Sensitivity Study on Top Quark Pair Resonance Search

Tatsuma Meguro

University of Tsukuba

Sensitivity Study on Top Quark Pair

Resonance Search

motivation

● The large mass of top quark could be sensitive to the physics beyond the standard model.● Many theories predict new particles preferentially coupled to the 3rd family, in particular top quark

● Model independent search is desirable.

Motivation 2•Current interest is in the high mass tail region (signal model can be anything)

• There are few background in the interested region (700~1000GeV) .• We can’t tell whether there is interference effect or not because of small data size.• Analysis is a little bit simpler without interference case.

•increase the signal acceptance• include pretag events

[~100%: assuming tag/jet=30% at high mass region]

• include >=5jet events [~30%]

(jet assignment become difficult)• include more lepton category [~20%]

(don’t need not-good quality lepton with very high pt)

• Separate events into different sensitivity categories• By tagging information

Event Selection (standard selection)

ttbar invariant mass reconstruction

Background

Process Pretag Tagged

QCD 169.93+-62.26 13.62+-11.68

ww 30.85+-2.07 2.53+-0.27

wz 5.87+-0.40 0.94+-0.08

zz 0.50+-0.04 0.09+-0.01

top 409.87+-55.05 48.23+-15.57

s-channel single top 3.33+-0.31 2.00+-0.20

t-channel single top 3.41+-0.27 1.94+-0.17

Z->l l 35.31+-3.17 2.39+-0.28

HF(Wbb,Wcc,Wc) 137.19+-31.32

LF(W+np) 425.74+-79.21 16.65+-3.64

Total Background 1222.00+-0.00 117.80+-19.38

Wbb(16.53+-6.66)

Wcc,Wc(12.88+-5.23)

<The background estimation>

Ex) Pretag LF background

from Method 2 tableUse Official MC sample of each process

0 parton + W(->e nu)

1 parton + W(->e nu)

2 parton + W(->e nu)

3 parton + W(->e nu)

4 parton + W(->e nu)

0 parton + W(->mu nu)

1 parton + W(->mu nu)

2 parton + W(->mu nu)

3 parton + W(->mu nu)

4 parton + W(->mu nu)

1,Apply MDL for each MC sample2,Calculate Acc.*Xs*Br and merge samples in this ratio3,Normalize according to method 2 table4,we can get background shape (ex: pretag LF background(below) )

Signal template

sqrt(s)

Neve black:Hepgred:Reconstructed

400GeV 600GeV

1000GeV

1200GeV

800GeV

Signal : Z' -> ttbar (width=1 GeV)

Old sampleodd tail in low mass region.I guess this comes from p.d.f distribution.

ttbar invariant mass distributiontaddeg sample pretag sample

mass(GeV)mass(GeV)

rough estimation

<Significance(pretag)/Significance(tag)>

Z' mass (GeV)

How to estimate:1. Count events in each Z mass window.2. Calculate significance = S/sqrt(B) pretag , tagged events separately. 3. Take ratio to see how much sensitivity improve by adding pretag events.

Low mass region ~ 30%High mass region ~ 40% !!

Likelihood Fit

Signal : qqbar -> Z' -> ttbarBackground : others(10~11 shapes)

mu : Number of expectation in i th bin (= Nbkg + Nsig(Xs x Br x Lumi. x Acc.) )ni : Number of pseudo data in i th binG(Nbkg,) for background constrainG(Nsig,) for signal systematics

We can retrieve 95% C.L. upper limit from L(x) plot

G:Gaussian

Cross section

Lik

elih

ood 95%CL

hMtt_ttop75 hMtt_zpr500

95%CL distribution

Cross section

Lik

elih

ood

normalized BG(left) , signal(right) distributionGeV

1, Make distribution from Pseudo experiment (w/o Z' event , Ndata = 500 )

2,Calculate the likelihood for Xs of Z' and set 95%C.L upper limit

Templete

95%CL

we can get one point from 1-pseudo experiment

How To Set Limit

Expected 95% C.L upper limit

cross section

Mz' = 600GeV

cross section

Mz' = 700GeV

cross section

Mz' = 400GeV

cross section

Mz' = 500GeV

cross section

Mz' = 800GeV

cross section

Mz' = 900GeV Mz' (

Z’)

BR

(Z’

ttba

r)

95% upper limit

Estimation of systematic uncertainty(JES)●JES●top mass●PDF●LO/NLO●hadronization(generator)●ISR/FSR●b-tag Efficiency

Suppose JES is dominant, I used only the uncertainty from JES in this analysis.

black:defaultred:JES + sigma_JESblue:JES - sigma_JES

Mttbar

Mttbar

Apply to G(Nbkg,sigma_bkg)

summary●made Signal MC●reconstruct ttbar inbariant mass (signal and background)●estimate 95% upper limit Z' cross section for each mass region , pretag ,tag event separately.

●Add pretag events to see how much improve the sensitivity... XX% Improvement in low mass region (500GeV) XX% Improvement in high mass region (1000GeV)●Other models will indicate same trend for sensitivity search

backup

rough estimate

Total pretag events : 851 Total tagged events : 371 (from method 2 table)

Sensitivity = Signal/sqrt(Background)

pretag : Nsig / sqrt(851) ~ N*0.6 / 29.17tagged : Nsig' / sqrt(371) ~ N*0.4 / 19.26

-> about 40% improvement in rough estimate!!

signal template350(GeV) 400 450 500 550

600 650 700 750 800

850 900 950 1000 1050

1100 1150 1200

sqrt(s)

Neve red:Ntag=0blue:Ntag>1

Identical Matched Jet (Old sample)

(Hepg-Rec)/HepgReconstructed

These plots are normalized

28762074

33742440

31802427

31722336

30292205

29332061

29932201

30102307

30462346

31692363

31262385

33142493

24921941

26061964

25151964

26352059

25591984

29002065

Red Identical JetBlue Not Identical

Problem of Pythia

sqrt(shat) GeV

num

ber

of e

vent

influence from p.d.f ?

So, I am planning to make new Z' template (Z' width = 1 GeV)

default width width = 1 Gev