F. Baptist | Searcharxiv

arXiv2026

Commissioning and Low Latency Operation of the Graph Neural Network Electromagnetic Calorimeter Trigger at the Belle II Experiment

We present the commissioning and operation of the Graph Neural Network Electromagnetic Calorimeter Trigger Module (GNN-ETM) of the Belle II experiment at the SuperKEKB collider. The GNN-ETM processes calorimeter trigger cells as graph nodes to perform clustering and feature extraction. We fully integrate the system with the successive stages of the first-level trigger, develop slow-control drivers, and add online monitoring capabilities. We optimise the existing FPGA-based architecture through hardware-algorithm co-design, achieving an overall system latency of 1.053 us. Our hardware implementation is validated through register-transfer-level simulations, achieving bit-accurate agreement with the offline reference model. Online monitoring enables the measurement of instantaneous trigger rates, providing a quantitative basis for trigger-level performance studies. In summary, we report on the GNN-ETM as a fully operational, low-latency trigger module with online control and monitoring capabilities, compatible with the latency requirements of the Belle II first-level trigger system.

hep-ex↗

arXiv2026

Real-time graph neural networks on FPGAs for the Belle II electromagnetic calorimeter

We present the development and evaluation of a real-time Graph Neural Network-based trigger module for the electromagnetic calorimeter of the Belle~II experiment at the SuperKEKB collider. The algorithm processes calorimeter trigger cells as graph nodes to perform clustering, feature extraction, and per-cluster signal classification with deterministic latency. The model predicts cluster positions and energies and provides a signal classification score, enabling a more flexible clustering strategy than the baseline trigger algorithm. Implemented on an FPGA and integrated into the Belle~II trigger readout infrastructure for synchronous operation, the system sustains the MHz trigger throughput with an end-to-end latency of $3.168\,\mu$s. The performance is evaluated on simulated events and collision data. The energy resolution is comparable to the baseline trigger, while the position resolution for high-energy clusters improves by up to 18% in the central detector region. Cluster purity increases by up to 20% at low energies for isolated clusters, and cluster efficiency improves by up to 20% for overlapping clusters. The signal classifier enables additional background suppression at fixed signal retention. These results demonstrate a first step towards GNN-based real-time reconstruction on FPGAs in a collider trigger. While the end-to-end latency exceeds the trigger decision budget, the system already sustains full operational conditions with 100% uptime.

physics.ins-det↗