Searches for strong production of supersymmetric particles with the ATLAS detector

Supersymmetry (SUSY) provides elegant solutions to several open questions in the Standard Model, and searches for SUSY particles are an important component of the LHC physics program. Naturalness arguments favour supersymmetric partners of the gluons and third-generation quarks with masses light enough to be produced at the LHC. With increasing mass bounds on more classical Minimal Supersymmetric Standard Model (MSSM) scenarios other variations of supersymmetry, including non-minimal particle content, become increasingly interesting. This proceeding will present the latest results of searches conducted by the ATLAS experiment at LHC at center of mass energies of $\sqrt{s}=13$ and 13.6 TeV which target gluino and squark production, including stop, in a variety of decay modes.

💡 Research Summary

The ATLAS Collaboration presents a comprehensive set of searches for strongly produced supersymmetric particles using the full Run 2 dataset at √s = 13 TeV (≈ 140 fb⁻¹) and a partial Run 3 dataset at √s = 13.6 TeV (≈ 52 fb⁻¹). The analyses target gluino, first‑ and second‑generation squark, top‑squark (stop), and charm‑squark (scharm) pair production in a variety of final states, including single‑lepton, all‑hadronic, charm‑tagged, and τ‑tagged signatures. Advanced object‑identification techniques (c‑tagging based on the DL1r algorithm, τ‑tagging, large‑R jet top‑tagging) and modern multivariate methods (neural networks, boosted decision‑tree classifiers, recursive jigsaw reconstruction) are employed to maximise sensitivity, especially in compressed mass spectra where the visible decay products are soft.

1. Stop pair production in the ℓ+jets+E_T^miss channel – The search separates events into a high‑E_T^miss “resolved” category (moderate‑p_T top quarks, no large‑R jets) and a “boosted” category (high‑p_T top quarks, at least one large‑R jet). A dedicated neural network discriminates signal from the dominant t t̄ and W+jets backgrounds, with control regions used to normalise these backgrounds. No excess is observed; the resulting 95 % CL limits exclude stop masses up to 1.23 TeV for a massless neutralino and, when combined with a zero‑lepton analysis, extend the exclusion to similar values across a broad Δm range.

2. Stop pair production in the t c+E_T^miss final state – This analysis assumes equal branching fractions for ˜t₁ → t ˜χ₁⁰ and ˜t₁ → c ˜χ₁⁰, probing models with non‑minimal flavour violation. A c‑tagging algorithm tuned for charm jets is used, and three Δm regimes (bulk, intermediate, compressed) are defined. In the compressed regime, an ISR jet requirement and a neural‑network classifier enhance sensitivity. The search sets limits of m(˜t₁) < 800 GeV (massless neutralino) and excludes compressed scenarios up to 600 GeV. The results are re‑interpreted to constrain flavour‑violating dark‑matter mediators up to 1.2 TeV.

3. Stop and scharm pair production in the cc+E_T^miss channel – Two kinematic regions are considered: a high‑mass region (m(˜t₁/˜c₁) ≳ 600 GeV, Δm ≳ 200 GeV) and a compressed region (Δm < 175 GeV). The high‑mass region requires a leading c‑tagged jet and exploits E_T^miss‑based variables; the compressed region relies on ISR jets and recursive jigsaw reconstruction while vetoing a c‑tagged leading jet. Backgrounds from Z+jets and W+jets are estimated with lepton‑based control regions. No excess is seen, and stop/scharm masses up to ≈ 0.9 TeV (neutralino massless) are excluded, improving previous limits by about 100 GeV.

4. Gluino and first/second‑generation squark production with τ‑leptons – Three τ‑multiplicity channels are defined: 1τ0ℓ, 1τ1ℓ, and 2τ. Both a traditional cut‑and‑count approach (optimising E_T^miss‑related selections) and a machine‑learning approach (multiclass BDTs trained on five‑fold cross‑validation) are applied. The BDTs classify events into signal, W+jets, Z+jets, diboson, top, and fake‑τ categories. No significant deviation from the Standard Model is observed. The resulting limits exclude gluino masses up to 2.25 TeV (neutralino up to 1.35 TeV for m_gluino≈2 TeV) and squark masses up to 1.7 TeV (neutralino up to 0.85 TeV). The cut‑and‑count method performs better in the compressed region, while the BDT approach provides superior reach at higher masses.

Overall, the suite of ATLAS searches finds no evidence for supersymmetry in the examined datasets. The analyses collectively push the 95 % CL exclusion limits for stops to ≈ 1.2 TeV, for scharms to ≈ 0.9 TeV, for gluinos to ≈ 2.25 TeV, and for first/second‑generation squarks to ≈ 1.7 TeV. The incorporation of c‑tagging, large‑R jet substructure, recursive jigsaw reconstruction, and sophisticated multivariate classifiers markedly improves sensitivity in compressed spectra, tightening previous bounds by up to 100 GeV. These results severely constrain natural‑SUSY scenarios that predict relatively light coloured superpartners and set the stage for future Run 3 and High‑Luminosity LHC analyses, where larger datasets and further refinements in object identification will be essential to probe the remaining viable parameter space of both minimal and non‑minimal supersymmetric models.