RADIOCAESIUM & PLUTONIUM ANALYTICAL METHODS

Gamma-spectroscopy

During nuclear reactor operations the fission of ²³⁵U results in the production of two smaller daughter nuclei which are in an excited state and the nuclei are said to be unstable. As the excited state nuclei return to their ground state energy levels, gamma ray photons are emitted which are characteristic of the difference in energy between the two levels (Brown & Firestone 1986). Thus, each fission product has a number of discrete gamma ray photon energies which can be detected and quantified using suitably calibrated counting equipment.

Gamma spectroscopic equipment

Gamma ray photons can be detected using either the scintillator Sodium Iodide Thallium-Drifted Nal(TI) or semiconductor Germanium (Ge) crystal detectors. The sodium iodide detector has good counting efficiencies at the low energy range of the spectrum around 200 keV with about 20% efficiency at 1 MeV (Livens 1985). The Nal(TI) type of detector is chiefly used for relatively simple spectra determination involving just a couple of nuclides. This detector type, however, suffers from poor peak resolution. Peak resolution is defined as the ability of the detector to resolve the main peak at full width at half maximum peak height (FWHM) i.e. peak resolution at an energy of 661.6 keV for ¹³⁷Cs is approximately 50 times lower using a Na(TI) relative to a High Purity Germanium (HpGe) semiconductor type.

The HpGe detector is a single large crystal of hyper-pure germanium which is capable of measuring gamma-ray photons of only a few keV (Radrem 1989). The detector type used in this work is a high purity germanium (HpGe) crystal (Gem Series EG&G Ortec) with a resolution of 2.2 keV (FWHM) at 1.33 MeV. Relative detector efficiency at 1.33 MeV for ⁶⁰Co was 86%. Relative detector efficiency is defined as the absolute germanium detector efficiency to the efficiency of a 3*3 NaI(Tl) scintillation detector at a source-detector distance of 25 cm which is calculated to be 1.2*10^-3 (EG&G 1992);

Relative Efficiency = (Peak area)/((activity)(live time))*100

• 1.2*10^-3

where peak area = number of counts in peak

activity = decays s^-1

• live time = real time - total system dead time (s)

  1.2*10^-3 = NaI(Tl) detector efficiency

The high counting efficiencies for this detector is a result of the relatively large size of the crystal which has a diameter of 80 mm and a length 64 mm. The hardware for this system is typical of modem gamma spectroscopic equipment. A 3000 v positive bias is fed to ‘activate’ the detector. The signals created from photon interactions with the crystal are amplified via a preamplifier and the signals are then passed through an analogue to digital converter (ADC). The pulse height of each event is related to energy and each pulse is assigned according to amplitude into a number of 2048 channels. When suitably calibrated each channel number can be related to the energy of the gamma-ray photon involved.

Energy calibration:

A mixed gamma-emitting standard from Amersham International was used to calibrate channel number with photon energy using data acquisition software Maestro. The Amersham International mixed nuclide calibration standard used for energy calibration for this work contained ²⁴¹Am, ¹⁰⁹Cd, ⁵⁷Co, ¹³⁹Ce, ²⁰³Hg, ¹¹³Sn, ⁸⁵Sr, ¹³⁷Cs, ⁶⁰Co and ⁸⁸Y. The low energy peak of ²⁴¹Am at 59.5 keV and a high energy peak at 1836.06 for ⁸⁸Y were used as markers to calibrate the system. After the system was energy-calibrated the next step was to determine efficiency of detection with energy range.

Efficiency calibration:

The efficiency calibrated standards used in this work were optimised with respect to energy range, sample mass, sample type and source-detector distance (Aycik 1992).

Of crucial importance in quantitative gamma rays spectroscopy is the physical distance between the sample container and the detector. The physical distance between sample and detector is optimised when using ‘Marinelli’ type beakers (Aycik 1992). It was not possible to use this type of container for this work because (a) the majority of samples (excluding bulk soils) were of low mass, possibly no more than 3 g - 5g maximum and (b) there were no commercially-available ‘Marinelli’ beakers available for a crystal this size.

For this particular detector advantage was taken of its large surface area by preparing a number of spiked sample standards in small 25 mm Petri dishes which gave a flat counting geometry. Efficiency losses from gamma ray self-absorption via sample density effects and further losses due to attenuation of gamma rays from the inverse-square law were thus minimised.

The gamma spectroscopy analysis package used in this work was the ^TMOmnigam software from EG&G Tennessee USA. The data from ^TMOmnigam was imported into the Jim Fitzgerald ^TMFitzPeaks UK which simplified the presentation of spectral data.

In essence the gamma ray analysis package identified each library peak for background, net area and peak centroid. Quantitative analysis was made possible using energy and efficiency correction files, background-stripping and decay-corrected values to give activity concentrations in units of Bq kg^-1.

The background for each peak was calculated by the 5-point, 3-point or 1 point averaging method. From the low energy side of a peak the 5-point average channel counts were calculated. The 5-point average is three times the calculated full width half maximum 3 FWHM on the low side of the peak and summing the contents of two channel counts either side of 3 FWHM and dividing by 5 (EG&G 1989). Peak-background subtraction can also be calculated for the 3 and 1-point method. The automatic mode was selected for this work where background subtraction was based on the characteristics of the desired spectrum e.g. it was decided that as small a number of nuclides were to be analysed which greatly simplified the background subtraction routines.

The calculation of gross peak area (assumed to be a singlet) was the sum of the channels between the low and high energy background channels and is defined below:

Gross area = h

å Ci

= I

where Ci is the data value of channel

I is the centre channel of the background calculation width at the low energy side of the spectrum h is the centre channel of the background calculation width at the high energy side of the spectrum

This method is known as total summation and precision is maintained even for small peaks. Count statistical error is the error in the gross area plus background in quadrature and the error in the gross area is the square root of the area. The program has a peak search based on the Maroscotti method (1967) and artifacts within a typical gamma rays spectrum due to backscatter peaks, sum peaks, the Compton plateau, double escape peaks, single escape peaks and the Compton edge were accounted for.

The reported uncertainty was composed of random and systematic errors. Random uncertainties relate to absorption corrections (not used in this work because of small sample mass) and count time. For this work errors were minimised by counting samples for a sufficiently long time as required. Systematic uncertainties are amplified as the sum of the user-defined nuclides increases, i.e. multiplet and deconvolution artifacts will be higher for a library containing 21 nuclides as opposed to 10 nuclides. Other systematic uncertainties relate to calibration source and geometry uncertainty. Calibration source uncertainty was specified by the supplier and geometry uncertainty was quoted at 2.5% (EG&G 1989).

The efficiency calibration file gave an efficiency value for each nuclide based on energy. This was defined as the fraction of gamma rays intercepted by the detector and counted within a specific energy channel. This value also included source-detector geometry configurations and a value of 1 means that detection efficiency was 100%. In practice, however, this was rarely achieved and average values for all sample matrices up to 800 keV for this work were between 12% - 19% depending on count geometry configuration.

A background correction file for all sample types was made using a low background matrix similar in composition to collected samples and counted for 250,000 s. The activities of ⁴⁰K and ²²⁸Ac were measurable and these activities were stripped from the activities of collected samples. The various stages of analysis are outlined below.

 (1) Spectrum File (2) Peak location and identification (3) Background subtraction

 (4) Efficiency corrected (5) Spectrum report

Preparation of air filter and soil standards

The soils around Sellafield are predominantly from the Wick 2 series which are deep well-drained coarse loamy soils (Soil Survey of England & Wales 1983). Soils with a similar composition (Essex series) to soils from around Sellafield were used to prepare a number of different standards in various count geometries.

Soil standards were made in 500 ml, 200 ml and 100 ml geometries for bulk surface soils. Count geometries of 50 ml and five 20 ml containers filled to different depths were prepared for soil size-fraction activities. The counting standard used for Frisbee material was prepared in 25 mm diameter Petri dishes. The count standard matrix for air samples was a glass fibre air filter that was spiked with a known amount of activity from a certified source. After spiking, the air filter was allowed to dry under an infra-red lamp (Bates personal communication 1993). The sample-detector distance was minimised and reproduced for all sample geometries by placing a ring of perspex on top of the material to be counted.

When preparing spiked standards it was important to ensure that the added tracer is homogeneously mixed as far as possible throughout the soil matrix. It is also clear that no two soils have the same physical and chemical characteristics, even from the same sample (Bowen & Volchock 1980). lt is probable, however, that the differences in composition and their effects on quantitative analysis between the low-level background soil used in this work and actual field samples will be negligible.

Soil standards were made up based on the methods of Sill et al (1974) and Bradley (personal communication 1993).

• (1) Soils were taken from an area of known low anthropogenic activity. The background activities of natural and anthropogenic radionuclides were determined by counting the unspiked samples for 250,000 s;

  (2) 500 g of soil was dried in an oven at a temperature of 40°C. until constant mass. The sample was lightly disaggragated by hand and passed through a 2 mm sieve to remove stones and vegetative matter;

  (3) The dried sample was ground using a mortar and pestle and passed through a 75 mm sieve and the remnants retained for future use;

  (4) The sieved material was quartered and coned and the sample re-screened a second time. This stage ensures that irregularly shaped particles do not go through to the next stage.

  (5) The sample was passed through a 45 mm sieve and the less than 45 mm material was re-combined with the material from the previous stage. Note mass of sample for this stage;

  (6) The material from stage 5 was transferred to a polyethylene sample bag and double-distilled water added to form a slurry;

  (7) A known amount of activity was carefully added to the sample solution and to ensure even distribution of activity the solution was stirred continuously with a silicone-coated rod; and

  (8) The spiked soil solution was placed in a freeze drier for a period of between 5-7 days to dry. After this period the standard was allowed to gradually warm up to 20⁰C. ambient temperature and transferred to a pre-weighed geometry container.

Spiked soil standards were analysed by ^TMOmnigam to form the efficiency calibration correction file. The efficiency calibration file was used to correct for detection efficiencies for all subsequent samples for that particular matrix.

Soil particle fractionation

The particle sizing technique used in this work was based on the high speed dispersion of fresh soil in distilled water and the physical isolation of particles within specific size ranges via wet-sieving and particle settling based on Stokes’ Law (Livens 1985).

Method:

Approximately 250 g of ‘fresh’ i.e. field-wet surface soil was lightly disaggregated by hand and dispersed in 1.5 l of distilled water by an electric stirrer for about 45 minutes. The dispersion of soils by chemical methods using either sodium pyrophosphate or other dispersants such as sodium hydroxide was not advisable because of possible desorption/activity losses, particularly for Pu species (Hetherington & Jeffries 1974).

The resulting suspension was applied to a series of 20 cm diameter sieves of aperture size 2 mm, 250 mm, 125 mm, 63 mm and 38 mm. The sieves were shaken mechanically for a period of 2 hours on a Gallenkamp shaker. Periodically, a small amount of distilled water was added to the top sieve to aid the fractionation process. Each sieve was washed in turn with the washings applied to the next sieve. Washings containing the less than 38 mm were retained for sedimentation. The material to all the other sieves were dried and transferred to weighed containers prior to g -spectroscopic analysis.

Material less than 38 mm was centrifuged and re-dispersed in 900 ml of distilled water by an electric stirrer. The beaker was covered and the suspension allowed to stand for a settling time of 7 hours 47 minutes for the less than 2 mm material. The first 10 cm of suspension was siphoned off and additional distilled water added to the beaker. The suspension was agitated further and the settling procedure repeated until the first 10 cm of the suspension was clear. This procedure was repeated at settling times of 29.2 minutes and 4.5 minutes to isolate the 2-8 mm and 20 mm fractions respectively. The remaining material contained the 20-38 mm fraction.

Plutonium analysis

One of the major problems of determining anthropogenic actinide activities in environmental samples is that they are present in very small quantities and for plutonium in particular, the sample matrices contain many interfering elements (Eakins 1984). Soils contain sodium, potassium and calcium whose concentrations are 3-4 orders of magnitude higher, with respect to mass than the actinides of interest. A further problem arises because some of the transition metals, rare earth’s and lanthanides behave similarly to the determinants. Soils also contain enhanced natural levels of radioactive materials such as ²³⁸U and ²³²Th and their decay-chain products. The activity of these naturally occurring radionuclides may be up to 2-3 orders of magnitude greater than the nuclides of interest. It is necessary to achieve a decontamination factor of all these interferences to at least 10⁴ in order to determine activity levels for background levels of plutonium (Eakins 1984).

Plutonium analysis requires a number of stages in which the matrix is dried, digested in mineral acids, put through ion exchange columns and finally electrodeposited onto stainless steel planchettes. The analytical procedure involves a number of stages in which sample losses can easily occur. A chemical yield tracer is added very close to the first stage of analysis and this enables any sample losses to be calculated. The chemical behaviour of the yield tracer must be the same as the determinant, the tracer must be in chemical equilibrium with the determinant and the tracer and determinant must be capable of being individually analysed (RADREM 1989). The tracer used in this work was ²⁴²Pu and the Pu separation method was based on the method used by Ballestra et al (1978) and Bates personal communication (1993).

Plutonium was separated from americium by anion exchange. Tetravalent Pu (IV) is strongly held by the anion exchange resin whilst trivalent Am (Ill) is eluted from the column. Plutonium is eluted from the column with a strongly reducing 9M hydrochloric-0.1M ammonium iodide solution. Americium (and curium if present) are co-precipitated with ferric hydroxide/phosphate, leaving in solution many interference elements such as aluminium titanium and anions in the phosphate form (Yamato 1982).

Other major contaminants such as uranium, thorium and iron are eluted from the column with 10M hydrochloric acid. A methyl alcohol - nitric acid solution was used to separate the actinides from the rare earth elements. A hydrochloric acid ammonium thiocyanate - methanol solution was used to separate the lanthanides from the actinides. Natural a -emitters such as ²¹⁰Po were strongly adsorbed onto the ion exchange column during the final elution stage of the americium separation (IAEA 295).

Following radiochemical separation plutonium was electrodeposited separately onto stainless steel planchettes from a sulphate salt electrolyte. The planchettes were counted for specific activities by alpha spectrometry.

Plutonium electrodeposition

Actinides have a short emission range in air and as such they must be deposited onto the planchette as a uniform deposit, counted under vacuum and should be free of contaminants such as iron, manganese or chlorides. Iron contamination results in a mottled dark appearance on the planchette surface as does microgram amounts of manganese and chlorine. Electrolytes containing chlorides for example, produce chlorine which etch the platinum wire and this results in gross contamination of the planchettes, the so-called platinum-black effect (Bennet personal communication 1991). This severely degrades the alpha spectrum. For best results the ideal deposit should be one atom thick, mono-element and almost weightless (Talvitie 1972 ;Kressin 1977).

During electrodeposition plutonium is deposited as a hydrous oxide onto a stainless steel planchette and in order to avoid premature hydrolysis during electrolysis, Pu is first fumed to dryness with sulphuric acid (Talvitie 1972).

Two electrolyte methods were used for the comparative electrodeposition of Pu. The International Atomic Energy Authority (IAEA) use a sulphuric acid -sodium sulphate solution which is carefully adjusted to a pH of 2 with the addition of ammonia. The second electrodeposition method is based on a 5% solution of sodium bi-sulphate buffered with it’s salt a 15% solution of sodium sulphate (Kressin 1977). The first method had the disadvantage that the pH of the electrolyte solution had to be checked frequently by dipping a narrow range pH paper into the solution. This may lead to unnecessary sample loss and lowered yield efficiencies. This first method was also time-consuming because it was very easy to miss the desired end-point pH.

The method used by Kressin (1977) was much simpler to use because the electrolyte solution is pre-adjusted to a pH of 2. The electrolyte is a solution of sodium-bisulphate and sodium sulphate. The ideal buffer for a strong acid such as sodium bisulphate is a salt of that acid such as sodium sulphate. Sodium sulphate is also a good electrolyte and this helps to keep the resistivity of the plating cell to a minimum. The excellent buffering qualities of the sodium bisulphate-sodium sulphate electrolyte ensures the pH range of the solution remained constant during electrodeposition. If the electrolyte solution turns alkaline during plating the efficiency of electrodeposition drops dramatically (Puphal et a/ 1972). A further advantage of this method is that the electrical resistance of the electrolyte is much lower than the IAEA method e.g. 52 W compared to 97 W .

Electrodeposition cell

The electrodeposition cell was based on a successful design previously used for the determination of actinides from Cumbrian soils (Evans 1991). The unit is in three parts (a) the body (b) the cap and (c) the electrode connections and ‘seat’ for the planchette. The body is made out of polypropylene and is commercially available from BDH. The major advantage of this unit is it’s cheapness, the body is simply thrown away after each run and this reduces the likelihood of cross contamination between samples.

The cap is made out of Perspex and it is a simple ‘push-fit’ onto the body. The central hole is used to position the platinum wire (0.015 mm) directly over the cathode regions of the cell. The two large holes in the cap provide a dual function of allowing evolved gases from the electrodeposition process to escape to atmosphere and also provide a means of injecting a sodium hydroxide solution into the cell prior to the end of electrodeposition.

The bottom cap provides a seated area for the planchette and the inclusion of a moulded lip within the inside of the cap allows an excellent leak-proof unit. A brass screw is pushed through the middle of the cap and this forms the cathode connection.

The diameter of the bottom cap is 23.5 mm wide and the planchette has a diameter of 22 mm with an active plating area of some 17 mm. The large active area of the planchette means that it’s inherent electrical resistance is low (Kressin 1977). The platinum wire is spiral shaped which helped to increase the surface area of the anode electrode and the anode-cathode distance was fixed at 5 mm. The unit is easily and quickly dismantled. The planchette has a smooth electropolished surface and it had a blue protective backing which was taken off immediately before use.

A number of electrodeposition experiments were carried out on this cell to ensure the unit worked satisfactorily without leaking and also to determine the optimum conditions for plating current and duration of electrodeposition.

ELECTRODEPOSITION PROCEDURE:

Two ml of 5% sodium bisulphate solution were added to a known activity of either ^239+240Pu or ²⁴¹Am and fumed to incipient dryness in a 50 ml beaker. An infra-red lamp was positioned over the beaker to ensure sample losses to the walls of the beaker were minimised. After fuming the beaker was allowed to cool and the inside walls of the beaker were washed with 2 ml of distilled water. One ml of concentrated hydrochloric acid was added to the beaker and the contents fumed to incipient dryness; this was repeated a second time. The contents of the beaker were dissolved in distilled water and allowed to stand for ten minutes before being placed in an electrodeposition cell.

The electrodeposition cells were prepared by assembling the units and rinsing the inside of the cells with two washes of a 50:50 mixture of nitric acid and distilled water. This was followed by at least five rinses with distilled water. Four ml of sodium sulphate was added to the assembled cell followed by the sodium bisulphate solution. The walls of the beaker containing the solution were washed with distilled water and this was added to the contents of the electrodeposition cell.

The first 6 samples were counted for 16 hours at 200 mA and plating recovery was 100%± 3% . The second batch of plating conditions (see figure 3.3) resulted in a reduced electrodeposition efficiency of ~72%± 4 which was possibly due to the electrolyte solution becoming too hot. Excessive heating of the electrolyte solution is caused by an excessive plating current relative to the surface area of the planchette. The third batch of plating conditions (see figure 3.4) resulted in a mean plating efficiency of 84%. The fourth batch of plating conditions, using 900 mA for 3 hours (see figure 3.5) resulted in a mean plating efficiency of 100%± 4.

To summarise the electrodeposition results, a current of 900 mA for 3 hours produced high reproducible electrodeposition efficiencies and these conditions were used for the electrodeposition of field samples. The electrodeposition cell is robust, leak proof and simple in operation. At the end of each electrodeposition run, one of the planchettes was turned over and counted to see if any activity had plated onto the wrong side of the disc: in all runs no activity was measurable. After electrodeposition a number of electrolyte solutions were re-acidified, buffered and electrodeposited a second time to see if any activity was left in solution; no activity was detected. A series of blanks were also run in two of the above experiments in which there was no detectable activity.

Alpha-spectrometric counting equipment

Eight ^TMUltra silicon-surface barrier detectors with an energy resolution of 20 keV (FWHM) and detector efficiencies of between 25-29% were used in this work. The equipment was supplied by EG&G (Ortec Octete PC^TM) USA. Data analysis was carried out via an eight-input multiplexer routed to a multichannel analyser (MCA) using Maestro ll data acquisition software. Each detector was first calibrated with respect to energy and efficiency using a certified source from UKAEA Fuel Services. System calibration was carried out in two parts using the energies of ²³⁹Pu at 5.16 MeV and ²⁴⁴Cm at 5.8 MeV to determine energy with channel using the following equation:

E_n = A + B * Chn

where A = the energy of the lowest channel

B = the gain in MeV/channel

E_n = the energy of channel Chn

The efficiency calibration was determined manually as the ratio between the number of alpha particles of a particular nuclide and energy emitted by the source and the number actually detected from that source in the corresponding spectral peak channel.

Each detector was mounted in an evacuated chamber and the sample to be counted was placed underneath the detector. The sample to detector distance was twice the diameter of the planchette to eliminate possible effects from energy re-coil (Holm 1984). Unlike gamma spectroscopy background counts for alpha spectrometry are normally present from the previous sample analysis, particularly if they are high activity samples.

Each chamber featured a manually operated interlock valve wired up to the detector bias on-off circuit. If vacuum is lost the bias is automatically cut-off to that detector. A vacuum is necessary for a-counting to prevent energy loss and spectral degradation due to gas-particle collisions (Livens 1985).

Sample activity was worked out via the following equation:

A = XP

Yw

where A = sample activity Bq kg-1

• X = is the number of counts in the peak from the sample after subtraction of background;

p = is the activity of yield determinant added (Bq);

• Y = is the number of counts in the peak from the yield determinant after subtraction of background;

w = is the mass of sample analysed (kg ashed weight).

The error for each sample is due to the uncertainty arising from the nature of the radioactive decay process and this was calculated according to the following:

s 1,2 = s 1² + s 2²

where s 1,2 = error in the ratio between tracer and sample peak

s 1 = error in tracer peak

s 2= error in sample peak

(Livens 1985)

Background correction files using just a counting tray and an unused planchette were also made for each individual chamber and an average value for a 24 hour count was less than 3E-04 s^-1 per chamber. Measured background values were subsequently subtracted from sample counts before sample activity was calculated relative to detector efficiency.

Background counts were taken in-between the counting of field samples and at least two out of twelve samples were either a blank or a known activity spike electrodeposited to a planchette.

Sample count times ranged from as little as three hours to five days and where sample activities were high, the chambers and detectors were de-contaminated before further samples analyses took place.

Decontamination of the chambers was carried out by placing a protective cover on the face of the detector prior to taking the detector out of the chamber. A cotton swab was dipped into methanol and the surface of the chamber was lightly cleaned and allowed to air-dry. The surface of the detector was lightly cleaned with a cotton swab moistened with acetone. After re-assembly the detector was left under vacuum for 45 minutes to remove all surface moisture. A background count of 250,000 s was carried out and background values subsequently stripped from the activity of field samples. Figure 3.6 and 3.7 illustrate typical alpha spectrum for a deposition and soil sample taken during run 2 at the Met. Station.

Figure 3.6

Figure 3.7

Quality control and interlaboratory comparison

It is important to know the accuracy of any methods used for radiochemical separations because this ensures continuous quality control and also served the purpose of providing a baseline for subsequent new methods (Bowen & Volchok 1980). A major problem in determining the ‘true’ value of a nuclide within a soil or other environmental matrices are the characteristics of sample homogeneity and how intimately the determinant is dispersed throughout the sample (see chapter 3.1.4).

There are two ways of determining the accuracy of a radiochemical method (a) Against an absolute standard e.g. the addition of a known amount of activity to a naturally low level sample (Sill & Hindman 1974) or (b) Obtain a large number of analyses of the same material from different workers using different methods to determine statistically a ‘true’ value for the determinants (Livens 1985).

The latter approach was used in this work and in 1993 triplicate alpha spectrometric analyses were carried out on a sediment sample (IAEA-300) for ^239+240Pu and ²³⁸Pu activities as part of a national and international interlaboratory comparison. Background details on how the IAEA-300 sample was prepared and where it came from can be found in IAEA/AL/064 (1994). All participants were informed that the expected concentration range for the artificial radionuclides in the sample were between 1-1000 Bq kg^-1 dry weight and less than 20 Bq kg^-1 dry weight for the transuranium elements.

One hundred and fifty-seven laboratories took part in this exercise and ²³⁸Pu activities were reported by 43 laboratories and ^239+240Pu by 50 laboratories. Over 74% of participants used conventional sample treatment e.g. partial dissolution of sample using concentrated nitric acid or Aquaregia. About 10% of laboratories used total dissolution methods such as alkaline fusion or a mixture of hydrofluoric, hydrochloric and nitric acid. Similarly, the majority of participants used ion-exchangers and or liquid-liquid extraction for radiochemical separation. Three laboratories used co-precipitation with lanthanum or cerium fluoride as opposed to the traditional method of electrodeposition as the final step.

Evaluation of results:

Sample averages were calculated as the arithmetical means with the standard deviation expressed at the 2 s count level. The statistical treatment of reported data included identifying and eliminating outliers, calculating the median and setting the confidence intervals at 0.05. The confidence intervals were taken from a non-parametric sample population representing a two-sided interval at a significance level of 0.05. The statistical treatment of alpha spectrometry results for the interlaboratory comparison assumes the data is not Gaussian in form and is suitable for analysis using non-parametric methods. Table 3.1 gives data on the statistically derived value based on measurements from 65 laboratories.

Table 3.1

Results for the interlaboratory comparison of Pu analyses

 Results for the interlaboratory comparison show that the plutonium activities for this work are on the high side of the statistically-derived median value of 0.15 Bq kg^-1 for ²³⁸Pu and 3.55 Bq kg^-1 for ^239+240Pu. The ²³⁸Pu activities, however, are within the confidence intervals of the dataset and were satisfactory. The value of ^239+240Pu for this work, however, was identified as an outlier. Of the total dataset, fourteen values were outside the confidence intervals outlined in Table 3.1 with three lower than 3.44 Bq kg^-1 and 11 higher than 3.65 Bq kg^-1. The high plutonium values for this work may indicate background interference from previously assayed samples (although 24 hour background counts were taken between each sample) or possibly, interference from natural thorium decay products particularly at the energies around 5.5 MeV - 5.8 MeV.

Top