First, the discrimination between central and peripheral events was studied with vacuum between the target and scintillator. This study indicated that even a single counter trigger would give good central/peripheral discrimination. Under experimental conditions this would probably mean that a four counter system would be adequate. The segmentation would be useful for rejecting fake triggers due to upstream interactions, etc. Studies with more counters were carried out, however no simple trigger algorithm gave better discrimination for events coming from the target than the one counter system.
A gold ion which passes entirely through the target without interacting
produces a cloud of delta rays. These events were simulated with GEANT
to determine the counting rate at the scintillator. The computer simulation
showed that 
200 delta rays per traversal enter the scintillation counter.
For the counter to work well it is desirable that the rate of delta rays from beam ions
should be less than the rate of particles from from central
and peripheral events. The delta rays are of relatively low energy. For
example, if we eliminate those delta rays with kinetic
energy less than 40 MeV, then the rate at the scintillator would be
, which is comparable to the rate from peripheral and
central events. Delta rays can be attenuated with a high-Z material.
For the monte carlo simulation we used 9.6 cm of lead, which
easily eliminates enough delta rays. Typically, depending on the
details of edge scattering, the average number of delta rays
going through the counter per beam gold ion is 0.05 -- 0.1.
We next studied how the addition of lead
affects the discrimination between central and peripheral events.
Figure 
is a scatter plot of integrated 
in the scintillator
per event versus impact parameter for 1000 Au-Au collisions. The
integrated per event will be referred to as simply for the remainder
of this section.
Figure: Correlation between integrated and impact parameter for
1000 events.
We wish to make the cut so that we accept approximately 10% of
the events. Because of the correlation between
and impact parameter,
these events will be quite central. Figure (a) is a plot of the
distribution of impact parameters that pass the .16 cut. This demonstrates
that approximately 10% of the events are selected, and those events which
are selected are the most central. Figure (b) is a plot of the
fraction of events passing the .16 cut as a function of impact
parameter.
Figure: (a) Distribution of impact parameters which pass a 
cut.
(b) Trigger probability for a particular impact parameter event to pass the
cut.