CDF Logo Search for the SM Higgs Boson using τ lepton
~ Simultaneous Search for WH/ZH/VBF/ggH in 2τ's+2jets Event ~
CDF Collaboration
UC Logo
General Remarks :
All results described below can be used in conference presentations with proper reference to the CDF Collaboration. They however should be considered preliminary until published in a refereed journal.


Documentation: CDF Note 9248

Figures and Tables :
Please click figures and/or tables to get clear images.

Contents

  1. Abstract
  2. Signal Processes
  3. Event Selection
    1. Signal Acceptance
    2. Background Estimation
    3. Systematic Uncertainties
  4. Neural Net Technique
    1. NN Input Choice
    2. NN Output Results
  5. 95% C.L. Upper Limit
  6. Concluding Remark


1. Abstract

    We present the results of a search for the Standard Model (SM) Higgs Boson using τ decay mode.
Four signal processes are considered:     Candidate events are selected by requiring one isolated lepton (e or μ) from leptonic τ decay (τlep), one hadronic τ (τhad) and at least 2jets in an event. In order to further improve search sensitivity, a technique we employ is a multivariate artificial neural network (NN). After optimizing a capability to discriminate signal against various backgrounds, four SM Higgs processes are simultaneously' searched with the final NN shape.
Using 2.0 fb-1 data accumulated at CDF Run II, there was no clear excess in our search region. Therefore we set a 95% confidence level upper limit on the cross section times branching ratio (σ× BR(H→τ+τ-)) relative to the SM prediction. The observed limit (σ(limit)/σ(SM)) is from 30 to 159 for Higgs mass (MH) from 110 to 150 GeV/c2, while the expected limit ranges from 24.2 to 111.7. This is not only the first result for SM Higgs search using τ decay mode at Tevatron but also the first approach to look for different Higgs production processes simultaneously' with final state basis.

2. Signal Processes

  Four Standard Model Higgs production processes are under consideration. The Feynman diagrams are shown below. Left diagram is association with vector boson (W/Z) in which vector boson decays into 2jets (BR:67/70% for W/Z). Middle one is so called Vector Boson Fusion (VBF) process in which 2jets coming from proton/anti-proton tend to have a large rapidity gap in η. The last process (right diagram) is the gluon fusion process that has the largest production cross section. In this round of analysis, we only focus on 2 or more jet bin (≥ 2jet). *Note that analyzing 0jet and 1jet events (from gluon fusion) is harder than 2jet bin due to much worse S/B so further study would be needed to maximize search sensitivity (listed in our future updates).

3. Event Selection

We select candidates events with the following criteria:

Signal Acceptance & Expectation

The expected number of events with 2 fb-1 data as a function of Higgs Mass.
The numbers in parenthesis correspond to acceptance in percent. (Note: denominator is σ× BR(H→τ+τ-)).


Background Estimation

Main backgrounds are Drell Yan(Z→τ+τ-), tt and QCD fake. For Drell Yan, tt, diboson(WW/WZ/ZZ), we use Monte Carlo simulation. The same sign data is used for estimate and model of jet→τhad fake. Due to a charge correlation between W and outgoing quark, there is an asymmetry in OS and SS. Therefore additional W+jets contribution is properly included as "add-on" W+jets.



Description gif file (just click figures for clear image) eps file
Background Summary
Njet Distribution after all cuts applied.
eps
Kinematics (1)
From left to right :
Lepton Pt,
Missing Et,
Visible Tau Pt
    eps1 eps2 eps3
Kinematics (2)
From left to right :
τ seedtrack Pt,
dφ(lep,Met),
Transverse Mass (Lep,Met)
    eps1 eps2 eps3
Kinematics (3)
From left to right :
Dijet Mass
dη(jet1,jet2)
  eps1 eps2
Kinematics (4)
From left to right :
M(lep,τ),
M(lep,τMet),
Full M(ττ) after Collinear Approximation
    eps1 eps2 eps3

Systematic Uncertainties

Systematics uncertainties are summarized below for background and signal separately.

A summary of systematic uncertainties for background

A summary of systematic uncertainties for Higgs Signal


4. Neural Net Technique

We train 3 Neural Nets and take minimum of those three score fit the data.

NN Input Choice

Starting from 16 kinematic variables shown below, the choice of input variables is optimized.

Description gif file (just click figures for clear image) eps file
Kinematics for NN input with Signalx30 (1)
eps
Kinematics for NN input with Signalx30 (2)
eps
Kinematics for NN input with Signalx30 (3)
eps


A summary of input variables used for each NN training.


NN Output Results

Description gif file (just click figures for clear image) eps file
NN output distributions
From left to right :
NN(Signal vs Z→ττ)
NN(Signal vs tt)
NN(Signal vs QCD)
at MH=120 GeV
    eps1 eps2 eps3
Final NN output distribution
After selecting minimum NN score out of three NN outputs above.

Results for MH=120 GeV is shown.
eps

5. 95% C.L. Upper Limit

Since there is no clear excess in signal region, we set a 95% CL limit relative to SM prediction.

Description gif file (just click figures for clear image) eps file
Likelihood Distribution to set the Limit (1)
From left to right :
MH = 110 GeV/c2
MH = 115 GeV/c2
MH = 120 GeV/c2
    eps1 eps2 eps3
Likelihood Distribution to set the Limit (2)
From left to right :
MH = 130 GeV/c2
MH = 140 GeV/c2
MH = 150 GeV/c2
    eps1 eps2 eps3
95% Confidence Level Exclusion
Expected and observed upper limit on cross section times branching ratio (H→ττ) relative to the Standard Model prediction.
eps

Expected and observed upper limit on cross section times branching ratio (H→ττ) relative to the Standard Model prediction.

6. Concluding Remark

This is the first result for Standard Model Higgs search using τ decay mode in Tevatron (i.e. in the world). Compared to the other analyzes such as WH→lνbb, ZH→ llbb and ZH→ννbb, expected sensitivity is roughly 50% (observed limit is ~30-40%). By combining with all other analyzes, this result will improve low mass Higgs search sensitivity (95% CL limit).
In addition, there are some more aspects that strongly motivate us to work on this channel.
Maintained by Kohei Yorita
Last modified: Tue Mar 25 14:38:24 CDT 2008