The above spectra were taken on a small subset of the total Radphi event sample that was selected to enhance fully-contained forward events. The selection entails three cuts: rejection of events with at least one cluster near the boundaries of the calorimeter (fiducial cut), rejection of events with neutral energy in the barrel (barrel gamma veto) and rejection of events with more than one pixel cluster in the barrel scintillator (single recoil). The above analysis was repeated with each of the above cuts suspended in order to examine their effects on the ability of the analysis algorithm to reconstruct an exclusive final state.
The fiducial cut is implemented using the software library function lgd_cluster_cleanup(). It removes events with any clusters whose center lies near the outer or inner boundary of the detector. It has essentially no effect on the sample of events with two clusters, but has increasingly visible effects on higher multiplicity samples. For example, only 2/3 of the five-cluster events survive this cut. When the cut is suspended the maximum of the total energy peaks are shifted downward relative to the above figures by about 4% and the r.m.s. of the peaks are increased by about 10%. The conclusion is that the fiducial cut is valuable for analyses involving higher multiplicity final states, and involves relatively modest cost in terms of statistics.
 |
The barrel gamma veto is implemented by looking for any bgd hit over ADC threshold and vetoing the event if no corresponding hit is found in one of the two straight scintillators overlapping the bgd counter. In contrast to the fiducial cut above, the barrel gamma veto is both very powerful and costly in terms of statistics. Fig. 9 shows the total forward photon energy spectrum corresponding to tagger channel 10 for all two-cluster events without (open histogram) and with (shaded histogram) the barrel gamma veto enabled. The figure shows that without this cut the total forward energy would bear little relation to the tagger energy; without the cut there is no real peak in the spectrum and the maximum occurs well below the energy of the tagging channel. The cut eliminates all but about 15% of the events, a fact which is more or less the same for all cluster multiplicities. Without it there is very little chance of knowing that the event was contained in the forward calorimeter. The conclusion is that either this cut or a refinement thereof must be strictly applied for all subsequent analysis or a method must be developed to include the barrel neutral clusters in the reconstructed event.
 |
In fact, the power of this cut for cleaning up fully-contained forward all-neutral events suggests that a piece of what appears in the low-energy tail of the peaks in Figs. 7-8 corresponds to events with neutral barrel energy that ``leaks'' through the barrel gamma veto cut by appearing in the two bgd counters that overlap the recoil pixel. These two counters are not included in the barrel neutral energy sum because they normally contain some energy from the stopping recoil track. A more sophisticated barrel gamma veto cut might make use of z-coordinate information in the bgd to separate the recoil track from additional neutral clusters in these counters. One way to test this is to repeat the above analysis with the zero neutral energy cut in place and in addition to require that among the two bgd counters behind the recoil pixel, only one of them may see a hit. This should reduce by a factor of two the leaks mentioned above. The result is shown in Fig. 10, which confirms the hypothesis. The r.m.s. of the peak is decreased by 10% by the additional requirement of at most one bgd hit, and the spectrum of the eliminated events shows no peak at the tagger energy. The cost of the additional requirement in terms of statistics is 20% in good agreement with expectations based upon the leaky veto hypothesis, leading to the conclusion that the barrel gamma veto should include the additional requirement of no more than one counter hit.
The single-recoil cut, by contrast, has a relatively weak effect on either the quality of the total energy signal or the statistics. It cuts roughly 20% of the statistics from the sample, regardless of multiplicity, and adds these events mainly to low-energy tail of the total energy peak. No other explanation for this effect would be required than simply that the additional bsd hits in the event allow more neutral energy in the bgd to ``hide'' from the barrel gamma veto cut. Since this cut is relatively cheap in terms of statistics and enables a cleaner identification of neutral clusters in the barrel, the conclusion is that this cut should be retained in subsequent analysis.