Probes of flavour symmetry and violation with top quarks in ATLAS and CMS

Results are presented of searches and measurements in the top quark sector by the ATLAS and CMS experiments. These analyses use data from proton-proton collisions at a centre-of-mass energy of 13 TeV, recorded during Run 2 at the Large Hadron Collider and corresponding to integrated luminosities of 138-140 fb$^{-1}$. Searches are carried out for charged lepton flavour violation, baryon number violation and the presence of neutral heavy leptons. A precise measurement of lepton flavour universality between electrons and muons originating from top quark-antiquark events is also presented.

💡 Research Summary

The paper presents a comprehensive set of searches and precision measurements performed by the ATLAS and CMS collaborations using the full Run 2 proton‑proton dataset at √s = 13 TeV (ATLAS 140 fb⁻¹, CMS 138 fb⁻¹). The focus is on probing flavour symmetries and possible violations in processes involving top quarks.

1. Lepton Flavour Universality (LFU) in W decays – ATLAS measures the ratio of W‑boson branching fractions to muons and electrons, B(W→μν)/B(W→eν), by exploiting t t̄ dileptonic events and inclusive Z→ℓℓ samples. By re‑weighting muon kinematics to match electrons, measuring lepton isolation efficiencies in‑situ, and performing a simultaneous likelihood fit to e μ, e e, and μ μ yields, the double ratio R_{μ/e}^{WZ} is extracted. Using the precisely known Z‑branching ratio from LEP/SLD, the final result is R_{μ/e}^{W}=0.9995 ± 0.0045, consistent with the Standard Model and representing the most precise test of e‑μ universality to date (0.45 % precision). Dominant uncertainties stem from PDFs, modelling, and lepton systematics.

2. Charged Lepton Flavour Violation (cLFV) – Three complementary analyses target cLFV in top‑quark production or decay, interpreted in a model‑independent Effective Field Theory (EFT) framework with dimension‑6 four‑fermion operators (vector, scalar, tensor).

   • CMS e μ trilepton: A Boosted Decision Tree (BDT) is trained separately for low (t t̄ decay) and high (single‑top production) invariant‑mass regions (m_{eμ}=150 GeV split). Each Wilson coefficient C/Λ² is fitted independently using binned likelihoods. No excess is observed; 95 % CL limits on |C|/Λ² range from 0.02 GeV⁻² (tensor) to 0.45 GeV⁻² (scalar). Corresponding branching‑fraction limits for t→qℓℓ′ are at the 10⁻⁶ level (e.g. u‑quark: 0.012–0.032 × 10⁻⁶).
   • ATLAS μ τ trilepton (τ_h): Hadronic τ decays are selected; the scalar sum of transverse momenta (H_T) serves as the discriminant. No signal is found, leading to limits that improve previous indirect bounds by factors of 7–41. The same dataset is reinterpreted for a scalar leptoquark S₁ model with hierarchical couplings; couplings λ_{LQ}=1.3–3.7 are excluded for masses 0.5–2 TeV at 95 % CL.
   • CMS μ τ hadronic: Both the W boson and τ lepton decay hadronically. A multiclass Deep Neural Network (28 kinematic inputs) classifies production‑cLFV, decay‑cLFV, and t t̄ background. Limits on Wilson coefficients improve by roughly a factor of two relative to earlier μ τ analyses, with branching‑fraction constraints again at the 10⁻⁶–10⁻⁸ level.

3. Baryon Number Violation (BNV) – CMS searches for t q q′ ℓ vertices that simultaneously violate baryon number and lepton number. Single‑top production and t t̄ decay topologies are considered, with dileptonic final states (e e, e μ, μ μ) and at least one b‑tagged jet. A dedicated BDT extracts the signal. No excess is observed; 95 % CL upper limits on the branching fractions of BNV top decays lie between 10⁻⁶ and 10⁻⁸, improving upon the 8 TeV results by 10³–10⁶.

4. Heavy Neutral Leptons (HNL) – ATLAS conducts the first search for HNLs produced in top‑quark decays: t → Nℓ, N → Wℓ, with N being a Majorana neutrino of mass 15–75 GeV. The signature consists of two same‑sign leptons (e e or μ μ) and two hadronically decaying W bosons. Separate BDTs for low and high HNL masses are trained, and profile likelihood fits are performed in the e e and μ μ channels (ττ channels are reinterpreted). The analysis extends existing ATLAS limits above m_N = 50 GeV, setting competitive constraints on the mixing parameter |V_{ℓN}|² as a function of the HNL mass.

5. Conclusions and Outlook – The combined ATLAS and CMS programme demonstrates that top‑quark processes provide powerful probes of flavour symmetries. LFU is confirmed at the sub‑percent level, while cLFV, BNV, and HNL searches set branching‑fraction limits in the 10⁻⁶–10⁻⁸ range. The EFT approach enables model‑independent constraints on dimension‑six operators, facilitating reinterpretation in a wide variety of BSM scenarios. With the upcoming Run 3 dataset and advances in machine‑learning techniques, further improvements in sensitivity and precision are expected, potentially revealing subtle deviations from the Standard Model.