Cascade reconstruction studies

I produced 1000 Xi's with GSTAR in four ways:

1 per event (1000 events)
10 per event (100 events)
100 per event (10 events)
1000 per event (1 event)

The intent was to get an idea of the reconstruction efficiency using clean cascade events as a function of how busy the event was. There is nothing in these events besides the cascades. In the first case, if all three tracks are reconstructed, then it is very likely that they'll combine to get the cascade mass. With more cascades in each event, you have more chance of screwing up which tracks are associated with which other tracks. In the case of all 1000 cascades in a single event, very few are reconstructed properly. The gaussian fit in this case is rather generous for indicating the number and efficiency of reconstructed cascades.

It appears that losses do not really worsen between having a single cascade in an event and having 10. But a downward turn at higher multiplicities is evident. It's clear that the background with these particular cuts will be a serious issue. It's also clear that clean single cascade events are quite poorly reconstructed (under 10% efficieny), using the cuts we have in place. However, even such an efficiency should be reasonable in determining spectra given enough events if our S/N is sufficient with these cuts. Using StMCEvent to evaluate losses will be very enlightening.

Note that these reconstruction efficiencies include the tracking reconstruction efficiency. I have taken into account the cascade and lambda decay branching ratios to the final state p,pi,pi = 0.99887 * 0.639 (I divided this out in the reconstruction efficiency). Also, the rapidity and pt distribution plots are on all cascade candidates, even if their mass is incorrect and/or they are not real cascades - such distributions are not very informative without using only truly reconstructed cascades, nor without the original generated distributions.

Details:

The 1 cascade per event plots actually only use 500 events, as the GSTAR file had some problem on the 528th event. This was accounted for in the efficiency calculation.
Generation was done following the probability distribution of rapidity between -1 and 1, and Pt between 0 and 2.5 GeV/c (with an inverse slope of 300 MeV). Note that while some cascades are lost at very low momentum, their decay imparts some momentum to each daughter as well (though this is not very much as the cascade mass is only about 120 MeV above the sum of the masses of its decay products).
Reconstruction was done using the default BFC in SL00b using the chain "Cy1d xin tfs" (that's the tfs TPC simulator), with a primary vertex always fixed at (0,0,0). All cuts for V0's and Xi's are those standard in this code library.
Fits were performed with a gaussian plus a line on the mass plots using log liklihood. In the 1 cascade per event plot, the line was constrained to zero.
Errors on the efficiency calculation are based on % error = 100%/sqrt(reconstructed counts).

For more detailed images, click on the image to download a postscript version.