Search for squarks and gluinos in $pp$ collisions at $\sqrt{s} = 13$ TeV and $13.6$ TeV in events with $τ$-leptons, jets and missing transverse momentum using the ATLAS detector

TL;DR

The ATLAS search probes R-parity-conserving SUSY in final states with taus, jets, and MET, focusing on gluino and left-handed squark production with cascades to $\tau$-sleptons or $\tau$-sneutrinos. It employs two parallel analysis strategies—a ML-based multivariate framework and a traditional cut-and-count approach—to define three orthogonal signal regions (1$\tau$0$\ell$, 1$\tau$1$\ell$, 2$\tau$) and corresponding control/validation regions, using data from Run 2 at $\sqrt{s}=13$ TeV (140 fb$^{-1}$) and Run 3 at $\sqrt{s}=13.6$ TeV (51.8 fb$^{-1}$). Backgrounds are constrained via data-driven methods for fake taus and MC-based estimations for genuine taus, with a comprehensive treatment of experimental, theoretical, and statistical uncertainties. No excess is observed; the results set the most stringent limits to date on gluino and squark masses in these tau-rich SUSY scenarios, excluding gluinos up to about 2.25 TeV and squarks up to about 1.7 TeV for specific LSP masses. The study demonstrates the power of combining ML techniques with conventional analyses to enhance sensitivity across compressed and high-mass regions, guiding future explorations of SUSY signatures involving tau leptons.

Abstract

A search for R-parity-conserving supersymmetry in events with large missing transverse momentum, jets and at least one hadronically decaying $τ$-lepton is presented. Both gluino and squark pair production are considered, with the cascade decay of each gluino or squark producing either a $τ$-slepton or a $τ$-sneutrino. Three channels are examined, requiring either exactly one hadronically decaying $τ$-lepton and no other leptons, exactly one hadronically decaying $τ$-lepton and at least one other lepton, or two or more hadronically decaying $τ$-leptons. Analyses in the three channels are optimised independently and combined statistically. Two separate analysis strategies, either a cut-and-count or machine-learning approach, are used. The search uses 140 $\mathrm{fb}^{-1}$ and 51.8 $\mathrm{fb}^{-1}$ of $pp$ collision data recorded by the ATLAS detector at the Large Hadron Collider during 2015-2018 at $\sqrt{s} = 13$ TeV and 2022-2023 at $\sqrt{s} = 13.6$ TeV, respectively. Gluino masses below $2.25$ TeV and squark masses up to $1.7$ TeV are excluded.

Figures (14)

• Figure 1: Diagrams showing the production and decay of \ref{['fig:gluinoFeynmanDiagram']} a gluino pair and \ref{['fig:squarkFeynmanDiagram']} a squark pair. In the gluino case, the two-step decay via an off-shell squark $\tilde{g} \to \tilde{q}^* q \to q q \hbox{$\tilde{\chi}^0_2$}\xspace/\hbox{$\tilde{\chi}^\pm_1$}\xspace$ is symbolised by a four-body vertex.
• Figure 2: Expected background composition across the range of signal scores. The numbers of events shown are normalised to their bin-width. Benchmark signal points are overlayed.
• Figure 3: Schematic definitions of the ML-based SRs, CRs and VRs. The diboson VR in the 1tau1lep channel additionally relaxes the signal score requirement to $>0.15~(0.05)$ for Run 2 (Run 3).
• Figure 4: Distributions of important variables used in ML-based CRs in Run 2 (top row) and Run 3 (bottom row). The total statistical and systematic uncertainty of the SM background is shown by the hatched band. The top-quark, diboson, $W(\tau\nu)$, and $Z(\tau\tau)$ background predictions are normalised according to the background-only fit. The leftmost and rightmost bins include the underflow and overflow entries respectively. The ratio of the observed yield to the expected yield is shown in the lower panel.
• Figure 5: Distributions of important variables used in ML-based VRs in Run 2 (top row) and Run 3 (bottom row). The total statistical and systematic uncertainty of the SM background is shown by the hatched band. The top-quark, diboson, $W(\tau\nu)$, and $Z(\tau\tau)$ background predictions are normalised according to the background-only fit. The leftmost and rightmost bins include the underflow and overflow entries respectively. The ratio of the observed yield to the expected yield is shown in the lower panel.