Advanced Analysis Methods for Higgs in Run II

1
Advanced Analysis Methods for Higgs in Run II

• Pushpa Bhat
• Fermilab

2

• Collaborators
• R. Gilmartin
• H. Prosper
• Run II SUSY/Higgs Working Group
• P.B., R.G., H.P., PRD 62, 074022 (2000)
• Run II Higgs Working Group Report,
• hep-ph/0010338
• For a discussion of Advanced Methods, see
• my video-streamed plenary talk at
• http//conferences.fnal.gov/acat2000/

3
Outline

• Low mass Higgs MH lt 130 GeV/c2
• WH, ZH Production
• Experimental Issues
• Signal-background discrimination
• b-tagging
• Mass resolution
• Conclusions

4
Low Mass Higgs Study

• Signals
• Channels

5
WH ? l?bb Single Lepton
Channel

• Signal
• WH
• Backgrounds
• Wbb
• WZ
• tqb
• tb
• ttbar
• Detector Simulation
• SHW
• Networks
• Jetnet (v3.0)
• 3 Networks/MH
• WH vs WZ
• WH vs Wbb
• WH vs tt
• 3 cuts

(7,9,1) Network
6
WH(100 GeV/c2) vs. WbbVariables
7
WH(100 GeV/c2) vs. WZVariables
8
WH(100 GeV/c2) NN Distributions
9
Optimizing Cuts (WH100)
10
WH(110 GeV/c2) NN Distributions
11
NN Learning
12
Discriminating Surface
13
Standard vs. NN
14
Combined Network
NN1 WH-Wbb
NN2 WH-WZ
NN4
NN3 WH-top
NN4 f (NN1,NN2,NN3)

• Use in fitting distributions
• Use in mass measurement

15
Combined Network (WH110)
Currently see no improvement in significance by
fitting the NN distributions for single lepton
channel. This is work in progress
16
b-Tagging

• LEP experiments have impressive b-tagging using
  Neural Networks
• Use Neural Networks to combine lifetime and
  kinematic variables
• Work in progress at CDF DØ
• R. Demina 60 better effpurity than
  conventional
• S. Repond (ACAT2K Proceedings)

P
b
b jets
Non-b jets
Channel-dependent b-tagging
17
Jet Energy Corrections
Example
Correct jet energy using transverse energy and
eta.
Use NN to fit the correction function
?
18
Mass
Another approach Find the parameters for energy
correction by minimizing the variance in Mbb
directly
19
Higgs Mass Fits
S 80 WH events Wbb used as Background
S/B 1/10 Mfit 114 ?11 GeV/c2 S/B 1/5 Mfit
114 ? 7 GeV/c2
MethodP.B., H.P., S.Snyder, PLB 407, 73 (1997)
20
Work in Progress

• Systematic selection of variables
• Continue study of b-tagging, mass resolution.
• Fit distributions instead of counting above cuts,
  and use new results in all channels
• Continue to develop Bayesian method for combining
  channels.
• Study effect of various systematic uncertainties
  on discovery reach.

21
Conclusions

• Finding Higgs at the Tevatron will be extremely
  challenging!
• As evidenced from work at LEP, we will need all
  the sophistication in analysis that we can
  acquire.

22
Back-up Slides
23
Mbb
HT
24
Higgs Mass Resolution
Use mjj and HT (? Etjets ) to train NNs to
predict the Higgs boson mass
13.8
12.2
13.1
11..3
13
11