Measurement of tt̄ → bb̄`τ cross-sectionin 13 TeV CMS data

Oleksii Toldaiev

supervised by Michele Galinaro and Joao VarelaLIP, CMS

25 May 2017

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 1 / 13

Motivation

Cross-section measurements intt̄ channels at 8 TeV CMSdata from 2012.

Measurement in tt̄ → bb̄lτ channelDevelopment of simultaneous likelihood fitwith improved uncertainty in estimation ofmain background

It serves as preliminary work for furthermeasurements in similar final states

E.g. improvement of selection of tauleptons based on characteristics of taudecay (main systematic error)

Yields a better measurement of ratio ofbranching fractions of W decay (tau VSother leptons) and puts limits on BSMprocesses

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 2 / 13

CMS detector, 2016 dataset

The luminosity for 2016. 36.8fb−1 is certified now (about 3.8fb−1 in 2015, 21.8fb−1 in2012 at 8TeV ).

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 3 / 13

Features of tt̄ → bb̄`τ channel, measurement methodThe decay has many particular finalproducts:

1 lepton (muon or electron, adistinct signal in the detector)

2 b-jets (displaced vertex of thejet)

2 neutrinos (high amount ofmissing ET )

1 hadronicaly decaying tau lepton(a narrow jet with distinctsignature)

A “cut and count” approach can be em-ployed on identification discriminants ofall final products to select a clean sampleof events.The main background comes from faketaus in tt̄ → lj channel. It is constrainedwith measurement of jet-tau fake rate incontrol regions.

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 4 / 13

Reconstruction algorithms

Particle Flow algorithm (PF) reconstructs products of a collision:muons, electrons, charged and neutral hadrons, photons.

Jets are reconstructed with anti-Kt clustering algorithm with conesize 0.4.

Combined Secondary Vertex MVA-based algorithm is used forb-tagging.

Additional requirements on characteristics of reconstructed electrons,muons, jets, b-jets are set to reduce mis-identification.

The objects are corrected according to studies of detectorperformance. The corrections are propagated to calculation ofMissing Transverse Energy in the event (MET).

Other characteristics of the event are taken into account in eventselection (state of the detector, presence of cosmic rays or other noisein systems) or in comparison of real data-taking conditions tosimulated data (Pile-UP, luminosity).

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 5 / 13

Reconstruction algorithms: tau identification

Taus decay into other leptons in one third of the cases or into hadrons in the restremaining cases. Hadronically decaying taus are reconstructed by the HadronPlus Strips algorithm. The major challenge is the mis-identification of jets astaus.Tau identification discriminators provide about 70% efficiency at 0.3%mis-identification.The study uses inclusively all hadronic decays of taus, without specificrequirements on constituents of the decay.

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 6 / 13

Cut and count method

Cuts on products of lepton-tau final state:

1 isolated lepton

3 or more jets

Missing ET > 40GeV

1 or more b-tagged jets

1 or more taus

opposite sign of tau and lepton

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 7 / 13

Event yields in tt̄ → bb̄`τ final stateComparison of data and simulated events in lepton+tau final state.

N eventsSource eτ µτ

DY+jets 203 318QCD 0 20

Single Top 580 902W+jets 130 168tt̄other 2802 4674

tt̄(signal) 6076 9963mcsum 9792 16045

data 9951 15355ratio 0.984 1.045

B-jet pT and eta. Tau pT and eta.

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 8 / 13

The study of tau mis-identification

Jet to tau fake rate is measured as fraction of jets matched to reconstructed taus(δR < 0.3) to all jets, in bins of pT , η, jet radius:

f .r .(pT , η,R) =jets ∩ taus

jets

∣∣∣jetpT ,η,R

Measurements in 4 different data regions yield these distributions:

3 control regions for measurement of the jet-to-tau fake rates are needed due tosignificant difference between fake rates of gluon jets, light-quark jets and b jets.The study is done in QCD, W+jets, µµ+jets, ee+jets and eµ+jets regions. QCD, µµand eµ are choisen for the estimation due to good statistics and negligible amount oftrue taus.

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 9 / 13

Data-driven estimation of mis-identified taus

The fake rate in the main selection is estimated from measured fake rates withcombination fixing the jet composition.

fractions of jets in control and signal samples

Sample b jets light-quark jets gluon jetsqcd 0.039 0.345 0.599

mumu 0.130 0.435 0.420elmu 0.535 0.184 0.274eτ 0.397 0.345 0.234µτ 0.407 0.339 0.236

weights for the measured fake rates, estimation of fake taus

Sample eτ weights µτ weightsqcd 0.0981 0.0981

mumu 0.3446 0.2755elmu 0.5573 0.6265

Selection eτ (% of data) µτ (% of data)tau 4250± 798 (7.2%) 6232± 1143 (6.5%)

opposite sign 2975± 558 (5.8%) 4431± 812 (5.3%)Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 10 / 13

Systematic errors

Uncertainty in corrections, scale factors and efficiency of ID requirements.

Uncertainty %Source eτ (MC) µτTau ID 5 5

Tau energy scale 3 3Tau fakes 6 6

pile-up 1.9 1.9JES 0.3 0.3JER 0.1 0.1

b-tagging 1.6 1.6Top pT 0.1 0.1

Other backgrounds 30% variation 2.3 2.4Other backgrounds stat. variation 0.3 0.3

Total 9 9Luminosity 2.6 2.6

Statistics 1 0.8

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 11 / 13

Current results with estimation of fake taus

N events (±stat)Source eτ µτ

DY+jets 181± 24 310± 34fake taus (from data) 2975± 558 4431± 812

Single Top 553± 8 879± 12tt̄other 462± 11 834± 17

tt̄ (signal) 5806± 42 9709± 59mcsum 9979± 560 16161± 815

data 9951± 99 15355± 123ratio 1.003± 0.057 1.052± 0.051

σtt̄ =data− bkgMC − bkgfakes

L× acceptance × BRW→lτResulting cross-sections (acceptance is measured in MC):

σeτ = 825± 7.8(stat) ± 70(sys) ± 20(lumi)pbσµτ = 762± 6(stat) ± 67.5(sys) ± 20(lumi)pb

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 12 / 13

Prospects

Measured tt̄ cross-section in final states with taus

In progress: measurement with simultaneous likelihood fit in dataregions with separated signal and background channels

Study on improvement of selection of tau leptons for precisemeasurement of ratio of W branching fractions

Limit on BSM processes with similar final state

Other measurements in tt̄ → lτ final state

Oleksii Toldaiev (supervised by Michele Galinaro and Joao VarelaLIP, CMS)tt̄ → `τ cross-section 25 May 2017 13 / 13

Prereview