You are on page 1of 5

Journal of Radioanalytical and Nuclear Chemistry, Vol. 264, No.

1 (2005) 193–197

Validation testing of the Genie 2000 Cascade Summing Correction


W. Russ,* R.Venkataraman, F. Bronson
Canberra Industries, 800 Research Parkway, Meriden, CT 06450, USA
(Received April 6, 2004)

To validate the accuracy and precision of the Cascade Summing Correction method, over 800 archived measurements of calibrated sources (filter
paper, 20 cm3 liquid scintillation vial, 400 ml beaker and Marinelli beaker) containing cascading (88Y and 60Co) and non-cascading isotopes from
133 different ISOCS/LabSOCS characterized high purity germanium detectors have been analyzed. Comparing the corrected results for the
cascading isotope activities to the known activities shows the method is effective and accurate. Evaluation of the accuracy as a function of the
amount of correction reveals a small systematic error for which a variable precision adjustment is recommended. Requirements to filter true
coincidence X-rays by are verified.

Introduction from a HPGe detector will reduce potential cascade


summing effects to less than 2%. However, the count
When a nuclide decays in cascade, producing more time would have to be increased by more than an order
than one gamma-ray from a decay, and if two or more of of magnitude to achieve the same precision or detection
these gamma-rays interact with the detector, and if they limits, which is normally not desirable. Therefore, a
cannot be discerned separately given the resolving time general methodology for correcting for cascade
of the measurement system, this is called cascade summing effects has been developed to enable more
summing. The measured energy deposited by a accurate measurements of such cascading nuclides.
cascading photon may be artificially increased by the Canberra Industries’ Genie 2000 spectroscopy software
amount of energy deposited by the cascading partner(s), provides such a Cascade Summing Correction (CSC)
up to the sum of all full energies. The effects of this analysis algorithm.1,2 To accurately compensate for true
phenomenon are summing out, where full energy peak coincidence summing losses, the total detection
counts are lost by being shifted above the photopeak, efficiency of each of the cascading gamma-rays must be
and summing in, where peak counts are gained in determined. The CSC algorithm does this by use of a
photopeaks that occur at the sum of the full energies of point-source peak-to-total efficiency (P/T) calibration.
the cascading photons. The probability of summing out Combining results from the P/T calibration with
depends only on the total efficiency for measuring any calculated peak efficiencies at various points within the
amount of energy from cascading photons. The sample yields the total efficiency for the sample. The
probability of summing in depends only on the peak CSC algorithm uses this data to calculate a correction
efficiency for measuring peak energies from cascading factor (COI) for each peak of cascading nuclides that is
photons. Therefore, the magnitude of cascade summing then applied to correct the final activity results. The
correction factor is solely a function of measurement angular correlation of gamma-rays from cascading
efficiency (detector size, sample size, and sample nuclides is not accounted for with the CSC algorithm.
location). It is not a function of activity (this is called However, the contribution of such anisotropy to the true
random summing). Some of the more commonly coincidence correction method is less than the
encountered cascading nuclides include 60Co, 152Eu, uncertainty of the correction in general. 7
133Ba, 88Y, and 139Ce, but there are about 75 other Initial verification and validation analysis of the CSC
nuclides that might require correction for cascade algorithm (using 8 detectors with source-based P/T
summing. Note that these are common calibration calibrations) estimated that the cascade correction factor
nuclides used for establishing the “correct” efficiency of introduces a systematic uncertainty of ±5% (1 ) in the
the detector; if cascade summing is not corrected during resulting corrected activities when using an
the calibration process, the efficiency will not be correct. ISOCS/LabSOCS characterization3 applicable to the
Higher measurement efficiencies, from larger specific detector used in the measurement. Further
detectors and closer sources, result in greater cascade investigation of the accuracy and precision of the CSC
summing effects. A 60Co or 88Y source at the endcap of algorithm has been made possible by the continuing
a large HPGe detector can result in cascade summing production of HPGe detector characterization
out losses of approximately 20–30%. Significant files for use by ISOCS/LabSOCS customers to do
cascade summing can be avoided by reducing mathematical (sourceless) efficiency calibrations.
efficiency. In general, keeping a source more than 10 cm

* E-mail: wruss@canberra.com

0236–5731/USD 20.00 Akadémiai Kiadó, Budapest


© 2005 Akadémiai Kiadó, Budapest Springer, Dordrecht
W. RUSS et al.: VALIDATION TESTING OF THE GENIE 2000 CASCADE SUMMING CORRECTION

This “characterization” production process and These P/T values were then converted into the P/T
subsequent Verification process uses calibrated sources calibration file for use by Genie 2000 in the CSC
measured in various geometries. This gives us a rich algorithm.
resource containing a large number of detectors of Each spectral file was analyzed with the appropriate
various types and sizes, with both cascade and non- energy, efficiency and P/T calibration. Peaks were
cascade nuclides. initially determined using the standard Sum/Non-Linear
LSQ fitting algorithm and then verified and refined
Experimental using Genie 2000’s Interactive Peak Fitting utility. After
applying the efficiency calibration to the peak search
This study used archived measurement data from results, Nuclide Identification with Interference
133 HPGe detectors, including over 800 separate Correction analysis was performed. This analysis step
spectra, with approximately 3200 separate peaks with included the application of the CSC algorithm using the
cascade summing. Seventy-four of the detectors were resident P/T calibration and an ISOCS geometry
standard coaxial detectors, and fifty-nine of them were template to specify the measurement geometry. Because
low energy detectors without an appreciable dead layer of the huge volume of analyses to be performed and to
that allows very low energy photons to be detected. For avoid human errors, this full spectral analysis was
most of this study, only the coaxial detectors were used automated. Automation included incorporating all of the
to avoid the additional complications of X-ray induced steps into an Analysis Sequence File (ASF) and running
true coincidence summing. Each of these detectors has this ASF using Genie 2000 Visual Basic programming
been fully ISOCS/LabSOCS characterized to allow components that subsequently extracted pertinent results
accurate calculation of peak efficiencies using a large from the spectral and efficiency files into a common
suite of geometry templates that can be tailored to spreadsheet for sorting and analysis. Results include
specific measurement geometries. To validate and verify count rate, count rate uncertainty, efficiency, efficiency
the ISOCS/LabSOCS efficiency calibrations, a series of uncertainty, and the CSC coincidence correction factor
calibrated volumetric sources were measured in a variety (COI) for each peak. In addition, the decay corrected
of geometries.4 These source geometries include a near certificate activity in gammas per second and
and far filter paper (active diameter 48 mm), near and uncertainty were calculated and assigned to each peak to
far 20 cm3 liquid scintillation vial (25 mm × 51 mm), serve as the correct activity.
near and far 400 ml cylindrical beaker (76 mm × The accuracy and precision of the CSC algorithm is
76 mm), and a 2.8l Marinelli beaker. Near geometries assessed by analyzing the average and standard
mean the source is 2–3 mm from the end cap and far deviation of the ratio of measured to certificate activities
geometries mean the source is 10 cm from the endcap. for the various geometries and nuclides. The
Each source contains the following nuclides: 241Am measured/certificate activity ratio is defined as:
(59 keV), 109Cd (88 keV), 57Co (122 keV), 139Ce
Measured/Certificate activity ratio =
(166 keV), 113Sn (392 keV), 137Cs (662 keV), 54Mn
(835 keV), 65Zn (1115 keV), 60Co (1173, 1332 keV), Measured counts per second
= (1)
and 88Y (898, 1836 keV). For the purposes of this study, Efficiency COI Certificate gammas per second
the nuclides of interest include 60Co and 88Y as the two
The non-cascading nuclides (56Mn and 65Zn) occur
cascading nuclides and also 54Mn and 65Zn as non-
at a similar energy region (835 and 1115 keV,
cascading nuclides with close energies for reference.
respectively) as those of the cascading nuclides (898,
Each of these sources was modeled with an ISOCS
1173, 1332, and 1836), and show expected agreement
geometry template to enable determination of peak
with the certificate activity, confirming the accuracy of
efficiencies at the energies of interest.
the peak efficiency calibration.
These measurements were for other purposes,
consequently source-based P/T calibrations were not
performed, and the detectors were no longer available.
Results and discussion
Therefore, Monte-Carlo methods (MCNP) were used to
create the P/T calibrations.5 Using the validated MCNP
The effect of the cascade summing correction is most
model for each detector, a monoenergetic point source
evident when applied to high efficiency geometries. Of
was modeled at 5 cm above the endcap in air, simulating
the available geometries, the higher efficiency
the normal customer measurements. The resulting
geometries include the near filter paper, near 20 cm3
spectral tally was analyzed to determine the P/T ratio for
vial, near 400 ml beaker, and the 2.8-liter Marinelli
each energy of interest. It has previously been shown
beaker. The measured/certificate activity ratios for these
that it is sufficiently accurate to use the MCNP P/T
geometries were calculated and are shown in Fig. 1.
calibration to enable the use of the CSC algorithm.6

194
W. RUSS et al.: VALIDATION TESTING OF THE GENIE 2000 CASCADE SUMMING CORRECTION

Fig. 1. Cascade summing correction results for non-cascading nuclides 54Mn (835 keV) and 65Zn (1115 keV) and cascading nuclides 60Co (1173,
1332 keV) and 88Y (898, 1836 keV) for the following sources near the detector endcap: filter paper (a), 20 cm3 liquid scintillation vial (b), 400 ml
beaker (c), and 2.8 liter Marinelli beaker (d)

Fig. 2. Cascade summing correction results for all cascading nuclides ( 60Co and 88Y) in all geometries as a function of the CSC Correction Factor
(COI), obtained by averaging each COI quadrant. Trend line added to guide the eye

195
W. RUSS et al.: VALIDATION TESTING OF THE GENIE 2000 CASCADE SUMMING CORRECTION

Uncorrected and corrected results are shown separately Equation (2) is a linear approximation of a fit to the
for all of the peaks for 54Mn, 65Zn, 60Co and 88Y. All of data in Fig. 2. This would not unnecessarily penalize
the error bars in this figure represent the measured minor corrections while still accounting for the highest
standard deviation of the data. However, the propagated severity bi-directional bias. This recommendation has
uncertainties were calculated and tracked well with the been drawn from results involving a wide range of
empirical data, generally matching within ±1%. Figure 1 detector and source geometries but only higher energy
shows that the CSC algorithm is accurate, with all gamma-rays. Extrapolating to lower energies should be
corrected line activities for all geometries resulting in valid because the likely source of the systematic bias,
activities within one standard deviation of the average the assumption of an intrinsic P/T calibration, is at least
non-cascading line activity ratios. as sound at lower energies. Previous work has also
It is hypothesized that any accuracy bias or precision validated the precision limit at lower energies.3
degradation caused by the CSC algorithm would become The preceding discussion was for standard coaxial
more evident as the amount of correction increases. To detectors, with a typical front dead layer of 0.3 mm Ge.
investigate this, the trend of performance as a function A greater bias for 88Y was observed when analyzing
of the CSC coincidence correction factor (COI) was results from detectors without a significant dead layer at
analyzed, with the data presented in Fig. 2. This data is the front of the detector and without any additional
the combined results of all 60Co and 88Y absorbing material, as shown in Fig. 3. Such low energy
measured/certificate activity ratios, along with the detectors without absorber can also detect the 15 keV X-
standard deviation of the data. To evaluate the trend rays that are in true coincidence with the 88Y gamma-
with COI, the results were divided into distinct regions rays. This X-ray true coincidence summing is not
where COI ranged from 0.80 to 0.85, 0.85 to 0.90, 0.90 corrected by the current CSC algorithm. This error can
to 0.95 and 0.95 to 1.00. In each region, the average COI simply be avoided by adding sufficient absorber material
was calculated for both the uncorrected and the between the source and detector to adequately attenuate
corrected data. the X-rays. This extra absorber material is required only
When comparing uncorrected and corrected results when using a low energy HPGe detector (e.g., LEGe,
for Fig. 2, there was no significant increase in the BEGe, XtRa, and REGe) when trying to correct for a
uncertainty at larger COI values. This implies that the cascading nuclide that decays by electron capture or
uncertainty of the CSC algorithm is small, compared to internal conversion. The data in Fig. 3 was collected
the propagated uncertainty of the other factors. using 3 mm acrylic attenuators. The total amount of
However, there does seem to be a distinct trend in the material needed to reduce the 15 keV X-rays from 88Y
data that suggests a slight bias in the magnitude of the by at least 90% would be approximately 2 cm of plastic,
COI value that would cause a systematic under- 2 mm of aluminum, 0.05 mm of iron or 0.04 mm of
correction. The magnitude of the bias suggests a linear copper. A standard aluminum endcap without a thin
increase from 0% with no correction (COI = 1.00) up to window has a thickness of about 1.6 mm. The typical
about 5% under-correction at COI = 0.80. An assumption dead layer for standard coaxial detectors (~0.3 mm Ge)
made with the current CSC algorithm is that the P/T attenuates 15 keV X-rays by over 99.9%. While the
calibration is an intrinsic function of the detector and is 15 keV X-rays from 88Y are easily eliminated with
invariant with sample size and shape. This is probably nominal attenuation, it may not always be reasonable to
not totally true, and is suggested as a cause for this small fully attenuate higher energy X-rays. However, even
bias, and an area of future investigation. Given that the nominal attenuation can reduce the impact of
direction of this systematic bias might be manifested in uncorrected X-rays to a reasonable level. As an example,
either direction for a particular detector and geometry, it an ISOCS-characterized 75% coaxial detector with
would be prudent to simply increase the uncertainty in nominal attenuation from a 0.53 mm dead layer and
all final corrected results in a manner that would account 1.6 mm aluminum endcap was modeled in MCNP with a
for the systematic error. This was the conclusion of the 152Eu point source sitting on the endcap. The total
original CSC algorithm validation and verification, with coincidence losses to the 1408 keV gamma-ray, caused
a recommendation that a 5% linear increase in by the 121 keV gamma-rays and the X-rays of 39–
uncertainty be added to all activity uncertainties that are 46 keV range, were found to be about 36.4% based on
corrected by any amount. However, this more extensive analytical and MCNP calculations. The coincidence
study suggests that the additional linear increase in losses caused by just the 121 keV gamma-rays were
uncertainty should lessen with less correction applied. found to be about 34%, based again on analytical and
Based on the data in Fig. 2, an appropriate uncertainty MCNP calculations. Therefore, the effect of not
assignment would be: including the X-ray contributions is an under-correction
of nuclide activity by less than 4% for this severe case.
% Corrected uncertainty =
= % Uncorrected uncertainty + 25(1 COI ) (2)

196
W. RUSS et al.: VALIDATION TESTING OF THE GENIE 2000 CASCADE SUMMING CORRECTION

error is caused by consistent differences between the


conditions for the P/T calibration and the actual
measurement and could trend in either direction. Further
study of the data is warranted to verify the source of this
systematic error. It is recommended that this bi-
directional systematic error be accounted for by adding a
linear correction to the resulting activity uncertainty.
Rather than unduly penalizing small corrections by
adding a constant, enveloping increment to all
uncertainties, the added uncertainty should be
proportional to the degree of correction. Because low
energy, cascading X-rays are not corrected by the
current CSC algorithm, measuring such sources with a
low energy detector currently requires a minimum
amount of absorbing material to physically attenuate
their effects.

Fig. 3. Cascaded summing correction results separately for 60Co and


88Y for low-energy detectors in all geometries as a function
References
of the CSC Correction Factor (COI), obtained by averaging
each COI quadrant. Trend lines added to guide the eye
1. V. P. KOLOTOV, V. V. ATRASHKEVICH, S. J. GELSEMA,
J. Radioanal. Nucl. Chem., 210 (1996) 183.
2. V. P. KOLOTOV, M. J. KOSKELO, J. Radioanal. Nucl. Chem., 233
Conclusions (1998) 95.
3. F. L. BRONSON, B. M. YOUNG, Mathematical Calibrations of Ge
Detectors and the Instruments that Use them, Proc. 5 th Annual
Using an extensive archive of ISOCS/LabSOCS NDA/NDE Waste Characterization Conference, Salt Lake City,
characterized detector measurements, it has been shown UT, Jan 11, 1997.
that the Cascade Summing Correction algorithm is an 4. R. VENKATARAMAN, F. BRONSON, V. ATRASHKEVICH,
effective and accurate method of compensating for B. M. YOUNG, M. FIELD, Nucl. Instr. Meth. Phys. Res., A442
cascade summing errors in cascading nuclide activities. (1999) 450.
5. J. F. BRIESMEISTER (Ed.), MCNP – A General Monte Carlo N-
Errors caused by cascade summing ranging over 20% Particle Transport Code, Version 4a, Los Alamos National
are typically reduced to 5% or less after correction. Laboratory Report LA-12625-M, November, 1993.
Using the CSC algorithm results in an insignificant 6. M. J. KOSKELO, R. VENKATARAMAN, V. P. KOLOTOV,
change in the measurement precision but does cause a J. Radioanal. Nucl. Chem., 248 (2001) 333.
discernable bias that is proportional to the amount of 7. F. DE CORTE, Ph.D. Thesis, Ghent University, The Netherlands,
1987.
correction. It is hypothesized that the small systematic

197

You might also like