Transcript 20120426_BhamAtlasWeekly_Diff

Update on Diffractive Dijet
Production Search
Hardeep Bansil
University of Birmingham
Birmingham ATLAS Weekly Meeting
26/04/2012
Contents
•
•
•
•
•
•
•
Theory & Motivation
Analysis
First look at systematics
CMS analysis & reconstructing diffractive variables
Forward gap distributions
Unusual acceptances
Next steps
2
Diffractive dijets
• Look for single diffractive events (pppX)
– Involve a rapidity gap due to colourless exchange
with vacuum quantum numbers: “pomeron”
• Search for diffraction with a hard scale set by 2
jets
– Described by diffractive PDFs + pQCD crosssections
• Previous measurements of hard diffractive
processes at HERA and Tevatron
– At Tevatron, ratio of yields of single diffractive to
inclusive dijets ≈ 1%
– Likely to be smaller than this at LHC
3
Motivation
• Understand the structure of the diffractive exchange by comparison
with predictions from electron-proton data and be able to get a
measure of FDjj
• Measure the ratio of the single diffractive to inclusive dijet events
• Gap Survival Probability – the chance of the gap between the intact
proton and diffractive system being lost due to scattering (affects
measured structure function)
– Tevatron have Gap Survival Probability of 0.1 relative to H1 predictions
– Khoze, Martin and Ryskin predict LHC to have GSP of ~ 0.03
Rescatter with p?
Comparison of Tevatron
results to H1 predictions
Gap destruction by
secondary scattering
ξ
4
Interesting variables
•
•
•
•
Calculate MX2 ≈ 2Ep·(E±pz)X  ξX = MX2 /s
Calculate zIP ≈ (E±pz)jj/(E±pz)X
Look at jet (η, ET, Mjj) and gap properties
Determine cross sections as a function of zIP
Mjj
Mx
ξX
5
Analysis
• Using L1Calo stream data10 period A and B ESDs
– Total ∫L dt = 7.22 nb-1 - calculated using online iLumiCalc tool with
L1_MBTS_2 live trigger and L1_J5 reference trigger
– Average <μ> for selected runs < 0.15
• POMWIG generator for Single Diffractive Dijets
– No rapidity gap destruction built in
– Generates QCD 22 process within diffractive system in different pT ranges
(8-17, 17-35, 35-70, 70+ GeV) for SD (system dissociating in ±z direction)
• Use PYTHIA 6 and PYTHIA 8 Dijet samples as Non Diffractive
(inclusive) Dijets
– Made in same pT ranges as Pomwig (8-17, 17-35, 35-70, 70-140 GeV)
• Anti-Kt jets with R=0.6 or R=0.4:
– Require >= 2 jets in event passing loose jet cleaning cuts
– ET Jet1,2 |η| < 4.4, ET Jet1 > 30 GeV, ET Jet2 > 20 GeV
– Jet ET & η cuts based on 2010 jet energy scale systematic and dijet analysis
• Ask for a forward gap: |ηstart| = 4.9, ΔηF ≥ 2.0
6
Gap Finding Algorithm
• Gap finding based on Soft Diffraction analysis
– Divides calorimeter into 49 rings of 0.2 in η
– Identifies calorimeter cells where energy significance (= cell energy/noise)
large enough that probability of noise cell studied in event is small
– Where no cells in ring found above ESig threshold  ring is ‘empty’
– Full details in blue box
• Determine the size of the biggest forward gap
Detector gap definition
•Calorimeter: no cell above threshold E/σ > Sth probability of noisy cell in ring smaller than 10-4
(electronic noise only, no pile-up environment),
cluster that cell came from has ET > 200 MeV
Example Single Diffractive Topology
ΔηF:3.4
|ηStart|:4.9
•Tracker: no good track above pT > 200 MeV, |η| <
2.5
Truth gap definition
•No stable particle above pT > 200 MeV
7
New Data Format
• Derived AOD format where clusters only have information for most
significant cell
– Cell sampling layer, cell significance
– No longer working directly with cells and ESDs
• Making NTUP_JETMET D3PDs from these
– Most of last 2 weeks went into getting code running
– Minor changes to gap algorithm for tracks and clusters
– Faster running time on Grid
• Currently short on statistics
(only 6M of 19M events
available across Periods A &
B due to production troubles)
8
Systematics
• Just a list at moment, not properly started!
• Jet Energy Scale and Jet Energy Resolution
– JESProvider and JERProvider software obtained for 2010 (need to set up)
• Gap Energy Scale (clusters)
• Gap Threshold Cut Uncertainty
– Gap related systematics based on Soft Diffraction analysis
•
•
•
•
•
•
MC Unfolding Uncertainty
Trigger
Tracking Uncertainty
Luminosity
Additional Material
Pile-up
• Are there others to consider?
• Focusing on the larger effects at the moment
9
CMS Diffractive Dijet Result
•
Analysis based on 2.7nb-1 of low-pileup data from early 2010
Trigger on single jets with pT>6 GeV
Require 2 jets with pT>20 GeV, |η|<4.4 based on Particle Flow
objects above noise threshold
Require a forward gap of 1.9 units in HF
•
•
Final cross-section extracted using the variable ξ (frac. p loss of scattered proton)
Sum over E and pZ of PF particles, with an MC-derived correction factor “C” = 1.45
•
Differential cross-section is determined from
a fit for the relative contribution of diffraction
(POMPYT) and non-diffraction (PYTHIA),
before gap cuts
•
•
•
~
10
Reconstructing ξ
•
Calculate Mx first using E±pz method then convert to ξ
•
Truth level: all final state particles excluding intact proton
(if there is one in event)
Reconstructed level: all caloClusters (not just above
significance thresholds)
•
•
Pomwig comparison (generator cut at ξ < 0.1)
Truth ξ v Truth ξ using E+Pz method
•
Truth ξ v Reconstructed ξ using E+Pz method
~
~
Harder to work out in data whether ξ+ or ξ- correct one to choose
11
Determining C
•
Determine C by comparing MC truth and reconstruction
•
Truth level: all final state particles excluding intact proton
(if there is one in event)
Reconstructed level: all caloClusters (not just above
significance thresholds)
•
•
Pomwig comparison (generator cut at ξ < 0.1) of ξ to gap size
Truth ξ v Truth Gap Size
•
Reconstructed ξ v Reconstructed Gap Size
Enough to divide gradients from two different plots? Very different for other MCs
12
Determining C
•
Determine C by comparing MC truth and reconstruction
•
Truth level: all final state particles excluding intact proton
(if there is one in event)
Reconstructed level: all caloClusters (not just above
significance thresholds)
•
•
Pythia 6 comparison of ξ to gap size
Truth ξ v Truth Gap Size
•
Reconstructed ξ v Reconstructed Gap Size
Enough to divide gradients from two different plots? Very different for other MCs
13
L1_J5 Efficiency
•
Look at L1_J5 efficiency on a run by run basis using MinBias stream and MBTS_2
trigger to gain independent sample
–
Will make this J5 & FJ5 (FJ3 in earlier runs)
–
Account for endcaps where RoI could have formed in FCAL using Δφ < 0.4
•
Ask that jet matches to Jet RoI in central region with ΔR < 0.4
•
•
Well aware of timing issues for early runs  improvement in later runs
Use MC combined jet results  account for MC to data comparison as systematic
η dependence
from EM TR
AntiKt4 Jets
AntiKt6 Jets
AntiKt4 jets
high efficiency
compared
to AntiKt6 jets
L1Calo Data AntiKt4 & AntiKt6 Jets
L1Calo Data AntiKt4 Jets (ET > 30 GeV)
•
Factor in to data so N data goes to N data  1 ( E trig )
T
•
Ignore variation in η and take correction factor only in ET
14
L1_J5 Efficiency – Data v MC
•
Pythia 6, Pythia 8 & Pomwig less affected by timing issues
– Around 30 GeV efficiencies are similar – Pythia 6 just about the best match
– MC reaches plateau faster (smaller correction factors) due to TR
L1Calo Data AntiKt4 Jets (ET > 30 GeV)
L1Calo Data AntiKt6 Jets (ET > 30 GeV)
Pythia 6 AntiKt4 Jets (ET > 30 GeV)
Pythia 6 AntiKt6 Jets (ET > 30 GeV)
15
Uncorrected Gap Size Distribution
POMWIG SD Jets,
PYTHIA 6 & 8 Jets
weighted relative to
luminosity of data
runs used and then
plotted against
L1Calo Data
L1Calo Data
Pomwig SD Jets
Pythia 8 Jets
Pythia 6 Jets
Drop in number of
events with ΔηF ≥ 6,
cuts into phase space
Biggest ND
contribution at
small ΔηF
•
•
Data agrees much better with Pythia distribution (best match Py 8 before ΔηF ≥ 5)
Ratio of MC to Data suggests a GSP of 0.3-0.4 in majority of bins
•
By ΔηF of 6, ξ = 10-4.5  MX = 39.4 GeV – cut out phase space for producing pair
of 20 GeV jets so get drop in events after this point
•
Dijet samples lacking statistics at higher gap sizes with new gap algorithm
– Still observe forward gaps with sizes of 4, 5, 6 in PYTHIA 6/8 but errors large
16
Differential Cross Sections
• MC samples weighted to lumi of data runs - Differential cross section
as a function of forward gap size
• Some odd bins due to acceptance but otherwise time to look at
systematics
L1Calo Data
Pomwig SD Jets
Pythia 8 Jets
Pythia 6 Jets
Combined Acceptance
Weights applied to different pT range
samples based to match to lumi of
data then samples combined
Migrations can then cause the
acceptances to be larger or smaller
than expected
Pomwig SD
Pythia 8 Jets
Pythia 6 Jets
17
Acceptances
• Combined acceptances look odd due to low stats & weights
– Look at acceptances for different samples
• Where stats become small, typically get more events passing cuts at
reconstruction level rather than truth level in samples
Combined
After Weights
Pythia 8
Pomwig
Pythia 6
18
Large Acceptances?
• Where stats become small, typically get more events passing cuts at
reconstruction level rather than truth level in samples
• Two major causes affecting large acceptances at ΔηF >4 so far:
• Far more events where 2 or more truth jets seen as 1 recon jet
– Seen in all MC samples
– Put into acceptance definition that must be 2 jets in recon & truth
• Differences in gap size between recon & truth
– More often the recon gap size is larger
– Some cases where Δ(ΔηF)
small, most where Δ(ΔηF)
large – greater than 2
• Improve with larger stats
or cut out odd events
Pythia 6 Jets
ΔηF Resolution = ΔηF Truth – ΔηF Recon
19
Next steps
• Decide on overall steps for analysis between Birmingham
and Prague
• Get official production of data & MC samples in new format
•
•
•
•
•
– Estimate sizes of MC samples necessary for analysis
Continue to develop systematics
Make ξ useable in data
Work out ways to reduce acceptances to smaller levels
Look into potential backgrounds
Work out relative amounts of diffraction and non-diffraction
present in data
20