Transcript 20140424_DiffDijets_SMPlenary

Diffractive Dijet Production with
2010 data
Hardeep Bansil(1), Oldřich Kepka(2),
Vlastimil Kůs(2), Paul Newman(1), Marek
Taševský(2)
(1) University of Birmingham
(2) Institute of Physics, Academy of Sciences of the Czech Republic
Standard Model Plenary Meeting
24th April 2014
Current Status
• Previously presented in SM Plenary meeting - July 2013
• CDS supporting note available since March 2014
• https://cds.cern.ch/record/1670320
• (also see H. Bansil thesis)
• http://cds.cern.ch/record/1696944
• Aiming for editorial board / paper
24/04/2014
Diffractive Dijet Production with 2010 Data
2
Introduction to diffraction
• Total cross section in hadronic scattering experiments at 7 TeV
– Total = Elastic + Diffractive + Non-diffractive (ND)
– 20% elastic, 80% inelastic (diffractive + ND)
SD
• Diffractive channels together – 25-30% of the σinel
– Single-diffraction (SD: pp  pX) – Main process of interest
– Double-diffraction (DD: pp  XY)
• Kinematic variables
– invariant mass of the dissociated system MX (MY)
DD
• at the LHC energy spans mp+mπ to approx. 1TeV
– fractional momentum loss ξX of the scattered proton
ξX = MX2 / s
[ξY = MY2 / s]
• Diffraction in the realm of soft QCD
– Best described by phenomenological models (e.g. Regge theory)
– Exchange of color singlet (Pomeron)  large angular region in which no outgoing particles
(soft QCD radiation) are detected  rapidity gap
24/04/2014
Diffractive Dijet Production with 2010 Data
3
Rapidity gaps in ATLAS detector
•
Large Rapidity Gap (LRG) … Δη ~ -log10ξX … smaller ξX (MX)  bigger gap
– Region in η devoid of hadronic activity due to the exchange of colorless object (Pomeron)
•
Detector-level LRG definition (ΔηF)
– First defined by soft diffractive rapidity gaps analysis (Eur.Phys.J. C72(2012) 1926)
– Biggest region in η (starting at the edge of the detector η=±4.9) absent of clusters and tracks
complying selection:
ΔηF ~ 6 ↔ ξ ~ 10-4
– no tracks with pT>pTcut (pTcut = 200 MeV)
– no TopoClusters; noise suppression …
•
•
(pTcluster > 200 MeV)
•
most significant cell in the cluster: Ecell/σnoise> Sthreshold
No pile-up environment required
– Pile-up could occupy the gap
– Use early runs from 2010 (Period B)
24/04/2014
Diffractive Dijet Production with 2010 Data
LRG
4
•
•
•
Hard diffraction
The aim
–
Study single diffraction at hard scale, i.e. in high pT dijet events for first time in ATLAS
–
Measure cross-section vs. gap size (ΔηF) and ξ (fractional momentum loss of intact proton)
–
Independent measurement of gap survival probability in terms of both ΔηF & ξ
Gap Survival Probability (S2)
–
Hard diffraction studied precisely at HERA (ep collisions). Diffractive PDFs (dPDFs) measured.
–
Then Tevatron (pp collisions). Structure function ~10x smaller than HERA-based dPDFs predictions for Tevatron
conditions. Explanation – rescattering of dissociated system with intact proton.
–
S2 typical for hadron-hadron collisions
What about S2 at LHC?
–
theoretical predictions (KMR) … S2 ~ 5-10 %
Rescatter with p ?
Predictions based on
HERA’s dPDFs
CDF data
ξ
24/04/2014
zIP, Momentum fraction of
parton emitted by Pomeron
Diffractive Dijet Production with 2010 Data
5
PYTHIA8 + POMWIG
• PYTHIA8 non-diffractive samples
• PYTHIA8 diffractive samples can use
different Pomeron flux models –
influences diffractive distributions
• Comparison to default Schuler-Sjostrand
• H1 DPDFs better described by
Donnachie-Landshoff model
• Can test against different models
• POMWIG samples
• Modified HERWIG, Based on picture to right, using H1 DPDFs
• Generated over kinematic range 10-6 < ξ < 0.1 and
10-6 < |t| < 10 GeV2
24/04/2014
Diffractive Dijet Production with 2010 Data
6
Event selection criteria
•
•
•
Starting with SM2010 inclusive measurement (next slide)
Good Runs List, Good primary vertex (ntracks>4), Triggers
Kinematic cuts (pT jet 1 > 30 GeV, pT jet 2 > 20 GeV, |ηjets|<4.4) , anti-kT R=0.4 or R=0.6
•
•
•
Focus on 2010 Period B data adding pile-up suppression cut
need for events with 1 interaction only … no PU vertices (having ntracks>1)
removes 5% of events in period B (σ-correction factors applied)
•
•
Going below trigger efficiency plateau for jet triggers (next slide)
improving the use of available statistics for large gaps
•
•
•
Going to lower jet-pT ranges
Increases statistics, allows smaller diffractive systems to be studied  larger gaps
pT jet 1 > 20 GeV, pT jet 2 > 20 GeV , |ηjets|<4.4, anti-kT R=0.4 or R=0.6
•
•
•
•
•
•
•
Forward gap definition (ΔηF)
after extensive studies  “hybrid” gap-definition method:
“Measurements of the total transverse energy in pseudorapidity bins in proton-proton collisions at √s with ATLAS”
η-region devoid of activity (starting at either η=-4.8 or η=+4.8)
detector-level noise cuts: tracks with pT track > 200 MeV
TopoClusters with cell significance Ecell/σnoise > Sthr (~5.5)
Corrected (truth) cross section definition: pch (n) particle > 500 (200) MeV OR pT > 200 MeV
24/04/2014
Diffractive Dijet Production with 2010 Data

L = 6.75 nb-1
7
Trigger strategy
•
Start from Implementation of selection cuts and
trigger scheme from inclusive SM2010 dijet
cross-section measurement
•
Invariant mass spectrum successfully reproduced
•
presented during Inclusive Jet + Dijet Cross-Section meeting in April 2012
•
However, event yields at large gaps unsatisfactory
•
Optimise inclusive SM2010 measurement
(OR of triggers associated with leading and
sub-leading jet according to their pT, η)
–
L1_J5 trigger for central jets (|η| < 2.8)
–
Lower pT and forward jets triggered by L1_MBTS_1
(L1 forward jets not commissioned by then)
•
Going below 99% trigger efficiency plateau (but
staying above 70%) as MBTS_1 highly prescaled
•
presented during Jet Trigger Signature Group Meeting in October 2012
(this allows to use jet triggers at lower pT)
 necessity to weight events by 1/εtrig
L1_J5
(0.3<|η|<1.2)
R=0.4 jets
R=0.6 jets
Fit to R=0.4
Fit to R=0.6
lower pT  lower invariant mass  larger gaps
24/04/2014
Diffractive Dijet Production with 2010 Data
8
ξproton reconstruction
• Truth level - fractional momentum loss of scattered proton
– ξproton = (3.5TeV – pZproton) / 3.5TeV
• Cross section measured in terms of observable ξ± –
approximation to real ξ
– ξ± = Σ pT e±y / √s (TopoClusters / stable particles)
– ξ+ … intact proton going in the +z
direction (system X in -z dir)
– ξ- … intact proton going in the -z
direction (system X in +z dir)
Pythia 8 SD
• Approximation performs very well
for ξ<0.01
24/04/2014
Diffractive Dijet Production with 2010 Data
9
ND, SD and DD ranges
• Data against PYTHIA8 ND, DD and SD – all scaled to L= 6.75 nb1
• Non diffraction dominant for ΔηF < 2, ξ± > 0.01
• Then diffractive contributions become more prominent
24/04/2014
Diffractive Dijet Production with 2010 Data
10
Comparison of groups
• Comparison of uncorrected cross sections between Prague and
Birmingham groups
– Completely independent code bases
– Minor discrepancies exist  work in progress
24/04/2014
Diffractive Dijet Production with 2010 Data
11
Unfolding
• Bayesian 2D unfolding technique
– RooUnfold 1.1.1
– input: pT of the leading jet vs. ΔηF or ξ±
• Detector-level PYTHIA8 ND:(SD+DD) ratio “fitted” to data to get the best
possible shape description in both ΔηF and ξ± distributions
– ΔηF distribution: ND×0.62, (SD+DD)×0.206
– ξ± distribution: ND×0.577, (SD+DD)×0.283
24/04/2014
Diffractive Dijet Production with 2010 Data
12
Performance of 2D Unfolding
•
Comprehensive tests of unfolding performance & stability done
•
closure tests
•
sensitivity tests to mixture of ND, SD and DD (different distribution shapes)
•
convergence of iterations (χ2)
•
stability against choice of binning
•
... (more information in the back-up note)
•
optimum  4 iterations
*Illustration based on 1D unfolding
2D unfolding actually uses several
inputs to account for migrations
24/04/2014
Diffractive Dijet Production with 2010 Data
13
Systematic uncertainties
•
•
•
•
•
•
•
•
•
•
•
•
Jet Energy Scale (JES)
Jet Energy Resolution (JER)
Jet Angular Resolution (JAR)
Jet Reconstruction Efficiency (JRE)
Jet Cleaning Efficiency (JCE)
Unfolding
Trigger efficiency
Cluster energy scale (CES)
Cell significance threshold cut (CTC) uncertainty
Tracking - negligible
Vertex requirement
Luminosity
•
•
•
•
•
Procedure:
1) Run uncertainty-adjusted analysis on MC
2) Produce new smearing matrix & reco-MC plots
3) Unfold uncorrected data with new smearing matrix & reco-MC
4) Compare new unfolding to standard procedure unfolding
24/04/2014
Uncertainty inputs from SM 2010
dijet analysis used here
Inherited from soft diffractive
analysis
Diffractive Dijet Production with 2010 Data
14
Selected systematic uncertainties
•
•
•
•
•
•
•
•
•
•
•
•
Jet Energy Scale (JES)
Inherited from 2010 SM dijet analysis
Split into 7 components
JES can vary as much as 15% for
components as a function of η, pT
Added in quadrature
Uncertainty range: typically 25-30%
continuously rising up to large gaps
ΔηF, Cell significance threshold cut (CTC)
Adjust noise suppression thresholds up and
down by 10%
Uncertainty typically 15-25%
ξ±, Cluster Energy Scale (CES)
Cluster EM energy scale in MC adjusted by
correction factors
Uncertainty typically 10-20%
24/04/2014
Diffractive Dijet Production with 2010 Data
15
Differential cross sections
•
d
N weighted

Differential cross sections calculated as
for ΔηF
F
F
d
L( )
• Nweighted accounts for trigger efficiency per data event, prescales and unfolding
• Compare against POMWIG and different PYTHIA8 flux models
R=0.4
R=0.6
• Requirement of hard scale affects kinematics  no rapidity gap plateau
• Cross sections slightly higher for R=0.6 but maintain the same shape
• For PYTHIA8, ND ~1.3x larger in first bin, SD+DD/ND fairly even for ΔηF >2.5
• POMWIG ~3x larger csx than data for ΔηF>3, slightly higher for R=0.4
24/04/2014
Diffractive Dijet Production with 2010 Data
16
16
Differential cross sections
•
d
N weighted
Differential cross sections calculated as d   L  for ξ±
• Nweighted accounts for trigger efficiency per data event, prescales and unfolding
• Compare against POMWIG and different PYTHIA8 flux models
R=0.6, no gap req.
R=0.6, ΔηF > 3
• Forward gap requirement of 3 units removes majority of large ξ± events
• POMWIG results significantly above data, PYTHIA8 roughly equal to data csx
• Could be used to determine S2
24/04/2014
Diffractive Dijet Production with 2010 Data
17
Differential cross sections
• Differential cross sections also calculated for jet variables
• Without gap requirements, data compatible with PYTHIA8 ND
• After Forward gap requirement of 3 units to enhance diffraction:
• Leading jet pt: diffractive states fall away faster than ND
• Leading jet η: limited statistically but shows hints of double peak structure observed
in diffractive MCs, ND shows single central peak
24/04/2014
Diffractive Dijet Production with 2010 Data
18
Summary
• Reproduced soft diffractive & inclusive dijet cross section
measurements
• New selection cuts & trigger strategy developed
• Analysis of 2010 period B data in ΔηF (gap size) and ξproton
(fractional momentum loss) distributions
• Good agreement between Birmingham-Prague
• Supporting note in advanced stage
• Looking to request editorial board as soon as possible
• Some additional studies and MC generation (NLO calculations)
planned to improve measurement, extract gap survival
probability (S2)
24/04/2014
Diffractive Dijet Production with 2010 Data
19
BACK UP SLIDES
24/04/2014
Diffractive Dijet Production with 2010 Data
20
Validation of new triggers
Period B
Uncorrected distributions
Agreement within 5%!
Significant grow of statistics at large gaps
RAW EVENTS
SM2010 B
New B
Gaps > 3
59
186
Gaps > 4
9
52
24/04/2014
Diffractive Dijet Production with 2010 Data
22
Diffractive dijet measurement by CMS
•
CMS measured diffractive contribution to dijet production at LHC
–
–
–
•
comparison to different MC models
–
–
–
–
•
ND (red): PYTHIA 6 & 8
SD (blue): PYTHIA 8, POMPYT, POMWIG
DD: PYTHIA 8
POWHEG for NLO comparisons
S2
Results
–
–
•
based on L = 2.7 nb-1
Measurement of ~
ξ (approximates fractional momentum loss of scattered proton)
region of interest: ΔηF>1.9
SD MCs predict more events than observed by
factor ≈5 in lowest ~
ξ bin (S2)
data also consists of proton dissociative events
(scattered proton excited into low mass state
escaping undetected into the forward region)
Gap Survival Probability
–
–
24/04/2014
S2 = 0.12 ± 0.05 (LO)
S2 = 0.08 ± 0.04 (NLO)
Phys. Rev. D 87 (2013) 012006
Diffractive Dijet Production with 2010 Data
23
Unfolding systematic
•
•
•
•
•
Method: creating “new” MC (by reweighting) describing reco-level data well
1) Fit DataReco/MCReco by continuous function (data-MC scaled agreement doesn’t have to be perfect)
2) Rerun MC analysis with weights from the fitting function
3) Unfold scaled reco-MC by standard procedure
4) Compare unfolded MC to truth MC (scaled) -> uncertainty
24/04/2014
Diffractive Dijet Production with 2010 Data
24
ΔηF measurement
• Soft diffractive measurement
• dσ/dΔηF ~ constant over several units of rapidity  plateau
Soft diffraction
Diffractive plateau
24/04/2014
Diffractive Dijet Production with 2010 Data
25