Scientists have publicly released a technique called Born‑spreading that assigns shower‑starting scales consistently for two popular merging methods, MC@NLO and FxFx. The method reduces the fraction of negative‑weight events in next‑to‑leading‑order (NLO) generators without biasing the physics predictions.

Born‑spreading redistributes the Born contribution over radiation variables, concentrating the negative part of the integrand. A single reference scale and a damped probability function D(q; μ1, μ2) define the shower‑starting scales, colour flows and counterterms. The same quantity also sets the start of the Sudakov re‑weighting in FxFx merging.

When applied to W+jets production with up to two jets, the new construction cuts the fraction of negative weights after merging and speeds up event generation by a factor of six compared with the earlier implementation. Comparative studies show that the fraction of negative‑weight events is halved for processes like Drell–Yan and Higgs production and reduced by up to three percentage points for top‑quark processes.

The technique incurs a fixed CPU overhead of 13–31 % for larger samples, mainly from fitting the ρ function and grid adaptation. This cost is smaller for very large event samples, and it does not affect the physical results.

These results demonstrate that Born‑spreading, now open‑source, improves the efficiency of LHC event simulations while preserving accuracy.