goal : try some b-tagging algorithms (used in CMS and originally Delphi for b-jets reconstruction)

The easiest one is the counting algorithm where , for each events, the number of tracks with a given impact parameter significance are counted. Still the idea is that tracks from displaced vertex may have larger DCA with respect the primary vertex.

Same arguments can be applied for the D0 reconstruction

1. count the number of track with a given impact parameter significance
2. associate only those tracks
3. use KFParticle for secondary vertex finder

A. impact parameter studies

Fig.1 shows the error of tracks DCA in XY (from covariance matrix + it includes the primary vertex error in XY) for tracks.

As seen before, tracks with at least (ssd+svt)>2 should be selected

The error of tracks is Pt dependant, as shown by the plot below (Fig. 2)

Even for tracks with 4 silicon hits, some bands are seen, traducing the fact that different phi regions have different material thickness (SVT electronic)

Fig 3. (below) is the radius of the first hit on track

Finally I found that for a radius < 6.6 cm, the error of DCA vs 1/Pt shows a unique band.

Of course it selects a small region in phi but it provides a clean error of tracks