Search for W'→tb in the hadronic final state at ATLAS · 2015. 1. 15. · 10 Hadronic top reconstruction Want to differentiate jets from hadronic top decays from QCD Top jet: high

1

Search for W'→tb in the hadronic final state at ATLAS

Ho Ling LiThe University of Chicago

January 5, 2015

2

Introduction

● What is W'?● A charged, heavy, Standard-Model-like W gauge boson

● Analysis channels● W'→lν, l = e, μ● W'→tb

● First result on W' in the decay channel of tb→qqbb● Available on http://arxiv.org/abs/1408.0886

● Why this analysis?● Why W'?● Why W'→tb?● Why W'→tb→qqbb?

● You may also ask:● Why hasn't it been done before?

3

Why W' and W'→tb?

● Why W'?● Consequence of many BSM theories

● Extra Dimension model, e.g. Kaluza-Klein tower of W● A new gauge sector, e.g. Little Higgs theories● Left-right symmetry: W'

R counterpart for SM W

L

● Why W'→tb?● Search for W'

R without assuming the existence and mass of νR

● Couple to the 3rd generation● Current mass limit: W'→lν (3.35 TeV, CMS) > W'→tb (2.03 TeV,

CMS), but SM couplings assumed● Weakly coupled to leptonic channel, or even leptophobic, while

strongly coupled to quarks?

4

Why W'→tb→qqbb?

● t → e/μ + ν ~ 2/9● t → qq ~ 2/3

leptonic top decay channelleptonic W' → tb

hadronic top decay channelhadronic W' → tb

● Before August 2014, only W'→tb→lνbb published● Current observed limits at ATLAS ~ 1.8 TeV

● Adv.: Existence of l significantly lowers background from light quarks and gluons (QCD)

● Disadv.: Lose sensitivity when mW'

> 1.75 TeV

7

5

Why W'→tb→qqbb?

● t → e/μ + ν ~ 2/9● t → qq ~ 2/3

● This is the first experimental result on W' → tb → qqbb● Advantages

● Br(t → qqb) ~ 3 x Br(t → lνb), l = e or μ● Able to reconstruct a sharp W' mass peak● Sensitivity maintained for high m

W'

● Disadvantage● Enormous background from QCD

leptonic top decay channelleptonic W' → tb

hadronic top decay channelhadronic W' → tb

6

W' → tb model

● Use a model independent effective field theory approach, keep this analysis as generic search● Independent of other phenomena in the models, only concern W'

● MC signal sample: SM W to fermion couplings (gSM

) as benchmark● Both W'

L and W'

R generated from 1 – 3.5 TeV in 250 GeV

increments● Set W' mass limit as a function of g'/g

SM

g'

7

The ATLAS detector

z-direction

8

Jet reconstruction

● Signal involves top quarks and b quarks; QCD as main background

● Top, b, QCD are identified as jets

● Additional criteria to distinguish top jets and b jets from QCD

9

b-tagging: the method to identify b jets

● B hadron has a very long lifetime (~10-12 s)

millimeters

10

Hadronic top reconstruction

● Want to differentiate jets from hadronic top decays from QCD● Top jet: high mass (173 GeV) and contains 3 showers (q, q, b)

● Search for W' with mass greater than 1.5 TeV● Top has high P

T → decay products tend to merge

● Reconstruct top decay products by large-R (radius para. R=1.0) jets● ATLAS standard jets have radius parameter R=0.4

● Use jet substructure information to distinguish top jets from light jets

high PT top

e.g. 800 GeV

low PT top

e.g. 100 GeV

11

Jet substructure variables for top-tagging

● Splitting scale √d12

, ratios of n-subjettiness τ21

and τ32

are used to distinguish top jets and light jets● Splitting scale √d

12

√d12

= min(PT(1), P

T(2)) x ΔR(1,2)

12

Jet substructure variables for top-tagging

● Ratio of n-subjettiness τ32 and τ21● Ratio related to number of constituents inside the jet

τ32

after applying √d

12

τ21

after applying √d

12

and τ32

τ32

→ 0 τ32

→ 1 τ21

→ 0 τ21

→ 1

top-like QCD-like top-likeQCD-like

13

W' top-tagger performance

2 TeV WL' → tb

QCD sample (2012 data)

● Initially named Love-Li top-tagger due to its simplicity and lovely performance● Performance compatible with more complex algorithms

46%

6%fa

ke r

ate

14

Analysis strategy● Looking for tagged dijet events

● Electron and muon veto applied● Still, leptonic W' → tb contributes 3.5 – 10% of signal yields

● Search for a bump in mtb spectrum

● Start the search at W' mass > 1.5 TeV

jetjet

45% b-tagging efficiencyif identified, 2 b-tag category; otherwise, 1 b-tag category

top-tagged large-R jet with P

T > 350 GeV

(always required)

b-tagged small-R jet with P

T > 350 GeV

(always required)

b

∆R(large-R, small-R) > 2.0

15

Selection efficiency● Normalized to 20.3 fb-1

● Leptonic W' → tb analysis has ~2% efficiency

total

2 b-tag category

1 b-tag category

QCD background ttbar background 2 TeV W'L signal 2 TeV W'

R signal

1 b-tag 16100 1900 58 85

2 b-tag 2600 300 45 74

16

Statistical strategy

● Parametrize signal distributions by analytic function● Fit background from data with analytic function● Unbinned likelihood fit to m

tb distribution

● Determine excess from probability for signal+background hypothesis● If no excess, set 95% Confidence Level limits

● Systematics taken into account as nuisance parameters

2 b-tag category2 TeV W'

L

17

Signal parametrization● Hadronic signal MC parametrized by Skew Normal + Gaussian

● Fit parameters can be parametrized into nice functions● Interpolate parameters between generated m

W' points

● Leptonic W' → tb signal MC parametrized by double Gaussian

central value of Skew NormalW'

L, 2 b-tag

2.5 TeV W'L

3 TeV W'L

2.5 TeV W'L

3 TeV W'L

18

Data-driven method for QCD background estimation

not b-tagged b-tagged

not top-tagged NA

NB

top-tagged NC

Nsignal

Extrapolate into signal region:● If N

B/N

A = N

signal/N

C → good

● Compare extrapolation with N

signal, example by using QCD

MC

b?

b? if yes, 2 b-tag category

top?

1 b-tag channel

● If found b-tagged small-R jet inside large-R jet → 2 b-tag category

extrapolation

19

Background parametrization

● Use data-driven QCD background sample + ttbar from MC to test functional form

2 b-tag category

parametrized by ExpPoly2: a

0*exp(a

1*m+a

2*m2)

1 b-tag category

parametrized by ExpPoly4

20

Background systematics “spurious signal”● Spurious signal: bias from choice of background function● Fit background with signal+background hypothesis to extract number

of “signal” events as spurious signal● Use data-driven QCD sample + ttbar from MC

ExpPoly2, 2 b-tag, W'L

choose simplest functions with low spurious signals

21

● Percent change of systematic uncertainties on event yield of 2 TeV W'

● b-tagging is the largest systematics in 2 b-tag category● Jet energy scale (JES) and jet energy resolution (JER) are small

Systematic evaluations

systematics W'L

W'R

1 btag 2 btag 1 btag 2 btagb-tagging +13/-20 +45/-37 +15/-21 +40/-34

top-tagging +11.0/-12.5 +8.8/-10.0 +10.3/-11.0 +8.0/ -8.8luminosity ±2.8 ±2.8 ±2.8 ±2.8

jet energy scale +1.0/-1.3 +1.5/-1.9 +0.6/-0.8 +1.4 / -1.9jet energy resolution -0.0 -0.2 -0.1 -0.5

22

Systematic example: b-tagging systematics

Change in event yields

● Example from b-tagging systematics for W'L in 2 b-tag category

23

Systematic example: jet energy scale (JES)

● Example from JES for W'L in 2 b-tag category

Shift in the width of skewnormal

Shift in the width of Gaussian

24

Now we open the box

25

Data● Background obtained from fitting data with functional forms

26

Data● Data consistent with SM

prob

abili

ty

prob

abili

ty prob

abili

ty

27

Observed (expected) 95% CL limits

● Observed (expected) limits on cross section x Br assuming gSM

● mass of W'L > 1.68 (1.63) TeV

● mass of W'R > 1.76 (1.85) TeV

28

Limits in terms of coupling strength g'

● Set limits on g'/gSM

up to 2 as a function of mW'

● At g'/gSM

= 2, mass limit is 2.18 (2.29) for W'L (W'

R)

● At mW'

= 1.5 TeV, g'/gSM

< 0.70 (0.55)

29

Leptonic and hadronic W' → tb

● Currently, comparable expected limits● Hadronic: flat sensitivity up to m

W' ~ 3 TeV

hadronic W' → tb

leptonic W' → tb

hadronic W' → tb

leptonic W' → tb

30

Summary and outlook

● First result on W' → tb in the decay channel of t → qqb● Dijets events with one jet top-tagged and the other one b-tagged

● Developed new and simple top-tagger with high performance

● Limits on cross section x Br (g' = gSM

)● Mass of W'

L > 1.68 TeV

● Mass of W'R > 1.76 TeV

● Limits on g'/gSM

as a function of mW'

● At g'/gSM

= 2, mass limit is 2.18 (2.29) for W'L (W'

R)

● Outlook● Sensitivity up to W' mass ~ 3 TeV● More important as LHC increases to 13 - 14 TeV for Run II (2015 –

2018)

● Thank you!

31

backup

32

Trimming

● Remove parton contamination in fat jets

● Remove subjets with a cone size R that have PT < P

Tjet * f

cut

● In our case, R is 0.3 and fcut

is 0.05

33

Splitting scale

34

N-subjettiness

35

W' top-tagger performance

● Shown compatible performance in ATLAS top tagger performance note

36

Trigger and HT performance

● Use HT trigger EF_j170_a4tchad_ht700

● Study the HT trigger plateau

● Offline HT calculated by summing over all AntiKt4 jets with P

T > 25

GeV and |η|<2.5

● Events used in this study are selected with the following cuts● Trimmed AntiKt10 jet with P

T > 350 GeV and |η| < 2.0

● AntiKt4 jet with PT > 350 GeV and |η| < 2.5

● ΔR(AntiKt10 jet, AntiKt4 jet) > 2.0

● Trigger efficiency measured in data from period A with respect to a looser pre-scaled trigger● Single jet trigger with P

T > 280 GeV

● EF_j280_a4tchad● H

T trigger with H

T > 600 GeV

● EF_j170_a4tchad_ht600

37

Trigger and HT performance

● Trigger efficiency in data from period A● 96% at offline H

T = 800 GeV

● 99% at offline HT = 840 GeV

● 100% at offline HT = 900 GeV

● Trigger efficiency is at least 97% at HT = 850 GeV for 1.5 TeV or

heavier left- and right-handed W'

● Require HT at 850 GeV

trigger efficiency with respect to single jet trigger with P

T > 280 GeV

in data from period A

trigger efficiency with respect to H

T trigger with H

T > 600 GeV in

data from period A

ATLAS internal

ATLAS internal

38

Selection efficiency: 2 TeV W'

● Absolute and relative selection efficiencies for 2 TeV W'L (W'

R)

● b-tagging requirement on high PT small-R jet is the tightest

39

Selection efficiency: data

● 8 TeV data taken in 2012● b-tagging requirement on high P

T small-R jet is the tightest

40

Acceptance interpolation

● Cross sections interpolated● Acceptance interpolated● Allows to set limits on cross

sections for all masses between 1.5 TeV and 3 TeV

acceptance x efficiency2 b-tag

acceptance x efficnecy1 b-tag

left-handed hadronic W' → tb → qqbbcross section with g' = g

SM

cros

s se

ctio

n (p

b)

41

● Percent change of systematic uncertainties on event yield of 2 TeV W'

● b-tagging is the largest systematics in 2 b-tag category● Jet energy scale (JES) and jet energy resolution (JER) are small

Systematic evaluations

42

Top-tagging systematics

● Example from top-tagging systematics for W'L in 2 b-tag category

Change in event yields

43

JER systematics

● Rerun the analysis with 1 sigma shift in variable and obtain the invariant mass spectrum

● Parametrize the shifted spectrum● Compare the nominal spectrum with the shifted one to determine

nuisance parameters● Example from JER for W'

L in 2 b-tag category

Shift in the width of Gaussian

44

Treatment of systematics

● Uncertainties treated correlated in the combination of two channels● Spurious signal uncorrelated between channels

source yield asymmetric uncertainties

shape constraint

top-tagging efficiency yes yes no log-normalb-tagging efficiency yes yes no log-normal

c- and τ-tagging yes no no log-normal

mis-tagging yes no no log-normal

JER no no yes log-normal

JES yes no yes log-normalspurious signal yes no no Gaussian

luminosity yes no no log-normal

45

p-value of signal region fits

● Largest excess 1.4 σ local significance● Consistent with SM

46

Effect of including 1 btag channel

statistics onlyW'

L

statistics onlyW'

R

statistics + systematicsW'

L

statistics + systematicsW'

R

● Clear gain from adding 1 btag at high mW'

when systematics included

statistics onlyW'

L