VALIDATION OF THE ACCURACY OF THE LABSOCS MATHEMATICAL

EFFICIENCY CALIBRATION FOR TYPICAL LABORATORY SAMPLES

Frazier L. Bronson CHP Ram Venkataraman Ph.D.

fbronson@canberra.com rvenkataraman@canberra.com

Canberra Industries, Inc. 800 Research Parkway Meriden CT USA

Presented at the 46^th Annual Conference on Bioassay, Analytical, and Environmental Radiochemistry.

November 12-17, 2000 Seattle, Washington

LabSOCS [Laboratory Sourceless Object Calibration Software] is a computer program that performs mathematical efficiency calibrations of Ge detectors, without any use of radioactive sources by the laboratory user. This allows quick and accurate calibrations of many geometries that are difficult to do [non-water samples], take time to obtain the reference sources and make them simulate the sample, require radiochemistry knowledge to make the simulation accurate, and cost money to purchase the radioactive calibration standards and pay for the labor. Thus, the advantages of an accurate computer program to do this are obvious. The purpose of the tests summarized in this document is to show how accurate the mathematical calibration is.

The LabSOCS software is an improved version of the successful and widely accepted ISOCS [InSitu Object Calibration Software] mathematical in-situ efficiency calibration product. The improvements relevant to laboratory applications include a totally revised detector characterization method for increased accuracy at very close distances, new computational algorithms to improve attenuation corrections at close energies, and also a method to better describe the contour of a complex beaker shape for improved accuracy.

To show the accuracy of the calibration method, several documents are provided to the user. First is the "Validation and Internal Consistency" document, containing the results of 120 intercomparisons between ISOCS/LabSOCS and a reference method. For laboratory geometries [sources < 100cm distance] there were 53 tests, with 52 of them using NIST traceable radioactive sources as the reference method. Each test typically had 7-10 energies [about 400 data points for the laboratory sources], and generally covered the energy range from 60-1400 keV.

The LabSOCS efficiency at each of the ~400 data points was first compared to the reference method. The difference contains contributions from 3 major sources of variability:

· calibration source inaccuracy

· counting statistics variability, and

· inaccuracies in the LabSOCS method.

The LabSOCS contribution to the total was estimated by removing the variance of the counting statistics and of the stated reference source activity uncertainty, with the remainder being attributed to the LabSOCS method. The results of this analysis, as concluded in the Validation document indicate that for laboratory geometries the LabSOCS calibration energy-efficiency datapoints should be assigned a 7.1% sd for energies <150 keV, 6.0% sd for 150-400 keV, and 4.3% sd for 400-7000 keV. The Genie gamma spectral analysis software assigns this uncertainty automatically and propagates it along with other errors into the final result.

In the testing for the creation of the Validation document, eight different detectors were used, that were created by the routine production method. Therefore, we believe that this process presents an accurate evaluation of the capability of the LabSOCS method. However, since this was done on a previous group of detectors, not the specific detector owned by the user, it doesn't tell the user how well his particular detector performs.

To help show the user that his detector is OK, he also gets a Detector Characterization document. This document describes the detailed tests done on each individual detector, gives the predicted performance maps, provides the NIST certificate of the traceable sources that were used in the process, and provides other useful information. One set of data that shows the user his detector's performance is several analyses of a NIST traceable source. The Detector characterization process starts with this NIST traceable point source, which is measured in several locations. MCNP is then used in conjunction with the source measurements to create a map of the detector's spatial efficiency response. At the conclusion of the process, the same NIST traceable source is analyzed as an "unknown" using the efficiency generated by the LabSOCS calibration software. The user is presented with the results of this process, as shown in Table 1 for a typical detector.

Source located at 0 degrees		Measured Activity using ISOCS efficiency.		True Activity from manufacturer		Ratio of measured act. over true act.
Nuclide	E (keV)	uCi/unit	1 sd %	uCi/unit	1 sd %	Ratio	1 sd %
AM-241	59.54	5.25E+00	10.41	5.07E+00	3.6	1.03	11.00
EU-152	121.78	4.96E+00	8.23	4.95E+00	3.3	1.00	8.87
	344.27	4.93E+00	5.95	4.95E+00	3.3	1.00	6.80
	778.89	4.82E+00	4.46	4.95E+00	3.3	0.97	5.55
	1112.02	4.81E+00	5.38	4.95E+00	3.3	0.97	6.31
	1407.95	4.80E+00	6.69	4.95E+00	3.3	0.97	7.46
				Weighted Average		0.98	2.91
Source located at 90 degrees		Measured Activity using ISOCS efficiency.		True Activity from manufacturer		Ratio of measured act. over true act.
Nuclide	E (keV)	uCi/unit	1 sd %	uCi/unit	1 sd %	Ratio	1 sd %
AM-241	59.54	5.26E+00	10.37	5.07E+00	3.6	1.04	10.97
EU-152	121.78	5.03E+00	8.11	4.95E+00	3.3	1.02	8.75
	344.27	5.09E+00	5.76	4.95E+00	3.3	1.03	6.64
	778.89	4.98E+00	4.31	4.95E+00	3.3	1.01	5.43
	1112.02	5.00E+00	5.18	4.95E+00	3.3	1.01	6.14
	1407.95	4.92E+00	6.55	4.95E+00	3.3	0.99	7.33
				Weighted Average		1.01	2.85

Table 1 Comparison of the measured activity (using ISOCS efficiency) with the true activity for s/n xxxx.

In this Detector Characterization report, the user is shown how accurately his detector analyzed a NIST traceable "unknown" point source, with this source counted at approximately 1 meter away, both on axis, and at the side of the detector. While this test is very useful to monitor the quality of the characterization process, and while it does provide some information to the laboratory user on his exact detector, it isn't too helpful show the quality of calibrations of typical laboratory samples.

That is why we started new series of tests early 2000. These tests provide data for 3 purposes:

· to give the user proof that LabSOCS works well on his detector for typical sample geometries;

· to provide updated data about the overall LabSOCS accuracy for groups of detectors;

· to provide data supporting the accuracy of our new cascade summing correction software.

Four NIST traceable sources in typical laboratory geometries were procured from a reputable commercial laboratory in the following shapes:

· 51mm diameter filter paper

· 20cc Liquid Scintillation Counter vial

· 350cc beaker

· 2800cc Marinelli beaker

Each source contained the following nuclides: Am-241, Cd-109, Co-57, Ce-139, Hg-203, Sn-113, Cs-137, Mn-54, Y-88, Zn-65, and Co-60. This gives 13 well known energy lines for efficiency calibration use, covering the energy range from 60 to 1836 keV. Mn-54 and Zn-65 were added, since 5 of the other energy lines exhibit cascade summing [Y-88, Co-60, and sometimes Ce-139]. Figure 1 shows these sources.

All 4 sources are then counted on the customer's detector at contact and at 10cm [except for MB] for a total of 7 different acquisitions. Each source is carefully centered by placing on a disc with concentric circles. The 10cm distance is created by placing the source on the hollow plastic cylinder shown in the figure. For each of the 7 geometries, a LabSOCS efficiency calibration is generated using the to-be-delivered software. Each of the 7 spectra is then analyzed as an "unknown". The reported activity for each of the 13 energy lines is then compared to the "true" decay-corrected activity from the source certificate, and a report is then generated. This is the process that happens when the customer orders the ISOXVRFY product. Table 2, and Figure 2 from a typical ISOXVRFY customer report shows the results from one of the 7 verification counts.

Figure 1 The 4 NIST traceable sources, centering ring and 10cm spacer

Nuclide l	Energy (keV)	Meas Activity (LabSOCS eff) gammas/s	Statistical uncertainty (1s)	True Activity 06/12/2000 gammas/s	Source uncertainty (1s)	Meas/True		rel. uncert (1s)	Specified LabSOCS Uncert.
Am-241	59.5	1342.5	0.17%	1349.28	1.67%	0.99	1.68%		7.0%
Cd-109	88	1087.5	0.21%	1053.60	1.40%	1.03	1.42%		7.0%
Co-57	122	490.4	0.36%	466.09	1.57%	1.05	1.61%		7.0%
*Ce-139*	166	285.8	0.52%	303.17	1.37%	0.94	1.46%		6.0%
Hg-203	279	39.7	4.36%	37.60	1.37%	1.05	4.57%		6.0%
Sn-113	392	322.0	0.71%	298.87	1.33%	1.08	1.51%		6.0%
Cs-137	662	1247.5	0.42%	1130.09	1.47%	1.10	1.53%		4.3%
Mn-54	835	1565.2	0.41%	1457.65	1.67%	1.07	1.72%		4.3%
*Y-88*	898	622.6	0.81%	678.58	1.50%	0.92	1.70%		4.3%
Zn-65	1115	1225.8	0.89%	1178.53	1.67%	1.04	1.89%		4.3%
*Co-60*	1173	1855.3	0.43%	1970.33	1.47%	0.94	1.53%		4.3%
*Co-60*	1332	1833.6	0.60%	1982.94	1.53%	0.92	1.65%		4.3%
*Y-88*	1836	630.9	0.82%	711.98	1.37%	0.89	1.59%		4.3%
*Activities of Ce-139, Co-60, and Y-88 are underestimated because of gamma ray cascade summing losses. The beaker file used in the LabSOCS calculations is FILTER.BKR. The diameter of the source matrix used in LabSOCS calculations is 48 mm.

Table 2 Results from analyzing reference source as unknown (1 of 7)

From the results presented in the plots and tables, one can observe that the measured activities of Co-60, Y-88, and in some cases, Ce-139, are lower than their true activities [shown by the open circles in Figure 2]. This is because of gamma ray cascade summing (or true coincidence summing) losses in these nuclide measurements. The severity of cascade summing errors is dependent upon the decay scheme of a given nuclide and the total efficiency of the measurement geometry. The higher the total efficiency, the greater is the loss due to cascade summing. In other words, cascade summing losses will be more severe at smaller source-detector distances and with larger detectors. The detector shown here is a rather small one [25%], but the cascade summing errors are very significant, approximately 10%. The datapoints in Figure 2 with the open circles are those with cascade summing.

While Co-60 and Y-88 are will known cascade summing emitters, the sources used in the verification tests also contain Ce-139. The energy of the principal gamma ray emitted from the decay of Ce-139 is165 keV. This gamma ray undergoes true coincidence summing with low energy X-rays emitted from Ce-139. Therefore, cascade summing losses for Ce-139 are observable primarily in the case of measurements with low energy detectors [as is the case here] because of the absence of the germanium dead-layer. Nuclides and geometries where cascade summing is a major contributor to the efficiency bias are not included in the bias calculation in the ISOXVRFY report, but they will be used for future reports to evaluate the quality of the cascade summing software. Canberra Industries has developed a patented algorithm that corrects for cascade summing effects. This algorithm and the supporting software will be incorporated into Genie 2000 Version 2.0.

The data from the 7 individual count tables are then grouped and summarized in a final table in the customer's report, as shown in Table 3, for a typical detector. The purpose of this table is to give the estimated overall performance and compare it to the Validation Document's estimated performance.

ISOXVRFY	Data < 150 keV				Data 150 - 400 keV				Data > 400 keV
Geometry	Meas/True Ratio (avg)	Bias	LabSOCS unc (1s)		Meas/True Ratio (avg)	Bias	LabSOCS unc (1s)		Meas/True Ratio (avg)	Bias	LabSOCS unc (1s)
Filter Paper	1.03	2.6%	7.0%		1.07	6.6%	6.0%		1.07	7.3%	4.3%
(close)
Filter Paper	0.97	-2.7%	7.0%		1.06	6.5%	6.0%		1.02	1.7%	4.3%
(far)
20mL Cyl.	0.92	-7.9%	7.0%		0.99	-0.6%	6.0%		0.98	-1.7%	4.3%
(close)
20mL Cyl.	1.02	1.6%	7.0%		1.03	3.3%	6.0%		1.05	5.4%	4.3%
(far)
350mL Cyl.	0.94	-6.3%	7.0%		0.99	-0.7%	6.0%		0.98	-2.5%	4.3%
(close)
350mL Cyl.	1.00	-0.1%	7.0%		1.09	9.3%	6.0%		1.04	3.7%	4.3%
(far)
2.8L Marinelli	0.95	-5.1%	7.0%		1.03	3.2%	6.0%		1.00	0.1%	4.3%
(close)
Average (all)	0.97				1.04				1.02
% Std Dev	4.4%				3.9%				4.4%
Average Bias	3.8%					4.3%				3.2%
LabSOCS Uncertainty (1s) this det				4.1%				1.1%				4.0%

Table 3 Summary data for all 7 counts for a typical detector

Within Table 3, the results for each geometry are grouped into three energy regimes; (i) less than 150 keV, (ii) 150-400 keV and (iii) greater than 400 keV. For each energy regime, the following results are presented.

· The average value of the measured to true activity ratio, for nuclides within the energy range

· The bias in the ISOCS efficiency of this detector, obtained by calculating the deviation of the average value of the ratio from its true mean, the true mean being unity. Nuclides exhibiting cascade summing are not included.

· The estimated uncertainty in LabSOCS efficiencies, derived from Canberra's validation test database, for a group of detectors.

· The average value of the measured to true activity ratio for a given energy range, computed by pooling together the ratios from all seven geometries.

· The standard deviation of the ratios for a given energy range, computed by pooling together the ratios from all seven geometries.

· The average uncertainty in LabSOCS efficiencies for this specific detector, computed as the difference between the standard deviation of the ratios and the measurement uncertainties.

· The LabSOCS uncertainty for this detector, as calculated in the same manner as described in the Validation and Internal Consistency document

It is hoped that the ISOXVRFY report will be helpful to the customer proving to himself and to others that efficiencies created using the LabSOCS software will give correct analysis results on his system, and that laboratory-specific testing can be minimized and the software can then be quickly used for production counting.

All of the data from this process are also summarized and reviewed periodically as a quality control measure, to improve the process, and to add to the validation database. For this document, we have summarized the first 13 detectors that have been completed using the standard production process and present that data in Table 4.

Summary of first 13 ISOXVRFY detectors
	<150 keV	150-400 keV	>400 keV
Average Bias	1.0%	1.9%	-1.9%
Average |Bias|	5.5%	5.7%	4.6%
sd from ISOXVRFY	5.1%	5.1%	4.2%
sd from Validation Doc	7.1%	6.0%	4.3%

Table 4 Summary of accuracy and precision from first 13 detectors

This data indicates that the performance of the new detector characterization process is about the same before for high energies, but better than before at medium and low energies. This are the first detectors done with this new process, and some improvement was expected, however detectors being produced today should be even better. Observations of the full data set show that the later ones are better than the early ones. And, we are about to incorporate some additional steps to the detector characterization process which should even further reduce the uncertainty at these close-up distances.

But, even as this data currently stands [4-5% sd], it is probably as good or better than most source based calibrations. Source-based calibrations, even using these carefully manufactured sources had problems. The 51mm source actually was 48mm in effective diameter, which is a 4% difference in efficiency. Two "identically manufactured" source sets were about 2-3% different from each other. The Hg-203 activity on a few [but not all] of the sources keeps "changing" with time, at a rate inconsistent with the half-life, and appears to be 10-20% different than the correct activity. Even though the sources are in an epoxy matrix, apparently the Hg being redistributed as the source ages. And, there are the 10% [20-30% for larger detectors] due to cascade summing if Y-88, Ce-139, Co-60, or Eu-152 are used without correction. So, the 4-5% sd from the LabSOCS process certainly seems acceptable for most laboratory sample assay processes, and is certainly more convenient, and quick than source-based calibrations.