21. Tracking


This documentation is under construction!

The main goal of the tracking reconstruction is to find and fit tracks using the hit informations provided by the main tracking sub-detectors: PXD, SVD, CDC. In general, tracking consists in two steps:

  1. pattern recognition, or track finding: recognise the hits belonging to a single charged particle

    See also

    The Track Finding at Belle II provides a nice introduction to Belle II tracking and describe the track finding algorithms and their performance on simulated events.

  2. track fit: extract the track parameters from a fit to the pattern (collection of hits).

The tracking reconstruction chain is built as a modular system that can be adapted to different background levels and detector performance. We combine stages of pattern recognition and fitting for the different detectors in order to improve efficiency and minimise fakes and clones.

SVD and CDC provide hits for the first steps of the pattern recognition. In order to optimise the performance of the pattern recognition we have different algorithms for the CDC and for the SVD. Hits in the PXD are added to existing tracks to improve the track quality.


Beside the offline reconstruction, the tracking reconstruction is performed also online on the High Level Trigger (HLT): the two reconstructions are slightly different. Beside the fact that PXD hits are not available on HLT, tracking on HLT is optimized to satisfy the execution time and memory limits with no significant impact on the physics performance.

Use add_tracking_reconstruction to append the tracking reconstruction to your path:

tracking.__init__.add_tracking_reconstruction(path, components=None, pruneTracks=False, skipGeometryAdding=False, mcTrackFinding=False, trackFitHypotheses=None, reco_tracks='RecoTracks', prune_temporary_tracks=True, fit_tracks=True, use_second_cdc_hits=False, skipHitPreparerAdding=False, use_svd_to_cdc_ckf=True, use_ecl_to_cdc_ckf=False, add_cdcTrack_QI=True, add_vxdTrack_QI=False, add_recoTrack_QI=False)[source]

This function adds the standard tracking reconstruction modules to a path:

  1. first we find tracks using the CDC hits only, see CDC Track Finding

  2. CDC tracks are extrapolated to SVD and SVD hits are attached, see CDC to SVD CKF

  3. remaining SVD hits are used to find SVD tracks, see SVD Track Finding

  4. SVD tracks are extrapolated to CDC to attach CDC hits, see SVD to CDC CKF

  5. SVD and CDC tracks are merged and fitted, see Track Fitting

  6. merged SVD+CDC tracks are extrapolated to PXD to attach PXD hits, see SVD to PXD CKF


PXD hits are not available on HLT. At the end of the tracking chain on HLT we have the PXD Region Of Interest Finding, that consists of extrapolating the tracks on the PXD sensors and defining regions in which we expect to find the hit. Only fired pixels inside these regions reach Event Builder 2.

  1. after all the tracks from the IP are found, we look for special classes of tracks, in particular we search for displaced vertices to reconstruct K-short, Lambda and photon-conversions, see V0 Finding


this last step is not run on HLT

  • path – the path to add the tracking reconstruction modules to

  • components – the list of geometry components in use or None for all components.

  • pruneTracks – if true, delete all hits except the first and the last in the found tracks.

  • skipGeometryAdding – (advanced flag) the tracking modules need the geometry module and will add it, if it is not already present in the path. In a setup with multiple (conditional) paths however, it can not determine, if the geometry is already loaded. This flag can be used o just turn off the geometry adding at all (but you will have to add it on your own then).

  • skipHitPreparerAdding – (advanced flag) if true, do not add the hit preparation (esp. VXD cluster creation modules. This is useful if they have been added before already.

  • mcTrackFinding – if true, use the MC track finders instead of the realistic ones.

  • reco_tracks – name of the StoreArray where the reco tracks should be stored

  • prune_temporary_tracks – if false, store all information of the single CDC and VXD tracks before merging. If true, prune them.

  • fit_tracks – if false, the final track find and the TrackCreator module will no be executed

  • use_second_cdc_hits – if true, the second hit information will be used in the CDC track finding.

  • trackFitHypotheses – which pdg hypothesis to fit. Defaults to [211, 321, 2212].

  • use_svd_to_cdc_ckf – if true, add SVD to CDC CKF module.

  • use_ecl_to_cdc_ckf – if true, add ECL to CDC CKF module.

  • add_cdcTrack_QI – if true, add the MVA track quality estimation to the path that sets the quality indicator property of the found CDC standalone tracks

  • add_vxdTrack_QI – if true, add the MVA track quality estimation to the path that sets the quality indicator property of the found VXDTF2 tracks (ATTENTION: Standard triplet QI of VXDTF2 is replaced in this case -> setting this option to ‘True’ will have some influence on the final track collection)

  • add_recoTrack_QI – if true, add the MVA track quality estimation to the path that sets the quality indicator property of all found reco tracks (Both other QIs needed as input.)

21.9. More ?

The tracking package includes the following sub-packages