Search for tau neutrino double cascades with 6 years of IceCube cascade data.

IceCube has discovered the flux of astrophysical high energy neutrinos of all flavors and characterized it. High energy (TeV-PeV) tau neutrinos contribute to the flux of astrophysical neutrino induced cascades, dominated by electron and tau neutrino interactions as measured with IceCube 6 years of cascade data. The identification of tau neutrinos is however challenging. This is due to IceCube strings and digital optical module spacing, where (low energy) tau neutrinos interacting via (charged) neutral current within detector volume are not distinguishable from electron neutrinos and both produce so-called single cascade signature. At high energies, in tau neutrino charged current interactions a high energy tau lepton is produced that decays predominantly into hadrons, as well as an electron-neutrino-electron pair. The event signature of such tau neutrinos is double cascade. We have developed an analysis which utilizes xgboost machine learning technique, utilizing single and double cascades reconstruction methods to detect tau neutrinos and have applied it to cosmic-ray background free 6 years of cascade data. This analysis is complementary to other tau neutrino searches with IceCube which identified a few tau neutrino candidates. In this talk we will present a method that improves double cascade tau neutrino identification. This method offers a high signal purity as well as maintains a good tau neutrino reconstruction quality.