Neutron beam experiments using nuclear research reactors: honoring the retirement of professor Bernard W. Wehring -II. 4. Accurate Characterization of the Shape of the HPGe Detector Peak Efficiency Curve for Application in PGNAA

In various situations, measurements in prompt gamma neutron activation analysis (PGNAA) are performed to determine the amount of an elemental impurity relative to that of a major constituent of the matrix. An example of this is the measurement of hydrogen concentration in a metallic matrix. In all such cases, a major contributor to the uncertainty in the measurement is the uncertainty in the ratio of the high-purity germanium (HPGe) detector full-energy peak efficiency for the gamma-ray lines of interest (i.e., impurity and matrix gammas). Usually, the ratio is derived from the relative peak efficiency curve, which is determined using isotopic standards that emit multiple gamma ray lines (e.g., 152Eu) in the energy range <3000 keV, or using prompt gamma radionuclides (e.g., 14N, 35Cl) in the energy range >3000 keV. In either case, the uncertainty in the ratio of the peak efficiency values derived from such measurements will be on the order of a few percent at best because the relative intensities of the calibration gamma lines are known to be ±2 to 3%. As a result, the best achievable measurement uncertainty will be dictated by that limit. In this paper, we consider the applicability of a high-accuracy method to determine the shape of the full-energy peak efficiency curve of an HPGe detector to PGNAA. To ensure maximum accuracy, this method uses calibration standards that emit at least two gamma-ray lines for which the relative intensity can be found to be better than ±0.1% (Refs. 2 and 3). If this condition is met, then the lower limit on the uncertainty in relative efficiencies measured using such a standard can be about ±0.1%. It has been shown that 24Na, 46Sc, 48Sc, 60Co, 94Nb, and 108mAg are appropriate calibration standards that can cover the energy range from 433 to 2754 keV (Refs. 3, 4, and 5). However, for this method to be usable in PGNAA, its applicability must be extended to ∼10 000 keV. The difficulty in doing this is the scarcity of radionuclides that can provide high-accuracy calibration points in the energy range from 2000 to 10 000 keV. Nevertheless, a hybrid experimental/computational approach can be used to address this problem. To achieve this, three steps must be completed: First, the shape of the full-energy peak efficiency curve using the high-accuracy calibration standards must be established up to 2754 keV. Second, a computational model that is benchmarked using the calibration data below 2754 keV must be constructed. Subsequently, this model is applied to extend the calibrated curve up to 10 000 keV. Reference 4 shows that if a detailed detector model is used in the simulation, agreement to within 60.2% can be achieved between the measurement and calculation. Finally, experimental benchmarking of the calculations at energies >3000 keV using prompt gamma calibration standards (e.g., 14N, 35Cl), albeit with an accuracy of about ±2 to 3%, can provide verification of the calculated curve. Recently, we initiated an effort to investigate detector calibration using the foregoing approach. We have used the MCNP-4C Monte Carlo code to perform simulations to establish the efficiency curve, using a simplified HPGe detector model, over the energy range extending from 400 to 10 000 keV (Ref. 6). Figure 1 shows the calculated relative full-energy peak efficiency curve for a 40% efficient closed-end coaxial HPGe. The curve was constructed using monoenergetic gamma lines covering the energy range from 400 to 10 000 keV. The result of each calculation is a differential pulse-height spectrum that is produced by the interaction of the gamma rays in the detector volume, and from which a full-energy peak efficiency value can be extracted. In addition, we used coupled neutron/photon calculations to investigate the production of prompt gammas from a source that contains 35Cl. Figure 2 presents the simulated spectrum, which upon evaluation was found to contain several inconsistencies with literature data. This includes missing gamma lines at 786, 788, 1951, 3062, 6628, and 7790 keV, which may be attributed to inaccuracies within the photon production data libraries included with MCNP-4C. Furthermore, for some gamma lines, the relative intensities can differ by as much as 50% from the measured values. Therefore, the simulation of any experiment that utilizes 35Cl should account for these discrepancies. As it stands, our current detector model requires benchmarking using measurements that utilize the sources listed in this paper. In addition, we will perform a measurement using 35Cl to benchmark the region above 2000 keV in Fig. 1. Note that once an accurate efficiency curve is produced, it could be used to validate and improve the accuracy of fundamental gamma-ray intensity data over the energy range extending up to 10 000 keV. (authors)