2022 Theses Doctoral

Applying Anomaly Detection to Search for New Physics with the ATLAS Detector at the Large Hadron Collider

Kahn, Alan

A search for a heavy new particle Y decaying to a Standard Model Higgs boson H and another new particle X is presented. The search is performed using 139 fb−1 of p−p collision data at √s = 13 TeV recorded by the ATLAS detector. The H boson is identified through its decays to bb, with the only assumption applied to X being that it decays hadronically. The X is identified through a novel anomaly detection method via the use of a Variational Recurrent Neural Network trained directly on data collected by the ATLAS detector.

This effort marks the first application of a fully unsupervised machine learning method to an ATLAS analysis. An additional benchmark based on interpreting the Y → XH process in the context of a heavy vector triplet model in which the X decays to two quarks defines an additional signal region in which upper limits on the HVT process cross section are reported at 95% confidence level.