- Email: [email protected]
A gamma-ray spectrometry analysis software environment
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Applied Radiation and Isotopes journal homepage: www.elsevier.com/locate/apradiso
A gamma-ray spectrometry analysis software environment ⁎
G. Lutter , M. Hult, G. Marissens, H. Stroh, F. Tzika European Commission, Joint Research Centre (JRC-Geel), Retieseweg 111, B-2440 Geel, Belgium
H I G H L I G H T S Carlo code has been developed and validated using EGSnrc. • AA Monte radionuclide decay generator has been written and validated. • Detection can be calculated for High Purity Germanium detector. • A flexible efficiency and user friendly analysis software using MS Excel has been developed. • Functionalities and performances of the software are presented. • ®
A R T I C L E I N F O
A B S T R A C T
Keywords: Gamma-ray spectrometry Monte Carlo Decay generator Analysis software
At the JRC-Geel's RadioNuclide Metrology sector, a Monte Carlo code based on EGSnrc, and a general purpose calculation sheet implemented in Microsoft Excel®, have been developed to make the quantitative gamma-ray spectrometry analysis of samples simpler and more robust. The further aim is that the software can be used by non-experts in gamma-ray spectrometry e.g. external researchers using JRC-Geel's facilities through the EUFRAT transnational access scheme. This paper presents the developed Monte Carlo software and the functionality included in the calculation sheet.
1. Introduction The JRC-Geel's RadioNuclide Metrology (RN) sector is operating low-background or ultra-low-background High Purity Germanium (HPGe) detectors. Currently, 11 of such detectors are operated in the underground laboratory HADES and 5 detectors are operated in the RN sector above ground laboratory at JRC-Geel. All detectors have distinct characteristics (type, size and manufacturer) as a wide range of samples with different geometries, masses or matrices have to be measured. Since 2014, the facilities of the RN sector are accessible to external researchers through the transnational (or rather, "external", as also Belgian institutes can apply) access scheme called EUFRAT (European facility for nuclear reaction and decay data measurements) [https://ec. europa.eu/jrc/en/eufrat]. Up to date (March 2017), 22 projects have been approved by the external programme committee, for execution in the RN sector, 17 of which involve to some extent gamma-ray spectrometry. The wide analytical capacity and large number of users of the RN sector gamma-ray spectrometry laboratories resulted in a clear need to employ a data analysis system that is robust and user-friendly so that external users can, following to a short introduction, perform their own
⁎
data analyses. A number of different approaches had been tested in the past. Although commercial computer codes for gamma-ray spectrometry are generally user friendly, it was found that they were not suitable for all types of samples/projects. Furthermore, it is of utmost importance for the RN sector to have full insight and control of the way the calculations are performed. This triggered the development of a general purpose MS Excel® based calculation sheet named “GLysis”. “GLysis” offers the possibility to automatically incorporate spectra from Canberra Genie-2000 and Monte Carlo (MC) calculated Full Energy Peak (FEP) efficiency output data. An important aspect was to make the analysis robust by minimising copy/paste errors and need for manually entering data, while still maintaining flexibility. Another aspect of the design was to enable compliance with the ISO11929-2010 ISO standard. This paper describes this integrated system and discusses its performance and limitations. Gamma-ray spectrometry using e.g. HPGe detector is a secondary method in the sense that the measurement of the activity in an unknown sample is conducted based on measurements of reference sources. The latter can be either activity standards or Certified Reference Materials (CRM). Availability of suitable reference sources is often limited. Computational methods using MC simulations have been
Corresponding author. E-mail address: [email protected] (G. Lutter).
http://dx.doi.org/10.1016/j.apradiso.2017.06.045 Received 10 March 2017; Received in revised form 28 June 2017; Accepted 29 June 2017 0969-8043/ © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
Please cite this article as: Lutter, G., Applied Radiation and Isotopes (2017), http://dx.doi.org/10.1016/j.apradiso.2017.06.045
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
Table 1 Comparison between the gamma-ray emission intensities generated by the decay generator and the DDEP data. The median values and the median absolute deviation (MAD) are calculated from the relative differences between the MC calculation and the DDEP. nb=number of particles/photons (beta-particles, gamma-rays and X-rays). Nucl.
227
Ac Ac 133 Ba 214 Bi 60 Co 134 Cs 137 Cs 152 Eu 234m Pa 214 Pb 223 Ra 226 Ra 235 U 212 Bi 212 Pb 208 Tl All nucl. 228
All
Only X-rays
Only gamma-rays (Pγ > 1%)
Only gamma-rays (Pγ ≤ 1%)
nb
Median (%)
MAD (%)
nb
Median (%)
MAD (%)
nb
Median (%)
MAD (%)
nb
Median (%)
MAD (%)
91 282 13 276 12 19 9 132 145 30 107 14 77 44 13 46 1310
−98 −3.3 −0.003 0.09 0.0005 −0.002 −0.005 −0.003 −0.13 −0.46 −0.60 0.02 −0.79 −0.27 0.0001 0.14 −0.073
2 3.2 0.092 0.11 0.1372 0.019 0.106 0.278 4.40 0.47 2.16 0.09 1.71 62.96 0.1509 0.21 0.465
4 4 4 4 4 4 4 8 4 4 4 4 4 4 4 4 68
−92.6 −6.912 0.086 0.28 0.05 0.003 −0.112 0.91 −96.63 −0.918 −0.475 −0.03 −1.71 −52.00 −0.156 0.07 −0.12
0.1 0.004 0.001 0.02 0.10 0.012 0.002 0.85 0.01 0.003 0.001 0.03 0.01 0.01 0.003 0.01 0.36
65 208 2 186 4 3 1 95 119 15 70 4 44 21 4 21 862
−98 −2.2 −1.6 0.19 −3.77 −0.06 0.24 0.01 −0.4 −0.01 −0.6 −0.2 −0.8 −0.07 0.004 0.14 −0.07
2 2.2 1.6 0.21 3.93 0.04 – 0.26 15.4 3.27 0.3 0.4 2.3 0.16 0.004 0.05 83.25
0 17 7 20 2 8 1 14 0 5 7 1 6 4 2 6 100
– 0.001 −0.7 0.004 0.0005 0.004 0.0005 −0.01 – −0.001 −0.61 −0.001 −3.4 −0.001 0.1 0.1 0.00008
– 0.079 0.1 0.005 0.0005 0.007 – 0.02 – 0.006 0.03 – 3.0 0.004 0.1 0.1 0.15000
2001) and may be downloaded directly from the DDEP webpage (DDEP, 2004–2016). The decay generator uses such files as input. As not all radionuclides are available (e.g. 227Th), the user may construct the missing nuclear data file, a simple ASCII file, following available instructions for its basic structure. The decay generator is able to reproduce alpha, beta-, beta+ and EC decays. To simulate the decay the generator needs, as input, for each energy level, the energy of the level and the type of particle (alpha or beta- or beta+) with the associated energy. For simplification, emitted beta particles are not generated as a full beta spectrum but only at their mean energy. For each gamma-ray the following parameters must be provided: the gamma-ray energy, the emission probability (Pγ), the Kand the total internal conversion factors (aK and aT, respectively). Only K-shell X-rays are simulated by providing their energy and emission probability and the fluorescence yield. The DDEP ENSDF data file does not provide the fluorescence yield. Thus, the fluorescence yields are written in a separate ASCII file. Once all the decay data are read, the developed algorithm searches for all possible de-excitation paths to the ground state. For each path, a probability (Ppath) is calculated using the total probability for gamma transition including the conversion electrons (Pg) of all the de-excitation gamma-rays. To simulate decay, a de-excitation path is selected randomly weighted by its Ppath value. For each energy transition a random number is produced to generate either a conversion electron or a gamma-ray, according to aT. In the case of a gamma-ray emission, a photon of the given energy is generated. If a conversion electron is selected, a second random number is generated to choose if the emitted electron is coming from the K-shell or not, depending on aK. If a K-shell electron is selected, a third random value gives the energy of the electron according to the emission probabilities of the different K-shell electrons. Then the gamma-ray or the conversion electron is propagated until its energy is below the set threshold. For each transition in the selected de-excitation path, the same process is repeated. Once the ground state has been reached a new path is selected and a new event starts. If applicable, all possible radionuclide decay modes are simulated (in e.g. 152Eu; 152Eu disintegrates 72% to 152 Sm by EC and beta+ and 28% to 152Gd by beta-). The simulations assume that the gamma-ray emissions are isotropic.
proven to be a good alternative (albeit not as robust) to determine the FEP detection efficiencies. Even when a CRM is used to set up a FEP efficiency curve, in most cases, there is still the need to perform MC calculations to derive FEP corrections (efficiency transfer method) accounting for the (small) differences between the reference source and the analysed samples. At the RN sector calculations of FEP efficiencies and efficiency transfer factors are performed using the specially developed code "hpge3" using EGSnrc (Kawrakow et al., 2015). The “hpge3″ code includes also a radionuclide decay generator, to account for true coincidence summing (TCS) effects. 2. The EGSnrc toolkit and “hpge3″ code EGSnrc is a general purpose simulation package including the coupled transport of electrons and photons. The approach of EGSnrc is different from a ready to use software. It provides a set of packages/ classes written in C++ and a general purpose geometry library, which includes a set of particle sources and different propagation algorithms to simulate the interaction of the ionising particles. It is up to the users to write a programme for their specific needs in order to get the results for the quantities of interest (FEP efficiency in this case). The “hgpe3″ is based on the “EGS_Advanced_Application” C++ class (Kawrakow et al., 2017). Thanks to the EGSnrc C++ class library called “egspp”, the model of the experimental setup and the definition of source(s) are defined in an input text file. No programming language and compilation are required to change setup. The structure of the input file is defined by. Kawrakow et al. (2017). The EGSnrc toolkit does not include any radionuclide decay. Simulation of radionuclide decay is necessary in gamma-ray spectrometry to take into account the TCS effects which may occur when radionuclides with cascading gamma-rays are measured. For that purpose, a decay generator has been developed and included in “hpge3″ software. 2.1. The decay generator The decay generator is a C++ class which is added to the EGSnrc code but it can easily be included in other C++ MC codes. The class has been written in a way that allows new radionuclides to be added and nuclear decay data to be updated easily. The Decay Data Evaluation Project (DDEP) provides carefully evaluated decay data for frequently used radionuclides. The data are available in the ENSDF format (Tuli,
2.2. Decay generator performance The performance of the decay generator was assessed by comparing the gamma-ray intensities derived from the simulated decay scheme 2
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
with the nominal ones provided by DDEP (DDEP, 2004–2016) for several commonly analysed radionuclides. Table 1 presents the median value and the median absolute deviation (MAD) of the relative differences between the MC calculation and the DDEP. The MAD corresponds to the median of the absolute deviations from the median. The values are calculated for the following cases: a) all the emitted/generated particles, b) only the X-rays taken into account by the decay generator, c) only the gamma-rays with an emission probability below 1% and d) only the gamma-rays with an emission probability of > 1%. For all the radionuclides taken into account in the table, 1310 particles/photons in total, the overall median is better than 0.08% with a MAD of 0.47%. For this assessment the median and the MAD were chosen as being less sensitive to outliers and more representative of the decay generator performance compared to the mean and the standard deviation (Leys et al., 2013). The large deviation for some cases is due to the fact that a high amount of emissions with very low probability are included in the decay scheme. Consequently, if we focus on the gamma-rays and distinguish the low emission ones (Pγ ≤ 1%) and the main gamma lines (Pγ > 1%), the overall agreement for the main gamma-rays is better than 0.00008% with a MAD of 0.15% whereas for the low emission gammarays the median agreement is better than 0.08% nevertheless with a wide spread of the values (MAD > 80%). The same effect may be observed on nuclide specific basis. For the study, 10 billion decays have been generated for each nuclide. The default pseudo-random generator RANMAR (Marsaglia et al., 1990) included in EGSnrc has been used. The smallest Pγ taken into account is the 166.5 keV from 234 mPa (Pγ = 2.4·10−7%) which would give 24 counts, indicating that the discrepancy cannot be entirely attributed to the statistics; the observed discrepancies are higher than the statistical uncertainties (Marouli et al., 2017). An observed contribution of the discrepancy between the DDEP and the MC results is in some cases due to high uncertainty on the nuclear data, i.e. Pγ and internal conversion coefficients. These uncertainties are not taken into account in the MC simulations.
Fig. 2. Like Fig. 1 but zoom in the energy window [1350–1430] keV.
Fig. 3. FEP efficiency as a function of gamma-ray energy for the NPL volume source on the end-cap of the “Ge-T2″ detector. The blue, red and green points correspond respectively to the uncorrected experimental FEP efficiencies, the MC FEP (monoenergetic mode) and the experimental FEP efficiencies corrected for coincidence summing. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).
2.3. Monte Carlo code validation
2700 keV is 130 counts per second. The measured background of “Ge-T2″ has been added to the shown MC spectrum (0.6 counts per second between 40 and 2700 keV). Fig. 2 focuses on two 134Cs sum peaks, the 1400.6 keV (604.7 keV + 795.86 keV) and the 1406.67 keV (604.72 keV + 801.95 keV). The MC simulation reproduces the main gamma lines of 134 Cs with an agreement of better than 3% compared to the experimental data. Fig. 2 indicates that some small discrepancies come from not including peak-tailings and random summing. Fig. 3 presents the FEP efficiency curve derived for a volume source, made of a multi-nuclide liquid solution from NPL, placed directly on the end-cap of the detector “Ge-T2″. The coincidence summing correction factors have been calculated using the “hpge3″ code. The red points correspond to the MC results of single gamma emission (monoenergetic mode). The relative difference between the experimental FEP efficiencies corrected for coincidence summing and the calculated ones has a median value of −1.2% with a MAD of 1.9%. Below 120 keV, the difference between the experimental values and the MC are more pronounced reaching 26% for the 53.2 keV gamma line. This is due to the fact that the imprecisions of the MC model (structure of the dead-layer thickness) are more important at low energies, especially for a source placed directly on the endcap (Lépy et al., 2001).
The “hpge3″ MC code, including the decay generator and detector/ source models, was validated using reference point as well as volume sources. The models of the HPGe detectors were derived from manufacturer's information on measurable dimensions of the Ge-detector, crystal position when cooled, derived from radiographs of the detectors, and thickness of Ge deadlayer using standardized point sources from PTB. Fig. 1 compares the simulated and experimental energy spectra from a 134 Cs point source measured on the end-cap of the “Ge-T2″ detector, which is a coaxial HPGe detector of the RN laboratory, with a relative efficiency of 20% and a sub-millimetre dead-layer. The count rate between 40 and
3. Data analysis and activity calculation At the RN sector, the data acquisition is based on the Canberra Lynx module and Canberra Genie-2000 acquisition software (Canberra, 2013). The Genie-2000 software is also used to perform the spectrum analysis via the Canberra Interactive Peak Fit package (Canberra, 2009). However, for the activity calculations, specific software has been developed as described in the introduction. The tool is using MS Excel® and its integrated script language MS Excel® VBA. MS Excel® is widely
Fig. 1. Experimental (blue) and MC-simulated with “hpge3″ (red) spectra emitted by a point source with 134Cs on the endcap of “Ge-T2″. The measured background has been added to the MC spectrum. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).
3
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
In the case of a peak analysis with Genie-2000, the results (peak area, its uncertainty, the continuum and its uncertainty) are used and a colour code warns the user about the quality of the peak fit. The quality parameter is the ratio between the expected energy resolution and the measured one. “GLysis” includes also the possibility to proceed to a peak area calculation. It uses the procedure described in the ISO11929 version 2010 (Annex C.1–3) (ISO, 2010). The peak area is determined by summing the channel contents in the peak region and subtracting a constant, linear or third order polynomial continuum around the region of interest. At this level of development it cannot de-convolute peaks. For each gamma line, the massic activity and uncertainty will also be calculated with the associated decision threshold and detection limit according to the ISO11929 version 2000 (ISO, 2000) and version 2010 (to be able to compare with results from projects carried out before the 2010 standard was released). A colour code is also applied on the activity value to warn the user if the activity value is below detection limit and/or decision threshold. No interference is corrected, except in the case of the 186 keV gamma line, which is a doublet with contributions from 235U and 226Ra. In this case, the number of counts from 226Ra in that peak is automatically calculated using the activities of the radondaughters 214Pb and 214Bi (assuming secular equilibrium). The default uncertainties are given for the coverage factor k = 1 and the detection limit and decision threshold are given at 90% confidence level. Both parameters may be modified by the user at any time. For a radionuclide with multiple gamma-rays, a mean weighted by the inverse of the associated uncertainty on the counting statistics, is calculated from the activities of the different gamma-rays. Once the activity calculation per gamma line is completed, it is up to the user to select or not the gamma line to be used for the calculation of the weighted mean. Table 2 compares the activities measured for the NPL source (from Fig. 3) using the Genie-2000 and the “GLysis” peak analysis. The activities are the weighted mean of the selected gamma lines. In both cases the activities were calculated using the same gamma-rays. The FEP efficiencies were estimated by using pure MC simulations. The uncertainties on the FEP efficiency, which is the main component, were fixed to 5% for gamma-rays over 100 keV and 10% below 100 keV. All results agreed within the uncertainty with the reference values given in the last two columns of the table. The standard deviation between the selected gamma lines showed a better agreement between them by using the Genie-2000 Peak fit compared to the method implemented in “GLysis”. This effect is not reflected in the results due to use of the weighted mean. In case of 134 Cs, a high standard deviation between the selected gamma lines is observed when the “GLysis” peak search is used. Indeed the 152Eu, 563.99 keV gamma line (Pγ = 0.457%) is interfering with the 563.246 keV of 134Cs (Pγ = 8.342%). The peak area calculation implemented in “GLysis” performs well for simple cases, clear Gaussian peak without neighbour peak or interferences.
used, provides a well-known user interface, allows user modification if needed and Canberra is providing libraries to read the Genie-2000 data files. This solution has some limitations: the need of licences and that MS Excel® and VBA are not internationalised, only English settings are supported. As input, the developed MS Excel® file, “GLysis”, is able to read directly the Genie-2000 data file including the results of the spectrum analysis (peak positions and areas stored in CAM_CLS_PEAK data block) and the output of the “hpge3″ MC software. The analysis software is divided in two parts, the activity calculation and the efficiency transfer. 3.1. Activity calculation The activity of the radionuclides is calculated based on individual gamma-ray using the following formula:
A=
bkg Ctot − Cpeak − Ccontinuum ε MC exp sample εref . ε MC . ref
Pγ . tl
. e λ . td .
λ. tm . K cs . Kbr . K eq 1−e−λ . tm (1)
where C is the number of counts in the peak in the sample spectrum (tot ), in the background spectrum (bkg ) and in the continuum under the peak in the sample spectrum (continuum ), ε , the full energy peak efficiency, Pγ , the gamma-ray emission probability,Pγ tl , the live time, λ , the decay constant, td , the decay time, tm , the real clock time of the measurement, K cs , the coincidence summing correction factor, Kbr , the branching correction factor, K eq , the equilibrium correction factor. The superscript on ε shows if the full energy peak is calculated using Monte Carlo simulation (MC ) or determined experimentally (exp). The subscript on ε shows if it is determined for the sample or a reference (ref ). The massic activity is obtained by dividing the activity by the mass of the sample. The nuclear data for the most common radionuclides present in environmental samples are all entered in “GLysis” using the DDEP reference data. Updates can be realised by editing manually the nuclear data sheet or automatically by loading the DDEP ENSDF file. After loading the spectrum of the sample and the associated background spectra, the selection of the identified radionuclides is possible in “Analysis Settings” sheet. The default settings are assuming no equilibrium in the complete decay chains but only between some important primordial radionuclides: (i) 238U, 234Th and 234mPa, (ii) 226Ra and 222Rn-daughters, (iii) 228Ra and 228Ac, (iv) 228Th and 224Ra, 212Pb, 212 Bi, 208Tl, (v) 227Ac and 227Th. In addition, by default, only the main gamma-rays with Pγ > 1% will be used, except for some nuclides (e.g. 234mPa). All the default settings may be changed for each single nuclide if necessary. To import the FEP efficiencies, two options exist: the output files of the “hpge3″ simulations may be uploaded directly or the FEP efficiency values may be entered manually. In the first case, the software will automatically search for the appropriate efficiency. According to the provided data and settings, “GLysis” generates a sample specific analysis sheet. During the generation of the analysis sheet, the software searches for possible interferences with other radionuclides present in the nuclear data sheet. For each case, the user will be informed of the possible interferences. It is up to the user to decide if the suggested interferences are relevant or not. On the analysis sheet, gamma-rays are grouped by radionuclide and the activity calculation can be done automatically nuclide by nuclide or gamma-ray by gamma-ray according Eq. (1). Nuclear data and FEP efficiencies are filled automatically by an MS Excel® VBA script. The user just needs to enter the reference date and the sample mass. All parameters can be changed by the user if necessary. Concerning the peak analysis, three options are available: the Genie2000 peak fit, the peak area calculation included in “GLysis”, or manual insertion of the peak information by the user if the data are provided by another method or software.
3.2. FEP efficiency transfer The MC efficiency transfer method provides more accurate results compared to calculation of absolute FEP efficiencies by pure MC simulations (Lépy et al., 2001). A specific calculation module has been implemented in “GLysis” to enable to use this method in the activity calculation. The procedure (input data, radionuclide selection and analysis settings) is similar to the one described previously. A specific sheet is created to calculate the experimental FEP efficiency curve of the detector, after selection of the suitable gamma-rays and introduction of the reference source(s) information. The experimental FEP efficiencies are automatically corrected for coincidence summing using MC simulations if the needed MC data are provided. Two kinds of fit are possible, both included in Genie-2000, the socalled dual curve fit and the empirical fit (Canberra, 2009). Once the curve has been set, it can be used to obtain the FEP efficiencies in the 4
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
Table 2 Massic activity calculation for the NPL multi-nuclide liquid solution. The massic activities are the weighted mean of the selected gamma lines. The standard deviation of the selected gamma lines is also given. The results for 134Cs are given with and without taking into account the 563 keV gamma line which is interfering with the 152Eu, 563.99 keV gamma-rays. Nuclide
60
Co Cs Cs 134 Cs (no 563) 133 Ba 152 Eu 137 134
“GLysis”
Genie-2000
Reference value (NPL)
Bq/g
unc. (k = 1)
Stdev
Bq/g
unc. (k = 1)
Stdev
Bq/g
unc. (k = 1)
5.23 15.55 2.77 2.77
0.26 0.78 0.14 0.14
0.02 – 0.02 0.01
5.20 15.50 2.75 2.74
0.26 0.78 0.11 0.11
0.01 0.78 0.96 0.23
5.23 15.41 2.80
0.01 0.06 0.02
7.36 15.35
0.36 0.74
0.12 0.28
7.37 15.31
0.37 0.73
0.09 0.28
7.41 15.26
0.05 0.10
focussing on RN's needs; low energy X-rays (from L,M,N,…) shells are not taken into account and neither the full beta energy spectrum. These limitations have negligible impact on our sample analysis, as all HPGe detectors have an aluminium or copper end-cap. However, HPGe detectors with a Be entrance window may require a more thorough treatment in this respect. From the user point of view, the next development on “hpge3″ would be the simplification of the experimental set-up in the MC simulation, by writing a geometry composer using database of the different containers used at the RN sector, detector models and sample and source holders. The user would only need to create the sample geometry if it cannot be placed in a standard container. As it regards “GLysis”, a next development would be to reduce the dependency of Canberra Genie-2000, to get the full control of the analysis chain, by improving the peak area calculation and to implement correction for interferences. Automating the analysis procedure reduces the time of analysis, makes the results more robust, prevents from data transfer mistakes and serves traceability. Nevertheless, it is needed to set optimal or relevant diagnostics and warnings to alert the user if something goes wrong in the analysis chain. Even if the automatic procedure is working well, it is important that the user still controls all the parameters.
Fig. 4. Comparison between the fit using Genie-2000 and “GLysis”, using the same experimental efficiency curve of Fig. 3. Red squares and blue triangles correspond, respectively, to the residuals of the Genie-2000 and “GLysis” fits. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).
sample analysis. Geometry and matrix correction factors are also automatically calculated if the MC data are provided. To help the user, a list of required MC simulations (nuclides and gamma-ray energies) may be exported in an ASCII file. Fig. 4 shows the FEP efficiency curve fitted with Genie-2000 and with “GLysis” using the MS Excel® “Linest” function and the relative residuals. For both methods, a dual fit was done with the same gammarays and with a crossover at the energy of 121 keV. The median values of the residuals are, respectively, −0.1% (MAD = 1.1%) for the “GLysis” fit and 0.5% (MAD = 1.5%) with Genie-2000. In this example, the fit with “GLysis” performs slightly better than in Genie-2000.
References Canberra, 2009. Model S506 Interactive Peak Fit: User’s Manual, Canberra. Canberra, 2013. The Genie-2000 Operations Manual, including the S501 Gamma Analysis option, Canberra. Croymans, T., et al., 2017. Radiological characterization and evaluation of high volume bauxite residue alkali activated concretes. J. Environ. Radioact. 168, 21–29. DDEP, 2004-2016. Table of Radionuclides, Vol. 1-8, Monographie BIPM-5 BIPM, Sèvres, website: http://www.nucleide.org/DDEP_WG/DDEPdata.htm. ISO, 2000. 2000. ISO 11929-3:2000 Determination of the Detection Limit And Decision Threshold For Ionizing Radiation Measurements – Part 3: Fundamentals and Application to Counting Measurements by High Resolution Gamma Spectrometry, without the Influence of Sample Treatment. ISO, 2010. 2010. ISO 11929:2010 Determination of the characteristic limits (decision threshold, detection limit and limits of the confidence interval) for measurements of ionizing radiation — Fundamentals and application. Kawrakow, I., et al., 2015. The EGSnrc Code System: monte Carlo Simulation of Electron and Photon Transport (Technical Report PIRS-701). National Research Council Canada. Kawrakow, I., et al., 2017. EGSnrc C++ Class Library (Report PIRS-898). National Research Council Canada. Lépy, M.C., et al., 2001. Intercomparison of efficiency transfer software for gamma-ray spectrometry. Appl. Radiat. Isot. 55, 493. Leys, C., et al., 2013. Detecting outliers: do not use standard deviation around the mean, use absolute deviation around the median. J. Exp. Social. Psychol. 49, 764–766. Marouli, M., et al., 2017. Measurement of Absolute γ-Ray Emission Probabilities in the Decay of Ac-227 in Equilibrium with Its Progeny. in preparation. Marsaglia, G., Zaman, A., Tsang, W.-W., 1990. Toward a universal random number generator. Stat. Prob. Lett. 9, 35. Tuli, J., 2001. Evaluated Nuclear Structure Data File: A Manual for Preparation of Data Sets. BNL-NCS-51655-01/02-Rev. Uematsu, S., et al., 2017. Foliar uptake of radiocaesium from irrigation water by paddy rice (Oryza sativa): an overlooked pathway in contaminated environments. New Phytol. http://dx.doi.org/10.1111/nph.14416.
4. Discussion and conclusion Since 2015 the RN is performing sample analysis using the two complementary software, “hpge3″, including a decay generator, for the MC simulation, and “GLysis” for the activity calculations. These software have been validated through several proficiency tests and interlaboratory comparisons. The two software packages compose an integrated system which is optimised for gamma-ray spectrometry, using HPGe detectors. This integrated system is flexible and robust enough to meet the demanding requirements of the wide variety of analytical conditions (different detector types and non-routine project specific sample parameters) occurring in the RN laboratories. In the frame of the EUFRAT access scheme, several students and guest scientists conducted the whole analysis chain and produced results with a limited training time (see e.g.: (Uematsu et al., 2017) and (Croymans et al., 2017)). At this level of development, the decay generator is mainly
5
Contents lists available at ScienceDirect
Applied Radiation and Isotopes journal homepage: www.elsevier.com/locate/apradiso
A gamma-ray spectrometry analysis software environment ⁎
G. Lutter , M. Hult, G. Marissens, H. Stroh, F. Tzika European Commission, Joint Research Centre (JRC-Geel), Retieseweg 111, B-2440 Geel, Belgium
H I G H L I G H T S Carlo code has been developed and validated using EGSnrc. • AA Monte radionuclide decay generator has been written and validated. • Detection can be calculated for High Purity Germanium detector. • A flexible efficiency and user friendly analysis software using MS Excel has been developed. • Functionalities and performances of the software are presented. • ®
A R T I C L E I N F O
A B S T R A C T
Keywords: Gamma-ray spectrometry Monte Carlo Decay generator Analysis software
At the JRC-Geel's RadioNuclide Metrology sector, a Monte Carlo code based on EGSnrc, and a general purpose calculation sheet implemented in Microsoft Excel®, have been developed to make the quantitative gamma-ray spectrometry analysis of samples simpler and more robust. The further aim is that the software can be used by non-experts in gamma-ray spectrometry e.g. external researchers using JRC-Geel's facilities through the EUFRAT transnational access scheme. This paper presents the developed Monte Carlo software and the functionality included in the calculation sheet.
1. Introduction The JRC-Geel's RadioNuclide Metrology (RN) sector is operating low-background or ultra-low-background High Purity Germanium (HPGe) detectors. Currently, 11 of such detectors are operated in the underground laboratory HADES and 5 detectors are operated in the RN sector above ground laboratory at JRC-Geel. All detectors have distinct characteristics (type, size and manufacturer) as a wide range of samples with different geometries, masses or matrices have to be measured. Since 2014, the facilities of the RN sector are accessible to external researchers through the transnational (or rather, "external", as also Belgian institutes can apply) access scheme called EUFRAT (European facility for nuclear reaction and decay data measurements) [https://ec. europa.eu/jrc/en/eufrat]. Up to date (March 2017), 22 projects have been approved by the external programme committee, for execution in the RN sector, 17 of which involve to some extent gamma-ray spectrometry. The wide analytical capacity and large number of users of the RN sector gamma-ray spectrometry laboratories resulted in a clear need to employ a data analysis system that is robust and user-friendly so that external users can, following to a short introduction, perform their own
⁎
data analyses. A number of different approaches had been tested in the past. Although commercial computer codes for gamma-ray spectrometry are generally user friendly, it was found that they were not suitable for all types of samples/projects. Furthermore, it is of utmost importance for the RN sector to have full insight and control of the way the calculations are performed. This triggered the development of a general purpose MS Excel® based calculation sheet named “GLysis”. “GLysis” offers the possibility to automatically incorporate spectra from Canberra Genie-2000 and Monte Carlo (MC) calculated Full Energy Peak (FEP) efficiency output data. An important aspect was to make the analysis robust by minimising copy/paste errors and need for manually entering data, while still maintaining flexibility. Another aspect of the design was to enable compliance with the ISO11929-2010 ISO standard. This paper describes this integrated system and discusses its performance and limitations. Gamma-ray spectrometry using e.g. HPGe detector is a secondary method in the sense that the measurement of the activity in an unknown sample is conducted based on measurements of reference sources. The latter can be either activity standards or Certified Reference Materials (CRM). Availability of suitable reference sources is often limited. Computational methods using MC simulations have been
Corresponding author. E-mail address: [email protected] (G. Lutter).
http://dx.doi.org/10.1016/j.apradiso.2017.06.045 Received 10 March 2017; Received in revised form 28 June 2017; Accepted 29 June 2017 0969-8043/ © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
Please cite this article as: Lutter, G., Applied Radiation and Isotopes (2017), http://dx.doi.org/10.1016/j.apradiso.2017.06.045
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
Table 1 Comparison between the gamma-ray emission intensities generated by the decay generator and the DDEP data. The median values and the median absolute deviation (MAD) are calculated from the relative differences between the MC calculation and the DDEP. nb=number of particles/photons (beta-particles, gamma-rays and X-rays). Nucl.
227
Ac Ac 133 Ba 214 Bi 60 Co 134 Cs 137 Cs 152 Eu 234m Pa 214 Pb 223 Ra 226 Ra 235 U 212 Bi 212 Pb 208 Tl All nucl. 228
All
Only X-rays
Only gamma-rays (Pγ > 1%)
Only gamma-rays (Pγ ≤ 1%)
nb
Median (%)
MAD (%)
nb
Median (%)
MAD (%)
nb
Median (%)
MAD (%)
nb
Median (%)
MAD (%)
91 282 13 276 12 19 9 132 145 30 107 14 77 44 13 46 1310
−98 −3.3 −0.003 0.09 0.0005 −0.002 −0.005 −0.003 −0.13 −0.46 −0.60 0.02 −0.79 −0.27 0.0001 0.14 −0.073
2 3.2 0.092 0.11 0.1372 0.019 0.106 0.278 4.40 0.47 2.16 0.09 1.71 62.96 0.1509 0.21 0.465
4 4 4 4 4 4 4 8 4 4 4 4 4 4 4 4 68
−92.6 −6.912 0.086 0.28 0.05 0.003 −0.112 0.91 −96.63 −0.918 −0.475 −0.03 −1.71 −52.00 −0.156 0.07 −0.12
0.1 0.004 0.001 0.02 0.10 0.012 0.002 0.85 0.01 0.003 0.001 0.03 0.01 0.01 0.003 0.01 0.36
65 208 2 186 4 3 1 95 119 15 70 4 44 21 4 21 862
−98 −2.2 −1.6 0.19 −3.77 −0.06 0.24 0.01 −0.4 −0.01 −0.6 −0.2 −0.8 −0.07 0.004 0.14 −0.07
2 2.2 1.6 0.21 3.93 0.04 – 0.26 15.4 3.27 0.3 0.4 2.3 0.16 0.004 0.05 83.25
0 17 7 20 2 8 1 14 0 5 7 1 6 4 2 6 100
– 0.001 −0.7 0.004 0.0005 0.004 0.0005 −0.01 – −0.001 −0.61 −0.001 −3.4 −0.001 0.1 0.1 0.00008
– 0.079 0.1 0.005 0.0005 0.007 – 0.02 – 0.006 0.03 – 3.0 0.004 0.1 0.1 0.15000
2001) and may be downloaded directly from the DDEP webpage (DDEP, 2004–2016). The decay generator uses such files as input. As not all radionuclides are available (e.g. 227Th), the user may construct the missing nuclear data file, a simple ASCII file, following available instructions for its basic structure. The decay generator is able to reproduce alpha, beta-, beta+ and EC decays. To simulate the decay the generator needs, as input, for each energy level, the energy of the level and the type of particle (alpha or beta- or beta+) with the associated energy. For simplification, emitted beta particles are not generated as a full beta spectrum but only at their mean energy. For each gamma-ray the following parameters must be provided: the gamma-ray energy, the emission probability (Pγ), the Kand the total internal conversion factors (aK and aT, respectively). Only K-shell X-rays are simulated by providing their energy and emission probability and the fluorescence yield. The DDEP ENSDF data file does not provide the fluorescence yield. Thus, the fluorescence yields are written in a separate ASCII file. Once all the decay data are read, the developed algorithm searches for all possible de-excitation paths to the ground state. For each path, a probability (Ppath) is calculated using the total probability for gamma transition including the conversion electrons (Pg) of all the de-excitation gamma-rays. To simulate decay, a de-excitation path is selected randomly weighted by its Ppath value. For each energy transition a random number is produced to generate either a conversion electron or a gamma-ray, according to aT. In the case of a gamma-ray emission, a photon of the given energy is generated. If a conversion electron is selected, a second random number is generated to choose if the emitted electron is coming from the K-shell or not, depending on aK. If a K-shell electron is selected, a third random value gives the energy of the electron according to the emission probabilities of the different K-shell electrons. Then the gamma-ray or the conversion electron is propagated until its energy is below the set threshold. For each transition in the selected de-excitation path, the same process is repeated. Once the ground state has been reached a new path is selected and a new event starts. If applicable, all possible radionuclide decay modes are simulated (in e.g. 152Eu; 152Eu disintegrates 72% to 152 Sm by EC and beta+ and 28% to 152Gd by beta-). The simulations assume that the gamma-ray emissions are isotropic.
proven to be a good alternative (albeit not as robust) to determine the FEP detection efficiencies. Even when a CRM is used to set up a FEP efficiency curve, in most cases, there is still the need to perform MC calculations to derive FEP corrections (efficiency transfer method) accounting for the (small) differences between the reference source and the analysed samples. At the RN sector calculations of FEP efficiencies and efficiency transfer factors are performed using the specially developed code "hpge3" using EGSnrc (Kawrakow et al., 2015). The “hpge3″ code includes also a radionuclide decay generator, to account for true coincidence summing (TCS) effects. 2. The EGSnrc toolkit and “hpge3″ code EGSnrc is a general purpose simulation package including the coupled transport of electrons and photons. The approach of EGSnrc is different from a ready to use software. It provides a set of packages/ classes written in C++ and a general purpose geometry library, which includes a set of particle sources and different propagation algorithms to simulate the interaction of the ionising particles. It is up to the users to write a programme for their specific needs in order to get the results for the quantities of interest (FEP efficiency in this case). The “hgpe3″ is based on the “EGS_Advanced_Application” C++ class (Kawrakow et al., 2017). Thanks to the EGSnrc C++ class library called “egspp”, the model of the experimental setup and the definition of source(s) are defined in an input text file. No programming language and compilation are required to change setup. The structure of the input file is defined by. Kawrakow et al. (2017). The EGSnrc toolkit does not include any radionuclide decay. Simulation of radionuclide decay is necessary in gamma-ray spectrometry to take into account the TCS effects which may occur when radionuclides with cascading gamma-rays are measured. For that purpose, a decay generator has been developed and included in “hpge3″ software. 2.1. The decay generator The decay generator is a C++ class which is added to the EGSnrc code but it can easily be included in other C++ MC codes. The class has been written in a way that allows new radionuclides to be added and nuclear decay data to be updated easily. The Decay Data Evaluation Project (DDEP) provides carefully evaluated decay data for frequently used radionuclides. The data are available in the ENSDF format (Tuli,
2.2. Decay generator performance The performance of the decay generator was assessed by comparing the gamma-ray intensities derived from the simulated decay scheme 2
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
with the nominal ones provided by DDEP (DDEP, 2004–2016) for several commonly analysed radionuclides. Table 1 presents the median value and the median absolute deviation (MAD) of the relative differences between the MC calculation and the DDEP. The MAD corresponds to the median of the absolute deviations from the median. The values are calculated for the following cases: a) all the emitted/generated particles, b) only the X-rays taken into account by the decay generator, c) only the gamma-rays with an emission probability below 1% and d) only the gamma-rays with an emission probability of > 1%. For all the radionuclides taken into account in the table, 1310 particles/photons in total, the overall median is better than 0.08% with a MAD of 0.47%. For this assessment the median and the MAD were chosen as being less sensitive to outliers and more representative of the decay generator performance compared to the mean and the standard deviation (Leys et al., 2013). The large deviation for some cases is due to the fact that a high amount of emissions with very low probability are included in the decay scheme. Consequently, if we focus on the gamma-rays and distinguish the low emission ones (Pγ ≤ 1%) and the main gamma lines (Pγ > 1%), the overall agreement for the main gamma-rays is better than 0.00008% with a MAD of 0.15% whereas for the low emission gammarays the median agreement is better than 0.08% nevertheless with a wide spread of the values (MAD > 80%). The same effect may be observed on nuclide specific basis. For the study, 10 billion decays have been generated for each nuclide. The default pseudo-random generator RANMAR (Marsaglia et al., 1990) included in EGSnrc has been used. The smallest Pγ taken into account is the 166.5 keV from 234 mPa (Pγ = 2.4·10−7%) which would give 24 counts, indicating that the discrepancy cannot be entirely attributed to the statistics; the observed discrepancies are higher than the statistical uncertainties (Marouli et al., 2017). An observed contribution of the discrepancy between the DDEP and the MC results is in some cases due to high uncertainty on the nuclear data, i.e. Pγ and internal conversion coefficients. These uncertainties are not taken into account in the MC simulations.
Fig. 2. Like Fig. 1 but zoom in the energy window [1350–1430] keV.
Fig. 3. FEP efficiency as a function of gamma-ray energy for the NPL volume source on the end-cap of the “Ge-T2″ detector. The blue, red and green points correspond respectively to the uncorrected experimental FEP efficiencies, the MC FEP (monoenergetic mode) and the experimental FEP efficiencies corrected for coincidence summing. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).
2.3. Monte Carlo code validation
2700 keV is 130 counts per second. The measured background of “Ge-T2″ has been added to the shown MC spectrum (0.6 counts per second between 40 and 2700 keV). Fig. 2 focuses on two 134Cs sum peaks, the 1400.6 keV (604.7 keV + 795.86 keV) and the 1406.67 keV (604.72 keV + 801.95 keV). The MC simulation reproduces the main gamma lines of 134 Cs with an agreement of better than 3% compared to the experimental data. Fig. 2 indicates that some small discrepancies come from not including peak-tailings and random summing. Fig. 3 presents the FEP efficiency curve derived for a volume source, made of a multi-nuclide liquid solution from NPL, placed directly on the end-cap of the detector “Ge-T2″. The coincidence summing correction factors have been calculated using the “hpge3″ code. The red points correspond to the MC results of single gamma emission (monoenergetic mode). The relative difference between the experimental FEP efficiencies corrected for coincidence summing and the calculated ones has a median value of −1.2% with a MAD of 1.9%. Below 120 keV, the difference between the experimental values and the MC are more pronounced reaching 26% for the 53.2 keV gamma line. This is due to the fact that the imprecisions of the MC model (structure of the dead-layer thickness) are more important at low energies, especially for a source placed directly on the endcap (Lépy et al., 2001).
The “hpge3″ MC code, including the decay generator and detector/ source models, was validated using reference point as well as volume sources. The models of the HPGe detectors were derived from manufacturer's information on measurable dimensions of the Ge-detector, crystal position when cooled, derived from radiographs of the detectors, and thickness of Ge deadlayer using standardized point sources from PTB. Fig. 1 compares the simulated and experimental energy spectra from a 134 Cs point source measured on the end-cap of the “Ge-T2″ detector, which is a coaxial HPGe detector of the RN laboratory, with a relative efficiency of 20% and a sub-millimetre dead-layer. The count rate between 40 and
3. Data analysis and activity calculation At the RN sector, the data acquisition is based on the Canberra Lynx module and Canberra Genie-2000 acquisition software (Canberra, 2013). The Genie-2000 software is also used to perform the spectrum analysis via the Canberra Interactive Peak Fit package (Canberra, 2009). However, for the activity calculations, specific software has been developed as described in the introduction. The tool is using MS Excel® and its integrated script language MS Excel® VBA. MS Excel® is widely
Fig. 1. Experimental (blue) and MC-simulated with “hpge3″ (red) spectra emitted by a point source with 134Cs on the endcap of “Ge-T2″. The measured background has been added to the MC spectrum. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).
3
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
In the case of a peak analysis with Genie-2000, the results (peak area, its uncertainty, the continuum and its uncertainty) are used and a colour code warns the user about the quality of the peak fit. The quality parameter is the ratio between the expected energy resolution and the measured one. “GLysis” includes also the possibility to proceed to a peak area calculation. It uses the procedure described in the ISO11929 version 2010 (Annex C.1–3) (ISO, 2010). The peak area is determined by summing the channel contents in the peak region and subtracting a constant, linear or third order polynomial continuum around the region of interest. At this level of development it cannot de-convolute peaks. For each gamma line, the massic activity and uncertainty will also be calculated with the associated decision threshold and detection limit according to the ISO11929 version 2000 (ISO, 2000) and version 2010 (to be able to compare with results from projects carried out before the 2010 standard was released). A colour code is also applied on the activity value to warn the user if the activity value is below detection limit and/or decision threshold. No interference is corrected, except in the case of the 186 keV gamma line, which is a doublet with contributions from 235U and 226Ra. In this case, the number of counts from 226Ra in that peak is automatically calculated using the activities of the radondaughters 214Pb and 214Bi (assuming secular equilibrium). The default uncertainties are given for the coverage factor k = 1 and the detection limit and decision threshold are given at 90% confidence level. Both parameters may be modified by the user at any time. For a radionuclide with multiple gamma-rays, a mean weighted by the inverse of the associated uncertainty on the counting statistics, is calculated from the activities of the different gamma-rays. Once the activity calculation per gamma line is completed, it is up to the user to select or not the gamma line to be used for the calculation of the weighted mean. Table 2 compares the activities measured for the NPL source (from Fig. 3) using the Genie-2000 and the “GLysis” peak analysis. The activities are the weighted mean of the selected gamma lines. In both cases the activities were calculated using the same gamma-rays. The FEP efficiencies were estimated by using pure MC simulations. The uncertainties on the FEP efficiency, which is the main component, were fixed to 5% for gamma-rays over 100 keV and 10% below 100 keV. All results agreed within the uncertainty with the reference values given in the last two columns of the table. The standard deviation between the selected gamma lines showed a better agreement between them by using the Genie-2000 Peak fit compared to the method implemented in “GLysis”. This effect is not reflected in the results due to use of the weighted mean. In case of 134 Cs, a high standard deviation between the selected gamma lines is observed when the “GLysis” peak search is used. Indeed the 152Eu, 563.99 keV gamma line (Pγ = 0.457%) is interfering with the 563.246 keV of 134Cs (Pγ = 8.342%). The peak area calculation implemented in “GLysis” performs well for simple cases, clear Gaussian peak without neighbour peak or interferences.
used, provides a well-known user interface, allows user modification if needed and Canberra is providing libraries to read the Genie-2000 data files. This solution has some limitations: the need of licences and that MS Excel® and VBA are not internationalised, only English settings are supported. As input, the developed MS Excel® file, “GLysis”, is able to read directly the Genie-2000 data file including the results of the spectrum analysis (peak positions and areas stored in CAM_CLS_PEAK data block) and the output of the “hpge3″ MC software. The analysis software is divided in two parts, the activity calculation and the efficiency transfer. 3.1. Activity calculation The activity of the radionuclides is calculated based on individual gamma-ray using the following formula:
A=
bkg Ctot − Cpeak − Ccontinuum ε MC exp sample εref . ε MC . ref
Pγ . tl
. e λ . td .
λ. tm . K cs . Kbr . K eq 1−e−λ . tm (1)
where C is the number of counts in the peak in the sample spectrum (tot ), in the background spectrum (bkg ) and in the continuum under the peak in the sample spectrum (continuum ), ε , the full energy peak efficiency, Pγ , the gamma-ray emission probability,Pγ tl , the live time, λ , the decay constant, td , the decay time, tm , the real clock time of the measurement, K cs , the coincidence summing correction factor, Kbr , the branching correction factor, K eq , the equilibrium correction factor. The superscript on ε shows if the full energy peak is calculated using Monte Carlo simulation (MC ) or determined experimentally (exp). The subscript on ε shows if it is determined for the sample or a reference (ref ). The massic activity is obtained by dividing the activity by the mass of the sample. The nuclear data for the most common radionuclides present in environmental samples are all entered in “GLysis” using the DDEP reference data. Updates can be realised by editing manually the nuclear data sheet or automatically by loading the DDEP ENSDF file. After loading the spectrum of the sample and the associated background spectra, the selection of the identified radionuclides is possible in “Analysis Settings” sheet. The default settings are assuming no equilibrium in the complete decay chains but only between some important primordial radionuclides: (i) 238U, 234Th and 234mPa, (ii) 226Ra and 222Rn-daughters, (iii) 228Ra and 228Ac, (iv) 228Th and 224Ra, 212Pb, 212 Bi, 208Tl, (v) 227Ac and 227Th. In addition, by default, only the main gamma-rays with Pγ > 1% will be used, except for some nuclides (e.g. 234mPa). All the default settings may be changed for each single nuclide if necessary. To import the FEP efficiencies, two options exist: the output files of the “hpge3″ simulations may be uploaded directly or the FEP efficiency values may be entered manually. In the first case, the software will automatically search for the appropriate efficiency. According to the provided data and settings, “GLysis” generates a sample specific analysis sheet. During the generation of the analysis sheet, the software searches for possible interferences with other radionuclides present in the nuclear data sheet. For each case, the user will be informed of the possible interferences. It is up to the user to decide if the suggested interferences are relevant or not. On the analysis sheet, gamma-rays are grouped by radionuclide and the activity calculation can be done automatically nuclide by nuclide or gamma-ray by gamma-ray according Eq. (1). Nuclear data and FEP efficiencies are filled automatically by an MS Excel® VBA script. The user just needs to enter the reference date and the sample mass. All parameters can be changed by the user if necessary. Concerning the peak analysis, three options are available: the Genie2000 peak fit, the peak area calculation included in “GLysis”, or manual insertion of the peak information by the user if the data are provided by another method or software.
3.2. FEP efficiency transfer The MC efficiency transfer method provides more accurate results compared to calculation of absolute FEP efficiencies by pure MC simulations (Lépy et al., 2001). A specific calculation module has been implemented in “GLysis” to enable to use this method in the activity calculation. The procedure (input data, radionuclide selection and analysis settings) is similar to the one described previously. A specific sheet is created to calculate the experimental FEP efficiency curve of the detector, after selection of the suitable gamma-rays and introduction of the reference source(s) information. The experimental FEP efficiencies are automatically corrected for coincidence summing using MC simulations if the needed MC data are provided. Two kinds of fit are possible, both included in Genie-2000, the socalled dual curve fit and the empirical fit (Canberra, 2009). Once the curve has been set, it can be used to obtain the FEP efficiencies in the 4
Applied Radiation and Isotopes xxx (xxxx) xxx–xxx
G. Lutter et al.
Table 2 Massic activity calculation for the NPL multi-nuclide liquid solution. The massic activities are the weighted mean of the selected gamma lines. The standard deviation of the selected gamma lines is also given. The results for 134Cs are given with and without taking into account the 563 keV gamma line which is interfering with the 152Eu, 563.99 keV gamma-rays. Nuclide
60
Co Cs Cs 134 Cs (no 563) 133 Ba 152 Eu 137 134
“GLysis”
Genie-2000
Reference value (NPL)
Bq/g
unc. (k = 1)
Stdev
Bq/g
unc. (k = 1)
Stdev
Bq/g
unc. (k = 1)
5.23 15.55 2.77 2.77
0.26 0.78 0.14 0.14
0.02 – 0.02 0.01
5.20 15.50 2.75 2.74
0.26 0.78 0.11 0.11
0.01 0.78 0.96 0.23
5.23 15.41 2.80
0.01 0.06 0.02
7.36 15.35
0.36 0.74
0.12 0.28
7.37 15.31
0.37 0.73
0.09 0.28
7.41 15.26
0.05 0.10
focussing on RN's needs; low energy X-rays (from L,M,N,…) shells are not taken into account and neither the full beta energy spectrum. These limitations have negligible impact on our sample analysis, as all HPGe detectors have an aluminium or copper end-cap. However, HPGe detectors with a Be entrance window may require a more thorough treatment in this respect. From the user point of view, the next development on “hpge3″ would be the simplification of the experimental set-up in the MC simulation, by writing a geometry composer using database of the different containers used at the RN sector, detector models and sample and source holders. The user would only need to create the sample geometry if it cannot be placed in a standard container. As it regards “GLysis”, a next development would be to reduce the dependency of Canberra Genie-2000, to get the full control of the analysis chain, by improving the peak area calculation and to implement correction for interferences. Automating the analysis procedure reduces the time of analysis, makes the results more robust, prevents from data transfer mistakes and serves traceability. Nevertheless, it is needed to set optimal or relevant diagnostics and warnings to alert the user if something goes wrong in the analysis chain. Even if the automatic procedure is working well, it is important that the user still controls all the parameters.
Fig. 4. Comparison between the fit using Genie-2000 and “GLysis”, using the same experimental efficiency curve of Fig. 3. Red squares and blue triangles correspond, respectively, to the residuals of the Genie-2000 and “GLysis” fits. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).
sample analysis. Geometry and matrix correction factors are also automatically calculated if the MC data are provided. To help the user, a list of required MC simulations (nuclides and gamma-ray energies) may be exported in an ASCII file. Fig. 4 shows the FEP efficiency curve fitted with Genie-2000 and with “GLysis” using the MS Excel® “Linest” function and the relative residuals. For both methods, a dual fit was done with the same gammarays and with a crossover at the energy of 121 keV. The median values of the residuals are, respectively, −0.1% (MAD = 1.1%) for the “GLysis” fit and 0.5% (MAD = 1.5%) with Genie-2000. In this example, the fit with “GLysis” performs slightly better than in Genie-2000.
References Canberra, 2009. Model S506 Interactive Peak Fit: User’s Manual, Canberra. Canberra, 2013. The Genie-2000 Operations Manual, including the S501 Gamma Analysis option, Canberra. Croymans, T., et al., 2017. Radiological characterization and evaluation of high volume bauxite residue alkali activated concretes. J. Environ. Radioact. 168, 21–29. DDEP, 2004-2016. Table of Radionuclides, Vol. 1-8, Monographie BIPM-5 BIPM, Sèvres, website: http://www.nucleide.org/DDEP_WG/DDEPdata.htm. ISO, 2000. 2000. ISO 11929-3:2000 Determination of the Detection Limit And Decision Threshold For Ionizing Radiation Measurements – Part 3: Fundamentals and Application to Counting Measurements by High Resolution Gamma Spectrometry, without the Influence of Sample Treatment. ISO, 2010. 2010. ISO 11929:2010 Determination of the characteristic limits (decision threshold, detection limit and limits of the confidence interval) for measurements of ionizing radiation — Fundamentals and application. Kawrakow, I., et al., 2015. The EGSnrc Code System: monte Carlo Simulation of Electron and Photon Transport (Technical Report PIRS-701). National Research Council Canada. Kawrakow, I., et al., 2017. EGSnrc C++ Class Library (Report PIRS-898). National Research Council Canada. Lépy, M.C., et al., 2001. Intercomparison of efficiency transfer software for gamma-ray spectrometry. Appl. Radiat. Isot. 55, 493. Leys, C., et al., 2013. Detecting outliers: do not use standard deviation around the mean, use absolute deviation around the median. J. Exp. Social. Psychol. 49, 764–766. Marouli, M., et al., 2017. Measurement of Absolute γ-Ray Emission Probabilities in the Decay of Ac-227 in Equilibrium with Its Progeny. in preparation. Marsaglia, G., Zaman, A., Tsang, W.-W., 1990. Toward a universal random number generator. Stat. Prob. Lett. 9, 35. Tuli, J., 2001. Evaluated Nuclear Structure Data File: A Manual for Preparation of Data Sets. BNL-NCS-51655-01/02-Rev. Uematsu, S., et al., 2017. Foliar uptake of radiocaesium from irrigation water by paddy rice (Oryza sativa): an overlooked pathway in contaminated environments. New Phytol. http://dx.doi.org/10.1111/nph.14416.
4. Discussion and conclusion Since 2015 the RN is performing sample analysis using the two complementary software, “hpge3″, including a decay generator, for the MC simulation, and “GLysis” for the activity calculations. These software have been validated through several proficiency tests and interlaboratory comparisons. The two software packages compose an integrated system which is optimised for gamma-ray spectrometry, using HPGe detectors. This integrated system is flexible and robust enough to meet the demanding requirements of the wide variety of analytical conditions (different detector types and non-routine project specific sample parameters) occurring in the RN laboratories. In the frame of the EUFRAT access scheme, several students and guest scientists conducted the whole analysis chain and produced results with a limited training time (see e.g.: (Uematsu et al., 2017) and (Croymans et al., 2017)). At this level of development, the decay generator is mainly
5