Observation%20of%20Electroweak%20Single%20Top%20Quark%20Production

1
Observation of Electroweak Single Top Quark
Production

• Bernd Stelzer

On behalf of the CDF Collaboration
Photo by Reidar Hahn Artwork by Jan Lueck
- Lawrence Berkeley National Lab , April 23rd
2009 -
On behalf of the CDF Collaborations
- Lawrence Berkeley National Lab , April 23rd
2009 -
2
Observation of Electroweak Single Top Quark
Production

• Bernd Stelzer

The Way To Single Top Discovery
- Lawrence Berkeley National Lab , April 23rd
2009 -
W. Wagner
3
An Idea was Born Production of Single Heavy
Quarks..
In 1985, ten years before the top quark
discovery, the idea of single top quark
production was born..
Symmetry magazine Jan./Feb. 2007
4
Top Quark History

• The search for the top quark lasted almost two
  decades.
• The big surprise was the huge mass!
• Single top quark production also predicted
• by the SM through an electroweak vertex

Discovered March 1995!
14 years!
?NLO 6.70.8pb (mt175GeV/c2)
Discovered March 2009!
s-channel ?NLO 0.880.07pb
t-channel ?NLO 1.980.21pb
B.W. Harris et al., Phys. Rev. D66, 054024 Z.
Sullivan, Phys. Rev. D70, 114012. Campbell/Ellis/T
ramontano, Phys. Rev. D70, 094012 (2004). N.
Kidonakis, Phys. Rev. D74, 114012 (2006).
5
Tevatron Experiments

• Fermilabs Tevatron Run II pp collider at 1.96
  TeV
• Surpassed design luminosity (Record Inst. Lum.
  3.18 1032 cm-2sec-1)
• Two multi-purpose detectors (CDF,D0) in highly
• efficient data taking mode since 2002

Fermilab
CDF
1 km
D0
Tevatron
6
A Long Way
CDF Single Top Results
4.2s Stronger Evidence ICHEP08
2.7 / fb
3.7s Strong Evidence Winter08 (PRL)
2.2 / fb
3.1s Evidence at LP07
PRD 71 (2005) 012005
1.5 / fb
0.16 / fb
2.5s Hint at DPF06
3.2/fb
Recorded Int. Luminosity (no GRL)
1.0 / fb
Moriond09 arXiv09030885 submitted to PRL
7
Single Top Observation
On March 4th, 2009! both, CDF and DØ submitted
observation papers to PRL
arXiv 09030850
arXiv 09030885
8
Top Quark in the Standard Model

• Top Quark is heaviest particle to date
• mt173.1 ? 1.3 GeV/c2 Teavtron, March 2009
• Close to the scale of electroweak symmetry
  breaking
• Special role in the Standard Model?
• Top Quark decays within 10-24s
• No time to hadronize
• Unique to study a bare quark

9
Why measure Single Top Production ?

• Source of 100 polarized top quarks
• Short lifetime, information passed to decay
  products
• Test (V-A) structure of W-t-b vertex
• Allows direct Measurement of CKM- matrix
  element Vtb
• ?single top Vtb2
• Is there a 4th generation?
• Altered production rate could indicate source
  from beyond SM physics
• Flavor changing neutral currents (t-channel),
  heavy resonances (s-channel)

Precision EW rules out simple fourth generation
extensions, but see J. Alwall et. al., Is
Vtb1? Eur. Phys. J. C49 791-801 (2007).
s-channel
t-channel
10
Single Top and Higgs

• Observing single top paves the way for low mass
  Higgs hunters
• Single top shares the same backgrounds and ?nal
  state as WHiggs
• Higgs cross section 10 smaller
• Single top is the last SM process to be observed
  at the Tevatron, before Higgs

single top
WHiggs
11

The Challenge
12
The Challenge

• Single Top production is a rare process at the
  Tevatron
• SignalBackground (SB) 1109 before doing
  anything
• First step
• Trigger and ID clean leptons/MET improves SB by
  a factor 106
• High pT lepton triggers (e,µ)
• MET jets triggers (recover non-fiducial
  leptons hadronic ? decays)
• Second step
• Topological event selection
• Efficient b-tagging
• Careful background estimates
• Third step
• Advanced analysis techniques to separate signal
  from background using discriminants
• SB gt 11 in most significant bins

13
Top Quark Event Signature
14
Single Top Candidate Selection

• High pT lepton MET jets Selection
• 1 Lepton, ET gt20 GeV, ?e(µ)lt 1.6
• Missing ET, (MET) gt 25 GeV
• 2 or 3 Jets, ET gt 20 GeV, ?lt 2.8
• Veto non-W, Z, Dileptons, Conversions, Cosmics
• At least one b-tagged jet (displaced secondary
  vertex tag optional jet probability tag)
• MET jets Selection
• MET gt 50 GeV
• Veto leptons
• ETgt 35 GeV (1st jet) andETgt 25 GeV (2nd jet)
• Same b-tag requirements as (1)
• Neural Network to suppress QCD background
• gt Orthogonal Event Selections (2) adds 33
  acceptance to (1)

CDF W(l?) 2 jets Candidate Event Close-up View
of Layer 00 Silicon Detector
12mm
ANNQCD
Run 205964, Event 337705 Jet 1 ET 62.8 GeV,
Lxy 2.9mm Jet 2 ET 42.7 GeV, Lxy 3.9mm
15
Background Dominated Sample

• W2 jet topology not very distinct
• Signal/Background W2jetbtag 117
• Counting experiment impossible
• Many sources of background
• Dominant background is WJets
• No golden variable
• Signal and background distributions look similar

single top signal
Need to make use of all single top
characteristics to distinguish it from large
backgrounds Use multivariate methods

• Mlvb shows top mass peak
• Neutrino ?MET (no pz)
• Suffers from Jet and MET
• resolution

Light quark jet follows beam direction Most
powerful for t-channel
16
CDF Single Top History
2006 Established sophisticated analyses Check
robustness in data control samples
2004 Simple analysis while refining Monte Carlo
samples and analysis tools
Phys. Rev. D71 012005
2 Years

• Development of powerful
• analysis techniques
• (Matrix Element, NN,
• Likelihood Discriminant)
• NN Jet-Flavor Separator
• to purify sample
• Refined background
• estimates and modeling
• Increase acceptance
• (forward electrons)
• 10x more data

2007 Evidence for single top quark production
using 1.5 fb-1 (expected and observed!)
First Tevatron Run II result using 162
pb-1 ?single top lt 17.5 pb at 95 C.L.
17

The Experimental Tools
18
CDF Detector
h 1.0

• CDF is working well
• integrated 5 fb-1 on tape
• - The analyses in this paper use 3.2 fb-1
• (All detector components ON)
• CDF is getting faster
• - 6 weeks turnaround time to calibrate,
  validate and process raw data

h 2.0
?
h 2.8

• Hadronic
• calorimeter
• Muon scintillator
• counters
• Muon drift
• chambers
• Steel shielding

• Silicon tracking
• detectors
• Central drift
• chambers (COT)
• Solenoid Coil
• EM calorimeter

Making best use of our detector and all
subsystems!
19
CDF b-quark Tagging and Jet Flavor Separation
b-tag Identified 2nd vertex
Even with a fully reconstructed secondary
vertex, 50 of the back-ground in the W jets
sample do NOT contain bottom quarks.
Charm tagging rate 9 Mistag rate 0.5-1
b like
c/l like

• Separate tagged b-jets from charm/light jets
  using NN trained on secondary vertex info
• Lxy, vertex mass, track multiplicity, impact
  parameter, semi-leptonic decay, etc...
• Used by all single top analyses as continuous
  analysis input
• 10-15 sensitivity increase (incl. systematics)

Neural Net Jet-Flavor Separator
20
Filling the CDF Cracks
Muons from METjets trigger
High pT Muon Trigger
METjet trigger
muon triggers
? rad
?
?

• Gain from METjets triggered events
• 30 gain in muon acceptance
• 10-14 gain in sensitivity

21
Background Estimate
Z/Dib
tt
Wbb
non-W
Mistags

• WHF jets (Wbb/Wcc/Wc)
• Wjets normalization from data and
• heavy flavor (HF) fraction from MC

Wcc
Wc
22
Expected Event Yield
lepton MET 2 jets event yield 3.2
fb-1 (single double SECVTX tag)
MET 2 jets event yield 2.1 fb-1 (single
double SECVTX tag, SECVTX JetProb.)
s-channel 58.1 8.4
t-channel 87.6 13.0
Single top 145.7 21.4
tt 204.1 29.6
Dibosons 88.3 9.1
Z jets 36.5 5.6
W bb 656.9 198.0
W cc 292.2 190.1
W cj 250.4 77.2
W light flavor 501.3 69.6
Non-W 89.6 35.8
Total background 2119.3 350.9
Total prediction 2265.0 375.4
Observed 2229 2229 2229
s-channel 29.6 2.7
t-channel 34.5 6.1
Single top 64.1 8.8
tt 184.5 30.2
Diboson 42.1 6.7
W HF 304.4 115.5
QCD multijet 679.4 27.9
Total background 1339.9 170
Total prediction 1404 172
Observed 1411 1411 1411
30
Predicted total backgrounds known to 13-17
23
Monte Carlo Validation
PTlepton
Untagged
single tagged
double tagged
Checking hundreds of plots!
MT(W)
Untagged
single tagged
double tagged
J1(?)
single tagged
double tagged
Untagged
24

Analysis Techniques
25
Analysis Flow Chart
CDF Data
Any technique to separate signal from
background Likelihood function Matrix element
Neural network Boosted Decision Trees etc
Analysis Technique
Analysis Event Selection
Apply MC Corrections
Monte Carlo Signal/Background
Cross Section

Signal Background
Template Fit to Data
Discriminant
Signal Significance
26
Multivariate Analyses
? ? ?
METJets Neural-Net (MJ) NeroBayes
Neural Net (NN) t-channel Likelihood
function (LF) Dedicated s-channel LF
(LFS) Boosted Decision Tree (BDT) Matrix
Element Method (ME)
27
Matrix Element Approach
Standard Event Simulation (Monte Carlo)
Matrix Element
Random Phase Space Point d?
Detector Simulation
Produces events with process specific Kinematics/
dynamics
Matrix Element Analysis
Event Probability
Transfer Functions
Matrix Element
For each possible underlying Process (Matrix
Element)
Jet resolution functions is the probability of
measuring a jet energy Ejet when Epart was
produced
28
Matrix Element Method
Event probability for signal and background
hypothesis
29
Event Probability Discriminant (EPD)
Take ratio of event probability densities as
discriminant (EPD)
b b-jet probability
Signal
Background
30
Validation in Data Control Samples
1 tag QCD enriched sideband
Untagged
Untagged
ME
METjets
NN
Untagged
Tagged W4jets
Untagged
ME
LFS
BDT
Extensive cross checks to validate MC modeling
before unblinding the signal region.
31
Binned Maximum Likelihood Fit
We perform a binned maximum likelihood fit of
discriminant templates to the data
Discriminant
Rate and
Shape systematics

• Rate systematics give fit templatesfreedom to
  move vertically only
• Shape systematics allow templates to slide
  horizontally (bin by bin)

Systematic uncertainties can affectrate and
template shape and are taken into account
32
Template Shape Uncertainties
Factorization and renormalization scale
Jet energy scale
Flavor Separator
µf,r
Mistags
signal
Wbb
Mistags

• A total of 370 shape uncertainties evaluated!
• Each template, each source of shape error, each
  channel (tags, jets, central and extended
  muon coverage)
• Shape uncertainties affect sensitivity - most
  are quite small but some appreciable

33

The Combination
Combine and conquer
34
Combination Strategy

• Combination using Super Discriminant (SD)
• Treats outputs of analysis discriminantsas input
  to new discriminant
• Evolutionary neural networks trained to give the
  best expected p-value, not classification error
  function
• Candidate networks compete witheach other
• Gained 13 over most sensitive input
• Optimization of
• - Network topology
• - Inter-node weights
• - Histogram binning

Neuro-Evolution of Augmenting Topologies (NEAT)
K O. Stanley and R. Miikkulainen, Evolutionary
Computation 10 (2) 99-127(2002)
2-jet 1-Tag Tight Lepton
Channels are divided up at least as finely as any
ingredient analysis (2 jets 3 jets), (1 tag
2 tags), (2 Lepton Categories) 8 Channels
35
Hypothesis Testing
p-value probability of upward fluctuation of
background to the data or something even more
signal-like Outcomes are rank ordered as
signal-like using -2lnQ
Background-Like Outcomes
Signal-Like Outcomes
? nuisance parameters
Neyman-Pearson Lemma Q is the uniformly most
powerful test
100M pseudo- experiments
sb outcome
Fit for WLF and WHF normalization. Fluctuate
all nuisance parameters in pseudo-experiments
Expected p-value
36

Results
Note All measurements were performed with the
discriminant distributions blinded until all
aspects of the analysis were under
control Essential requirement for a
statistically limited analysis like single top
37
Discriminants of Individual Analyses
5.2?
4.9?
4.0?
2.0?
5.2?
Discriminants normalized to prediction expected
sensitivities
38
Combination
Perform combined cross section fit over the two
orthogonal analyses (SD MJ)
High pT lepton MET jets
MET jets
1.4?
gt5.9?
39
Cross Sections
CDF Combination (All Channels)
40
Direct Measurement of Vtb

• Using cross section result measure Vtb
• Assume Standard Model (V-A) coupling
• and Vtb gtgt Vts, Vtd
• (from BR(t ?Wb) measurements)?

Vtb
CDF combined fit
Vtbgt0.71 at 95 C.L.
Z. Sullivan, Phys.Rev. D70 (2004) 114012
Vtb 0.91 0.11 (statsyst) 0.07 (theory)?
41
Signal Significance
5s Observation!
Analysis Significance Std.Dev. (s) Sensitivity Std.Dev. (s)
NN 3.5 5.2
ME 4.3 4.9
LF 2.4 4.0
LFS 2.0 1.1
BDT 3.5 5.2
SD 4.8 gt5.9
MJ 2.1 1.4
Combined 5.0 gt5.9
3.1 x 10-7
400 Mio pseudo-experiments!
Expected p-value xxx x10 -10 gt5.9s Observed
p-value 3.1 x10-7 5.0s
42
A Golden Event
Event taken 2007/05/27
light jet ET 52 GeV
electron PT 66 GeV/c
b-tagged jet ET 38 GeV
43
Signal Features
SDgt0.72
HT /GeV
rec
Q x ?
MTop /GeV
rec
Purity S/B 1.2
44
Observation of Electroweak Single Top Quark
Production

• Bernd Stelzer
• UCLA, SFU

The Way To Single Top Discovery
- Lawrence Berkeley National Lab , April 23rd
2009 -
On behalf of the CDF Collaborations
- Lawrence Berkeley National Lab , April 23rd
2009 -
45
Summary and Outlook

• We report Observation (5s) of electroweak single
  top quark production
• Important milestone for the Tevatron Run II
  program
• Direct measurement of the CKM matrix element
  Vtb0.91 0.11 (statsyst) 0.07 (theory)
  (CDF)
• Established new source of top quarks which allows
  to measure
• properties of top quark
• Advanced analysis tools were essential to
  establish the small
• signal buried underneath large backgrounds
• We developed and applied new experimental tools
• Important milestone along the way to the Higgs!
• All Standard Model backgrounds for Higgs searches
  are
• now firmly established
• Exciting times! Moving on to the Higgs Boson..!

Wresonances
preciseVtb
top FCNC production
top quark polarization
46
D0 Results
5s Observation!
47
Progress on Higgs at the Tevatron

• Recent progress
• Added acceptance from complementary trigger paths
  
• Increase acceptance by adding new ?nal state
  signatures
• Improved lepton identification
• Better b-quark tagging and flavor purifiaction
• Improved reconstruction Jet corrections
• Improved QCD rejection
• Advanced multivariate techniques MEBDT/NN

Combine all channels using information about SM
ratios