Transcript qcdbkg_intro

Introduction to the workshop
on QCD background to W
Pierre-Hugues Beauchemin
University of Oxford
Workshop on QCD background to W, June 14th, 2010
Objectives of this workshop
•
Premises:
– QCD background to W or W+jets cross section measurements is one of the largest
sources of systematic uncertainty
– Many people worked, is working or plan to work on QCD background estimates
 Need dedicated and organised efforts
•
The workshop intends to gather the different contributions on the topic to:
– Review the status of different background estimate strategy
– Exchange experience with MC and data
– Bring the experts together to discuss improvements
– Compare results
•
At the end of the workshop, we aim to be able to:
– Concrete strategy to estimate QCD background and uncertainty
– Set a list of missing elements
– Share the tasks among the various analyzers
► Keep in mind: the focus is 1 pb-1 W (+jets) cross section measurement(s)
2
Subjects to be covered today
•
Background expectations from MC and first data
•
Samples and tools available and projected
•
Understanding the behaviour of such background
– Shape studies
– Different modelling of the background
•
Data-driven strategies to estimate the QCD background and uncertainties
– Various template approaches
– Met vs Iso method
– Estimate using TRT information
•
Studies specific to muon channel
•
Trigger issues
•
Discussion of the plans for first cross section measurements
•
Combination of various predictions
3
Few Monte Carlo numbers for Wen
•
Consider the following MC samples:
mc09_7TeV.105802.JF17_pythia_jet_filter.merge.AOD.e505_s765_s767_r1215_r1210/
mc09_7TeV.106043.PythiaWenu_no_filter.recon.AOD.e468_s765_s767_r1215/
•
Only 180 JF17 events survive electron candidate selections
– There is a parton of at least 17 GeV
– There is a filter to preselect electron-like jets (e = 0.086)
– Sample contains ~10M events after filter
•
Numbers from
J-B Blanchard
QCD background to Wen events is ~ 35% in this study
– Really low statistics of QCD background events
– fQCD = NQCD / (NQCD + NW )
process
Wen
QCD
LO s
(nb)
L
(nb-1)
Axe
Nevent
8.9
1000.
0.38
3380
1.1E6
1000.
Relative e (Wen)
(%)
Relative e (QCD)
(%)
78.5
7.2
Eta cut
92
92
author
99
97
Cluster ET
77
8.7
Medium ID
89
1.9
ETmiss
81
2.9
Isolation
96
53
Total
38
1.7E-4
Selection
1.7E-6
1870
EM3
4
Needs for data-driven estimates
•
QCD multijet contribution to W samples come from the cancellation of two big
numbers:
– QCD multijet cross section is ~5 order of magnitude higher than Wen cross section
– High rejection factors from Wen events selections
•
•
•
•
Probability of ~7/100 to pass EM3 trigger
Extra probability of ~1/1000 of having a medium electron with ET > 25 GeV
Extra probability of ~3/100 of having ETmiss > 25 GeV
With 1pb-1 of data, we expect:
– O(3000) Wen events  ~2% statistical uncertainty
– fQCD~ 10%
•
 ~20% on the QCD background prediction
In order to have such a precise prediction of QCD background from MC, we need
– High precision on fragmentation and hadronization models (O(<1%))
– High precision on description of interactions with matter and in shower model (O(<0.1%))
– Good description of hadronic activity (O(<1%))
– Precise estimate of the cross section
 Such precision is not achievable in MC
 Need data-driven estimate to hope to get such an accuracy on the prediction
5
Monte Carlo studies
•
Despite the need for data-driven estimate of the QCD background to Wen,
MC studies are important
– Useful to develop background estimate methods
– Give good idea of the level of correlation between variables
– Allow to develop tests to estimate systematic uncertainties
•
In the following:
– Show example from MC of data-driven estimates
– Relative effects of Wen event selections on QCD background contamination
– Breakdown of the origin of the QCD background
– Expectations for various jet multiplicities
6
Example of data-driven estimate
From J. Hartert, K. Lohwasser,
and L. Weithofer
•
QCD bkg template obtained from data by
reverting some of the standard IsEM cuts
•
Data = fsig x sig + sbkg x bkg
•
Fit in region where #sig ~ #bkg
•
QCD =  > 25 f(bkg) dETmiss
Dedicated session to this topic!
7
Examples on data
•
First studies on 7 TeV data-driven (6.69 nb-1):
 Methods look promising!!!
From J. Hartert, K. Lohwasser,
and L. Weithofer
More on that in later sessions!
8
Further QCD rejection?
•
Table in slide 4 presents the effect of cut flow on QCD multijet contamination to Wen
– Best cuts are: EM3, Cluster ET > 20 GeV, IsEM = medium, ETmiss > 25 GeV
Can we cut harder???
Df(lept – ETmiss)
–
Interesting potential for 0-jet
–
Less relevant in W+jets
Balance between ETmiss and leading jet PT
 Potentially interesting too
Black = QCD ALPGEN mc08
Red = W+jets ALPGEN mc08
From M. Dundord
Df(lept – ETmiss)
9
Origin of the background
•
Classification of electron candidates:
– Hadrons (27%)
– Electron from photon conversion (50%)
•
•
•
QED ISR/FSR
Bremsstrahlung
Pi0 decay
– Electron from heavy flavor decay (23%)
M. Dundord, I. Jen-Laplante
Medium electron candidate
All medium tracking cut pass
Some calo cuts fail
Reversing cuts change the bkg
composition a little bit
10
QCD background to W+jets
From S. Whitehead
•
QCD background relatively small in inclusive W but
significantly increases with jet multiplicity
 Precise estimate even more crucial for W+jets
•
Statistics too low  relaxe ETmiss cut and study
trends
– Only 41 events with ≥ 1-jet
•
No significant dependence on the PT of recoiling
jets
11
Generator dependence
From M. Dundord
•
Reasonable agreement between
Pythia FJ17 and ALPGEN QCD
samples
– Normalization issue in 0-jet bin
– ALPGEN have better stats in
high multiplicity
Black = QCD ALPGEN mc08
Red = W+jets ALPGEN mc08
12
A first look at data
From P. Malecki
•
ETmiss description not good in MC, but small
amount of background
 Relatively good data-MC agreement for
isolation and MT distribution
•
ETmiss template not ideal
– Need to use MC for signal
•
Need to renormalize the background
prediction by a factor of 0.45
13
Today’s Meeting
•
There are many speakers and contributions:
– Need to leave enough time for discussion (purpose of meeting)
 Please keep up to 15min!!!
•
Lunch will be from 13h to14h
– Will start at 14h sharp in theory conference room (4-03-006)
•
Morning session:
– Monte Carlo samples
– Shape studies
– Data-driven methods: template methods
•
Afternoon session:
–
–
–
–
TRT method
Met vs Iso method
Trigger issues
Summary
14