Anal. Chem. 1997, 69, 2247-2250
The Accuracy of Instrumental Neutron Activation Analysis of Kilogram-Size Inhomogeneous Samples Menno Blaauw,* Olof Lakmaker, and Paul van Aller
Interfaculty Reactor Institute, Mekelweg 15, 2629 JB Delft, The Netherlands
The feasibility of quantitative instrumental neutron activation analysis (INAA) of samples in the kilogram range without internal standardization has been demonstrated by Overwater et al. (Anal. Chem. 1996, 68, 341). In their studies, however, they demonstrated only the agreement between the “corrected” γ ray spectrum of homogeneous large samples and that of small samples of the same material. In this paper, the k0 calibration of the IRI facilities for large samples is described, and, this time in terms of (trace) element concentrations, some of Overwater’s results for homogeneous materials are presented again, as well as results obtained from inhomogeneous materials and subsamples thereof. It is concluded that large-sample INAA can be as accurate as ordinary INAA, even when applied to inhomogeneous materials. Trace element analysis is usually performed on samples in the 100 mg range. Analysis of larger samples can be advantageous, e.g., when a representative 100 mg sample is difficult to obtain, when the trace element distribution is of interest, or when the sample must remain intact. Because of the penetrating power of both neutrons and γ rays, neutron activation analysis is the only likely candidate to offer possibilities in this field. At our institute, facilities and methods for instrumental neutron activation analysis (INAA) have been developed with mainly the representativity problem in mind. Others have focused on the determination of the element distribution by neutron-induced γ ray emission tomography1 or on the analysis of large, valuable objects with prompt γ ray neutron activation analysis (PGNAA).2 The feasibility of quantitative INAA without internal standardization of samples in the kilogram range has been demonstrated earlier by Overwater et al.3 In their studies, however, they demonstrated the agreement between the “corrected” γ ray spectrum of four large samples and that of small samples of the same materials. To this end, only interference-free γ ray lines could be used. Moreover, the materials used were homogeneous on the 100 mg scale. For such materials, the analysis of kilogramsize portions is unnecessarysresults from 100 mg subsamples would be equally valid and much more easily obtained. In this paper, results obtained from three materials are presented in terms of trace element concentrations. The first material, homogeneous phosphate ore, was also used by Overwater et al.,3 the second is mercury-contaminated soil, expected (1) Balogun, F. A.; Spyrou, N. M.; Adesanmi, C. A. Nucl. Instrum. Methods 1996, B114, 387-393. (2) Sueki, K.; Kobayashi, K.; Sato, W.; Nakahara, H.; Tomizawa T. Anal. Chem. 1996, 68, 2203-2209. (3) Overwater, R. M. W.; Bode, P.; De Goeji, J. J. M.; Hoogenboom, J. E. Anal. Chem. 1996, 68, 341-348. S0003-2700(96)01280-2 CCC: $14.00
© 1997 American Chemical Society
to be inhomogeneous with respect to trace elements distribution, and the third is wet river sludge, possibly inhomogeneous with respect to trace element concentrations but certainly with respect to water content since several centimeters of water were visibly standing on top of the sludge by the time of irradiation. Trace element inhomogeneity might result in less accurate results because of incorrect γ ray self-absorption computation, and water inhomogeneity leads to an inhomogeneous neutron density distribution in the sample during irradiation and might, therefore, also lead to inaccurate results. The large-sample facilities were calibrated using the k1 method,4 which is closely related to the k0 method.5 For all three materials, trace element concentrations in the large sample were determined by large-sample INAA as well as from numerous subsamples by traditional INAA using different irradiation and measurement facilities than were used for the large samples. The γ ray spectra were interpreted with the holistic method,6 where all full energy peaks are taken into account. The concentrations presented in this paper are the first concentrations ever measured in large samples by reactor-based INAA. EXPERIMENTAL SECTION Calibration of the Large-Sample Facilities. The efficiency curve of the 97% relative efficiency HpGe γ ray detector used for the large samples was determined using a certified 152Eu source. The center of mass of the source was positioned at a 20 cm distance from the end cap of the detectorsthe same position as used for the large samples. A polynomial defined by Gunnink7 was fitted to the measured efficiencies. Subsequently, a theoretical γ ray catalog in terms of counts per disintegration4 was computed for this geometry, without coincidence corrections. It was decided to disregard coincidence summing in this geometry because of the small geometrical efficiency of the detector. The validity of the resulting catalog was verified by measuring neutron-activated point sources both on the 17% relative efficiency detector to be used for some of the small subsamples and on the detector for the large samples and comparing interpretation results. The k0 flux parameters f (the thermal to epithermal neutron flux ratio) and R (needed to describe the epithermal flux distribution) were determined in both the BP3 irradiation facility of the Hoger Onderwijs Reactor (HOR) used for the irradiation of the small samples and the large-sample irradiation facility BISNIS. Zirconium and gold were irradiated in both facilities and measured to this end.5 The BP3 parameters were found to be f (4) Blaauw, M.; Bode, P. J. Radioanal. Nucl. Chem. 1993, 169, 201-208. (5) Simonits, A.; De Corte, F.; Hoste, J. J. Radioanal. Chem. 1975, 24, 31-46. (6) Blaauw, M. Nucl. Instrum. Methods 1994, A353, 269-271. (7) Gunnink, R. Nucl. Instrum. Methods 1990, A299, 372.
Analytical Chemistry, Vol. 69, No. 13, July 1, 1997 2247
Table 1. Sample Preparation Details for the Three Materials and Their Subsamples material
sample mass (kg)
water content (%)
primary subsample mass (mg)
secondary subsample mass (mg)
no. of subsamples
homogenized before subsampling
phosphate ore soil sludge
1.67 1.60 2.06
0 5 32
300 40 × 103 200
300 300 200
7 9 14
yes no yes
) 57.9 and R ) 0.04; for BISNIS, f ) 104, indicating a purely thermal flux in the BISNIS facility and rendering the R parameter superfluous. These flux parameters in conjunction with the k0 parameters listed in ref 8 describing the neutron capture cross section were used to convert the BP3 effective cross sections measured in the past9 to BISNIS effective cross sections. Neutron and γ Self-Attenuation Correction. Neutron and γ self-attenuation corrections were performed on the basis of the measured values of the neutron diffusion coefficient and the neutron diffusion length of the sample, determining the neutron density distribution during irradiation, and the γ ray transmission coefficients were measured separately. Analysis of Samples. Three sample materials were analyzed both by large sample INAA and by traditional INAA of the subsamples. Sample preparation details are given in Table 1. All large samples were irradiated in the BISNIS facility at a neutron flux of approximately 2.5 × 1012 m-2 s-1, and the γ ray spectra were acquired using the 97% relative efficiency detector. The subsamples were irradiated in the BP3 facility at a neutron flux of 4.5 × 1016 m-2 s-1, and the γ ray spectra were acquired using various independently calibrated detectors. The spectra obtained were interpreted6 using the nuclear data mentioned before.8,9 A blank correction was applied only to chromium results obtained through traditional INAA. Phosphate Ore. Prior to the large-sample INAA, this material was homogenized. Seven subsamples of this material were taken and analyzed by traditional INAA as described by Overwater et al.3 From the subsample results, it was concluded that the material was, in fact, homogeneous and suitable for use in the validation of the self-shielding correction method.3 Since a year had elapsed between Overwater’s experiments and the calibration of the large-sample facilities, during which the calibration of the detectors used for the subsamples as well as of the BP3 irradiation facility had slightly changed for some elements, the subsample spectra were interpreted in terms of element concentrations once more at the same time as the spectrum of the large sample. Mercury-Contaminated Soil. Two γ ray spectra of the soil sample were acquired 5 days and 5 weeks after the 24 h irradiation in the BISNIS facility, during 14 and 8 h, respectively. Normally, 16 flux monitors are used to determine the neutron flux close to the sample in this facility,3 but this time only eight were used. Two months after irradiation, the sample material was dried, weighed, and spread out over a surface of 25 cm × 19 cm. Nine subsamples of 50 g were taken at computer-generated random positions. Gravel, roots, etc. were not excluded. These 50 g subsamples were homogenized in a porcelain mortar, after which 300 mg quantities were taken from each subsample to be analyzed by traditional INAA. (8) Blaauw, M. The k0-consistent IRI γ-ray Catalogue for INAA; IRI: Delft, The Netherlands, 1996. (9) Blaauw, M. J. Radioanal. Nucl. Chem. 1995, 191, 387-401.
2248
Analytical Chemistry, Vol. 69, No. 13, July 1, 1997
River Sludge. The sample of river sludge was taken from the Nieuwe Waterweg close to Hook of Holland in a marine environment. Apart from segregated sediment, the sample bottles also contained a considerable amount of water. By the time the sample was irradiated, the sediment had settled, and a layer of water several centimeters high had become visible on top. In that state, the sample was irradiated during 24 h, and the γ ray spectrum was acquired 12 days later during 4 h. After 2 months, the sample was dried in air and roughly homogenized, and 14 subsamples were taken and analyzed by traditional INAA. Data Handling. From the subsample results, weighted average concentration jc were computed for each element, as well as the internal and external standard error of the mean (SEM). The average concentration jc was computed with N
jc )
ci
N
∑ ∑ i)1 s
/
2
i
1
(1)
2
i)1 s i
the internal standard error of the mean SEMint was computed with
1 SEMint ) 1/ N
x
N
∑
1
i)1 s
2
(2)
i
and the external standard deviation of the mean SEMext was computed with
SEMext )
1 N
x∑ N
(ci - jc)2
(3)
i)1
where N is the number of subsamples, ci the measured concentration in subsample i, and si its 1σ uncertainty computed from counting statistics. When the results from the large-sample INAA were to be compared with the results from the subsamples, z-scores were computed with
z)
cL - jc
xsL2 + SEMext2
(4)
where cL is the concentration obtained from the large sample and sL its 1σ uncertainty. RESULTS Neutron Diffusion Parameters. In Table 2, the neutron diffusion parameters10 D (the diffusion constant) and L (the (10) Overwater, R. M. W.; Hoogenboom, J. E. Nucl. Sci. Eng. 1994, 117, 141157.
Table 2. Neutron Diffusion Parameters Measured at Different Heights Relative to the Center of Mass of the Samples for the Three Sample Materials sample
height (cm)
D (cm)
L (cm)
phosphate ore
11.5 6.5 1.5 -3.5 2.5 -2.5 11.5 6.5 1.5 -3.5
1.5 1.4 1.6 1.7 0.90 0.93 0.37 0.25 0.57 0.95
11.7 10.7 11.3 10.4 11.5 12.1 5.3 3.9 6.2 9.7
soil sludge
Table 3. Subsample (SS) and Large Sample (LS) Concentrations, Uncertainties, and z-Scores Obtained for Phosphate Ore SS element
concn (mg/kg)
K Ca Sc Cr Fe Zn As Br La Ce Sm Eu Yb Lu Th U
1.90 × 102 3.60 × 105 1.45 × 100 6.43 × 101 8.67 × 102 1.11 × 102 5.04 × 100 1.94 × 100 1.76 × 101 1.97 × 101 2.20 × 100 6.34 × 10-1 1.77 × 100 2.73 × 10-1 1.37 × 100 5.44 × 101
LS
SEMint SEMext (%) (%) 15 1 0.7 2 7 6 2 2 0.2 5 2 5 2 13 7 0.3
20 2 1.4 3 13 6 2 3 1.5 5 10 3 3 13 7 2
concn (mg/kg) 1.76 × 102 3.22 × 105 1.48 × 100 6.04 × 101 8.84 × 102 1.12 × 102 4.88 × 100 1.80 × 100 1.70 × 101 2.11 × 101 2.73 × 100 6.90 × 10-1 1.75 × 100 2.70 × 10-1 1.45 × 100 5.12 × 101
precision z(%) score 15 5 1.1 4 15 10 2 3 0.3 13 2 5 4 2 14 0.9
-0.3 -2.2 1.3 -1.3 0.1 0.1 -1.2 -1.7 -2.3 0.5 2.1 1.3 -0.2 0.0 0.4 -2.6
diffusion length) obtained from the flux measurements at different heights relative to the center of mass of the samples are presented. Uncertainties, estimated from the reproducibility of the phosphate ore and soil, are about 8% for D and 5% for L. Concentrations. In Tables 3-5, the results are presented that were obtained for phosphate ore, mercury-contaminated soil, and river sludge, respectively. Concentrations with associated uncertainties exceeding 30% were omitted from the tables. DISCUSSION Neutron Diffusion Parameters. The results shown in Table 2 indicate that the water distribution in the river sludge sample is very inhomogeneous: The higher the vertical position, the more these parameters approach the values for water (D ) 0.14 cm, L ) 2.7 cm); the lower, the more they resemble the values of soil as observed in the other two samples. In those two samples, no such inhomogeneity can be observed. Since irradiation takes place in a graphite environment, a high concentration of water results in a severe neutron flux depression around and in the sample, so in the sludge sample, the neutron flux distribution appears to have been inhomogeneous, indeed. Concentrations. The results indicate good agreement between concentrations found in the large samples and the small samples, even though nine statistically significant discrepancies occur at the R ) 0.05 level and one at the R ) 0.01 out of a total of 61 results. The fact that large-sample INAA involves extra
Table 4. Subsample (SS) and Large Sample (LS) Concentrations, Uncertainties, and z-Scores Obtained for Mercury-Contaminated Soil SS element
concn (mg/kg)
K Ca Sc Cr Fe Co Zn As Br Rb Sb Cs Ba La Ce Sm Eu Tb Yb Lu Hf Ta Au Hg Th U
1.01 × 104 3.85 × 103 4.47 × 100 3.23 × 101 1.09 × 104 5.41 × 100 6.59 × 101 1.38 × 101 1.40 × 100 4.01 × 101 9.77 × 10-1 1.83 × 100 1.99 × 102 1.10 × 101 2.44 × 101 1.93 × 100 4.35 × 10-1 2.29 × 10-1 1.05 × 100 1.87 × 10-1 2.91 × 100 2.86 × 10-1 3.70 × 10-3 5.97 × 101 3.40 × 100 5.79 × 10-1
LS
SEMint SEMext (%) (%) 9 5 0.2 1.1 0.2 0.4 1.3 0.5 4 1.0 0.5 1.4 1.2 0.2 0.9 1.1 1.1 2 0.6 6 0.6 3 3 0.2 0.6 3
20 20 14 7 10 10 4 6 12 10 14 8 6 14 12 15 15 13 10 10 6 1 70 5 12 20
concn (mg/kg) 8.93 × 103 3.96 × 103 4.71 × 100 3.32 × 101 1.37 × 104 6.19 × 100 7.90 × 100 1.41 × 101 1.36 × 100 3.31 × 101 8.92 × 10-1 1.35 × 100 1.40 × 102 1.16 × 101 2.75 × 101 2.29 × 100 4.75 × 10-1 1.69 × 10-1 1.08 × 100 1.44 × 10-1 3.26 × 100 2.66 × 10-1 3.94 × 10-3 6.64 × 101 3.42 × 100 3.73 × 10-1
precision z(%) score 2 30 0.9 3 1.1 1.5 4 0.5 1.4 8 4 8 30 2 8 2 4 20 5 10 4 13 20 2 3 40
-0.6 0.1 0.4 0.4 2.4 1.5 3.1 0.3 -0.2 -1.5 -0.6 -2.6 -1.3 0.4 0.8 1.3 0.6 -1.3 0.2 -1.9 1.7 -0.4 0.1 2.2 0.0 -1.1
Table 5. Subsample (SS) and Large Sample (LS) Concentrations, Uncertainties, and z-Scores Obtained for River Sludge SS element
concn (mg/kg)
Ca Sc Cr Fe Co Zn Br Rb Zr Cs Ba Ce Sm Eu Tb Yb Hf Th
4.18 × 104 2.36 × 100 4.22 × 101 7.64 × 103 3.33 × 100 5.20 × 101 1.98 × 101 4.75 × 101 2.57 × 102 1.63 × 100 2.45 × 102 2.56 × 101 2.21 × 100 5.11 × 10-1 3.01 × 10-1 1.00 × 100 6.28 × 100 3.36 × 100
LS
SEMint SEMext (%) (%) 0.7 0.09 0.3 0.12 0.4 0.9 0.3 0.8 0.9 0.8 0.7 0.3 0.4 0.5 1.3 0.4 0.2 0.3
1.0 0.7 4 0.9 0.9 1.4 5 0.8 4 2 0.8 2 2 1.5 2.6 3 3 5
concn (mg/kg) 4.22 × 104 2.41 × 100 3.90 × 101 7.15 × 103 3.13 × 100 5.66 × 101 1.92 × 101 4.38 × 101 2.98 × 102 1.32 × 100 2.34 × 102 2.72 × 101 2.44 × 100 4.69 × 10-1 2.15 × 10-1 1.00 × 100 6.86 × 100 3.46 × 100
precision z(%) score 6 2 5 1.1 3 5 5 6 12 9 20 10 8 7 20 6 2 4
0.3 0.8 -0.5 -2.0 -1.7 1.6 -0.2 -1.4 1.2 -2.7 0.3 -0.6 1.2 -1.3 -2.2 0.0 0.9 -0.2
sources of uncertainty, such as the neutron flux distribution and γ ray self-absorption correction data and methods, that were not propagated to the uncertainties given for the large-sample results probably explains these discrepancies. For the subsamples, differences between SEMint, based on counting statistics only, and SEMext, based on the observed variation of the results, indicate the degree of inhomogeneity of the material as well as the presence of sources of variation not accounted for by counting statistics, such as the precision of the determination of the neutron flux (∼1%) and the reproduciblity of the sample positioning for the γ ray spectrum acquisition (also Analytical Chemistry, Vol. 69, No. 13, July 1, 1997
2249
Figure 1. It is clear that the gold concentrations are not normally distributed here and that the average of the subsample concentrations does not mean much. Nevertheless, the result from the large sample seems to be a reasonable value for the sample volume as a whole. Considering all results together, it seems that large-sample INAA is applicable not only to homogeneous materials but also to materials as inhomogeneous as the materials in this experiment, be it trace element distribution inhomogeneity or large-scale inhomogeneity with respect to neutron diffusion. Further research is required, however, to establish the relationship between the different types of inhomogeneity and the accuracy of largesample INAA.
Figure 1. Gold concentrations measured in subsamples from mercury-contaminated soil. The horizontal lines indicate the result obtained from the large sample.
∼1%). It turns out that both the phosphate ore and the river sludge were well homogenized, the ore from the start and the sludge at least prior to subsampling. The mercury-contaminated soil shows different degrees of inhomogeneity for different elements, with gold as the extreme case, further illustrated in
2250
Analytical Chemistry, Vol. 69, No. 13, July 1, 1997
CONCLUSIONS After calibration, the IRI facilities for large sample INAA prove to yield accurate trace element concentrations, even when applied to inhomogeneous materials.
Received for review December 18, 1996. Accepted March 7, 1997.X AC961280X X
Abstract published in Advance ACS Abstracts, May 1, 1997.
