Search for Pair Produced Stops Decaying to a - PowerPoint Presentation

Slide1

Search for Pair Produced Stops Decaying to a Dileptonic Final State at CMS

David Kolchmeyer

OverviewIntroduction

SUSY OverviewThe CMS ExperimentThe dileptonic stop search -

Event Selection

Data-driven

estimation of background (Tight to Loose method

)

My Summer Project

Potential Obstacles

Fun Slide

IntroductionMe: David Kolchmeyer, Rutgers University (New Brunswick, NJ)

My Advisor: Alberto Graziano, Instituto de

Física

de

Cantabria (

Santander,

Spain)

Working in a collaboration with UMD, IFCA, Oviedo, Cantabria,

CIEMAT

Part of CMS SUSY sub-group

I will be involved in offline physics analysis (work will mostly involve programming analysis code)

SUSY Overview

SUSY posits that each SM particle has a superpartnerFermions paired to bosons, bosons paired to fermionsCan solve Hierarchy/naturalness problem

Higgs mass is too low

Stops can cancel top quark loop corrections

The CMS Detector

Tracker surrounds collision point; charged particles leave tracks

Large magnetic field curves particles

ECAL absorbs photons/leptons

HCAL absorbs hadrons

Muon

detector detects long lived

muons

Jets are streams of particles created from individual quarks that emerge from the collision

Dileptonic Stop Decay

stop

top

W

b

lepton

neutrino

Neutralino

We consider the scenario:

Hence, we expect:

Two opposite sign leptons

Two b-tagged jets

Missing transverse energy (MET)

x 2 since the stop is pair-produced with an

antistop

Note: Largest background is from

ttbar

production, which has similar end-products

Event SelectionStandard

muon and electron selection criteria, determined by Physics Object Group (POG)Includes cuts on PT, eta, impact parameter, isolationLook for pair of opposite charged leptons in a

dilepton

triggered sample

If

ee

or

mumu

final state, veto in Z window (76 to 106

GeV

)

Require

m

ll

> 20

GeV

to eliminate low mass resonances

At least two jets, at least one b-jetMET > 40 GeV

in ee, mumu final states to further eliminate Z background Define signal region for MT2(

ll) > 80, control region MT2(ll) < 80Blinding policy

The MT2 variable

Data Driven Background EstimationBackgrounds include

ttbar, Drell-Yan, W+Jets

,

tW

, WW, ZZ, WZ, … (dominated by

ttbar

)

My focus: estimate

Wjets

background via data-driven method

Lepton from W and fake lepton from jet can contaminate signal region

‘Prompt’ lepton is from W/Z decay (the leptons we care about)

‘Fake’ lepton is from meson decay, actual fake from jet, photon conversion,

etc

(these contaminate our signal)

Tight to Loose Method

Define a tight lepton as a lepton candidate satisfying all selection criteriaDefine a loose lepton as a lepton candidate that satisfies all criteria except for the isolation criteria (instead it satisfies a looser isolation cut)

So a lepton can be categorized

as tight/prompt, loose/prompt, tight/fake, loose/fake

In a sample of fake leptons (i.e. in a QCD

dijet

sample) calculate the ratio of tight leptons to all leptons

 fake rate (f

)

Require jet that is separated from the lepton (“Away Jet”) to pass minimum energy threshold

With the fake rate, go back to the

dilepton

sample and observe the number of loose leptons to calculate the number of tight leptons

We can validate the estimation in a same-sign control region

Prompt rate (p) is ratio of tight to all among prompt leptons (measure in

Drell

-Yan sample using tag and probe), usually is close to 1

My Summer Project

Find the best definition of a loose electron and muonSee how estimated background varies depending on this definition

Determine best criteria of QCD sample to measure fake rate

Try different values of “Away

jet“ threshold

Validate the background prediction

Define control region, plot relevant variables

Compare with MC

Potential Obstacles

Trying to determine fake rate with samples from different triggers of different luminosities (how to combine trigger luminosities to weight leptons appropriately?)Electroweak contamination in fake rate estimation

Possibility of fake rate estimation being overly sensitive to definition of loose lepton

Not being able to properly predict backgrounds in control region or MC

Technical Difficulties

Jam Session

sur

le Bateau

Backup Slides

The MT2 variable

How can you measure masses of parent particles when daughter particles escape undetected?Consider W

l v

Note that

η

= ½ log

(

pseudorapidity

)

m

W

2

= m

l

2

+ m

v

2

+ 2 (

E

lT

EvT cosh (

Δη) – plT

* pvT)In our case, m

v2= 0 and Ev

T = |pv

T| = Emiss

T MT2(p

missT) = ml2 + 2 (ElT

EmissT – pl

T

*

p

miss

T

)

Note that MT is a lower bound of

m

W

The MT2 variable (contd)

Now consider two cases of W

l

v happening at once

We only know the sum

p

T

of the two neutrinos

So, let us try all the possibilities of dividing up

p

miss

T

into the

p

1

T

and p2T of each invisible neutrino, calculate MT twice to get the highest (best) lower bound for that possibility, and take the minimum over all the possibilities to get a lower bound on the W mass

The distribution should not cross the W mass (80 GeV)

The MT2 variable (contd)Formally, we can now define

MT2 = min { max { MT(p1T

),

MT(

p

2

T

) } }

for

p

1

T

+

p

2

T = pmiss

TFor ttbar (largest background), neutrinos are only source of MET, so MT2 should stay below 80 GeV

For dileptonic stops, neutralinos can add more MET, so MT2 distribution can exceed 80 GeV

Tight to Loose Method – 1 Lepton Example

Assume the only kinds of events are single electron events (as before, electrons can be fake/prompt and tight/loose)Define

ε

=

(tight over loose only for fakes)

Define

η

=

(loose only over tight for prompt)

Tight to Loose Method – 1 Lepton Example

Number of fake leptons identified as tight is ε times number of fake leptons identified as loose

Number of fake leptons identified as loose is number of loose leptons minus number of loose prompt leptons

Number of loose prompt leptons is number of tight prompt leptons times

η

Number of tight prompt leptons is number of tight leptons minus number of tight fake leptons

In symbols: LF = N

t0

–

η

[ N

t1

–

ε

[ LF ] ]

So fake tight =

ε

LF =

η N

t1 ] Use this to weight events with loose or tight leptons