Dijet Resonances (Update) Kazım Z. Gümüş and Nural Akchurin Texas Tech University

1

Dijet Resonances (Update)

Kazım Z. Gümüş and Nural Akchurin

Texas Tech University

Selda Esen and Robert M. Harris

Fermilab

SUSY/BSM Meeting

September 06, 2005

Kazim Gumus, Texas Tech University 2

Introduction

Search:

We search for narrow resonances decaying to dijets:

pp X Jet Jet (inclusive)

We use Zprime as a model for dijet resonances.

Motivation:

There are lots of new resonances which decay to dijets. (See next slide)

LHC is a parton parton collider. Resonances made from partons will decay to partons, giving jets.

Update:This update includes estimates of CMS capability with 1 fb-1 to

o Exclude dijet resonances at 95% CL

o Discover dijet resonances with 5 sigma significance

3

Review: Cross Section for Dijet Resonances

Resonances produced with strong force, or from valence quarks in proton, have

highest cross sections.


Jet Reconstruction & Correction

Iterative cone jet algorithm with R=0.5 and E scheme.Generated Jets: particles in the jet cone, excluding pileup.Reconstructed Jets: EcalPlusHcalTowers with ET>500 MeV.Corrected Jets: Reconstructed jets with jetCalibV1 applied.Correction back to particles in jet cone before pileup.

Event Selection

Find the two jets in the event with highest PT: leading jets.Require each leading jet have |eta|<1. Enhances sensitivity to new physics which is produced at low |eta|. Also, Ecal end caps may not be there, or understood, on day 1.Dijet mass: M = sqrt( (E1+E2)2 - (px1+px2

)2 – (py1+py2)2 – (pz1+pz2)2 ).

Cross Section

Bin the dijet mass in ~10% wide bins: bin size increases with mass.Divide rate by the luminosity and bin width: differential cross section

Review: Analysis

(from “New Physics with Dijet” talk given by R.Harris at HCAL/JetMet Workshop, 11/12/2004)


Review: Cross section for QCD and Z’ Signals

(|eta|<1 )

(Lum=1fb-1)


We are constructing the likelihood for a data sample resulting from a QCD background and a resonance signal as a function of the resonance signal cross section. Briefly;

ni : Observed events (We consider first the case where observed events are QCD only and next the case where observed events are QCD plus a mass resonance. )

Nsignali : Predicted signal events in mass bin i

Nqcdi : Predicted qcd background events in mass bin i

Let the signal to be multiplied by an unknown parameter alpha and add in the background to obtain the predicted number of events in the mass bin i:

i = alpha * Nsignali + Nqcdi

Probability Pi of observing ni events when i are predicted is given by Poisson statistics

Pi = ini * exp( - i ) / factorial( ni )

The product of Pi over all bins in the mass spectrum is the likelihood function L for seeing the observation given the prediction

L = P1 * P2 * .....Pn

Beginning of update: Binned Likelihood Calculation


QCD Cross Section and Fit

21 /)1(*0pp m

s

mp

dm

d

We fit the QCD background to a smooth parameterization to remove fluctuations that would distort our likelihood. We are smoothing the background.


Likelihood Distributions (No Signal)

When there is no signal, the likelihood peaks at zero cross section for the signal. The 95% area point gives us the limit.


Method:

To produce 5 sigma discovery cross section, we formed a new distribution of "signal + background“.

We found the likelihood for this distribution.

We increased the signal cross section until we had a gaussian likelihood that was above zero cross section by 5 sigma (conservative approach).

The result was very close to a 5 sigma exclusion from the likelihood distribution without signal, as it should be for large statistics."

5 sigma discovery cross section


Likelihood Distributions (Signal)

The likelihoods are gaussian and the most likely cross section is 5 sigma above 0.


Resonance Bumps at 95 %CL and 5 Sigma Discovery

The size of the resonances for a 95% CL exclusion and a 5 sigma discovery is compared with the size of the statistical error bars from QCD.

Cross sections appear to be conservative estimates of what we can exclude or discover (but no systematics!).


95%C.L. Cross Section Exclusion and 5 Sigma Discovery Cross section compared to cross section predictions

With 1 fb-1 we can exclude at 95% CL all models above the dotted line.

With 1 fb-1 we can discover at 5 sigma significance all models above the solid line

We need to do estimates for more resonance points, and make the line into a realistic curve. We also need systematics.


Mass Limits (95% CL) and Discovery Range (5)

We will not be sensitive to Zprime, Wprime and RS Graviton with only 1 fb-1.

Object 95% CL Exclusion Range

( 1 fb-1, stat err only)

5 Discovery Range

(1 fb-1, stat err only)

Excited Quarks up to ~ 4.5 TeV up to ~3.5 TeV

Axigluons up to ~4.5 TeV up to ~3.5 TeV

E6 diquark ~0.7 TeV to ~ 5 TeV ~1.5 TeV to 4 TeV


Conclusions and Future Plans

• We have made a first estimate of CMS ability with 1 fb-1 to exclude (at 95% CL) or discover (at 5) resonances in the dijet mass distribution.

• Although the 5 sigma discovery plot has three points for the moment, It gives us an idea about the cross section limits we can bring in search of new resonances.

• We are working on interpolating more points.

• Selda is preparing a new qcd dataset which we will have binning corresponding to the dijet mass resolution.

• Plan to use the new jet trigger table prepared by Robert and Selda.

• Plan to include the first estimates of systematic uncertainties.


BACKUP SLIDEResidual Plot

We test the goodness of our fit to the qcd data.

Dijet Resonances (Update) Kazım Z. Gümüş and Nural Akchurin Texas Tech University

Documents