On the reduction of negative weights in MC@NLO-type matching procedures

We show how a careful analysis of the behaviour of a parton shower Monte Carlo in the vicinity of the soft and collinear regions allows one to formulate a modified MC@NLO-matching prescription that reduces the number of negative-weight events with respect to that stemming from the standard MC@NLO procedure. As a first practical application of such a prescription, that we dub MC@NLO-$\Delta$, we have implemented it in the MadGraph5_aMC@NLO framework, by employing the Pythia8 Monte Carlo. We present selected MC@NLO-$\Delta$ results at the 13 TeV LHC, and compare them with MC@NLO ones. We find that the former predictions are consistent with the latter ones within the typical matching systematics, and that the reduction of negative-weight events is significant.
Comment: 44 pages, 13 figures; several minor text amendments and one figure added