Some miscellaneous thought regarding improving experiment reliability
follow. There are undoubtedly many more that should be added to the
list.
The use of 6 micron Mylar film between the cathode and CR-39 in
SPAWAR experiments reduced the CR-39 particle tracks by about 90
percent, but had the advantage of (at least mostly) eliminating
chemical and mechanical damage caused tracks. The effect of this 90
percent track count reduction can be overcome by simply running 10
times as long, provided there is a continuous production of events.
In fact, some of the primary goals of such experiments,
reproducibility and repeatability, can likely also be increased by
simply running the experiments longer, and thus eliminating the
variability due to small counts. When using a geiger counter, the
expected standard deviation for N counts is N^0.5. If you only get
64 counts, the standard deviation is 8. If the background for the
time of counting is estimated at 16 then you estimate the mean count
as 64 - 16 = 48. The variance of each count, being from a Poisson
distribution, is the count itself. Since variances add, we have the
variance on the composite mean of 64 + 16 = 80, and thus a standard
deviation of (80)^0.5 = 8.9, which is 18.5% of 48. The count of 64
is thus 5 sigma away from a null result. However, experimental
variations can wipe that out quickly. By increasing the counting time
we increase the reliability of both our estimate of the background
and the live count.
The presence of a during experiment background count is not too
important when it comes to alpha or proton counting, because these
are fairly non-existent except in cases of contamination (e.g.
radon), and can in some cases, be shielded. However, when it comes
to neutron counting, there is indeed a run time background which can
be important to error bar calculations because the neutron production
from CF experiments is so low and shielding difficult. Also, it has
been established that neutron background can go up substantially
during solar flares, and this has affected CF experiments in the
past. It is thus important to always use controls, and, if positive
neutron results are obtained, especially in an event, to check the
status of solar radiation.
Beyond this is a third counting random variable, the baseline
background exposure of the particle counter material itself, which is
relatively independent of the run time of the experiment, and more a
function of time since curing. The mean of this probably does not
vary much during the course of an experiment, but there can be
variance over the areas chosen for chip cutting, i.e. across the
areas of the larger piece of plastic from which the CR-39 is cut, and
sub-areas if each chip. If multiple control chips are used across
multiple experiments, the counts taken on those chips may be useful
in establishing the area normalized mean and variance of the baseline
background count. If the area of a chip is large, this might be
accomplished by dividing up the chip into at minimum 6 equal sized
sub-areas and counting and determining the area normalized mean and
the sigma for that mean.
So what does this say about procedures for increasing results
reliability? To increase reliability, with minimal change to the
materials or experimental protocol: (1) run counts for longer times,
(2) break chip areas into subsets, and count subsets separately, (3)
either use as large a cathode area as feasible or make multiple runs,
(4) use control chips to establish area normalized baseline
background mean count and sigma.
In addition, it is obviously advisable to try to increase the mean
counter events per unit area. It appears, from published
experimental results, that one way to do this, for some cathode
types, is to (5) increase the electrostatic field at or parallel to
the cathode surface and (6) increase the imposed magnetic field
intensity.
And of course, while we are at it, there is the old standby, (7) use
rigorous procedures to eliminate variability due to the experimenter,
or experiment variables, like run time. Some things related to this
may be: (7a) making a fresh solution for each experiment, (7b)
pumping or siphoning the electrolyte through the cell from a larger
reservoir in order to maintain concentration consistency, or (7c)
frequently topping off with distilled water to compensate for
evaporation, or (7d) using a semi-closed cell with vapor condenser,
or a closed cell with recombination catalyst, especially for long
runs; (7e) controlling temperature, (7f) applying the above
procedures, where appropriate, to the etching solution and process,
(7g) avoiding pre-etching of chips or exposure to heat or chemicals
that effectively nullifies the manufacturers curing procedure and
changes the chip's exposure rate (dE/dx) and etching rate by depth,
(7h) using a control chip or control area on each chip to check for
post experiment contamination, especially in the etching procedure,
(7i) etching in a closed oven-like accurately temperature controlled
environment for an accurate time period, and (7j) of course meeting
good laboratory standards for cleanliness, measurement, and safety.
If a rigorous procedure is available, preferably the manufacturer's
procedure, for curing the CR-39 or other particle counting plastic,
then it should be tested to see if re-curing can significantly
eliminate baseline background counts and/or improve the uniformity of
the tracks generated for a given energy particle type, across
experimenters and/or plastic manufacturers, and if so, the curing
scenario employed within a limited time interval prior to
experimentation. This is somewhat in conflict with the need to (8)
include on a portion each chip a controlled exposure to particles of
a known energy, so as to be able to determine the effects of the
etching procedure used.
Best regards,
Horace Heffner
http://www.mtaonline.net/~hheffner/
