Standard Model backgrounds to h®gg are :

- Z/g*®ee

- Wg (W®en)

- QCD containing photons and/or jets misidentified as EM objects: gg , gj,  jj

 

ee,Wg and gg are estimated from Monte Carlo while gj and jj are estimated from data.

 

 

Estimating ee,Wg ,gg  backgrounds from Monte Carlo

 

For each of the three processes I used MC samples to count number of events with at least 2 loose EMcandidates and to obtain mass distributions and then scaled those taking into account cross-sections of individual processes, integrated luminosity of di-EM data sample, trigger and offline EMID efficiencies.                                                                          

 

In an attempt to reproduce the width of the Z-peak as it appears in the data the energies of loose EMcandidates are smeared according to the data-derived energy resolution described at the very bottom of this page

                                                                                                                                       

This table contains description of the samples,

values of the factors that were used for scaling and event counts at different stages.

 

 

 
Estimating gj+jj background from data

 

The idea is to estimate gj+jj background contribution by using “EM+jets” data sample of known luminosity and EMFakeRate obtained with an independent measurement.                                                                                                    

 

This method can be used if EM-Rate » EM-Fake-Rate, i.e.

ratio of EM clusters (coming from both photons and misidentified jets) to jets » ratio of EM clusters coming from misidentified jets to jets

 

I’ll try to show in a crude way that for current measured EM-Rates (>= 4*10^-3)

this is indeed the case

 

I assume LO cross-sections ( s(gj)=1.4*10^4 pb ;   s(jj)=3.5*10^7 pb)

and consider objects above 20GeV

Then

EM-rate = NofEMclusters/NofJets

              = [N_of_”EM+Jets”_events]/ [N_of_multijet_events*<NofJets>]

  = [s(gj)*e + s(jj)*2*EM-Fake-Rate]*Lint/ ([s(jj)*<NofJets>]*Lint)    ( where e=EM efficiency )

              = s(gj) *e /s(jj) /<NofJets> + 2*EM-Fake-Rate/<NofJets>

              = 0.4*10^-3*e/<NofJets> + 2*EM-Fake-Rate/<NofJets>

 

from this equation for EM-rate=0.4*10^-3,  e=1 and  2 < (<NofJets>) <3

I obtain |EM-Rate - EM-Fake-Rate|/EM-Rate < 10%

 

Having shown that EM-Rate » EM-Fake-Rate I’ll try to explain how gj+jj background

can be derived from “EM+jets” sample:

 

     N_of_”EM+Jets”_events = [s(gj) *e + s(jj)*2*EM-Fake-Rate]*Lint(”EM+Jets”)

                                              » [s(gj) *e + s(jj)*2*EM -Rate]*Lint(”EM+Jets”)

 

Ţ [s(gj) *e + s(jj)*2*EM-Rate] = N_of_”EM+Jets”_events/ Lint(”EM+Jets”) (1)

 

whereas for di-EM events due to gj+jj:

 

N_of_”di-EM”_events 

               due to gj + jj  = [s(gj) *e *EM-Fake-Rate + s(jj)* EM-Fake-Rate^2]*Lint(”di-EM”)

                                     » [s(gj) *e *EM-Rate + s(jj)* EM-Rate^2]*Lint(”di-EM”)

                                     = EM-Rate* [s(gj) *e  + s(jj)* EM-Rate]*Lint(”di-EM”)

                                     = (1/2)*EM-Rate* [2*s(gj) *e  + s(jj)* 2*EM-Rate]*Lint(”di-EM”)

                                     = (1/2)*EM-Rate* [s(gj) *e +s(gj) *e  + s(jj)* 2*EM-Rate]*Lint(”di-EM”)

 

  { in the same way as in EM-Rate » EM-Fake-Rate argument can show that

               s(gj) *e/(s(jj)* 2*EM-Rate) <10% , then approximate by dropping s(gj) *e term  }

 

 = (1/2)*EM-Rate* [s(gj) *e  + s(jj)* 2*EM-Rate]*Lint(”di-EM”)

now use equation (1)

 = (1/2)*EM-Rate* Lint(”di-EM”)* N_of_”EM+Jets”_events/ Lint(”EM+Jets”)

 

This is final equation : scaling factor is (1/2)*EM-Rate* Lint(”di-EM”)* / Lint(”EM+Jets”) 

 

So I selected “EM+jets” sample from single good run 149330 requiring

-        “EM_HI” trigger pass

-        at least one EMcandidate

-        at least one Jet not matched to EMcandidate

 

and calculated luminosity using lm_access package (0.051 pb^-1)

Number of thus estimated gj + jj  events was : 57 (CCCC),  67 (CCEC)