Wavelet Analysis: Event De-noising, Shower Evolution and Jet Substructure Without Jets

Abstract: Wavelet decomposition is a method that has been applied to signal processing in a wide range of subjects. The decomposition isolates small scale features of a signal from large scale features, while also maintaining information about where in the signal those features occur. Wavelets obey particular scaling relations and are especially suited to the analysis of systems that are self-similar and scale invariant. They are therefore a natural tool for the study of hadron collisions. This paper introduces the use of wavelets for de-noising (removal of soft activity), studying the scaling behaviour of a shower, and recognising jets according to this behaviour. This is demonstrated by processing a sample of boosted W boson Monte Carlo events together with their QCD background. The method is quite general and can be used as a pre-processing step in conjunction with any jet-finder or other event-shape algorithm. The result in this simple example is a significant enhancement in both the size and shape of the W boson mass peak, together with an improved separation of the background distribution.