Escolar Documentos
Profissional Documentos
Cultura Documentos
Counting Statistics
March 27, 2009
PAUL KILGO
DR. JEROME BUSENITZ
MODERN PHYSICS LAB
THE UNIVERSITY OF ALABAMA
Kilgo 2
ABSTRACT
In many fields of science, many different phenomena can be described in terms of a statistical function.
In this laboratory, we'd like to measure the Gaussian nature of Cesium-137 radioactive decay. If we can
successfully prove that Cesium decay is in some way Gaussian, then we can verify using Gaussian
statistics as an appropriate approximation for the model.
METHOD
The most sophisticated device we will use is a Geiger-Muller (GM) tube which is a way to count the
radioactivity. We will place Cesium-137 under the GM tube, suspending the tube over the material by
hanging it with a ring stand. From here, we can vary the applied voltage and measure the resulting
change in the counting rate.
Procedure
1. Set up the experiment. Place the GM tube directly over the Cs source to ensure it will collect
radiation.
2. Plateau the counter by starting the voltage at 600V. Step upwards in quantums of 100V and find
where the rate of counting changes the least in between steps. Do the same thing for lower
resolutions of voltage until you find a plateau voltage.
3. Move the GM tube a distance above the Cs source so that it collects close to 30 counts per
Kilgo 3
minute.
4. Begin collecting data. Set the counter to manual mode and allow to run. Every ten seconds,
write down the reading on the counter. Do this until at least 500 measurements are collected.
DATA
See Appendix A.
ANALYSIS
For analysis, we subdivide our data into four different sets where we measure the rate over intervals of
10 seconds, 30 seconds, 1 minute, and 2 minutes. See Appendix B for the organized data.
We can see that as a sanity check that the average increases by a factor indicative of the change in time
unit. For instance, from the 10 second to 30 second interval, which differ by a factor of 3, the average
increases by a factor of 3. The standard deviation doesn't seem quite so related, but we can expect the
standard deviation to increase as our data points decrease.
We can check the weight of our distribution by measuring the probability that we will take a
measurement that is one sigma over the average and one sigma under the average. For each
distrubution:
Sigma weights
Interval Avg + Sigma Avg – Sigma
10s 0.196 0.111
30s 0.176 0.110
1m 0.160 0.160
2m 0.123 0.226
For each of these graphs, it does suggest a Gaussian trend. For the 10-second interval, Gaussian
statistics aren't great for an approximation since the frequency does not approach 0 as we tend toward
the left of the graph. We can't have readings which are less than zero, so there's not a possible way to
extend our graph. If we were to “pretend” that we could, the distribution does follow that of a Gaussian
curve.
The 30-second interval graph is marred beyond belief. Our data appears to be a little too rambunctious
to even consider a Gaussian trend. We could possibly mend this by increasing the bin size, but this bin
size was the smallest to produce a mode with a frequency greater than 15.
Kilgo 4
The 1-minute interval is actually quite symmetric. If we were able to take more data points at this
interval, the graph would look much more Gaussian, despite our graph having 'wings' on either end of
it.
The 2-minute interval graph does appear to follow a Gaussian trend, though there are too few bins to
really tell much at all about it. Judging by the above sigma weights, it's possible that if we were to take
more measurements and decrease the bin size to produce a mode that's greater than 15, then this
interval would also appear Gaussian.
Overall, to increase the observation-expectation correlation we could solidify our setup a little better.
We could have had disturbances in our setup that introduced non-statistical error into our system.
[ Binomial Statistics ]
Kilgo 5
Appendix B
Appendix C: Histograms
75 20
70
18
65
60 16
55
14
50
45 12
Frequency
Frequency
40
10
35
30 8
25
6
20
15 4
10
2
5
0 0
1 2 3 4 5 6 7 8 9 1011121314151617 10 12 14 16 18 20 22 24 26 28 30 32 34 36
Counts per 10s Counts per 30s
20 16
18 15
14
16
13
14
12
12 11
Frequency
Frequency
10 10
8 9
8
6
7
4
6
2 5
0 4
[26,28] [32,34] [38,40] [44,46] [50,52] [56,58] [65,73] [74,82] [83,91] [92,100]
Counts per 1m Counts per 2m