Quantification of uranium-238 in environmental samples using gamma-ray spectrometry

A large number of environmental samples are routinely measured world-wide using gamma-ray spectrometry, some of its assets being easy sample preparation and comprehensive data for many radionuclides in one analysis. Although other techniques can be considered more suitable for analysing U in environmental samples, it is also routinely done by gamma-ray spectrometry. One mainly uses γ-ray emissions following the decay of the first daughter, Th, for determining the U activity. However, the low-energy gamma-rays at 63 keV and 92.5 keV are very difficult to quantify in a robust way due to high attenuation and interferences. This paper quantifies parameters affecting the possibility of making robust quantification of U via Th using gamma-ray spectrometry. It addresses the use of correct decay data, suitable detectors, optimised sample size, enhanced spectral amplification, correction for peak interferences and control of background.


Introduction
Gamma-ray spectrometry using HPGe-detectors is a widely used technique in laboratories monitoring environmental radioactivity.The possibility to quantify the activity of several radionuclides in one single analysis makes it an efficient analytical tool.However, the detection limits for various radionuclides vary greatly depending on the matrix in which they are.The detection limits of radionuclides emitting gamma-rays of low energy (< 100 keV) suffer from being in a matrix with other radionuclides emitting gamma-rays of higher energy due to interferences with X-rays and scattered gamma-rays.This is generally the case when quantifying 238 U in an environmental sample like soil.
Uranium-238 is a primordial radionuclide that is important to monitor.It decays by alpha-decay to 234 Th, which generates low-energy gamma-rays with very low emission probability.It can be measured with low detection limits using alpha-particle spectrometry and mass spectrometry.However, in order to save time on sample preparation and analysis, it is very common to measure 238 U using gamma-ray spectrometry [1].The γ-ray transitions in 234 Th (following the decay of 238 U) are too weak for this purpose.Instead the gamma-rays following the β-decays of 234 Th (T 1/2 = 24.1 d) and its daughter 234 Pa m (T 1/2 = 1.17 m) are used for identification and quantification of 238 U. Care must of course be taken to ensure that secular equilibrium exists between 234 Th and 238 U 1 .
In pure uranium samples the detection limits (in mBq) are low and the following gamma-rays can be used: 63 keV (doublet), 92.5 keV (doublet), 767 keV and 1001 keV.The gamma-rays at 767 and 1001 keV are robust for quantification but due to their low emission probabilities (0.317% and 0.842%, respectively) they are often swamped by the background from other radionuclides in the environmental sample itself.It is thus necessary to use the two low-energy gamma-lines.This introduces several problems that are outlined and quantified in this study, which was initiated through participation in the IAEA-CRP "Benchmarking Calibration for Low-Level Gamma Spectrometric Measurements of Environmental Samples".The first problem is that the four low-energy peaks appear as two doublets in a "normal" HPGe-detector spectrum.In this paper we will simply refer to the two doublet peaks as if they were singlet peaks at 63 keV and 92.5 keV.

1
In case it can be proven that equilibrium between 238 U and 226 Ra exist (e.g. in certain geological samples), the latter can be used as a measure of the 238 U activity.Furthermore, if one knows that there is a natural isotopic abundance of uranium isotopes in the sample and 235 U can be quantified, this provides an alternative way to calculate the 238 U activity without having to consider the radioactive equilibrium.

Analytical considerations 2.1 Use of correct decay data
Numerous measurements of the γ-ray intensities and energies following the decay of 234 Th to 234 Pa have been performed, and best values were derived from evaluations of the available data sets (see [2,3] and references therein).Apparently the emission probabilities vary significantly from one source to another.For example, for the most intense γ-ray at 63. 290 [2,3] have to be preferred, as they must have gained accuracy due to the recent measurement of the main peak intensity by Abousahl et al. [12].The intensities of the other gamma-rays were derived from relative intensity measurements by Godart and Gizon [13] in review [3], while review [2] relied on Chu and Scharff-Goldhaber [14], mainly because in this case the U KX-rays contributions were resolved from the gamma-ray peaks situated in the (90-115) keV energy region.In this paper we use I γ = 3.75 (8)% [2] for the 63.29 keV emission probability, although it is very common to find recent articles using much higher values, up to 4.8%.The review [2]

Use of a suitable detector
It is well-known in some fields of radionuclide analysis, like e.g.safeguards and nuclear fuel studies, that low energy gamma-ray emitters like 234 Th, 238-242 Pu and 241 Am, are best measured using a thin HPGe-detector with a thickness of only a few mm.However, when measuring 234 Th in environmental samples, one often has to make use of available instrumentations not optimised for the purpose of measuring gamma-rays below 100 keV or to compromise with the detection of other radionuclides emitting gammarays with higher energies.In recent years the sales of large volume p-type HPGe-detectors with thin upper deadlayer have increased.Such detectors provide high detection efficiency for gamma-ray energies over a wide range, which to some extent has taken over measurements previously performed with small n-type detectors.Two main advantages using "thin" HPGe-detectors for measurement of low energy gamma-rays are: (i) The low energy gamma-rays do not penetrate very deep.Figure 1 shows that for the 63 keV line, already with a detector thickness of 3 mm, the full-energy peak (FEP) efficiency has reached 96% of the FEP efficiency of a 30 mm thick detector.The rear part of the Ge-crystal, in this case, has no added value since it mainly increases the background at the lower energies.A Compton suppression system will help only to some modest extent, since its effect below 100 keV is limited unless a complicated arrangement with an extra detector "behind" the Ge-crystal is used [15].(ii) The resolution of a HPGe-detector degrades with increased size and particularly at low energies a high resolution is needed to resolve the many γ-rays, X-rays, backscatter and escape peaks lying close in energy.
The use of thin deadlayer detectors improves the robustness of quantification of low-energy gamma-rays since the slope of the FEP efficiency curve is much less steep compared to a detector with thick deadlayer (see figs. 2 and 3).Furthermore, the extent and effect of a deadlayer is difficult to assess accurately since it depends on the electrical field, which depends on doping profile and implantation/diffusion depth, and this can vary along a crystal axis or height.Therefore having a thin deadlayer reduces the impact of such changes as was shown by Johnston et al. [16].

Optimising the sample
Figure 4 shows the count rate of the FEP as a function of the sample thickness for different gamma-rays from a soil sample with a density of 1.3 g/cm 3 in a cylindrical container with an inner diameter of 50 mm.Also the continuum background from higher energy gamma-rays increases with sample size.However, for a "low-Z sample" with relatively low density, say around 2 g/cm 3 , it is possible to use relatively thick samples (50-100 mm) as the count rate of the 63 keV and 92.5 keV peaks might increase more than the square root of the background.It is important to know the concentration of heavy elements at percent level since they will influence the self-attenuation at low energies strongly.

Use of suitable amplification
When measuring with an HPGe-detector that is not very small, the amplification is often made such that one can detect the 2614 keV gammaray from 208 Tl.Using 8192 channels this results in a bin-width of about 0.33 keV.Such an amplification is not optimal for measuring gamma-rays below 100 keV, where the resolution (FWHM) of the detector can be 0.5 keV.In case of poop counting statistics, this can result in "peaks" of one or two channels.Such peaks are not suitable to fit with any gamma-ray spectrometry software.Furthermore the potential non-linearities of the energycalibration that are known to be greater at low energies will have a greater impact when a peak covers fewer channels.A more suitable amplification is to use e.g. the 662 keV line from 137 Cs as an upper limit near to channel 8192, giving a bin-width of 0.08 keV.Then a peak with a full-width-at-halfmaximum of 0.5 keV (FWHM) would be covered by roughly ten channels.The bin-width must also be considered when performing Monte Carlo simulations.At low energies 8 important detector response features need to be resolved in a narrow energy range (within 53 keV in the case of the 63 keV peak).Starting from the highest energy and going down, those features are: (i) the FEP, (ii) single low angle Compton scattering in the sample generating photons that can end up in the FEP, (iii+iv) the Ge K α and K β escape peaks, (v) the backscatter peak, (vi) the Compton edge (note that it is at lower energy than the backscatter peak), (vii+viii) Ge K α and K β peaks.

Subtraction of interfering peaks
Depending on the sample composition, there are many X-rays and gammarays that can interfere with the 63 and 92.5 keV lines.Since gamma-ray spectrometry of 238 U is a very common activity, there are numerous papers (see e.g.[17][18][19][20][21][22]) dealing with this and listing interfering peaks and ways to resolve this.For robust quantification of 238 U it is important that the interfering peaks are not too dominating.The main interference for the 92.5 keV doublet of 234 Th is the thorium X-ray K α1 (93.351 keV).As for the gamma-rays, the values available in literature for the intensities of Th X-rays present significant differences.The values used in this article are from the DDEP-website [2].In environmental samples, thorium X-rays can be originated from the decay of: 228 Ac (4.1(11)%), 235 U (5.76(14)%) and 238 U (0.0018(5)%).Considering the low emission probability of 238 U and a nominal 235 U/ 238 U activity ratio (0.046) the main interference come from 228 Ac Th K α1 .If the activity ratio 238 U( 234 Th)/ 228 Ra( 228 Ac) equals 0.95 in a sample, the count rate from the 92.5 keV doublet and the 93.3 X-ray are equal.Activity ratios near to 1 are often encountered in environmental samples and results thus in almost equally sized peaks.

Use of suitable stable reference materials
It is very important to calibrate measurements using good quality reference materials of a geometry and matrix that is similar to the samples that are being measured.Most reference materials do, however, suffer to a greater or lesser degree from aging due to chemical reactions, diffusion and evaporation.There are also known cases where reference materials have inhomogeneous distribution of 238 U, which can occasionally be localised in hot-spots in e.g.soil samples.It is therefore important not to be too reliant on one reference material.Furthermore in the case of the 63 and 92.5 keV peaks, there is very much influence of the sample composition on the self-absorption and interference corrections.Therefore it is dangerous to blindly use the detection efficiency obtained from a single reference sample.
An important consideration is to look at how a reference value was determined.For many primordial radionuclides in environmental samples it is not so easy with spiking of primary standards and reference values may have been determined as mean values from intercomparisons of non-expert laboratories, where systematic errors may have been hidden.

Calculating efficiency transfer and absolute efficiency
When calculating efficiency using Monte Carlo codes one needs to know geometrical details of the detector and the sample.Low energy gammarays are more sensitive to errors in these geometrical factors.Gasparro et al. [23] showed that failing to include e.g. the rounded edges of a crystal in a computer simulation, can alter the calculated efficiency of several tens of percent for energies below 100 keV.The other most crucial parameters are: Sample composition, container bottom thickness, detector window thickness, distance from crystal to window, deadlayer thicknesses (side and top).For low-energy gamma-rays it is common that calculated FEP efficiencies can be off with up to 100%.Therefore, it is strongly recommended to use efficiency transfer calculations as errors in the computer model more or less cancel out.Furthermore, Vidmar et al. [24] showed that different Monte Carlo codes generate the same results when calculating efficiency transfer factors.

Background
Perhaps the most obvious parameter for performing robust measurements is a good control of the background.There is a strong trend in placing HPGe-detectors underground [25], meaning at a depth of at least 10 m water equivalent [26].It is clear that by reducing the cosmic ray background one can reach lower detection limits.However, when the cosmic ray continuum reduces, the peaks from radionuclides in the background appear.It is thus easier to see if the background changes.Performing measurements with low-background is thus important not only for activities in the mBq range but also for activities in the range of a few hundreds of mBq to some Bq.These are activities frequently encountered in monitoring laboratories.Furthermore, having a lower background means that one can measure smaller samples, which has many practical advantages with regards to sampling and storage.
The lowest detection limits for 238 U using gamma-ray spectrometry is obtained in small swipe samples containing only uranium, measured in the underground laboratory HADES (225 m deep) [27].The detection limit derived from that measurement is 0.1 mBq corresponding to 10 ng U.

Conclusions
In order to improve robustness in determination of 238 U using gamma-ray spectrometry we can identify a few steps that may not be unrealistic to introduce although they all involve investments and spending longer time on each sample.
(i) Use of underground laboratory to reduce background in order to have better control of the background and reduce detection limits and robustness.
(ii) Use of two separate HPGe-detectors optimised for determination of low energy and high energy gamma-rays, or alternatively.
(iii) to make use of one large volume p-type detector with thin deadlayer with two chains of electronic with high amplification for low energies and low amplification for high energy gamma-rays.
It is important to use correct decay data and being aware that the peaks interfering with the 63 keV and 92.5 keV gamma-rays vary depending on the sample matrix.The problems with quantifying 238 U discussed in this paper are of course more severe at even lower energies like the 59.5 keV from 241 Am and 46.5 keV from 210 Pb.This was manifested e.g. in a recent IAEA intercomparison on 210 Pb in soil [28].All in all, one needs to exercise great care when analysing data below 100 keV and investigate any automatic peak fitting and interference correction carefully.Note also the abundant uncertain decay data in use in different laboratories.The discrepant decay data is to some extent a consequence of all the other sources of errors at low energies mentioned in this paper.

Figure 1 :
Figure 1: Monte Carlo calculations of the FEP efficiency as a function of the detector thickness for a planar crystal with a radius of 50 mm and a top deadlayer thickness of 0.5 μm.The sample was a 40 mm high cylinder of dried soil with density 1.3 g/cm 3 , located directly on the endcap.

Figure 2 :
Figure 2: (a) The FEP efficiency for a detector with a thick (0.7 mm) deadlayer and (b) the derivative of the curve in (a).The sample was dried soil with the following atomic composition: H(6), C(5), O(4), Al(4), Si(5) and Fe(1) and 1.5 g/cm 3 density.The sample was placed in a Teflon container with 0.2 cm bottom thickness and 5 cm internal radius.The height of the sample was 1 cm.The Teflon container was centered directly on the endcap.

Figure 3 :
Figure 3: (a) The FEP efficiency for a detector with a thin (0.5 μm) deadlayer and (b) the derivative of the curve in (a).The same sample as for figure 2.

Figure 4 :
Figure 4: The count rate of four gamma lines for a fixed detector from a sample with density 1.3 g/cm 3 as a function of the sample thickness.The assumed massic activity was 10 mBq/g of 238 U and 226 Ra.