The project, carried out at the Georg‑August University of Göttingen under the leadership of Prof. Dr. Steffen Schumann, aimed to advance the SHERPA event generator for high‑energy collider physics. Funded by the German Federal Ministry of Education and Research for the period 1 July 2018 to 30 June 2021 with a budget of €128 809.87, it was part of the ErUM‑FSP T02 consortium that supports the ATLAS experiment at the Large Hadron Collider. The work was carried out by three doctoral students—Stephan Bräuer, Daniel Reichelt, and Simon Luca Villani—supported by a visiting student, Timo Janssen, and involved close collaboration with the ATLAS Analysis Group Siegert at TU Dresden and with the broader theoretical community, including groups working on POWHEG and MadGraph5_aMC@NLO.
Technically, the project delivered several significant improvements to SHERPA. First, electroweak corrections were systematically incorporated into parton‑shower simulations. The virtual approximation of electroweak effects was extended and applied to reference processes such as Higgs‑boson pair production, enabling next‑to‑leading‑order accuracy in both the strong and weak couplings while retaining the parton‑shower framework. Real photon emissions were treated with the soft‑photon approximation of Yennie–Frautschi–Suura. Parallel to this, an electroweak Sudakov approximation was implemented and validated against the virtual method for Z‑boson pair production, providing an alternative, faster route to electroweak corrections. These developments directly increase the precision of Standard Model predictions used by ATLAS and CMS.
The project also explored machine‑learning techniques to enhance event generation efficiency. Normalising flows—a class of bijective neural‑network mappings—were investigated for phase‑space sampling. Early results show promising gains in unweighting efficiency, suggesting a viable path toward faster, higher‑statistics event samples. In addition, SHERPA was extended to perform semi‑analytic resummation of soft‑gluon emissions to next‑to‑leading logarithmic accuracy. This enabled the calculation of global event‑shape observables, such as Soft‑Drop‑groomed thrust, and jet‑substructure variables, including jet angularity. Comparisons of these predictions with CMS measurements demonstrated good agreement, validating the new resummation implementation and providing benchmarks for future analyses.
The outcomes of the project were released as open‑source SHERPA versions, ensuring immediate availability to the experimental collaborations. The new releases were integrated into ATLAS production software and analysis frameworks, and the improved electroweak and resummation modules are now routinely used in Standard Model studies. The project’s contributions also enriched the theoretical toolkit, offering the first multi‑purpose event generator implementation of machine‑learning‑based phase‑space sampling and a robust electroweak correction framework that can be adopted by other generators.
Overall, the project achieved its goals of enhancing SHERPA’s physics reach and computational performance, while fostering collaboration across experimental and theoretical communities. The work has strengthened the precision of collider predictions, supported the ATLAS and CMS analyses, and laid groundwork for future developments in event‑generation technology.