The inseparable three and four tops

In measurements of four-top-quark production ($tttt$), LHC collaborations observe a significant degeneracy with three-top-quark production. We compute the dominant three-top-production mode, namely associated production with a $W$ boson ($tttW$), at complete next-to-leading order (NLO), including all possible QCD and electroweak (EW) corrections. Beyond leading order (LO), $tttW$ production with the radiation of an additional $b$-flavoured quark contributes to the same final state as $tttt$ production with a $t \to bW$ decay. Away from the on-shell top-quark limit, the usual overlap removal of resonant contributions in the non-resonant computation either breaks gauge invariance and generates unitarity violation or involves a significant arbitrariness in the required reshuffling of momenta. To overcome these issues, we introduce a novel window-removal prescription that produces consistent predictions for the inseparable $tttW+tttt$ process, with both components described at NLO accuracy. We argue that such a joint prediction should be used in comparisons with experimental selections targeting $tttt$ production, since the on-shell $tttt$ component can not be isolated in practice. Such a joint prediction has an inclusive rate more than 10% higher than the purely on-shell $tttt$ one. We also study an idealised veto on additional hard and central $b$-jet radiation, which suppresses the contributions of resonant $tttt$ diagrams as well as their interference with non-resonant $tttW$ ones and therefore defines a relatively pure $tttW$-like signal region. Formally subleading coupling orders are numerically important at LO, while the corresponding subleading NLO corrections largely cancel both inclusively and differentially. Consequently, the complete-NLO prediction is well approximated by retaining the first three LO coupling orders together with the leading QCD NLO correction.