Search for Leptoquark at D

1
Search for Leptoquark at DÆ using Neural
Networks

• Silvia Tentindo-Repond
• Florida State University
• for the
• DÆ Collaboration
• Workshop on Neural Computation in
• High Energy Physics
• Maale Hachamisha, Israel
• Oct 13-15, 1999

2
Outline

• Introduction
• Brief overview of Tevatron and DØ
• Leptoquark Phenomenology
• Leptoquark Searches
• First Generation Searches
• Second Generation Searches
• Results and Conclusions

3
Introduction

• Leptoquarks (LQ)
• Fractional electric charge
• Color charge
• baryon and lepton number
• can be scalar or vector particles

First Generation
Second Generation
Third Generation
Experimental limits on rare decays and flavor
changing neutral currents imply that leptoquarks
dont mix generations
4
Leptoquark Searches at Dzero

• DØ has conducted several searches for Leptoquarks
  for the last several years
• Searches considered
• electron channels (eejj e?jj)
• muon channels (µµjj µ?jj)
• neutrino channels (??jj)
• scalar and vector couplings
• Motivation for recent searches
• HERA 97 - Excess reported

5
HERA Excess

• Observation of high Q2 events in ep collisions
  at HERA, by both H1 and ZEUS
• H1 7 events (1 pred.) Z.Phys.C74 (1997) 191
• ZEUS 4 events (0.9 pred). Z.Phys.C74 (1997) 207
• Possible interpretation as LQ1
• LQ1 mass 200 GeV

6
Experimental Apparatus The Tevatron

• Fermilab Tevatron
• proton - antiproton collider
• center-of-mass energy 1.8 TeV

7
The DØ Detector
Tracking ? lt 3.5 ?(vertex) 6mm ?(r?) 60?m
(VTX) 180 ?m (CDC) 200 ?m
(FDC)
Calorimetry ? lt 4.0 ????? 0.1 ? 0.1 ?(EM)
15/?E ?(?) 50/?E ?(jet) 80/?E
Muon Spectrometer ? lt 3.3 ?(1/p)0.18(p-2)/p2 ?
0.003
8
Leptoquark Phenomenology

• At the Tevatron, Leptoquark would be dominantly
  pair-produced via strong interactions
• decay within the detector
• Pair production process insensitive to the
  coupling l
• decay parametrized by ß ( branching ratio to
  charged leptons)

pp LQLQ (l-j)(lj) b2 (lj)(nj) 2b(1 -
b) (nj)(nj) (1 - b)2
9
Leptoquark Cross Sections
LO Vector
NLO Scalar
J. Blümlein, et. al., Z. Phys. C 76, 137 (1997)
Krämer, et. al., PRL 79, 341 (1997)
at 200 GeV/c2 ?slq 0.16 pb ?YM 10 pb
?MV 2 pb ?Min 0.6 pb
10
Search Strategy
11
Neural Networks

• Neural networks used in this search
• feed-forward
• back-propagation
• supervised learning
• single output
• single hidden layer
• sigmoid transfer function
• JETNET 3.0 (Lund)
• adapted to D0 Software environment
• ( P. Bhat, H.Prosper, C.Stewart )

12
First Generation Leptoquarksee jets

• Accept events with
• ETem gt 20GeV, ?eem ? 1.1 OR 1.5 ? ?e ? 2.5)
• nj(ET ? 15 GeV, ?e ? 2.5) ? 2
• ?R(ej) ? 0.7
• VETO (82 GeV/c2 ? mee ? 100 GeV/c2 )
• electron ID
• 101 events remaining from 123? 7 pb-1

13
First generation Leptoquarks (Conventional
Analysis)

• 101 candidate events - base sample
• Backgrounds
• Drell-Yan 66.8 13.4
• QCD 24.3 3.6
• ttbar 1.8 0.7
• Total 92.8 13.8
• Optimization - use kinematic parameters of MC
  events for signal and MC/data events for
  background to develop optimal set of cuts
• Best results obtained with cuts on ST
• ST HTjet HTe
• HTjet S ETjet( if gt 15 GeV) HTe S ETe
• ST gt 350 GeV maximizes the signal for LQ masses
  of 200 to 220 GeV/c2 above a total background of
  0.44 /- 0.06 events.

14
First Generation Leptoquarks (Conventional
Analysis)
Most effective cut
Events / 25 GeV
Signal Systematic Errors
Main Backgrounds Drell-Yan 2j, multijet, and
Top
LQ Signal Efficiency 36 at 225 GeV/c2
15
First generation Leptoquarks Neural Network
Analysis
First Generation Leptoquarks(Neural Network
Analysis)
HTe SETe HTj SETj
Training Signal LQ(200 GeV) MC Mixture of QCD,
D-Y and Top backgrounds Set desired output -gt 1
for signal -gt 0
for background
16
First Generation Leptoquarks ee jets
17
First Generation Leptoquarks eejets Mass Fit
(MC)

• 3 constraint (3c) mass fits
• Energy balance and equal LQ masses
• Monte Carlo studies
• MLQ1 225 GeV

18
First Generation Leptoquarks eejets Mass Fit
(data)

• ST gt 0 GeV

blue points - data solid black line -
background red dashes - 225 GeV/c2 LQ sample
19
First Generation LeptoquarksST vs Mfit

• ST gt 0 GeV
• B S
• Data

LQ1 Mass events expected
with STgt350 GeV 160 20
200 6 225
3 No evidence for LQ signal is
seen in mass analysis as well
20
First Generation Leptoquarks eejets Mass Limit
MLQ1 gt 225 GeV/c2 , b1.0
21
First Generation Leptoquarkse? jets
Final selection made using a 2 input Neural
Network
Network inputs
22
First Generation Leptoquarkse? jets
DNN gt 0.85 gives 0.41 ? 0.270 to 0.54 ? 0.25
expected background.
gives 0.60 ? 0.27 expected background
Signal Efficiencies
14.7 ? 1.1 mslq 180 GeV/c2
Backgrounds W2j, multijet, tt
23
First Generation Leptoquarkseejj e? ??
combined results
D0 Limits on First Generation leptoquarks
24
Second Generation Leptoquarks?? jets

• Separate signal and background
• SCM ? 1.5 (?1 ?2)
• ?1 lt ?2 lt ?3 are normalized eigenvalues of
• DNN selection

• DNN output of a 7?15?1 Neural Network

Inputs
25
Second Generation Leptoquarks?? jets

• Again
• careful examination of the distributions shows no
  evidence for a signal
• apply a selection on a neural network which
  maximizes the significance

4?9?1 NN with inputs
26
Second Generation Leptoquarks Exclusion Contours

• Combine limits at various values of ?
• obtain mass limits at each point
• exclusion contours in ? vs. mslq plane

Summary of 2nd Generation Leptoquark Mass limits
at DØ
27
Conclusions

• DØ has performed LQ searches in all 3 decay
  channels
• ?? jets with complete data in progress
• Tevatron Combined Limit (DØ CDF)
• (ee jets ? 1) ? 242 GeV/c2

First Generation
Phys. Rev. Lett. 79 4321 (1997) Phys. Rev. Lett.
80 2051 (1998) PRD in progress for complete search
Second Generation
2 PRLs nearing submission