Identifying Multi-Top Events from Gluino Decay at the LHC

TL;DR

This paper analyzes LHC signals from a light gluino decaying predominantly into third-generation quarks, producing top-rich final states. It shows that early discovery is feasible through multi-lepton, multi-bottom channels, especially same-sign dileptons, even with modest data. Instead of relying on direct top reconstruction, the authors develop a template-based fitting approach to extract gluino decay branching ratios and constrain the production cross section, using inclusive channel counts across many final states. The method demonstrates robustness to mass assumptions and indicates that, with about 5–10 fb^-1, one can glean key features of the SUSY spectrum, including the LSP nature and third-generation squark hierarchy, via a combination of branching-ratio fits and simple mass-scale observables like the effective mass.

Abstract

We study the LHC signal of a light gluino whose cascade decay is dominated by channels involving top, and, sometimes, bottom quarks. This is a generic signature for a number of supersymmetry breaking scenarios considered recently, where the squarks are heavier than gauginos. Third generation final states generically dominate since third generation squarks are typically somewhat lighter in these models. At the LHC we demonstrate that early discovery is possible due to the existence of multi-lepton multi-bottom final states which have fairly low Standard Model background. We find that the best discovery channel is 'same sign dilepton'. The relative decay branching ratios into tt, tb and bb states carry important information about the underlying model. Although reconstruction will yield evidence for the existence of top quarks in the event, we demonstrate that identifying multiple top quarks suffers from low efficiency and large combinatorial background, due to the large number of final state particles. We propose a fitting method which takes advantage of excesses in a large number of channels. We demonstrate such a method will allow us to extract information about decay branching ratios with moderate integrated luminosities. In addition, the method also gives an upper bound on the gluino production cross section and an estimate of the gluino mass.

Figures (9)

• Figure 1: Distributions of the lowest four values of separations, $\Delta R$, between reconstructible partons (which here we take as any first/second generation quark or b-quark) obtained using the Pythia parton-level (truth) information for benchmark model A. For each event in the simulation, $\Delta R$ was computed for all pairs of reconstructible objects, ranked, and the lowest four values binned into histograms. The resulting distributions showing the lowest, 2nd, 3rd, and 4th lowest $\Delta R$ values encountered are given for (a) four-top events of model A, and also (b) hadronic $t\bar{t}$ decays. All distributions are normalized to unity. The partons present in four-top events are significantly closer to one another relative to those in SM $t\bar{t}$ events, increasing the liklihood of overlap, as well as a lower reconstruction efficiency.
• Figure 2: As a demonstration of the combinatoric background, in figure (a), we show the observed number of possible top 'candidates' obtained for our benchmark model A. (a) We consider combinations with one $b$-jet and two light jets, where the invariant mass falls within the top mass window, and where the combination satisfies a minimal set of selection criteria (see text). Due to the combinatorics, the same $b$-quark can be combined with other partons to form multiple "top candidates". In (b) we show the resulting number of top candidates after attempts are made to remove combinatorics by isolating distinct 3-jet combinations. For this basic study we only require that events have at least 4 $b$-tagged jets.
• Figure 3: Results for fits of ratio Br$(\tilde{g}\rightarrow t \bar{t})/$Br$(\tilde{g}\rightarrow b \bar{b})$ for benchmark Model-B. The left panel shows the result with the correct mass hypothesis. In the right panel, the efficiencies are calculated for five different mass templates show in the legend. The solid horizontal line gives the actual values of the branching ratios. Errors are 1$\sigma$ and include errors for subtracting off the $t\bar{t}$ background.
• Figure 4: Results for fits of ratio Br$(\tilde{g}\rightarrow t \bar{t})/$Br$(\tilde{g}\rightarrow b \bar{b})$ for benchmark Model-C. The left panel shows the result with the correct mass hypothesis. In the right panel, the efficiencies are calculated for five different mass templates shown in the legend. The solid horizontal line gives the actual values of the branching ratios. Errors are 1$\sigma$ and include errors for subtracting off the $t\bar{t}$ background.
• Figure 5: Results for fits of ratio Br$(\tilde{g}\rightarrow t \bar{b}+\bar{t} b)/$Br$(\tilde{g}\rightarrow t \bar{t})$ for benchmark Model-D. The left panel shows the result with correct mass hypothesis. In the right panel, the efficiencies are calculated for five different mass templates show in the legend. The solid horizontal line gives the actual values of the branching ratios. Errors are 1$\sigma$ and include errors for subtracting off the $t\bar{t}$ background.