Downstream vertexing and two-track downstream lines for Lambda -> P Pi and Ks0 -> Pi Pi
Need lhcb-datapkg/ParamFiles!80 to be merged first to run CI/CD test in GitLab
Introduction
This MR introduces a downstream tracks vertexing algorithm and HLT1 lines for long-lived particles based on it, utilizing the brand-new downstream reconstruction algorithm (!1095 (merged)).
This MR involves:
- Adding supplementary information to downstream track particles, such as IPChi2 and the magnet correction factor.
- A new downstream tracks vertexing algorithm: pairing downstream track particles to create downstream composite particles.
- A new composite particle validator, suggested by @dcampora, to check the quality of composite particles in Allen.
- Two HLT1 downstream track lines, one for triggering Ks0 and another for Lambda0.
1. Additional Information to Downstream Tracks Particles
The following additional information is added:
- The IPChi2 is computed and used to select the best PV for each downstream track particle and downstream composite particle.
- The magnet correction factor is computed to extrapolate the downstream tracks from UT to VELO, taking into account the non-negligible magnet effect on UT.
- The fit uncertainties of downstream track states are explicitly obtained as a function of qop from MC simulation, which is essential for the vertexing algorithm.
2. Downstream Tracks Vertexing Algorithm
The new vertexing algorithm for downstream tracks considers the non-linear extrapolation of downstream tracks from UT to VELO. This algorithm will substantially improve the mass resolution of composite particles and provide a fit chi2, which is highly discriminant for background rejection in selection lines.
The non-linear extrapolation is described as follows:
x(z) = x0 + tx (z-z0) + \gamma (z-z0)^{2}
y(z) = y0 + ty (z-z0)
where \gamma = \gamma(qop).
The vertexing algorithm first estimates the POCA for two tracks, then extrapolates the state non-linearly to the POCA z position, and finally performs a traditional vertexing algorithm, assuming the extrapolation is linear around this point.
3. New Composite Particle Validator
The new Allen checker is designed to validate the truth-matching categories of composite particles. The definitions are as follows:
- '-1'=> not defined (this is just a placeholder)
- '0' => true composite
- '1' => low energy composite, one of the tracks has bremsstrahlung radiation
- '2' => combinatorial background, two tracks have different mothers
- '3' => partial ghost, one of the tracks is a ghost
- '4' => ghost, both tracks are ghosts
This new checker also provides a tuple for essential information of composite particles such as p, pt, ip_chi2, doca, chi2, etc. This information is used to develop the selection lines.
4. Two HLT1 Downstream Track Lines
The Ks0ToPiPi and L0ToPPi lines are developed to select the DD Ks0 and DD Lambda0. Both lines are based on a single hidden layer fully-connected neural network, trained using the tuple produced by the composite particle validator.
The training sample is /MC/Upgrade/Beam7000GeV-Upgrade-MagDown-Nu7.6-25ns-Pythia8/Sim10aU1/13104012/XDIGI, with the training variables being:
- Track: P, PT, IPChi2
- Composite: Chi2, DOCA, p, pt, ovtx_z, IPChi2
The performance of these lines is verified with upgrade_mc_minbias_scifi_v5_retinacluster_000_v1_newLHCbID_new_UT_geometry.mdf in the online machine:
| Ks0 | Lambda |
|---|---|
 |
|
Armentero Podolanski Plot (Using all downstream composites + loose mass window cuts)
Expect efficiency impact with new downstream lines
Checking with HLT1 Dumpped Data
Note: We use a very loose threshold to make these plots with enough statistics.
| Lambda | Kshort |
|---|---|
 |
 |