After delivery, the CsI crystals must pass three tests to be selected for use in the
calorimeter: the ratio of the fast component to total scintillation light output must
be greater than seventy percent; the linear and angular dimensions of a given crystal
must be within predefined tolerances; the light
collection non-uniformity must be better than three percent. Various surface treatment
methods have been investigated for maximal light output and maximum non-uniformity and
a cosmic ray tomography apparatus has been set up to measure the light non-uniformity
throughout the volume of each module. A particular surface treatment consisting of
three layers of teflon plus one layer of aluminized mylar was arrived at by the time of
the development run of 1994; as a result, the modules used during that run were wrapped
as described above.
The results of the development runs aided in the optimization of the design of the
apparatus, revealed the performance of the main detector systems and validated the
experimental technique. During the test of 1992, the effort was concentrated
exclusively on the beam and its parameters such as the spot size, the horizontal and
the vertical profiles, the contaminations and the beam particle rates. The results of
this test affected the design of the active target and its segmentation. The target was
fabricated at the University of Virginia and inserted in the beam during the test run
of 1993. By that time, some CsI modules were delivered by Bicron Corporation
and it was possible to build up a section of the calorimeter and to study the responses of
the array to monoenergetic beams of positrons at various momenta. The line shapes of
the responses of the array were compared to simulations. Excellent agreement was
observed after the effects of photoelectron statistics and light collection
non-uniformity were accounted for. At 70 MeV/c (this is the energy of interest since the
's from 
have 
)
the measured energy resolution of the calorimeter was 5.6 MeV FWHM. The study of the
stopping rate in the active target revealed that the number of stopped pions scaled with
the area of twenty-five, nine and four fibers. This allowed the redesign of the active
target with 
fibers instead of 
. The
redesigned active target was also built in Virginia and tested in beam during the
test run of 1994: the beam distributions measured with the active target suggest a beam
spot of 
FWHM. Another study conducted during the 1993 run was a test of
the wrapping method. Of the various surface treatments considered, black paper at the
front faces of the crystals with the lateral surfaces wrapped in teflon gave the best
light output and uniformity. Further studies after the run led to the wrapping system used in the tomography
apparatus and during the test run of 1994. In the tomography system, the directions of
the cosmic muons are defined by three drift chambers. The path-lengths of the cosmic
rays and their energy depositions in the CsI modules are used in a reconstruction
algorithm which finds the light per unit path-length in the volumes of the modules ---
the crystals are subdivided into cubic cells and the reconstructed light yields per
path-lengths in the cells constitute a measure of the optical non-uniformity in the
crystals. The CsI modules used during the test of 1994 were all examined in the
tomography system. More crystals were delivered by that time and it was possible to
set up a much bigger portion of the calorimeter than the one used in the 1993 experiment.
The line shapes of the responses of the calorimeter were remeasured with not only
monoenergetic 
's but monoenergetic 
's as well. With a good control of the
temperature and the humidity, a better wrapping system, and a bigger size of the
array, a better energy resolution was achieved, i.e., 4.2 MeV FWHM. In addition, the
response of the array to 
decays confirmed the need
for MWPC's with good double track resolution for additional suppression factors of
double Michel events. The 
decays were well separated
from the Michel background and the number of 
events
collected was consistent with the correct branching ratio.