The neural network first-level hardware track trigger of the Belle II experiment

We describe the principles and performance of the first-level (“L1”) hardware track trigger of Belle II , which uses the information of Belle II ’s Central Drift Chamber (“CDC”) and provides three-dimensional track candidates based on neural networks. The inputs to the networks are “2D” track candidates in the plane transverse to the electron–positron beams, obtained via Hough transforms, and selected information from the stereo layers of the CDC. The networks then provide estimates for the origin of the track candidates in direction of the colliding beams (“$z$-vertex”), as well as their polar emission angles $\theta$. Using a suitable cut $d$ on the $z$-vertices of the “neural” tracks allows us to identify events coming from the collision region ($z\approx 0$), and to suppress the overwhelming background from outside. Requiring $\left|z\right|<d$ for at least one neural track in an event with two or more 2D candidates will set an L1 track trigger. The networks also enable a minimum bias trigger, requiring a single 2D track candidate validated by a neural track with a momentum larger than 0.7 GeV in addition to the $\left|z\right|$ condition. We also sketch our concepts for upgrading the neural trigger in view of rising instantaneous luminosities, accompanied by increasing backgrounds.