This paper reexamines a recent ATLAS search for doubly charged Higgs bosons using the full Run 2 dataset at 13 TeV and finds that the four-lepton part of that search appears to overstate how many signal events it should see. The four-lepton channel is nearly free of background and so strongly controls the reported mass limits. The authors derive a strict, model-independent upper bound on how often pair-produced doubly charged scalars can produce four reconstructed electrons or muons, and they find the ATLAS auxiliary cutflow reports larger fractions than that bound allows.

Doubly charged Higgs bosons are hypothetical particles that would decay into two same-sign charged leptons (electrons, muons, or taus). The ATLAS result was interpreted assuming equal branching fractions into the six leptonic final states (each set to 1/6). Using the known tau leptonic branching fractions (tau→e: 0.18, tau→μ: 0.17, total leptonic tau 0.35) and the ATLAS lepton reconstruction efficiencies (εe = 0.88, εμ = 0.95), the authors compute per-arm probabilities f(ee)=0.20, f(eμ)=0.24, f(μμ)=0.20 and find a truth-level four-light-lepton fraction (before detector effects) Ptruth4 = 0.41. Folding in reconstruction efficiencies gives a strict upper bound on the fraction of pair-production events that can yield four reconstructed electrons or muons of Preco4 = 0.29.

The ATLAS auxiliary table quoted cumulative retained fractions for the four-lepton signal that are larger than these bounds. Across four benchmark masses (700, 900, 1100, 1300 GeV) the fractions quoted are about 0.50–0.51 for a ‘‘loose’’ selection and 0.43–0.46 for a ‘‘tight’’ selection. The ‘‘tight’’ numbers even exceed the truth-level maximum of 0.41, so the discrepancy cannot be explained simply by imperfect modeling of lepton reconstruction. The authors also argue that filling the gap with misidentified hadronic taus or jets would require unrealistically large fake rates.