Date: Mon, 08 Mar 1999 17:19:12 -0600 (CST)
From: Frederic Stichelbaut <STICHEL@d0ssc1>
Subject: vertex group report - draft
To: MEENA@d0ssc1
Message-id: <990308171912.2a400199@D0SSC1.FNAL.GOV>


Hi Meenakshi,

I have written in a hurry the veretx group report for Dec/Jan 99 (this
is due today!). May I ask you to have a look at it and send me your
comments/criticisms. One topic which is missing is the L3 and I don't
know what to say. Maybe indicate that John Zhou has written a D0 note
on his tool but has left the vertex group after?

Regards,
         Frederic.

===================================================================

			Vertex Group Report (Dec 98- Jan 99)
				Meenakshi Narain
				Frederic Stichelbaut


   o During the last two months, the vertex reconstruction code has been
     revisited in deep to provide the functionalities needed for primary
     vertex (pv) and secondary vertex (sv) reconstruction. 
     This reconstruction can now be divided into four steps:

      -1) Initial vertex determination: The search for pv or sv starts
          =============================
          from some initial vertex with respect to which the track impact
          parameters are computed. In case of pv search, this initial
          vertex can be:
          - a fixed/random position (for MC events);
          - the beam-spot position extract from a file or the database 
            (for real data);
          - a set of vertex seeds found along the z axis (to handle 
            multiple interactions).
          For sv search, one should start by the reconstructed primary 
          vertices identified at the previous step.

          We have developed the vertex_init package to handle these
          initial vertices for pv/sv search. For pv search, the initial
          vertex can be a random number or read from an external file.
          For sv search, one reads back the pv's stored in the 
          VertexCollChunk.

     -2) Track Selection: The user must define the set of tracks used
         ================
         to find vertices. These tracks are selected with respect to the
         initial vertex (close to it for pv search, far from it for sv 
         search). We modified the track selectors to take into account
         the initial vertex provided by the vertex_init package. In case
         of sv search, we also modified the code such that the track 
         selector can know what tracks were attached to the selected primary 
         vertices and veto these tracks. This request forced us to modify
         the organization of our packages and create a new package, called
         vertex_evt, that contains the definition of the VertexCollChunk 
         and the VertexInfo namespace allowing to access the vertex chunks' 
         content.

     -3) Vertex Reconstruction: this is the central piece of the code that
         ======================
         is supposed to reconstruct the vertices given a set of selected 
         tracks and an initial vertex. Most of the code was already ready
         in December and we just propagated the initial vertex from the 
         steering class VertexAlg to the other classes that need that
         initial vertex.

     -4) Vertex Selection: At this stage, one selects the good vertices
         =================    
         among those reconstructed before. Various algorithms can be
         developed both for primary and secondary vertices. We implemented
         one algorithm based on the track multiplicity who attempt to 
         identify the best primary vertex and discard the fake ones, 
         usually close to the good pv but with a much smaller multiplicity.
         For sv selection, we have started to implement selection tools
         based on the position of the primary vertex and the associated
         tracks:
         - angle between the line joining the pv and the sv and the 
           momentum sum of the tracks attached to the sv;
         - decay length significance.

   o In addition, we have released an analysis package who produces ntuples
     with the important information at each step of the reconstruction.
     This package also contain the MC information so that comparisons between
     generated and reconstructed vertices can be done.

   o The framework needed to reconstruct pv and sv is now almost completed.
     We thus have started to do some performance studies using tracks 
     obtained by fitting the hits they have generated in the SMT and CFT 
     detector. This way, the track finding had an efficiency close to 100%
     and our vertex reconstruction is not affected by the performance of the
     track finding algorithm. This is essential in order to understand the
     intrinsic performance/limitations of the vertexing algorithm.
     We have studied the performance of the primary vertex reconstruction
     in t-tbar and b-bbar events:
       - global reconstruction efficiency, pt dependence
       - fraction of fake vertices
       - resolution in x,y,z + effect of the beam-spot size.
     We have also simulated multiple interactions by mixing together tracks 
     belonging to one ttbar and n bbbar events and looking how well one
     can reconstruct the different vertices as a function of the number of
     interactions.
     These results were presented in front of the Von Ruden review in January
     and the collaboration in February.

   o Finally, we have switched to CTBUILD with the release t00.47.00.