Search for NMSSM Higgs in Top Quark Decays

Theoretical Motivation

The minimal supersymetric standard model (MSSM) contains five Higgs bosons: a light and a heavy Standard Model like CP-even Higgs (denoted h and H respectively), a CP-odd Higgs (denoted by A), and a pair of charged bosons (H^±). The next-to-minimal supersymetric standard model (NMSSM) further extends the MSSM to include an additional CP-even and CP-odd neutral higgs boson, as well as an additional neutralino. The extension of the MSSM to NMSSM resolves the conflict that under the MSSM it is theoretically expected that the h should have a mass below 82 GeV/c2 [*], but decay mode independent limits from the LEP experiments exclude this possibility. Although a large range of masses for the A boson have been experimentally excluded under the NMSSM, no limits exist for the scenarios that the mass of the A is less than twice the b-quark mass. Additionally, for many regions of parameter space, especially tanβ<≈2.5 the H^± bosons could likely be within reach of the Tevatron, especially considering the H^± mass has little dependence on other SUSY particle masses. If m(A)<2m(b) then the branching ratio of H^±→W^±A is always greater than 0.5 for m(H^±)≈100 GeV/c². A good explanation can be found in Phys. Rev. D 79, 055014 (2009) (or alternatively here). Further experimental motivation for the possible existence of a ≈100 GeV/c² H^± boson comes from a 2.8σ excess of W^±→τν, as compared to decays to electrons and muons from the LEP experiments [*], which if H^± was the cause would not (and is not) be observed at the Tevatron experiments. At the Tevatron direct production of the H^± is small compared to that of it's background, the W^±. Similarly, the same is true for the A boson. So instead of a direct production search, we take advantage of the relatively large tt pair production cross section, and low backgrounds to search for t→H^±b→W^±Ab→W^±bττ.

Experimental Methodology

In the kinematic regions allowed in the decays of top quarks, the taus from the A boson typically have too low p_T to allow for efficient identification. Additionally the decay kinematics and topological acceptance of events with an H^± and A in them appear very similar to standard model top quark events. Therefore we look for the presence of a low p_T (3<p_T<20 GeV/c) isolated track in the event that is not associated with any jets or the primary lepton. For signal events this track is typically created by the τ decay products. We search in the b-tagged, lepton+≥3 jets channel, requiring events to have at least 20 GeV/c² of missing transverse energy (met), sum of transverse energy (including met) greater than 250 GeV/c², and jets defined as having a corrected energy of greater than 20 GeV/c². We veto cosmic muons, electron conversions, and events were the lepton and track form an invariant mass near the Z mass pole. Our primary discriminant between Standard Model top and signal events is the presence of the low p_T isolated track, of which the dominant Standard Model background is poorly modeled by Monte Carlo simulation. The majority of the isolated tracks which are background to the signal are not part of the hard interaction which took place (so not from the top or anti-top quark decays), but rather from so called Underlying Event sources. The Underlying Event can be thought of loosely as the result of ripping the parton, which participates in the hard process, out of the (anti)proton. When the parton is removed, the (anti)proton must "heal" to make up for color and other imbalances, and in this process may produce charged hadrons which may leave tracks in the detector. Other sources of these isolated tracks include low p_T leptons from W^± or Z bosons, or hadronic jets which may not deposit enough energy in the calorimeter to be reconstructed as a jet, but instead may leave a single isolated track in the tracking region of the detector. Tracks from Underlying Event are poorly modeled in Monte Carlo simulation, so we rely on jet triggered data to model this background. Additionally the rate of events having an Underlying Event track varies according to the hard interaction which took place, for among other reasons how many gluons participated in the interaction, making it difficult to perform just a counting experiment to search for signal. However the p_T spectrum of underlying event tracks is found to be invariant to the hard interaction that took place, as can bee seen in the following data plots:

[png] [pdf] [gif]	[png] [pdf] [gif]
[png] [pdf] [gif]	[png] [pdf] [gif]

Where the the Lepton+Jet data corrected for bosons is Lepton+Jet data which has been corrected for the presence of tracks left by low p_T leptons from W^± or Z bosons. This correction is determined from Monte Carlo, and consists mostly of Z events in the lower jet multiplicity events. Additionally we find that the track p_T spectrum of jet triggered data is invariant to things like lead jet E_T, event ∑E_T, number of jets, b-tagging, etc. This allows us to use jet triggered data to model Underlying Event tracks in the signal region. Tracks originating from boson daugters (Z/W^±/A) are modeled using Monte Carlo. An important control region validation which shows we can sufficiently model Underlying Event via jet triggered data, as well as boson daughter tracks from Monte Carlo, is a fit to the track p_T spectrum in lepton + 1 or 2 jet events, in order to fit for the Z/γ* cross section. Although leptonic decays of Z/γ* events typically have 2 reconstructed leptons, if one lepton has a p_T<20 GeV/c, or the Z decays to τ leptons, the events may only have one reconstructed lepton, with the other lepton getting reconstructed as a low p_T track. Taking advantage of the fact these lepton tracks have a slightly different p_T spectrum than Underlying Event tracks, allows us to fit for the Z/γ* cross section in the lepton + jets channel using the p_T spectrum of the tracks. These fits can be seen below:

[png] [pdf] [gif]

[png] [pdf] [gif]

Which shows a very good consistency with the Z cross section measured in the dilepton channel. In these fits the Underlying Event and Z/γ* contributions are left unconstrained. The errors on the plots are from the fit. and statistical only. As can be seen in the results section of this page, isolated tracks from signal have a different p_T spectrum than Underlying Event tracks, thus to search for the presence of signal we perform a fit to the p_T spectrum of isolated tracks. Performing a fit to the p_T spectrum of tracks allows us to not be sensitive to our relatively large uncertainty on the rate of Underlying Event Tracks. The estimated and observed number of tracks can be seen below:

[png] [pdf]

Where the first column is the expected number of events before the track requirement, the second column is the expected number of tracks, the third column is uncertainty on expected number of tracks due to either experimental or theoretical production rate uncertainties. The JES Acceptance column denotes the uncertainty on track yield due to kinematic effects of event selection due to the jet energy scale uncertainty. The JES Track Rate column represents the effect of jets being 'created' or 'destroyed' by varying the jet energy scale, thus either allowing or removing an additional track from the analysis, since we require tracks to not be inside of a jet. The final column represents the uncertainty on the efficiency to measure a track in the detector. The estimates for number of Underlying Event tracks is derived from b-tagged jet triggered data with ≥3 jets, which has a similar Underlying Event structure and average number of gluons in the event as tt events, thus a similar number of tracks is expected. The Underlying Event systematic uncertainty is determined by the extremes in track rates between different event selection cuts (≥4 jets in ∑E_T triggered data, instead of ≥3 jet, jet triggered data), added to the extreme difference between data of different hard interactions, namely 1 jet Z/γ* rich data and 1 jet W^&plusm; rich data. Additionally a shape systematic is taken which was chosen by the variation of events selection cuts which caused the extremes in hardening or softening of the track p_T spectrum of jet triggered data. The shape systematic consisted of raising the required ∑ET of jet triggered events by 61 GeV/c2, the average missing transverse energy of signal events, and also requiring ≥2 jets, instead of ≥3 jets, we note that the shape systematics are a small effect. The Underlying Event systematics are only used for throwing pseudo-experiments; in fitting to (pseudo)data the Underlying Event rate is never constrained, and nominal shape is always used. Since signal events are are a subset of the Standard Model tt events, we implore a data and Monte Carlo hybrid method (the so called Method II prescription, see Phys. Rev. D 71, 052003 (2005)) to estimate the number of tt events in the data sample, taking into account the slight reduction of acceptance for signal events for a given branching ratio. This allows us to greatly reduce systematic uncertainties such as the total integrated luminosity, and theoretical production rate uncertainties of signal Signal is modeled using standard CDF tt PYTHIA Monte Carlo, where the decay of a top quark may be to a H^± and a b-quark, where the H^± then decays to a W^± and A boson. Since we consider masses of A only less than twice the b-quark mass, and above 4 GeV/c², our results assume A→ττ branching ratio of 100%. Also since signal tracks must come from a τ at truth level, there is some probability that a Underlying Event track could have had a higher p_T, leaving the signal track p_T spectrum dependant on the Underlying Event spectrum and rate. We therefore correct for this effect, as well as correcting for similar effects due to the signal branching ratio. The p_T spectrum of signal events can be seen in the results section of this page.

Results

Since data in the signal region agrees well with expectations, we therefore set 95% C.L. exclusion limits on the branching ratio of top to H^±b for various H^± and A masses:

[png] [pdf] [gif]

Where the shaded bands show the region we would expect 68% of experiments to set limits, given signal did not exist. The band for masses of A of 7, 8, and 9 GeV/c² is only approximate, more exact bands are showed below for each mass of A individually. These limits represent first limits in this previously unprobed parameter space of top quark decays. The observed p_T spectrum of tracks in data can be seen below, with signal plotted at the level we exclude at the 95% C.L.: