ARTICLE IN PRESS

Applied Radiation and Isotopes 67 (2009) 678–682

Contents lists available at ScienceDirect

Applied Radiation and Isotopes

0969-80

doi:10.1

� Tel.:

E-m

journal homepage: www.elsevier.com/locate/apradiso

A UK comparison for measurements of low levels of gamma-emitters inwaste drums

Julian Dean �

National Physical Laboratory, Hampton Road, Teddington, Middlesex, TW11 0LW, UK

a r t i c l e i n f o

Keywords:

Radioactive waste

Gamma spectrometry

Standards

Comparisons

43/$ - see front matter Crown Copyright & 2

016/j.apradiso.2009.01.009

+44 20 8943 6278; fax: +44 20 8943 8700.

ail address: julian.dean@npl.co.uk

a b s t r a c t

Much of the work of the UK nuclear industry is now concerned with decommissioning many of the

existing power stations and other facilities. An important aspect of this work is the accurate

measurement of low levels of radioactivity in waste forms such as building materials in order that these

materials can be assigned to the correct waste streams. This has led to a call for suitable standards and

reference materials, and the specific needs of UK users were identified at an NPL workshop in 2005. One

of the highest priorities was for ‘soft waste’ spiked with g-emitters in a 200 L drum format, with an

activity concentration of just under 0.4 Bq g�1. In response, NPL prepared a single reference drum

meeting this specification. The low density was achieved by loading the drum with plastic bottles, each

partially loaded with ion-exchange resin. The resin in each bottle had been previously spiked with a

mixture of 241Am, 137Cs and 60Co, all traceable to national standards. The drum would be used primarily

as the basis of a comparison exercise, but feedback on its usefulness as a calibration standard would also

be sought.

The drum was measured by 17 radioassay groups at 15 UK sites. The monitors used were mostly

commercial g-spectrometry systems designed to accommodate waste drums. Some groups measured

the drum on more than one monitor and some used more than one efficiency calibration. Many of the

groups used mathematical modelling to derive their efficiencies. The results of the exercise were

discussed at a second NPL workshop (2007), after which the participants were allowed to submit

supplementary or replacement results (with reasons for any changes clearly stated). In total, 88 results

were submitted. A total of 51 results were in agreement with the NPL values; of the remaining results,

24 were explained by the participants concerned (or were revised to provide supplementary values), but

the other 13 results were either clearly discrepant or questionable. The exercise demonstrated

differences between laboratories who had used the same modelling software for their efficiency

calibrations, and indeed facilitated an exchange of models between these laboratories. The importance

of an accurate knowledge of the form and structure of the sample in efficiency modelling was clearly

demonstrated. Uncertainty budgets were of variable quality. Some participants quoted MDA values for

one or more of the radionuclides. A follow-up comparison was requested and this is now being planned.

Crown Copyright & 2009 Published by Elsevier Ltd. All rights reserved.

1. Introduction

The UK civil nuclear industry evolved over several decadesduring the second half of the 20th century. Many of the reactorsand other facilities built during this time have now reached theend of their working lives and are at various stages of thedecommissioning process. An important aspect of this process isthe clearance and sentencing of very large volumes of potentiallycontaminated (or potentially activated) waste materials such asbuilding materials, soil and plastics. This must be underpinned byaccurate radioassay to ensure that materials are assigned to thecorrect waste streams. However, there is a lack of suitable

009 Published by Elsevier Ltd. All

reference materials and calibration services to support measure-ments such as these in the UK.

In 2005, the National Physical Laboratory (NPL) commissioned asurvey of the UK nuclear industry (Dean and Jerome, 2005) toestablish the exact needs of the user base. This revealed a wide rangeof requirements with no particular sample type being dominant, so aworkshop of industry stakeholders was convened (June 2005) todiscuss this further and to prioritise the needs. Chief among thefindings was a need for reference materials for g non-destructiveassay of concrete and laboratory ‘soft waste’. The preferred formatswere large-volume containers (e.g. 200 L drums and 1 m3 plasticbags) and the required activity concentration was around 0.4 Bq g�1,this being the upper limit of the ‘exempt waste’ category in UKlegislation. The preferred radionuclides were 241Am, 60Co and 137Cs.

On this basis, NPL prepared a single ‘prototype’ 200 L drumstandard for soft waste at the required activity level. This was

rights reserved.

ARTICLE IN PRESS

Table 1Principal radionuclides in NPL standard source.

Radionuclide Activity

concentration

(averaged across

contents of

drum) at 1200

GMT 1/12/07

(Bq g�1)

Standard

uncertainty

(k ¼ 1) (Bq g�1)

Standard

uncertainty

(k ¼ 1) (%)

241Am 0.2594 0.0015 0.57137Cs 0.0295 0.0003 0.8760Co 0.03731 0.00023 0.62

J. Dean / Applied Radiation and Isotopes 67 (2009) 678–682 679

circulated as a blind sample to interested laboratories both toprovide them with a traceable comparison sample against which tovalidate their systems and to allow them to comment on theusefulness of the prototype design as the basis for further standardsof this type. Following the comparison, NPL convened a secondworkshop to discuss the results, to identify any measurementproblems and any further work required, and to obtain feedback onthe prototype. In this paper, the preparation of the standard drumand the conduct of the comparison are described, and the results ofthe comparison are presented and discussed.

2. Preparation of standard source

‘Soft waste’ is a poorly defined waste form and can refer to, forexample, plastics, paper, ion-exchange resins and vermiculite. Itvaries greatly in density, but in uncompacted form it is typicallyaround 300 kg m�3. Ion-exchange resins are ideal for a standard ofthis type in terms of stability of retention of the active species, butthe problem is that their densities tend to be around 700 kg m�3.It was therefore decided to use resin, but to reduce the overalldensity of the drum’s contents to 300 kg m�3 by firstly loading theresin into a series of plastic bottles (only partially filling eachbottle) and then packing the bottles into the drum. Dowex‘Marathon C’ Cation Exchange Resin1 would be used. The radio-active ‘spike’ would be a mixed radionuclide standard solutioncontaining the preferred radionuclides given above.

Firstly, the standard solution was prepared by mixing weighedaliquots of stock solutions of the individual radionuclides with aknown mass of carrier solution. The stock solutions either hadbeen directly measured in an NPL secondary standard ionisationchamber or had been prepared by quantitative dilution of moreactive solutions previously measured on an NPL chamber. Thesedilutions were validated by counting samples of diluted andundiluted solution using a NaI(Tl) g-counter. The activityconcentrations of the three radionuclides in the final mixturewere confirmed by high-resolution g-spectrometric assay and aregiven in Table 1. This technique was also used to check for g-emitting impurities but none were detected.

Secondly, bulk resin (approximately 90 kg) was dried, and aweighed portion (nominally 190 g) was loaded into each of 240plastic bottles (each having a volume of 500 ml). Each bottle wasthen spiked with approximately 0.1 g (accurately weighed) of thestandardised solution before being sealed and then tumbled for6 h (to homogenise the activity distribution within the bottle). Thebottles were then loaded into a 205 L drum (with weighing).Finally, the g-emission rate was measured at a few points aroundthe drum as a fairly crude check on the overall homogeneity of theactivity distribution.

1 The Dow Chemical Company, Michigan, USA. Note that use of this particular

resin is not meant to imply that NPL recommends it over other similar products.

3. Conduct of the comparison

Seventeen UK laboratories (for example, at nuclear powerstations and other nuclear sites, and at sites of analytical serviceproviders and instrument manufacturers) participated. The drumwas sent ‘site-to-site’ according to a predetermined deliveryschedule and each participant was given approximately fiveworking days to carry out all their measurements. The partici-pants were told which radionuclides were present and theapproximate overall activity concentration. They were also givendetails of the (empty) drum’s dimensions and mass, and details ofits contents other than the activity (e.g. the resin used, the massand the overall density).

After the initial results had been returned, the assigned valuesof the activity concentrations were disclosed to allow theparticipants to check for any transcription errors and to submitfurther data and/or explanatory information) if they wished. Allsubmitted data (including any initial values later amended by theparticipants) were summarised in a draft report in which eachparticipant was assigned a code number to preserve anonymity.The results were discussed further at a second NPL workshop(2007) and a final version of the report was then published (Dean,2007).

4. Results and data analysis

The results were analysed using the method devised by Harms(Harms et al., 2006).

The deviation ‘D’ of each participant’s result from the assignedvalue was calculated as follows:

D ¼ 100L� N

N¼ 100

L

N� 1

� �(1)

where L is the participant’s result and N is the assigned value (i.e.the NPL value).

The standard uncertainty (k ¼ 1) of the deviation, uD, wascalculated from

uD ¼ 100L

N

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiuL

L

� �2

þuN

N

� �2r

(2)

where uL is the standard uncertainty of the participant’s result anduN is the standard uncertainty of the assigned value.

The results were then evaluated using three parameters: the‘zeta score’ (z), the relative uncertainty of the participant’s result(RL) and the ‘z-score’ (z). These were calculated as follows:

z ¼L� Nffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiu2

L þ u2N

q (3)

RL ¼uL

L(4)

z ¼L� N

RmedN(5)

where Rmed is the median value of RL.The zeta scores and z-scores were used to determine whether

the difference between the participant’s result and the assignedvalue was significantly different from zero. One-sided Dixon’s Qoutlier tests (Rorabacher, 1991) were used to determine whetherthe RL value for a particular result was significantly larger than theother values in the overall set of results.

Results for which the absolute values of the zeta score and thez-score were both p2.58 (corresponding to a significance levels ofa ¼ 0.01) and for which the relative uncertainty RL was notsignificantly larger than the other values in the data set were

ARTICLE IN PRESS

J. Dean / Applied Radiation and Isotopes 67 (2009) 678–682680

regarded as being ‘in agreement’ with the assigned value. In caseswhere (i) the relative uncertainty RL was significantly larger thanthe other values in the data set, or (ii) either the zeta score or thez-score (but not both) was 42.58, the result was classified as‘questionable’. If both the zeta score and the z-score were 42.58,the participant’s result was classified as ‘discrepant’ from theassigned value regardless of the value of the relative uncertaintyRL.

The deviations of all the final results from the assigned valuesare illustrated in ascending order in Figs. 1–3. Data symbols areused as follows:

�

Figlarg

‘qu

Figlarg

‘qu

Filled squares—result in agreement

� Unfilled squares—result questionable � Filled triangle—result discrepant.

Some participants reported results for more than one detector, orreported more than one result for one detector (for example, usingdifferent efficiency models). This is discussed in Section 5. Initial

241Am(results 4B and 7A

-100

0

100

11A11B17B 1D 1B 1C 1A 4A 2 7B 17

E 8A

Partic

Dev

iatio

n (%

)

. 1. Deviations from the assigned value for 241Am. The smaller dotted line represents

er dotted line represents the pass/fail limits of the z-score. Filled squares denote data

estionable’.

137C(results 10A and 10

-100

0

100

5 211A 1D 1H 1C 6 1B 1A 1G 11

B17A17C

Partic

Dev

iatio

n (%

)

. 2. Deviations from the assigned value for 137Cs. The smaller dotted line represents

er dotted line represents the pass/fail limits of the z-score. Filled squares denote da

estionable’ and filled triangles denote data which are ‘discrepant’.

results and uncertainties which were later amended are notincluded in Figs. 1–3 but are discussed in Section 5. Someparticipants reported that the activities of some or all theradionuclides were below the limit of detection (see Table 2)and one laboratory reported a total activity for the drum (Table 3).

5. Discussion

5.1. Methods

Thirteen of the participants used high-resolution g-spectrometry,using any of a number of commercial systems designed for waste-drum assays. In systems such as these, the drum is mounted on aplatform (often a rotating platform) adjacent to one or moredetectors. Some systems have collimated, moveable detectors whichallow segments of the drum to be separately assayed (e.g. segmentedgamma scanners).

Two participants used NaI(Tl) scintillator-based systems andtwo used plastic scintillation detectors.

results outside limits of plot)

8B 1H 1F 617G 1G 17

C 3 1E 17A17D17F 5

ipant code

the percentage relative uncertainty on the assigned value multiplied by 2.58. The

‘in agreement’ with the assigned value and unfilled squares denote data which are

s resultsB outside limits of plot)

1F 1E 8A 8B 17B17G 4A 4B 17

F17D17E 7B 7A 3

ipant code

the percentage relative uncertainty on the assigned value multiplied by 2.58. The

ta ‘in agreement’ with the assigned value, unfilled squares denote data which are

ARTICLE IN PRESS

60Co results(results 13 and 15 outside limits of plot)

0

-100

100

6 511A11B 1D 1F 1H 4B 1G 1A 1B 1C 1E 8A 8C 8B 8D 4A 17

A17C 7B 17

B17F17G17D17E10A10B 7A 3

Participant code

Dev

iatio

n (%

)

Fig. 3. Deviations from the assigned value for 60Co. The smaller dotted line represents the percentage relative uncertainty on the assigned value multiplied by 2.58. The

larger dotted line represents the pass/fail limits of the z-score. Filled squares denote data ‘in agreement’ with the assigned value, unfilled squares denote data which are

‘questionable’ and filled triangles denote data which are ‘discrepant’.

Table 2Reported minimum detectable activities for 241Am, 137Cs and 60Co.

Participant code Reported 241Am

MDA (Bq g�1)

Reported 137Cs

MDA (Bq g�1)

Reported 60Co

MDA (Bq g�1)

2 – – o0.01674

9 o0.004 o0.07 o0.06

12 o0.002997 o0.07118 o0.05529

14 – o0.102 o0.227

Table 3Reported total activity concentrations.

Participant code Reported total activity

concentration (Bq g�1)

Reported uncertainty

(k ¼ 1) (Bq g�1)

16 0.081 0.003

J. Dean / Applied Radiation and Isotopes 67 (2009) 678–682 681

Many participants derived detection efficiencies by modellingusing various types of software. In some cases, these efficiencieshad been ‘validated’ using standard sources of various types.

2 ORTEC.3 CANBERRA.

5.2. 241Am results

The inclusion of this radionuclide was interesting because ofthe low energy of the principal g-emission and the resultant needfor accurate modelling of both the drum itself and its contents toensure that attenuation factors were correctly calculated. Some ofthe results demonstrated this well. For example, the originalresults from Participant 17 (not included in Fig. 1) were found tobe discrepant because the wrong drum wall thickness had beenused in the participant’s efficiency model. The revised results(discussed further below) were based on the correct thickness andwere mostly in agreement with the assigned value.

Participant 7’s first result (7A, outside the limits of Fig. 1) wasdiscrepant and this fact was attributed by them to their use of aMonte Carlo efficiency model which assumed that a range ofmaterial types were present in the drum; also, a drum wallthickness of 1.2 mm had been assumed (compared with a truevalue of 0.8 mm). Their second result (7B) was based on a moreaccurate model and resulted in much closer agreement with theassigned value.

Other participants who tried to model the drum’s contentsinitially misunderstood the internal ‘structure’, a crucial aspect ofwhich was that the plastic bottles were only half-full of resin,resulting in (to a rough approximation) ten alternating layers ofair and resin. These participants each submitted multiple resultsbased on efficiency models assuming different ‘degrees oflayering’, revealing some interesting effects. For example, Partici-pant 1 initially submitted results 1A to 1D (using four differentdetectors and based on misunderstood layering) and later results1E to 1H (based on correct layering). Although all these resultswere in agreement with the assigned value, the second set wascloser to that value. Participant 17, meanwhile, used only onedetector but modelled the efficiency (using Monte Carlo techni-ques) using six different assumptions regarding the compositionof the contents, these being that it consisted of ion-exchange resin(result 17A), polyethylene (17B), PVC (17C), ‘layered’ ion-exchange(17D), layered polyethylene (17E) and layered PVC (17F). Thereason for this approach was that polyethylene and PVC are usedby this participant as ‘extremes cases’ for soft waste (when thematerial is ill-defined) and are therefore considered as providingreasonable bounds on the detection efficiency for soft waste. The‘layered’ models were used after it became clear from transmis-sion source measurements that layers were present. The sixmodels all yielded different results (apart from 17D and 17F,which were identical); as noted above, all but one of the resultsagreed with the assigned value. A seventh result (17G) wassubmitted, based on a mean of various soft-waste materialcalibrations using ISOTOPIC2 software.

Participant 4 expected their result 4B to be discrepant becausethey were still evaluating the instrument used and knew therewere problems with the software, but another factor was that theinstrument used a transmission source for matrix corrections andthis was affected by the layering of the resin.

Several of the participants (1, 8 and 11) had used ISOCS3

software to calculate detection efficiencies but did not achieve thesame level of agreement with the assigned value. This seemed tobe partly because they had modelled their efficiencies on the basisof different assumptions regarding the internal structure of thedrum and partly because of differences in uncertainty estimation.

ARTICLE IN PRESS

J. Dean / Applied Radiation and Isotopes 67 (2009) 678–682682

This led to information exchange between these laboratories andsome improvement in performance (for example, Participant 11revised their result from 11A to 11B by using Participant 8’smodel).

Regarding uncertainty estimations, it is clear from Fig. 1 that arange of values had been quoted. This is an area where currentmeasurement guidance (e.g. Measurement Good Practice Guide34) requires revision to better assist users. Note that Participant 8revised their initial result 8A to 8B (same activity concentrationbut different uncertainty) because the first submission hadassumed the presence of high-density materials and it had notbeen possible, at the time of measurement, to ‘remove’ the high-density component from the model. Note also that Participant 1had initially quoted only the random components of theuncertainties on the results 1A–1D, resulting in them beingdiscrepant (these are not shown in Fig. 1).

Of the 27 (final) results submitted for 241Am, 21 were inagreement with the assigned value.

5.3. 137Cs results

Of the 29 final submitted results for this radionuclide, 17agreed with the assigned value. Some of the comments in Section5.2 above again apply (e.g. Participants 1, 7 and 11 revised theirmodels, yielding more accurate results); however, the results forParticipants 2 and 5 were low and discrepant for this radionuclideand this may have been due (at least in part) to the fact that theycarried out segmented scans. The ‘layering’ of the drum’s contentsmay have resulted in some segments registering very littleactivity, possibly below the limit of detection. This effect mayhave been compounded by the low activity concentration of the137Cs. Participant 5 pointed out that in everyday use their monitorwas set up for activities typically 410 Bq g�1 and that their aim inthis exercise had been to merely confirm the presence of theradionuclides in the drum. The low 137Cs activity present resultedin three other participants reporting this radionuclide as belowtheir limit of detection.

Participant 10’s results were outside the limits of the plotbecause they had reported gross g measurements (using a plasticscintillator detector).

5.4. 60Co results

For this radionuclide, 13 of the 32 results agreed with theassigned value. The data follow a similar pattern to 137Cs.Participant 6’s result was low and discrepant and they attributedthis to the low activity of this radionuclide and the low detectionefficiency at higher g energies; the software had been set toidentify this radionuclide in a segment only if both the main gemissions had been detected, and sometimes the confirming peakwas not seen. The results for Participants 13 and 15 were outsidethe limits of the plot (in the case of Participant 15 by some twoorders of magnitude). The reasons for this are not clear. Again, thelow activity resulted in several participants reporting valuesbelow the limit of detection.

6. Conclusions

In the final data set (i.e. excluding initial values corrected later,MDA submissions and the one result for the total activity in thedrum), 51 of the 88 results were found to be in agreement withthe assigned value. Of the remaining 37 results, 24 were eitherexplained by the participants concerned or had been revised bythe participants to provide supplementary values. However, 13results were either discrepant or questionable for no obviousreason. Most of these problems arose with 137Cs and 60Co.Contributory factors seemed to be the low activities present ofthese radionuclides and (in the case of measurements involvingSegmented Gamma Scanners) the presence of what were, ineffect, horizontal layers of sample within the drum.

A key conclusion of the exercise was that an accurateknowledge of a sample’s dimensions and composition is essentialto successful modelling of g detection efficiencies.

A wide range of uncertainties was quoted (even within sub-groups using the same technique and software), suggesting a needin some cases for guidance in the compilation of budgets formeasurements of this type. The exercise also illustrated thebenefits of information exchange between laboratories.

Those participants who had attended the second NPL work-shop were asked if the problems highlighted by this exercise wereindicative of day-to-day measurement problems with ‘real’samples, but none felt that they were; however, they consideredthe exercise very worthwhile and requested a second comparisonbased on a more challenging sample (e.g. with a heterogeneousactivity distribution). This was seen as the best use of theprototype standard design. The second comparison will be run in2009.

Acknowledgements

The author wishes to thank the participating organisations forthe time and effort they have put into analysing the drum and forthe information provided. He also thanks his colleagues Mr PeteBurgess, Dr Arvic Harms, Mr Andy Stroak and Miss Jean Wong fortheir help with preparing and transporting the drum and writingthe final report. Finally, the author gratefully acknowledges thefinancial support of the National Measurement System Pro-grammes Unit of the UK Department for Innovation, Universitiesand Skills (DIUS).

References

Dean, J.C.J., Jerome, S.M., 2005. A survey of user requirements for a UK radioactivitymeasurement infrastructure for nuclear decommissioning. NPL Report DQL RN009.

Dean, J.C.J., 2007. A comparison of procedures used at UK nuclear sites for gammaassays of potentially contaminated or activated materials. NPL Report IR2.

Harms, A.V., Gilligan, C.R.D., Arinc, A., Collins, S.M., Jerome, S.M., Johansson,L.C., MacMahon, T.D., Pearce, A.K., 2006. Environmental radioactivity compar-ison exercise 2005. NPL Report DQL RN 015.

Rorabacher, D.B., 1991. Statistical treatment for rejection of deviant values: criticalvalues of Dixon’s ‘‘Q’’ parameter and related subrange ratios at the 95%confidence level. Anal. Chem. 63, 139–146.