Providing fast information for a real-time event selection (“trigger”) is crucial for an efficient event recording in high-energy-physics experiments. Triggering on charged particles (“tracks”) at the first deadtime-free (“pipelined”) trigger level, in particular, is becoming more and more important for high selectivity at the extreme luminosities envisaged at future particle colliders. We propose here a new approach for deadtime-free track triggering, emphasizing the track origin (“vertex”), based on artificial neural networks. Due to their inherent parallel architecture, neural networks are ideal for super-fast execution. Our preliminary studies, concentrating on the estimation of the vertex for single-track events without explicit reconstruction of the tracks, showed that this new approach is suitable for a fixed-latency first level trigger. The vertex determination can be used for a fast rejection of events not coming from the interaction point, for example in electron-positron experiments, or for secondary vertex determination in hadron collider experiments. To study the ultimate timing performance in a pipelined mode, this new approach needs to be implemented in hardware, for instance on FPGAs, exploiting the inherent parallelism of neural computation. The methods to be developed will be of great importance for future track triggering concepts at Belle II and the high luminosity upgrade of the LHC detectors. In addition, the hardware and the algorithms could also be of use to applications beyond particle physics.
Search for New Physics at High Luminosities
Searching for physics beyond the Standard Model (SM) using accelerator-based experiments requires the investigation of small deviations from the SM expectation and of extremely rare processes. In either case high instantaneous luminosity is mandatory, leading to large total event rates. Here, the performance of the trigger, enriching the recorded data sample with the events of interest, is crucial for a successful search for New Physics. The present generation of experiments at the LHC, such as the multi-purpose detectors ATLAS and CMS, currently use information only from the calorimeters and the muon chambers in order to meet the stringent timing requirements of the deadtime-free first level trigger. The selection mechanism can, however, be made much more powerful if information from the tracking detectors is used already at this stage. In fact, the silicon trackers of the next generation will provide some “intelligence”, allowing the implementation of track-based selection algorithms at the first trigger level. The need for providing fast track information for event selection is not limited to the LHC experiments. The future Belle II experiment at the SuperKEKB electron-positron collider will face a similar challenge, mainly because of the high level of background that is expected once the design luminosity is reached. The project described in the following is motivated by the requirements of Belle II. However, its basic concept is general enough to produce generic results that will be useful also for the upgraded experiments at the future high-luminosity LHC.
The Belle II Experiment at SuperKEKB
Belle II [1] is an experiment on the asymmetric electron-positron collider SuperKEKB [2,3], which is currently under construction at the KEK laboratory in Tsukuba, Japan. Belle II is an upgrade of the Belle experiment [4], which was instrumental in exploring CP violation in the B meson system. The success of this program has led to the rapid approval of an upgrade for both the detector and the KEKB collider. The new facility, Belle II at SuperKEKB, aims for an amount of recorded physics events fifty times larger than in Belle, i.e., 50 ab −1 at the energy of the Υ(4S). To reach this goal, the instantaneous luminosity will be exceeding L = 0.8 × 10 36 cm −2 s −1 , 40 times larger than the world record achieved by KEKB. The pure physics rate at this luminosity is around 10 kHz. An unfortunate side effect of the high luminosity is a much higher level of machine background, dominated by Touschek scattering [5, 6]. This produces a high rate of undesirable background events with vertices outside of the nominal interaction region, where the physically interesting reactions from the e + e − collisions are produced. As an illustration, Fig. 1 shows the distribution of interaction vertices in the beam direction z as measured in Belle. The wide background around the narrow peak at z = 0 cm, marking the interaction point, is clearly visible. Since Belle had no fast detection of the z-vertex at the first trigger level, these events could not be rejected there. Note that the level of this background is expected to be much higher at SuperKEKB. The showcase of the proposed project is the reduction of the Touschek scattering background by using track information from the central drift chamber (CDC) of Belle II. This requires a precise estimate (in the order of 1 to 2 cm) of the event z-vertex in real time, sufficiently fast for the first level trigger.