Abstract

This web page documents the search for the Higgs boson in the all-hadronic higgs channel (W/ZH → qqbb and Hqq → qqbb). The search is performed on 9.45fb^-1 of data recorded at CDF.
The values of Expected and Observed 95% Confidence Limit for 120GeV/c² Higgs boson are:

• Expected = 8.9
• Observed = 8.4

Introduction

The Higgs boson plays a central role in the Standard Model (SM) as it endows particles with mass. We search for a Higgs boson dacaying to two bottom-quark jets (bb) accompanied by two addtional quark jets (qq') for Higgs mass 100 ≤ m_H ≤ 150GeV/c². The search is most sensitive to a Higgs boson with mass < 135 GeV/c² where the Higgs boson dacay to bb is dominant. The two production channels studies are the associated production and vector boson fusion.

PDF PNG

PDF PNG

Event Selection

The events selected for this analysis must pass the CDF multi-jet trigger and the good run selection, the events must not have leptons, they have small missing transverse energy significance and they must present 4 or 5 jets. We require events with two b-tagged jets and we classify them in two catagories: SS and SJ. The events in SS category are defined as events with exaclty two jets tagged with SecVtx algorithm. The events in SJ categories are defined as events with exactly one jet tagged with SecVtx algorithm and exactly one tagged with JetProb algorithm.

The number of signal and background events which pass the events selection in the signal region and for each b-tagging category are given in the tables below.

PDF PNG

PDF PNG

bjet Energy Neural Network Correction

The di-jet invariant mass has a straight correlation to jet energy measurament and to optimize the di-jet mass we need to improve the resolution of jet energy measurament. A Neural Network (NN) is trained to correct the measured b-jet energy to the b-parton energy. The NN was trained with b tagged jets matched to b-partons. The match criteria is defined as ΔR between the b-jet and the b-parton is ≤ 0.4. We started to train, for each sample (VBF, WH and ZH), a dedicated NN. To identify the best NN performance, we applied to each sample the others NN. We obtained the best performance with the VBF NN. For the MH120, we have an improvement of value of M_bb, the mean value shift from 110.2GeV/c² to 124GeV/c², the RMS value reduces by 5-6% and the resolution increase by ~12-13%.

PDF PNG

PDF PNG

Background

The backgrounds considered in this analysis are:

• QCD Multi-jets : this is the dominant background as it consists of about 98% of the total background
• Top quark pair production.
• Single Top quark production.
• W and heavy-flavor jets production (W+HF).
• Z plus jets production, where Z decays into b and c (Z+JET).
• Diboson production : WW, WZ, ZZ.

Neural Network Training

For the Higgs analysis, a multivariate discriminant has the ability to combine the information from several variables. This improves the ability to separate a Higgs signal from background events far greater than a standard cuts analysis.

A dedicated Neural Net was trained for each process, WH, ZH and VBF. The output of this first Neural Net was used as the input of a second one (SuperDiscriminant - SD). We used the output of the last Neural Net for the analysis.

The Neural Net was trained at three target Higgs masses: 100 GeV, 120 Gev and 140 GeV. These three trained neural nets were used to search for a Higgs boson between 100 GeV to 150 GeV. For each mass point, the NN which give the best expexted limit was used:

• 100 GeV Higgs used Neural Net trained on 100 GeV Higgs sample
• 105, 110, 115,120,125,130GeV Higgs used Neural Net trained on 120GeV Higgs sample
• 135,140,145,150GeV Higgs used Neural Net trained on 140GeV Higgs sample

Neural Network Output

The following plots show the Neural Network Output for the mass point MH120 in SS and SJ categories.
For the others mass point for the SS and SJ category can be found here

PDF PNG

PDF PNG

Systematics

We consider systematics which affect the shapes & normalisation of the Neural Network distributions for the signal & background. The systematics which are considered are:

• TRF Systematic
• TRF Interpolation Uncertainty
• M_qq Tuning Uncertainty
• VBF-NN Tuning uncertainty
• Signal Systematics
• Jet Energy Scale
• ISR/FSR Systematics
• Jet Width Uncertainties

PDF PNG

Results

After examining the data, no indication of a Higgs signal is observed in the data and 95% confidence limits(CL) are quoted. The results for the combination of SS and SJ channel are summarized in the table and plot below.

PDF PNG

PDF PNG

Notes

More details can be find in the notes:

• Analysis Note
• Public Note