Boosted Multijet Resonances and New Color-Flow Variables

TL;DR

The paper addresses the challenge of locating fully hadronic multijet resonances from RPV gluino decays at the LHC by combining jet-substructure tagging with two novel color-flow observables, radial pull and axis contraction, to exploit the distinctive soft radiation patterns of color-singlet $R$-hadrons. By performing a shape-based analysis of fat-jet masses and applying aggressive color-flow and substructure cuts, the authors demonstrate a substantial improvement in signal discrimination against QCD and $t\bar t$ backgrounds, achieving a prospective gluino mass reach up to about $m_{\tilde g} \sim 750$ GeV at LHC8 with 20 fb$^{-1}$, and ~650 GeV with 5 fb$^{-1}$. The study covers both heavy gluinos ($m_{\tilde g} \gtrsim 500$ GeV) and top-mass gluinos ($m_{\tilde g} \sim m_t$), showing that color-flow observables can dramatically enhance resonance visibility and shape, while remaining robust across multiple Monte Carlo generators. These methods offer a complementary, largely orthogonal approach to existing counting-based searches and bear potential for generalization to other color representations and decay topologies. The work also emphasizes the need for experimental assessment of pile-up effects on color-flow observables, while providing a framework for future substructure studies in boosted colored objects.

Abstract

We use modern jet-substructure techniques to propose LHC searches for multijet-resonance signals without leptons or missing energy. We focus on three-jet resonances produced by R-parity-violating decays of boosted gluinos, showing that shape analyses searching for a mass peak can probe such gluinos up to masses of ~ 750 GeV (650 GeV) with 20/fb (5/fb) at the LHC at 8 TeV. This complements existing search strategies, which also include counting methods that are inherently more prone to systematic uncertainties. Since R-parity-violating gluinos lighter than all squarks hadronize before decaying, we introduce new color-flow variables, "radial pull" and "axis contraction", which are sensitive to the color structure of the R-hadron's decay. The former measures the inward pull of subjets in a fat jet, while the latter quantifies the inward drift of the $N$-subjettiness axes when changing the distance measure. We show that they can dramatically improve the discrimination of a boosted gluino signal versus QCD, ttbar and combinatoric background for m_gluino ~ m_top. Cuts on axis contraction also noticeably improve the resonance shape for heavy gluinos with m_gluino > ~500 GeV. With minor adaptations, these variables could find application in substructure searches for particles in different color representations or with other decay topologies. We also compare how several different Monte Carlo generators model the high-multiplicity QCD background. This provides evidence that the discriminating power of our color-flow observables are robust, and provides useful guidance for future substructure studies.

Figures (13)

• Figure 1: The signal studied in this paper: pair production of gluinos, with each decaying to three quarks via an intermediate off-shell squark.
• Figure 2: (a) The color flow in $R$-meson $(\tilde{g} qq)$ decay due to RPV gluino decay via an intermediate off-shell squark. (b) and (c): The two physically distinct possibilities for color flow in $R$-baryon $(\tilde{g} qqq)$ decay. Forward/backward arrows indicate color/anticolor, which are also indicated with $r,g,b$.
• Figure 3: Exaggerated schematic representation of the hypothesized radiation pattern in the $\eta-\phi$ plane produced by the $R$-hadron decays shown in Fig. \ref{['f.Rhadron']}. The radiation pattern from the shower & hadronization of the gluino is shown in purple at the center. The left-ward slant indicates (in this example) a color-connection to the beam, but this radiation pattern is extremely soft, hard to detect, and susceptible to uncertainties in the $R$-hadron formation model. The centers of the hard jets from RPV gluino decay are shown in red, with orange indicating the shape of the soft radiation pattern resulting from a second round of showering & hadronization of the three quarks produced in the gluino decay. Squares and triangles indicate the location of axes minimizing the N-subjettiness variables $\tau_3^{\beta = 1}$ and $\tau_3^{\beta = 2}$, respectively, and the cross marks the fat-jet centroid, i.e. the original $R$-hadron direction. There are two possibilities for the radiation pattern: the left plot corresponds to the $R$-meson and the $R$-baryon in Fig. \ref{['f.Rhadron']}(a) and (b), respectively; the right plot only occurs for the $R$-baryon in Fig. \ref{['f.Rhadron']}(c) and is of limited importance, since $R$-baryons make up only $\sim 1\%$ of produced $R$-hadrons pythiaRhadrondiscussion.
• Figure 4: Left: Normalized distributions of the new color-flow variable radial pull ($t_{\rm r}$, Eq. (\ref{['e.radialpull']})) for a gluino $R$-hadron with $m_{\tilde{g}} = 175$ GeV decaying to three jets via RPV (red and orange); QCD (green) and $t\bar{t}$ (cyan) backgrounds are also shown. The signal is divided into two components: red (orange) corresponds to events in which the hardest two fat jets are aligned (are not aligned) within $\Delta R = 0.3$ of the two gluino $R$-hadrons at the MC truth level. The red distribution is dominated by events where both fat jets reconstruct a decaying gluino, and thus constitute "good" (T) signal events, whereas the orange distribution shows events where the fat jets do not reconstruct a decaying gluino and are "bad" (F) signal events. For each event with two hard fat jets, the larger of the two radial pulls is shown in the histogram. The inset table shows the absolute sizes of the different samples, normalized to the number of expected events at LHC8 with 20 $\mathrm{fb}^{-1}$. Basic generator- and trigger-level cuts have been included, same as for Fig. \ref{['f.cutflow.light']} (a). Error bars indicate MC statistical uncertainty. Right: Same as plot on left, but now showing the normalized distribution of the new color-flow variable axis-contraction ($A^{\beta \beta'}_N$ with $\beta=2, \beta'=1, N=3$) for $m_{\tilde{g}} = 650$ GeV. An equivalent set of cuts has been applied, same as for Fig. \ref{['f.cutflow.heavy']} (a). In both cases the distribution of the good signal events differs markedly from the other (background) distributions.
• Figure 5: Comparing the normalized distributions of $\mathrm{max}(\tilde{t}_{\rm r})$ (maximized over the two hardest fat jets in the event) for 175 GeV gluinos with and without formation of $R$-hadrons before decay. The kinematic cuts (though not the radial-pull cut) of the top-mass gluino analysis outlined in Section \ref{['ss.topmass']} have been applied. Comparing the first and last bin occupations readily distinguishes between gluinos that either form or do not form $R$-hadrons before decaying.