Overview
The LHCb experiment has implemented a fast Kalman filter within its first-level trigger system to perform track fitting of charged-particle trajectories. This system operates at the full LHCb collision frequency of 30 MHz, which is critical for the LHC Run 3 data taking period. The approach replaces traditional magnetic field map lookups and numerical integration with analytical parameterizations, while retaining the mathematical framework of Kalman filtering.
Research Context
Prior to the LHC Run 3 data taking period, the LHCb experiment at the Large Hadron Collider underwent a significant upgrade. This upgrade included the adoption of an all-software approach for its trigger system. The presented Kalman filter implementation forms a key component of this upgraded system, specifically in the first trigger stage, and has been in use since the 2025 data taking period. It replaced a previous partial track fitting algorithm within the real-time trigger environment.
Approach
The methodology centers on a fast implementation of a Kalman filter. This implementation was designed to determine parameter estimates of charged-particle trajectories. A core aspect of this approach is the use of fast analytical parameterizations. These parameterizations serve to replace computationally intensive magnetic field map lookups and numerical integration methods, while preserving the underlying mathematical structure of Kalman filtering. The algorithm is deployed on approximately 500 GPUs within the first-level trigger of the LHCb experiment.
Findings
- The adapted Kalman filter operates at the full LHCb collision frequency of 30 MHz.
- It superseded the prior partial track fitting algorithm in the real-time trigger environment.
- The implementation incurred a 2% increase in processing time compared to the previous partial track fitting algorithm.
- The parameterised Kalman filter demonstrated a significantly improved momentum resolution.
- This improvement in momentum resolution resulted in a factor of two enhancement in the invariant mass resolutions for reconstructed D0 and J/ψ hadrons.
- The system displayed greater robustness against detector misalignment effects.
- It substantially sharpened the discrimination between genuine particle trajectories and accidental background.
- Specifically, at a standard selection working point, the algorithm achieved more than double the rejection of accidental background without compromising genuine-track efficiency.