A simple method of seperating pi0 signal from background in a Monte Carlo sample is presented, followed by some results using a Pythia QCD "background" production, partonic pT bin 9-15 GeV.

The Method

Associated the Monte Carlo tracks with reconstruted pi0s is diffucult.  Each pi0 has 4 clusters, each cluster could be due to one or more MC tracks and each MC track can result in zero, one, or many recontructed clusters.  While efforts are ongoing to make this "bottom up" type association, the results in this blog take a simpler "top down" approach.

Reconstructed pi0 candidates are labeled by doing the following:

• Loop over all McTracks and find the pi0s
• For each pi0 McTrack
  • Associate all reconstructed pi0 candidates within [tex]\Delta R < 0.04[\tex], with [tex]\Delta ^2 = \Delta \eta^2 + \Delta \phi^2[\tex].
  • If there is only one reconstucted candidate, label it (the reconstructed pi0 candidate) as a 1-1 pi0.
  • If there are multiple reconstructed candidates
    • Label the highest energy candidate as "1-many, max"
    • Label the other candidates as "1-many, other"
• Let all other candidates be labeled "all else"

Cuts Used

After labeling all the pi0 candidates, the following cuts are applied (same as the "new" cuts from this other blog)

• photon energies > 1.5 GeV,
• pi0 p_T > 4 GeV
• M_gg < 0.6 GeV
• Distance between photons in < 25 cm
• Preshower energy for the towers under the two photons each < 40 MeV
• Energy in SMD for each photon / photon energy > 0.012 for each photon

The attached pdf files show a histogram with the number of candidates for each case before "few cuts" (defined as the first three cuts), followed by the same type of plot but for candidates within "all cuts" (the above six cuts).  Next, invarient mass plots are presented for each of the 5 "labels".

Source Files

http://www.star.bnl.gov/cgi-bin/protected/cvsweb.cgi/offline/users/sgliske/StRoot/StEEmcPool/EEmcTree/McPi0TypeSelector.h

http://www.star.bnl.gov/cgi-bin/protected/cvsweb.cgi/offline/users/sgliske/StRoot/StEEmcPool/EEmcTree/McPi0TypeSelector.cxx

http://www.star.bnl.gov/cgi-bin/protected/cvsweb.cgi/offline/users/sgliske/StRoot/StEEmcPool/EEmcTree/macros/runMcPairTypeSelector.C

Comments

The good 1-1 signal is quite wide, as expected from previous studies.  The "1-many max" case doesn't look too bad, and could reasonably be considered pi0 signal.  The "1-many, other" case does not look so good, and should probably be considered background.  Note, the overall signal to background in 0.1-0.2 GeV is not terrible, probably around 6:1.  To improve these studies, I should seperate the plots into pT bins, check the pT resolution in the various cases, and possibly divide the "all else" catagory into a few more specifics.   Thus far, just the TSP algorithm results are attached.