Last Updated: April 14, 2003
At the cluster level, I iterate through two-cluster combinations looking for splitoff candidates. Separation between two clusters less than 6-7 cm seems to signal a high probability of a splitoff cluster, evidenced by the 3-gamma mass of 3-cluster events where the closest two clusters are less than 6 cm apart (Fig 1). The peak towards the 175 MeV looks (to me) like a tail of the 2-gamma pi0 mass peak.
I've written a routine called in make_lgd_clusters that goes through 2-cluster combinations, finds cluster pairs closer than a nominal separation (6 cm, for now), and merges one cluster into the other by summing the energies, adding the energy weighted spatial coordinates, and summing the energy widths in quadrature. The cluster that has been merged away is given a flagged value of energy.
After all candidates are found, the routine goes through the cluster list, finding flagged clusters. It then compresses the list by moving every higher cluster down one spot in the clusters->cluster[#] hierarchy, and lowers the value of clusters->nClusters. After the list of clusters has been compressed the routine rewrites the clusters group.
As advised, I've looked at the effect of splitoff reclamation on mass walk in the pi0 and eta as a function of cluster multiplicity.
Using the new cluster seed energies:
Using the old cluster seed energies:
It seems that lowering the cluster seed energies does decrease the mass walk, as expected. The splitoff reclamation doesn't seem to have much effect on mass walk at this point (less than 1 MeV change in 2 -> 6 cluster walk), but I'll check back in on this effect as I further refine the splitoff identification and merging procedures.
As requested, I've examined two-gamma mass vs cluster separation. In hopes of maintaining a clean pi0, and to keep the code simple for a first look, I've only looked at 2 cluster events. In the following postscripts I've plotted two-gamma mass for different separations: less than 6cm, 6cm - 10cm, 10cm - 15cm, and greater than 15cm. It seems clear to me that there is no clean pi0 in the cluster pairs separated by less than 6 cm. As expected, the eta shows up in higher separations....
I've also plotted the energy ratio of the two photons vs separation, hoping to find some other (smarter) tool for locating splitoff clusters. As previous results have led me to believe, the splitoffs don't seem to be dominated by pairs of large and small clusters.
I've started to look into bringing the clusterizer to the block level, but am still working out the code. My hope is that I can make simple changes to the code, perhaps looking at the energy distribution as the blocks come in. My goal is to identify splitoffs "before they happen" without throwing out the advantages of the lowered seed energies.
The following plots are intended to mirror those presented as Benchmark Samples on April 1, 2003. Ideally the parallel use of cluster reclamation will produce converging results....
The signal-to-noise in these results seems better than those presented in the past.... One major difference between our analyses is the way we identify mesons. I've set relatively wide windows on my mass measurements (20, 40, and 50 MeV for the pi0, eta, and omega, respectively), which would produce higher statistics. As can be seen in the final plot, and I've confirmed in other plots, the cut on "one and only one BSD pixel" doesn't seem to affect signal-to-noise or peak width. Removing this cut increases statistics by about 80% in uncut plots, and by about 50% after a 4.5 total energy cut is applied.
Changed low cluster threshhold energy from 150 MeV to 50 MeV; Things look a lot closer to Mihajlo. We also differ in that he uses prune_cluster_singles and supress_dead_lgd_channels.
