Contributions to the Search for Single Top Production at DZero

1
Contributions to the Search for Single Top
Production at D-Zero

• Matt Tilley
• University of Washington
• Masters Examination Presentation
• March 13, 2006

2
Outline

• Introduction
• Standard model and top quark
• Top quark production
• Tevatron and D-Zero Run II detector
• Analysis
• Strategy for single top search
• B-tagging
• Cross section limit calculation
• Current status of single top search

Matt Tilley Masters Examination
Presentation March 13, 2006 2
3
Top quark and standard model

• Standard model
• Three flavors of matter
• Three corresponding forces (no gravity, yet)
• Higgs search ongoing

• Top quark properties
• Third generation, up type quark
• t mass 172.7 2.9 GeV/c2 (world avg.)
• ttop 4 x 10-25 seconds
• Charge 2/3, spin ½
• Discovered in 1995 at FNAL stt
  7pb

Matt Tilley Masters Examination
Presentation March 13, 2006 3
4
Electroweak single top quark production

• via W-boson exchange

t-channel (tqb) pp ? tqb X
s-channel (tb) pp ? tb X
st-channel 1.98 0.30 pb _at_ NLO and vs 1.96
TeV
ss-channel 0.88 0.14 pb _at_ NLO and vs 1.96
TeV
st-channel 1.98 0.30 pb _at_ NLO and vs 1.96
TeV
ss-channel 0.88 0.14 pb _at_ NLO and vs 1.96
TeV
Current limits from Run II D-Zero s95s-channel
6.4/4.5 pb (obs/exp)
Current limits from Run II D-Zero s95t-channel
5.0/5.8 pb (obs/exp)
Matt Tilley Masters Examination
Presentation March 13, 2006 4
5
Single top motivation

• Why is single top important?
• Production mechanism is an excellent tool to test
  the validity of the SM
• Opportunity to study the Wtb vertex
• Direct measurement of Vtb as s Vtb2
• Test CKM unitarity
• Precision measurement of top mass and other top
  properties
• Background for beyond SM processes
• e.g. Higgs production

Matt Tilley Masters Examination
Presentation March 13, 2006 5
6
The Tevatron at Fermi National Laboratory

• Tev produces collisions of 980 GeV protons and
  anti-protons
• vs 1.96 TeV
• Highly relativistic
• p p collisions
• Occur every 396 ns
• Approx. 2.5 MHz rate
• Luminosity 1032 cm-2 s-1
• Approximately O(1 day) to generate a single top
  event

Matt Tilley Masters Examination
Presentation March 13, 2006 6
7
D0 Run II Detector

• Tracker
• Detect trajectories of the collision output
• Silicon wafer interior, scintillating wire
  tracker exterior 2T solenoid
• Calorimeter
• U bathed in liquid Ar U causes interaction Ar
  detects interaction and gives signal from which
  we can measure energy deposited
• Muon system
• Muons are not absorbed in the calorimeter, and
  are short-lived. Lifetime is long enough that
  they can escape the detector, however. Their
  energy is measured by proportional counters and
  they are detected by firing the scintillators.

Tracker
Matt Tilley Masters Examination
Presentation March 13, 2006 7
8
D0 Run II Detector

• Photons no tracks in tracking system, deposit
  energy in EM calorimeter
• e - leave tracks in the tracking system,
  deposits energy in the EM calorimeter
• µ - leave tracks in the tracking system, MIP
  trace in the calorimeter, fire scintillators,
  energy deposited in the proportional counters
• Jets showers of hadrons by partons, may leave
  tracks, deposit energy in the hadronic
  calorimeter
• ? non interacting, detected as missing
  transverse energy

Matt Tilley Masters Examination
Presentation March 13, 2006 8
9
2.5 MHz _at_ 20kB/event? Thats 50GB/sec!
Well, until we apply triggers anyway.
980 GeV protons
980 GeV anti-protons
Physics!

• Detect hits in
• muon scintillators
• calorimeter
• silicon layers

• Trigger Level 1
• calorimeter energy
• muon hits
• central tracks

396 ns 2.5 MHz
2 kHz
50 Hz to tape

• Trigger Level 2
• clustered calorimeter energy
• missing transverse energy
• matched muon segments

• Trigger Level 3
• jets
• primary vertex
• muon matched to tracks

1 kHz
Matt Tilley Masters Examination
Presentation March 13, 2006 9
10
Integrated Luminosity to Date
Included analysis is 230pb-1
Matt Tilley Masters Examination
Presentation March 13, 2006 10
11
Single top analysis
Event signature
If t-channel, light quark jet observed
High PT muon
High PT neutrino
High PT b-tagged jet
b-tagged jet (not seen most of the time, high ?)
Matt Tilley Masters Examination
Presentation March 13, 2006 11
12
Analysis overview (230 pb-1)
Selection cuts Select events with similar final
states
B-tagging Seek jets from b-quarks
Determine backgrounds Determine backgrounds,
estimate after b-tagging
Multivariate analysis Discriminate single top
from background
Limit calculation Calculate x-sect limit with
binned likelihood
Matt Tilley Masters Examination
Presentation March 13, 2006 12
13
Selection cuts

• Muon Selection
• Exactly one w/PT gt 15 GeV
• Reject cosmic muons
• Isolated from jets and other muons
• Require 3 hits in the tracking system
• Require 3 hits in the muon system

• Jet selection
• 2 of jets in event 4
• PT gt 25 GeV (leading), PT gt 15 GeV (others)
• ?det lt 2.5 (leading), ?det lt 3.4 (others)

• MET selection
• 15 GeV ET 200 GeV

• Event quality selection
• Remove mis-measured events
• Pass level 3 trigger
• Require SMT validity

• Extra lepton veto
• Require no EM object in muon events

Matt Tilley Masters Examination
Presentation March 13, 2006 13
14
Muon isolation study
Purpose? Study the effect of loosening muon
isolation parameters to discern any possible
increase in single top signal acceptance while
minimizing the increase in the number of
additional background events.

• Muon isolation parameters
• loose - ?R(µ,jet) gt 0.5 defined during muon
  reconstruction, not easily modified (not included
  in this study)

• tight Track halo in a cone ?R lt
  0.5
• tight Cal halo in a conical shell
• 0.1 lt ?R lt 0.4

Calorimeter cell
Matt Tilley Masters Examination
Presentation March 13, 2006 14
15
Muon isolation study
Originally, isolation parameters were defined as
0.06 and 0.08 for track halo and cal halo,
respectively.
Chosen value of (0.15, 0.15) gives 10.4
increase in signal yield
Cal Halo Et By Pt
Z-scale is the factor of signal acceptance
increase from the previous choice of (0.06, 0.08)
Previous value of (0.06, 0.08)
Track Halo Pt By Pt
Matt Tilley Masters Examination
Presentation March 13, 2006 15
16
Muon isolation study
Problem! The highest magnitude background
increase when loosening muon isolation parameters
is produced via QCD multijet events, i.e. muons
from bb or cc are mis-identified as a muon
resulting from a W decay. How do we calculate the
mis-IDd fraction?
Solution The matrix method
Real loose- isolated muons from single top,
Wjets, tt
Fake loose-isolated muons from multijet events
Muons passing tight isolation parameters (the
ones being loosened!)
Probability that a fake muon passes from loose to
tight sample
Probability that a real muon passes from loose to
tight sample
Matt Tilley Masters Examination
Presentation March 13, 2006 16
17
Muon isolation study
Black dashed area is same area of signal plot
Cal Halo Et By Pt
Z-scale is factor of fake muon increase from the
choice of (0.06, 0.08)
Previous value of (0.06, 0.08)
Track Halo Pt By Pt
Chosen value of (0.15, 0.15) gives factor of
2.13 increase in fake rate
Matt Tilley Masters Examination
Presentation March 13, 2006 17
18
Analysis overview (230 pb-1)
Selection cuts Select events with similar final
states
B-tagging Seek jets from b-quarks
Determine backgrounds Determine backgrounds,
estimate after b-tagging
Multivariate analysis Discriminate single top
from background
Limit calculation Calculate x-sect limit with
binned likelihood
Matt Tilley Masters Examination
Presentation March 13, 2006 18
19
What is b-tagging and why does it matter?

• What is b-tagging?
• When a b-quark is produced in the collision,
  they almost immediately hadronize and form b
  mesons
• These mesons have a relatively long lifetime and
  travel 1 mm before decaying
• The decay products leave tracks that point back
  to a point that is not the primary vertex
• Several algorithms, I will be focusing on
  Secondary Vertex Tagger (SVT)

• Why does it matter?
• Strong discriminator to distinguish between
  single top signal (always containing b-quarks
  from top decay) and background (events without
  b-jets)
• efficiency is 40 for high PT central jets
  (like from single top)
• light jet mis-tag rate is approximately 1.0
  (u,d,s,c)

Matt Tilley Masters Examination
Presentation March 13, 2006 19
20
SVT algorithm

• Select the primary vertex of event
• Find multiple tracks within cone ?R 0.5 that do
  not point back to the PV
• Trace tracks back, seek displaced vertex
• Calculate the decay length Lxy (xPV xDV) and
  its error (sLxy)
• If the significance ratio Lxy/ sLxy gt 7, count as
  tagged b-jet
• For MC, assign TRF (tag rate function)value to
  the jet based on PT and ?.

p
p
Matt Tilley Masters Examination
Presentation March 13, 2006 20
21
Proper MC event weighting from TRFs
Problem Since the tagging algorithm returns the
TRF value, which is a probability that a jet is
tagged by the detector and not a certainty, one
needs to take into account all possible jet-tag
combinations for each event. Then assign a weight
to that event associated with the total
probability from all combinations.

• Enter the permuter software.
• Grabs all good jets for the event
• Cycles through all possible combinations of
  tagged jets, marking 0 tags to njet tags for each
  combination
• For each jet-tag combo, a weight is calculated
  from TRF values assigned to the event which is
  then passed to subprocessors for the calculation
  of various variables (tmass, WT mass, various
  shape variables, etc.)

Jet 2
Jet 1
p
p
Jet 4
Jet 3
Matt Tilley Masters Examination
Presentation March 13, 2006 21
22
Analysis overview (230 pb-1)
Selection cuts Select events with similar final
states
B-tagging Seek jets from b-quarks
Determine backgrounds Determine backgrounds,
estimate after b-tagging
Multivariate analysis Discriminate single top
from background
Limit calculation Calculate x-sect limit with
binned likelihood
Matt Tilley Masters Examination
Presentation March 13, 2006 22
23
Backgrounds for singletop tt Wjets
Top pair production
W Jets (incl. Diboson)
Todays discovery is the background for
tomorrows.

• Two channels dilepton (both Ws decay to
  leptons) and lepton jets (one W decays to
  lepton, one hadronically)
• stt 3 times the size of single top production
• Higher b-tagging and trigger efficiency.
• Dilepton has similar jet content to single top,
  lepjets has more.
• Branching fraction of lepjets is dominant.
• Modeled with Monte Carlo

• Problematic Wbb is the exact final state of
  s-channel, and Wqbb is the final state of
  t-channel
• Modeled with MC, normalized to data
• Cross section for W2jets is 1000pb
• Two Wjets processes modeled Wjj and Wbb, also
  WW, WZ, etc.
• Wjj Wbb combined according to the NLO SM cross
  sections

Matt Tilley Masters Examination
Presentation March 13, 2006 23
24
Backgrounds for singletop Mis-IDd muons (QCD)

• Mostly bb events
• One of the b quarks decays into a muon which is
  mis-IDd
• The muon is inside of a jet and the jet is not
  reconstructed
• Also produces MET
• Use data instead of MC
• Larger statistics
• Difficult to model in MC
• Two different processes for estimating dependent
  upon lepton µ/e

Matt Tilley Masters Examination
Presentation March 13, 2006 24
25
Estimating the Yields
Matt Tilley Masters Examination
Presentation March 13, 2006 25
Matt Tilley Masters Examination
Presentation March 13, 2006 25
26
Analysis overview (230 pb-1)
Selection cuts Select events with similar final
states
B-tagging Seek jets from b-quarks
Determine backgrounds Determine backgrounds,
estimate after b-tagging
Multivariate analysis Discriminate single top
from background
Limit calculation Calculate x-sect limit with
binned likelihood
Matt Tilley Masters Examination
Presentation March 13, 2006 26
27
Discriminating variable determination

• Same variable pool is used for all three
  analyses.
• Optimized separately for s- and t-channel.
• Seek maximum discrimination in variables

• Look for best combination of cuts which maximize
  SB
• Standard variables
• HT
• MT(W)
• Jet Multiplicity
• Top Mass Window

Matt Tilley Masters Examination
Presentation March 13, 2006 27
28
Two single top distributions
Matt Tilley Masters Examination
Presentation March 13, 2006 28
29
Event shapes are nifty

• Developed by Bowen, Ellis, Strassler
• Shapes are an interesting tool because they
  avoid the massive systematic errors found in
  counting experiments.
• Interesting prospect for t-channel single top
  discrimination due to the high ? light flavor
  jet, unshared by the backgrounds.

• pp is a CP eigenstate, therefore so are its
  final states
• So,
• Define and the CP invariant
• cross section as

• Correlations for the various backgrounds and
  signal for discrimination strength?

Matt Tilley Masters Examination
Presentation March 13, 2006 29
30
Correlations for UW Theory variables

• If leading untagged jet and lepton are
  uncorrelated, then
• Furthermore, if both uncorrelated and P
  invariant
• Resulting in the consequence

• t-channel single top is P-asymmetric and
  correlated so violates the above relation

• tt satisfies above at 90 level P symmetric no
  correlation
• multijet events are similar to tt, excepting
  lepton-jet correlations in events with heavy
  flavor. Disregarded due to small occurrence.

• Wjets events have moderate asymmetry and
  correlations accords to above roughly - still
  problematic

Matt Tilley Masters Examination
Presentation March 13, 2006 30
31
Define shape variables for discrimination
Matt Tilley Masters Examination
Presentation March 13, 2006 31
32
A look at F-

• s-channel and tt clearly removed
• t-channel now roughly the same size as Wjets
  events
• SB ratio for shapes vs. SB ratio for pure
  counting
• SBshapes 12
• SBcounting 130

All is not well 1) Wjets changes shape
drastically with varying flavor content, needs to
be studied further 2) number of MC events at
D-Zero was lt 1 of events produced by UW theory,
so any shapes seen could be statistical
Matt Tilley Masters Examination
Presentation March 13, 2006 32
Matt Tilley Masters Examination
Presentation March 13, 2006 32
Matt Tilley Masters Examination
Presentation March 13, 2006 32
33
Multivariate Analyses

• Same pool of variables used for all three
  analyses
• Optimized separately for s-channel and t-channel
• Focused on the two dominant backgrounds Wbb
  and tt lepjets

Matt Tilley Masters Examination
Presentation March 13, 2006 33
34
Cut based analysis
Matt Tilley Masters Examination
Presentation March 13, 2006 34
35
Decision trees
Matt Tilley Masters Examination
Presentation March 13, 2006 35
36
Decision tree output
Matt Tilley Masters Examination
Presentation March 13, 2006 36
37
Neural networks
Matt Tilley Masters Examination
Presentation March 13, 2006 37
38
Neural network output
Matt Tilley Masters Examination
Presentation March 13, 2006 38
39
Analysis overview (230 pb-1)
Selection cuts Select events with similar final
states
B-tagging Seek jets from b-quarks
Determine backgrounds Determine backgrounds,
estimate after b-tagging
Multivariate analysis Discriminate single top
from background
Limit calculation Calculate x-sect limit with
binned likelihood
Matt Tilley Masters Examination
Presentation March 13, 2006 39
40
Limits from binned likelihood
Matt Tilley Masters Examination
Presentation March 13, 2006 40
41
D-Zero Results
Matt Tilley Masters Examination
Presentation March 13, 2006 41
42
Conclusion and present status

• D-Zero has the current best published
  upper-limit on the search for single top quark
  decay
• s95s-channel 6.4 pb , s95t-channel 5.0 pb
  (neural networks)
• With 1 fb-1 data recorded (roughly 4 times the
  amount for this limit) and more coming in every
  day, evidence and discovery are just around the
  bend
• Improvements ongoing
• Better estimating Wjets heavy flavor fractions
• Better object ID (muon isolation issues, etc.)
• Combining multiple b-tagging algorithms in NN
• Increase in data will help systematics

Matt Tilley Masters Examination
Presentation March 13, 2006 42
43
Thanks

• Henry Lubatti and Gordon Watts
• Aran Thomas as well as the entire D-Zero
  Single Top Working Group at Fermilab
• Our in-house theorists Bowen, Ellis Strassler

Matt Tilley Masters Examination
Presentation March 13, 2006 43