NAZAR LUTSKIV 1
Determining Sr-90 Activity with Geometrical
Corrections
Nazar Lutskiv
Abstract—In this experiment, we investigated the effect of To assess our assumption of geometrical influence, we run
distance on the count rate measured by a silicon detector exposed Monte Carlo simulation with a point source. We then plotted
to a Sr-90 source. We hypothesized that a plot of the adjusted a graph of the ration of emmited particles to detected against
n
count rate, ∆t · d2 , as a function of distance d would yield a
slowly decreasing exponential due to air absorption. However, distance, and indeed found that at small distances there is no
due to the solid angle efficiency of the detector averaged over longer a sharp rise, while the shape of the graph at other
the source’s surface, the count rate significantly diverges from distances was the same as real data.
1
d2
dependence at short distances. After applying corrections for To determine the geometrical efficiency of the source as a
these geometrical effects, we fit an exponential model to the data, function of distance, we first modeled the source as a disk of
yielding a reduced chi-squared value of χ2red = 0.74 and an
estimated source activity of 1.75 ± 0.01 MBq. radius r0 with a point detector. Another assumption we made
was that the efficiency of each infinitesimal point on the disk
is related to the angle θ, as illustrated in the Figure 1.
I. I NTRODUCTION
M Easuring the activity of a radioactive source is of
utmost importance in many fields like physics or
medicine. If extremely precise measurements are needed, It
is critical the geometrical relation of a source and detector are
considered.
While at far distances the source may be treated as a singular
point, the issue of geometry becomes apparent at shorter
distances. In this experiment, we measured the activity of the
source while taking in to the account a complex geometrical
model of the detector and source. To provide evidence for the Fig. 1. Model used to find the geometrical efficiency, with a point detector
validity of our model, we fit an exponential with a reduced and disk source
chi-squared value of χ2red = 0.74 and a resulting activity of
1.75 ± 0.01 MBq. Further we assume that the efficiency is proportional to
cos(θ), so that at θ = 0◦ , the efficiency is at its maximum
II. M ETHOD value of 1, while at θ = 90◦ , it is zero. Based on aforemention
To determine the activity of the source, we used a silicon assumptions, we were able to derive a formula to adjust our
detector mounted on a meter track, with a Sr-90 sample whose fitting model that had r0 as a parameter. While we were able to
distance to the detector could be adjusted. Strontium-90 was fit the line though the data points in the Figure 2, the predicted
chosen as it has a long half-life of 29 years, so changes in value of r0 was 7 ± 0.4 cm which is significantly higher than
its activity over the experiment’s duration can be considered the true value of 1 ± 0.05 cm. We needed a more realistic
2
negligible. We then measured the value of n·d 2 model, which is discussed in further sections.
∆t [Bq · m ].
For each different distance, we did 3 separate measurements
because of the random nature of air absorption and decay. III. D ISUSSION AND R ESULTS
During the measurements of higher distances, we noticed that 2
the random effect started to dominate, leading to a larger For the measurements of n·d ∆t of Sr-90, we obtain the
spread of n·d
2 following graph.
∆t at the same distance d. To make the data more
stable we increased the sample time the further we went. The sharp rise is due to the geometrical relation of the
We also assessed the influence of the inherent limitation source and emitter. At a close distance, we can no longer treat
of the silicon detector—its dead time of approximately Td = the source as a point. Instead, we have to calculate the solid
10−6 s, as obtained from the interval plot of the Data Acqui- angle subtended by a disk detector averaged over the surface
sition (DAQ) program at the smallest interval. At the shortest of the source Ref. [2]. The value of the solid angle is given
distances, the count rate is sufficiently high to introduce a by a formula (1).
significant systematic error of 5% losses, estimated using a Z ∞
4πa exp(−dk)J1 (sk)J1 (ak)
paralyzable detector model, with a ratio of lost counts equaling Ω= dk (1)
Td ·n s 0 k
to 1 − e ∆t . For an overview of dead time behavior models
see Ref.[1]. However, this error quickly becomes negligible at where a is the detector radius (derived from the known
greater distances. detector area of 1 cm2 ), d is the distance between the source
NAZAR LUTSKIV 2
Fig. 2. Plot of an adjusted activity of the source on distance, based on the
assumption of a d12 dependence. Fig. 4. Plot of measured activity corrected for geometrical effects, scaled on
Y axis to compare with the first figure
and the detector, s is the radius of the source, k = ad , and J1 performed a chi-squared test on the corrected data to quanti-
is the Bessel function of the first kind. tatively evaluate the fit quality. Due to significant measurement
The efficiency of the source on the detector is defined as errors at short distances, the data is slightly overfitted, resulting
Ω
ε = 4π . By numerically computing the solid angle for our set in a reduced chi-squared value of χ2red = 0.74 Despite this, the
up values a = 0.56 ± 0.05 cm, s = r0 = 0.1 ± 0.05 cm, (with χ2red value provides strong statistical support for the validity
uncertainties due to ruler measurements), and thus efficiency of the geometrical correction applied. Lastly, we obtained
for a range of distances, and then dividing it by the expected reasonable correction for the data points, using the true value
efficiency of our point source model Aπd detector
2 we obtain the of the radii of the detector and the source, which implies more
following Figure 3. evidence for the correctness of the geometrical correction for
our model.
Also, we obtain an air absorption coefficient µ = 0.3 ±
0.07 m−1 . The coefficient is dependent on the initial energy
of the electron, which is a spectrum for beta decay. Thus, our
coefficient represents the average absorption rates of electrons
with different energies and is assumed to be constant with
distance.
While the silicon detector is precise at low count rates, at
smaller distances the systematic error of dead time plays a
crucial role in the estimation of total error. Because of lost
counts at the higher rates, the points at lower distances are
below the trend line.
IV. C ONCLUSION
Fig. 3. Graph demonstrating the difference between new correction coefficient
In this experiment, we measured the activity of a Sr-
ε to the one previously used ( Aπd
detector
2 )
90 source using a silicon detector, resulting in a value of
1.75 ± 0.1 MBq. Rather than applying a simple inverse-square
It is clear that at smaller distances there is a high variation correction, we proposed a refined model incorporating geo-
which is aligned with what we see at small distances in metrical factors, validated with a reduced chi-squared value of
Figure 2, while at larger distances the solid angle efficiency χ2red = 0.74. The systematic error, primarily due to the detec-
approaches the point source model. tor’s dead time, affected counts at shorter distances, resulting
Instead of multiplying the count rate by d2 , we can now in an estimated 5% loss. These adjustments provide evidence
use the calculated efficiency to improve attenuation at smaller supporting the geometrical correction model’s accuracy and
distances. The plot of ∆t n
· 1ε on distance is shown in Figure demonstrate the necessity of accounting for solid-angle effects
4. in activity measurements of sources at short distances from the
As the rate of air absorption with distance is proportional detector.
to the number of emitted particles, ( dN dx = −µN0 ), we fit an R EFERENCES
exponential to the resulting data point to obtain an activity
of 1.75 ± 0.1 MBq from our Co coefficient, with the error [1] G. F. Knoll, “Chapter 4 - General Properties of Radiation
Detectors,” in *Radiation Detection and Measurement*, 3rd ed.
calculated using error propagation. Since our objective was New York City, NY, USA: John Wiley Sons, Inc., 2000, pp.
to introduce a more accurate adjustment factor than d2 , we 118–120.
NAZAR LUTSKIV 3
[2] Michael Fox (Ed.) The Second Year Laboratory: radioactivity
experiment’, Imperial College London, 2024.
A PPENDIX
A PPENDIX : U SE OF AI
The AI was used for minor grammar corrections.