The LCFIVertex Package: vertex detector-based Reconstruction at the ILC

The contribution gives an overview of the LCFIVertex package, providing software tools for high-level event reconstruction at the International Linear Collider using vertex-detector information. The package was validated using a fast Monte Carlo simulation. Performance obtained with a more realistic GEANT4-based detector simulation and realistic tracking code is presented. The influence of hadronic interactions on flavour tagging is discussed.

💡 Research Summary

The paper presents the LCFIVertex software package, a C++‑based toolkit designed for high‑level event reconstruction at the International Linear Collider (ILC) that exploits the excellent spatial resolution of the ILC vertex detector. The package implements the ZVTOP family of topological vertex finders originally developed for the SLD experiment, providing both the ZVRES and ZVKIN branches. ZVRES constructs a three‑dimensional vertex function from “probability tubes” representing tracks and searches for its maxima, while ZVKIN introduces a “ghost track” approximating the flight direction of a B‑hadron to recover one‑prong and short‑lived B decays that would otherwise be missed.

Flavor tagging is performed with a neural‑network approach based on the algorithm of R. Hawkings. LCFIVertex incorporates a full neural‑network library written by D. Bailey, allowing users to modify input variables, network architecture, and training method. By default the network uses a set of variables that have proven effective in previous studies: the impact‑parameter significance of the most significant track, the joint probability that all tracks originate from the primary vertex, the vertex multiplicity within a jet, and the Pt‑corrected vertex mass. The package also includes tools for determining the heavy‑quark charge, extending the SLD charge‑dipole technique to neutral hadrons via the ZVKIN information.

The validation proceeds in two stages. First, a fast Monte‑Carlo simulation (SGV) is used to compare LCFIVertex against the legacy FORTRAN implementation that had been employed with the BRAHMS reconstruction chain for the TESLA TDR. Using identical e⁺e⁻ → qq̄ events at the Z‑pole, the purity‑efficiency curves for b‑ and c‑tagging are found to be in excellent agreement, confirming that the new C++ code reproduces the physics of the older version.

Second, a more realistic study employs the GEANT4‑based full detector simulation Mokka (version 0.6‑03) together with the LDC tracking package and MarlinReco for reconstruction. The detector model (LDC01Sc) features cylindrical vertex‑detector layers; photon conversions are switched off, and tracks from hadronic interactions in the beam pipe and detector material are suppressed using Monte‑Carlo truth information. Tracks are digitised with a simple Gaussian smearing, fed into the Wolf particle‑flow algorithm, and clustered into jets with the Durham kₜ algorithm (y‑cut = 0.04). The same “track cheater” (truth‑based hit assignment) is used for the baseline, and it is shown that replacing it with a realistic pattern‑recognition algorithm does not significantly alter the tagging performance.

The full‑simulation results show that LCFIVertex matches or slightly exceeds the performance obtained with the BRAHMS chain. The improvement is attributed to a finer assumed vertex‑detector resolution (2 µm versus 3.5 µm), and to the explicit suppression of K⁰_S, Λ decays, photon conversions, and hadronic‑interaction tracks. Figure 3 demonstrates the enhanced b‑ and c‑tagging purity at the Z‑pole, while Figure 4 illustrates the degradation of b‑tagging at √s = 500 GeV when hadronic‑interaction tracks are retained; at the Z‑pole the effect is negligible.

In summary, LCFIVertex provides a flexible, well‑documented implementation of ZVTOP vertex finding, Hawkings‑style neural‑network flavor tagging, and heavy‑quark charge determination. Validation against both fast and full simulations confirms its reliability, and the package is ready for integration into ILC physics analyses. Future work will need to replace the current MC‑based suppression of hadronic‑interaction tracks with a data‑driven approach and to incorporate more sophisticated pattern‑recognition algorithms, but the current version already offers a solid foundation for precision measurements at the ILC.