The InterLeptons project was successfully carried out at the Austrian Academy of Sciences in Vienna, with the aim of searching for new phenomena beyond the Standard Model of particle physics using data from the Belle II experiment in Japan. Belle II studies high-intensity electron–positron collisions produced by the SuperKEKB accelerator, allowing extremely rare particle decays to be observed with high precision. The project focused in particular on decays involving leptons, such as electrons, muons, and tau particles, which are especially sensitive probes of new physics.
A major challenge of the project was the strong underperformance of the accelerator, which delivered about one hundred times less data than originally foreseen. Faced with this limitation, I adapted the strategy of the project, focusing on extracting the maximum possible information from the available data and developing machine-learning methods to enhance the sensitivity of Belle II at low luminosity.
One key objective was the search for new light particles that could mediate interactions between ordinary matter and dark matter. Such particles would reveal themselves through missing-energy decays, where some of the decay products escape detection. Although no signal was observed, the analysis resulted in world-leading constraints, significantly restricting the possible properties of these hypothetical particles. Several related studies of dark-sector physics were also completed in parallel.
Building on these developments, similar machine-learning techniques were applied to precision tests of lepton flavour universality, a fundamental principle stating that different leptons should decay in the same way, apart from well-understood mass effects. This was studied through leptonic decays of the tau particle, comparing decays to electrons and muons. Despite limited statistics and challenging systematic uncertainties, the final result became the most precise test of this principle to date, confirming its validity at the per-mille level.
In addition to data analysis, I contributed to improving the Belle II detector itself, helping to introduce a machine-learning-based trigger system that enables the experiment to record new classes of rare events, including decays with very simple visible signatures. This development significantly expanded the experiment’s discovery potential.
Finally, intriguing hints observed in certain rare decays involving missing energy, such as decays of B mesons, motivated further investigations and the design of a follow-up program. This new phase extends the methods developed in InterLeptons to a broader set of rare decays in particle physics and, more recently, to gravitational-wave research, where similar challenges arise in identifying weak signals in noisy data. This unified approach to machine-learning-based discovery across different areas of fundamental physics now forms the basis of my next major research project.
