Recent results on vector-like quarks and excited fermions at CMS

The most recent results on the searches for the vector-like quarks and excited fermions from the CMS Collaboration are presented. These results are based on the full Run 2 $\sqrt{s} = $13 TeV proton-proton collision data collected by the CMS Collaboration at the LHC from 2016 to 2018, which corresponds to an integrated luminosity of 138 fb$^{-1}$. No significant excess above the Standard Model expectation is observed. Exclusion limits are set at the 95% confidence level on various benchmark models.

💡 Research Summary

**
The CMS Collaboration has completed a comprehensive set of searches for vector‑like quarks (VLQs) and excited fermions using the full Run 2 dataset collected at √s = 13 TeV (2016‑2018), corresponding to an integrated luminosity of 138 fb⁻¹. The paper summarizes three VLQ analyses and two excited‑top (t*) analyses, each employing state‑of‑the‑art machine‑learning techniques and dedicated reconstruction strategies to maximize sensitivity across a broad mass range.

In the VLQ sector, the focus is on a heavy top‑partner T that can decay to a top quark and a new scalar boson ϕ (or the SM Higgs H). Two final‑state topologies are examined: (i) a single‑lepton channel where the top decays leptonically and ϕ/H → b b̄, and (ii) an all‑hadronic channel where both the top and the scalar decay hadronically. The single‑lepton analysis introduces ParticleNet‑MD, a graph‑convolutional neural network that provides mass‑decorrelated tagging of H/ϕ → b b̄, together with a boosted‑decision‑tree top tagger. These tools extend the reach for T masses above 1.3 TeV and allow exclusion of σ·B(T → t ϕ) values as low as 0.15 fb for a light ϕ (25 GeV) and up to 15 fb for a heavy ϕ (250 GeV) across the full 1‑3 TeV T‑mass scan.

The all‑hadronic search uses a two‑dimensional binned maximum‑likelihood fit in the reconstructed (m_T, m_ϕ) plane. For a benchmark ϕ mass of 125 GeV, cross‑section limits range from 4.6 fb (for T ≈ 0.8 TeV) to 300 fb (for T ≈ 3 TeV). Under the weak‑isospin singlet hypothesis with a width of 5 % of the mass, T masses below 1.2 TeV are excluded.

A complementary VLQ search targets particles with charge +2/3 (T) or –4/3 (Y) that decay exclusively to Wb. The analysis exploits the recursive‑jigsaw reconstruction algorithm and the charge asymmetry of the W boson to improve mass resolution. Events are split by lepton charge, and limits are set on the production cross‑section times branching fraction as well as on the coupling κ_W. Assuming a 100 % branching fraction to Wb and κ_W between 0.15 and 0.20, Y masses from 0.7 to 2.4 TeV and T masses from 0.82 to 2.15 TeV are excluded at 95 % confidence level.

The excited‑fermion program investigates heavy top excitations (t*) that can be spin‑½ or spin‑3/2. The dominant decay t* → tg (97 % branching) motivates a pair‑production search in the tg tg final state. Variable‑radius jets are used to capture the boosted gluon jet, and a deep neural network (DNN) classifier, decorrelated from the scalar sum of transverse momenta (S_T) via the DDT technique, provides strong background rejection. The resulting limits exclude spin‑½ t* masses up to 1.05 TeV and spin‑3/2 t* up to 1.70 TeV.

A second excited‑top search focuses on the rarer t* → tγ decay (3 % branching) which yields a clean high‑p_T photon plus large‑radius jets from hadronic top decays. Backgrounds are categorized into prompt photons, jet‑fakes, and electron‑fakes, and a dedicated reconstruction of the tγ invariant mass is performed. This channel achieves comparable sensitivity to the tg tg search, excluding spin‑½ t* up to 0.94 TeV and spin‑3/2 t* up to 1.33 TeV.

Across all analyses, no statistically significant excess over the Standard Model expectation is observed. The paper demonstrates that modern deep‑learning taggers (ParticleNet‑MD, DNN, DDT) and sophisticated kinematic reconstruction (recursive jigsaw, variable‑radius jets) substantially improve the reach compared with earlier Run 2 results based on 35.9 fb⁻¹. The derived limits considerably shrink the viable parameter space of models featuring vector‑like top partners (e.g., composite‑Higgs scenarios) and of compositeness models predicting excited tops. Looking ahead, the techniques validated here will be essential for probing higher mass scales in the upcoming Run 3 and future high‑energy LHC upgrades.