Mattione Reconstruction Uncertainties

Introduction

• I would like to update the current reconstruction uncertainties to improve the particle-ID algorithms and support the kinematic fitting analyses. Right now the tracking uncertainties are yielding dubious chi-squares and kinematic fits with too many events with low confidence levels. Also, the timing uncertainties used for calculating the particle-id FOM are one-size-fits-all numbers, instead of incorporating any momentum or energy dependence.

• I would like to fix these problems with the reconstruction uncertainties, and do so in a framework that can be used for experimental running. That way we will get a better understanding of the eventual experimental reconstruction uncertainties, and we won't need to rewrite the framework later.

Proposal

• During experimental running, I think that the measurement uncertainties should be determined during offline detector system calibrations. Specifically, I think that the hit time and hit energy of the detector systems (BCAL, FCAL, TOF, etc.) should be determined from calibrations. The hit position uncertainties can be determined using these uncertainties, the known detector geometries, and the drift chamber residuals for the tracking. Determining the uncertainties at the hit-level (where possible) instead of at the shower-level allows localized detector inefficiencies to be taken into account (weak PMTs, etc.). These uncertainties can then be directly propagated through the reconstruction code for the particle-id FOM determination (for both charged and neutral particles) and kinematic fitting. However, if you think there are systems for which it would be more appropriate to determine the uncertainties at the shower-level, please let me know!

Current Uncertainty Status

• No detector systems have hit uncertainties directly implemented.
• For particle-id, one-size-fits-all values are used for the time uncertainties from the TOF, FCAL, and BCAL systems.
• The BCAL & FCAL systems have shower uncertainties, but there are no correlations between them (in the lab frame). For the FCAL, the shower time uncertainty is absent. Also, for the neutral tracks the uncertainties are all overridden later by a 7x7 diagonal (no correlations) error matrix.
• The start counter times are smeared with a fixed sigma of 300 ps in mcsmear.
• The tracking code (DTrackFitterKalmanSIMD) incorporates CDC and FDC hit uncertainties determined from fits to the simulated drift chamber residuals. These uncertainties are then propagated throughout the tracking code and into the final tracking covariance matrix.

Workflow

• Get feedback from the offline group as to whether the measured time & energy uncertainties should be implemented at the detector hit-level, and the position uncertainties from these and geometric considerations.
• Integrate these uncertainties into the reconstruction code, using mcsmear to set the simulated uncertainties at the hit-level where applicable.
• If possible, get help from detector system experts on propagating these uncertainties through their systems (e.g. FCAL hit uncertainties -> FCAL shower uncertainties).
• Propagate these uncertainties to the chi-squared and FOM calculations.