Speaker
Description
At the Large Hadron Collider, the ATLAS detector experiences a collision rate of about 1 billion proton-proton collisions per second. This rate of collisions is far too large for us to store all observed events, so only interesting events are stored. The ATLAS trigger system reduces this input rate to a manageable 3 kHz via the use of the hardware-based Level 1 trigger and the software-based high level trigger (HLT). Neural networks classifiers have previously been used in the identification of b-quark initiated jets in order to further reduce the rate within the HLT. Here, a similarly structured classifier has been used for hadronically decaying tau leptons. A first iteration of this classifier has been active since early 2024, achieving a rate reduction of 32% for a 𝑏 + 𝜏 trigger chain used in the 𝐻𝐻 → 𝑏𝑏𝜏𝜏 analysis. Further development of this hadronic tau classifier has yielded an additional 46% rate reduction.