Resonance Searches with an Updated Top Tagger G. Kasieczka, T. - - PowerPoint PPT Presentation

▶
resonance searches with an updated top tagger
SMART_READER_LITE
LIVE PREVIEW

Resonance Searches with an Updated Top Tagger G. Kasieczka, T. - - PowerPoint PPT Presentation

Aug 15, 2023 181 likes •319 views

Resonance Searches with an Updated Top Tagger G. Kasieczka, T. Plehn, T.S., T. Strebler, G. P. Salam [arXiv:1503.05921] Torben Schell Institute for Theoretical Physics, Heidelberg University Pheno 2015 May 4, 2015 T. Schell (ITP U

slide-1
SLIDE 1

Resonance Searches with an Updated Top Tagger

  • G. Kasieczka, T. Plehn, T.S., T. Strebler, G. P. Salam [arXiv:1503.05921]

Torben Schell

Institute for Theoretical Physics, Heidelberg University

Pheno 2015 May 4, 2015

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 1 / 12

slide-2
SLIDE 2

HEPTopTagger

HEPTopTagging

reconstruction of boosted hadronic tops collimated decay products → fat jets → reduced combinatorial problems SM: number of top quarks vs. collimation substructure analysis based on subjet masses

1 10

2

10

3

10

[GeV]

T

P 200 400 600

bjj

R Δ 1 2 3

[Plehn et al. arXiv:1006.2833]

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 2 / 12

slide-3
SLIDE 3

HEPTopTagger

HEPTopTagger – Algorithm

fat jet: C/A R = 1.5, pT > 200 GeV

1

hard substructures: mass drop fdrop = 0.8, mi < msub = 30 GeV

2

filtering: filter triplets of hard substructures → 3 jets (j1, j2, j3) 150 GeV < m123 < 200 GeV

3

mass plane cuts: 0.85 mW

mt < mij m123 < 1.15 mW mt

m23 ≈ mW : 0.2 < arctan m13

m12 < 1.3; else m23 m123 > 0.35

4

triplet selection: choose triplet closest to mt

5

consistency: p(tag)

T

> 200 GeV

[arXiv:1006.2833]

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 3 / 12

slide-4
SLIDE 4

HTT Resonance Reconstruction

Resonance Reconstruction

Heavy neutral Z ′–gauge bosons decaying to top quarks at LHC run II Event generation: Pythia8, LHC √s = 13 TeV signal: Z ′ → th¯ th, mZ′ = 1500 GeV, Γ(Z ′) = 65 GeV background: QCD-dijet & th¯ th, both pT > 400 GeV no detector simulation Event selection: 2 hardest C/A, R = 1.5 fat jets (FastJet) require pT,fat > 400 GeV and |yfat| < 2.5

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 4 / 12

slide-5
SLIDE 5

HTT Resonance Reconstruction

Decay Kinematics

HTT working point + mtt window

Z′ → t¯ t t¯ t QCD generator level 105 105 (1.76 pb) 8 · 106 (1.93 nb) ≥ 2 fat jets with pT > 400 GeV and |y| < 2.5 69142 85284 (1.50 pb) 6.7 · 106 (1.62 nb) hardest 2 fat jets HTT [JHEP1010] tagged 9679 11706 (0.21 pb) 4426 (1.07 pb) mtt ∈ [1200, 1600] GeV 7031 2817 (0.05 pb) 978 (0.24 pb)

include additional kinematic variables in BDT analysis decay kinematics well described by { mtt, pT,j }

2 4 0.05 0.1 0.15 y| ∆ | y| ∆ d| σ d σ 1 Z' t t QCD

S

ε 0.05 0.1 0.15 0.2

B

ε 1 /

3

10

4

10

5

10 HTT[JHEP1010]

tt

m y| ∆ + |

tt

m

T,i

+ p

tt

m y| ∆ + |

T,i

+ p

tt

m

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 5 / 12

slide-6
SLIDE 6

HTT Resonance Reconstruction

Final State Radiation

HTT reconstructs on–shell tops → misses final state radiation → sizeable tail in mtt distribution consider HTT tagged fat jets

Z′ W W b b

500 1000 1500 2000 0.05 0.1 [GeV]

tt

m GeV 1

tt

dm σ σ d Z' t t QCD 500 1000 1500 2000 0.05 0.1 0.15 [GeV]

ff

m GeV 1

ff

dm σ σ d R=0.3, N=5 Filtering: Z' t t QCD

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 6 / 12

slide-7
SLIDE 7

HTT Resonance Reconstruction

Final State Radiation & Variable Masses

filtered fat jet momenta instead of reconstructed tops → no improvement add filtered fat jet information: { mtt, pT,j, m(filt)

ff

, p(filt)

T,fj }

going beyond HTT working point: variable masses in HTT cuts + corresponing variables in BDT {mtt, pT,j, m(filt)

ff

, p(filt)

T,fj , mmin rec , mmax rec , f max rec },

frec = min

ij

|(mij/mrec)/(mW /mt) − 1|

S

ε 0.05 0.1 0.15 0.2

B

ε 1 / 10

2

10

3

10

4

10

5

10 HTT[JHEP1010] decay kinematics (2) filtered fat jets (3) QCD t t

S

ε 0.05 0.1 0.15 0.2 0.25

B

ε 1 /

3

10

4

10

5

10 HTT[JHEP1010] filtered fat jets (3) variable masses (4)

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 7 / 12

slide-8
SLIDE 8

HTT Resonance Reconstruction

OptimalR Mode

there is an optimal fat jet size Ropt reduce R until leaving top mass plateau |m(R)

rec − m (Ropt) rec

| < 0.2 m

(Ropt) rec

→ Ropt estimate as R(calc)

  • pt

→ additional variable Ropt − R(calc)

  • pt

OptimalR {mtt, pT,j, m(filt)

ff

, p(filt)

T,fj , mmin rec , mmax rec , f max rec , max(Ropt − R(calc)

  • pt

) }

T,filt

p 200 400 600 800 1000

bjj

R 1 2 >200, 400, 600 GeV combined

T

p

S

ε 0.05 0.1 0.15 0.2 0.25

B

ε 1 /

3

10

4

10

5

10 HTT[JHEP1010] filtered fat jets (3) variable masses (4)

  • ptimalR (6)

= 13 TeV s

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 8 / 12

slide-9
SLIDE 9

HTT Resonance Reconstruction

N–Subjettiness

  • ptimalR working point

mrec ∈ [150, 200] GeV, frec < 0.175, Ropt − R(calc)

  • pt

< 0.3 two different filterings and BDT analyses passed: Rfilt = 0.3, Nfilt = 3 rejected: Rfilt = 0.2, Nfilt = 5 N–Subjettiness [Thaler, Van Tilburg] τN = 1 R0

  • k pT,k
  • k

pT,k min (∆R1,k, · · · , ∆RN,k)

S

ε 0.05 0.1 0.15 0.2 0.25

B

ε 1 /

3

10

4

10

5

10 HTT[JHEP1010] filtered fat jets (3) variable masses (4)

  • ptimalR (6)

N-subjettiness (8) = 13 TeV s

BDTs: {mtt, mff, pT,t1, pT,t2, pT,f1, pT,f2, mmin

rec , mmax rec , f max rec , Ropt − R(calc)

  • pt

, τfi ,N, τ (filt)

fi ,N }

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 9 / 12

slide-10
SLIDE 10

HTT Resonance Reconstruction

Qjets

[Ellis, Hornig, Roy, Krohn, Schwartz]

deterministic clustering → set of weighted histories each possible merging (ij) gets a weight ω(α)

ij

= exp

  • −α dij − dmin

ij

dmin

ij

  • clustering history weight

Ω(α) =

  • mergings

ω(α)

ij

=

mergings

exp

  • −dij − dmin

ij

dmin

ij

α

S

ε 0.2 0.4 0.6

B

ε 1 /

2

10

3

10

4

10

5

10 HTT[JHEP1010] filtered fat jets (3) variable masses (4)

  • ptimalR (6)

N-subjettiness (8) Qjets (11, 0.1x0.1 cells) = 13 TeV s

use leading tagged Qjets history + statistical information from tagged histories { mtt, mff, pT,t1, pT,t2, pT,f1, pT,f2, mmin

rec , mmax rec , f max rec , Ropt−R(calc)

  • pt

, {τN}, εmin

Qjets, {mQjets rec } }

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 10 / 12

slide-11
SLIDE 11

HTT Resonance Reconstruction

Comparison

Event Deconstruction[Soper, Spannowsky] likelihoods based on up to 9 C/A microjets per fat jet (R = 0.2 and pT > 10 GeV) soft and/or collinear approximation of QCD event classification based on likelihood ratio

S

ε 0.2 0.4 0.6

B

ε 1 /

2

10

3

10

4

10

5

10 ED[PRD89] HTT[JHEP1010] filtered fat jets (3) variable masses (4)

  • ptimalR (6)

N-subjettiness (8) Qjets (11, 0.1x0.1 cells) = 14 TeV s

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 11 / 12

slide-12
SLIDE 12

Summary

Summary

Resonance search with an updated HEPTopTagger and additional kinematic variables: fat jet kinematics to account for final state radiation algorithmically optimized size of used fat jets and its prediction (optimalR) N–subjettiness probing more general substructures inside the fat jet Qjets with a global picture of the most likely clustering histories giving a top tag → factor 30 improvement compared to the previous HEPTopTagger version

  • T. Schell (ITP – U Heidelberg)

HEPTopTagging Pheno 2015 12 / 12