Transcript 20110519_BhamAtlasWeekly_Diff

Searching for Diffractive Dijets
Hardeep Bansil
University of Birmingham
Birmingham ATLAS Weekly Meeting
19/05/2011
Contents
•
•
•
•
•
Theory & Motivation
MC Generators
Analysis
Plots
Next steps
2
Diffractive dijets
• A mix of single diffractive events
(with rapidity gap due to colour
singlet exchange – “pomeron”)
• With dijet events
• To get diffractive dijet events
– Hard diffraction
– Two high pT jets amongst other hadronic
activity + gap on one side
• Shows proton has DPDF (diffractive parton
density function)
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
• Gap Survival Probability – the chance of the gap between the
intact proton and diffractive system being lost due to
scattering (affects measured structure function e.g. Tevatron
results a factor of 10 smaller than H1 predictions)
Rescatter
with p?
(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
Meaning of E±pz
6
zIP & xP reconstruction method
• Based on E±pz method, calculate xP and zIP using jets and
calorimeter clusters on the correct side of the gap
• If a proton dissociates to +z (C side of ATLAS)
zIP = (E+pz)jj/(E+pz)clus
xP = (E-pz)jj/(E-pz)clus
• If a proton dissociates to -z (A side of ATLAS)
zIP = (E-pz)jj/(E-pz)clus
xP = (E+pz)jj/(E+pz)clus
7
Monte Carlo Generators
• Currently using Pomwig LO generator - modifies Herwig
ee+γ to recreate pp+IP with CTEQ H1 pomeron flux
• No rapidity
gap survival
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) + DD
• Only available files on Grid have √s = 10 TeV and old
reconstruction so generated new MC samples (1000 events of
each) of as well as in a new pT range (5-8 GeV)
– Event generation: AP-15.6.13.9 (MC10JobOptions)
– Simulation: AP-15.6.13.9
– Reconstruction: AP-16.0.3.5
• Will need to get official Monte Carlo production done soon
8
Monte Carlo Generators
• Pomwig – scattered parton ET distribution scaled by csx
• Csxs agree with each other but not necessarily correct ?
– Still also see some events where partons generated out of pT range
• Rapgap - Used a lot at HERA but not implemented in Athena
– Still trying to get this set up with Rivet
– R. Zlebcik (Prague) looking at this from theory perspective and
looking to do NLO calculations
• Have Pythia 6, Pythia 8 and Phojet SD and DD samples so can
try to find diffractive dijets within them
• Also told Herwig++ can do this as well but very little
information on this available at the moment
9
Analysis
• Gap finding inherited from Tim’s MinBiasPackage
• Divides calorimeter into 10 rings of unit rapidity
• Identifies calorimeter cells where energy significance
(= cell energy/noise) large enough that probability of
noise cell studied in event is small
• Where no cells with high energy significance found
in ring is determined to be ‘empty’
• Determine the biggest gap and where it starts
Example Single Diffractive Topology
Largest Gap
|Gap Start| = 5
Gap Size = 4
+pi
A B C D E E D C B A
-5
-4
-3
-2
-1
0
+1
+2
+3
+4
-pi
+5
10
Analysis
• Anti-Kt jets with R=0.6: Require >= 2 jets with ET > 7 GeV
– Currently no requirement to ask about jet quality cuts
– Currently no asymmetric jet ET cuts (NLO) e.g. ET1 > 12, ET2 > 8
• Ask for a forward gap: |start| = 5, gap ≥ 2 units
• Using first seven runs of data10 period A1 (MinBias stream,
latest reprocessing)
– 152166, 152214, 152221, 152345, 152409, 152441, 152508
– Total ∫L dt = 0.198 nb-1 (half of total lumi for period A1) – calculated
using online iLumiCalc tool with L1_MBTS_2 ref. trigger
– Average <μ> for selected runs < 0.01  currently ignore pile-up
• Using Pomwig Single + Double Diffractive MC
• Using Pythia 6 Non Diffractive MC for comparison (where
possible)
11
First Truth Level Comparisons
• Compared truth parton level with truth hadron level (final
state particles)
• Then went on to truth hadron level with reconstruction (in
particular applying cuts to pick out zIP, xP more easily)
MX parton v hadron
MX hadron v reconstructed
12
Resolutions (Jet ET & Mjj)
• Resolutions calculated with Pomwig as (Truth – Recon)/Truth
used to determine appropriate bin widths for variables
• RMS around 15% for both jets in ET, 20% for Mjj
13
Resolutions (Jet η)
• Resolutions calculated with Pomwig as (Truth – Recon) used
to determine bin widths for variables
• Resolutions in η have RMS of 0.6 for leading jet, 0.8 for subleading jet – need to also investigate jet mismatches (|Δη|>1)
14
Resolutions
• Resolutions calculated as (Truth – Recon)/Truth used to
determine bin widths for variables
• zIP shows some correlation but xP does not really work - (E±pz)
method does not work well in opposite direction in opposite
direction to dissociation
xP
zIP
15
Uncorrected Data
• Combined Pomwig SD+DD, Pythia 6 ND weighted relative to
luminosity of data runs used and then plotted against data
• Where statistics are plentiful it suggests between Pomwig
and data that there is a gap survival factor of 10-15
• Pythia 6 distributions smaller in jet ET, η but odd in Mjj  not
yet understood
16
Uncorrected Data
• Combined Pomwig SD+DD, Pythia 6 ND weighted relative to
luminosity of data runs used and then plotted against data
• Where statistics are plentiful it suggests between Pomwig
and data that there is a gap survival factor of 10-15
• Pythia 6 distributions smaller in gap size, zIP  not yet
understood
zIP
Gap size
17
Shape Comparison
• Combined Pomwig SD+DD, Pythia 6 ND weighted relative to
luminosity of data runs used and then scaled to data integral
(from plot) to make comparison of distribution shape
ET Jet 1
η Jet 1
Gap size
zIP
18
Differential Cross Sections
• Combined Pomwig SD+DD weighted to lumi of data runs Differential cross section as a function of leading jet ET along
with acceptance
Note in this and the
following slides that
the acceptances are
not well understood
just yet
MC/Data ratio suggests GSF of 10-15
19
Differential Cross Sections
• Differential cross section – as a function of leading jet η
Acceptance higher in
negative η compared
to positive η 
difference in MC
simulation?
MC/Data ratio suggests GSF of 10-15
20
Differential Cross Sections
• Differential cross section – as a function of gap size
Acceptance really
high in one bin
compared to rest?
Gaps being
reconstructed larger
than actual
generation?
MC/Data ratio suggests GSF of 10-15
21
Differential Cross Sections
• Differential cross section – as a function of zIP
0.8 < zIP < 1.0 has a
big acceptance →
likely to be due to
big migration (also
seen in resolution
plots)
MC/Data ratio suggests GSF of 10-15
22
Next steps
• Get official production of Pomwig MC
• Get cross sections from Rapgap / Herwig++ to
compare with Pomwig and NLO theory
• Run over remaining data in 2010 Period A1
• Improve resolutions between truth and
reconstruction levels for important variables
• Investigate changes to analysis
– Gap selection - different energy significances, η ring sizes,
are there gaps on other side of detector
– Jet quality cuts – are there bad jets being used and how
many events are lost as a result
– Asymmetric jet cuts – necessary but how much signal lost
– Pile-up – how to deal with events with 2+ rec. vertices
23