New Jet Substructure Techniques at ATLAS

The Q-jets technique introduces the idea of interpreting jets through multiple sets of possible showering histories. This approach allows jet observables, such as the jet mass, to be evaluated not simply as single values, but rather as distributions. The resulting distributions can be interpreted statistically to form new observables, allowing the separation of boosted, hadronically-decaying particles from light quark and gluon backgrounds. We present a study of Q-jets in boosted, hadronically-decaying $W$ boson and dijet samples, demonstrating the discriminating power of this technique. Different Q-jet parameters and observables are studied, and an optimal configuration based on physics performance and computational efficiency is proposed, leading to a factor of 15 in dijet rejection at a 50\% efficiency for jets from boosted, hadronically decaying $W$ bosons. The impact of pile-up on the performance of this method is tested up to an average of 40 additional interactions per event and found to be weak. A performance comparison between the Q-jets algorithm and $N$-subjettiness, a previously measured substructure observable which determines the compatibility of a jet with the $N$-subjet hypothesis, is presented.