 −

10 1 10

)

• 1
• Phys. Rev. D 98, 032016 (36.1 fb

)

• 1
• Phys. Rev. D 98, 032016 (Scaled to 139 fb

)

• 1

Current Result (139 fb

ATLAS = 13 TeV s = 0.25, 2 b-tag

q

), g b DM mediator Z'(b 1.5 2 2.5 3 3.5 4 [TeV]

SSM Z'

m

4 −

10

3 −

10

2 −

10

1 −

10 BR [pb] × ∈ × A × σ

Theory Observed 95% CL Expected 95% CL σ 1 ± σ 2 ±

ATLAS

• 1

= 13 TeV, 139 fb s ), 2 b-tag b SSM Z'(b Josu Cantero (OSU) Heavy resonances 30 / 36

Analysis results: Z ′ → tt

Models including heavy resonances decaying into tt pair are studied, such as, top-color-assisted-technicolor (TC2), two-Higgs-doublet model (2HDM) and Randall-Sundrum (RS) models of warped extra dimensions. For large resonance masses, decay products of top and anti-top quark decays become more collimated, leading to final states with two high pT large-R jets.

→ DNN top tagging 80% WP is used to identify jets from boosted top and anti-top quark decays. → b-tagging requirements applied to VR trackjets found within large-R jets.

Signal bump expected in the invariant mass of the top and anti-top large-R jet candidates, mtt. L = 139 fb−1 of data used to perform this search.

Josu Cantero (OSU) Heavy resonances 31 / 36

Analysis results: Z ′ → tt

Two SR are defined depending on the number of b-tagged jets found in the final state (nb = 1 or 2).

→ For both SRs (SR1b and SR2b) top-candidates must fulfill 80% top tagger WP. → 51% (90%) background contribution from tt SM production in SR1b (SR2b) → Remaining background coming from multijet production.

Background contribution in SRs estimated from fits to parametric function: f(x) = p0(1 − x)p1xp2+p3 log x+p4 log x2.

→ Fitting function validated using the expected mtt in SR from a data-driven estimation of multijet contribution and tt MC distribution. → Wilk’s test to determine the optimal number of parameters to describe the function: most optimal function found for p4 = 0.0. → Spurius signal studies by performing S+B fits on a background only distribution.

Top tagging SFs plus uncertainties included in the MC predictions.

Together with large-R jet related uncertainties, b-tagging uncertainties.

Bumphunter tool to look for local excesses in the mtt distribution.

→ No (significant) local excess was found.

Josu Cantero (OSU) Heavy resonances 32 / 36

Analysis results: Z ′ → tt

Bumphunter tool to look for local excesses in the mtt distribution.

→ No (significant) local excess was found. → Global p-values of 0.45 and 0.56 for SR1b and SR2b respectively. → Local excesses less than 2-σ away from the SM prediction.

Exclusion limits at 95% CL performed using a test statistic based on the profile likelihood ratio.

2000 3000 4000 5000 6000

[GeV]

reco t t

m

3 − 2 − 1 − 1 2 3 Significance BumpHunter

3 −

10

2 −

10

1 −

10 1 10

2

10

3

10 Events / GeV Data Background fit Fit parameter unc. x5

TC2

2 TeV Z' x5

TC2

4 TeV Z' interval (5440 - 5690 GeV) Most significant deviation

ATLAS

• 1

= 13 TeV, 139 fb s SR1b BH global p-value = 0.45 2000 3000 4000 5000 6000

[GeV]

reco t t

m

3 − 2 − 1 − 1 2 3 Significance BumpHunter

4 −

10

3 −

10

2 −

10

1 −

10 1 10

2

10

3

10 Events / GeV Data Background fit Fit parameter unc. x5

TC2

2 TeV Z' x5

TC2

4 TeV Z' interval (5440 - 5820 GeV) Most significant deviation

ATLAS

• 1

= 13 TeV, 139 fb s SR2b BH global p-value = 0.56

Josu Cantero (OSU) Heavy resonances 33 / 36

Analysis results: Z ′ → tt

Signal bump expected in the invariant mass of the top and anti-top large-R jet candidates, mtt.

→ No (significant) local excess was found. → Global p-values of 0.45 and 0.56 for SR1b and SR2b respectively. → Local excesses less than 2-σ away from the SM prediction.

Exclusion limits at 95% CL performed using a test statistic based on the profile likelihood ratio.

→ Limits on topcolor-assisted-technicolor model, resulting in the exclusion of Z ′ masses up to 3.9 and 4.9 TeV for decay widths of 1% and 3%, respectively.

1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 [TeV]

Z'

m

3 −

10

2 −

10

1 −

10 1 10 ) [pb] t t → B(Z' × Z') → (pp σ

Observed 95% CL upper limit Expected 95% CL upper limit σ 1 ± Expected 95% CL upper limit σ 2 ± Expected 95% CL upper limit 1.3 × /m=1.2%) cross-section Γ ( t t →

TC2

LO Z' /m=1%) cross-section Γ ( t t →

TC2

NLO Z' /m=3%) cross-section Γ ( t t →

TC2

NLO Z'

ATLAS

• 1

= 13 TeV, 139 fb s

2 2.5 3 3.5 4 4.5 5 [TeV]

Z'

m

3 −

10

2 −

10

1 −

10 ) [pb] t t → B(Z' × Z') → (pp σ

)

• 1
• Phys. Rev. D 99, 092004 (36.1 fb
• 1

Current analysis with 36.1 fb

• 1

Current analysis with 139 fb 1.3 × /m=1.2%) cross-section Γ ( t t →

TC2

LO Z'

ATLAS = 13 TeV s Expected 95% CL upper limit Josu Cantero (OSU) Heavy resonances 34 / 36

Conclusions

Searches for heavy resonances decaying into third generation quarks have been presented.

→ W ′ → tb with L = 36.1 fb−1; SSM m(W ′

R) (m(W ′ L)) < 3.0 (2.85) TeV

excluded. → Z ′ → bb with L = 139 fb−1; SSM m(Z ′) < 2.8 TeV excluded. → Z ′ → tt with L = 139 fb−1; SSM m(Z ′

TC2) < 3.9 (4.9) TeV excluded for

Γ/m = 1% (3%)

In general, these limits on mW ′,Z′ are relaxed by assuming smaller couplings to SM quarks.

→ Is there any deep reason to assume gqqV ′ ≈ gSM

qqV ?. I guess this will depend

• n the particular theoretical model ...

→ 2D limits (mW ′,g′) searching for W ′ → tb using the leptonic decay of the top.

New techniques included in the b-tagging and top-tagging algorithms.

→ Better performance compared to the low level algorithms. → Important to be able to properly compute the systematic uncertainties associated to these new WPs.

A lot of work done in the performance size for a thoroughly estimation of the systematic uncertainties and its correlations associated to jets, top-tagging and b-tagging.

→ Important role in the profile likelihood fit. → A measurement must be always accompanied by its error.

Josu Cantero (OSU) Heavy resonances 35 / 36

Thank you

Josu Cantero (OSU) Heavy resonances 36 / 36

Backup

Josu Cantero (OSU) Heavy resonances 1 / 0

W ′ → tb leptonic channel

[TeV]

R

W'

m 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 g'/g

1 −

10 1 ATLAS

• 1

= 13 TeV, 36.1 fb s ν l b b → b t →

R

W'

Observed 1 s.d. ± Expected

Josu Cantero (OSU) Heavy resonances 2 / 0