Analysis Stategy

Standard model Higgs boson production at the Tevatron can occur a number of ways. While Higgs events are expected to be produced via gluon fusion more often, associative production with a Z or W boson leaves distinct signatures in the detector which better identify "interesting events". This analysis focuses on a ZH search in which the Z decays to neutrinos. Such events produce a large amount of missing energy in the transverse direction, as neutrinos are undetectable. Near the LEP limit of 114/c² GeV, Higgs bosons are expected to primarily decay to a pair of bottom quarks, which can be identified by secondary vertices displaced from the primary point of interaction due to the long lifetime of b hadrons. Therefore the primary event selection for this analysis requires large missing E_T recoiling against 2 jets with identified secondary vertices. While a high P_T lepton veto is applied in this analysis to reject leptonic W events, a significant amount of WH is also expected to contribute to this channel due to events in which the lepton is unidentified, providing a way to recover some of the signal that cannot be measured in dedicated WH searches.

The single largest background to a Higgs signal in the missing E_T+jets channel involves QCD heavy flavor production. This is a difficult background to model, due to the fact that the events passing event selection are likely severely mismeasured, faking the signature large missing E_T from neutrinos. Past analyses have generated hundreds of millions of QCD Monte Carlo events to model this process. Therefore, we have developed a technique to model this process directly from the data. The main focus of this analysis is concerned with events containing two secondary vertex tagged jets. However, a substantial amount of heavy flavor exists in data in which only one of the jets has an identified secondary vertex. This data sample is used as the base of our heavy flavor QCD model. Light flavor, top and heavy flavor electroweak backgrounds are subtracted out at their expected rate. The leftover shape is used as our heavy flavor QCD template for the double tag data. This technique is tested in a heavy flavor QCD dominated region, where the Missing E_T is aligned with second jet. We achieve good agreement between our heavy flavor QCD shapes and the double tag data. Additionally, control regions have been developed to study the modeling of electroweak and top backgrounds, as well as falsely tagged light flavor events.

Two separate neural network have been developed for this analysis. The first focuses on discerning real missing E_T resulting from electroweak decays involving neutrinos from fake Missing E_T generated by mismeasurement. The main goal of this network is to provide a method for reducing the heavy flavor QCD background from a ZH signal without tightening event selection. The Track-based Discriminant is solely trained with tracking information. While calorimeter mismeasurement can generate fake missing E_T, it is uncorrelated with the P_T of charged particles measured in the tracking chamber. However, the missing E_T in events involving high energy neutrinos is highly correlated to the missing P_T from tracking. This neural network is highly efficient at separating ZH signal from the QCD background. In addition, a Neural Network Discrminant is utilized to combine tracking information with calorimeter-based quanities. This network is trained to optimize the separation of both ZH and WH events from QCD and tt backgrounds. The output of this network is used in the signal region to set expected and observed limits given the predictions of our Higgs signal and all relevant background processes. Separate networks have been trained to optimize background rejection for Higgs masses ranging from 110 to 150 GeV/c².