search for new phenomena in the cdf top quark sample

Search for New Phenomena in the CDF Top Quark Sample

Kevin LannonThe Ohio State UniversityFor the CDF Collaboration

ANL HEP Seminar 10-11-06 K. Lannon 2

Why Look in Top Sample?

Top only recently discovered Turned ~10 years old last year Samples still relatively small Still plenty of “room” for

unexpected phenomena

Top is really massive Comparable to gold nucleus! Yukawa coupling near unity Special role in EWSB?

Many models include new physics coupling to top

0.001

0.01

0.1

1

10

100

1000

u d

sc

b

t

Quark Masses

GeV/c2

5 orders of magnitude between quark masses!


What Might We Find? Nothing but the Standard Model . . . .

Not as bad as it sounds Test our abilities to calculate signal and background properties

Important at the LHC top becomes background to other searches

Constrains models that put new physics in the top sample It’s not Standard Model top at all!

Spin not 1/2? Charge not 2/3? Not likely . . . .

It’s not only Standard Model top Additional heavy particles decaying to high pt leptons, jets and

missing energy (t ’) Heavy resonance decaying to tt tH+b ttH production


The Tevatron and CDF Tevatron accelerator

Highest energy accelerator in the world (Ecm = 1.96 TeV)

World record for hadron collider luminosity (Linst = 2.29E32 cm-2s-1)

Only accelerator currently making top quarks

Central Cal

Plug Cal

CentralTracker

Silicon Tracker

Muon Detectors

CDF Detector Trigger on high pT leptons,

jets and missing ET

Silicon tracking chamber to reconstruct displaced vertices from b decays


Tevatron Performance

Integrated luminosity at CDF Total delivered: ~2.0 fb-1

Total recorded: ~1.6 fb-1 (~ 15 Run I!) So far for top analyses, used up to 1 fb-1

More analyses with 1.0-1.2 fb-1 in progress for fall and winter Doubling time: ~1 year Future: ~2 fb-1 by 2006, ~4 fb-1 by 2007, ~8 fb-1 by 2009

Peak Luminosity

Today’s Presentation:200 pb-1 ~ 1 fb-1

Integrated Luminosity at CDF


Triggering on Top Need high efficiency, low fake

rate trigger for high pT leptons Relies on track trigger XFT

Tevatron bunch spacing Original plan: switch from 396 ns

to 132 ns at high lum Now: staying at 396 ns

3x more interactions per crossing! ~ 9 interaction per crossing at

highest luminosity Z ee at low lum.9 add. Int./crossing

fakeFake tracks can be made from segments of different real physical tracks.

Trigger for muons without upgrade

Missing segments

Instrumenting additional layers reduces fake rate. Efficiency stays high.

Upgrade passed final review last week Efficiency = 97% for high pT tracks Fake track rejection factor = 5-7

Reduction factor ~ 4


Top Quark Production at Tevatron

~85% ~15% QCD pair production

NLO = 6.7 pb First observed at

Tevatron in 1995

EWK single-top production s-channel: NLO = 0.9 pb

t-channel: NLO = 2.0 pb Not observed yet

s-channel t-channel

Other?: ???,0 ttX Htt

(and LHC)

833 pb

~13% ~87%

10.6 pb

247 pb

Associated tW

62 pb


Top Production Rates

Like finding a needle in a haystack . . . .

pb7.6)175@( GeVMttpp top

One top pair each 1010 inelastic collisions at s = 1.96 TeV

Efficient Trigger ~90% for high pT leptons

Advanced analysis techniques Artificial Neural Networks

b-tagging

Needle in haystack (approx.)


SM Top Quark Decays

Particular analysis usually focuses on one or two channels New physics can impact different channels in different ways Comparisons between channels important in search for new physics

BR(tWb) ~ 100%


Top SignaturesElectron or

muon: pT > 20 GeV

Neutrino: Missing ET > 20-25 GeV

Jet: ET > 15-20 GeVcone = 0.4

b-jet: identified with secondary vertex tag

Dilepton Lepton + Jets All Hadronic

,e

bbtt

,e

bqqbtt

bqqqqbtt


Top Event Yields

To give an idea of CDF sample sizes . . . . Based on top cross section of 6.7 pb Background and signal numbers based on event yields from

current analyses, scaled by luminosity Assume no changes in event selection, efficiency, etc.

Luminosity 1 fb-1 4 fb-1

Total Top Events 6700 26,800

Decay Mode Dil. L + J L + J (b-tag) Dil. L + J L + J (b-tag)

Before Event Selection

330 1985 1325

7940

Selected Signal Events

50 480 290 190 1910 1140

Expected Background

40 2290 160 150 9150 670 L+J: ~2k signal events with 4 fb-1 (signal:background ~ 1 : 5) L+J (b-tag): ~1k signal events with 4 fb-1 (signal:background ~ 2:1)


Searching for New Physics Precision study of

top properties Non-SM behavior from

top quark Evidence of

something other than top in sample

Direct search for new phenomena in top sample Resonant production Non-SM decays New particles with

“top-like” signature New particles

produced in association with top

Vtb


Top Properties Working Group

Studying all properties of top quark (except mass) ~ 150 faculty,

postdocs, students ~15 papers (so far) ~50 active analyses

Vtb


Precision Study: Cross Section Cross section

Measured in different final states New physics can affect different

final states differently Different techniques used in

same final state Results combined at end for

most precise answer tt production calculated to NLO

Central value: 6.7 pb — 6.8 pb Uncertainties: 5.8pb — 7.4 pb For mtop = 175 GeV/c2

Combined result: 7.3 0.9 pb

A

NTop NTop = Nobs- Nbackground, or from fit


Two Best Measurements Both in Lepton + Jets Channel

Vertex Tag (weight = 0.50, pull = + 0.88) Uses b-tagging to increase ratio of signal to background Counting experiment

Count W+jets events with a b-tag Subtract expected background Excess attributed to top

Kinematic Artificial Neural Net (weight = 0.32, pull = -1.14) Uses kinematic variables to separate signal from background Combines several variables in a neural network to increase sensitivity Fit for the number of top events Does not use b-tagging (lower signal to background ratio)

Consistency Overall combination has 2 of 4.96 for 5 degrees of freedom Only Vertex Tag and ANN: 2 of 2.91 for 1 degrees of freedom

7% probability of getting worse agreement Not statistically significant yet, but keep watching!


B-Tagging

b-tagging: Identifying jets containing a b quark Take advantage of

long b lifetime Look at precision

tracking information for tracks within jet

Reconstruct secondary vertices displaced from primary

Efficiency Per jet

40% for b jet 9% for c jet 0.5% for light jet

Per event (tt ) 60% for single tag 15% for double tag


Sample Composition

W+light flavor:From pretag using mistag matrixW+heavy flavor:From pretag using MC for HF fraction and b-tagging eff.

Single Top and Diboson: Estimated using theoretical cross section

Non-W QCD: Estimated from MET and lepton isolation side-bands

Difference between observed and predicted background attributed to top

Event count before applying b-taggingNumber of events with an identified W + 1 jets


Lepton + Jets Vertex Tag Result

One Tag + HT Cut 8.2 ± 0.6 (stat.) ± 1.0 (sys.)

pb

Two tags, no HT Cut Cross check 8.8 +1.2 -1.1 (stat.) +2.0 -1.3 (sys.)

pbHT = scalar sum of lepton, jet, and missing ET


Using Kinematics to Identify Top

Look for central, spherical events with large transverse energy Signal: PYTHIA tt monte carlo Background: ALPGEN + HERWIG W + 3p monte carlo

• Normalized to unit area

• HT scalar sum of lepton, jet, and missing ET

• Aplanarity uses lepton, jet and missing ET

• Max jet uses 3 highest ET jets; all others use 5 highest


Statistical Sensitivity Evaluate expected fit

fractional error using MC-based pseudo experiments Single variable fits: fit signal

fraction using distributions of a single kinematic variable

Plotted Points: median fit fractional

error Error bars: 68% interval


Multivariate Approach: Neural Nets

Structure Composed of nodes modeled

after neurons in nervous system Organized into layers

Input layer: initialized by input variables

Hidden layer: takes information from each input node and passes to output layer

Output node: new discriminating variable with range [0,1]

Training Neural net output determined by

exposure to training data Iteratively adjust parameters to

minimize error:

Training accomplished through JETNET program(Peterson et al. CERN-TH/7135-94)

7 kinematic variables 7 input nodes

Output node—range [0,1]—signal = 1

1 hidden layer, 7 hidden nodesInformation flow


Statistical Sensitivity Evaluate expected fit

fractional error using MC-based pseudo experiments Single variable fits: fit signal

fraction using distributions of a single kinematic variable

NN: fit NN output of data to NN templates

Plotted Points: median fit fractional

error Error bars: 68% interval

NN Fit performs significantly better than single variable fits


Using NN to Fit Data Basic Approach

Train NN to distinguish tt signal from backgrounds

PYTHIA tt MC as signal model ALPGEN + HERWIG W + 3p MC as

background model Use this NN to make templates

for fitting the data Use same signal model as above Also extract QCD multijet

template from data Supplement electroweak template

with contributions from other processes: WW,WZ, Z + jets, single top

Fit templates to NN distribution from data

Binned maximum likelihood fit Three component fit

Signal and electroweak float QCD constrained to value

estimated using isolation vs missing ET method


Lepton + Jets Kinematic ANN Result

Sample Events Fitted tt (tt )

W + 3 Jets 2102 324.6 31.6 6.0 0.6 0.9 pb

W + 4-Jet 461 166.0 22.1 5.8 0.8 1.3 pb


Kinematics of b-Tagged Events

Looks like top!


Systematic Uncertainties

Main Systematic Uncertainties uncorrelated Lepton + Jets Vertex Tag

b-tagging efficiency: 6.5%Background estimation: 3.4%

Kinematic ANNBackground shape modeling: 10.2% Jet Energy Scale: 8.3%

For both results, uncertainty dominated by systematics

Both are working to reduce for 1 fb-1 publications


Search for t H+b Compare top yield in four different channels

Measurements consistent with SM Consider correlated effect of tH+b decays on four channels Exclude when changes make expectation inconsistent with data Limits for 6 sets of MSSM parameters and less model-specific scenarios

Varying model parameters changes:BR(tH+b)

BR(H+)BR(H+cs)BR(H+t*b)BR(H+W+h0)BR(H+W+A0)

Shown here: Variations as a function of tan particular set of MSSM parameters

Phys.Rev.Lett. 96 (2006) 042003


MSSM Limits

Calculate BR(tH+b) and H+ BR’s as a function of MH+ and tan()

Use 6 different MSSM “benchmarks” Results for “Benchmark #1” shown below


Less Model Dependent Limit “Tauonic Higgs” Model

Assume BR(H+) = 1 i.e. MSSM with high tan()

“Worst” Limit Find arbitrary combination of

H+ BR’s that give least stringent limit


The Search for Single Top Standard Model

Rate |Vtb|2

Spin polarization probes V-A structure

Background for other searches (Higgs)

Beyond the Standard Model Sensitive to a 4th generation Flavor changing neutral

currents Additional heavy charged

bosonsW’ or H+

New physics can affect s-channel and t-channel differentlyTait, Yuan PRD63, 014018(2001)


Signal and Backgrounds

W + Heavy Flavor

W + Light Flavor (Mistags)

Multi-jet QCD

tt

Other EWK

Total Background: 64696 events

Expected Single-Top: 28 3 events

Signal / Background ~ 1/20

e or : pT > 20 GeV: MET> 20 GeV

2 jets: ET > 15 GeV, 1 b-tag

Single-top Signature Backgrounds

Must use multivariate, kinematic techniques to separate signal from background


Current Results

Current limits are getting close to SM expectation

Apriori projections (ignoring systematics) 2.4 excess with 1 fb-1

3 excess around 1.5 fb-1

1 fb-1 in progress

Stay tuned!

Channel s+t-channel

t-channel s-channel

SM expectation (NLO)

2.9 0.4 pb

2.0 0.3 pb

0.9 0.1 pb

Likelihood 95% 4.3 pb 2.9 pb 5.1 pb

NN 95% 3.4 pb 3.1 pb 3.2 pb

Statistical errors only

Based on SM single-topcross section

Result w

ith 695 pb

-1


t’ Production

Consider possible contribution to “top” sample from heavier particles with “top-like” signature (t’)

Examples 4th chiral generation consistent with precision EWK data

[Phys. Rev. D64, 053004 (2001)] “Beautiful Mirrors” Model: additional generation of quarks that

mix with 3rd generation [Phys. Rev. D65, 053002 (2002)]

Consider decay of t’Wq Happens when mt’ < mb’ + mW

Precision EWK data suggests mass splitting between b’ and t’ small

Search for by fitting HT vs Mreco

HT = sum of transverse momenta of all objects in event

Mreco = Wq invariant mass reconstructed with a 2 fitter (same technique used in top mass reconstruction)


t’ Search Results

No evidence for t’ observed

Set 95% confidence level limits on t’BR(t’Wq)2

Exclude mt’ < 258 GeV for BR(t’Wq) = 100%

Interesting behavior in high mass tails


Summary This is an exciting time to be at the Tevatron

1 fb-1 sample currently in hand and being analyzed Top sample has grown from ~30 events in Run I to ~ several hundred

Larger samples coming soon Analysis techniques becoming increasingly mature and sophisticated Look forward to 1 fb-1 publications this winter

No evidence for new physics in top sample so far Have many more top measurements than covered in this talk (see CDF

public results webpage) Increasing precision continues to test consistency of measurements in

different channels Many new analyses on their way (as well as updates of current results)

Improved cross section measurements Single-top Top charge Flavor changing neutral currents Direct search for tH+b


The Future: Top at LHC “Top physics will be easy at

the LHC” Top cross section increases

by factor of ~ 100 Background cross sections

increase by factor of ~10 Top used for LHC detector

calibrations Searches for new physics

Mtt distribution Associated Higgs production:

ttH Above assumes top quark

holds no “surprises” at the LHC Precise understanding of top

properties and mass will speed the process

If there’s more than top in the top sample . . . .

Test and validate ability to model backgrounds


Extra Slides


W Helicity in Top DecayW Helicity in Top Decay

V-A Forbidden

W0 Longitudinal fraction

F0

W- Left-Handed fraction F-

tb

W

+1/2-1/2

+1

W+ Right-Handed fraction F+

tW

b

+1/2+1

-1/2

Helicity of W determined by V-A structure of EWK interaction 70% longitudinal 30% left-handed Right handed forbidden

tW+1/2

+1/2

0W

b


W Helicity in Top Decay

Can be tested by measuring W helicity angle: *

* = angle of the lepton relative to negative the direction of the top in the W rest frame.

Can also use Mlb

2 0.5(mt2-mW

2)cos *


W Helicity Results

Two CDF results with 955 pb-1

Use different kinematic fitters to reconstruct tt system: cos*

Very consistent measurements of F0 and limits on F+

F0 = 0.61 0.12(stat) 0.04 (syst) and F+ < 0.11 at 95% C.L.

F0 =0.59 0.12(stat) +0.07-0.06(syst)

and F+ < 0.10 at 95% C.L.

One measurement with 750 pb-

1

Uses Mlb and measures fraction of V+A

FV+A < 0.29 at 95% C.L.

Assuming F0 = 0.7 F+ < 0.09 at 95% C.L.


Top Quark Lifetime Measure impact parameter of lepton

from Lepton + Jets top decay Evidence of displaced top suggests

Production via decay of long-lived particle New long-lived particle in top sample Anomalous top lifetime

Templates for SM processes

Result: c < 52.5 m at 95% confidence level


Multivariate Discriminants

Improve signal discrimination by combining several variables into a multivariate discriminant Neural Network and multivariate likelihood function both used

Variables: ℓb and dijet invariant masses, HT, Q, angles, jet ET and , W-boson , kinematic fitter quantities, NN b-tag output

ZOOM


Search for Resonant Production Motivation

Some models predict particles decaying to top pairs

Should be visible as resonance in tt invariant mass spectrum

Example model: Topcolor assisted technicolor Extension to technicolor

that includes new strong dynamics

Couples primarily to 3rd generation

Includes new massive gauge bosons: topgluons and Z’

pp X 0 tt


Search for Resonant Production Look for generic, spin

1 resonance (X0) decaying to top pairs Assume X0 =

1.2%MX0

Test masses between 450 GeV and 900 GeV in 50 GeV increments

Results No evidence for

resonance Set 95% confidence

level limit for X0 at each mass

Exclude leptophobic Z’ with Mz’ < 725 GeV

search for new phenomena in the cdf top quark sample

Documents

high pt leptonsrelies

high lumnow

topneed high efficiency

lannontop quark production

low fake rate trigger

tevefficient trigger

different channels

quark masses