ANALYSIS OF ENVIRONMENTAL RADIONUCLIDES
RADIOACTIVITY IN THE ENVIRONMENT A companion series to the Journal of Environmental Radioactivity Series Editor M.S. Baxter Ampfield House Clachan Seil Argyll, Scotland, UK Volume 1: Plutonium in the Environment (A. Kudo, Editor) Volume 2: Interactions of Microorganisms with Radionuclides (F.R. Livens and M. Keith-Roach, Editors) Volume 3: Radioactive Fallout after Nuclear Explosions and Accidents (Yu.A. Izrael, Author) Volume 4: Modelling Radioactivity in the Environment (E.M. Scott, Editor) Volume 5: Sedimentary Processes: Quantification Using Radionuclides (J. Carroll and I. Lerche, Authors) Volume 6: Marine Radioactivity (H.D. Livingston, Editor) Volume 7: The Natural Radiation Environment VII (J.P. McLaughlin, S.E. Simopoulos and F. Steinhäusler, Editors) Volume 8: Radionuclides in the Environment (P.P. Povinec and J.A. Sanchez-Cabeza, Editors) Volume 9: Deep Geological Disposal of Radioactive Waste (R. Alexander and L. McKinley, Editors) Volume 10: Radioactivity in the Terrestrial Environment (G. Shaw, Editor) Volume 11: Analysis of Environmental Radionuclides (P.P. Povinec, Editor)
ANALYSIS OF ENVIRONMENTAL RADIONUCLIDES
Editor
Pavel P. Povinec Faculty of Mathematics, Physics and Informatics Comenius University Bratislava, Slovakia
AMSTERDAM – BOSTON – HEIDELBERG – LONDON – NEW YORK – PARIS SAN DIEGO – SAN FRANCISCO – SINGAPORE – SYDNEY – TOKYO
Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK First edition 2008 Copyright © 2008 Elsevier B.V. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email:
[email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress ISBN: 978-0-08-044988-3 ISSN: 1569-4860
For information on all Elsevier publications visit our website at books.elsevier.com
Printed and bound in Hungary 08 09 10 11 12
10 9 8 7 6 5 4 3 2 1
v
Contents
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.
1
Statistical sampling design for radionuclides by E. Marian Scott and Philip M. Dixon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
2.
Sampling techniques by Fedor Macášek . . . . . . . . . . . . . . . . . . . . .
17
3.
Detection and quantification capabilities in nuclear analytical measurements by L.A. Currie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
Radiometric determination of anthropogenic radionuclides in seawater by M. Aoyama and K. Hirose . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
137
Monte Carlo simulation of background characteristics of gamma-ray spectrometers—a comparison with experiment by Pavel P. Povinec, Pavol Vojtyla and Jean-François Comanducci . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
163
Underground laboratories for low-level radioactivity measurements by Siegfried Niese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
209
Accelerator mass spectrometry of long-lived light radionuclides by A.J. Timothy Jull, George S. Burr, J. Warren Beck, Gregory W.L. Hodgins, Dana L. Biddulph, Lanny R. McHargue and Todd E. Lange . . . . . . . . . . . . . . .
241
Accelerator mass spectrometry of long-lived heavy radionuclides by L.K. Fifield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
263
Analysis of radionuclides using ICP-MS by Per Roos . . . . . . . . . . . . . .
295
4.
5.
6.
7.
8.
9.
10. Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides by N. Erdmann, G. Passler, N. Trautmann and K. Wendt . . . . .
331
vi
Contents
11. Environmental radioactive particles: A new challenge for modern analytical instrumental techniques in support of radioecology by Maria Betti, Mats Eriksson, Jussi Jernström and Gabriele Tamborini . . . . . . . . . . . . . . . . . . .
355
12. Activation analysis for the determination of long-lived radionuclides by Xiaolin Hou . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
371
13. In situ and airborne gamma-ray spectrometry by Andrew N. Tyler . . . . . .
407
14. Underwater gamma-ray spectrometry by Pavel P. Povinec, Iolanda Osvath and Jean-François Comanducci . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
449
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
481
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
525
1
Foreword The Radioactivity in the Environment series is an ambitious project covering recent progress in this rapidly developing field, which has included aspects such as the behavior of radionuclides in the environment, the use of natural and anthropogenic radionuclides as tracers of environmental processes, marine radioactivity studies, radiation protection, radioecology, etc. to mention at least a few. State of the art radioanalytical environmental technologies have always been a limiting factor for environmental radioactivity studies, either because the available sensitivity was not high enough to get meaningful results or the required sample size was too big to carry out such investigations, very often with limiting financial resources. There has in recent years been great progress in the development of analytical tools related to sampling strategies, development of rapid and efficient radiochemical separation methods, radiometric counting systems utilizing high sensitivity Ge detectors often working underground, and mass spectrometry technologies based on ICPMS (inductively coupled plasma mass spectrometry) and AMS (accelerator mass spectrometry) for sensitive analysis of natural and anthropogenic radionuclides in the environment. For example, in the marine environment, where research work has been heavily dependent on the new technologies, we have seen a replacement of time-consuming and expensive large volume water sampling (500 L) from several km water depths by Rosette multisampling systems enabling high resolution water sampling within one or two casts with 12 L bottles only. The sampling strategies are often developed and controlled using satellite information for the optimization of the sampling programs. Further, the philosophy of sampling and laboratory measurements has changed, where appropriate, to in situ analysis of radionuclides in the air, on land, in water and in the sediment, thus developing isoline maps of radionuclide distributions in the investigated environment. In the field of analytical technologies we have moved from simple radiochemical methods and gas counters to robotic radiochemical technologies and sophisticated detectors working on line with powerful computers, often situated underground or having anticosmic and/or anti-Compton shielding to protect them against the cosmic radiation, and thus considerably decreasing their background and increasing their sensitivity for analysis of radionuclides in the environment at very low levels. The philosophy of analysis of long-lived radionuclides has also changed considerably from the old concept of counting of decays (and thus waiting for them) to the direct counting of atoms (as if they were stable elements) using highly sensitive mass spectrometry techniques such as AMS, ICPMS, TIMS (thermal ionization mass spectrometry), RIMS (resonance ionization mass spectrometry) and SIMS (secondary ionization mass spectrometry). There have also been considerable changes in the philosophy and organization of research as institutional and national investigations have been replaced by global international projects
2
P.P. Povinec
such as WOCE (world ocean circulation experiment), CLIVAR (climate variability and predictability study), PAGES (past global changes), WOMARS (worldwide marine radioactivity studies), GEOTRACES (global marine biochemistry of trace elements and isotopes), SHOTS (southern hemisphere ocean tracer studies), to mention at least a few. Although the topic of the analysis of environmental radionuclides has already been covered in several reviews, there has not been available a book covering critical progress in recent years. The present collection of review papers covers a wide range of topics starting with the development of statistically based sampling strategies to study radionuclides in the environment (Chapter 1 by Scott and Dixon), followed by description of sampling techniques and pre-concentration of samples (Chapter 2 by Macášek). Statistical evaluation of data has been a crucial point in correct interpretation of measurements, especially when dealing with counting rates very close to the detector background (Chapter 3 by Currie). Recent progress in environmental studies is documented by the analysis of 137 Cs, 90 Sr and Pu isotopes in the seawater column (Chapter 4 by Aoyama and Hirose). Monte Carlo simulations of detector background characteristics have been an important pre-requisite when designing low-level counting systems (Chapter 5 by Povinec et al.), also important when working in laboratories situated hundreds of meters underground, where radioactive purity of construction materials and radon concentration in the air become dominant factors controlling the detector background (Chapter 6 by Niese). AMS has been a revolutionary breakthrough in analytical methodologies for long-lived environmental radionuclides, as described by Jull et al. in Chapter 7 for light elements, and Fifield in Chapter 8 for heavy elements. However, the most widely used mass spectrometry technique for analysis of long-lived environmental radionuclides has been ICPMS, as documented by Ross in Chapter 9. Another new trend in analytical techniques has been an introduction of resonance ionization mass spectrometry for radionuclide analysis (Chapter 10 by Erdemann et al.), and a change from bulk sample analysis to particle sensitive analysis, as described by Betti et al. in Chapter 11 using SIMS, scanning electron microscopy (SEM), and synchrotron based techniques like µ-XRF and 3D-µ tomography. Neutron activation analysis (NAA) has been contributing in specific applications with long-lived radionuclides, and usually this is the only alternative technique for certification of reference materials (Chapter 12 by Hou). In situ techniques represent a new approach to analysis of environmental radionuclides and these have been recently widely applied for surface monitoring of radionuclides using either mobile gamma-ray spectrometers, helicopters and airplanes (Chapter 13 by Tyler) or measurements carried out under the water, e.g., for radionuclide mapping of seabed sediments and/or stationary monitoring of radionuclides in the aquatic environment as described in Chapter 14 by Povinec et al. The Editor would like to thank all authors for their fruitful collaboration during preparation of this compilation and Prof. Baxter, the Radioactivity in the Environment Series Editor, for his patience when working on this book. In publishing this book we hope to further stimulate work in the exciting field of environmental radioactivity and the use of radionuclides as tools for investigations of environmental processes. Pavel P. Povinec Editor Comenius University Bratislava, Slovakia
3
Statistical sampling design for radionuclides E. Marian Scotta,∗ , Philip M. Dixonb a Department of Statistics, University of Glasgow, Glasgow G12 8QW, UK b Department of Statistics, Iowa State University, Ames, IA 50011-1210, USA
1. Introduction This chapter presents some of the key ideas for designing statistically based sampling strategies to study radionuclides in the environment. Environmental samples will naturally vary in their specific activity, no matter how precise the radionuclide measurement. This variability is caused by natural variations in the processes that control radionuclide transport and uptake in the environment. A statistically based sampling design quantifies this variability and allows information from specific samples to be generalized to a larger population. The statistical sampling principles discussed here are detailed in many textbooks and papers about environmental sampling, such as the general sampling textbooks by Cochran (1977) and Thompson (2000), the environmental statistics textbook by Gilbert (1987), and government agency guidance documents (US EPA, 2002). A recent ICRU report (2006) presents a thorough presentation of sampling issues for environmental radionuclides. This chapter draws heavily on these reference works. It provides only a taster to the issues; the reader is strongly encouraged to read more in-depth descriptions. Environmental sampling should not be considered as a ‘recipe’ based activity. The best (most efficient, valid and reliable) sampling schemes use environmental knowledge to guide the sampling. Changes in objectives (apparently small) may also lead to quite significant changes to the sampling scheme. 1.1. General sampling concepts and principles Statistical sampling is a process that allows inferences about properties of a large collection of things (the population) to be made from observations (the sample) made on a relatively small number of individuals (sampling units) belonging to the population. The population is the set of all items that could be sampled, such as all deer in a forest, all people living in the UK, etc. A sampling unit is a unique member of the population that can be selected as an individual ∗ Corresponding author. E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11001-9
© 2008 Elsevier B.V. All rights reserved.
4
E.M. Scott and P.M. Dixon
sample for collection and measurement. The sample is then the set of sampling units that are measured. Sampling units might be individual deer, individual people, trees, garden plots, or soil cores of a given dimension. An essential concept is that a statistically based sample of a sufficient number of individual sampling units is necessary to make inferences about the population. Statistical sampling also allows a quantification of the precision with which inferences or conclusions can be drawn about the population. The focus of this chapter is on statistical sampling design, namely how to select specific sampling units from a population or sampling locations within a larger area, and how to determine the number of individual units to collect. Sampling has many purposes, including estimation of the distribution (and mean) concentration of a radionuclide (Bq l−1 ) in a river, or in fruit in a region (Bq kg−1 ), or a map of radionuclide deposition (Bq m−2 ). Different purposes require different sampling strategies and different sampling efforts in order to be effective and efficient, so it is important that the purpose(s) of the sampling program be clearly specified. The environmental context also plays an important part in determining the choice of sampling method. Statistical sampling requires information about the nature of the population and characteristics to be described. 1.2. Methods of sampling A statistical sampling design is based on probability sampling, in which every sampling unit has a known and non-zero probability of being selected. The actual sample (set of sampling units to be measured) is chosen by randomization, using published tables of random numbers or computer algorithms. Selecting a probability sample is easy when the population can be enumerated. As a simple example, imagine sampling 10 adults from a specified geographic area for whole body monitoring. We could use an electoral register or census information to enumerate all individuals. Suppose that the population comprised 972 such individuals, then we could generate 10 random numbers lying between 1 and 972, such as 253, 871, 15, 911, 520, 555, 106, 83, 614, 932 to identify the 10 individuals. If the same number was generated more than once, then we would simply continue the process till we had 10 unique random numbers and these would then identify the individuals to be called for monitoring. The actual numbers (253, 871, etc.) are read from random number tables or may be generated by statistical software. There are many sampling designs. We describe simple random sampling, stratified random sampling and systematic sampling because these are the three most common in environmental studies. For each design, we discuss how to select a sample and how to estimate the population mean and its sampling error. A brief review of the advantages and disadvantages of the different methods is also included. More detail can be found in ICRU (2006). 1.2.1. Simple random sampling In a simple random sample, every sampling unit in the population has an equal probability of being included in the sample and all pairs of sampling units have the same probability of being included in the sample. One way to select a simple random sample is to enumerate all sampling units in the population, then use random numbers to select the desired number of sampling units. Simple random sampling is easy to describe but may be difficult to achieve
Statistical sampling design for radionuclides
5
in practice. Some common problems include lack of response from some individuals, inaccessibility of some plots of ground, and long travel times between sampling locations when sampling large areas. Example: Estimation of the average baseline 14 C level in the food-chain An estimate of the dose to the general public due to 14 C in the food-chain is an important radiological quantity for regulatory impact assessment since many nuclear power stations discharge 14 CO2 which is rapidly taken up. At a given station, the task would be to select representative environmental samples that enter the food-chain, e.g., root or cereal crops or fruits. For this particular problem, definition of the population should include identification of the species and information on where and when it grew and its spatial context. For the choice of species, it should be widely available and a suitable material for 14 C assay; a material such as soft fruit, mushrooms or grain would be ideal. The analysis requirements would then define how much material needed to be collected for each sampling unit. In terms of the temporal extent, it would be logical for the samples to be selected from a single growing season and in a specific year such as 2004. This results in a clear definition of the population, namely all selected crop in the vicinity of the site growing in a specific year and of a sampling unit, namely a bulked sample of berries, vegetables or wheat harvested from a specific location. The next step requires identification of the locations at which the samples will be collected and determination of how many sampling units will be required to satisfy the objectives of the study. A map of the vicinity in terms of all locations where the crops grow, would allow the numbering of all the locations from 1 to N , and random numbers would then be used to identify which actual locations would be sampled. The reader might care to consider whether there are more efficient but equally valid sampling approaches to this problem. Analysis of the results from a simple random sample Suppose that the 14 C activity density is measured in each of the n sampling units in the sample. ¯ given by Equation (1) The value from sampling unit i is denoted as yi . The sample average, y, is a good (unbiased) estimate of the population mean 14 C activity density and the sample variance, s 2 , given by Equation (2) is a good estimate of the population variance: yi , y¯ = (1) n and ¯ 2 (yi − y) . s2 = (2) n−1 The sample average is a random quantity; it will be a different number if different sampling units were chosen for the sample, because of the natural variation in 14 C activity densities among sampling units. The uncertainty in the sample average is quantified by the sampling variance, given by Equation (3) or its square root, the estimated standard error, e.s.e. The sampling fraction, f , is the fraction of the population included in the sample, n/N, which is usually very small. 2 1−f . Var(y) ¯ =s (3) n
6
E.M. Scott and P.M. Dixon
In the 14 C example, there is no environmental evidence to believe that the population of crop is non-homogeneous. This means that we have no reason to expect any sub-groups with distinctly different 14 C levels and so simple random sampling is a reasonable sampling strategy. However, there may be spatial information which could prove important, such as distance and direction from the stack and which might lead the scientist to believe that the population is heterogeneous and so a more directed sampling scheme might use contextual information such as wind rose data to determine a different sampling scheme. Next, we consider an example with a similar objective, but where the environmental context would suggest that the population could be non-homogeneous, hence a different sampling scheme could be better (Dixon et al., 2005). 1.2.2. Stratified sampling Stratified sampling designs provide two important advantages over simple random sampling designs, namely, efficiency and improved estimates for meaningful subdivisions of the population. We must assume that the population can be divided into strata, each of which is more homogeneous than the entire population. In other words, the individual strata have characteristics that allow them to be distinguished from the other strata, and such characteristics are known to affect the measured attribute of interest, namely the radioactivity. Usually, the proportion of sample observations taken in each stratum is similar to the stratum proportion of the population, but this is not a requirement. Stratified sampling is more complex and requires more prior knowledge than simple random sampling, and estimates of the population quantities can be biased if the stratum proportions are incorrectly specified. Example: 60 Co activity in an estuary Mapping radioactive contamination in specific locations or areas is a common objective in radioecological investigations (ICRU, 2001). Suppose one wished to map 60 Co in the sediments of an estuary. The population could be conceived to be all possible sediment cores (depth 30 cm, diameter 10 cm) (N in total) within the estuary; a sampling unit would be a single core. A simple random sample, i.e. a random selection of n of the N possible core locations, is a valid sampling design, but it may not be the best. Stratified sampling, using additional information about the estuary and the environmental behavior of 60 Co, can provide a more precise estimate of the mean activity density in the estuary. The distribution of 60 Co in the estuary may not be homogeneous because sediment type and particle size distribution are associated with 60 Co activity. If a map of the sediment type within the estuary is available, it will indicate areas of mud, sand, etc., each of which would be expected to have a different 60 Co activity density. A stratified random sample estimates the mean 60 Co activity in each sediment type. Estimates from each stratum and the area of each sediment type in the estuary are combined to estimate the overall mean activity for the estuary and the sampling error. In stratified sampling, the population is divided into two or more strata that individually are more homogeneous than the entire population, and a sampling method is used to estimate the properties of each stratum. Usually, the proportion of sample observations in each stratum is similar to the stratum proportion in the population. In the example above, we might consider the estuary as being composed of areas of mud, sand and rock. These would then define the
Statistical sampling design for radionuclides
7
strata. This is an example where environmental knowledge and the problem context lead to a better sampling scheme. In stratified sampling, the population of N units is first divided into sub-populations of N1 , N2 , . . . , NL units representing the sampling units in each of the different strata. These sub-populations are non-overlapping and together comprise the whole population. They need not have the same number of units, but, to obtain the full benefit of stratification, the subpopulation sizes or areas must be known. In stratified sampling, a sample of units is drawn from each of the strata. Often, simple random sampling is used in each stratum. The number of sampling units allocated to a stratum is often proportional to the population size or area of that stratum. When each stratum has the same within-stratum variance, proportional allocation leads to the most precise estimate of the population mean (Cochran, 1977). For the sediment example, the strata might be defined as distinct sediment types. Knowledge of the fractional areas of each sediment type within the estuary would be needed to ensure appropriate sampling fractions within each stratum. Simple random samples of size n1 , n2 , . . . , nl would be taken from each strata. Thus if the estuary was 60% sand, 30% silt and 10% mud, then 60% of the sampling units would be selected in the sandy areas, 30% in the silty areas and 10% in the muddy areas. To estimate the average and variance of each stratum, one would use Equations (1) and (2). The population mean activity, Ac , and its sampling error, Var(Ac ), are weighted averages of the average, y¯1 , and variance, sl2 , for each stratum, l. The weights, Wl , are the fractions of each stratum in the population, i.e. Wl = Nl /N . (Nl y¯l ) Ac = l (4) , N and Var(Ac ) =
l
Wl2
sl2 (1 − fl ) . nl
(5)
The equation for the sampling error (5) assumes that Wl , the stratum weight, is known. This would be the case when the strata are areas on a map. When proportional allocation for the sampling fraction is used (i.e., nl /n = Nl /N ), then in Equation (5), Nl is replaced by nl and N is replaced by n. It is not necessary for the sediment map to be accurate, but inaccuracy in the definition of the strata increases the sampling error and decreases the benefit of stratification. Stratified random sampling will, with appropriate use, provide more precise (i.e. less uncertain) estimates than simple random sampling, but more information is required before the specific strategy can be carried out. Simple random and stratified random sampling may be impractical, say in the sediment example, if finding precise sampling locations is difficult. A more practical method of sampling might involve covering the area in a systematic manner, say in parallel-line transects and this final sampling method, systematic sampling, which is often easier to execute than simple or stratified random sampling, and which in some cases is more representative than a random sample is described below. One disadvantage of systematic sampling is that the analysis of the results is often more complex. Again this example also illustrates how the theoretical description of the sampling scheme needs to be modified by the practical reality of sampling in the environment.
8
E.M. Scott and P.M. Dixon
1.2.3. Systematic sampling Systematic sampling is probably the most commonly used method for field sampling. It is generally unbiased as long as the starting point is randomly selected and the systematic rules are followed with care. Transects and two dimensional grids are specific types of systematic samples. Consider sampling sediment in an estuary. One possible systematic sample of 15 locations might be obtained by randomly choosing a transect across the estuary and taking 5 core samples along this transect, and then placing a further two transects equally spaced along the estuary. Systematic sampling is often more practical than random sampling because the procedures are relatively easy to implement in practice, but this approach may miss important features if the quantity being sampled varies with regular periodicity. Systematic sampling differs from the methods of random sampling in terms of practical implementation and in terms of coverage. Again, assume there are N (= nk) units in the population. Then to sample n units, a unit is selected for sampling at random. Then, subsequent samples are taken at every k units. Systematic sampling has a number of advantages over simple random sampling, not least of which is convenience of collection. A systematic sample is thus spread evenly over the population. In a spatial context such as the sediment sampling problem, this would involve laying out a regular grid of points, which are fixed distances apart in both directions within a plane surface. Data from systematic designs are more difficult to analyze, especially in the most common case of a single systematic sample (Gilbert, 1987; Thompson, 2000). Consider first the simpler case of multiple systematic samples. For example, 60 Co activity in estuary sediment could be sampled using transects across the estuary from one shoreline to the other. Samples are collected every 5 m along the transect. The locations of the transects are randomly chosen. Each transect is a single systematic sample. Each sample is identified by the transect number and the location along the transect. Suppose there are i = 1, . . . , t systematic samples (i.e. samtransects in the estuary example) and the yij is the j th observation on the ith systematic i yij /ni . ple for j = 1, . . . , ni . The average of the samples from the ith transect is y¯i = nj =1 The population mean is estimated by t ni t i=1 j =1 yij i=1 ni y¯i = . y¯sy = t (6) t i=1 ni i=1 ni The estimator of the population mean from a systematic sample is exactly the same as the estimator for a simple random sample but it is more difficult to estimate the variance. When there are multiple systematic samples, each with n observations, the variance of the mean can be estimated by Var(y¯sy ) =
t 1 − t/T (y¯i − y¯sy )2 , t (t − 1)
(7)
i=1
where T is the number of transects in the population (Gilbert, 1987; Thompson, 2000). The term in the numerator, 1 − t/T , is a finite population correction factor that can be ignored if the number of transects in the systematic sample, t, is small relative to the number in the population. The variance estimator given by Equation (7) cannot be used in the common case of a single systematic sample, i.e. when t = 1. Many different estimators have been proposed
Statistical sampling design for radionuclides
9
(summarized in Cochran, 1977; Gilbert, 1987). If the population can be assumed to be in random order, then the variance can be estimated as if the systematic sample were a simple random sample, i.e. using Equation (3). That equation is not appropriate when the population has any form of non-random structure. More details of these and other problems are given in Cochran (1977), Gilbert (1987) and Thompson (2000), and in ICRU (2006). Other sampling schemes exist, but they are often intended for rather specialized situations. These include cluster sampling, double sampling and adaptive sampling. These are beyond the scope of this chapter but details can be found in ICRU (2006). Although estimation of the average is probably the most common objective for a sampling campaign, there are other quantities that are of interest in the population, and the basic sampling designs are equally applicable. Perhaps one of the most common sampling purposes is to map the spatial extent of a pollutant, or to estimate the spatial pattern. This is described in more detail in the next section since most radionuclide problems have a spatial context and there is growing use of geographic information systems (GIS) within the radioecology community.
2. Sampling to estimate spatial pattern 2.1. Introduction In many sampling problems, especially in the environmental context, we must consider the spatial nature of the samples collected. It is only common sense that samples, e.g., plants, animals, or soil that are close together are more similar to each other than other samples that are farther apart. Euclidean distance between them can be measured in one dimension if samples are taken along a single transect, or in two dimensions if samples are taken over an area. Spatial sampling methods use the possible relationship between nearby sampling units as additional information to better estimate the quantities of interest. One other important consideration in spatial sampling concerns the ‘spatial size’ of the sampling unit, e.g., a soil sample is of small ‘spatial size’ in the context of mapping a valley. In remote sensing applications, the spatial size’ of the sampling unit may be several hundred square meters and small scale features and variation within a sampling unit would not be observable. In environmental radioactivity, as in general spatial problems, there are two quite different general cases: C ASE 1. We assume that in principle it is possible to measure the radionuclide at any location defined by coordinates (x, y) over the domain or area of interest. This case would generally be appropriate for radionuclides in soil, water and air, such as mapping the Chernobyl fallout, where x and y would typically be latitude and longitude or some other positioning metric. C ASE 2. We assume that in principle it is not possible to measure the radionuclide at all locations defined by coordinates (x, y) over the domain or area of interest, but that it can be measured only at specific locations. For example, consider 137 Cs concentrations in trees. It can only be measured at locations of trees.
10
E.M. Scott and P.M. Dixon
In this chapter, we focus on Case 1, and refer the reader to ICRU (2006) and the more specialized textbooks such as Webster and Oliver (2001) or Cressie (2000) for further discussion of Case 2. Methods of statistical sampling and analysis that take a spatial perspective, such as kriging (Cressie, 2000), form part of the broad topic known as geostatistics, which has a long history (from mining engineering) and is becoming increasingly popular (Wackernagel, 2003; Webster and Oliver, 2001) in radioecology. 2.1.1. Spatial scale Consider the Europe-wide mapping of Chernobyl fallout, where maps for different countries were produced based on very different sampling techniques with very different spatial extent. In some countries, there was a detailed and intensive sampling scheme based on some of the schemes described in Section 1, thus quite detailed inferences could be drawn. In other countries, helicopters and fixed wing aircraft were used to map the fallout in a more continuous and systematic manner. This case is closest to that described in Case 1 above. These two scenarios also emphasis the importance of the spatial scale. If we look at this example in more detail, taking as an example the use of soil samples, consider a survey designed to explore the levels of spatial variation in Chernobyl fallout radionuclides in a small area. A sampling unit is defined to be a 38-mm diameter soil core. Consider nine such cores collected at random within a 1-m2 area. This experiment provides information about small-scale spatial variation within that 1-m2 area. The results could also be combined to provide an areal average and to complete the map of presumably quite a small area, there would need to be many 1-m2 areas surveyed in such a way. A second approach might use a sampling unit of larger dimension (e.g., a 400 m2 area) to sample from a much bigger area, e.g., of 10,000 m2 . This second approach could provide detail about variation over a larger spatial scale and would more easily lend itself to providing a country map. A third study might be to map the entire continent of Europe where the sampling units are ‘counties’ or ‘regions’ within countries. If the environment is heterogeneous at the moderate spatial scale, perhaps because 137 Cs in the soil cores is influenced by soil type and vegetation cover that vary considerably within the sampling area, then these three studies provide quite different descriptions of the spatial pattern and variability. The small scale study might find no spatial pattern, i.e. no correlation between nearby samples, within the 1-m2 area. The intermediate scale study might find a strong spatial pattern, because nearby samples are from the same soil type and vegetation cover. The large scale study will identify a different spatial pattern that depends on the large scale geographic variation between countries. If there were only a few samples per country, the large scale study could not detect the moderate scale pattern due to vegetation and soil characteristics within a country. The use of aerial radiometrics to map large scale radioactivity and its relationship to smaller spatial scale techniques is dealt with in Sanderson and Scott (2001). The extent of the area to be sampled defines the population for which inferences will be made. This will be defined by the context and purpose of the study. The physical area of the study could range from locating hot spots in a former radium factory site to mapping Chernobyl fallout throughout Europe. Practicality and feasibility also affect the optimal sampling design. Although sampling grids and sampling effort may be prescribed on statistical grounds according to some criterion of optimality, the sampling design must also be practical.
Statistical sampling design for radionuclides
11
2.1.2. Sampling objectives Specific sampling objectives that arise in the spatial context are usually similar to those described earlier but with added consideration of area, for example estimation of the average activity over a specified area, or estimation of the inventory within an area (or volume). Other objectives could include mapping the radionuclide distribution over an area, estimating the percentage of an area that exceeds a given level and mapping these locations, estimating the scale of spatial variation, and detecting hot spots. These are only possible with spatially referenced data. Finally, on the basis of observations made one might also wish to predict activity at unsampled (target) locations using a spatial interpolation scheme based on the observed measurements. There is a great deal of specialized terminology and notation used in spatial sampling, and these are best studied in some specialized texts, including Cressie (2000), Webster and Oliver (2001), Wackernagel (2003), and Burrough and McDonnell (1998). 2.2. Classical spatial sampling methods The classical sampling designs and analyses described in Section 1 can be used in a spatial context. The sampling procedure remains probabilistic (random) with the assumption that we have identified the full set of ‘individuals’ or locations within the target population, and sampling involves a selection of individual sites to be examined. In random sampling, a random sample of locations at which the attribute is to be measured is chosen from the target population of locations. If there is knowledge of different strata over the sampling domain (such as soil type), the use of a stratified sample would be recommended and a random sample of locations would be selected within each strata. The number of samples to be collected in each stratum would be defined based on the relative areas of each stratum. The data set is then given by the spatial coordinates of each measurement location and the measured value of the attribute at that location. However, systematic sampling is more commonly used in the spatial setting due to its practicality. Usually, for systematic sampling the region is considered as being overlaid by a grid (rectangular or otherwise), and sampling locations are at gridline intersections at fixed distance apart in each of the two directions. The starting location is expected to be randomly selected. Both the extent of the grid and the spacing between locations are important. The sampling grid should span the area of interest (the population), so that any part of the population could become a sampling location. A systematic grid may also include additional points at short distances from some of the grid points. These points can provide additional information about small-scale spatial correlations. Some commonly used sampling schemes are based on quadrats and transects. A quadrat is a well-defined area within which one or more samples are taken; it is usually square or rectangular in shape, with fixed dimensions. The position and orientation of the quadrat will be chosen as part of the sampling scheme. A line transect is a straight line along which samples are taken, the starting point and orientation of which will be chosen as part of the sampling scheme. When it is necessary to sample a large area with closely spaced sampling locations, a systematic grid can require a very large number of sampling locations. Transect sampling can provide almost as much information with many fewer sampling locations. A transect sample is a random or, more commonly, a systematic sample along a line. Two types of transect
12
E.M. Scott and P.M. Dixon
samples are common. In one, transects are short relative to the size of the study area. These transects are randomly located within the study area. Both the starting point and directional orientation of the transect should be randomly chosen. Ideally, data are collected from more than one transect; each transect has a new starting point and direction. In the second form of transect sampling, transects extend completely across one dimension of the study area, and all transects have the same orientation. Often the spacing of sampling locations along the transect is much shorter than the distance between transects. This form of transect sampling results in a rectangular grid sample. The measurements generated from such sampling can be used to estimate population averages, proportions and other percentiles of the distribution. 2.3. Geostatistical based sampling methods Some specialized techniques have been developed for spatial data. In statistical terminology, these are model-based techniques, since underlying the analysis there is a well-defined statistical model. The most commonly used approach is that of kriging, which is an optimal interpolation method, often used to provide estimates of an assumed continuous attribute, based on a measurements made at a sample of locations (such as a map of Chernobyl fallout). Estimates of the map surface are based on a weighted average of the attribute of interest at neighboring sites. The weights are based on the spatial correlation, which must be estimated. The spatial correlation is often presented as a function of the distance separating the sites and presented graphically in a variogram or semi-variogram. The kriging algorithm can provide a spatially continuous map and can also be used to provide predictions of the attribute at unsampled locations. An important part of the calculation includes production of a map of the uncertainty which can then be used to better design subsequent sampling schemes. Geostatistics is a large and active research topic in its own right and the interested reader is referred to the specialized literature referenced at the start of the section for further detail. 2.4. Conclusions Spatially referenced data are becoming an increasingly common part of many radioecological problems. Many choices of sampling design are possible. The most common classical choices are simple random sampling, systematic sampling using transects or grids, or stratified sampling. Other sampling methods, based on geostatistical ideas, can be used both to estimate population characteristics and to predict values at unobserved locations using interpolation schemes such as kriging. A dense grid of predicted values can be used to draw maps. Visualization of the sampling locations and also the measurements are useful exploratory tools, and with increasing availability of GIS software, these methods will gain increasing use in radioecology. 3. Other sampling methods and some common problems 3.1. Other sampling methods Other, more specialized and hence less common sampling methods include two-stage sampling, which involves definition of primary units, some fraction of which are selected ran-
Statistical sampling design for radionuclides
13
domly. Then, the selected primary units are sub-divided and a fraction of the sub-units are selected randomly. Cluster sampling is most frequently applied in situations where members of the population are found in clusters or colonies. Then, clusters of individuals are selected randomly and all individuals within each cluster are selected and measured. Double sampling can be useful when one characteristic may be difficult or expensive to measure but another related characteristic is simple or easy to measure. This might involve making a relatively large number of analyses using the more efficient technique, and selecting a few specimens from this sample on which to make the more expensive analysis. These and other sampling schemes can be found in Thompson (2000) and Thompson and Seber (1996). 3.2. Number of replicate samples One of the most commonly asked questions is “how many individual samples are required?” To answer this question, we must ask a further series of questions. The questions given here focus on the mean, but the approach can be applied for any population parameter. (a) First we must determine if it is the population mean or the difference in two population means which is of interest. If it is the difference in two means, then (b) we must ask how much of a difference would be of real world importance and hence is important to be able to detect. (c) Next, how variable is the quantity in the population, i.e. what is the variance in the population? (d) Finally, how sure do we want/need to be in the answer, i.e. what is the desired standard error (for an estimate) or statistical power (for detecting a difference)? If a quantity is very variable in the population, then we are likely to detect only very large effects. A precise estimate of the population mean will require a large sample size. Alternatively, if we want to study a small effect, then we will need to increase the sample size perhaps to an unmanageable level. In both cases, there is a trade-off between sample size, precision and size of effect. The exact relationships depend on the sampling design and the quantities of interest. Formulae for sample size determination can be found in many statistical textbooks; the computations are implemented in general statistical software and specialized programs (e.g., Visual Sample Plan; Pulsipher et al., 2004). 3.3. Practical sampling issues One common practical issue that might arise concerns the problem of not being able to follow exactly the pre-determined statistical sampling design. This happens for various good reasons, such as problems with sample selection (e.g., material not present at the selected site), the presence of obstacles or conditions preventing a sample being taken at a given location, and analytical problems (e.g., insufficient amount of material for analysis). Absence of suitable material is a common source of missing values in environmental sampling. For example, if lichen are being sampled, what should be done if the designated sampling location has no lichen? One common solution is to select a nearby location that does contain appropriate material. Another is to ignore the location and reduce the sample size. A less common solution is to select another location using the original sample design. This
14
E.M. Scott and P.M. Dixon
approach differs from the first in that the alternate location may be far from the original location. There are no definitive guidelines about a best strategy to adopt. Good sense and knowledge of the environmental context will provide guidance.
4. Conclusions Sampling for radionuclides in the environment is similar to sampling for other attributes of environmental media. Statistical sampling is pertinent and necessary in radioecology because of the natural stochastic variation that occurs, and the fact that this variation is usually much larger than variations associated with measurement uncertainties. The environmental context of the problem affects the nature of the sampling to be carried out. Individual cases vary in objective and environmental context. Design of sampling schemes requires problem-specific environmental knowledge, statistical knowledge about the choice of the sampling design and practical knowledge concerning implementation. Frequently, practical issues can limit sampling designs. Good discussions of the statistical aspects of sampling include the general sampling textbooks by Cochran (1977) and Thompson (2000), the environmental statistics textbook by Gilbert (1987), and many scientific papers including ICRU (2006).
Acknowledgements This work draws heavily on the preparation work for the ICRU report (2006). The drafting committee included Ward Whicker, Kurt Bunzl, and Gabi Voigt. The many helpful discussions with them are gratefully acknowledged.
References Burrough, P.A., McDonnell, R.A. (1998). Principles of Geographical Information Systems. Oxford Univ. Press, Oxford, UK. Cochran, W.G. (1977). Sampling Techniques, third ed. John Wiley and Sons, New York. Cressie, N. (2000). Statistics for Spatial Data, second ed. John Wiley and Sons, New York. Dixon, P.M., Ellison, A.M., Gotelli, N.J. (2005). Improving the precision of estimates of the frequency of rare events. Ecology 86, 1114–1123. Gilbert, R.O. (1987). Statistical Methods for Environmental Pollution Monitoring. Van Nostrand Reinhold Company, New York. International Commission on Radiation Units and Measurements, ICRU (2001). Quantities, Units and Terms in Radioecology. ICRU Report 65, ICRU 1 (2). Nuclear Technology Publishing, Ashford Kent, UK, pp. 1–48. International Commission on Radiation Units and Measurements, ICRU (2006). Sampling for Radionuclides in the Environment. ICRU Report 75. Oxford Univ. Press, Oxford, UK. Pulsipher, B.A., Gilbert, R.O., Wilson, J.E. (2004). Visual Sample Plan (VSP): A tool for balancing sampling requirements against decision error risk. In: Pahl-Wostl, C., Schmidt, S., Rizzoli, A.E., Jakeman, A.J. (Eds.), Complexity and Integrated Resources Management, Transactions of the 2nd Biennial Meeting of the International Environmental Modelling and Software Society. iEMSs, Manno, Switzerland. Sanderson, D.C.W., Scott, E.M. (2001). Special issue: Environmental radiometrics. J. Environ. Radioact. 53 (3), 269–363.
Statistical sampling design for radionuclides
15
Thompson, S.K. (2000). Sampling. John Wiley and Sons, New York. Thompson, S.K., Seber, G.A.F. (1996). Adaptive Sampling. John Wiley and Sons, New York. United States Environmental Protection Agency, US EPA (2002). Guidance on Choosing a Sampling Design for Environmental Data Collection for Use in Developing a Quality Assurance Project Plan. EPA QA/G-5S, EPA/240/R02/005. EPA, New York. Wackernagel, H. (2003). Multivariate Geostatistics: An Introduction with Applications, third ed. Springer-Verlag, New York. Webster, R., Oliver, M.A. (2001). Geostatistics for Environmental Scientists. John Wiley and Sons, New York.
This page intentionally left blank
17
Sampling techniques Fedor Macášek∗ Department of Nuclear Chemistry, Faculty of Natural Sciences, Comenius University, Mlynská dolina CH-1, SK-84125 Bratislava, Slovakia
1. Introduction There are several ways to analyze radionuclides in the environment. Direct radioactivity measurements for environmental surveys have been carried out by placing a detector near the media being surveyed and inferring radionuclide levels directly from the detector response. Scanning is a measurement technique performed by moving a portable radiation detector at a constant speed above a surface to assess areas of elevated activity. However, there are certain radionuclides that will be impossible to analyze using this simple approach because of their decay properties. Examples of such radionuclides include pure beta emitters such as 3 H, 14 C, 90 Sr and 63 Ni and low-energy photon emitters such as 55 Fe and 125 I. Analysis of alpha emitters is also restricted to surface contamination of relatively smooth, impermeable surfaces. Although direct measurements are practical for space–time monitoring and obtaining averaged radionuclide levels they cannot provide information on structural peculiarities of radioactive contamination. Therefore, a detailed environmental survey usually starts by sampling and preparation for laboratory analyses and is characterized by inspection of the technically and financially feasible fraction of the entire population of objects under investigation. The purpose of sampling is to achieve a mass reduction by collection of a certain subpopulation of the objects of analysis while preserving complete similitude of the crucial parameters of the sample and object. The ideal sample has the same composition as the sampled object; in the case of radiometric and radiochemical analysis of environmental samples, this means the same massic activity (e.g., Bq kg−1 ) or activity concentration (e.g., Bq m−3 ) of radionuclides, and the same relative abundance of their physical and chemical forms. The persistent difference between the true (or total population) value and the sampled one (subpopulation) is an exhibition of bias—a systematic distortion of the measurement in one direction. Sampling for radionuclide analysis is thus defined as the process of collecting a portion of an environmental medium which is representative of the locally remaining medium. Represen∗ Present address: BIONT, Karloveská 63, SK-84229 Bratislava, Slovakia. E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11002-0
© 2008 Elsevier B.V. All rights reserved.
18
F. Macášek
tativeness is characterized as a measure of the degree to which data accurately and precisely represent a characteristic of a population, parameter variations at a sampling point, a process condition, or an environmental condition (EPA, 2002a, 2002b, 2002c, 2002d; ANSI, 1994). Still, the obstacles in practical interpretation of such a definition are obvious when you imagine any sampling of your actual close-by environment. Therefore the definition of sampling as the process of gaining information about a population from a portion of that population called a sample (Hassig et al., 2004) is more universal and acceptable. The sampling is the foundation block on which any analytical result is built. Without reliable out-of-lab and in-lab sampling plans no quality assurance of results can be achieved. Most traditional strategies of environmental sampling were implanted from the field of geochemical assay, which should provide precise information on the ore and other raw material resources within the area designed for exploitation (Journel and Huijbregts, 1978; Isaaks and Srivastava, 1989). Even more, the same random sampling tests are incorrectly applied in the same way as in the chemical and pharmaceutical industry where their goals are to provide evidence of declared identity and homogeneity of the whole batch sampled (Duncan, 1986; Schilling, 1982). For such purposes the reliable statistical standards are well elaborated (ISO, 1991, 1995, 2000, 2001; ASTM, 2002). Specific sampling analysis plans are necessary for survey of the environment where the field is random by its nature (Gilbert, 1987; Cressie, 1993; Myers, 1997; Byrnes, 2000; USACE, 2001; EPA, 2001). The MARSSIM (the EPA’s MultiAgency Radiation Surveys and Site Investigations Manual) guide focuses specifically on residual radioactive contamination in surface soils and on building surfaces (EPA, 2002a, 2002b, 2002c, 2002d). The natural and industrialized environment is neither homogeneous nor unchanging. Even the most illusive homogeneous system like the atmosphere is in fact a colloidal/particulate system—typical indoor air contains about 107 –108 particles in the 0.5 to 100 µm size range per cubic meter. Air sampling needs the size fractionation and characterization of particles, the radioaerosols specifically: at least, distinguishing the radioaerosol size below 1 µm and in the interval 1–5 µm is necessary for radiation protection assessment purposes (ICRP, 1966; Koprda, 1986; ISO, 1975; IAEA, 1996). The measurement of pollutants, the radioactive ones included, may be oriented less to their inventory and more towards the estimation of their role in litho- and biosphere migration which cannot be expressed unequivocally either in space or time. Monitoring of contamination of soil by radiocesium serves well as an illustration. The global contamination with 137 Cs characterizes the radiation field above the ground and it can be easily determined by in situ gamma spectrometry and aerospace scanning. However, when there is the concern for radiocesium bioavailability or soil remediation (Cremers et al., 1988; Navratil et al., 1997; Wauters et al., 1996) the assay of the soil should include determination of the mobile forms of cesium (Macášek and Shaban, 1998; Bartoš and Macášek, 1999, 2002). The same applies to other radionuclides and metal ions (Ure et al., 1992; Smith-Briggs, 1992; Hlavay et al., 2004) that can exist in several different soil or sediment phases, e.g., in solution (ionic or colloidal), organic or inorganic exchange complexes, insoluble mineral/organic phases, precipitated major metal oxides or in resistant secondary minerals. Very often, environmental authorities and decision-makers long for reproducible and reliable (= accurate) data on pollution, i.e. smooth data with a narrow statistical deviation. From this request the traditional features of the sampling and preparatory treat-
Sampling techniques
19
Table 1 Requested parameters for food and environmental radioactivity monitoring (IAEA, 1988) Monitoring
Tolerable bias
Assessment time
Screening Very fast Fast
10× 2–3× 20–50%
5–15 min 1–6 h 6–24 h
ment of environmental samples for trace and radiochemical analysis follow (Green, 1979; Alberts and Horwitz, 1988), such as the concern for representativeness of the results. However, the principal question in environmental assay, and that for radioactive pollution in particular, is whether all the requirements for a radiation survey should be satisfied with representative samples and the indication of a low uncertainty of the analytical techniques or by a scattered plenitude of data on naturally occurring samples? It is evident, for example, that the subsamples containing hot particles of burnt nuclear fuel after the Chernobyl accident (Tcherkezian et al., 1994; Salbu et al., 1994) should be statistically found irrelevant by going far beyond the confidence limits of a normal distribution of activity of samples collected from the contaminated areas. The total variance of sample assay σt2 is σt2 = σd2 + σs2 + σl2 ,
(1)
where σd2 , σs2 and σl2 are the total variances of the sampling design, field sampling and laboratory technique, respectively (Smith et al., 1988; Kirchhoff et al., 1993). The time factor is significant in general sampling design. For various decision-making tasks the required accuracy/precision of radionuclide monitoring is compromised by the sampling, processing and measurement times. For example, during a nuclear incident, some compromise must be reached between the ideal sample coverage and the need to gather larger numbers of samples as rapidly as possible, as in Table 1 (IAEA, 1988). 1.1. Sampling homogenization versus speciation Ordinary application of mean values of massic activities and standard deviations characterizing the laboratory technique is imperfect because variation in the object of sampling is usually much larger than the variation of laboratory techniques. It could resemble a situation when a tailor would use his customer’s family’s average size and the tape measure distortion as a legitimate bias instead of the individual and changing differences between folk—most probably his garments will fit nobody. The effort for representativeness incorporates: (1) Taking of “sufficiently large” samples (Mika, 1928; Baule and Benedetti-Pichler, 1928; Gy, 1992), which is a must when the specific activity is below the detection limit of radiometry and pre-concentration (matrix removal) is necessary (2) homogenization and/or other processing of samples (Remedy and Woodruff, 1974), and (quite often)
20
F. Macášek
(3) spiking or isotopic labeling of the sample (isotope dilution) if the reprocessing of the sample leads to losses of analyte (Tölgyessy et al., 1972; Macášek, 2000). As shown below, the homogenization in principle leads to a loss of information contained in the sample. Hence, it is substantiated only when the sample is defined as homogeneous and its specific activity, as an average value, is deemed important, the assay of the layers of ground investigated for penetration of radionuclides to depth being an example. An averaged radionuclide content in the sample results from a population of n different sorts of entities. Each ith sort of entity represents a homogeneous component, such as a chemical compound, species or particulate form of radionuclide, of identical composition. When i indexes these constitutive species, the average concentration can be expressed as a function of the massic activities of radionuclide species ai and the fractional abundances of the species xi , a¯ =
n
xi ai .
(2)
i=1
The massic activity of any j th subsample is given by a subset of entities, a¯ j =
ν
yij ai ,
(3)
i
where “i”s are ν random values from the interval 1, n and yij is the abundance of the ith entity in the j th subsample. The subsample is representative in respect of massic activity, when a¯ ≈ a¯ j which is merely a statistical task except for a simple form of radionuclide in a homogeneous matrix (i = 1), a true solution of radionuclide present in a single particulate and chemical form. Also the evaluation of various species in heterogeneous objects usually proceeds through replacement of the full set of parameters by assuming that yij = xi .
(4)
This traditional access is still most widespread and representativeness is ensured by a large sample and random mass (volume) reduction to ensure its validity. Also, a composite sample is formed by collecting several samples and combining them (or selected portions of them) into a new sample which is then thoroughly mixed. The approach is strongly favored, e.g., by the tasks of geological resource assessment, technological material balances, artificial preparation of modified materials and standards for validation of analytical methods. The approach always looks attractive because of the ability of current analytical techniques to supply reproducible data of low uncertainty for a set of replicate samples—the relation to true value of radionuclide concentration (the massic activity) remains the ultimate test for quality of measurements. In fact, the objects of environmental origin are far from being either homogeneous or uniform and it is therefore difficult to present precise data. More sophisticated procedures may include: (i) an estimation of specific forms of radionuclide in various fractions of the representative sample (i > 1, j = 1), i.e. a physico-chemical speciation that is the “species distribution or abundance studies which provide a description of the numerical distribution (or
Sampling techniques
21
abundance) of different species containing the same central element, in a given sample” (Pickering, 1995; Macášek, 1994, 1996). (ii) a treatment of a more complete subset of analytical data on the partial concentrations and matrix composition (m > 1, j > 1). Such a procedure is applied when variability of the samples appears to be too high, e.g., in analysis of acid extracts of sediments, or for stratified (segregated) random sampling (ACS, 1980). Reduced uncertainty of environmental radionuclide data is artificially achieved in the course of thoroughly applied but entropy generating homogenization procedures. The grinding, milling, mincing, blending, chopping, mixing, melting, mineralization, burning and dissolution procedures are substantiated when the analyte concentration is below the detection limit of an analytical technique and a pre-concentration is necessary. However, in the course of homogenization, considerable information content is lost according to the Shannon information entropy (Eckschlager and Danzer, 1994), pi ln pi H =− (5) i
(information can be considered as negative entropy). H becomes minimal (zero) for an ideally homogenized sample when radionuclide should be in a single form (xi = 1) and the probability of its presence is either pi = 1 or pi = 0 (in the latter case, conditionally, pi ln pi = 0). A sample preserving, e.g., two particulate forms of radionuclide at equal abundance (pi = 1/2) has the information content for 0.693 entropic units higher. 1.2. Ontology of environmental analysis The question is “why?” and “when?” the sample should be “representative” and “precise”. This attribution is transformed to the problem of “what?” should be determined in the sense of what information (not identical to the “activity” or “massic activity”!) should be extracted from the analysis of environmental objects and how should it be properly presented. The final question is a real ontological problem of the purpose of analysis (Macášek, 2000). Therefore, analytical scenarios must be designed from the very beginning by co-operation of environmentalists and (radio)analytical chemists or other nuclear specialists. In other words, the way to determine radionuclides in environmental objects strongly depends upon the mode of application of the data received; such an assessment may be shown to vary from a random sampling and sample homogenization to a stratified sampling and sophisticated speciation of radionuclides in the environment. The aim of the analysis is conditioned by the purpose for which the data are being collected. Obviously in many cases, instead of an averaged value of analyte determination in a homogenized sample and an uncertainty derived from standard deviations of the analytical procedures, more complex issues should be derived. It should be stressed that quality assurance, good laboratory practices, reference methods and reference materials should be applied throughout the analytical procedures (Povinec, 2004). We shall characterize (Table 2) the goals of data collection and related sampling and sample treatment by the preservation of information content, i.e. the consequent change of entropy in
22
F. Macášek
Table 2 Links between the objectives of analysis, respective sampling and pre-analytical treatment strategy Sampling and treatment Homogenization Composite sampling
Random replicate sample analysis
Stratified random sampling and analysis Speciation
Strongly entropic mode
Medium entropic mode
Low entropic mode
Objectives (examples) (1) Radioactivity inventory, regional abundance and availability of radionuclide, differences between areas and contour maps of GIS (2) Regional environmental impact evaluation (3) Regional spatial dynamics of radionuclide in pseudocontinuous environment (4) To determine source and undeclared releases of radioactivity
(1) Natural distribution and environmental heterogeneity indication by a non-parametric evaluation
(1) Physico-chemical disclosure of species and their relations
(2) Test statistics for hypothesis tests related to particular parameter (3) Assessment of a stochastic action of radionuclide
(2) Specific action of radionuclide species
(4) Average mobility of radionuclide in multicomponent media
(3) Distribution between environmental compartments and food chains (4) Specific mobility and transfer of species (5) Future behavior forecast
the course of reprocessing—see Section 1.2. Obviously, the associated costs increase to the right and down the list. Direct physical and chemical speciation is desirable, e.g., in ecotoxicological evaluation (function) of analyte (Remedy and Woodruff, 1974): 1. functionally – biological availability – toxicity – mobility and transfer 2. operationally – leachability – exchangeability 3. particulate, morphological and chemical state. The native state of species can be seriously disturbed not only by a destructive pre-analytical treatment but also in the course of exposure of samples to air oxygen and microbial flora, light, heat and contact with sampling devices and vessels. Then the task of analytical speciation is either to distinguish “stable” species or to get a fingerprint of original native abundance (Macášek, 1994, 1996). The time factor is also important; due to their heterogeneity, photochemical and metabolic processes the natural objects occur much more frequently in a steady-state than in thermodynamic equilibrium.
Sampling techniques
23
2. Optimization of sampling 2.1. Data quality objectives Data quality objectives (DQOs) are “qualitative and quantitative statements derived from the process that clarify study technical and quality objectives, define the appropriate type of data, and specify tolerable levels of potential decision errors that will be used as the basis for establishing the quality and quantity of data needed to support decisions” (EPA, 2000, 2002a, 2002b, 2002c, 2002d). Data quality objectives derive from a systematic scientific planning that defines the type, quality and quantity of data needed to satisfy a specified use (EPA, 2000). The key elements of the process include: • • • • • • •
concisely defining the problem identifying the decision to be made identifying the inputs to that decision defining the boundaries of the study developing the decision rule specifying tolerate limits on potential decision errors, and selecting the most resource efficient data collection design.
Data validation is often defined by the following data descriptors: • • • • • •
reports to the decision maker documentation data sources analytical method and detection limit data review, and data quality indicators.
The principal data quality indicators according to EPA (1998) are: • • • • •
precision bias representativeness comparability, and completeness.
Uncomplicated spikes, repeated measurements and blanks are used to assess bias, precision and contamination, respectively. Other data quality indicators affecting the radiation survey process include the selection and classification of survey units, uncertainty rates, the variability in the radionuclide concentration measured within the survey unit, and the lower bound of the gray region. Of the six principal data quality indicators, precision and bias are quantitative measures, representativeness and comparability are qualitative, completeness is a combination of both qualitative and quantitative measures, and accuracy is a combination of precision and bias. The selection and classification of survey units is qualitative, while decision error rates, variability, and the lower bound of the gray region are quantitative measures. Data qualifiers (codes applied by
24
F. Macášek
the data validator) help quickly and critically to judge the collected data and give the manager a sign on how to use them (EPA, 2002c). Environmental data verification and the validation processes test for non-conformance in project planning, field activities and sample management. The process for determining the utility of the obtained data is based on scientific and statistical evaluation of whether they are of the right type, quality and quantity to support their intended use (EPA, 2002d). 2.2. Sampling plan The sampling plan can be divided into design and field sampling components. The design part of sampling solves the questions • • • • • •
what is the objective of sampling what type of samples to pick in respect of radionuclides of interest minimal amount of sample necessary for its laboratory and statistical assay how many samples to collect where to sample, and when and at what intervals to sample.
Special care should be addressed on choice of blank samples. The field sampling phase is concerned with • • • • • • • • • • • •
site preparation who takes the samples how to describe them how to identify and label samples how to collect them how to avoid cross-contamination how to stabilize samples how to pack samples how to transport samples how to store, and how to advance samples to the laboratory, and last but not least what are the sampling and treatment costs.
The sampling plan is usually a compromise between various demands of data users, sampling enterprise and analytical laboratories on accessibility and availability of samples and cost-effectiveness of sampling and measurement procedures. The last strongly differs for low and high activity samples, type (alpha, beta and gamma) of radionuclides, type of matrix and required precision of individual data. Basically, two alternate procedures for sampling plans and test hypothesis are developed: (1) Classical procedures of random sampling when nothing is considered about the population under investigation (i.e., absence of prior data or “zero hypothesis”), each individual sample in a population has the same chance of being selected as any other, and just the information received experimentally is taken into account (“frequentist” approach). (2) Bayesian methods when the prior data (i.e., information collated prior to the main study) are used in the establishment of a priory probability laws for sample parameters, i.e. new
Sampling techniques
25
Table 3 Number of samples for a frequentist approach Number
n
Substantiation
Fixed
1 2 3 5 10 21 or 28
Minimally imaginable Enabling a rough assessment of average value and standard deviation Median can be assessed with its approximate deviation A non-parametric estimate of deviation becomes possible Adapted for populations 10 N 100 Evaluation of median and distribution in percentiles (quantiles)
Population N related N/20 . . . N/10 √ N
To be used when N < 10 A “reasonable” fraction (5–10%) of a medium size of population, 30 N 100 Deviations diminish with size of analyzed samples, suitable for larger populations (N > 100) 20 + (N − 20)/10 Good evaluation of median and distribution in percentiles (quantiles) in populations (N > 100)
samples are planned according to available information or a working hypothesis on the former ones. Bayesian methods might be further classified according to their approach to prior information. The uninformative approach implies there is no information concerning the assayed prior sample parameters. Empirical methods extract the prior distributions from the available statistical information on the studied samples. Subjective Bayesian methods use the prior information based on expert personal beliefs (Carlin and Louis, 1996). Application of the Bayesian approach enables minimization of the number of samples necessary for confirmation of a working hypothesis. The determination of the number of samples that should be gathered is one of the most important practical tasks of a sampling plan (Aitken, 1999). The number of samples may issue from the allowed decision uncertainty and estimated variability of the samples. The variance of parameters x in a population of n samples is usually derived from the standard deviation based on a Gaussian (“bell-shaped”) distribution, n (xi − x) ¯ 2 2 σ = i=1 (6) , n−1 i.e. n > 1 is necessary. However, the Student coefficient as large as 12.7 for a 95% confidence interval should be considered for the standard deviation obtained from two samples! Thus, two-item replicates are the customer’s favorite as a minimal number to estimate variability. Five sample replicates may be sufficient for a non-parametric evaluation—see below. A replicate set of 28 samples is considered suitable for non-Gaussian estimates of the 10-, 25-, 50-, 75- and 90-percentile points (the P -percentile point is that value at or below which P percent of total population lies)—see Table 3. When the estimate is not available from a pilot study of the same population, a conservatively large preliminary value of the population variance should be estimated by another study conducted with a similar population, or based on a variance model.
26
F. Macášek
Table 4 Number of samples to be taken for one-sample t-test (Tukey, 1977) Significance
Test power
10%
Relative size of gray region 20%
30%
40%
50%
Risk level 5%
95% 90% 80%
1084 858 620
272 216 156
122 97 71
69 55 40
45 36 27
Risk level 10%
95% 90% 80%
858 658 452
215 166 114
96 74 51
55 42 29
36 28 19
To determine the minimum replicate sample size needed to estimate a population mean (for example, a mean contaminant concentration), in the total absence of prior information, the rough standard deviation expectation can be calculated by dividing the expected range of the population (between expected maximal and minimal values of x) by six, i.e. xmax − xmin . (7) 6 Using either Equation (6) or (7), the minimum sample size needed to achieve a specified precision for estimates of the population means and proportions is further evaluated. Table 4 gives the number of samples necessary for a hypothesis test for the number of samples to be taken with acceptable risk level of a wrong decision. It illustrates the reliability of a test when the relative width of the gray region towards the standard deviation is within 10 to 50% (broader data are available for various sampling designs; EPA, 2002a, 2002b, 2002c, 2002d). As seen, the number of samples strongly increases when the gray zone of an ambiguous decision is formulated too narrowly. For example, the limit for contamination of construction walls with 239 Pu is 3 kBq m−2 . In preliminary investigation of a building, the activity of samples did not exceed a rough value of 2.6 kBq m−2 . The sampling should ensure that neither false contamination under a true value of 3.03 kBq m−2 nor a false clean facility, actually contaminated above 2.8 kBq m−2 , will be announced with an acceptable level of risk 5%. From the last figures, the gray area width is 3.03 − 2.85 = 0.18 kBq m−2 . According to Equation (7), the standard deviation may be expected to be (2.6 − 0)/6 = 0.43 kBq m−2 . The relative width of the gray area is 100 × 0.18/0.43 = 45%. If the test reliability (test power) should be 90%, we found from Table 4 that the necessary number of random samples should be between 36 and 55 (say 45) to confirm the hypothesis of the mean contamination level of facility walls. The same value will harmonize with a higher, 95% reliability of test, but also a higher acceptable 10% level of risk. Two statistical tests are usually used to evaluate data from final status surveys. For radionuclides also present in background, the non-parametric Wilcoxon rank sum (WRS) test is used. When contaminants are not present in background, the sign test is used (EPA, 2002a). A desired sample size in the Bayesian approach is derived from the hypothesis of the occurrence of non-conforming units of population (e.g., the fraction of those contaminated by σ =
Sampling techniques
27
Table 5 Required sample size n to guarantee with 95% confidence that at least a fraction θ = q/n of population units is clean if all samples are clean (q = n) Population, N
n θ = 50%
θ = 70%
θ = 90%
10 20 30 40 50 90 100 1000 5000 10000
3 4 4 4 4 5 5 5 5 5
5 6 7 7 8 8 8 9 9 9
8 12 15 18 19 23 23 28 29 29
radioactivity) which should be accepted or rejected on an empirical base. Let the zero hypothesis says that in the whole population of samples there are Q clean units and the rest R = N − Q is contaminated. The question is what is the probability to find q clean samples in the set of n samples (and r = contaminated ones), while the sample set is representative, i.e. the fraction θ is the same as in the whole population, θ = Q/N = q/n? Such conditional probability P can be obtained as Q R P (q|N, Q, n) =
q
r
N .
(8)
n
Some data are illustrated for the case when it is necessary to confirm various fractions θ of clean population units by obtaining n totally conforming (clean) samples in Table 5. It follows that for confirmation of, e.g., area cleanness, it is advantageous to withdraw random samples from a large set of population units, and with almost the same effort applied for their analysis, the statement on contamination may cover a much larger area. For large populations (N > 100), the calculation by hypergeometric distribution (Equation (8)) can be easily approximated by a binomial distribution, n q P (q|θ, n) = θ (1 − θ )n−q , (9) q which for the case of all negative samples (n = q) gives P = θ n and n can be calculated for probability P and fraction θ simply as n = log P / log θ . For example, to guarantee with 95% confidence (P = 0.95) the occurrence of 0.2% or less of contaminated units (θ = 0.998), the number of all negative samples from a large population should be at least n = log(0.95)/ log(0.998) = 2.23 × 10−2 /8.69 × 10−4 ≈ 27. 2.2.1. Simple random sampling A simple random sampling design is the most common way of obtaining data for statistical analysis methods, which also assume that the data were obtained in this way. For a simple
28
F. Macášek
random sampling plan it is typical that the population (field) is considered systematic (not random) and its premises are irrelevant; it is the mode of sampling that is kept random. It is clear that simple random sampling is appropriate when the population is indeed relatively uniform or homogeneous. Visually, the classical simple random sampling is the “basket of apples” method. If one wants to measure the parameters of apples (activity in particular) without using the whole population of apples on a tree, just a basket of apples is taken for analysis. The randomness lies in arranging the choice of the apples and random numbers are used to determine which sample of the whole set is to be taken. Then the averaged parameter of the subpopulation in the basket and its variance is supposed to perfectly imitate the “true” parameters of the whole population—see discussion for Equation (4). Random sampling would meet numerous logistic difficulties in reaching the sites picked on a map by random numbers. Sometimes further on-site randomization (classically, the blind throwing of a rock hammer) is necessary in this case. From the information point of view, any spatial interrelation existing in the original population of samples (“apples on a tree”) is destroyed by randomization and information is discarded. 2.2.2. Systematic-random sampling A special random sampling technique to obtain regional coverage is systematic sampling. The sampling locations vary in a systematic manner and the distance between the samples is uniform. However, the first location in the tissue is random; a rectangular Cartesian grid is laid over the investigated region, the origin and axis orientation of co-ordinates being obtained from random numbers. Both systematic-random and pure random sampling are used to converge to the expected parameter. However, estimates with systematic-random sampling converge to the expected value sooner. Systematic-random sampling makes more precise estimates in less time; therefore, systematic-random sampling is more efficient than pure random sampling. 2.2.3. Stratified-random sampling Stratified random sampling consists in a preliminary subdivision of the samples into more homogeneous groups, subgroups or strata. There can be no sample that does not belong to any of the strata and no sample that belongs to more than one stratum. The subgroups are expected to exhibit smaller variance than the whole population and the weighted combination of strata means gives the value for the total population—see Equation (2). Preserving the rhetoric of apples, the strata can be composed of the fruits separately collected from the top and bottom of a tree or from its north and south sides. In the case of biota they may be lichens or mushrooms chosen from the flora, etc. Most verbatim, various soil layers can be sampled as strata. In most situations the vegetative cover is not considered part of the surface soil sample and is removed in the field. A specific stratum for sampling is a cell of a Cartesian (also an irregular) grid, from which a specified number of random samples are collected. GPS (Global Positioning System) technology facilitates irregular grid sampling. In the working environment, strata are more imaginary, e.g., exposure zones based on similarity of jobs, environment, etc. Temporal strata permit dif-
Sampling techniques
29
ferent samples to be selected for specified time periods and support accurate monitoring of trends. The stratification of objects happens to be affected by the strong personal opinions of planners led by their intuition. Such discretionary samples may be helpful but certainly cannot be used validly within the random sampling plans. 2.2.4. Representative sampling As discussed in Section 1.1, the representativeness of a particular sample cannot be quantified because the “truth” is not known. It can be verified through application of statistical approaches based on the actual measurements. When the average concentration of analyte is of main interest, Mika’s formula (Mika, 1928) can be applied for estimation of the minimal mass m representing an inhomogeneous object, m>K
d3 , δ2
(10)
where K (g mm−3 ) depends on relative size and relative weights of species, d is the diameter (mm) of the largest particles and δ is the tolerated relative uncertainty. This equation is simplified in the “Gy’s safety rule” (Pitard, 1993) which calls for the minimum mass of solid samples m 125d 3 .
(11)
For example, for the sieved soil samples with particle diameter d 1 mm the minimum representative size of sample is calculated as 125 g, and when 0.25 g subsamples are assayed the sample should be powdered below 0.13 mm. 2.2.5. Geostatistical (random field) sampling When there is a spatial or temporal dependence, samples close together in space and time scale will tend to have more similar values than samples far apart. This is often the case in an environmental setting. Geostatistical sampling resides in such a sampling procedure that is guided by the assumed properties of some region via random space–time coordinates and prior estimates of the covariance functions (statistical measures of the correlation between two variables) to account for the pattern of spatial continuity (Borgman and Quimby, 1988; Armstrong, 1998). Initial estimation of the covariance structure of the field and its stationary or non-stationary character is of principal importance. Therefore, the design phase of a geostatistical sampling plan when a model for probability law is to be developed is rather laborious but the field phase and gross sampling are less costly than for random sampling. A primary advantage of geostatistical sampling method is that sample sites may be relatively freely selected by personal judgment to best cover the investigated area. On the other hand, the sample sites should allow estimation of the covariance structure, the correlation of data in space and time. Site preparation involves obtaining consent for performing the survey, establishing the property boundaries, evaluating the physical characteristics of the site, accessing surfaces and land areas of interest, and establishing a reference coordinate system. A typical reference system spacing for open land areas is 10 meters and can be established by a commercially available GPS, while a differential GPS provides precision on the order of a few centimeters.
30
F. Macášek
2.2.6. Ranked set sampling In this two-phased approach, r subsets of m subsamples each are selected and ranked according to some feature that is a good indicator of the parameter of interest using professional judgment or a rough estimate (Bohn and Wolfe, 1994; Patil et al., 1994). From the first subset only the first ranked unit (rank m = 1) is chosen and measured. Another set is chosen, and the (m + 1)th ranked unit is chosen and measured, etc. The advantage is that only r samples are sufficient to estimate an overall mean and variance, instead of the full set r × m. For example, suppose that nine samples would be randomly selected and grouped into three groups of three each. The three samples in each group would be ranked by inspection (assumed to be correlated with the parameter of interest). The sample with rank 1 in group 1, the sample in group 2 with rank 2, and the sample in group 3 with rank 3 would be composited and analyzed. The initial group of nine samples reduces to only one composite sample of size three. In terms of the precision of the estimated mean, such an approach should perform better than a simple random sample of size three (though worse than a simple random sample of size nine). 2.2.7. Adaptive cluster sampling Choosing an adaptive cluster sampling design has two key elements: (1) choosing an initial sample of units and (2) choosing a rule or condition for determining adjacent units to be added to the sample (Thompson, 1990, 2002; Seber and Thompson, 1994). Initial samples are selected randomly and evaluated; it is most useful when the field radioactivity measurement can be used for this step. Then additional samples are taken at locations surrounding those sites where the measurements exceed some threshold value. Several rounds of such sampling may be required. Adaptive cluster sampling is similar in some ways to the kind of “oversampling” done in many geostatistical studies. Therefore, selection probabilities are used to calculate unbiased estimates to compensate for oversampling in some areas via either declustering techniques, polygons of influence, or kriging. Kriging is the extrapolation method to estimate a field at an unobserved location as an optimized linear combination of the data at the observed locations (Stein, 1999). Kriging also allows an estimate of the standard error of the mean once the pattern of spatial covariance has been modeled. 2.2.8. Hot spot identification The Bayesian method can be also demonstrated in the identification of hot spots, relatively small areas of elevated activity (EPA, 1989). The sampling plan for identification of a hot spot of the radius R is made with a distance D between adjacent quadratic grid points. It is supposed that at least one grid point in any square of area D 2 will fall inside the hot spot. The prior probability P (H |E) of hitting a hot spot at its existence E by such sampling (at large distance from sampling points as compared with the hot spot area) is P (H |E) = 1, when D 2R. The most frequent sampling situation is that the grid distance is large compared with the hot spot radius. If R < D/2 then the probability to hit the spot is P (H |E) = (πR 2 )/D 2 .
(12)
Sampling techniques
31
√ For a less frequent case of medium distances R 2 D 2R the probability is √ R 2 [π − 2 arccos(D/2R) + (D/4) 4R 2 − D 2 ] P (H |E) = (13) D2 where the angle θ = D/(2R) is expressed in radians. Geometrical similitude exists for the same θ . When the probability based on previous experience that a hot spot exists is P (E), then the posterior probability that it does not exist can be found by the Bayes formula. The probability of existence of the hot spot E when there was no hit by the sampling (H¯ ) is P (E|H¯ ) =
P (E)P (H¯ |E) . ¯ (H¯ |E) ¯ P (E)P (H¯ |E) + P (E)P
(14)
¯ = 1, and Because the probability that the hot spot is not hit when it does not exist is P (H¯ |E) ¯ the sum of the prior probabilities to find or not find the spot is P (E) + P (E) = 1, and also P (H |E) + P (H¯ |E) = 1, there is P (E|H¯ ) = P (E)
1 − P (H |E) . 1 − P (E)P (H |E)
(15)
For example, let the sampling plan to find hot spots of radius 2 m have a grid distance of 10 m. The prior probability P (H |E) = 0.126 according to Equation (12). Previous experience from the equal sub-area sampling indicated contamination in 8% of the collected samples, i.e. the probability of hot spot existence is P (E) = 0.08. When at last sampling there was no hit of a hot spot, still there is the probability of its hidden existence, though only slightly lower, 1 − 0.126 = 0.071, 1 − 0.08 × 0.126 and the absence of hot spot probability is 1 − 0.071 = 0.929. P (E|H¯ ) = 0.08
2.2.9. Hot particle sampling Hot particles originated mostly from world-wide fallout after the nuclear weapon tests carried out in the atmosphere (Mamuro et al., 1968), and also from nuclear reactor accidents, principally that at Chernobyl (Sandalls et al., 1993). Their size varies between 0.1 up to the “giant” particles of 30–1000 µm with the occurrence of 1–100 particles per 1000 m3 . Hot particles have much smaller dimensions than a hot spot, but their estimation is important to radiation exposure of a population after a fallout event (ICRU, 2000). A statistical evaluation for detecting hot particles in environmental samples by sample splitting was performed by Bunzl (1997). The presence of hot particles in the environment could be detected with fairly high probability in replicate or collocate samples. The wider the frequency distribution of the activities of the hot particles, the smaller the number of parallel sample measurements is necessary to detect their presence. 2.2.10. Visual Sample Plan VSP 3.0 The Pacific Northwest National Laboratory offers to upload the Visual Sample Plan software (VSP 3.0) (Hassig et al., 2004) elaborated for the U.S. Department of Energy and the U.S. Environmental Protection Agency. VSP is designed for selecting the right number and location
32
F. Macášek
of environmental samples so that the results of statistical tests performed on the data collected via the sampling plan have the required confidence for decision making. VSP allows selection of a design from the following list (all but the judgment sampling are probability-based designs): • simple random sampling • systematic grid sampling on a regular pattern (e.g., on a square grid, on a triangular grid, along a line) • stratified sampling • adaptive cluster sampling • sequential sampling, requiring the user to take a few samples (randomly placed) and enter the results into the program before determining whether further sampling is necessary to meet the sampling objectives • collaborative sampling design, also called “double sampling”, uses two measurement techniques to obtain an estimate of the mean—one technique is the regular analysis method (usually more expensive), the other is inexpensive but less accurate. It is actually not a type of sampling design but rather a method for selecting measurement method • ranked set sampling • sampling along a swath or transect—continuous sampling to find circular or elliptical targets is done along straight lines (swaths) using geophysical sensors capable of continuous detection • sampling along a boundary in segments, which combines the samples for a segment, and analyzes each segment to see if contamination has spread beyond the boundary • judgment sampling—the sampling locations are based on the judgment of the user such as looking in the most likely spot for evidence of contamination or taking samples at predefined locations. 2.3. Replicates, composite and collocated samples An averaged value of the analyte estimation in the homogenized sample, and the uncertainty derived from the standard deviations of the analytical procedures are the regular presentation format for analytical results. However, the distribution of the analyte in fractions of the original sample and its non-parametric statistical (Horn, 1983; Efron, 1981) and physico-chemical speciation (Macášek, 1994, 2000; Pickering, 1995) should be referred to as reflecting the true similarity between the subsamples and virgin matrix. Field replicates are samples obtained from one location, homogenized, divided into separate containers and treated as separate samples throughout the remaining sample handling and analytical processes. They are used to assess variance associated with sample heterogeneity, sample methodology and analytical procedures. Conversely, composite samples consist of several samples, which are physically combined and mixed, in an effort to form a single homogeneous sample, which is then analyzed. They are considered as the most cost effective when analysis costs are large relative to sampling costs. However, their information value is dubious. Collocated samples are two or more specimens collected at the same spot (typically, at a distance of about 10–100 cm away from the selected sample location) and at the same time and these can be considered as identical. Collocated samples should be handled in an
Sampling techniques
33
identical way. Analytical data from collocated samples can be used to assess site variation in the sampling area. From the total variance (Equation (1)) the measure of identity of collocated samples can be obtained as σs2 = σt2 − σd2 − σl2 .
(16)
The real situation is well documented by the fact that has resulted from a growing number of intercomparison analysis—the reliability of the overall means decreases though the statistical uncertainty reaches unrealistically small values, without relation to the real variability of the individual samples (Seber and Thompson, 1994; Thompson, 2002). This indicates that a similarity of averaged values of total and sub-populations is not sufficient, and a similarity of population fields should be considered instead, the latter not being achievable using homogenized gross samples. Anyone’s effort to reduce the uncertainty of radionuclide data does a disservice to the purpose of analytical quality assurance. As usual, the larger the set of samples the better it reflects the abundance of species in the total population of entities. However, the uncertainty expressed by the expected standard deviation (s) of the mean (x) or median (μ) gives just a vague idea of the natural sample variance and creates an illusion of high certainty in the assessment. The treatment of unevenly distributed data by a normal (and even more so for a lognormal) distribution approach means an underestimation of each species abundance. Hence, not only sample homogenization but also normal distribution statistics exhibit the tendency to smooth out the picture of the radionuclide distribution in the investigated area. Still, an environmental decision-maker should be provided not only with averaged values but also with their realistic statistics of environmental abundance. The results of replicate sampling analysis when expressed through percentiles and a Tukey (1977) box plot better reflect the distribution of an analyte species (or the matrix heterogeneity). In routine analysis, usually the criteria derived from a normal distribution are applied to indicate the outlying results for 2s (95% confidence) or 3s (99% confidence) from a mean value. In the Hoaglin statistics which considers the normal distribution of non-rejected results (Patil et al., 1994; Meloun and Militký, 1994) the lower and upper fences, BL and BU , for outliers are established from the 0.25 and 0.75 percentiles as follows, BL∗ = x¯0.25 − K(x¯0.75 − x¯0.25 ),
(17)
BU∗
(18)
= x¯0.25 + K(x¯0.25 − x¯0.75 ),
where K is found for 8 n 100 without any posterior data, just from their number n as 3.6 . K∼ = 2.25 − n
(19)
For small subsets, the limits are non-parametrically assessed as a distance from the median on the value sμ =
x(n−k+1) − x(k) 2u
(20)
34
F. Macášek
which is the non-parametric mathematical expectation of the median dispersion (Meloun and Militký, 1994). For the 95% confidence interval u = 1.96, and k is found from
n+1 n k = int (21) − |u| . 2 4 The dispersion of the values indicates the real situation in the samples, but the fetishism for a mean value mostly prevails and the result would probably be issued and accepted by the client with a standard deviation, but may be corrected by Student coefficients, as indicating a minor “error”. A small set of the smallest subsamples, e.g., 5 randomly collected entities, can be easily treated by four different non-parametric methods: (i) Finding a median and the fences according to Equations (17) and (18); for the set of n = 5 data, k = 1 and n − k + 1 = 5, i.e. the dispersion of the median at 95% confidence is easily obtained from the set. (ii) Marritz–Jarret evaluation (Meloun and Militký, 1994) of the small set of data median dispersion is performed as n 2 2 sμ = (22) wi x − wi xi , i
i
where the weights wi of measurements xi are found from the Bessel functions J , J i−0.5 n wi = n j −0.5 . j =1 J n
(23)
For the set of n = 5 sorted values i
1
2
3
4
5
J wi
0.24 0.05
1.32 0.26
1.88 0.37
1.32 0.26
0.24 0.05
(iii) The median is assessed by pivot range according to the Horn statistics (Patil et al., 1994; Meloun and Militký, 1994) for small subsets of data; indexes of sorted pivot data are derived for 1 n+1 h = int (24) 2 2 or n+1 1 +1 h = int (25) 2 2 —depending on which one is an integer—and iL = h,
(26)
iU = n + 1 − h.
(27)
Sampling techniques
35
For n = 5 the pivot indexes are simply iL = 2 and iU = 4, and from the corresponding xi values the median and deviations can be obtained. (iv) The median and its dispersion obtained by the “bootstrap” method (Bohn and Wolfe, 1994; Thompson, 1990) may provide a large set of subsets created by multiple (200– 1000 times) random repeating of n original data to calculate more reliably the median and the dispersion. Most surprisingly, it may be just the small random set of the “most non-representative subsamples” which gives the smallest bias of the median from the true value of the total population. Also its ordinary dispersion assessment by Student statistics or non-parametric methods looks much more realistic in reflecting the true variance of samples composition (Macášek, 2000).
3. Field sampling 3.1. Sampling protocol and standard operation procedures The sampling operation should be based on Standard Operating Procedures (SOPs) or protocols especially developed for the specific problem. Design of protocols is very important for the quality assessment and comparativeness. Sample takers should be trained to understand and fulfill the protocols carefully. All deviations from the survey design as documented in the standard operation procedures should be recorded as part of the field sample documentation. 3.2. Field sample preparation and preservation Proper sample preparation and preservation are essential parts of any radionuclide-sampling program. Storage at reduced temperatures (i.e., cooling or freezing) to reduce biological activity may be necessary for some samples. Addition of chemical preservatives for specific radionuclides or media may also be required (Kratochvil et al., 1984; DOE, 1997). Water samples may need filtering and acidification. Samples taken for tritium analysis should be sealed in air-tight glass or HPPE containers to avoid isotope exchange with atmospheric water vapor. Convenient and economical containers for soil samples are the polyethylene bottles with screw caps and wide mouths. Glass containers are fragile and tend to break during transportation. Soil and sediment sample protocols for organic-bound radionuclide analysis require cooling of soil samples at least to 4 ◦ C within the day of collection and during shipping and storage though it is not a practice normally followed for radiochemical analysis. When storage to −20 ◦ C is demanded, resistant plastic sample bottles should be used and be no more than 80% full to allow for expansion when freezing. Specific demands for sample conservation arose for long-term environmental specimen banking, such as “shock freezing” by cooling to below −150 ◦ C by liquid nitrogen and parallel high-pressure mineralization of a portion of the samples (Rossbach et al., 1992).
36
F. Macášek
3.3. Packaging and transporting, traceability Chain-of-custody procedures are necessary for quality assurance of the sampling process. The sample collector is responsible for the care and custody of the samples until they are properly transferred or dispatched. This means that samples are in his/her possession, under constant observation, or secured. All samples should be accompanied by a chain-of-custody record, which documents sample custody transfer from the sampler, often through another person, to the laboratory. Quality assurance of all sampling procedures should be recorded in sufficient detail to enable the finding of the records on space, time, mode of sampling, storage, treatment and transfer of samples. Appropriate photographs and drawings should be provided in the quality assurance report. Field personnel are responsible for maintaining field logbooks with adequate information to relate the sample identifier (sample number) to its location and for recording other information necessary to adequately interpret the results of sample analytical data. Identification and designation of the sample are critical to being able to relate the analytical result to a site location. However, samples are typically collected by one group, and analyzed by a second group. Shipping containers should be sealed (metal containers can present a problem) and include a tamper-indicating seal that will indicate if the container seal has been disturbed. Even if the identity of samples is well declared and communications between the field personnel and laboratory is ensured, an accredited analytical laboratory should not certify the link of the analytical result to a sampled field object and can only account for the uncertainties that occur after sample receipt. For the field aspect is always the responsibility of the sampling body (a sample taker). Global quality assurance means that the field sampling phase was performed or supervised and recorded by the same staff or under the same project manager. 3.4. Sample spiking and isotope labeling Environmental samples are spiked with a known concentration of a target analyte(s) to verify percent recoveries. This procedure can be done during the sampling, but laboratory procedures are more practical and reliable. Spiking is used primarily to check sample matrix interference but can also be used to monitor radionuclide behavior in laboratory performance. Spiking in analysis of radionuclides is applied either as labeling with non-active carrier (reverse isotope dilution analysis) or adding a well measured (usually gamma-emitting) radioisotope (Tölgyessy et al., 1972; Macášek, 2000; Navrátil et al., 1992). It enables a simple assessment of many components of the combined uncertainty of the final result, especially those caused by a varying yield of analyte recovery R. This can be evaluated from the added isotope label amount S0 (in mass or activity units) and isolated amount S as R = S/S0 . A well-known formula is applied to calculate the unknown radionuclide activity A0 in the original sample from the amount A measured in reprocessed sample, A0 = A
S0 . S
(28)
Most customary gamma indicators are 85 Sr for 90 Sr, 243 Am for 241 Am, 232 U for 235,238 U, and 236 Pu for 238,239,240 Pu.
Sampling techniques
37
A further advantage of labeling (spiking) with a gamma radionuclide is to assay the realistic variability of the contaminant distribution in environmental objects (Macášek, 2000). To determine the actual level of natural variance of contamination by radioactive and trace substances is a very difficult task because large samples are to be processed and the measurements are performed very often in the vicinity of detection limits. The chance that the variance of label distribution and isotope exchange rate will lead to indistinguishable massic activities of different species is sufficiently low even for two species. Now, the advantage of radioisotope labeling has been made plain; it is the technique which may enhance laboratory replicate sampling and non-parametric assessment, which normally sharply increases the cost of analysis because of the great number of analyzed subsamples. No doubt this procedure deserves more attention in the future development of radiochemical analysis and radioanalytical methods, especially when combined with rapid monitoring techniques and new statistical methods.
4. Sampling technologies Improper sampling devices and inexperienced staff may cause serious bias in sampling operations. The important considerations for field work should be fulfilling a sampling protocol to avoid disturbance of the samples, especially when volatile, labile or metabolized species are collected. Luckily, in the case of environmental radioactivity, there is no danger of serious contamination of samples by the device used, except if it is carelessly transferred from a zone of heavy contamination to a clean area. Cross-contamination is more probable and undetectable in systematic sampling. However, it can be minimized by executing the sampling plan from the clean or less contaminated sectors. For determining bias resulting from cross-contamination, the field blanks that are certified clean samples of soil or sand are used. Background samples of soils are collected from a reference area offsite of a contaminated section. The “reference background water” from deep wells and glacial aquifers is the basis of any tritium sampling program. For an increased tritium assay, a method blank sample of deionized water is simply used in the case of water samples. Probably, the most comprehensive list of recommended sampling procedures for natural materials is contained in the US Department of Energy manual HASL-300 (DOE, 1997). 4.1. Air and aerosols Except for integral in-line/flow-through detection, there are the following methods for obtaining samples or measurements of airborne radioactivity concentrations: • • • • • •
filtration container filling impaction/impingement adsorption on solids absorption in liquids condensation/dehumidification.
38
F. Macášek
These techniques can be designed as integral and size-selective in nature (Hinds, 1982). In all cases, to calculate the concentrations of radionuclides in air, it is necessary to accurately determine the total volume of the air sampled. The criteria for filter selection are good collection efficiency, high particle loading capacity, low-flow resistance, low cost, high mechanical strength, low-background activity, compressibility, low-ash content, solubility in organic solvents, non-hygroscopicity, temperature stability, and availability in a variety of sizes. Dense cellulose and cellulose-asbestos filters, glass fiber filters, or membrane filters are used. The latter have the advantage of dissolving in organic solvents and then analyzed in a counter, e.g., by liquid scintillation, or they can be burnt. An air mover, such as a vacuum pump, should be used to draw air through the removable filter medium. To sample large volumes of air and obtain total particle collection in a filter battery it is necessary to use an appropriate filter material and an air mover. A suitable air mover should reach a flow rate of 0.5–2 m3 min−1 at pressure drops across the filter ranging from ∼5 to ∼20 kPa, but for small filters (up to 5 cm in diameters) a flow of 5–20 dm3 min−1 is sufficient. To generate complete particle size spectra for the chemical species of interest (radon progeny, Aitken particles, etc.), multichannel (parallel) or multistage (series) screens or disk filters batteries are used. The sizes of particles passing through a stack of filters are calculated from the hydrodynamic parameters of sampling (Cheng et al., 1980; Maher and Laird, 1985). When evacuated containers are used for air sampling, they are opened at the sample location to draw the air into the container. The sample is sealed in the container and removed for analysis or its activity is measured directly in the vessel. To ensure the sample is representative, the flow rate in the sample device inlet must be the same as the flow rate in the system, such as the duct or stack. When the sample line velocity is equal to the system velocity at the sampling point, it is called isokinetic sampling. In other cases discrimination can occur for smaller or larger particles. This occurs because the inertia of the more massive particles prevents them from following an airstream that makes an abrupt directional change. If the velocity of the sample airstream is bigger than the velocity of the system airstream, then the larger particles cannot make the abrupt change and are discriminated against in the sample, i.e., the smaller particles are collected more efficiently, and vice versa. Voluntary discrimination of particles is used in impingers or impactors. Particles are collected on a selected surface as the airstream is sharply deflected. The surface on which the particles are collected must be able to trap the particles and retain them after impaction. This is achieved by coating the collection surface with a thin layer of grease or adhesive or wetting by water or a higher alcohol. High volume cascade inertial impactors operating with filter paper treated with light mineral oil are used to measure the particle size distribution between 0.2 and 20 µm for both indoor and outdoor aerosols (Fuchs, 1964; Theodore and Buonicore, 1976). A dichotomous sampler is capable of separating the “inhalable” particles <10 µm into two fractions collected on membrane filters. By means of virtual impaction, the sampled particles can be separated into two fractions, the fine fraction, <2.5 µm, and the coarse fraction between 2.5 and 10 µm. Condensation or dehumidifier sampling devices employ a “cold trap” to condense water vapor in the sampled atmosphere and provide a liquid sample for further analysis. Liquid nitrogen or a refrigeration unit is utilized to cool the condensation surface, but if sampling
Sampling techniques
39
involves tritium oxide vapor, cooling by dry ice and calcium chloride solutions is applied to avoid freezing of carbon dioxide. The collected water is frequently analyzed using a liquid scintillation spectrometer. Evaluation must include the relative humidity and temperature of the air at the time the sample is taken to determine the concentration of water vapor. Adsorbers, such as activated charcoal, silica gel, alumina and silver zeolite, are commonly used to collect organic vapors and non-reactive gases and vapors. Silica gel is primarily used for humidity removal and tritium oxide vapor sampling. Activated charcoal is used for radiohalogen sampling, but does also trap noble gases, such as radon, xenon (133,135 Xe, representing by the far biggest activity fraction of NPP releases) and krypton (85,85m,87,88 Kr) when cooled with liquid nitrogen (Cimbák et al., 1986). Silver zeolite or silver alumina is used for radioiodine sampling. For stratified sampling of radioiodine species a May-Pack is used (Wilhelm and Schuettelkopf, 1970; Kovach, 1998; Tschiersch et al., 2004). It consists of a stack of columns or cartridges filled by alumina (retains I2 ), alumina impregnated with phenol (for HIO), silver alumina (for CH3 I) and tertiary amine on alumina to collect all the radioiodine forms. 4.2. Water and precipitation Sample-collection equipment and related supplies differ, depending on the chemical nature of the target analyte and on whether samples are collected for surface water or groundwater (Wilson, 1995); surface water samples from a lake or ocean collected with buckets below the water surface are quite different from those collected by screens because of the surface microlayer, enriched by bacterial, organic and heavy metal species. As for gas sampling, isokinetic sampling means that the stream water which approaches and enters a bottle or bag sampler intake does not change in velocity. To keep this condition, the minimal stream velocity should be 0.5–0.7 m s−1 (Lane et al., 2003). Isokinetic hand-held or cable-and-reel samplers are designed to accumulate a representative water sample continuously and isokinetically from a vertical section of a stream moving vertically at a uniform rate during the sampling (depth integrating). For non-isokinetic samplers, hand-held open mouth bottles or weighted-bottle samplers, the water enters the device at a velocity that differs from the ambient stream velocity. Thief samplers are used to collect instantaneous discrete samples. They have been used primarily to collect samples from lakes, reservoirs and some areas of estuaries. The most commonly used thief samplers are the Kemmerer sampler, Van Dorn sampler and double check-valve bailer with bottom-emptying device (DOE, 1997). Water samples (usually 10–50 liters) must be processed as quickly as possible after collection. Their pH is adjusted by diluted nitric or hydrochloric acid to pH 1–3 and 10–50 mg amounts of carriers of radionuclides (if available) may be added to avoid adsorption. The filtration assemblies use standard 0.45 µm membrane filters (cellulose, polycarbonate, and fluorocarbon or polyethersulfone polymers) which also do not interfere with assay of major ions, trace elements, dissolved organic carbon and bacteria. Five-stage cartridges, combining ion exchangers, inorganic sorbents and anchored extraction agents, are convenient for rapid and selective fractionation of fission products in reactor cooling water (Moskvin et al., 1973).
40
F. Macášek
Tap drinking water represents one of the less radioactive media in the artificial environment. Usually, a month composite sample obtained by sending a hundred liters of tap water to a column filled with absorber (a mixed bed ion exchange resin) is prepared for counting in a Marinelli beaker by a gamma spectrometer (Hess et al., 1985). Rainwater and snow samples can be collected for a month in fluorocarbon polymer or polyethylene reagent bottles (2–10 liters) of defined mouth opening (about 30 cm in diameter) and previously containing a few milliliters of diluted nitric or hydrochloric acid. 4.3. Groundwater Groundwater is most commonly collected using either a pump designed specifically for water sampling from monitoring wells, pumps installed in supply wells, or a bailer or other smaller versions of a thief-type sampler. Special multi-layer samplers were designed to collect ground water under natural gradient flow conditions (Margaritz et al., 1989). Suction-lift and positive-displacement pumps are commonly used to collect water samples from monitoring wells. The positive-displacement pumps and submersible pumps are preferred because they do not create a vacuum. Automatic pumping samplers (autosamplers) with fixed-depth intake are sometimes used to collect samples at remote sites from ephemeral, small streams or from urban storm drains. 4.4. Soil Maintaining a constant surface area and depth for soil samples can eliminate problems associated with different depth profiles, unless stratified sampling is projected. Depths of the soil sampling issues from the conceptual site model. The top few centimeters are available for resuspension in air and the top 15 cm is homogenized by agricultural activities like plowing. In most situations the vegetative cover and plant roots are not considered part of the surface soil sample and are removed in the field together with foreign material (concrete, glass, metal). Also soil particles greater than 2 mm are removed by sieving. The sampling record should indicate what was and what was not considered part of the soil sample. Samples are collected from an area of 30 × 30 up to 50 × 50 cm, usually stratified to 0–5 and 5–25 cm depth. In general, quartering reduces the surface soil samples from 15–30 kg down to 10–100 g. Soil sampling equipment is reviewed in Table 6 (EPA, 1991). Homogenization of dry soil in the field is performed by simple techniques, such as cone and quarter or riffle splitter. Wet soil, which forms clumps, is better homogenized for replicate samples in a laboratory. 4.5. Archaeological and geological samples The isotope dating methods require careful sampling strategies. They are best accomplished by an excavation team, archaeologists and/or geochronologists working together to elaborate a strategy for an optimal timing of the site and hence discuss details of samples spanning the stratigraphic range of the deposit. Conventional geochronological methods such as mass-spectrometric uranium series and 40 Ar/39 Ar dating have produced a major advance for all of the uranium series dating methods.
Sampling techniques
41
Table 6 Soil sampling equipment (EPA, 1991) Equipment
Application
Features
Tier Scoop or trowel Bulb planter
Soft surface soil Soft surface soil Soft soil, 0–15 cm
Soil coring device
Soft soil, 0–60 cm
Thin-wall tube sampler
Soft soil, 0–3 m
Split spoon sampler
Soil, to bedrock
Shelby tube sampler
Soft soil, to bedrock
Bucket auger
Soft soil, 7.5 cm–3 m
Hand-operated power auger
Soil, 15 cm–4.5 m
Inexpensive, easy to decontaminate Inexpensive, easy to decontaminate Easy to decontaminate: uniform diameter and sample volume; preserves soil core; can be difficult to decontaminate Preserves soil core; limited depth capability; can be difficult to decontaminate Preserves soil core; easy to decontaminate; can be difficult to remove cores Excellent depth range; preserves soil core; useful for hard soils; conjunction with drill rig for obtaining deep cores Excellent depth range; preserves soil core; tube may be used for shipping core to laboratory; conjunction with drill rig for obtaining deep cores Easy to use; uniform diameter and sample volume; may disrupt and mix soil horizons greater than 15 cm Good depth range; generally used in conjunction with bucket auger; destroys soil core; requires two or more operators; can be difficult to decontaminate
Unlike the older technology of alpha spectrometry, for thermal ionization mass spectrometry (TIMS) only milligrams of sample are needed at high uranium concentrations (tens of ppm), though grams are still required when U concentrations are only a few ppb (Faure, 1986; Edwards et al., 1987). For uranium series dating of calcite it is best to collect 10–20 g of clean, dense, hard material, which will provide enough sample in most cases, even if the isotopic concentrations are very low and the age is quite young. For dating of bone, it is desirable to collect both the dense cortical portions and the more porous elements. If the bone is important from an archaeological point of view, a small sample can be first studied to determine the minimum amount needed, but where use of abundant faunal elements is possible, 5–10 g is generally sufficient. In uranium series dating of teeth, only a portion of the dentin, cement or enamel in a tooth is generally needed when mass-spectrometric methods are used. Alpha-spectrometric U-series dating often requires a whole tooth; hence the non-destructive gamma-spectrometric method of whole sample assay is useful (Rink, 2000). With modern 40 Ar/39 Ar methods, single grains to be dated can be isolated by hand-picking under a binocular microscope. The amount of sample to be collected depends upon the approximate age of the sample and the potassium content of the minerals to be dated (the potassium rich feldspars are preferred). Rarely is the collection of more than several kilograms of sample necessary to obtain a sufficient abundance of datable grains. Accelerator mass spectrometry (AMS) dramatically reduced the sample size needed for 14 C dates, which accounts for milligrams of carboniferous materials like charcoal (Li et al., 1989).
42
F. Macášek
For fission track dating, tiny subsamples are used for microscopic analysis, but their number must be sufficiently large so that a statistically valid number of fission tracks can be counted. When tracks must be counted in multiple grains, the area of sample containing the grains is highly variable (Wagner and Van den Haute, 1992). 4.6. Sediments Accurate coring, e.g., with minimal disturbance of sediments, is of great importance to paleolimnological investigations, especially those aimed at study of vertical profiles of the concentrations of fallout radionuclides. Larger area coring devices are of such size and weight that they can only be used on large oceanographic vessels (Burke, 1968), making them unfit for studies of most lakes and reservoirs. All of the sediment cores can be taken with a sphincter corer mounted in a tripod frame that provides stability to keep the corer in a vertical position while it penetrates. The core barrel is driven into the sediment only by the force of weights mounted on the weight stand. The sediments are always extruded immediately after retrieval. Water retained above the core is siphoned or pipetted off and may be reserved for analyses. Once the surface sediment is removed, the remaining sediment is usually firm and the rest of the core can be sectioned. In most cases, contiguous 1 cm increments are taken from the top of the core down to ∼40 cm. 4.7. Human samples Urine samples are most convenient for inner contamination control of personnel (Stewart et al., 1972). The baby teeth (deciduous teeth) that have fallen out, or have been extracted by dentists, are used for monitoring of 90 Sr, which accumulates since fetus growth. The stratified collective sample should consist of at least 100 teeth for a given birth year. Specific human organs are used when looking for specific assessments (e.g., 129 I). 4.8. Biota and food As a rule, samples of natural biota are taken from vegetation and less conveniently from fauna (e.g., deer). The types of foodstuff sampled are chosen on a site-by-site basis to reflect local and seasonal availability, and to provide information on the main components of diet (milk, meat, potato apples, cereals, etc.) and products most likely to be contaminated by disposals. Samples of leafy vegetables (i.e., cabbage, broccoli, beet and lettuce) are sampled because of the potential deposition of airborne contaminants. Minor foods such as mushrooms and honey, which under certain circumstances are known to accumulate radioactivity, may also be sampled when available (SEPA, 2000; Kammerer et al., 1994). Outdoor reared cows are excellent, seminatural grass and other flora sample collectors from a large area of pasture. Grass, hay and alfalfa are efficient sampling tools for atmospheric contaminants and important radiotoxic nuclides (radiocesium, radioiodine and radiostrontium) which rapidly pass into milk. Therefore the monitoring of milk provides a method of carrying out surveillance of large areas. Weekly or monthly collections are combined (bulked) to provide four quarterly samples for analysis each year, although some analyses may be carried out more frequently, such as weekly 131 I analysis. Annual bulking of some samples is carried
Sampling techniques
43
out for analysis of tritium, 14 C and Cs isotopes. Stratified sampling may take place near farms up to 8 km from the site, and far from farms at 8–16 km from site. Tree-ring samples can be used for assessment of annual radionuclide concentrations in the environment (e.g., 3 H, 14 C, etc.) 4.9. Dust and hot particles The sampling objectives for hot particles (of burned nuclear fuel) may include information about the size and activity distribution of particles in the environment. The detection of individual nuclear fuel particles, their isolation and characterization are complicated and require well-designed sampling and tailored analytical methods, e.g., autoradiography (Pöllänen et al., 1999; Zeissler et al., 1998). However, information on hot particles in the environment is very scarce (Vajda, 2001). Direct identification of the presence of hot particles in the environment requires devices that are able to scan the inhomogeneities of the contamination (Khitrov et al., 1994). The presence of non-volatile nuclides, such as 154 Eu, detected by traditional bulk sample analysis, may give a hint of the presence of hot particles on repeated sample mixing or sample splitting into smaller subsamples and subsequent counting (Bunzl and Tschiersch, 2001). For radiometric in-field measurements, namely alpha activity determination, thin samples less than 100 µm should be collected on aerosol filters. Thin samples of soils and sediments are obtained by spreading ground and milled samples on solid supports. Microparticle classifiers are used to gather a series of size fractions. Sampling of marine samples (seawater, biota, sediment, particulate matter) has recently been described in detail by Povinec (2004).
References Aitken, C.G.G. (1999). Sampling—How big a sample? J. Forensic Sci. 44, 750–760. Alberts, R., Horwitz, W. (1988). Coping with sampling variability in biota: percentiles and other strategies. In: Keith, L.H. (Ed.), Principles of Environmental Sampling. Am. Chem. Soc., Washington, DC, pp. 337–354. American Chemical Society, ACS (1980). Guidelines for Data Acquisition and Data Quality Evaluation in Environmental Chemistry. Anal. Chem. 52, 2242. American National Standards Institute (ANSI) and American Society for Quality Control, ANSI/ASQC (1994). Specifications and Guidelines for Environmental Data Collection and Environmental Technology Programs (E4). Armstrong, M. (1998). Basic Linear Geostatistics. Springer-Verlag, Berlin. ASTM (2002). Standards Related to Environmental Site Characterization. Baule, B., Benedetti-Pichler, A. (1928). Sampling of granular materials. Z. Anal. Chem. 74, 442–456. Bartoš, P., Macášek, F. (1999). Radiochemical analysis and speciation of radiocesium in soils by leaching. Czechoslovak J. Phys. 49 (S1), 641–647. Bartoš, P., Macášek, F. (2002). Desorption techniques for determination of metals mobility in soils. Sci. World J. 2, 573–577. Bohn, L.L., Wolfe, D.A. (1994). The effect of imperfect judgment ranking on properties of procedures based on the ranked-set samples analog of the Mann–Whitney–Wilcoxon statistics. J. Am. Statist. Assoc. 89, 168–176. Borgman, L.E., Quimby, W.F. (1988). Sampling for tests of hypothesis when data are correlated in space and time. In: Keith, L.H. (Ed.), Principles of Environmental Sampling. Am. Chem. Soc., Washington, DC. Bunzl, K. (1997). Probability of detecting hot particles in environmental samples by sample splitting. Analyst 122, 653–656.
44
F. Macášek
Bunzl, K., Tschiersch, J. (2001). Detection of radioactive hot particles in environmental samples using a Marinellibeaker measuring geometry. Radiochim. Acta 89, 599–604. Burke, J.C. (1968). A sediment coring device of 21-cm diameter with sphincter core retainer. Limnol. Oceanogr. 13, 714–718. Byrnes, M.E. (2000). Sampling and Surveying Radiological Environments. CRC Press, Boca Raton. Carlin, R.P., Louis, A.T. (1996). Bayes and Empirical Bayes Methods for Data Analysis. Chapman and Hall, London. Cheng, Y.-S., Keating, J.A., Kanapilly, G.M. (1980). Theory and calibration of a screen-type diffusion battery. J. Aerosol Sci. 11, 549–556. Cimbák, Š., Cechova, A., Grgula, M., Povinec, P., Sivo, A. (1986). Anthropogenic radionuclides 3H, 14C, 85Kr, and 133Xe in the atmosphere around nuclear power reactors. Nucl. Instrum. Methods Phys. Res. B 17, 560–563. Cremers, A., Elsen, A., De Preter, P., Maes, A. (1988). Quantitative analysis of radiocesium in soils. Nature 335, 247. Cressie, N.A. (1993). Statistics for Spatial Data. John Wiley & Sons, New York. Department of Energy, DOE (1997). EML Procedures Manual HASL-300, 28th ed. Environmental Measurements Laboratory, New York, Section 2. Duncan, A.J. (1986). Quality Control and Industrial Statistics, fifth ed. CRC Press, Boca Raton. Eckschlager, K., Danzer, K. (1994). Information Theory in Analytical Chemistry. John Wiley, New York. Edwards, R.L., Chen, J.J., Wasserburg, G.J. (1987). Earth Planet. Sci. Lett. 90, 371. Efron, B. (1981). Nonparametric standard errors and confidence intervals (with discussion). Canad. J. Statist. 9, 139–172. Environmental Protection Agency, EPA (1989). Risk Assessment Guidance for Super-Fund, vol. II, Environmental Evaluation Manual. Report EPA/540/89/001. U.S. Environmental Protection Agency, Washington, DC. Environmental Protection Agency, EPA (1991). EPA Removal Program Representative Sampling Guidance, vol. 1, Soil. EPA 9360.4-10. U.S. Environmental Protection Agency, Washington, DC. Environmental Protection Agency, EPA (1998). EPA Guidance for Quality Assurance Project Plans. EPA QA/G-5. U.S. Environmental Protection Agency, Washington, DC. Environmental Protection Agency, EPA (2000). EPA Guidance for the Data Quality Objectives Process. EPA QA/ G-4. U.S. Environmental Protection Agency, Washington, DC. Environmental Protection Agency, EPA (2001). EPA’s Guide for Choosing a Sampling Design for Environmental Data Collection; http://www.epa.gov/quality/qa_docs.html. Environmental Protection Agency, EPA (2002a). EPA Multi-Agency Radiation Surveys and Site Investigation Manual (MARSSIM), Rev. 1; http://www.epa.gov./radiation/marssim/obtain.htm. Environmental Protection Agency, EPA (2002b). EPA Guidance on Choosing a Sampling Design for Environmental Data Collection for Use in Developing a Quality Assurance Project Plan. EPA QA/G-5S. U.S. Environmental Protection Agency, Washington, DC. Environmental Protection Agency, EPA (2002c). EPA Guidance on Environmental Data Verification and Data Validation. EPA QA/G-8. U.S. Environmental Protection Agency, Washington, DC. Environmental Protection Agency, EPA (2002d). EPA Guidance for Data Quality Assessment: Practical Methods for Data Analysis. EPA QA/G-9. U.S. Environmental Protection Agency, Washington, DC. Faure, G. (1986). Principles of Isotope Geology, second ed. John Wiley, New York. Fuchs, N.A. (1964). The Mechanics of Aerosols. McMillan, New York. Gilbert, R.O. (1987). Statistical Methods for Environmental Pollution Monitoring. John Wiley & Sons, New York. Green, R.H. (1979). Sampling Design and Statistical Methods for Environmental Biologists. John Wiley, New York. Gy, P.M. (1992). Sampling of Heterogeneous and Dynamic Material Systems. Elsevier, Amsterdam. Hassig, N.L., Gilbert, R.O., Wilson, J.E., Pulsipher, B.A. (2004). Visual Sample Plan (VSP) Version 3.0, PNNL14970. Pacific Northwest Laboratory, Richland; http://dqo.pnl.gov/vsp/document.htm. Hess, C.T., Michel, J., Horton, T.R., Prichard, H.M., Coniglio, W.A. (1985). The occurrence of radioactivity in public water supplies in the United States. Health Phys. 48, 553–586. Hinds, W.C. (1982). Aerosol Technology: Properties, Behavior and Measurement of Airborne Particles. Wiley– Interscience, New York. Hlavay, J., Prohaska, T., Weisz, M., Walter, W.W., Stingeder, G.J. (2004). Determination of trace elements bound to soils and sediment fractions. Pure Appl. Chem. 76, 415–442. Horn, J. (1983). Some easy T-statistics. J. Am. Statist. Assoc. 78, 930–934.
Sampling techniques
45
IAEA (1988). Report of the Consultants Meeting on Rapid Instrumental and Separation Methods for Monitoring Radionuclides in Food and Environmental Samples. IAEA/AL/019. IAEA, Vienna. IAEA (1996). International Basic Safety Standards for Protection Against Ionizing Radiation and for the Safety of Radiation Sources. IAEA SS-115. IAEA, Vienna. ICRP (1966). Task Group on Lung Dynamics. Health Phys. 12, 173. ICRU (2000). Sampling for Radionuclides in the Environment. ICRU Report; http://www.icru.org/n_991_3.htm. Isaaks, E.H., Srivastava, R.M. (1989). Introduction to Applied Geostatistics. Oxford Univ. Press, Oxford, UK. ISO (1975). General Principles for Sampling Airborne Radioactive Materials. ISO 2889, Geneva, Switzerland. ISO (1991). Sequential Sampling Plans for Inspection by Variables for Percent Nonconforming (known standard deviation. ISO 8423, Geneva, Switzerland. ISO (1995). Sampling Procedures for Inspection by Attributes. ISO 2859, Geneva, Switzerland. ISO (2000). Acceptance Sampling Plans and Procedures for the Inspection of Bulk Materials. ISO 10725, Geneva, Switzerland. ISO (2001). Statistical Aspects of Sampling From Bulk Materials, Part 2: Sampling of Particulate Materials. ISO 11648-2, Geneva, Switzerland. Journel, A.G., Huijbregts, C.J. (1978). Mining Geostatistics. Academic Press, New York. Kammerer, L., Hiersche, L., Wirth, E. (1994). Uptake of radiocesium by different species of mushrooms. J. Environ. Radioact. 23, 135–150. Khitrov, L.M., Cherkezyan, V.O., Rumyantsev, O.V. (1994). Hot particles after Chernobyl accident. Geochem. Int. 31, 46–55. Kirchhoff, K., Mende, O., Michel, R. (1993). Statistische Aspekte bei der Probenentahme. In: Winter, W., Wicke, A. (Eds.), Environmental Radioactivity, Radioecology, Radiation Effects, vol. 1. Verl. TucV Rheinland, Köln, pp. 461–465. Koprda, V. (1986). Internal Contamination with Radioactive Substances. Veda, Bratislava (in Slovak). Kovach, J.L. (1998). History of radioiodine control. In: Proceedings of the 25th DOE/NRC Nuclear Air Cleaning and Treatment Conference, Minneapolis, pp. 304–319. Kratochvil, B., Wallace, D., Taylor, J.K. (1984). Anal. Chem. 56, 113R. Lane, S.L., Flanagan, S., Wilde, F.D. (2003). Selection of equipment for water sampling. In: Wilde, F.D., Radtke, D., Gibs, J., Iwatsubo, R.T. (Eds.), National Field Manual for the Collection of Water-Quality Data. U.S. Geological Survey, Washington, DC. Li, W.-X., Lundberg, J., Dickin, A.P., Ford, D.C., Schwarcz, H.P., McNutt, R.H., Williams, D. (1989). High precision mass spectrometric dating of speleotherm and implications for paleoclimate studies. Nature 339, 534–536. Macášek, F. (1994). Separation methods for chemical speciation of radionuclides and metals in environmental matrices. J. Radioanal. Nucl. Chem. 183, 5–18. Macášek, F. (1996). Speciation fingerprints of binary mixtures by optimized repeated two-phase separation. J. Radioanal. Nucl. Chem. 208, 5–22. Macášek, F. (2000). Isotope dilution and sampling factors of the quality assurance and TQM of environmental analysis. J. Radioanal. Nucl. Chem. 246, 709–718. Macášek, F., Shaban, I.S. (1998). Cesium speciation in solid matrix and its specific ion adsorption by soils. J. Radioanal. Nucl. Chem. 229, 79–82. Maher, E.F., Laird, N.M. (1985). EM algorithm reconstruction on particle size distributions from diffusion battery data. J. Aerosol Sci. 16, 557–570. Mamuro, T., Matsunami, T., Fujita, A. (1968). Radionuclide fractionation in fallout particles from an air burst. Health Phys. 14, 223–239. Margaritz, M., Wells, M., Amiel, A.J., Ronen, D. (1989). Application of a multi-layer sampler based on the dialysis cell technique for the study of trace metals in ground water. Appl. Geochem. 4, 617–624. Meloun, M., Militký, J. (1994). Statistical Evaluation of Experimental Data. PLUS, Prague (in Czech). Mika, J. (1928). Theoretical notes on sample taking. Z. Anal. Chem. 73, 257–264. Moskvin, L.N., Miroshnikov, V.S., Mel’nikov, V.A., Slutskii, G.K., Leontiev, G.G. (1973). Opredelenie produktov delenia v vode pervogo kontura metodom gruppovogo khromatograficheskogo razdelenia. Atom. Energia 35 (2), 83–88 (in Russian). Myers, J.C. (1997). Geostatistical Error Management: Quantifying Uncertainty for Environmental Sampling and Mapping. Van Nostrand Reinhold, New York.
46
F. Macášek
Navratil, J.D., Greenwell, R.D., Macášek, F. (1997). Leaching or size separation process of soil decontamination— INEEL approach. In: 6th Int. Conf. on Radioactive Waste Management and Environmental Remediation ICEM ’97, Singapore, Oct. 12–16, 1997. Am. Soc. Mech. Eng., New York, pp. 769–772. Navrátil, O., Hála, J., Kopunec, R., Macášek, F., Mikulaj, V., Lešetický, L. (1992). Nuclear Chemistry. Ellis Horwood, New York, p. 279. Patil, G.P., Sinha, A.K., Taillie, C. (1994). In: Patil, G.P., Rao, C.R. (Eds.), Handbook of Statistics, vol. 12, Environmental Statistics. North-Holland, New York, p. 167. Pickering, W.F. (1995). In: Ure, A.M., Davidson, C.M. (Eds.), Chemical Speciation in the Environment. Blackie, London, p. 9. Pitard, F.F. (1993). Pierre Gy’s Sampling Theory and Sampling Practice, second ed. CRC Press, Boca Raton, pp. 25– 44. Pöllänen, R., Ikäheimonen, T.K., Klemola, S., Juhanoja, J. (1999). Identification and analysis of a radioactive particle in a marine sediment sample. J. Environ. Radioact. 45, 149–160. Povinec, P.P. (2004). Developments in analytical technologies for marine radionuclide studies. In: Livingston, H.D. (Ed.), Marine Radioactivity. Elsevier, Amsterdam, pp. 237–294. Remedy, W.R., Woodruff, J.F. (Eds.) (1974). Role of Homogeneity in Powder Sampling, ASTM Special Technical Publication 540. Am. Soc. Testing & Materials, Philadelphia. Rink, W.J. (2000). Beyond 14 C dating: A user’s guide to long range dating methods in archaeology. In: Goldberg, P., Holliday, V.T., Ferring, C.R. (Eds.), Earth Sciences and Archaeology. Plenum, New York, pp. 385–417. Rossbach, M., Schladot, J.D., Ostapczuk, P. (Eds.) (1992). Specimen Banking; Environmental Monitoring and Modern Analytical Approaches. Springer-Verlag, Berlin. Salbu, B., Krekling, T., Oughton, D., Oestby, G., Kashparov, V.A., Brand, T.L., Day, J.P. (1994). Hot particles in accident releases from Chernobyl and Windscale nuclear installations. Analyst 119, 125–130. Sandalls, F.J., Segal, M.G., Victorova, N. (1993). Environmental radioactivity. In: Sampling for Radionuclides in the Environment. ICRU Report; http://www.icru.org/n_991_3.htm. Schilling, E.G. (1982). Acceptance Sampling in Quality Control. Marcel Dekker, New York. Scottish Environment Protection Agency, SEPA (2000). Radioactivity in Food and the Environment, RIFE-5. Food Standards Agency, Sterling, London. Seber, G.A.F., Thompson, S.K. (1994). In: Patil, G.P., Rao, C.R. (Eds.), Handbook of Statistics, vol. 12, Environmental Sampling. Elsevier Science, New York. Smith, F., Kulkarni, S., Myers, L.E., Messner, M.J. (1988). Evaluating and presenting quality assurance sampling data. In: Keith, L.H. (Ed.), Principles of Environmental Sampling. Am. Chem. Soc., Washington, DC, pp. 157– 170. Smith-Briggs, J.L. (1992). Review of Speciation, Solubility of Radionuclides in the Near and Far Field. Report DOEHMIP-RR-92.096. AEA, London. Stein, M.L. (1999). Interpolation of Spatial Data: Some Theory for Kriging. Springer-Verlag, Berlin. Stewart, M.L., Kleine, J.R., Jordan, C.F. (1972). Int. J. Appl. Radiat. Isotopes 23, 387. Tcherkezian, V., Shkinev, V., Khitrov, L., Kolesov, G. (1994). Experimental approach to Chernobyl hot particles. J. Environ. Radioact. 22 (2), 127–139. Theodore, L., Buonicore, A.J. (1976). Industrial Air Pollution Control Equipment. CRC Press, Boca Raton. Thompson, S.K. (1990). Adaptive cluster sampling. J. Am. Statist. Assoc. 85, 1050–1059. Thompson, S.K. (2002). Sampling, second ed. John Wiley & Sons, New York. Tölgyessy, J., Braun, T., Kyrš, M. (1972). Isotope Dilution Analysis. Pergamon, Oxford, UK. Tschiersch, J., Shinonaga, T., Heuberger, H., Bunzl, K., Pliml, A., Dietl, F., Keusch, M. (2004). Unterschiede bei der Ablagerung von Radionukliden auf verschiedene Blattgemüsearten, BMU-2004-635. Bundesministerium für Umwelt, Naturschutz und Reaktorsicherheit, Bonn. Tukey, J.W. (1977). Exploratory Data Analysis. Addison-Wesley, Reading, UK. USACE (2001). Requirements for the Preparation of Sampling and Analysis Plan. EM 200-1-3. USACE, Washington, DC. Ure, M., Thomas, R., Littlejohn, D. (1992). Int. J. Environ. Anal. Chem. 51, 65–84. Vajda, N. (2001). Radioactive Particles in Environment, Review. IAEA Task Group Report BC: 5380.170, Budapest, Hungary. Wagner, G.A., Van den Haute, P. (1992). Fission-Track Dating. Kluwer Academic, Dordrecht.
Sampling techniques
47
Wauters, J., Madruga, M.J., Vidal, M., Cremers, A. (1996). Solid phase speciation of radiocesium in bottom sediments. Sci. Total Environ. 187, 121–130. Wilhelm, J.G., Schuettelkopf, H. (1970). Inorganic absorber materials for trapping of fission product iodine. In: Proceedings of the 11th AEC Air Cleaning Conference, CONF-700816 (vol. 2), pp. 568–578. Wilson, N. (1995). Soil Water and Ground Water Sampling. CRC Press, Boca Raton. Zeissler, C.J., Wight, S.A., Lindstrom, R.M. (1998). Detection and characterization of radioactive particles. Appl. Radiat. Isotopes 49, 1091–1097.
This page intentionally left blank
49
Detection and quantification capabilities in nuclear analytical measurements L.A. Currie∗ National Institute of Standards and Technology, Gaithersburg, MD 20877-8370, USA
1. Introduction The ability to detect or quantify tiny amounts of chemical or radioactive species has long been a fascinating and challenging aspect of basic analytical science. By the mid- and late 20th century, however, detection and quantification capabilities began to assume widespread practical importance, in such applications as: • • • •
protecting health and the environment from noxious substances, monitoring sources and transport of both natural and anthropogenic pollutants, assuring the purity of materials involved in trade, industrial production, and manufacturing, sensing chemical and nuclear precursors of natural disasters, and trace species having potential impacts on global climate, • and detecting characteristic chemicals and nuclides that may signal international treaty violations or terrorist activities. Beclouding the vital theoretical and practical aspects of detection and quantification limits, however, were vague and inconsistent terminology, absence of accepted fundamental metrological definitions, and varying algebraic and numerical expressions for these basic measurement process performance characteristics. The disastrous state of the art, when this author first explored the field, is captured in Figure 1. This figure shows the ordered results of applying eight alternative expressions for the detection limit, found in the analytical and radiochemical literature of the time (1968), to exactly the same radioactivity measurement process (Currie, 1968). The years (decades) following the 1968 publication saw the gradual evolution of a more consistent approach to the matter of detection and quantification limits, particularly in nuclear analytical measurements. It was not until the last decade of the 20th century, however, that coherent sets of international recommendations appeared: for chemistry (International Union of Pure and Applied Chemistry; IUPAC, 1995), for ionizing radiation (International Organization for Standardization; ISO, 2000), and for metrology in general (ISO, 1997). ∗ E-mail address:
[email protected] (Scientist Emeritus)
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11003-2
© 2008 Published by Elsevier B.V.
50
L.A. Currie
Fig. 1. Detection limits: exponential distribution of definitions (1968).
The objective of this chapter is to provide a glimpse of the above evolution (Section 2); and, more importantly, to provide an overview of the fundamental concepts underlying detection and quantification capabilities, their algebraic and numerical expression, and their application to problems of practical import—particularly in the context of nuclear analytical measurements. Central to the realization of these goals are important matters related to assumptions, specification of the measurement process, distribution functions, the blank, and the dichotomy between the characterization of the measurement process (MP) and the reporting of measurement results (MR). Complementing the recommendations for evaluating detection and quantification performance characteristics of the MP, are international guidelines for the evaluation of uncertainty for MR (ISO, 1995). A critical link between the two is the inter-relationship of measurement errors, parameters of distribution functions, and uncertainties of measurement results. The chapter content derives from a number of earlier works of the author, in particular Chapter 1 in the recently issued International Atomic Energy Agency (IAEA) document, Quantifying Uncertainty in Nuclear Analytical Measurements (Currie, 2004c). Two other documents, because of their special significance (1) for the early exposition of detection and quantification concepts in chemistry (Currie, 1968) and (2) for the presentation of the official nomenclature recommendations of the International Union of Pure and Applied Chemistry (IUPAC, 1995), have provided the basis for much of the material in this chapter. The first has
Detection and quantification capabilities in nuclear analytical measurements
51
become a basic, and widely cited text in the disciplines of analytical chemistry and nuclear science; and it served as one of the resources for the development of the IUPAC recommendations. Reference to specific figures and sections in both documents will be critical for the development of several of the topics in this chapter. An introduction to some of the varied perspectives that demonstrated the need for a consistent, international approach to the formulation of detection and quantification limits is given in the historical overview, which follows. 2. Historical perspective 2.1. Early history Although interest in the underlying meaning of detection limits in analytical chemistry is perhaps as old as the discipline, itself, it is instructive to examine its “pre-history”, beginning with the seminal work of Fritz Feigl on the use of spot tests for detecting chemical species. In fact, in the very first volume of the journal Mikrochemie, some 80 years ago, Feigl proposed a system for quantifying this capability as the Erfassungsgrenze (Feigl, 1923). A central figure in the development of detection concepts in the mid-twentieth century was Heinrich Kaiser of the Spectrochemistry Institute, University of Dortmund. Kaiser’s Nachweisgrenze (Kaiser, 1947), which was introduced as a critical value for distinguishing a measured signal from noise (detection decision), has often been used subsequently in its English translation (detection limit) as both the critical level (or decision threshold) for judging the significance of a measurement result (MR) as well as an indication of the underlying detection capability of a measurement process (MP). Decades of confusion followed from this early failure to distinguish the MR critical level, used to control the probability of a false positive (type-1 error), from the MP minimum detectable value, which takes into account also the probability of a false negative (type-2 error). The type-2 error appears to have been “discovered” in 1966 with the introduction of the term Garantiegrenze für Reinheit. Papers in the journal Z. für Analytische Chemie of that year make interesting reading (Ehrlich and Mai, 1966; Kaiser, 1966). The type-2 error is given prominence; and earlier works by Nalimov et al. (1961) and by Roos (1962) are acknowledged. It would seem that this recognition may have evolved in part from vigorous discussions among some of the most noted analytical chemists of the time, in a “kleinen Diskussionskreis” at the Institut für Metalphysik und Reinstmetalle in Dresden. (See the Schlussbemerkung at the very end of Kaiser (1966).) The early state of understanding is summarized in Table 1, where we see that there are two distinctive questions, the one related to the measurement process (detection capability) and the other, to the measurement result (detection decision). Failure to distinguish clearly between the two, leads to the diverse perceptions indicated in the lower part of the table. By the mid-1980’s it was becoming clear that assessment of MP detection capabilities, as opposed to simply assessing the statistical significance of MR (detection decisions), was of increased national and international importance. That is, very serious public health and safety matters were arising regarding the ability to detect dangerous levels of nuclear and chemical contaminants. Closely linked were sociopolitical regulatory issues of major economic consequence. Documentation of international needs and progress on these topics, with sociopolitical as well as technical input, grew out of a Symposium on Detection in Analytical
52
L.A. Currie Table 1 Detection—the concept Dual questions: Q1 How little can be detected? (Measurement Process [MP]) Q2 Has something been detected? (Measurement Result [MR]) Popular responses: • Intuitive (sound experience, non-quantifiable) • Ad hoc (rigid formula, dictum, vote, . . .) • Signal/noise (assuming white noise, Q2 only) • Avoidance (small signals not worth considering) • Hypothesis testing (false positive [α], false negative [β] risks)
Chemistry organized at the request of the Chairlady of the Analytical Chemistry Division of the American Chemical Society. The introductory chapter of the Symposium volume (Currie, 1988) gives a partial compilation of terms, all referring to the same, detection limit concept—ranging from “Identification Limit” to “Limiting Detectable Concentration” to “Limit of Guarantee for Purity”, for example. The most serious terminological trap indicated was the use of the expression “detection limit” (a) by some to indicate the critical value (LC ) of the estimated amount or concentration above which the decision “detected” is made; but (b) by others, to indicate the inherent “true” detection capability (LD ) of the measurement process in question. The first, “signal/noise” school, explicitly recognizes only the false positive (α, type-1 error), which in effect makes the probability of the false negative (β, type-2 error) equal to 50%. The second, “hypothesis testing” school employs independent values for α and β, commonly each equal to 0.05 or perhaps 0.01. In the most extreme case, the same numerical value of the “detection limit” is ˆ when employed, e.g., 3.3σo , where σo is the standard deviation of the estimated net signal (S) its true value (S) is zero. The ratio β/α for the first (single/noise) group, assuming normality, is thus 0.50/0.0005 or one thousand, far in excess of the unit ratio of the second group! Special Nuclear Regulatory Commission and IAEA issues are addressed in Chapter 9 of the Symposium Volume, including an extreme example of the severity of low-level interlaboratory measurement problems (Currie, 1984; Currie and Parr, 1988). Specifically, for the measurement of trace levels of arsenic in an IAEA horse kidney reference material (µg/g level), detection limits quoted for the nuclear analytical methods employed spanned nearly 5 orders of magnitude! 2.2. Recent history; toward international consensus The chronology of events culminating in international harmonization is shown in Table 2. The “flawed IUPAC definition”, exposed by Long and Winefordner (1983), resulted from the early failure to take into account the type-2 error (false negative) in the definition of the detection limit. What led to this situation is that the Kaiser Nachweisgrenze definition (trans. “detection limit”) had been incorporated into the early editions of the IUPAC “Orange Book” or Compendium of Analytical Nomenclature (IUPAC, 1987). As this was, and is, the official international guide for nomenclature in analytical chemistry, it served as the official international chemical basis for expressing/assessing detection limits. The insidious part of the issue
Detection and quantification capabilities in nuclear analytical measurements
53
Table 2 Some milestones toward international harmonization • • • • • • • • • •
“IUPAC Definition” of detection limits flawed (Long and Winefordner, 1983) Lower Limit of Detection: Definition and Elaboration . . . Radiological Effluent (Nuclear Regulatory Commission; Currie, 1984) CODEX requests “Measurement Limit” guidance from ISO and IUPAC (1990) ISO-IUPAC Harmonization Meeting, Washington, DC (1993) IUPAC Recommendations on Detection and Quantification Capabilities (1995) ISO/TC69 Standard 11843-1,2 Capability of Detection (1997) IUPAC Compendium of Analytical Nomenclature (1998) IAEA Manual (TECDOC-1401, 2004; initiated 1998) Quantifying Uncertainty in Nuclear Analytical Measurements ISO Standard 11929-7,8 Bayesian Characteristic Limits (initiated 1999) (Michel, 2000) ISO Standard ISO 11929-1/-4 (2000) Detection limit and decision threshold for ionizing radiation measurements
was that there then existed two sets of meanings for “detection limits”—the official IUPAC meaning, which Kaiser introduced originally as a detection criterion to be applied to measurement results; and the other, as presented in Currie (1968), related to the underlying detection capability of a measurement process. Kaiser’s equivalent term “limit of guarantee for purity”, has scarcely appeared in the analytical literature, and it was gone from the second edition of the IUPAC Compendium. 2.2.1. ISO-IUPAC harmonization efforts: 1993–1998 By 1990 the attention of both ISO and IUPAC was focused on the issue in response to requests from the Committee on Methods of Sampling and Analysis of the Joint FAO/WHO Food Standards Program. The Committee sought guidance on the meaning and assessment of “measurement limits” in connection with methods for the assay of toxic and essential elements in foods involved in international trade (CX/MAS, 1984). They had long been beset with problems of meaning, terminology, and realization of fundamental detection/quantification capabilities— problems known to chemists of the world for decades. In July of 1993, representatives of the ISO and IUPAC came together in Washington, DC to begin efforts to develop a harmonized international chemical–metrological position on detection and quantification capabilities (ISO-IUPAC, 1993). These efforts began with some of the same challenges, since the independent paths taken by the two organizations led initially to contradictory symbols and terminology. They did, however, adopt the same framework (hypothesis testing) for the specification of detection capabilities. (IUPAC, but not ISO, also addressed quantification capabilities.) The efforts were rewarded in just a few years, however, when IUPAC published its official recommendations for the international chemical community (IUPAC, 1995), and ISO published its standard (ISO 11843) for the international metrological community (ISO, 1997). More recently IUPAC has incorporated the 1995 recommendations into its basic nomenclature volume, the Compendium on Analytical Nomenclature (IUPAC, 1998). A discussion of the harmonization efforts appears with a reprinting of the IUPAC Recommendations in Currie (1999).
54
L.A. Currie
2.2.2. A second ISO detection limit series: ISO 11929 Five years after the publication of the IUPAC recommendations, a second series of ISO norms treating detection decisions and detection limits began to appear (ISO, 2000). In the original series (ISO-11843, Capability of Detection, ISO/TC 69) the emphasis was placed on basic metrology, with special attention to calibration-based detection decisions and limits. The later series (ISO-11929, Determination of the Detection Limit and Decision Threshold for Ionizing Radiation Measurements, ISO/TC 85) focuses specifically on radiation measurement, with special attention to the Poisson–normal distribution.1 The final standards published to date (ISO-11929, parts 1 through 4) provide important, detailed guidance for such measurements as pulse counting (with and without the influence of sample treatment), high resolution gamma spectrometry, and the use of linear scale, analog ratemeters. According to Michel (2000), other upcoming parts of ISO-11929 will utilize a Bayesian theory of uncertainty. Happily, all three sets of documents (IUPAC, 1995; ISO, 1997; and ISO, 2000) are harmonized with respect to concepts. Hypothesis testing, with its type-I (α) and type-II (β) errors, serves as the foundation in each case, although there are minor differences in the proposed values for α and β. Unfortunately, however, there were no directed efforts to achieve full consistency between the two earlier publications and the more recent series; and as a result, the terminology used in the ISO-11929 norm differs somewhat from that in the earlier documents. Future efforts by the three groups to include proper cross-referencing and to achieve harmonized terminology and notation would be extremely beneficial to the international metrological community.
3. Detection decisions and detection, quantification capabilities: introduction to the basic concepts and relationships The official recommendations of the International Union of Pure and Applied Chemistry (IUPAC, 1995, 1998) treat Detection Limits (LD ) and Quantification Limits (LQ ) as performance characteristics of the (Chemical) Measurement Process. As such, they demand a fully defined, and controlled measurement process, taking into account such matters as types and levels of interference, and even the data reduction algorithm. “Interference free detection limits” and “Instrument detection limits”, for example, are perfectly valid within their respective domains; but if detection or quantification characteristics are sought for a more complex measurement process, involving for example sampling, analyte separation and purification, and interference and matrix effects, then it is essential that all of these factors be considered in deriving values for LD and LQ for that process. Otherwise the actual performance of the MP (detection, quantification capabilities) may fall far short of the requisite performance. Perhaps the most important purpose of these performance characteristics is for planning— i.e., for the selection or development of MPs to meet specific scientific or regulatory needs. In both the IUPAC and ISO documents mentioned in Section 2, the conceptual basis for the detection limit (minimum detectable amount) is tied to the theory of hypothesis testing, and 1 The assumption of approximate normality is inadequate for extreme low-level counting. As a result, a number of workers have begun to consider the detailed impact of the discrete, Poisson distribution (Hurtgen et al., 2000, and references therein).
Detection and quantification capabilities in nuclear analytical measurements
55
the probabilities of false positives α, and false negatives β. The Quantification Limit (minimum quantifiable amount) is addressed in the IUPAC, but not ISO, documents; where LQ is defined in terms of a specified value for the relative standard deviation (RSD). Default values for α and β are 0.05, each; and the default RSD for Quantification is taken as 10%. As MP Performance Characteristics, both Detection and Quantification Limits are associated with underlying true values of the quantity of interest; they are not associated with any particular outcome or result. The detection decision, on the other hand, is result-specific. It is made by comparing the experimental result with the Critical Value (LC ), which is the minimum significant estimated (observed) value. For presentation of the defining relations, L is used as the generic symbol for the quantity of interest. This is replaced by S when treating net analyte signals; and x, when treating analyte concentrations or amounts. The symbol A is used in the IUPAC recommendations to represent the Sensitivity, or slope of the calibration curve; it is noted that A is not necessarily independent of x, nor even a simple function of x. Subscripts C, D, and Q are used to denote the Critical Value, Detection Limit, and Quantification Limit, respectively. Symbols, and alternative terminology recommendations of IUPAC are: LC [α] (critical value, critical level, minimum significant [observed] signal); LD [α, β] (minimum detectable [true] signal, detection limit); LQ [kQ ] (minimum quantifiable [true] signal, quantification limit).2 3.1. Defining relations; algebraic formulation The basic concepts are given algebraic expression below. Each case requires parameter specification: α and β, for the probabilities of the errors of the first and second kinds (false positives and false negatives); and RSDQ , for the relative standard deviation at the Quantification Limit. Mandatory values of these parameters are not incorporated in the defining expressions; recommended default values are shown in parentheses. Detection Decision (Critical Value) (LC , α = 0.05): Pr(Lˆ > LC | L = 0) α.
(3.1)
Detection Limit (Minimum Detectable Value) (LD , β = 0.05): Pr(Lˆ LC | L = LD ) = β.
(3.2)
Quantification Limit (Minimum Quantifiable Value) (LQ , RSDQ = 0.10): L Q = kQ σ Q
where kQ = 1/RSDQ .
(3.3)
As noted above, LD , LQ , and σQ denote MP “true” parameter values, referring to net signals or concentrations above the background or blank. In practice the L’s may be derived from estimates (or assumptions), such that calculated values themselves will be characterized by uncertainties. Also, for LD and β to be meaningful, LC [α] must be employed for detection 2 Brackets are used in this chapter to denote arguments—e.g., L [α, β], as opposed to products (multiplication). D
56
L.A. Currie
decision making. The relation, β vs LD is known as the “Operating Characteristic” (OC) of the (detection) significance test performed at significance level α; (1 − β) is the “power” of the test. Equation (3.1) is given as an inequality, because not all values of α are possible for discrete distributions, such as the Poisson. The IUPAC recommended default values for α, β, and kQ serve as a common basis for measurement process assessment, but they may be adjusted appropriately in particular applications where detection or quantification needs are more or less stringent. Also addressed in the IUPAC document is the issue of reporting. The recommendation is ˆ together with its unceralways to report both the estimated value of the measured quantity (L) tainty, even when Lˆ < LC results in the decision “not detected”. Otherwise, there is needless information loss, and, of course, the impossibility of averaging a series of results. The practice of quoting upper limits for “non-detects” can be helpful, but this still impairs future uses of the data. Quoting LD as an a posteriori upper limit is not recommended, as this is really an a priori performance characteristic. Although it can be viewed as the “maximum upper limit” in the case of the null hypothesis, LD does not reflect the new information contained in an experimental result. (See also Section 8.2 on the Reporting of Low-Level Data.) 3.2. Graphical representation: detection of earthquake precursors Complementing the algebraic representation of the underlying Detection Limit concept is a graphical representation, shown in Figure 2 as a metaphorical depiction of earthquake detection needs and capabilities—adapted from IUPAC (1995). In the graphical representation the “driving force” in this hypothetical example is the ability to detect specific chemical, physical or biological precursors of earthquakes (e.g., radon, elevation shifts, animal behavior) at levels corresponding to earthquakes of magnitude LR and above. Thus LR is the “requisite limit” or maximum acceptable limit for undetected earthquakes; this is driven, in turn, by a maximum acceptable loss to society. (Derivation of LR values for sociotechnical problems, of course, is far more complex than the subject of this chapter!) The lower part of the figure shows the minimum detectable value for the chemical precursor LD , that must not exceed LR , and its relation to the probability density functions (pdf) at L = 0 and at L = LD together with α and β, and the decision point (Critical Value) LC . The figure has been purposely constructed to illustrate heteroscedasticity—in this case, variance increasing with signal level, and unequal α and β. The point of the latter construct is that, although 0.05 is the recommended default value for these parameters, particular circumstances may dictate more stringent control of the one or the other. Instructive implicit issues in this example are (1) that a major factor governing the detection capability could be the natural variation of the radon background (blank variance) in the environment sampled, and (2) that a calibration factor or function is needed in order to couple the two abscissae in the diagram. In principle, the response of a sensing instrument could be calibrated directly to the Richter scale (earthquake magnitude); alternatively, there could be a two-stage calibration: instrument response–radon concentration, and radon concentration– Richter scale. Among the links to public needs and perceptions that are implied in this diagram, is what might be called “The Zero Myth”. In many cases the lay public believes, given sufficient effort or funding, that a concentration of zero can be detected and/or achieved. Not unlike the
Detection and quantification capabilities in nuclear analytical measurements
57
Fig. 2. Graphical representation of detection needs and capabilities. Top portion shows the requisite limit LR ; bottom shows detection capability LD .
Third Law of Thermodynamics, however, neither is quite reachable, even in concept. A policy of reporting “zero” when Lˆ < LC , yielding the decision “Not Detected”, compounds this lack of understanding. These are issues of considerable national and international importance, especially in the context of environmental monitoring, and related legislative and regulatory issues, where necessarily, and appropriately, many of the policy makers have critical sociopolitical expertise, but not necessarily scientific or technical expertise. The solution to this sociotechnical dilemma is, once again, very careful communication; and, beyond that, mutual understanding and education among the complementary disciplines. One possible solution to the zero dilemma is the substitution of a science-based, non-zero value Lo for characterizing the null hypothesis. Equation (3.1) would then take the form Pr(Lˆ > LC | L = Lo ) α.
(3.4)
Further discussion of the sociotechnical issues and the role of such a non-zero null is given in (Currie, 1988) in connection with discrimination limits. Also relevant is Section 8.2 of this chapter, which addresses bias and information loss in the Reporting of Low-Level Data.
58
L.A. Currie
3.3. Simplified relations3 Under the very special circumstances where the distribution of Lˆ can be taken as Normal, with constant standard deviation σo (homoscedastic) and the default values for parameters α, β, kQ , the foregoing defining relations yield the following expressions: LC = z1−α σo → 1.645σo , LC = t1−α,ν so → 2.132so
(3.5) [4 df],
(3.6)
LD = LC + z1−β σD → 2LC = 3.29σo , LD = δα,β,ν σo ≈ 2t1−α,ν σo → 4.26σo LQ = kQ σQ → 10σo ,
(3.7) [4 df],
(3.8) (3.9)
ˆ under the null where σo , σD , and σQ represent the standard deviation of the estimator L: hypothesis, at the Detection Limit, and at the Quantification Limit, respectively; so2 represents an estimate of σo2 , based on ν degrees of freedom (df); z1−α (or zP ) and t1−α,ν (or tP ,ν ) represent the appropriate 1-sided critical values (or percentage points) of the standard normal variate and Student’s-t, respectively; and δ represents the non-centrality parameter of the noncentral-t distribution. The symbol ν indicates the number of degrees of freedom. For the illustrative case (4 degrees of freedom), the actual value for δ[0.05, 0.05, 4] appearing above (Equation (3.8)) is 4.07.4 The above relations represent the simplest possible case, based on restrictive assumptions; they should in no way be taken, however, as the defining relations for detection and quantification capabilities. Some interesting complications arise when the simplifying assumptions do not apply. These will be discussed below. It should be noted, in the second expressions for LC and LD given above, that although so may be used for a rigorous detection test,5 σo is required to calculate the detection limit. If so is used for this purpose, the calculated detection limit must be viewed as an estimate with an uncertainty derived from that of σ/s. (See Equation (4.8).) Finally, for chemical measurements a fundamental contributing factor to σo , and hence to the detection and quantification performance characteristics, is the variability of the blank. This is introduced formally below, together with the issue of heteroscedasticity. 3 Once the defining relations are transformed into algebraic expressions for particular distributions, it becomes necessary to introduce critical values of those distributions—in particular: zP , tP ,ν , δα,β,ν , and χP2 ,ν . Meaning and
relevance will be indicated as the symbols are introduced; and a table of critical values used in this chapter is given in Appendix A (Table A1). 4 When ν is large, 2t is an excellent approximation for δ. For ν 25, with α = β = 0.05, the approximation is good to 1% or better. For fewer degrees of freedom, a very simple correction factor for 2t is 4ν/(4ν + 1). This takes into account the bias in s, and gives values that are within 1% of δ for ν 5. For the above example where ν = 4, δ would be approximated as 2(2.132)(16/17) which equals 4.013. 5 The use of ts for multiple detection decisions is valid only if a new estimate s is obtained for each decision. In o o the case of a large or unlimited number of null hypothesis tests, it is advantageous to substitute the tolerance interval factor K for Student’s-t. See Currie (1997) for an extended discussion of this issue.
Detection and quantification capabilities in nuclear analytical measurements
59
4. The signal (S-) domain 4.1. Basic equations In many cases the smallest signal SD that can be reliably distinguished from the blank, given the critical level SC , is desired—as in the operation of radiation monitors. Assuming normality ˆ simple expressions can be given for the two quantities involved in and knowledge of σ [S], Signal Detection. Equation (3.5) takes the following form for the Critical Value, SC = z1−α σo → 1.645σo ,
(4.1)
where the expression to the right of the arrow results for α = 0.05. In the Signal Domain, L is equated to the net signal S, which in turn equals E(y) − B. In this expression, y represents the observed (gross) signal or response, and B represents the expectation of the background or blank. The variance of the estimated net signal is then given by ˆ = V [y] + V [B] ˆ → V [B] + V [B] ˆ = Vo . V [S]
(4.2)
The quantity to the right of the arrow is the variance of the estimated net signal Sˆ when its true value (expectation) S is zero—i.e., when E(y) = B. If the variance of Bˆ is negliof the Blank. If B is estimated in a “paired” gible, then σo = σB , the standard deviation √ √ ˆ = VB , then σo = σB 2. Note that σo = σB , and σo = σB 2, are experiment—i.e., V [B] √ ˆ limiting cases. More generally, σo = σB η, where η = 1 + (V [B]/V [B]). Thus, η reflects different numbers of replicates, or, for particle or ion counting, different counting times for ˆ or the “sample” vs blank measurements. Taking n to represent the number of replicates for B, ratio of counting times, we find that η equals (n + 1)/n. The Minimum Detectable Signal SD derives similarly from Equation (3.7), that is, σo2 ,
SD = SC + z1−β σD ,
(4.3)
where σD2 represents the variance of Sˆ when S = SD . For the special case where the variance is constant between S = 0 and S = SD , and α = β = 0.05, the Minimum Detectable Signal √ SD becomes 2SC = 3.29σo = 3.29σB η, or 4.65σB for paired observations. The treatment using an estimated variance, so2 and Student’s-t follows that given above in Section 3.3 (Equations (3.6), (3.8) with LC,D set equal to SC,D ). 4.2. Heteroscedasticity and counting experiments The above result that equates SD to 2SC is not correct if the variance of Sˆ depends on the magnitude of the signal. A case in point is the counting of particles in radiation detectors, or the counting of ions in accelerators or mass spectrometers, where the number of counts accumulated follows the Poisson distribution, for which the variance equals the expected number of counts. Taking B to be the expectation for the number of background counts, σo2 = ηB; and for the normal approximation to the Poisson distribution with α = β = 0.05 and kQ = 10, resulting expressions for SC , SD , and SQ (units of counts) are given by (Currie, 1968) SC = zP (ηB) = 1.645 (ηB), (4.4)
60
L.A. Currie
SD = zP2 + 2zP (ηB) = 2.71 + 3.29 (ηB), 2 2) S Q = kQ /2 1 + 1 + (4ηB/kQ = 50 1 + 1 + (ηB/25) .
(4.5) (4.6)
Asymptotic expressions for SQ , for negligible √ B and for B 1, are simply 100 counts and 10σo , respectively. For SD , we get 3.29 (ηB) for the large B asymptotic expression.6 As shown in Currie (1968), these large B asymptotes for SD and SQ lie within 10% of the complete expressions (Equations (4.5), (4.6)) for (ηB) > 67 and 2500 counts, respectively. Equating the background variance with the expected number of counts B, often leads to an underestimation—i.e., it ignores non-Poisson variance components. When B is large, over 104 counts for example, additional components tend to dominate; and these can and should be represented in VB and Vo . A common practice is to add an extra variance component for background variability, VxB = (ϕxB B)2 , where ϕxB represents the relative standard deviation of the non-Poisson B-variation. VB then becomes B + VxB , and the variance of the null signal becomes 2 B . Vo = ηVB = η B + (ϕxB B)2 = ηB 1 + ϕxB (4.7) In more complicated cases where net signals are estimated in the presence of chromatographic or spectroscopic baselines, or where they must be deconvolved from overlapping peaks, the limiting standard deviations (σo , σD , and σQ ) must be estimated by the same procedures used to calculate the standard deviation of the estimated (net) signal of interest. Examples can be found in Currie (1988) (see also Sections 5 and 6). Other cases, to be considered in later sections, include the treatment of the discrete, non-normal Poisson distribution function for small numbers of counts, and the fitting of empirical variance functions. Regarding the latter, an added quadratic term, (ϕxS S)2 , representing asymptotic constant relative variance, is often appropriate. (See Section 7.1.2.) Accuracy of the (Poisson–normal) expressions for SC , SD Expressions for SC and SD above treat the discrete Poisson distribution with parameter μ (mean and variance) as though it were continuous normal. The question posed here is: How accurate is the Poisson–normal approximation for small values of μB , where μB is equal to B, the expectation of the background counts? For the well-known background case, where η = 1 in Equations (4.4) and (4.5), an exact comparison can be made. To illustrate, taking B as 10.3 counts, the Poisson critical value for the observed number of counts, nC , is found to be 16, necessarily an integer. The corresponding value for the false positive probability α is 0.034, consistent with the inequality requirement of Equation (3.1). SC for the exact Poisson distribution is thus 16 − 10.3 = 5.7 counts. The normal approximation for SC is given by √ Equation (4.4) (with η = 1)—i.e., 1.645 10.3 which equals 5.3 counts. The detection limit, calculated from the exact Poisson distribution, derives from the value of (SD + B) for which the lower tail area of the Poisson distribution equals β (0.05) for n nC or 16 counts. The result is 24.3 counts, so SD = 24.3 −√10.3 or 14.0 counts. The normal approximation, as given by Equation (4.5), is 2.71 + 3.29 10.3 = 13.3 counts. In each case (SC and SD ) we find that the result given by the Poisson–normal approximation lies within 6 Note that S , which equals z σ √η, is the same quantity that appears in the 1-sided Prediction Limit for the C P B + zP σB √η). (For σ 2 estimated as s 2 , the corresponding expression is tP ,ν sB √η.) background, (B
Detection and quantification capabilities in nuclear analytical measurements
61
one count of the exact Poisson result. Similar results obtain for other small values for B. For B = 5.0 counts, for example, exact (Poisson) values of SC and SD equal 4.0 and 10.7 counts; the normal approximation gives 3.7 and 10.1 counts, respectively. For B = 20.0 counts, exact (Poisson) values of SC and SD equal 8.0 and 18.4 counts; the normal approximation gives 7.4 and 17.4 counts, respectively The case of extreme low-level counting, involving (1) the difference between a Poisson variable and a known or presumed mean background, and (2) the difference between two Poisson variables with unknown means, is treated in more detail in Section 7.4. 4.3. Uncertainty of estimated values for SD , SQ , and α For normally distributed variables, the expressions in Equations (3.5)–(3.9) give expected (true) values for LC [α], LD [α, β], and LQ [kQ ]—immediately useful for judging the absolute and comparative capabilities of alternative measurement processes as a function of the background or blank, the type-I, II risks, and the quantifiability parameter. Evaluation of these performance characteristics using experimental data from a particular measurement process, however, transforms these quantities into data-based estimates, with associated uncertainties. Uncertainties will addressed here for two classes of measurements: those for which σˆ is based on replication variance (s 2 ), and those for which it derives from “counting statistics” and the Poisson assumption. Applications requiring a combination can be met through variance (“error”) propagation, taking into account effective degrees of freedom (ISO, 1995). First, we consider the case of replication, where variance is estimated as s 2 (with ν degrees for freedom), SC is given correctly as t1−α,ν so , given and the selected values for ν and α—i.e., an uncertainty in the estimated value is not an issue, provided that each SC decision is based on an independent variance estimate. (The critical value SC in this case is a random variable, but the overall probability that (Sˆ > SC ) remains α, through the application of the critical value of the t-distribution.) SD and SQ , on the other hand, can only be estimated, barring knowledge of σo . Considering the homoscedastic case (constant-σ ), the relative uncertainties of the estimates—SˆD = δα,β,ν so and SˆQ = kQ so —are given by those of estimated standard deviations (so , sQ ). For normal data, bounds for σ 2 , given s 2 and ν, come directly from the chi-square distribution. If the observations are distributed normally, s 2 /σ 2 is distributed as χ 2 /ν. A 95% interval estimate for this ratio is therefore given by (χ 2 /ν)0.025 < s 2 /σ 2 < (χ 2 /ν)0.975 .
(4.8)
A useful approximation, excellent for large ν, for rapidly estimating the standard uncertainty √ of s/σ is 1/ (2ν). Thus, about 50 degrees of freedom are required before the relative standard uncertainty in σˆ is decreased to about 10%. Using Equation (4.8) and taking roots, we get a quite consistent 95% (expanded) uncertainty (ν = 50) of 0.804 < s/σ < 1.20. For more modest, perhaps more typical, degrees of freedom, e.g., ν = 10, this can be a major source √ of SD , SQ uncertainty. (The approximate relative standard uncertainty of σˆ for ν = 10 is 1/ 20, or 22%.) 4.3.1. Confidence intervals (normal, Poisson) for B, given Bobs For the Poisson case, experiment-based estimates for the expressions for SC , SD , and SQ require, at a minimum, estimated values of B. (If non-Poisson variance must also be considered, then ϕxB and the variance function may require additional replication-based variance
62
L.A. Currie
estimates.) In the context of uncertainties in SC , SD , and SQ , this section (4.3.1) has two purposes: (1) to determine the uncertainty interval for B, given the observed number of counts Bobs (integer), as this will directly impact the uncertainty of estimated values for α, SD and SQ ; (2) to evaluate the accuracy of the Poisson–normal approximation for that interval, in comparison to the exact Poisson limits. We first consider the latter approximation. Poisson–normal lower (BL ) and upper (BU ) bounds for the 90% confidence interval for B (background expectation and Poisson mean, variance) are given as the solutions to the 2-sided equality BL + zP BL = Bobs = BU − zP BU . (4.9) The dual (Poisson–normal) solutions to Equation (4.9) are given in columns 4 and 5 of the table below, taking zP = 1.645 and the series of Bobs integer counts in column 1. The exact (Poisson) 90% confidence limits are given in columns 2 and 3. Although the normal confidence limits in the table were calculated from√ the exact solution to Equation (4.9), they can 2 /2 (Cox and Lewis, 1966). The be very closely approximated by Bobs ± z0.95 Bobs + z0.95 2 7 Poisson limits were derived from the χ distribution. Although Bobs values are necessarily integers for the discrete Poisson distribution, the confidence limits for the Poisson parameter are real numbers.
Bobs
BL (Poisson)
BU (Poisson)
BL (normal)
BU (normal)
4 counts 9 counts 25 counts 100 counts
1.37 4.70 17.38 84.01
9.16 15.70 34.97 118.28
1.80 5.24 18.02 84.85
8.91 15.47 34.69 117.86
The first conclusion from this small study is that the Poisson–normal approximation for the 90% confidence interval is remarkably good, differing from the exact Poisson interval by no more than 10%, for as few as 4 counts (Bobs ). The second conclusion is that estimates of SC , SD based on the observed count, Bobs , may be seriously misleading, especially for small numbers of counts observed—e.g., <10. Taking Bobs = 9 counts, for example, we see √ that 2 the 90% interval for the “true” detection limit (η = 2) would range from z0.95 + 2z0.95 9.40 √ 2 to z0.95 + 2z0.95 31.40—i.e., 12.80 to 21.15 counts. The 90% interval for the quantification limit would range from 108.6 to 125.1 counts. (SQ is less susceptible to small count estimation error, because it is relatively insensitive to Vo < 25 counts.) SC is a special matter. Since decisions (detected, not detected) must be made, using √ a defined discriminator, here, SC = 1.645 (ηB), this discriminator cannot be treated with uncertainty. Rather SC will be calculated with an observed Bobs , subject √ to uncertainty. The uncertainty, however, is transferred to α, such that S = 1.645 (ηBobs ) = C √ √ z+ (ηBL ) and z− (ηBU ). Taking the case of Bobs = 25 counts, and the Poisson– √ normal approximation, for example, z− becomes 1.645 (25/34.97) = 1.391, and z+ equals 7 90% Poisson confidence limits are given as follows: B [0.05] equals (χ 2 )/2 with 2B L obs degrees of freedom, 0.05 2 )/2 with 2B and BU [0.95] equals (χ0.95 obs + 2 degrees of freedom.
Detection and quantification capabilities in nuclear analytical measurements
63
√ 1.645 (25/17.38) = 1.973. The corresponding 90% uncertainty interval (normal approximation) for α is (0.024, 0.082). 4.3.2. Use of the normal approximation, not taking into account heteroscedasticity The very common application of the normal approximation to the Poisson distribution, to calculate a confidence interval for the mean (expectation), given Bobs counts, uses the expression BL , BU = Bobs ± zP Bobs . For Bobs = 25 counts, for example, the 90% confidence interval (zP = 1.645) is given by (16.78, 33.22) counts. An approximate tP ,ν -multiplier is sometimes substituted for zP in the equation above, to account for the uncertainty in σ , linked to the uncertainty in the mean (expectation of Bobs , i.e., B). One approach for deriving the equivalent number of degrees of freedom is to equate the rsd (or relative standard uncertainty) of √ the replication-based estimate for σ —i.e., s, with Bobs ). The asymptotic expression for√the former that of the Poisson–normal based estimate ( √ √ is 1/ (2ν); for the latter, 0.5/ B. (0.5 results from the fact that the variance of Bi equals 1/4.) Equating the two expressions for rsd gives ν = 2B, for the degrees of freedom to be associated with Bobs counts. Again taking Bobs = 25 counts, we find tP ,ν = t0.85,50 = 1.676. In the case, the 90% confidence interval is given as (16.62, 33.38) counts. The simple approximations for the confidence interval of the “true” mean (Poisson parameter) give slightly poorer results than that based on the direct Poisson–normal approximation (Equation (4.9)), in part as a result of the incorrect treatment of heteroscedasticity, which is reflected in the positive shift of zP2 /2 in the assigned confidence limits (Cox and Lewis approximation). For α = 0.05, this shift is equivalent to +1.35 counts. 4.3.3. Minimum B for acceptable uncertainties √ in SD and α The fact that the standard uncertainty (u) of B = 1/2 for Poisson variables provides a rapid means to estimate the minimum value for B to limit the uncertainty of α to 0.01, for example, or to limit the relative standard uncertainty (ur ) of SD to an √ √ acceptable value, such as 10% 2 or 5%. Since SC √= zP B and SD = z + 2SC ≈ 2zP B, ur [SC ] and, asymptotically, √ [SD ] equal u [ B]. Finally, inversion to get B[u ], follows from the relations: u [ B] = ur √ r r r √ √ u[ B]/ B = 0.5/ B. Thus B[ur ] = 1/(2ur )2 . Consider first the uncertainty of α, arising from the effect of B uncertainty on SC . As indicated in Section 4.3.1, in the case of SC , the uncertainty is projected onto α via the standard uncertainty of z1−α . Thus, to constrain α to the range 0.04 to 0.06, zP must be constrained to the range of 1.555 to 1.751. This is equivalent to a relative standard uncertainty for the default z0.95 (1.645) of approximately 6.0%. The minimum value of B to achieve that is 1/(2·0.060)2 , or 69.4 counts. To summarize, to control the uncertainty in α, ˆ For SC [α],
−2 Bmin = 2ur [z]
where ur [z] = (z1−α+ − z1−α− )/(2z0.95 ).
(4.10)
The impact √ SD is more direct. For the asymptotic approximation, √ of B uncertainty on SD = 2z (ηB), ur [SD ] = ur [ B], so the minimum B (Bmin ) to guarantee an SD with a
64
L.A. Currie
relative standard uncertainty no greater than 10%, would be 1/(2 · 0.1)2 or 52 or 25 counts. 2 2 For ur = 5%, the corresponding minimum value for B would be 1/(2 · 0.05) √ or 10 or 100 2 counts. Using the improved Poisson–normal approximation for SD , z + 2z (ηB), requires a√small adjustment. Defining the asymptotic result as B o [ur ], the adjusted result is given as √ ( (B o [ur ]) − zP /(2 η ))2 . For ur = 10%, and η = 2, the adjusted minimum B becomes √ ( 25 − 0.582)2 = 19.52 counts. Thus, the asymptotic results for Bmin are somewhat conservative. To summarize, to control the uncertainty in SD , −1 √ 2 For SˆD , Bmin = 2ur [SˆD ] (4.11) − zP /(2 η ) . 4.3.4. Uncertainty in α and use of a threshold While the acceptance of a modest α uncertainty is not a severe drawback for significance testing, it poses a small dilemma in terms of the meaning of SD , which has been defined in terms of specified values for α and β. One approach is to reconsider SD from the perspective of a fixed threshold—i.e, to change from SD [α, β] to SD [ST , β], where ST is the threshold ˆ which SD must exceed with probability (1−β). Besides circumventing the small count SC [B] problem, this approach applies to the vast number of measurement systems that impose a fixed hardware or software threshold to all results. Generally, in fact, such thresholds are set so high that α 1, with the consequence that the detection capability is severely degraded compared to its potential. To illustrate, in a “chemometrics” interlaboratory detection limit exercise in which this author participated, the gas chromatographic instrumental threshold was set so high that the inherent detection capability was degraded by about a factor of ten (Kurtz, 1985). The background noise limited detection limit for the system to detect the pesticide was 4.2 pg, whereas the imposition of a threshold equivalent to 42 pg, increased the LD to 46 pg. As a consequence zP was increased to 34.2, such that the theoretical α-risk was indistinguishable from zero. Balancing the degradation of detection capability, however, are certain benefits from setting a threshold well in excess of the presumed critical value, SC [0.05]—i.e., resistance to unsuspected external factors, such as excess random, or even erratic, variation in background, interference, matrix effects, and sampling errors (Kurtz et al., 1988; McCormack, 1988). In effect, such “real-world” issues underlie the first “intuitive” response indicated in Table 2, related to the impact of expert knowledge and sound experience with the specific measurement problem. The transformation from an (α, β)-based detection limit to an ST -threshold limit can be expressed as ˆ . Vy + V [B] SD [ST , β] = ST + z1−β σD , where σD = (4.12) ˆ is constant (homoscedastic case), then σD = σo ; otherwise, SD can be calculated If σ [S] in an iterative fashion, using the variance function, Vy . For the Poisson–normal case, σD = √ ˆ which equals (SD + Vo ). In this case there is a closed solution (Vy [SD ] + V [B]) 1/2 SD [ST , β] = ST + zP2 /2 + zP zP2 /4 + (ST + Vo ) (4.13) . To illustrate, let us take B = 28.7 counts, η = 1, z√P = z1−β = 1.645, and ST = 20 counts. The conventional SD [α, β] limit is then 2.71+3.29 28.7 = 20.3 counts. The threshold-based
Detection and quantification capabilities in nuclear analytical measurements
65
limit SD [ST , β] is about 60% higher, at 32.9 counts. Barring large unsuspected B fluctuations, the false positive risk and its uncertainty are negligible, with z1−α ≈ 3.7. Now, if SD [ST , β] is estimated by an experimental background observation Bobs (necessarily an integer), the standard uncertainty u[SˆD ] of the estimated detection limit can be shown to be 0.63, or about 2%. (u[SˆD ] follows from the application of variance propagation to Equation (4.13), using 28.7 for the variance of B.) 4.4. A conservative approach to SC ; a lesson from accelerator mass spectrometry Current practice in AMS laboratories sets the stage for an elegant, but potentially over conservative, method for avoiding the uncertain-α problem that arises from the use of an imprecise estimate Bˆ in place of B (expectation) as required in the expression for SC (Equation (4.4)). The problem becomes more severe when Bˆ derives from just a few counts, or when it is influenced by an unrecognized source of non-Poisson variance (Equation (4.7)). The accepted (AMS) solution to the second problem is the use of the “maximum sigma rule” to calculate confidence intervals (Donahue et al., 1987). The standard procedure requires n-replicate counts {Bi } to generate both a Poisson vari and a replication variance estimate (s 2 ). The standard uncertainty (u) is ance estimate (B) √ s). A small problem with the procedure is the fact that the rule leads then taken as max( B, to an inflated uncertainty, on average, for the null case (absence of non-Poisson variance) (Currie, 1994). This occurs because, in the null case, roughly half the time the standard uncertainty will be fixed at the Poisson value, and half the time it will be set to the positive tail of the s-distribution. This positive bias diminishes, of course, when the non-Poisson variance dominates. An added dividend from this procedure is the possibility of testing for ex can be tested with χ 2 /ν. For n = 10, the cess (non-Poisson) variance, as the ratio (s 2 /B) P 2 P = 0.95 critical value for (s /B) would be 16.92/9 = 1.88. (More rigorously, the test would be made with the F -distribution, taking into account the uncertainty of the Poisson variance estimate.8 However, except for very small values B, the denominator-ν will be far larger than the numerator-ν.) The lesson for moderately low-level counting is to include replication as standard practice, and to use t0.95,ν so for making detection decisions (not the maximum-σ rule) if there is any chance of unrecognized non-Poisson variance, or if the uncertainty of the estimate for B is unacceptable. Then, the α-risk no longer uncertain, and the detection limit is defined through the use of the non-central-t. Moderately low-level counting may benefit also from the fact that the t-distribution is robust with respect to modest deviations from normality. It would be very interesting to quantify this robustness for small numbers of counts.9 8 The degrees of freedom for the denominator, σˆ 2 , is expected to be given by equating u [σˆ ] which equals r B √ √B 1/(2 B) with ur [σˆ -replication] which equals 1/ (2ν). Then, ν = 2B, which generally will be significantly larger than the replication-ν. 9 A small, preliminary Monte Carlo study was made using a Poisson distribution having a mean of 2.00 counts— to evaluate the comparative one-sided α performance of (1) the “closest” Poisson critical value (n c = 4 counts √ √ (where each B was based on 5 α = 0.053), (2) z1−α B (where B = 2, the exact Poisson parameter, (3) z1−α B replicates, such that E( Bi ) = 10 counts), and (4) t4,1−α s (where each s was computed from the dispersion of each subset of five Bi replicates). The results were excellent for (1) and (2), but (3) and (4) gave somewhat larger α’s ¯ with
66
L.A. Currie
4.5. SC as a random variable; limits of the Poisson–normal approximation When the background variance is estimated as sB2 , by replication, SC becomes a random vari√ able: tν,1−α so , where so = sB η. Using the more conventional formulation for the (detection) ˆ o is compared to the critical value of t, which is not random. significance test, the ratio S/s The validity of the test depends on the independence of the estimated net signal, and the (root of the) variance estimate—a condition that is guaranteed when s 2 is based on replication. An alternative is to use the observed background counts as an estimate √ for the Poisson variance. Formally, the structure would look the same, but with so = (ηnB ) and ν ≈ 2nB . This version of the random SC test has the advantage that it can be applied successfully with relatively small numbers of background counts—ca. 10 or more, in contrast to the fixed SC approach that requires 70 or more background counts (Section 4.3.3). Unfortunately, in practice the independence requirement is often ignored, leading to a ratio that is not (approximately) distributed as t. For the paired observation detection test, where Sˆ = (y − nB ), use of the √
same background observation to calculate so as (ηnB ) introduces correlation, and a null test ˆ o having significant positive skew, resulting in excessive false positives if a normal ratio S/s (or t-) distribution is assumed. (A solution to the problem is to use independent observations of the blank counts for the estimates of Sˆ and sB .) The false positive problem is illustrated in MARLAP (2004) and references therein; a more complete exposition of the small count random SC issue, including limits for the Poisson–normal approximation, is given in Currie (2008).
5. The concentration (x-) domain Application of the defining equations (3.1)–(3.3) to the concentration domain requires knowledge of the distribution of the concentration estimator xˆ or G(y), where G is the “evaluation operator” that transforms the observed response (or response vector or response matrix, in the case of multicomponent analysis) into an estimated concentration. Here, we treat the simplest case only, where G represents the inverse of the straight line calibration curve. Comments on more general cases, including multicomponent analysis and the treatment of interference and matrix effects, are found in (IUPAC, 1995; Sections 3.3, 3.7.5, and 3.7.6). Thus, for the simplest (single component, straight line) calibration function we obtain ˆ A. ˆ xˆ = (y − B)/
(5.1)
Although this is the simplest case, it nonetheless offers challenges, for example in the application of error propagation for the estimation of the xˆ distribution if ey is non-normal, and for the non-linear parts of the transformation (denominator of Equation (5.1)). In what follows, we consider three cases for concentration (amount) detection decisions and limits: (1) the special category of “direct reading” instrument systems, where the response xi is given directly in units of concentration (or amount). Under these circumstances, except for the non-linear issue, the distinction between the treatment for signal domain and the concentration domain (4) being marginally better than (3). If there were unrecognized extra, non-Poisson variance, method 3 could seriously underestimate the critical value, whereas method 4 would remain unbiased. (See also “α-control”—Section 7.4.2.)
Detection and quantification capabilities in nuclear analytical measurements
67
vanishes, and the expressions in the preceding section apply, with y, ey , B, and S already expressed in concentration units. One finds this, for example, in selected systems for liquid scintillation counting or accelerator mass spectrometry where an {xi } series of independent results is produced from (y, B, A [standard]) observation triplets. ˆ of the calibraThe other two cases differ in the nature of the error in the estimated slope (A) tion curve: (2) eA negligible (A known) or fixed (systematic), and (3) eA random. As before, the development of particular values for the critical level and detection limit requires distributional assumptions, such as normality, which should be tested. More generally, the transformation to the minimum detectable concentration (amount) involves one or more multiplicative (or divisive) factors, each of which may be subject to error. For example, one divides the net signal by a theoretically or experimentally determined sensitivity factor or efficiency to convert a gamma ray counting rate into an emission rate; and factors taking into account neutron monitor responses and irradiation and decay times are generally needed in activation analysis. Collectively, these factors comprise the sensitivity A which relates the net signal to the physical or chemical quantity of interest x, as indicated in Equation (5.1). 5.1. Calibration function known or assumed (eA negligible or fixed) In this case expressions for concentration detection and quantification limits follow immediately from those derived for the signal domain. One simply divides SD or SQ by the known or assumed sensitivity—i.e., xD = SD /A,
(5.2)
xQ = SQ /A.
(5.3)
For normal data with constant, known variance, and α = β, the Minimum Detectable Concentration xD is thus 2SC /A. Taking the default value for α and β, this becomes (3.29σo )/A, where σo is the standard deviation of Sˆ when S = 0. For paired observations this is equivalent to (4.65σB )/A, where σB is the standard deviation of the blank. Since only the numerator in Equation (5.2) is subject to random error, the detection test will still be made using SC . When variance is estimated as s 2 , Student’s-t (central and non-central) must be used as shown in Section 3.3. Similarly, xQ is kQ σQ /A, or 10σQ /A using the default value (kQ = 10) for the quantification parameter. When the assumed value of the sensitivity A is fixed, but biased—as when an independent estimate of the slope from a single calibration operation, or a calibration material or a theoretical estimate having non-negligible error, is repeatedly used—the calculated detection limit will be correspondingly biased. If the relative uncertainty is acceptably small, its bias bounds can be applied to compute bounds for the true detection and quantification limits. ˜ for the blank is not recommended, unless V ˆ VB , Repeated use of a fixed estimate [B] B as that may introduce a systematic error comparable to the Detection Limit, itself. This is of fundamental importance in the common situation, especially in trace analysis, where a relatively precise estimate of the sensitivity is derived from instrument calibration, but where the blank and its variance depend primarily on non-instrumental parts of the measurement process such as sample preparation and even sampling.
68
L.A. Currie
5.2. Calibration function estimated (eA random) When the error in Aˆ is random, then its effect on the distribution of xˆ must be taken into ˆ In what follows, we treat the problem from account, along with random error in y and B. three perspectives: (case I) use of a Taylor expansion to estimate the variance of x, ˆ assuming approximate normality; and (case II) use of a common, quasi-concentration domain treatment that preserves strict normality, but which results in a distribution of “detection limits”—a matter of some confusion that has been widely misunderstood. Brief mention will be made also of the new treatment (case III) that takes into account the exact, non-normal distribution of the xˆ estimator. 5.2.1. Case I: The xˆ distribution; application of the Taylor expansion (IUPAC, 1995; Kurtz, 1985) Here, the defining relations are applied directly to the estimated quantity of interest, the concentration. That is, Pr(xˆ > xC | x = 0) α,
(5.4)
Pr(xˆ xC | x = xD ) = β,
(5.5)
xQ =
(x) kQ σ Q .
(5.6)
For Case I, we treat xˆ as approximately normal, using the Taylor expansion to generate its variance by “error-propagation”.10 The Taylor expansion for x is given as
x(y, ˆ B, A) = x + xy δy + xB δB + xA δA + xBA δB δA + · · · ,
( )
(5.7)
(
)
and second derivatives, and all derivatives, including where the primes represent first the zeroth, are evaluated at the expected values E(y), B, and A. Higher order derivatives (with respect to A, only) exist but have no consequence in calculating the variance of x, ˆ because y, B, and A are taken as normal, and moments of higher order than three are neglected in this approximation. (The third moment for normal variables is zero.) The second moment (variance) of xˆ follows by calculating E(x − x) ˆ 2 using the above expansion in first and second ˆ Aˆ covariance is derivatives. The result, which takes into account possible B, Vx = (1/A2 ) (Vy [x] + VBˆ )J + x 2 VAˆ + 2xVBA , (5.8) 2 ). (NB: ϕ denotes the relative standard deviation of A.) ˆ It should where J is equal to (1 + ϕA A be noted that (1) (Vy [0] + VBˆ ) is, by definition Vo , the variance of the estimated net signal under the null hypothesis, and (2) for small ϕA , such that J ≈ 1, Vxˆ ≈ Vy−yˆ /A2 —a result that will be of interest when we consider Case II. Equation (5.8) provides the basis for deriving expressions for xC,D,Q , under the normal approximation. The basic relations are √ √ xC = z1−α σo(x) = z1−α σo J /A = SC J /A, (5.9) (x)
xD = xC + z1−β σD ,
(5.10)
10 The parenthetical exponent notation (x) in Equation (5.6) is used to indicate an x-domain (concentration-domain)
standard deviation. Also, to reduce notational clutter, the circumflex is omitted from subscripts in the more extended expressions.
Detection and quantification capabilities in nuclear analytical measurements (x)
xQ = k Q σ Q .
69
(5.11) (x)
In Equation (5.9) σo (standard deviation of xˆ when x = 0) is expressed in terms of σo (standard deviation ofSˆ when S = 0) using Equation (5.8) with x set to equal to zero. (Note that σo in this case is (Vy [0] + VBˆ ).) The Detection Decision is made by comparing xˆ with xC , according the defining relation (5.4). That is, √ ˆ Aˆ = S/ ˆ Aˆ is compared with xC = (SC /A) J . xˆ = (y − B)/ (5.12) No longer is the decision made in the signal domain—i.e., we do not compare Sˆ with SC to make the decision; and the x-critical value is, in fact, increased in comparison with the de facto value (SC /A) when A has negligible error. The increase derives from the variance of the denominator estimator Aˆ which affects the dispersion of xˆ even when the true value of x is zero. In the limit of negligible Aˆ variance, the quantity J goes to unity, giving the result xC = SC /A. A conceptually very important distinction between this treatment of the concentration detection decision and the following one (case II) is that J is absent from the latter, where in fact, the decision remains in the signal domain, regardless of the magnitude of ϕA . The absence of J —i.e., J → 1—would make the numerator in Equation (5.8) identical with the variance of (y − y). ˆ The Detection Limit requires the solution of Equation (5.10), which in turn requires knowledge of Vy [x] if the response is heteroscedastic. Although an analytic solution may be possible in certain cases of heteroscedasticity, an iterative solution is generally required. Here, we treat only the homoscedastic case where Vo = Vy + VBˆ is constant. However, even in this case, Vxˆ is not constant, because of the x-dependence shown on the right side of Equation (5.8). Taking this into account, the solution of Equation (5.10) leads to the following expression for the minimum detectable concentration11 : √ √ xD = (2z1−α σo J /A)(K/I ) = (2SC J /A)(K/I ), (5.13) √ √ where K = 1 + z1−α VBA /(Aσo J ) = 1 + r[B, A](σB /σo )(z1−α ϕA )/ J , I = 1 − (z1−α ϕA )2 ,
constraint: I > 0.
When B and A are estimated independently, VBA equals zero, so K = 1; and in the limit of negligible A uncertainty, J , K, and I all equal unity; in that case xD is identical to the expression given in Section 5.1. At the other extreme, xD will be unattainable if the relative standard deviation of Aˆ becomes too large. In particular, the factor I in the denominator goes to zero when ϕA approaches 1/z1−α , or 0.61 using the default value for z1−α . When Vy is estimated as s 2 , SC is replaced with t1−α,ν so in Equation (5.9), and 2SC is replaced with δ1−α,ν σo in Equation (5.13). Also, in this equation, the z’s in K and I are replaced with δ1−α,ν /2. In this case, the ϕA limit for finite xD is 2/δ1−α,ν or 0.49 for the default α (0.05) and 4 degrees of freedom. 11 When V is constant, the change in V with concentration can be both negative and positive. At low concentrations, y xˆ negative correlation between Bˆ and Aˆ tends to decrease it; at higher concentrations the term x 2 VAˆ tends to dominate, causing Vxˆ to increase. Except for bad calibration designs, these effects tend to be small. Variation of Vy [x], however, has the potential to cause significant differences among Vy [0], Vy [xD ], and Vy [xQ ]. (See Section 7.1.2 for discussion of variance function models and non-conventional weighting schemes for Vy [x] estimation.)
70
L.A. Currie
The Quantification Limit is given as the solution of Equation (5.11), using Equation (5.8) with the substitution of xQ for x. For the homoscedastic case the resulting expression is √ 1/2 + gQ , xQ = (kQ σo J )/(A IQ ) 1 + gQ (VBA /Vo ) 2 2 2 2 where gQ = kQ VBA (A IQ ), IQ = 1 − kQ ϕA , and ϕA = σA /A. (5.14) When independent estimates are made for A and B, VBA = 0, and xQ takes the simpler form √ xQ = kQ σo J /(A IQ ). (5.15) As with xD , xQ may be unattainable if the relative standard deviation of Aˆ becomes too large. In particular, the factor in the denominator IQ goes to zero when ϕA approaches 1/kQ , or 0.10 when using the default value for kQ . In the limit of negligible A uncertainty, xQ reverts to the form given in Equation (5.3). For the heteroscedastic (Poisson) case, where Vy = S + B counts, a closed solution obtains when VBA = 0—i.e., in the case of independent estimates for σB and σA . Then, 1/2 . xQ = k 2 /(2AIQ ) 1 + 1 + (4Vo IQ )/(k 2 J ) (5.16) For k = 10, the asymptotic solutions for xQ are then 100/(AIQ ) and (10σo )/(A (J IQ )). The latter is approached more rapidly for large background counts—i.e., when Vo > 100 counts. 5.2.2. Case II: Quasi-concentration domain: calibration-based limits Undoubtedly the most popular, “calibration-based” approach to detection limits is that inspired by the established method for calculating the confidence interval for x through inversion of the linear calibration curve. By extending the equations given for computing lower and upper confidence bounds for x (Natrella, 1963), given an observed response y, Hubaux and Vos (1970) developed expressions for the critical y-value (yC ) and an x-“detection limit” for a particular realization of the linear calibration curve. The beauty of the method is that it is based on strictly linear relationships, so that normality is fully preserved. This comes about because the operations are actually performed in the signal domain, hence the “quasiconcentration” domain designation. The major drawback with the method is that it produces “detection limits” that are random variables, as acknowledged in the original publication of the method (Hubaux and Vos, 1970; p. 850), different results being obtained for each realization of the calibration curve, even though the underlying measurement process is fixed. For that reason, in the following text we label such limits “maximum upper limits”, denoted by symbol xu . The defining relations, which are linked in this case to the estimated (fitted) calibration curve and its limit xu , are Pr(y > yC | x = 0) α,
(5.17)
Pr(y yC | x = xu ) = β.
(5.18)
The conditions are thus equivalent to points on the fitted calibration curve corresponding to x = 0 and x = xu .
Detection and quantification capabilities in nuclear analytical measurements
71
To derive the critical (signal) value SC and maximum (concentration) upper limit xu , we ˆ need the variance of (y − y) ˆ as a function of concentration. Noting that (y − y) ˆ = (y − Bˆ − Ax), and applying variance propagation to the relation that, unlike case I, is linear in the random variables, we obtain V [y − y] ˆ = Vy [x] + VBˆ + x 2 VAˆ + 2xVBA .
(5.19)
If y is normally distributed, and if Bˆ and Aˆ are derived from linear operations, then (y − y) ˆ is also normal, and SC and xu can be calculated using the central and non-central-t distributions, ˆ follows from Equation (5.17) with x = 0, respectively. SC or (yC = SC + B) SC = t1−α,ν so , or z1−α σo if σo is known, (5.20) √ where σo = Vo , and Vo = Vy [0] + VBˆ = VB + VBˆ = VB η. The development thus far is similar to that in Sections 3 and 4, as it must be, because ˆ does not appear of the correspondence between Equations (3.1) and (5.17). Since A (or A) in Equation (5.20), its value does not influence the detection decision, made in the signal domain; so, in contrast to Case I, the α of Case II has no A-induced uncertainty. The maximum (concentration) upper limit (“detection limit”) xu , for the known-σ situation, derives from the signal domain variance of (y − y) ˆ given by Equation (5.19) with x = xu . This limit may then be derived from the solution of Equation (5.21): ˆ u ) − (z1−α σo + B) ˆ (yu − yC ) = (Bˆ + Ax 1/2 . = z1−β Vy [xu ] + VBˆ + xu2 VAˆ + 2xu VBA
(5.21)
The solution to Equation (5.21) for the homoscedastic case (σy = const.), with α = β, is straightforward. The result is ˆ ˆ xu = (2z1−α σo /A)(K/I ) = (2SC /A)(K/I ), ˆ , where K = 1 + r[B, A](σBˆ /σo ) z1−α (σAˆ /A) ˆ 2, I = 1 − z1−α (σAˆ /A) constraint: xu > 0.
(5.22)
The constraint is imposed to suppress rare, but physically meaningless negative values for xu ˆ Unlike Case I, the above solution that could arise from the effects of the random variable A. does not depend on approximate normality of x, ˆ nor does it depend on an unknown sensitivity parameter, since the estimated value Aˆ is used in the solution. If σo is not assumed known, the detection decision is made using (t1−α so ) for SC , and the non-central-t distribution is employed to compute xu . (In that case substitute t1−α for z1−α for approximate correction of K and I .) Notes 1. The distinction between case I and case II is important. The former gives a result xD that represents the fixed detection limit for a fully specified measurement process, albeit with unknown A. The latter gives a variable upper limit xu that is directly calculable, given the ˆ but which is applicable only to the specific calibration experiment that proobserved A, duced it. This means for the measurement process as a whole that there is a distribution of ˆ corresponding to the distribution of A’s. ˆ When A is used in Equation (5.22), limits xu [A]
72
L.A. Currie
the resulting xD can be shown to be approximately equal to the median value of the distriˆ bution of the maximum upper limits. (The mean does not exist for 1/A.) 2. When B and A are estimated from the same calibration data set, the estimates will be negatively correlated with r[B, A] = −x/x ¯ q , xq being the quadratic mean (or root mean square). The ratio K/I may then range from slightly less than one to very much greater, depending on the calibration design and the magnitude of σy . The effect of the factor I in particular, can cause xD (Case I) or xu (Case II) to differ substantially from 2t1−α,ν σo /A. The extreme occurs when the relative standard deviation of Aˆ approaches 1/t1−α,ν (or 1/kQ ); then xD (or xQ ) is unbounded. When B and A are estimated independently, then r[B, A] = 0, and K = 1. If the relative standard deviation of Aˆ is negligible compared to 1/t1−α,ν , then K and I both approach unity, and xD reduces to the form given in Equation (5.2). 3. A note of caution: If the parameters used in Equations (5.20) and (5.22) derive from a calibration operation that fails to encompass the entire measurement process, the resulting values for SC and xu are likely to be much too small. Such would be the case, for example, if the response variance and that of the estimated intercept, based on instrument calibration data alone, were taken as representative of the entire measurement process, which may have major response and blank variations associated with reagents and the sample preparation environment. This makes a strong argument for estimating B from an independent background/blank study, rather than relying heavily on the B-magnitude and uncertainty from the intercept of a linear calibration experiment. 4. An alternative approach for estimating xD , developed by Liteanu and Riˇca (1980), is based on empirical frequencies for the type-II error as a function x. Using a regression– interpolation technique these authors obtain a direct interval estimate for xD corresponding to β = 0.05, given xC . This “frequentometric” technique for estimating detection limits is sample intensive, but it has the advantage that, since it operates directly on the empirical xˆ distribution, it can be free from distributional or calibration shape assumptions, apart from monotonicity. It foreshadows the theoretical approach to the x-distribution ˆ that comprises case III. 5.2.3. Case III: Exact treatment for xC , xD A special transformation procedure developed for the ratio of random variables (Eisenhart and Zelen, 1958) was adapted to treat the exact, non-normal distribution of x, ˆ in an attempt to overcome certain limitations of the foregoing methods. Details of the method and the relatively complicated results for the general case involving covariance are given in Currie (1997). Here, we present results for xC and xD for the simpler case involving independent estimates ˆ can be taken of B and A (i.e., VBA = 0). In the dominant background situation, where V [S] as approximately constant between zero and xD , the results for xC and xD are √ xC = (z1−α σo )/(A I ), where I = 1 − (z1−α ϕA )2 , (5.23) xD = 2xC .
(5.24)
The remarkable, exact relationship of Equation (5.24) holds despite large deviations from normality, as related to ϕA , and also in the presence of covariance (VBA = 0). As a consistency test, a modest Monte Carlo experiment was performed twice to examine the empirical distribution of xˆ for both the null state (x = 0) and at the detection limit (x =
Detection and quantification capabilities in nuclear analytical measurements
73
Fig. 3. Concentration domain detection limits (case III). Null and alternative hypothesis empirical density histograms (top) and distributions (bottom). When x = 0 the distribution is symmetric but kurtotic; increasing concentration leads to increasing asymmetry as shown by the distribution on the right. The expected medians equal the true concentrations (x = 0 and x = xD = 0.52), and xC = 0.26 marks the upper 0.95 tail of the null distribution and the lower 0.05 tail of the alternate (detection limit) distribution.
ˆ A, ˆ based on two sets of 1000 xD ). The empirical distribution of xˆ was obtained as the ratio S/ ˆ ˆ variables, simulating the normal random samples of the net signal (S) and sensitivity (A) “direct reading” (or self-calibrating) type of experiment mentioned at the beginning of this ˆ was purposely pushed section. The relative standard deviation of the non-linear variable (A) close to its limit, with σAˆ /A = 0.33; resulting in strikingly non-normal distributions for x. ˆ The results proved to be consistent with Equation (5.24), despite an (empirical) null distribution that was kurtotic, though symmetric, and a distribution for x = xD that had sub-
74
L.A. Currie
stantial positive skew (Figure 3) (Currie, 1997; Figure 3). Such distributional deviations from normality are expected to give smaller values for xD than the preceding methods. Comparisons with the exact expression for xD (Equations (5.23), (5.24)), including significant nonlinearity from 1/Aˆ (with σAˆ /A = 0.33), confirmed this. Both methods I and II overpredicted xD (xu , expected median): case I (Equation (5.13)), overpredicted xD by 26%, whereas for case II (Equation (5.22)) the overprediction (xu ) was about 20%. With more precise calibration (σAˆ /A = 0.10), the overpredictions were trivial: 1.9% for case I, 1.4% for case II. 6. Multiple detection decisions; multicomponent and multivariate detection limits Emphasis in this section is placed on concepts underlying multiple detection decisions and the corresponding detection limits, such as arise in the interpretation of multicomponent chemical chromatograms and nuclear and optical spectra. The simplest case, which might be described as multi-univariate detection, relates to the detection of a series of non-overlapping gamma ray peaks, for example, or a series of independent monitoring tests for a specific radionuclide or chemical pollutant over time. When the null hypothesis dominates, such multiple detection tests can lead to an overall false positive probability far in excess of the default 5% (Section 6.2). When spectral peaks overlap, a matrix solution is required, where the critical levels and detection limits for each component depend on the levels of other components—this is the case where pure component detection limits can be seriously degraded as a result of multicomponent interference. Finally, the problem of multivariate detection is addressed where a measured component is characterized by a multiple variable (multivariate) pattern or fingerprint (Section 6.3). Only the basic concepts, and approaches for normal variates with known or assumed covariance matrices will be presented in any detail. 6.1. Multicomponent detection When a sensing instrument responds simultaneously to several analytes, one is faced with the problem of multicomponent detection and analysis. This is a very important situation in chemical analysis, having many facets and a large literature, including such topics as “errorsin-variables-regression” and “multivariate calibration”; but only a brief descriptive outline will be offered here. For the simplest case, where blanks and sensitivities are known and signals additive, S can be written asthe summation of responses of the individual chemical Sij = Aij xj , where the summation index j is the chemical components—i.e., Si = component index, and i, a time index (chromatography, decay curves), or an energy or mass index (optical, mass spectrometry). In order to obtain a solution, S must be a vector with at least as many elements Si as there are unknown chemical components. Two approaches are common: (1) When the “peak-baseline” situation obtains, as in certain spectroscopies and chromatography, for each non-overlapping peak, the sum Ax can be partitioned into a one component “peak” and a smooth (constant, linear) baseline composed of all other (interfering) components. This is analogous to the simple “signal–background” problem, such that for each isolated peak, it can be treated as a pseudo one component problem. (2) In the absence of such a partition, the full matrix equation, S = Ax, must be treated, with xkC and xkD computed for component k, given the complete sensitivity matrix A and concentrations of all
Detection and quantification capabilities in nuclear analytical measurements
75
other (interfering) components. These quantities can be calculated by iteratively computing, from the Weighted Least Squares covariance matrix, the variance of component k as a function of its concentration, keeping all interfering components constant, and using the defining equations (3.1) and (3.2), or their normal forms, Equations (3.5) and (3.7). Further discussion can be found in Currie (1988) and references therein. An additional topic of some importance for multicomponent analysis is the development of optimal designs and measurement strategies for minimizing multicomponent detection limits. A key element for many of these approaches is the selection of measurement variable values (selected sampling times, optical wavelengths, ion masses, etc.) that produce a sensitivity matrix A satisfying certain optimality criteria. Pioneering work in this field was done by Kaiser (1966, 1972); a review of later advances is given by Massart et al. (1988). 6.2. Multiple and collective detection decisions When several independent null or alternative hypothesis tests are made, error probabilities necessarily magnify. This is a common occurrence in spectrometry or chromatography at extremely low levels where null hypothesis dominance is the rule. A striking illustration is a quality assurance test spectrum for gamma ray spectrometry, distributed by the International Atomic Energy Agency, where an international comparison resulted in up to 23 false positives and 22 false negatives from a single (simulated) gamma ray spectrum (Parr et al., 1979; Currie, 1985a). In what follows we present a brief introduction into two ways for controlling these hypothesis testing errors when treating the limited, collective decision problem, such as occurs in the analysis of a single nuclear, optical or mass spectrum, a single liquid or gas chromatogram, or a limited space–time field study. Also included is a treatment of a logical (and essential) extension to the case of unlimited detection decisions. In order to convey basic concepts and dilemmas in this brief treatment of the multiple decision problem, we restrict discussion in this section to known (or assumed) distribution functions, including as a consequence, known σ . Realistic examples, that come close to meeting this restriction, are drawn from counting experiments where the Poisson distribution is nominally satisfied. There is a growing literature on the topics of multiple and multivariate detection decisions and limits, and the closely related topic of simultaneous inference, both in the statistical and chemometrics literature (Miller, 1981; Gibbons, 1994; Davis, 1994). 6.2.1. Setting the stage As a focus for developing the topic, we present in Figure 4 a portion of a measured spectrum of the NIST Mixed Emission Source Gamma Ray Standard Reference Material, SRM 42156. The spectrum extract shows two of the energy-scale calibration peaks at 1332 keV (60 Co) and 1836 keV (88 Y, 41.7% of the 60 Co peak height), as well as a small impurity peak at 1460 keV (40 K, 1.6% of the 60 Co peak height). Were it not for the impurity, the region between the two major peaks should have nothing other that the Compton baseline and detector background, the sum of which amounts to about 265 counts per channel in the region of the 40 K peak, or 1.3% of the 60 Co peak height. (See also Appendix A.4.) There are several relevant lessons to be drawn from this snapshot of a gamma ray spectrum. First, the detection limit (counts per peak) is determined by the counting statistics of the baseline and the peak estimation algorithm, when no impurity peak is present. The baseline is the
76
L.A. Currie
Fig. 4. Portion of a spectrum of the NIST Mixed Emission Standard gamma ray spectrum, SRM 4215f, showing energy calibration peaks at 1332 keV (60 Co) and 1836 keV (88 Y), together with an impurity peak at 1460 keV (40 K). The counting data should be approximately Poisson, so the square root transformed data shown here should have variance ≈0.25.
summation of the Compton counts and the detector background counts, possibly perturbed by the presence of the counting sample. The baseline counts derive from single and multiple scattering of “sample” (measurand) and interference gamma rays in the sample matrix and detector system. Added to the 265 baseline counts per channel (at 1460 keV) is the 40 K impurity peak, which probably has dual sources: background potassium in the detector system, and a chemical impurity in the master solution for the standard, possibly derived from leaching of potassium from the glass container. (40 K, which has a half-life of 1.2 × 109 years, is found in all naturally occurring potassium.) The magnitude and variability of each of these three types of “blank” must be considered when deriving critical levels and detection limits. Estimation algorithm dependence is simply demonstrated by considering alternatives for the response:
Detection and quantification capabilities in nuclear analytical measurements
77
peak height, peak area (given fraction or full integration), or fitted peak parameter(s). To further illustrate, we take one of the simplest “square wave” (trapezoidal) estimators, which uses as a weighting function the symmetric 12 channel filter (−1 −1 −1 1 1 1 1 1 1 −1 −1 −1). This leads to an estimate of the net peak area as the summation of the central 6 channels corrected with baseline counts from the 6 channels from the left and right regions adjacent to the peak—here totaling 1589 counts. (The filter width obviously must adapt to the peak width, which is often a slow and smoothly changing function of gamma ray energy.) In the absence of the 40 K peak and using this filter, the standard deviation of the estimated net signal (null hypothesis) would be approximately (2 · 1589)1/2 = 56.37 counts (σo ). Assuming approximate normality for the Poisson distribution, and taking α = β = 0.05, gives a critical level SC of 1.645σo or 92.7 net peak counts; SD would be twice that, or 185.4 counts, which is just 0.34% of the 60 Co peak area estimated by the same “3/6/3” filter algorithm. Since Sˆ for the 40 K peak (839 counts) exceeds S , it is considered “detected” above the (Poisson) baseline. C (One small assumption has been made in the foregoing calculation; namely, that the integrated number of baseline counts (1589) is a good approximation for its expected value which, in turn, equals the Poisson variance; see Section 4.3.) 6.2.2. Multiple, independent detection decisions In many cases, in both the research and the regulatory environments, it is necessary to make a number of detection decisions simultaneously. An example comes from the Technical Specifications required by the US Nuclear Regulatory Commission for environmental radioactivity in the vicinity of operating nuclear power reactors. These specifications mandate the capability to detect prescribed levels of, for example, up to 9 fission and activation products in the same gamma ray spectrum.12 Let us suppose that the 9 detection decisions are based on an “empty” spectral region, such as depicted in the baseline portion of Figure 4. For k = 9 independent Ho tests with the error of the first kind set equal to α , the binomial distribution gives the probability of r false positives as (α )r (1 − α )k−r . The collective risk α of one or more false positives is thus α = 1 − Pr(r = 0) = 1 − (1 − α )k , or approximately kα , if α is sufficiently small. Thus, for k = 9, α must be adjusted to 1 − (1 − α)1/k = 1 − (0.95)1/9 = 0.00568 to attain a collective risk α of 0.05. For normal data, that results in an expansion of zC from 1.645 to 2.531. (For the 40 K example in the preceding paragraph, the critical level would be increased from 92.7 counts per peak to 2.531 · 56.37 = 142.7 counts, and the detection limit would increase from 185.4 to (2.531 + 1.645) · 56.37 = 4.176 · 56.37 = 235.4 counts per integrated peak.) If α is not adjusted, the collective false positive risk α in an empty spectrum then becomes 1 − (0.95)9 or 0.37; and the expected number of false positives for a nine-fold independent test scenario is then 9 · (0.05) or 0.45. A 90% two-sided interval for the number of false positives would range from zero to two counts, as given by the binomial distribution with N = 9 and p = 0.05. A small extension to this example would be to inquire not just whether a specific set of (9) peaks is present, but rather whether any gamma ray peaks, besides 40 K, are present in the apparently empty region between the 60 Co and 88 Y gamma ray peaks. This can be addressed by determining the maximum number of independent null tests that can be made using the 12 An early draft for detection capabilities for environmental sample analysis for radionuclides in water lists requisite
“lower limits of detection” ranging between 1 and 30 pCi/L for Mn-54, Fe-59, Co-58,60, Zn-65, Zr-Nb-95, I-131, Cs-134, Cs-137, and Ba-La-140 (U.S. Nuclear Regulatory Commission, 1982).
78
L.A. Currie
selected algorithm. Taking a search region of 216 channels (1.40–1.60 MeV) in the baseline portion of Figure 4, and using the 3/6/3 algorithm, we can perform a total of 17 independent tests of the null hypothesis. To control the false positive error rate to 0.05 would then require α = 0.0030, or zC = 2.745. That means an increase in detection limit by a factor of (2.745 + 1.645)/3.29 = 1.33 over the single peak test value. Life, in fact, may be a bit more complicated: the number of potential gamma rays in the search region from nuclides with half-lives of an hour or more is approximately 54; so totally independent detection testing is impossible. What this means, in such cases of intensive peak searching, is that the multiple independent test technique must be replaced by one that takes into account the full covariance matrix. This issue will be touched upon at the end of this section. 6.2.3. Multiple false negatives (quasi-Bayesian approach) Thus far, we have used the foreknowledge concerning the number of null states likely present to adjust α. Although an analogous treatment of multiple false negatives was not found in the chemical literature, it is clear that this situation also must be controlled if we are to achieve an overall error rate of, say, 0.05. If “subjective” probabilistic information were available on relative HA abundance, we could give a Bayesian formulation for the detection limit. The prefix “quasi-” is used here to illustrate the extension of the treatment of Ho state multiplicity to include that of HA state multiplicity where the relative abundance of each is known (or assumed) in advance. A case in point is the IAEA quality assurance gamma ray spectrum mentioned earlier. Here, it was known (by those who created the data) that the 2048 channel spectrum was empty (baseline only) except for 22 peaks. Given such knowledge, and defining the search window as 12 channels, a zeroth order method of analysis would comprise (2048 − 12)/12 ≈ 170 tests, of which k = 148 would be applied to the null state and m = 22 to the alternate state. (See the following sub-section for commentary on so large a number of tests of the null hypothesis.) To preserve an overall (expected) error rate of 0.05 for signals lying outside the deadzone between zero and SD , we would need to set α to 0.00034 and β to 0.0023. The corresponding values of z1−α and z1−β are 3.39 and 2.83, resulting in nearly a doubling of the detection limit (6.22σo vs 3.29σo )—the price of adequately limiting the probability of both false positives and negatives.13 The coefficient 2.83 is derived for the worst case, where the HA state is m-fold degenerate, all 22 peaks having the same amplitude. In effect we are developing an operating characteristic for the doubly degenerate case (Ho : k-fold; HA : m-fold). The inequality (β < 0.05) applies to all other (HA ) cases: this is clear when one or more of the 22 peaks actually present exceeds the detection limit SD ; if one or more is less than SD , then m has been overestimated, and the probability of one or more false negatives for peaks of magnitude SD must again be <0.05. 13 The example has, of course, been oversimplified, for the sake of illustration. The “170 tests” that comprise the
zeroth order analysis are taken to be independent, applied to contiguous but non-overlapping spectral (channel) windows. Furthermore, each spectral peak (actually or potentially present) is considered to lie within just one of the 170 search windows. Thus, overlapping peaks and overlapping search windows are, by definition, excluded. More realistic and more complex search strategies introduce interesting covariance issues, but they will not be treated here. Beyond mathematical statistical issues, and perhaps even more important, is the specialized knowledge that would be employed by the experienced spectroscopist to guard against spurious conclusions by utilizing information related to peak shape, possible contaminants, matrix effects, instrument artifacts, and sample history.
Detection and quantification capabilities in nuclear analytical measurements
79
6.2.4. Extended detection decisions There are two situations that limit the utility of the above α-adjustment approach for multiple detection decisions: (1) when decisions must be continually made during a low-level monitoring exercise of indefinite extension—i.e., where k may not be known in advance; and (2) where k is so large, and the corresponding multiple decision α so small, that it would make the resultant critical levels and detection limits unreasonably large, putting an excessive burden on the measurement process capabilities, and at the same time severely straining the validity of assumptions concerning the tails of the distributions employed.14 An alternative, when k is large or indefinite, but where Ho nevertheless dominates, is to treat α as the population parameter p (binomial distribution) giving the asymptotic proportion of null hypothesis decisions that will be false positives. Having fixed α in this manner, one can specify the corresponding detection limit in the usual way by setting the single decision β to the desired level, assuming that HA is rare. Using this approach, it is interesting to ask first, for a selected value of k, the likely range of false positives; and second, what value of k corresponds to the “break-even” point where the k-multiple decision α equals the extended decision α . To illustrate, let us assume that the (1-sided) extended decision α is taken as 0.005. That means that in the long run 0.5% of the null tests will be seen as false positives. For a run of length 500, for example, using the binomial distribution, we find that the expected number of false positives would be 2.5, with a 90% interval extending from 0 to 5 false positives, the most probable number being 2 which has a 25.7% chance of occurring. The occurrence of one or more false positives, on the other hand, is almost a sure thing, with probability 1 − (1 − 0.005)500 = 0.918. Regarding the break-even k, it is given as the solution to: 0.05 = 1 − (1 − 0.005)k , or 10.23, where the desired multiple decision α is 0.05. The nearest integer 10 is conveniently given by the approximate relation, α ≈ α/k. 6.3. Multivariate detection decisions and limits Another way to view the multiple null hypothesis test is to treat it as a single collective (vector) decision based on a k-variate null hypothesis. Referring again to the nine-fold search for specific environmental radionuclides, the collective questions become: (1) for a given spectrum or appropriate set of observations for the 9 radionuclides considered as a 9-variate vector, does it differ significantly from the corresponding 9-variate background vector? If normality can be assumed, the vector difference can be tested against the critical value of the F -distribution, or, if the covariance matrix is assumed known, against the critical value of the χ 2 distribution with 9 degrees of freedom. The quantity for testing is the sum of squares of the nine standardized differences, where each term in the sum is equivalent to the square of the respective (S/σo )i that was used for the single peak tests by comparison to (SC /σo )i . Adjustment of the critical value for the number of variates is now automatic through the specification of the number of degrees of freedom for χ 2 . There is a special point regarding the comparison with 14 Caution should be exercised in over-reliance on the central limit theorem to guarantee “normal” tail probabilities.
With finite tailed distributions, for example, distributions of variate sums (or averages) will always remain finite, except in the asymptotic limit. If extreme tail areas are to be used, it is mandatory that their validity be demonstrated for the particular measurement process in question.
80
L.A. Currie
the multiple independent (multi-univariate) detection test scenario: namely, that the individual terms in the sum bear a 1-to-1 comparison with the squares of the individual detection test terms. However, this applies to the independent case only. If the covariance matrix is not diagonal, it is no longer true; and the multiple independent decision approach is not valid. The collective multivariate test remains valid, however, provided that the full covariance matrix is used. The quantity to be tested against the critical value of χ 2 (ν) is then S Σ −1 S, where S denotes the net signal vector and Σ denotes the covariance matrix. Because of the squared, central relation, sign information for the central χ 2 distribution is lost; the univariate z1−α or 2 2 . Thus, for the multivariate null hypothesis the z0.95 emerges as a multivariate χ1−2α or χ0.9 critical value is defined by Pr(χ 2 χC2 ) = 1 − 2α. 2 equals 14.68. An infinite set of values for the For the 9-variate case, taking α = 0.05, χ0.9 individual standardized estimated net signals could give rise to this value, but it is interesting to consider two limiting combinations: one where one variate is totally responsible, and the other √ where all are equal. In the first case, the equivalent critical value for Si would be (σo )i 14.68, or 3.83(σo )i , somewhat greater that the critical value of 2.53σo found for the 9-fold univariate, independent test. The second limiting case would be where the sum includes nine√ equal squared terms. Each root mean square standardized net signal is thereby equal to √ χC / ν—i.e., 3.83/ 9 or 1.28, somewhat less than the critical value for the multi-univariate test. To calculate the multivariate detection limit, we must use the non-central chi-squared distribution χ 2 , somewhat analogous to the earlier use non-central t-distribution. Unlike the central (χ 2 ) distribution, sign information is not lost from the non-central (χ 2 ) distribution. The non-centrality parameter λ is defined as S Σ −1 S, which if Σ is diagonal is simply the sum of squares of the true standardized net signals at the detection limit, such that Pr(χ 2 χC2 ) = β. For the present 9-variate case, assuming that both the central and non-central sums of squares share the same covariance matrix, we find that at the detection limit, the non-centrality parameter λD has the value 19.9. The value 19.9 used in this example illustrates the use of the Patnaik approximation.15 The multivariate detection limit is thus given as the solution to the equation ⎛ ⎞ S1 S ⎜ 2⎟ ⎟ (S1 S2 . . . S9 )Σ −1 ⎜ (6.1) ⎝ ... ⎠ = λD = 19.9. S9 As noted above, if Σ is diagonal (no correlation) this equation reduces to the simple sum of squares, (d12 +d22 +· · ·+d92 ) = 19.9, where di = Si /σi . Once again, we may compare the two limiting cases with the multi-univariate signal detection limit of 4.18σi . The extreme case, all 15 Unfortunately, tables for the non-central statistics (t , χ 2 , F ) are not found in every statistics or chemometrics
textbook; nor is the corresponding software particularly abundant. The latter may be obtained, however, from the NIST Statistical Engineering Division (Reeve, 1986); also, high resolution charts have been published by Pearson and Hartley (1951), and good approximations have been developed by Patnaik (1949) that utilize the respective central distributions. Patnaik’s χ 2 approximation, utilized for this example, is based on the fact that χ 2 /C has approximately a χ 2 distribution with m degrees of freedom. The parameters C and m are given as C = (ν + 2λ)/ (ν + λ) and m = (ν + λ)2 /(ν + 2λ). By way of comparison, rigorous calculation of λD for use in Equation (6.1) gives 20.31 rather than 19.9.
Detection and quantification capabilities in nuclear analytical measurements
81
√ d’s zero save one dj , gives√(SD )j = σj 19.9 or 4.46σj . The uniform case, all d’s equal, gives all (SD )i ’s equal to σi 19.9/9 or 1.49σi . Because of the λD constraint in Equation (6.1), the nine true net signals at the detection limit lie on an eight-dimensional hypersurface which spans the nine-fold positive sector originating at the blank. The possibly surprising result is that there is not a single vector representing the detection limit but an eight-fold infinity of vectors with endpoints at the detection hypersurface. Particular endpoints or patterns (“fingerprints”) may serve to identify specific elements or compounds or sources giving rise to the spectral or chromatographic data. To better illustrate the geometric concepts given above, the simplest situations where ν = 1 2 and 2, respectively, are shown in Figure 5. For the univariate case χ0.90 = 2.71, and λD
2 corresponding to χ0.05 equals 10.8. These correspond to the well known single decision re√ √ sult of SC = σo 2.71 = 1.645σo and SD = σo 10.8 = 3.29σo . The detection limit is 2 = 4.605 corresponding to a stanzero-dimensional. For the uncorrelated bivariate case, χ0.90 dardized vector of length 2.15 which defines the critical region for the detection decision.
2 , equals 13.02 which is equivalent to The non-centrality parameter λD , corresponding to χ0.05 a standardized vector of length 3.61. The Patnaik approximation, which is poorer for fewer degrees of freedom, in this case gives λD ≈ 12.1. This defines the one-dimensional detection limit, which is represented by the arc subtending the doubly positive quadrant in Figure 5. The two limiting combinations, where the arc intersects one or the other of the axes, and where equal-length (rms) vectors intersect the arc, give SD /σo vector√extremes of 1: (3.61, 0) or (0, 3.61) and 2: (2.55, 2.55), respectively, where 2.55 = 3.61/ 2. For comparison, the multi-decision standardized critical value and detection limit, with k = 2 and α = 0.0253, are 1.95 and 3.60, respectively. The multivariate blank and estimated covariance Because of space–time limitations, the development in this section was restricted to the known variance, covariance matrix situation, and emphasis was placed on the concepts appropriate to multiple and multivariate detection decisions and limits. Also, translations into specific formulas or values for critical levels and detection limits were limited to normal variates. Here, we briefly introduce the more general situation where degrees of freedom are few; multiple correlation, common; and distributions, non-normal. We shall not attempt detailed treatment of what might be labeled heteroscedastic correlation matrices, where the correlation structure varies with signal level or analyte concentration; nor shall we discuss creative applications of principal component techniques (Boqué and Rius, 1996). Our intent, rather, is simply to highlight areas deserving further explication and exploration, and to introduce some scientific applications from our laboratory entailing bi- and multivariate isotopic and chemical blanks. The first issue concerns estimated variance (covariance). Provided that normality can be assumed univariate and multivariate critical values can be set using t- and F -distributions, respectively. Detection limits can be given using the non-central versions of these distributions. That requires not only normality, however, but also a known or assumed variance (covariance) matrix for the HA state (Tiku, 1972; Eisenhart and Zelen, 1958). For the multi-univariate case, one uses tν,1−α where α = 1 − (1 − α)1/k ≈ α/k for k-independent detection decisions, and the non-central-t to calculate detection limits. (See footnote 14 regarding over-reliance on distributional tails.) The multivariate case is treated by replacing the use of the central/noncentral χ 2 distributions, discussed above, with a corresponding use of the central/non-central
82
L.A. Currie
Fig. 5. Univariate and Bivariate Detection Limits (standardized units). The univariate case (ν = 1, left side of the figure) has a 1-dimensional blank vector extending from the origin (x1 = 0) to point B, and net concentration critical level and detection limit vectors extending from B to points C and D, respectively. For normal data, zC = z1−α = 1.645 and zD = (z1−α + z1−β ) = 3.29 (α = β = 0.05). For this illustration, to aid in construction, B has been assumed known. The right portion of the figure depicts the bivariate critical region (dashed arc) and detection limit region (solid arc), both extending from the endpoint of the two-dimensional background vector [B]. The detection limit vector extends into the doubly positive quadrant with origin at point B; it traces an infinite set of detection limit solutions lying along right circular arc D. The vectors to arcs C and D have lengths χ2,0.90 = 2.15 and √ λ = 3.61, respectively, where λ is the non-centrality parameter of the non-central chi-squared distribution where
equals 2.15. (The ν = 1 z’s are the corresponding lengths χ1,0.90 the square root of the lower critical value χ2,0.05 √ and λ, respectively, where λ is now the non-centrality parameter of the non-central chi-squared distribution where
the square root of the lower critical value χ1,0.05 equals χ1,0.90 .) When correlation is present, the right circular arcs become elliptical arcs.
F -distribution. This change is accompanied, and caused by, a change from a known to an estimated blank covariance matrix derived from an n-fold replication of the k-variate background vector (multivariate background data matrix). Here, the multivariate detection test is essentially equivalent to the multivariate t-test, using Hotelling’s T 2 statistic (Jackson, 1959). More complete discussion must be deferred, but two examples from our laboratory will serve to highlight the nature of the problem, and perhaps offer a challenge for those who
Detection and quantification capabilities in nuclear analytical measurements
83
may wish to extend the detection question to new domains. In our investigations of trace chemical and isotopic levels in atmospheric aerosols, we found that processing of the filter materials employed in sampling resulted in covarying chemical and isotopic blanks showing strong positive correlation for the isotopes 12 C and 14 C, and exhibiting large relative standard deviations. The objective of the measurement was to determine the isotope ratio 14 C/12 C, important for discriminating biomass from fossil fuel air pollution. The issue is to develop critical and minimum detectable isotope ratios, in light of the estimated bivariate isotopic blank. An interesting “twist” to the problem is that a severely non-linear transformation is required to convert the observed quantities into an estimated isotope ratio, leading of course to non-normality (Currie et al., 1994). In fact, tri- and tetra-variate extensions of this problem deserve investigation, for example for application to isotopic triplets and quartets that arise in geochemical exploration and source attribution studies. The second research problem relates to the long range transport of atmospheric contaminants from populated continental regions to remote areas such as Greenland and Antarctica. Among the multivariate challenges in our recent research has been the exploration of 8-variate time series of trace anions and cations, in search of patterns that differ substantially from the baseline or background trace chemical patterns found in Greenland air, snow, and ice (Currie et al., 2005).
7. Background, baseline, blank (ultimate limitations) 7.1. The blank as an ultimate limiting factor; blank manifestations, dimensionality For a given method of low-level measurement, in the region of the detection limit, the blank (B) becomes the ultimate limiting factor.16 In fact, as shown in Section 4, the distribution and standard deviation of the blank are intrinsic to calculating the Detection Limit of any measurement process (MP). Standard deviations are difficult to estimate with any precision (ca. 50 observations required for 10% RSD for the Standard Deviation); distributions are harder! It follows that extreme care must be given to the minimization and estimation of realistic blanks for the over-all MP, and that an adequate number of full scale blanks must be assayed, to generate some confidence in the nature of the blank distribution and some precision in the estimate of the blank relative standard deviation (rsd). Provided that the assumption of normality is justified, an imprecise estimate for the Blank standard deviation is taken into account without difficulty in detection decisions, through the use of Student’s-t. Detection Limits, however, are themselves rendered imprecise if σB is not well known (see IUPAC, 1995; §3.7.3.2). Blanks or null effects have several manifestations, depending on their origin (Figure 6). The Instrumental Background is the null signal (which for certain instruments may be set to zero, on the average) obtained in the absence of any analyte- or interference-derived signal; the (spectrum or chromatogram) Baseline comprises the instrumental or environmental background plus signals in the analyte (peak) region of interest due to interfering species; the Chemical (or Analyte) Blank is that which arises from contamination from the reagents, 16 When used without a modifier, the term “Blank” indicates the generic “B” that occurs in the expressions for the
critical level, and detection and quantification limits.
84
L.A. Currie
Fig. 6. The several manifestations of the blank.
sampling environment, or sample preparation steps—corresponding to the very chemical or nuclear species being sought. A fourth, “pseudo-blank”, is the intercept of the calibration curve, as estimated by ordinary least squares (OLS) or weighted least squares (WLS) applied to a calibration model. The intercept-blank and its variance offer certain pitfalls, however, as noted in Section 5.2.2. (See Appendix A.4, for a detailed numerical example that treats the “3 B’s”—Background, Baseline, Blank.) Examples of varied B-frequency distributions are given in Figure 7. These are seen to range from more or less normal, or at least symmetric for low-level counting background (Figure 7a), to decidedly asymmetric (positive skew, TIMS “atom counting” of sulfur and 127 I laboratory blanks: Figures 7b, 7e). The decay counting distributions of environmental 85 Kr (Figure 7d) and laboratory carbon (manometric) blanks for low-level 14 C atom counting (Figure 7c) are more nearly normal, but still with a hint of positive skew, suggestive of contamination. When degrees of freedom are few, however, as with Figure 7e, small departures from normality tend to be statistically insignificant. Added dimensionality potentially brings insight as to the nature, sources and improved control of the background or blank. Isotope and element ratios are cases in point. For example, the 2-dimensional B-distribution including 127 I and 129 I (Figure 8) has several apparent clusters that may provide guidance in identifying and controlling specific sources of contamination. Another example, of some practical importance, relates to correction of the bivariate 12 C–14 C blank in connection with the detection and apportionment of (fossil, biomass) sources of mutagenic atmospheric particles. Details of the low-level metrological and non-linear statistical challenge are given in Currie (2001, Section 2). A multivariate representation of “B” as a reference state is given in Figure 9. In this case, the reference state is the 7-dimensional isotopic-chemical urban background, for the
Detection and quantification capabilities in nuclear analytical measurements
85
Fig. 7. Some B-distributions.
town of Elverum, Norway, the site of a fossil-biomass 14 C apportionment study (Currie, 2000a). The urban background “signature” is directly analogous to the previously discussed 9-dimensional nuclear background signature (Section 6.3), except for the pronounced covariance sub-structure in Figure 9. The detection (discrimination) issue is the same, however, regarding the capability of detection of a differing pattern (source)—in this case the incursion of carbonaceous aerosol from long-range transport. The urban background and the discrimination problem is perhaps easier to grasp when presented in a 3-dimensional perspective (Figure 10). Explorations of the dimensionality of Figures 7a and 7d, which have special significance to the interpretation of B-distributions, will be given in Sections 7.5 and 7.3, respectively. 7.1.1. Assessment of the blank Determination of the magnitude and variability of the blank may be approached by an “external” or “internal” route, in close analogy to the assessment of random and systematic error components (Currie, 1988). The “external” approach consists of the generation and direct evaluation of a series of ideal or surrogate blanks for the overall measurement process, using samples that are identical or closely similar to those being taken for analysis—but containing none of the analyte of interest. The measurement process (MP) and matrix and interfering species should be unchanged. The surrogate is simply the best available approximation to the ideal blank—i.e., one having a similar matrix and similar levels of interferants.
86
L.A. Currie
Fig. 8. Bivariate isotopic blank distribution.
The “internal” approach has been described as “Propagation of the Blank” (IUPAC, 1995; Kelly et al., 1994). This means that each step of the MP is examined with respect to contamination and interference contributions, and the whole is then estimated as the sum of its parts—with due attention to differential recoveries and variance propagation. This is an important point: that the blank introduced at one stage of the MP will be attenuated by subsequent partial recoveries. Neither the internal nor the external approach to blank assessment is easy, but one or the other is mandatory for accurate low-level measurements; and consistency (internal, external) is requisite for quality. Both approaches require expert scientific knowledge concerning the MP in question. Special care must be taken when the blank has large relative standard deviation (rsd). This invariably means that the blank is not normal, for the negative values that must occur with a normally distributed variable with large rsd, such as a net signal near the detection limit, cannot obtain, and conventional error propagation can be misleading. Non-normal blank distributions have their greatest impact on the Critical Level and Detection Limit because of the one-sided character of the false positive and false negative probabilities. Clearly, if the gross signal (y) is normal and the blank (B), lognormal, for example, then the distributional ˆ will change with its magnitude. character of the net signal (Sˆ = y − B)
Detection and quantification capabilities in nuclear analytical measurements
87
Fig. 9. Multivariate blank distribution.
7.1.2. Calibration based detection limits: the blank as the intercept of the calibration function, and σB2 as the intercept of the variance function The variance of the blank, which is essential for the critical level, may be estimated by direct replication, or from the number of background counts, provided that the Poisson assumption holds. In the first case, the precision of the estimate improves with increased degrees of freedom (ν); in the second, with increased number of counts. A less direct approach, that requires a constant variance (homoscedasticity), comes from the dispersion about a fitted linear calibration curve, where σB2 is estimated as the residual variance of the fit s 2 . Such an estimate is invalid, of course, for simple Poisson data, where the expected number of counts (and variance) increases with concentration. Also, the cautions of Sections 4.2 and 5.2 apply, especially Note 3 at the end of Section 5.2.2. If σy is not constant, then the variance function Vy (x) = σy2 (x) must be estimated in order to calculate LD and LQ (ASTM, 1997, 2000; Gibbons and Bhaumik, 2001; IUPAC, 1995). This is straightforward in Poisson–normal case, where LD and LQ are given by simple algebraic expressions (Equations (4.5), (4.6)). More generally, an appropriate variance model must be assumed, with parameter estimation based on replication data as a function of concentration. The model must be fit using iterative weighted least squares, because the statistical weights depend on √ the fitted data—i.e., the rsd of an estimated standard deviation (s) is approximately 1/ (2ν), so the weights for each point of an s(x) function are (2νs −2 ). For the commonly used linear model, σ (x) = b + ax, the intercept provides an estimate of σB while the slope corresponds to the relative standard deviation. A detailed ex-
88
L.A. Currie
Fig. 10. Trivariate urban fingerprint (B-reference state).
ample is given in ISO-11843-2 (ISO, 1997), using the linear model for σy as a function of x, with the objective of estimating the minimum detectable amount of toluene by GC/MS (gas chromatography/mass spectrometry). Model selection for the variance function should be based on scientific considerations and thorough knowledge of the measurement process, rather than simply the “best fit” to alternative empirical models. The problem is that an acceptable fit to an inappropriate, empirical model can lead to biased estimates of detection and quantification limits. We observed this in an investigation of the ultimate detection/quantification capabilities of 14 C accelerator mass spectrometry (AMS), using replicate data covering the range from 0.5 to 40 µg modern carbon (MC). Acceptable (or marginally acceptable) fits to three different models for the variance function gave σB estimates ranging from 0.04 to 0.13 µg MC, and LQ /σB ratios ranging from 10.2 to 26 (Currie, 2004a). 7.1.3. Sensitivity as an ultimate limiting factor; the transition from decay counting to atom counting Complementing the blank as the ultimate limiting factor for low-level measurement is sensitivity, or overall measurement efficiency. The successful development of accelerator mass spectrometry constituted a revolution in sensitivity, for the ratio of the number of atoms counted to the number of decays counted for a long-lived radionuclide is propor-
Detection and quantification capabilities in nuclear analytical measurements
89
Fig. 11. Aerosol 14 C: comparative capabilities for decay counting (llc) and atom counting (AMS).
tional to the ratio of the mean life to the (typical) counting time. Meanwhile, there have been evolutionary (incremental) gains in counting efficiency, sample preparation and blanks. The impact on sample size requirements is illustrated by 14 C: Libby’s original “screen wall counter” which used 8 g of carbon, whereas a quantification limit for modern carbon for environmental studies has been reported recently as 0.9 µg (Currie et al., 2000; Currie, 2004b). Realization of such a capability, however, brings us back to the blank! The driving force for our efforts to reduce sampling and sample preparation blanks for 14 C AMS came in the context of tracing sources and transport of fossil/biomass carbonaceous (combustion) aerosol in urban, regional, and hemispheric atmospheres. Work began, just before the appearance of AMS, with miniature proportional counters, capable of quantifying ≈10 mg modern carbon. AMS improved these capabilities by roughly three orders of magnitude. The tale is captured in Figure 11. The lower right corner shows a Monte Carlo simulation of the bivariate isotopic data mentioned earlier, while the comparative limiting factors for decay (llc) and atom (AMS) counting appear to the left. The “take home messages” are that: All limiting factors are dramatically reduced for AMS. For decay counting the Poisson limit (µg MC to achieve 100 counts with 1000 min counting) and the counter background each limit the quantification mass to several mg MC, with a negligible chemical blank. For atom counting, the situation is reversed, the blank being the ultimate limiting factor, and the machine background, negligible. Applications to remote atmospheres, µ-molar 14 C speciation in trace constituents (elemental carbon, individual PAHs and amino acids), particle size fractions, and recent and paleo atmospheric particles in ice cores provide the challenge, for total particulate carbon in Arctic air is typically 0.3+ µg/m3 while concentrations in firn and snow are ≈9+ µg particulate carbon per kg (Currie et al., 2005).
90
L.A. Currie
Fig. 12. Laboratory vs field air filter blanks.
For these studies, very special measures have been taken to control the laboratory (AMS sample preparation) blank to ca. 0.16 µg carbon, with a variability (sd, not standard uncertainty) of roughly 60 ng carbon (Figure 7c) The field blank is another matter: For Greenland snow, we sent specially cleaned glass and laboratory ware to Summit for direct sample processing in the clean Arctic environment. Air filter samples, however, were subject to occasional local field contamination, and the (quartz) filters themselves showed increased blanks simply by being mounted in the sampling apparatus.17 The contrast between laboratory and field filter blanks is shown in Figure 12, where it is seen that the field blanks have a distinctive positive skew, and the critical value for the net signal is an order of magnitude greater than the mean laboratory processing blank. The skew distribution, and modest number of blank determinations (n = 28) pose a serious problem in estimating the expectation of the critical value, or conversely, the expectation for false positives (α). Treating the frequency histogram as a sample of an unknown distribution, we can form an estimate of the positive tail area (α) above a given level, together with a confidence interval, using the intervals for proportions. The empirical distribution in Figure 12b shows, for example, 3 of the 28 blanks 4.0 µg C/cm2 . The 17 The actual level of contamination seen in the field blanks is not necessarily representative of Greenland, since the
study took place in northern Canada.
Detection and quantification capabilities in nuclear analytical measurements
91
estimated value for the positive tail area above this is thus 0.11. The 90% confidence interval, however, extends from 0.04 to 0.23. As is often the case, we find that the basic limitation is the environmental blank, rather than either the laboratory processing blank, or the inherent instrumental (AMS) quantification capability. A final note regarding asymmetric blank distributions: When they are clearly or potentially positively skewed, the earlier recommendation of paired (sample, blank) measurements should be reinforced, in that the approach to normality is enhanced for the symmetric (Bi −Bj ) distribution (Currie, 2001; Appendix). 7.2. Detection vs Discrimination Limits: impact of the environmental background There is a profound difference in the task (and relevant limits) of detecting a minute amount of a nuclear or chemical species against a laboratory or instrumental background, and the discrimination of a minute change in amount against an environmental background. As noted in Section 3.2, this type of distinction is characterized by the terms “detection limit” applied to the zero null hypothesis and “discrimination limit” applied to the non-zero null (Currie, 1988). In the context of nuclear measurements, the contrast between these two types of limits can alter the comparative performance of alternative low-level measurement techniques—e.g., regarding the choice of atom vs decay counting.18 To illustrate, we consider two methods of detecting Kr isotopes: (1) low-level decay counting (LLC), and (2) detection of individual atoms of 85 Kr and 81 Kr by the recently developed technique of Atom Trap Trace Analysis (ATTA) (Chen et al., 1999). ATTA depends on the production of a beam of “long-lived” metastable noble gas atoms with a discharge source, followed by laser cooling and confinement in a magneto-optical trap where the “photon burst” technique allows the atoms to be detected individually with effectively “zero” background. The background is not really zero, because of a blank or memory effect from krypton atoms implanted in the walls of the discharge chamber (Du et al., 2004). ATTA has excellent isotopic selectivity, as shown in Figure 13 (Bailey et al., 2000); and, as with AMS, the technique is based on the measurement of isotope ratios. ATTA excels at measuring extremely low-levels of 81 Kr and 85 Kr especially as applied to groundwater dating and circulation (Sturchio et al., 2004). Decay counting of 85 Kr is possible also for these purposes, despite background and sample size limitations, but decay counting of 81 Kr is no longer possible. Counting of 81 Kr decays was already extremely difficult when it was first accomplished, in the early nuclear era (Loosli and Oeschger, 1969), but now it is ruled out because of the overwhelming 85 Kr background.19 Although ATTA is currently the method of choice for 81 Kr dating, the first successful atom counting of natural 81 Kr was accomplished by Collon and coworkers, performing AMS on fully stripped Kr ions with a superconducting cyclotron (Collon, 1999). For the detection of small 85 Kr excursions above the atmospheric baseline, which itself is gradually increasing, the situation is reversed. ATTA is limited in its ability to accept large 18 Importance of the non-zero null hypothesis extends beyond environmental baselines, per se. In human biology,
for example, it is important to be able to detect changes from normal homeostatic levels of essential elements such as Zn (2-sided test) and the crossing over of other elements, such as Se, into the toxic region (1-sided test). 19 Nuclear parameters for the Kr “dating” isotopes: half-lives: 2.3 × 105 a (81 Kr), 10.76 a (85 Kr); abundances: 5.2 × 10−13 (81 Kr/Kr), ≈2 × 10−11 (85 Kr/Kr) (from Collon et al., 2004; Du et al., 2004).
92
L.A. Currie
Fig. 13. Single atom counting of Kr isotopes (ATTA) (Bailey et al., 2000).
krypton samples, and its precision is limited by the atom counting statistics of the rarer isotope, 81 Kr, essential for the isotope ratio measurement. The atmospheric 85 Kr discrimination limit by decay counting is superior, with routine monitoring of changes in the baseline level being measured with high precision (large count) on large krypton samples (ca. 7 mLSTP ) extracted from 10–13 m3 of air (Steinkopff et al., 2004). Small sample (100 µL Kr) measurements at and above the atmospheric baseline have been reported using for both atom (ATTA) and decay (LLC) counting. Comparative results for the two techniques are as follows: ATTA 10 samples of purified Kr with 85 Kr concentrations at and below ambient (ca. 1.2 Bq/mL Kr) were measured by Du et al. (2003). For the 5 samples equivalent to the ambient level, relative standard uncertainties for the ratio, 85 Kr/81 Kr ranged from 10 to 15%; counting times, from 10 to 15 h; volumes of Kr, 50–171 µL; and counts, ca. 2400 (85 Kr) and 120 (81 Kr, the precision-limiting isotope). Uncertainties were reported to be Poisson statistics (81 Kr), plus a 5% systematic error related to laser setting uncertainties, plus small contributions from 85 Kr counting statistics and memory effects (from discharge source). Larger ATTA samples (for a single run) were precluded by Ar outgassing.
Detection and quantification capabilities in nuclear analytical measurements
93
LLC Dual, paired measurements were made on about 2.2 µL of Kr in each case. The krypton, which had been extracted from the atmosphere had an activity concentration equivalent to ca. 1.2 Bq/m3 air (Currie and Klouda, 2001). Measurements, performed in a surface low-level laboratory, yielded consistent results, with a final relative standard uncertainty of 9.0%, The 40 mL GM counters used had backgrounds and efficiencies of ca. 0.07 cpm and ca. 90%, respectively; and counting times for the dual, paired measurements were about 1000 min each. Thus, both techniques satisfied the quantification limit for small ambient samples, but the LLC samples were some fifty times smaller than the ATTA samples. Small sample discrimination of minor excursions above the atmospheric or environmental baseline (Be ) may be limited to decay counting because of (1) increased counting precision requirements, and (2) decreased benefit from small or even zero instrumental backgrounds (Bi ). These effects can be expressed quantitatively in terms of the Poisson variance inflation factor, VIF—the factor that converts the zero background Poisson variance VP for the estimated sigˆ including the effects of the B-components. That nal Sˆ to the actual counting variance V (S), is, ˆ = S + Vo = VP · VIF = S · 1 + η(Bi + Be )/S . V (S) (7.1) ˆ to (SQ /10)2 ; For the minimum quantifiable signal (amount, or difference), we equate V (S) thus, SQ = 100 · 1 + η(Bi + Be )/SQ = 100 · VIF, (7.2) where S, Bi , and Be are expressed in counts; and η = 2 for paired sample-background measurements. Note that Vo is simply the expression given in Equation (4.7), for the case where ϕxB is negligible (not necessarily true), and where B now has two components. The two limitations above are now (perhaps) transparent: For discrimination above (or below) baseline counts, the advantage of a zero instrumental background decreases, because now “B” cannot be reduced below Be ; and, similarly, a non-zero background (Bi ) may become negligible compared to Be . To quantify (to 10%) a given excursion above the environmental baseline—e.g., 50% (SQ = Be /2), the variance of S must increase above the zero background value (VP ) by the factor VIF to compensate for the effect of the environmental background (Be ). At the same time, SQ is increased to 100·VIF counts, and it becomes necessary to achieve a relative standard deviation for the “sample” (gross) count (y = S + B) below 10%. For example, taking η = 2, Bi = 0, and S = 0.5Be (50% excursion above ambient), we find that the relative counting “error” for gross counts y is reduced from 10 to about 2.6%. Including our non-zero instrumental background has a relatively small impact, as then the precision requirement for y is further reduced only to 2.0%. (SQ for these two cases is 500 and 664 counts, respectively.) A comprehensive view of the impact of the environmental background and its variability on small sample 85 Kr detection and discrimination capabilities for the NIST low-level facility is shown in the following table. The excellent detection limit of the smallest counter (5 mL, 0.02 cpm background, 65% counting efficiency) is quickly eroded for detection or quantification of small increments above ambient 85 Kr. Also, under those circumstances where Be dominates, the optimal counter changes from the 5 mL proportional counting tube to the 40 mL
94
L.A. Currie
Table 3 Kr-85: detection, quantification limits (Bq m−3 ) (2 L air; T = 1000 min; ambient = 1.4 Bq m-3 ) >BG
>Ambient
>Ambient (rsd = 0.12)
LD 40 cm3 (SG) 40 cm3 (NaI) 5 cm3 (NaI)
0.79 0.40 0.30
0.95 0.67 0.71
1.23 1.03 1.06
LQ 40 cm3 (SG) 40 cm3 (NaI) 5 cm3 (NaI)
2.83 1.68 1.68
3.32 2.48 2.79
4.17 3.56 3.83
Excursions (Bq m−3 ): Prague (June 1986): 1.4; Freiburg (April 1995): 3.2.
(NaI anticoincidence) GM tube with its higher background (0.07 cpm) but improved counting efficiency (90%). The complete discussion, including basic equations and data is found in Currie and Klouda (2001). In particular, Figure 3 in that publication highlights the diminishing returns that set in for background reduction with increased counting time and increased precision requirements—in connection with the conventional “Figure of Merit”, S 2 /B. Of special interest is the comparison of an ideal “zero background” counter, which shows little advantage over the background of the 5 mL counter with increasing counts (counting time). The transition is seen also in Table 3, where the detection limit advantage of the smaller (5 mL) counter nearly disappears at the quantification limit. To complete the ultimate detection discussion for 85 Kr, mention should be made of the need to measure 85 Kr contamination at extreme low levels in connection with the BOREXINO solar neutrino experiment (Simgen et al., 2004). The contamination issue relates to the potential problem of gaseous radionuclide impurities in N2 used for sparging the organic liquid scintillator used to detect 7 Be solar neutrinos (expected rate: 34 ν-induced events per day). The 85 Kr impurity limit for the experiment is set at 0.13 µBq/m3 of nitrogen. Such a low detection limit demands both massive sampling (here, 750 m3 ) and ultra-low level counting. The latter is being achieved in an exceptional underground laboratory with 1 cm3 proportional counters having a detection limit of just 100 µBq. This raises another interesting detection capability comparison, for the NIST 5 cm3 counter (Table 3) has an estimated detection limit of about 0.30 Bq/m3 air, which for 2 L air is equivalent to 600 µBq. This gives a direct measure of the improvement that may be expected by adopting the smallest practical counter and moving to an underground laboratory. The comparative detection limits above, however, leave a question that is most pertinent to the central topic of this chapter: “What is meant by the “Detection Limit?”20 The 600 µBq value is based on the IUPAC definition, with α and β equal to 0.05, the Poisson–normal distribution, η = 2, and T = 1000 min, with a background of 0.02 cpm and 65% counting efficiency. If, for example, the 100 µBq were based on the minimum significant signal (critical 20 That question may be answered in work referenced by Simgen and coworkers, but such references were not
searched.
Detection and quantification capabilities in nuclear analytical measurements
95
level), that would yield a factor of two between the two “detection limits”. Changing from √ paired measurements to η = 1 could account for another factor of 2, and extension of counting time beyond 1000 min could easily yield another factor of two. 7.3. Further exploration of laboratory (blank) and environmental (baseline) distributions Empirical “B” distributions, as shown in Figure 7, may be used directly to estimate critical levels and detection limits, especially when there are many observations, or degrees of freedom—for S (Figure 7b), n = 179; for 85 Kr (Figure 7d) n = 109. When such frequency histograms display positive skewness, however, caution is advised, for it may be an indication of non-stationarity and contamination. Time series analysis and physicochemical analysis are powerful tools in such cases. We illustrate below some insights derived from the S and 85 Kr distributions. A third example, explored in Section 7.5 (low-level counter backgrounds), demonstrates that, even with symmetric frequency histograms, exploration of “hidden” dimensions—beyond the blank frequency distribution, per se—can provide very important insight concerning non-random phenomena. 7.3.1. Distribution of the sulfur laboratory blank In a remarkable series of experiments on the laboratory blank associated with trace analysis of sulfur by thermal ionization mass spectrometry, driven by the impacts of trace amounts of sulfur on fuel emissions (coal) and materials properties (steel), Kelly and coworkers have systematically collected blank data over nearly a 20 year period (Kelly et al., 1994). A histogram showing the data distribution for the first decade of this record (n = 179) is given in Figure 7b. The distribution is clearly skewed. Although there is no rigorous theoretical basis for the form of the distribution, the distribution of sulfur blanks is well fitted with a 2-parameter lognormal function: x¯ = 0.27 µg S, s = 0.20 (p = 0.39).21 A normal distribution yields a very poor fit (p < 0.000001). At this point, it is important to inject a note of caution, regarding assumptions. Use of the fitted distribution for significance testing, or more generally for generating uncertainty intervals for low-level measurements, depends on assumptions that: (1) the presumed form of the distribution function is correct22 and (2) the blank is stationary (fixed mean, dispersion) and random (independent, or “white” noise). If the blank has multiple components (generally the case for environmental blanks), stationarity implies that the “mix” of components is fixed (not generally the case for environmental blanks). Even in application of a multi-step sample preparation process, the structure of the blank distribution may change as a result of multiple injections and multiple losses of blank components along the way. Testing for stationarity and randomness requires additional information, such as a meaningful time/space series, or exploration of some other type of informing variable, or “hidden dimension”. More generally, such external information relates to potential sources and the 21 Unless otherwise indicated, all “p-values” for lack of fit refer to the χ 2 test. Unless the histogram fit is very poor, p(χ 2 ) should interpreted cautiously, because the p-value will change with the number of classes and location of class boundaries, sometimes substantially. When there are too few data classes to perform the χ 2 test, the Kolmogorov–
Smirnov test, indicated p(K–S), is applied. 22 Note that a “good fit”, per se, does not constitute proof, except perhaps in the case that all other models can be
ruled out. If the randomness assumption is valid, however, distribution-free techniques may be applied, such as the use of non-parametric tolerance limits (see also the following footnote).
96
L.A. Currie
chemistry and physics of the blank, i.e., the science underlying the introduction of contaminant. For sulfur there are several interesting sources, transport routes, oxidation states, and phase changes to be considered. To quote Kelly, regarding the shape (positive skew) of the empirical frequency distribution of the sulfur blank: “. . . this distribution is probably a combination of a normal distribution plus random events that add sulfur to our system. These events could be changes in SO2 gas concentration in the air in the laboratory, changes in sulfate particulates in the laboratory, and other unknown events. I would expect all blank distributions to be positively skewed to some degree. We have noticed that on days when it rains, that the blanks appear to be lower, and we are tracking this. The large Dickerson Power Station and the new county incinerator are located about 20 miles to our west. Negative skewness in a blank distribution is an impossibility in my mind” (W.R. Kelly, personal communication, 14 September 2000). The essential point is that for rigorous application, the empirical blank frequency distribution must be at least tested for randomness. Far better, however, for reduction and control, is the injection of scientific understanding of the sources and pathways for the introduction of the blank. 7.3.2. Distribution of the 85 Kr environmental background The distribution of a second series of blanks, also noteworthy for its length, is given in Figures 7d and 14a. As with the sulfur blanks, the data distribution is skewed to the right, perhaps lognormal, and it again represents an accumulation of data over a period of about a decade, by Wilhelmova et al. (1994) (see also Csongor et al., 1988). In contrast to the sulfur data, however, Figure 14a represents a field blank, in this case atmospheric (85 Kr) background radioactivity in central Europe (Prague). The data, which were collected during the period 1984–1992, contain hints of some of the most important events in modern European history. The histogram of the complete dataset (n = 109) can be fit to a lognormal distribution (p = 0.12), whereas a normal fit is unacceptable (p ≈ 0.0002). Omitting the “unusual” point at 2.23 Bq m−3 brings about an improved lognormal fit (p = 0.59) which is the one shown in Figure 14a. The fitted parameters (x, ¯ s) are (0.930, 0.185) Bq m−3 , equivalent to a relative standard deviation (rsd) of 19.9%. Although the frequency distribution is still noticeably skew, a normal distribution cannot be rejected.23 Looking beyond summary statistics and empirical frequency distributions can be remarkably revealing. Through the added dimension of time (Figure 14b) the 85 Kr data project a multicomponent background structure that is neither stationary, nor simply random. Visually, even, there appear to be sharp, positive excursions, and a complex secular trend, including quasi-seasonal variations and an apparent level shift of nearly 30% during 1989. In Figure 14b it is seen that the unusual datum at 2.23 Bq m−3 resulted from sampling in spring 1986, apparently shortly after the fateful event of 26 April 1986 at Chernobyl! Trend modeling of the 85 Kr background series was done by Wilhelmova et al. (1994) with linear and quadratic baselines, but the resulting residuals were far from random. In principle, much more might be learned via scientific and “political” modeling, linked, for example, 23 A “good fit” of an environmental blank frequency distribution to any model distribution function must be treated
with some circumspection. Not only does it not constitute proof that the blank data have been drawn from such a distribution, it does not even imply that the data are random or independent. Since environmental blanks generally reflect complex, non-stationary processes that are multicomponent in nature, it is risky to consider fitted model distributions more than smooth descriptions for the particular blank data set. Even the concept of outliers, apart from outright blunders, deserves cautious interpretation.
Detection and quantification capabilities in nuclear analytical measurements
97
Fig. 14. Environmental (field) background data for 85 Kr (n = 109) in the central European atmosphere covering the period 1984–1992. (a) (upper plot) shows the radioactivity concentration frequency distribution and the best fit lognormal distribution (excluding the highest point). Estimated parameters are (x, s) = (0.930, 0.185) Bq m−3 , equivalent to a relative standard deviation (rsd) of 19.9% (p = 0.59). (b) (lower plot) shows the 9 year time series of measured concentrations, revealing a non-stationary mean and several positive excursions. Prior to 1989, data were derived from daily samples taken at monthly intervals; later data correspond to integrated monthly samples. (Figure (b) is adapted from Figure 1 in Wilhelmova et al. (1994), with permission.)
98
L.A. Currie
to tracer applications and to anthropogenic releases from spent nuclear fuel, combined with transport and mixing in the troposphere. An interesting alternative, as a first step in understanding such complicated data series, is the application of the techniques of Exploratory Data Analysis as pioneered by the late J.W. Tukey (1977, 1984). We have taken the latter course. The result of iteratively applying Tukey’s “5RSSH” resistant, non-linear smoother to the 85 Kr baseline data of Figure 14b is shown as the smooth in Figure 15a. The 5RSSH algorithm is resistant to outliers and generally more flexible than simple moving averages; it is based on 5 point running medians, supplemented by “splitting” and “Hanning” running weighted averaging (Velleman and Hoaglin, 1981). Because of the magnitudes of the excursions above the baseline, the smoother was applied twice, with those values replaced by corresponding smoothed values produced by the first pass of 5RSSH on the raw data. The resulting temporal pattern is striking. The principle features displayed by the smooth in Figure 15a include: (1) a secular trend that rises from a mean level of 0.81 Bq m−3 prior to 1989, to an average level of ca. 1.04 Bq m−3 after 1989; (2) a quasi-annual cycle with maxima tending to occur in mid-year; and (3) a dramatic, nearly 30% level shift (increase) during a single year, 1989.24 One wonders whether there might be an indirect historical link with the tumultuous sociopolitical events in Eastern Europe during that same year, when the “Iron Curtain” was rent. About the time that the smooth began to rise above values of the preceding 5 years, the collapse of Communism in Eastern Europe began—with the dismantling of the barbed wire fence separating Hungary and Austria on 2 May 1989, and culminating with the destruction of the Berlin Wall on 9 November 1989. Complementing the 5RSSH description of the systematic background variation is the Tukey rough or residual component (raw data minus the smooth, excluding artifact zeroes); this provides an approximate description of the random component of the background variations. In contrast to the skew frequency distribution of Figure 14a, the distribution of the rough (Figure 15b) is consistent with a normal distribution (x, s, p = 0.00, 0.084, 0.50, respectively). Normal tolerance limits (P , γ : 0.90, 0.90 bounding lines in the figure) show that gross concentrations that exceed the local value of the smooth by ca. 0.16 Bq m−3 would be judged significant. The rough is not homoscedastic, however, as seen by the pre- and post-1989 tolerance limits in Figure 15b. The relative standard deviations (rsd) before and after 1989, however, are approximately the same, at ≈8.7%. This value serves as the fundamental constraint to the capability to detect excursions above the smoothed environmental baseline. (See Section 7.2 and Currie and Klouda (2001) for a detailed theoretical treatment of the impacts of Poisson, instrumental, and environmental “noise” on detection and quantification limits for atmospheric 85 Kr.) Although the rsd of the rough is about twice the rsd from counting statistics, as shown in Figure 14b, removal of the structure underlying the histogram of Figure 14a has resulted in an rsd reduction (from 19.9%) by more than a factor of two. Conclusion. An interesting and central feature of laboratory and environmental blank distributions is their tendency to fall into two classes: those that tend to be controlled (endogenous) and more or less normally distributed, and those that are uncontrolled (exogenous) that tend to 24 A significant level shift (nearly 20%) remains even if the large excursions from the baseline are included in the
pre-1989 mean.
Detection and quantification capabilities in nuclear analytical measurements
99
Fig. 15. Decomposition of the 85 Kr background time series into “signal” and “noise” components. Exploratory data analysis was utilized to extract a complex signal with minimal assumptions from the raw time series data of Figure 14b, using the Tukey resistant non-linear smoothing procedure 5RSSH (Tukey, 1977, 1984; Velleman and Hoaglin, 1981). (a) (upper plot) shows the estimated signal (the smooth). Important features are the significant shift in mean concentration during the momentous year 1989, and the quasi-periodic structure that tends to favor higher atmospheric concentrations in mid-year. (b) (lower plot) shows the residuals (the rough) remaining after the smooth is extracted from the raw baseline data of Figure 14b.
100
L.A. Currie
exhibit positive skewness. The first class is found in generally controlled physical and biological systems, such as (1) self-regulating vegetative and human organisms whose health depends on limited concentration ranges of “essential” elements or compounds, and (2) environmental (geochemical) compartments and laboratory reagents having stabilized blank components that tend to be well mixed and uniformly distributed. The second class is the “wild class” where toxic or non-essential contaminants are found in widely varying concentrations, depending of the whims of contaminant (source) strengths and environmental transport processes. A striking contrast in this regard is found in the distributions and multivariate patterns of toxic and essential elements found in different environmental/global regions in the Daily Diet Study of the International Atomic Energy Agency (Currie, 1992). 7.4. Extreme low-level counting: the Poisson perspective This final part of the “B” section of the chapter treats the situation where few counts are observed, and the Poisson distribution is far from normal. Detection decisions and detection limits are considered in Sections 7.4.1 and 7.4.2 for the two asymptotic cases where tB
ty = tS+B (well-known blank, η = 1) and tB = ty (paired counting, η = 2), respectively. While not presented explicitly here, the intermediate cases are readily treated by the theory in Section 7.4.2. The last segment, Section 7.5, considers the nature of the distribution of (background) counts, which is not necessarily Poissonian. 7.4.1. Poisson detection decisions and limits for the well-known background asymptote (η = 1) (Currie, 1972) When the background is “well-known,” with expectation μB , the exact Poisson formalism for SC and SD (Section 7.4.1.1) is quite straightforward, since Sˆ is represented by the difference between a single Poisson variable and a constant (μB ). Because of the discrete nature of the Poisson distribution, however, the defining equation for the detection decision must be cast as an inequality (Equation (3.1)). If one wishes to realize a specific target value for α (as 0.05), a little-used Bernoulli weighting procedure can be employed to overcome the inequality (Section 7.4.1.2). Finally, to complement the algebraic solutions presented, a graphical representation of η = 1, extreme Poisson critical values and detection limits is given in Section 7.4.1.3. 7.4.1.1. Exact Poisson solution Notation. For the treatment of the extreme Poisson distribution the following notation is adopted.25 • Expectation of background counts: μB (real). • Expectation of “sample” (gross counts) at the detection limit: μD (real)—alternatively: yD . • P (n | μB ) = (μB )n /n! · exp(−μB ): Poisson density function, Prob of n counts (integer), given μB . • P (n | μD ) = (μD )n /n! · exp(−μD ): Poisson density function, Prob of n counts (integer), given μD . 25 Note that the Poisson parameter (μ) is continuous (real), whereas the observable counts (n) are necessarily discrete
(integers).
Detection and quantification capabilities in nuclear analytical measurements
101
Table 4 Exact Poisson confidence limitsa Observed counts (n)
Lower limit (μ− )
Upper limit (μ+ )
0 1 2 3 4 5 6 7 8 9
0 0.0513 0.355 0.818 1.37 1.97 2.61 3.28 3.98 4.70b
3.00 4.74 6.30 7.75 9.15 10.51 11.84 13.15 14.44 15.70b
a P = 0.05, 0.95; μ = χ 2 /2 (ν = 2n, ν = 2n + 2). ± − + P b Poisson–normal approximation (Section 4.3.1): 5.24 (μ ), 15.47 (μ ). The mid-point of the confidence interval − + (solution of Equation (4.9)) equals (n + z2 /2), or (n + 1.35) for the 90% CI.
• P (n | μ) = (μ)n /n! · exp(−μ): general expression for the Poisson density function. k Pr(n k | μ) = (7.3) P (n | μ): • 0
Poisson (cumulative) distribution, for integers zero to k. Given the expectation for the background counts (μB ), we can apply the exact cumulative Poisson distribution to calculate the critical number of counts nC , considering the error of the first kind (α), and then the detection limit for the expectation of the gross sample counts μD , considering the error of the second kind (β). These values follow from the defining equations (3.1) and (3.2), adapted to this special case. For the net signal, it follows that SC = nC − μB , and SD = μD − μB . Critical value: Pr(n > nC | μB ) 0.05
(7.4)
(an inequality because of the discrete Poisson distribution). Detection limit: Pr(n nC | μD ) = 0.05.
(7.5)
The solutions to Equation (7.4) and (7.5), for any particular choice of μB , present no problem, due to the wide availability of mathematical–statistical software. For a convenient overview, however, we present brief tabular and graphical renditions for the problem. The solutions, in fact, can be easily extracted from a table of the confidence limits for the Poisson parameter vs the integer values of n. These confidence limits can be determined readily from the chi-squared table. (See Section 4.3 for a more complete exposition, including a detailed comparison with the results of the Poisson–normal approximation, which gives remarkably good interval estimates for μ, even for very few counts observed.) Table 4 gives exact Poisson lower (0.05) and upper (0.95) confidence limits for μ for integer values of n, ranging from zero to 9. Table 5 gives the derived values for the critical number of counts (nC ) and the gross count detection limits (μD ).
102
L.A. Currie
Table 5 Poisson critical values and detection limits (η = 1; tB tS+B ) α (minimum)a
μB (range)
nC = SC + μB (α 0.05)
μD = SD + μB (β = 0.05)
– 0.0013 0.0060 0.0089 0.013 0.016 0.018 0.020 0.021 0.022
0–0.051 0.052–0.35 0.36–0.81 0.82–1.36 1.37–1.96 1.97–2.60 2.61–3.28 3.29–3.97 3.98–4.69 4.70–5.42
0 1 2 3 4 5 6 7 8 9
3.00 4.74 6.30 7.75 9.15 10.51 11.84 13.15 14.44 15.70
a For each μ
B range, α varies monotonically from αmin to 0.05.
To illustrate, take the expectation of the background count to be 2.30. Referring to Table 5, we find that nC = 5 counts, and μD = 10.51 counts (expectation). Thus, SC = 5 − 2.30 = 2.70 counts; and SD = 10.51 − 2.30 = 8.21 counts. The net count detection limit, expressed in units of background equivalent activity (BEA) is 8.21/2.30 = 3.57. The first column of Table 5 shows us that the actual value of α falls in the range of 0.016 to 0.050. The expression for the cumulative Poisson distribution (Equation (7.3)), with μB = 2.30 and k = nC = 5, gives the result α = 0.030. For k = 4, α = 0.084. If only an estimate for μB is available—e.g., from nB counts observed, we can compute an exact Poisson confidence interval (CI) for μB , and from that, an interval for nC and μD . If the μB CI falls entirely within one of the μB bands in column 2 of Table 5, then nC and μD have unique values, but the actual value for α is uncertain. Such an event is unlikely for small numbers of counts observed, however. For example, if nB = 4 counts, then the 90% CI for μB equals 1.37 to 9.15 counts. The “Bayesian” estimate of (nB + 1) is no less uncertain, but it does give a somewhat better point estimate for μB as the approximate mid-point of the (90%) confidence interval. (Compare: Est(μB ) = 4 + 1 = 5 counts, with mid-CI = (1.37 + 9.15)/2 = 5.26 counts.) The exact Poisson–normal solution for the mid-point of the μB CI (solution to Equation (4.9)) gives the mid-point as (nB + z2 /2) = (4 + 1.35) = 5.35 counts for the 90% CI. 7.4.1.2. Overcoming the inequality Equation (7.4) can be transformed into an equality, such that α = 0.05, by applying Bernoulli weights for the α’s from the cumulative Poisson distribution for k = nC and k = nC − 1. Taking the previous example where μB = 2.30 counts (nC = 5), we can force a value α = 0.05, by selecting a value of p such that p · α(k = 4) + (1 − p) · α(k = 5) = 0.05. Using the α’s (k = 4, k = 5) found above, the expression becomes α = 0.05 = p · (0.084) + (1 − p) · (0.030),
giving p = 0.37.
In practice we would use a series of Bernoulli random numbers (0’s and 1’s), with p = 0.37, to select 4 counts as a modified critical value 37% of the time, and 5 counts 63% of the time.
Detection and quantification capabilities in nuclear analytical measurements
103
Fig. 16. Poisson critical values (nC ) and detection limits (ρD ) vs background expectation (μB ). The y-axis corresponds to the background equivalent activity (BEA or ρ)—i.e., the ratio of the net signal to the background, S/μB . The sawtooth envelope represents the detection limit SD /μB , with nC indicated by the integers just above the envelope. The dashed line corresponds to the Poisson–normal approximation for ρD .
7.4.1.3. Graphical representation; approach to normality A global perspective is presented in Figure 16, which shows, as a function of μB : (1) the critical numbers of counts nC , (2) the detection limit, expressed here as the ratio SD /μB —i.e., ρD , in units of background equivalent activity (BEA), and (3) the approach to normality. The latter is indicated by the dashed line that is, in effect, the Poisson–normal extension of the discontinuous “sawtooth” envelope of the exact Poisson detection limits (ρD ). The results for previous example (μB = 2.3 counts), based on the limited tabular (Table 5) equivalent of Figure 16, are evident also from the plot: the ordinate for the ρD curve corresponding to 2.3 counts on the abscissa is seen visually to be approximately 3.6 (SD /μB ), giving a net signal detection limit of about 8.3 counts. The corresponding value for nC , indicated just above the ρD envelope is, again, 5 counts. The Poisson–normal approximation is already reasonably good for μB = 2.3 counts; application of Equation (4.5) gives SD = 7.7 counts, or ρD = 3.4. The asymptotic function for ρD (dashed curve), corresponding to Equa√ tion (4.5), equals (2.71/μB + 3.29/ μB ). 7.4.2. Extreme low-level (Poisson) detection decisions and limits for paired counting (η = 2) The solution for the other asymptotic case, where detection decisions and limits must be evaluated for pairs of Poisson variables, was published more than 65 years ago, by Przyborowski and Wilenski (1939). The stimulus for the work of these authors was the practical problem of
104
L.A. Currie
detecting small numbers of dodder seeds in clover. This was of some practical import, because the contaminating seeds belong to the class of twining herbs that are parasitic to plants. In the context of this chapter, the detection of the rarely occurring dodder seeds is the analog of the detection of rarely occurring nuclear particles or decays. Unlike the previous section, where the challenge was to detect a significant signal above a well-known background, we now address the problem of detecting a significant (1-sided) difference between two extreme low-level Poisson variables. The relative simplicity of the previous section is gone, since the distribution of the difference between two Poisson variables is no longer Poissonian. In fact, the solution space is now 2-dimensional, with a critical region replacing the critical level of the single Poisson variable. Przyborowski and Wilenski formulate the problem by first expressing the joint probability law for observations x, y as y p(x, y | μx , μy ) = μxx μy /x!y! · exp −(μx + μy ) ,
(7.6)
where, in the context of low-level counting, x represents counts observed from the background variable (B) having expectation (mean) μx , and y represents counts observed from the gross count (signal + background) variable having expectation μy . The density function follows from the fact that the distribution of the sum of Poisson variable is itself Poissonian.26 Equation (7.6) can be transformed into a more interesting form (7.7) using the following substitutions: ρ = μy /(μx + μy ), μ = μ x + μy , n p(x, y | ρ, μ) = (μ /n!) · exp(−μ) n!/ y!(n − y)! ρ y (1 − ρ)n−y .
n = x + y,
(7.7)
Critical region. The sample space is necessarily a 2-dimensional (integer) grid, with the possible sample points (E) defined by the discrete observable values of x (background counts) and y (gross sample counts). For a given n, the partition into y and x = n − y is governed only by the second factor in Equation (7.7), which is a term in the binomial expansion of [(1 − ρ) + ρ]n . For the null hypothesis, μy = μx , so ρ = 1/2; thus, for each n, the critical value for y is given by Pr(y > yC | n, ρ = 0.5) 0.05, independent of μ. The yC are simply the 1-sided critical values for proportions, which may be determined from the binomial (cumulative) distribution (n, ρ). To give a specific illustration, consider an observation pair for which x + y = n = 12 counts. Then the integer yC derives from the 95+ percentile of the binomial distribution (12, 0.5), which equals 9 counts. (The probability that y > 9 counts, given n, ρ, is 0.019.) The full set of gross count critical values (yC ) for background counts from x = 0 to x = 30 for α 0.05 is given below.27 26 For more convenient application to the low-level counting problem, slight changes have been made from the
original formulation (in notation, only). 27 Extension to larger values of x can be accomplished using the Poisson–normal approximation, including the cor-
rection of 1/2 for “discontinuity” (Przyborowski and Wilenski, 1939; p. 323), also called the correction for “continuity” (Cox and Lewis, 1966, pp. 21 and 225, respectively). This correction has not been applied to the Poisson–normal approximations appearing elsewhere in this chapter.
Detection and quantification capabilities in nuclear analytical measurements
105
(x, yC ) (0, 4)
(1, 6) (11, 21) (21, 34)
(2, 8) (12, 22) (22, 35)
(3, 9) (13, 23) (23, 36)
(4, 11) (14, 25) (24, 37)
(5, 12) (15, 26) (25, 39)
(6, 14) (16, 27) (26, 40)
(7, 15) (17, 29) (27, 41)
(8, 17) (18, 30) (28, 42)
(9, 18) (19, 31) (29, 44)
(10, 19) (20, 32) (30, 45)
Detection limit. Evaluation of the detection limit for the exact paired count Poisson problem is not so simple, in that the full probability equation (7.7) must be considered. Przyborowski and Wilenski calculate the power function (1 − β, given α) from the expression ∞ P {E ∈ w | ρ, μ} = n!/ y!(n − y)! ρ y (1 − ρ)n−y , (7.8) (μn /n!) · exp(−μ) n=0
w
where (E ∈ w) refers to all observable pairs of observations (x, y) that lie within the critical region w—i.e. beyond the critical contour. Numerical data given for power (1 − β) = 0.95, for α 0.05, were combined with critical values yC to construct plots showing the critical values (yC , discrete; dashed curve) and detection limits (yD , continuous; solid curve) as a function of μx (which is μB ). This information is given in the upper plot in Figure 17a. Note that although the detection test is based strictly on the observed count pair (x, y), the detection limit necessarily depends on μB . The lower plot, Figure 17b, gives similar information for the well-known blank case (η = 1), based on the derivations of Section 7.4.1. To give a clearer representation of the two-dimensional distribution of x, y for the null case, with a given mean μ, a Monte Carlo simulation is shown in Figure 18. The three-dimensional frequency histogram was created for the null case, with a common mean (μx , μy ) equal to (5.00, 5.00), and 1000 random samples from the pair of Poisson distributions. The lower part of the figure shows a series of (x, y) pairs extending beyond the critical region; these constituted 2.3% of the random pairs. This is consistent with 0.05 as the upper limit for the actual α of the null hypothesis test, and the value of 0.024 given for the true significance level when μ = 5 (Table IIa in Przyborowski and Wilenski, 1939). α-Control. Like the well-known blank problem, the paired Poisson variable problem is characterized by discrete observables (counts), resulting in discontinuities in α. An alternative to the “Bernoulli trick” to force the equality, α = 0.05, is the possibility of exercising modest control over the wide, small count swings at the critical value (nC ) by the selection of that member of the bracketing pair having α closer to 0.05. Such a practice moderates the swings to extremely small α, without the need to generate and apply series of Bernoulli random numbers—but at the cost of modest dispersion (+, −) about the target value of 0.05. To illustrate, the bracketing (1 − α)’s for n = x + y from 10–15 counts, for (nC , nC − 1) are: n = 10 (0.989, 0.949), n = 11 (0.967, 0.887), n = 12 (0.981, 0.927), n = 13 (0.954, 0.866), n = 14 (0.971, 0.910), n = 15 (0.982, 0.941). Using the conventional inequality (α 0.05) gives (α±sd) ¯ = 0.026±0.013; for the alternate, “closest approach” rule, (α±sd) ¯ = 0.049±0.017. Evidently the quite simple, alternate rule is a good choice, at least for this part of the critical region. (See also footnote 9.)
106
L.A. Currie
Fig. 17. Extreme low-level counting: critical regions (dashed curves) and detection limits (solid curves) for (1) paired counting (η = 2)—upper plot (a), and (2) well-known background (η = 1)—lower plot (b). In (a), both axes have integer values only for the dashed (critical value) curve, where x = observed background counts, and y = observed (gross) sample counts. The solid curve represents the detection limit as a function of the background expectation; both axes are then continuous. In (b), the x-axis represents the expectation of the blank counts, and is continuous. y (gross counts) has integer values only, for the dashed (critical value) curve, but it is continuous for the solid curve which represents the detection limit (expectation) as a function of the background expectation.
7.4.3. Some closing observations, and reference to other work Some summary observations can be drawn from the two sets of yC critical boundaries and two sets of yD detection limit curves in Figure 17. For the paired case (η = 2, Figure 17a), the minimum value for yC (4 counts) occurs when x = 0; thus, the smallest integer pair that would indicate “detection” would be (x, y) = (0, 5). Similarly, for x = 4 counts, the smallest significant value for y would be 12 counts. Looking at the detection limit curve, we see that for a truly zero background, the minimum detectable gross count (yD ) is approximately 9.0 counts, which here also equals SD . For μB = 4.0 counts, the minimum detectable yD ≈ 20.0 counts (SD ≈ 16.0 counts). To detect a net signal (S) equal to 4 times the background, the intersection of line of slope 5 (μy /μx ) with the (gross signal) detection limit curve is needed. That occurs at μB = 4.0 counts. Considering the same background values (μB = 0, μB = 4.0 counts) for the well-known background case (η = 1, Figure 17b), the minimum detectable gross counts are yD = 3.00
Detection and quantification capabilities in nuclear analytical measurements
107
Fig. 18. Empirical (Monte Carlo) test of the critical region for extreme low-level counting (paired Poisson variables, η = 2).
counts and yD = 14.44 counts, respectively. The corresponding values for the minimum detectable net signal are SD = 3.00 counts and SD = 10.44 counts, respectively. So, for this small count range, excellent knowledge of the background expectation carries a benefit (reduction in SD ) of a factor of two to three. (The Poisson–normal approximation, Equation (4.5), applied to μB = 4.0, gives SD = 9.29 counts (η = 1) and SD = 12.02 counts (η = 2).) Although the critical value boundary curve allows rigorous testing for significant signal, background differences in the two asymptotic (η = 1, 2) extreme Poisson cases, the corresponding detection limit curves are essential for planning for successful extreme low-level studies. A statement from Przyborowski and Wilenski (1939) captures the thought: “It might with reason be regarded as undesirable to plan an experiment in which the chance was less than 0.5 of detecting from two random samples that differences . . . of practical importance existed”. The extreme Poisson, well-known blank case has achieved practical importance in very low-level monitoring and spectroscopic applications, such as in the large scale gamma ray monitoring activities of the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) (De Geer, 2004). Three factors are at play in the low-level monitoring activities in question:
108
L.A. Currie
(1) the backgrounds (and gamma ray baselines) are generally empty and smooth, such that large regions can be utilized for background estimation, yielding a background that is rather well known; (2) significant peak areas may fall in the very small count region, such that the rigorous Poisson distribution should be applied; (3) continual monitoring with null hypothesis dominance means that large numbers of false positives will occur, unless the effect of multiple detection decisions is taken into account. For these special circumstances, De Geer proposes application of the rigorous Poisson distribution with the well-known blank, with the α reduced from 0.05 to reflect the large number of null decisions. Table 2 of De Geer (2004) is basically an extension of our Table 5 (Section 7.4.1) for type-1 risk levels (α) ranging from 5% to 0.0001% and LC (Poisson critical counts) ranging from 1 to 50. Multiple null decision α reduction is governed by the binomial distribution (Section 6.2.2); for small α, the reduction required is approximately α/n, where n is the expected number of null hypothesis tests. There has been a resurgence of interest in recent years in the low-level counting communities in the problem addressed by Przyborowski and Wilenski so long ago, in part because of major advances in, and needs for, low-level radiation measurement, and in part because of then un-dreamed of advances in computing power. A review of such work is impracticable in this chapter, but a recent inter-agency resource document may be consulted for an excellent critical review with pertinent references, exhibits, and examples (MARLAP, 2004). For special topics related to extreme low-level gamma-ray spectrometry, see also the Proceedings of the Conference on Low-Level Radionuclide Measurement Techniques (ICRM, 2004), CELLAR (2004), Laubenstein et al., (2004); Povinec et al. (2004), and Méray (1994). 7.5. On the validity of the Poisson hypothesis for background counts It seems to be conventional wisdom, at least for low to moderate counting rates, that counting background can be described as a Poisson process. Such an assumption, of course, affects uncertainty estimates, as well as computed critical levels and detection limits, where the variance of the background is dominant. For those background components from long-lived radioactivity, the assumption for the decay process is doubtless valid, but the assumption for the counting process deserves rigorous testing for possible environmental and/or measurement process artifacts. The following summary is drawn from a comprehensive study of time series of some 1.4 × 106 individual low-level GM background events (coincidence and anticoincidence) spanning a period of nearly one month, where the amplitude and time of occurrence of each pulse was recorded to the nearest 100 µs. Additionally, the system provided pulse-pair resolution to the nearest 1 µs, and full waveform analysis (Currie et al., 1998). 7.5.1. Background as a Poisson process; expected characteristics Nuclear decay can be described as a Bernoulli process, which, in the limit of a large pool of decaying nuclei each with a small but fixed decay probability, can be approximated as a Poisson process. Measurements of long-lived radionuclides are therefore expected to follow: (1) the Poisson distribution of counts, which is asymptotically normal, but discrete and positively skewed for small numbers of counts, and for which the expectation of the Index of Dispersion (variance/mean) is unity (Cox and Lewis, 1966); (2) the Uniform distribution of arrival times; and (3) the Exponential distribution of inter-arrival times, also positively skewed. Thus, we
Detection and quantification capabilities in nuclear analytical measurements
109
have the situation where skewed data distributions are grounded in the very nature of the physical process; and tests of the fundamental assumptions for actual measurement processes may be made by performing significance tests on the corresponding empirical data distributions. For low-level measurements of radioactivity, where background is dominant, one finds that the assumption of an underlying Poisson process is commonly extended to the background radiation. The validity of such an assumption is crucial for making low-level detection decisions and in estimating detection limits. It is central also, in the “propagation of uncertainty” for low-level radioactivity results. Problems are not expected for background components derived from long-lived radioactive contaminants. If, however, there are other significant background components, in particular cosmic radiation and background artifacts characteristic of the real, laboratory measurement process, the assumption should be approached cautiously. A comprehensive investigation of the validity of the Poisson assumption for low-level, anticoincidence background radiation was carried out with the NIST individual pulse analysis system, with 40 cm3 cylindrical gas counters operated in the Geiger mode (Currie et al., 1998). The individual pulse analysis capability of the NIST system is unique, in that it captures the shape and time of arrival of individual low-level coincidence (muon) and anticoincidence pulses with the capability of time stamping each counting event and accumulating up to 105 events in a single file. Without such a capability it would not be possible to construct the arrival and inter-arrival time distributions. Essential characteristics of the system are: (1) ca. 80% of the gross background was eliminated by surrounding the counting tube with 25 cm of pre-nuclear era steel (World War I naval gun barrel); (2) the bulk (98%) of the residual background (ca. 20 cpm), due to penetrating cosmic ray muons, was “canceled” by anticoincidence shielding. This is achieved by placing the sample counting tube within an outer cylindrical tube (guard counter), such that an external cosmic ray (muon) which penetrates both tubes results in a coincidence pulse, whereas the lower energy net background (BG) radiation in the inner sample tube does not. The fundamental assumptions to be tested are whether the muon (coincidence) background and the net (anticoincidence) background can be treated as independent Poisson processes. 7.5.2. Empirical distributions and tests of the Poisson (background) hypothesis Distributional highlights from the NIST study are given in Figures 19–21. The basic information is generated from the individual pulse arrival and inter-arrival times, as shown in Figure 19 which represents a 150 s glimpse of the pulse data stream (arrival times) for coincident (C) and anticoincident (A) events from a 980 min background counting experiment. (The event labeled “G” represents a rare giant pulse, atypical of the ordinary GM process.) Considering first the Poisson distribution of counts and the corresponding count rates, we examined results for C (muon) and A (net background) events spanning a period about sixteen thousand times as long as that shown in Figure 19, nearly a month (7 May–5 June 1997). At the beginning of this period the GM counting tube was filled with purified counting gas; following that, background (C and A) counts were aggregated for 21 individual counting periods, 16 of which has relatively uniform counting times of about 1600 min. The Poisson hypothesis was tested with χ 2 , considering the ratio of the observed variance to the theoretical variance based on the number of counts accumulated in each counting period. (This variance ratio can be viewed as a weighted estimate for the index of dispersion Iw .) The A-events passed the test (Iw = 1.36, p = 0.16) but the C-events failed (Iw = 4.92, p < 0.000001).
110
L.A. Currie
Fig. 19. A 150 s snapshot of the individual pulse data stream from the low-level GM counting tube, showing times of arrival (TOA, x-axis) and pulse amplitudes (z-axis), extracted from a 980 min background measurement during May 1997. Two anticoincidence pulses (A) occurred during this interval, the remainder being coincidence pulses (C). DT represents the interval between successive anticoincidence background events, while dt represents the interval between anticoincidence events and preceding (coincidence) events. (DT and dt distributional analysis for the entire, 588 A-count sequence is shown in Figure 21.)
Insight is gained by looking beyond the summary statistics. The presumably random, symmetric distribution of the 1600 min C-background counts, shown in Figure 7a, is repeated in Figure 20a, together with further exploration in two additional dimensions: time (Figure 20b) and barometric pressure (Figures 20c, 20d). The time series shows immediately that the extra C-background variance (above Poisson) is not due simply to an added component of normal, random error. Rather, a distinctive quasi-periodic pattern is evident, well in excess of the Poisson–normal standard uncertainty bars. In fact, the source of the non-Poisson, non-random C-background variance represents an interesting blend of meteorology and physics, with time as surrogate: increased barometric pressure leads to increased attenuation of mesons in the atmosphere. Removal of the barometric pressure effect reduces Iw to 2.19. Arrival time distributions for the C- and A-pulses (not shown) from the 980 min experiment referenced in Figure 19 showed no significant difference from a Poisson process (uniform distribution), but inter-arrival times were a different matter. The inter-arrival times (DT) for the low-level anticoincidence background did show a good fit (p = 0.40) to the exponential distribution as shown in Figure 21a, but the A vs C inter-arrival times (dt) did not. As seen in Figure 21b, there was an excessive number of very short intervals falling within the first (1 s) histogram bin. Upon expansion of the time scale, it was found that the excess was concentrated almost entirely between dt = 150 µs and dt = 350 µs (43 counts). The physics of this phenomenon is well known. It relates to a counting process artifact whereby a secondary event (“afterpulse”), which occurs within a few hundred microseconds of the primary event, is occasionally generated by a complex photo-electric process within
Detection and quantification capabilities in nuclear analytical measurements
111
Fig. 20. Upper left (a): frequency distribution of gross background pulses in the NIST low-level gas counting system. (Each of the 16 background measurements corresponds to a counting period of 1600 min.) Although the histogram appears symmetric that does guarantee that the background events are normally distributed. The time series of background events (b, upper right) reveals a non-random, quasi-periodic structure significantly in excess of the bounds of the Poisson–normal error bars. (c, lower left) and (d, lower right) demonstrate a clear dependence on barometric pressure. (N and p curves represent the count rate and barometric pressure time series in (c), and the y- and x-axes in (d).)
the counting tube. For the NIST system the afterpulse probability is ca. 0.1%. Because of their absolute abundance and time constant, the spurious pulses are scarcely detectable in the DT distribution. They have a good chance of escaping cancellation during the C-pulse time gate, however, and appearing in the dt distribution where intervals may be as short as the intrinsic deadtime of the GM tube (≈140 µs). Since the primary C-pulses are more abundant than the valid A-pulses by a factor of 70, the contamination probability is raised to ca. 7% for the anticoincidence background. A further, interlaboratory complication is that the afterpulse contamination of the A-background depends on the electronic deadtime of the low-level counting system. Conclusion. Fundamental (theoretical) skewed data distributions are firmly established for radioactive decay, in the form of the Poisson distribution of counts and the exponential distribution of inter-arrival (decay) times. Extension of this model to low-level background radiation is widely assumed, but it does not necessarily follow, especially for the major gas counting background component (cosmic ray muons) and for the extremely important low-
112
L.A. Currie
Fig. 21. Empirical (exponential) distribution of inter-arrival times: (DT, left) between anticoincidence background events, and (dt, right) between anticoincidence and preceding coincidence events. The histogram on the left (a) summarizes DT inter-arrival times between zero and 1000 s; mean = 100.7 s, equivalent to a mean background rate of 0.60 cpm (total for the two GM counters that were operational during the 980 min period). The fit is good (p = 0.40). The histogram on the right (b) summarizes dt inter-arrival times between zero and 30 s; mean = 2.72 s, equivalent to a mean rate of 22.0 cpm. The fit is not good (p = 0.0006), showing excessive counts in the first 1 s class (histogram bin).
level anticoincidence background component. Because of the enormous excess of C-events over A-events in low-level anticoincidence counting, it is essential to provide extremely effective muon (anticoincidence) shielding and afterpulse control via special timing and pulse shape circuitry. Note that the afterpulse artifact was manifest as a departure from the theoretical skewed (exponential) data distribution. The non-random character of the muon background, together with its effect on afterpulse amplification, makes it clear that strict validity of the Poisson assumption for counting background requires systems where “meson leakage” (Theodórsson, 1992) can be minimized by extremely efficient anticoincidence shielding, or better, by largely eliminating the cosmic ray meson background by going deep underground (CELLAR, 2004).
8. Two low-level data quality issues 8.1. Detection limits: intralaboratory vs interlaboratory perspectives The issue of single laboratory vs multi-laboratory Detection and Quantification Limits has, on occasion, become contentious, resulting in what might be considered “two cultures”. Some of the statistical and conceptual issues, and controversy involved may grasped from an article by
Detection and quantification capabilities in nuclear analytical measurements
113
Fig. 22. Sampled (S) and target (T ) populations.
Gibbons (1995), which treats the detection of environmental pollutants. Of particular interest are insightful and provocative contributions from the Discussants. The purpose of this brief section is to give some perspective on the two views—both of which are correct, and incorrect; and to suggest how they might be reconciled. To begin, it is useful to consider the error structure of the Compound Measurement Process which, in its simplest manifestation, has been presented by Natrella (1963). In this view, the measurement process is treated in terms of two populations which represent, respectively, the population (of potential measurements) actually sampled (S-population), and that which would be sampled in the ideal, bias-free case (T -population). The corresponding S and T populations are shown schematically in Figure 22, for a two-step measurement process. When only the S-population is randomly sampled (left side of the figure), the error e1 from the first step is systematic while e2 is random. In this case, the estimated uncertainty is likely to be wrong, because (a) the apparent imprecision (σS ) is too small, and (b) an unmeasured bias (e1 ) has been introduced. Realization of the T -population (right side of the figure) requires that all steps of the MP be random—i.e., e1 and e2 in the figure behave as random, independent errors; T thus represents a Compound Probability Distribution. If the contributing errors combine linearly and are themselves normal, then the T -distribution also is normal. The concept of the S and T populations is absolutely central to all hierarchical measurement processes (Compound MPs), whether intralaboratory or interlaboratory. Strict attention to the concept is essential if one is to obtain consistent uncertainty estimates for Compound MPs involving different samples, different instruments, different operators, or even different methods. In the context of (material) sampling, an excellent exposition of the nature and importance of the hierarchical structure has been presented by Horwitz (1990). From the perspective of a single laboratory, the S-population represents that for which the type-“A” uncertainty component can be estimated via replication. The link to the T -
114
L.A. Currie
population, and the total uncertainty, is the uncertainty resulting from the unmeasured (nonsampled) distribution (upper left, error e1 ). Difficulties arise when that uncertainty component is ignored, and the single laboratory’s result is accompanied only by the replication-based type-“A” uncertainty. In particular, uncertainty (confidence) intervals would be unrealistically small, and derived values for detection and quantification limits would be overly optimistic. The approach taken by the National Metrological Institutes (NMIs) leads to the resolution of the problem. That is, the mandated policy for all NMIs is that all uncertainty components must be taken into account, for all measured results. The process begins with a complete uncertainty (error) budget for the measurement process, and all non-sampled component(s), deemed type-“B” uncertainties, must be estimated and combined with the “A” components to provide a total combined standard uncertainty (uC ). The expanded uncertainty (U = kuC ) may be given, using a “k”, commonly 2, to generate an approximate 95% confidence interval. Note that neither absolute nor relative uncertainties are necessarily the same for different instrumentation or laboratories. Each measured result, when accompanied by a valid combined standard uncertainty, is expected to be consistent with the “truth” or assigned value for the measurand. A complete presentation of the principles and methods for deriving combined standard uncertainties may be found in the “Guide to the Expression of Uncertainty in Measurement” (ISO-GUM) (ISO, 1995). From the interlaboratory perspective, the first population in Figure 22 (e1 ) would represent the distribution of errors among laboratories; the second [S] would reflect intralaboratory variation (“repeatability”); and the third [T ], overall variation (“reproducibility”) (ISO, 1993). If the sample of collaborating laboratories can be taken as unbiased, representative, and homogeneous, then the interlaboratory “process” can be treated as a compound MP. In this fortunate (doubtless asymptotic) situation, results from individual laboratories are considered random, independent variates from the compound MP population. For parameter estimation (means, variances) in the interlaboratory environment it may be appropriate to use weights—for example, when member laboratories employ different numbers of replicates (Mandel and Paule, 1970). As in the case of the individual laboratory following the ISO Guide (ISO-GUM), unbiased T -distribution results would be expected from the multilaboratory consortium that satisfies the foregoing conditions. In these very special circumstances, the different perspectives of the two cultures should be reconciled—each providing meaningful measurement uncertainties and unbiased estimates of (detection, quantification) measurement capabilities. For detection decisions and limits, a small problem exists when there are residual systematic error components—intra- or inter-laboratory. If a measurement error is fixed—e.g., through the repeated use of a faulty instrument or calibration source, or background- or blank- or σ -estimate, then the ideal random error formulation of hypothesis testing cannot apply. For a single measurement, the question is moot; but for a suite of data from the same (biased) laboratory, or sharing the same numerical estimate of the blank, or imprecision, or calibration factor, then the independence (randomness) assumption is not satisfied. A conservative treatment of the problem, resulting in inequalities for α and β, is given in Currie (1988). In Figure 23, we attempt to reconcile the intra- and inter-laboratory environments, by presenting an explicit partitioning of error components that can be grasped from both perspec-
Detection and quantification capabilities in nuclear analytical measurements
115
Fig. 23. Partitioning of method, interlaboratory, and intralaboratory error. (Adapted from Currie (1978).)
tives. Here, we see that specification of the Performance Characteristics of a compound or hierarchical MP depends upon one’s viewing point or position in the hierarchy. That is, at least for the “tree” structure, all segments below the viewing (or “null”) node consist of multiple branches or replicates—essential for direct assessment of random error. Only a single path lies above the null node; this path necessarily fixes the bias of the MP. By moving up in the hierarchy, one has an opportunity to convert bias into imprecision—put another way, what is viewed as a fixed (albeit unknown, perhaps “type-B”) error at one level of a compound MP, becomes random at a higher level. This is very important, for random error may be estimated statistically through replication, but bias may not; yet inaccuracy (total error) necessarily comprises both components. Collaborative or interlaboratory tests, which under the best of circumstances may be found at the uppermost node of the Compound MP, provide one of the best means for accuracy assessment. In a sense, such intercomparisons epitomize W.J. Youden’s recommendation that we vary all possible factors (that might influence analytical results), so that the observed dispersion can give us a direct experimental (statistical) measure of inaccuracy (Youden, 1969). The basic concept, as indicated in Figure 23, is that fixed intralaboratory biases are converted into random errors from the interlaboratory perspective. If the overall interlaboratory mean is free from bias, then the observed interlaboratory dispersion is the measure of both imprecision and inaccuracy. An apt metaphor that has been applied to the two perspectives is that the intra-laboratory approach (following ISO-GUM) uses a “bottom up” evaluation of total uncertainty, whereas the inter-laboratory approach uses “top down” total uncertainty evaluation. Guidelines to the derivation of detection (quantification) limits that would apply to the
116
L.A. Currie
individual laboratory situation have been prepared by IUPAC (1995, 1998) and ISO (1997, 2000), while ASTM (1997, 2000) documents treat the interlaboratory situation. Total reconciliation between the two cultures is guaranteed, if the two approaches are routinely combined, somewhat in the spirit of the NMI “key comparisons” policy. That is, (1) individual laboratories would be advised always to consider the complete error budget for their specific MPs, taking into account both type-A (statistical) and type-B (experiential, theoretical) errors in assessing detection, quantification capabilities, and (2) at the same time, the consortium of such laboratories should adopt the Youden philosophy, to utilize interlaboratory data, to provide a cross-check on the intralaboratory type-B uncertainty bounds. If the two approaches are self-consistent, meaningful estimates of uncertainty and measurement limits can be expected; if not, research into the source(s) of the discrepancy is called for.28 8.2. Reporting of low-level data 8.2.1. Statement of the problem; values and non-values Quantifying measurement uncertainty for low-level results—i.e., those that are close to detection limits—deserves very careful attention: (a) because of the impact of the blank and its variability, and (b) because of the tendency of some to report such data simply as “nondetects” (Lambert et al., 1991) or “zeroes” or “less than” (upper limits). The recommendations of IUPAC (1995, 1998) in such cases are unambiguous: experimental results should not be censored, and they should always include quantitative estimates of uncertainty, following the guidelines of ISO-GUM (ISO, 1995). When a result is indistinguishable from the blank, ˆ with the critical level (LC ), then it is important also based on comparison of the result (L) to indicate that fact, perhaps with an asterisk or “ND” for not detected. But “ND” should never be used alone. Otherwise there will be information loss, and possibly bias if “ND” is interpreted as “zero”. Data from an early IAEA intercomparison exercise on radioactivity in seawater illustrates the point (Fukai et al., 1973). Results obtained from fifteen laboratories for the fission products Zr-Nb-95 in low-level sample SW-1-1 consisted of eight “values” and seven “non-values” as follows. (The data are expressed as reported, in pCi/L; multiplication by 0.037 yields Bq/L.) Values: 2.2 ± 0.3 9.5 9.2 ± 8.6 77 ± 11 0.38 ± 0.23 84 ± 7 14.1 0.44 ± 0.06. Non-values: <0.1 ND <7.3 ND <40 <20.9 <26.8. The data as a whole is uninterpretable, particularly the “non-values”. A simple mean and standard uncertainty calculated from the “values” is 25±12 pCi/L. Because of missing uncertainty information, however, a weighted mean of the “values” cannot be calculated. Also, in several cases the meanings of the stated uncertainties or upper limits were not made clear. Ironically, for this particular sample, the true value of the activity concentration for the radionuclide pair was negligible because of unintended shipment delays that were long compared to the longer half-life (95 Zr, 65 d). 28 The proposed combined approach is inspired, in part, by the NMI philosophy for establishing traceability and
certifying reference materials (ISO, 1989a, 1989b).
Detection and quantification capabilities in nuclear analytical measurements
117
8.2.2. Virtual reality: bias and uncertainties in reporting of low-level data There is near universal agreement that results of measurements and their uncertainties should be reported for all experimental data, including data in the region of the detection limit and below (ASTM, 1997, 2000; ISO, 1995; IUPAC, 1998). Even in the case of what might be considered subliminal data (non-detects, nd), the act of measurement adds information that ought to be preserved by avoiding truncation or censoring, if possible. An abundance of non-detects reflects a measurement process having inadequate detection/quantification capabilities, or excessive censoring; and the cost, when using certain substitution rules, can be considerable. Two recent case studies make the point (Currie, 2000b). In the first, “virtual absence” in an emissions database was guaranteed by the substitution of zeroes for some 80,000 nd trace organic field data from small scale ordnance disposal tests. The cost: the database was intended as input for modeling the impact of large scale environmental disposal by open burning or detonation. The counter example (“virtual presence”) showed a low-level nuclear waste repository “filling up with virtual radioactivity”. In this case positive bias occurred when upper limit symbols, for shipments containing little radioactivity, were stripped as the data were entered into the inventory database. In certain cases of censored data or questionable data, robust estimators and data completion schemes may be applied (Cohen, 1991; Hoaglin et al., 1983); in others, human or software or hardware discriminators operate on the data as they are generated, inserting substitute values for non-detects. Such a process is strongly discouraged, but if it occurs, the resulting “nonvalues” need to be flagged, and bounds for the resulting bias need to be indicated. In this context, a tentative approach to the assessment of substitution bias functions is given. Substitution bias functions. Taking x and xC to represent the observed (non-censored) result and the critical value, respectively, we consider three common practices, plus an alternative having lower average bias: x1 , substitution of zeroes for non-detects; x2 , substitution of the critical value xC ; and x3 , substitution of xC /2. The new alternative x4 is defined as the value that would produce zero bias for a true value equal to xC /2, the motivation being to better balance positive and negative non-detect bias. Assuming normality and taking xC = 1.645σ , we find x4 ≈ 0.28xC . For the present exercise, σ is assumed constant—i.e., equal to σo , over the range covered in Figure 24. If all observable values x were to be reported for a low-level concentration having true value m, then the resulting mean would be unbiased. That is, ! xϕ(x − m) dx = m, (8.1) where ϕ(x − m) is the normal density function for a random variable having mean m and unit variance. A logical motivation for a simple substitution rule would be to take a value x , when x xC , such that the resulting bias is zero when m takes on a value m that is “guessed”, or “believed” in the Bayesian sense, to be likely. This is equivalent to taking the variable x out of the integral, x
"xc
−∞
"xc
ϕ(x − m ) dx = −∞
xϕ(x − m ) dx.
(8.2)
118
L.A. Currie
Fig. 24. Bias functions for non-detects. b(x , m), which equals bias in units of σ , is plotted for four substitution choices (x ) vs m/xC , which equals the true mean in units of xC . Substitution values, when x xC , are (top to bottom): x = xC , xC /2, 0.28xC , and 0, where xC = 1.645σ . (NB: “1” on the abscissa corresponds to the critical value; “2”, to the detection limit; σ = σo assumed constant.)
When Equation (8.2) is an equality, x substitution below xC is unbiased. Replacing m by m, the bias function b(x , m), which equals bias in units of σ , is given by the difference between the left and right sides of Equation (8.2). The resulting four functions are shown in Figure 24. It is seen that x = 0 (bottom curve) gives negative bias for all values of m except those very close to zero (below about 0.2σ ). Maximum (absolute) bias (b ≈ −0.5) occurs for m ≈ 1.2σ , or 0.7xC . The other three x’s have maximum (absolute, and positive) bias when m = 0. The rule x = xC (top curve) is unattractive, since its bias is always positive, and rather large for m xC . The rule x = xC /2 is quite good, producing negligible bias for m > xC . For smaller m it leads always to positive bias, becoming practically significant for absent analytes. The alternative rule, where x ≈ xC /4 (solid line) is attractive in that its average bias is small and its bias extrema, more acceptable at b− = −0.20 and b+ = +0.54. It is perhaps close to optimal in the Bayesian sense, if the m values for non-detects are believed to be uniformly distributed between 0 and xD . Impact of hardware/software thresholds. Major negative bias may result from data-destroying thresholds or discriminators that are set well in excess of the intrinsic critical level. Such thresholds are not uncommon in instrumental measurement systems, particularly chromatographic systems designed for monitoring organic pollutants, where the threshold discriminator rejects signals below a set point, and returns non-values, or zeroes.29 Although such an oper29 Because of the widespread use of “black box” chromatographic monitoring systems for the detection of organic
pollutants in air and water, it seems likely that the severe negative bias problem may be responsible for the spate of
Detection and quantification capabilities in nuclear analytical measurements
119
ation can be viewed as applying xC σo , such that α 0.05, the substitution of zero for an observed value xˆ and its uncertainty makes for severe negative reporting bias. Referring to the discussion of the impact of thresholds in Section 4.3.4, with a slight change in notation (xT in place of ST ), it was seen that setting xT far above the 0.05 critical level xC caused a significant degradation in detection capability. Failure to report numerical values (“non-detected” observations, x) ˆ compounds the problem, by creating negative bias. We consider here the magnitude of that bias, based on Equations (8.1) and (8.2), and the chromatographic data mentioned in Section 4.3.4 (Kurtz, 1985). The measurement problem in that case was the detection of residues of the synthetic pesticide, fenvalerate, in chickens and eggs. Electron capture gas chromatographic data were produced from a system that included a “black box” hardware/software peak integrator, that had a built-in threshold that returned nonzero areas only for peaks exceeding 1.00 cm2 —equivalent to a fenvalerate mass (xT ) equal to 42 pg. (The measurement precision at this artificial, de facto critical level, was significantly better than that of the conventional quantification limit!) By comparison, using σo and the variance function, with α = β = 0.05, the intrinsic critical value and detection limit were 2.0 and 4.2 pg, respectively. With zero-substitution for non-reported signals, the xT -induced negative bias function equals "xT b(0, m) = −
xϕ(x − m) dx.
−∞
Thus, when the actual amount m is well below the threshold, e.g., −3σ , the negative bias is equal to the mean—i.e., b(0, m) = −m (0 m xT ). For m = xT , and xT σ , the bias approaches −m/2, and it vanishes when m approaches zero and when m xT . For the specific fenvalerate measurement process, taking into account the calibration and variance functions, the maximum bias (−37 pg) occurs for m ≈ 37 pg, and it becomes negligible for m > 49 pg. Thus, the combination of an artificially high threshold and the suppression of data below that threshold has resulted in a maximum negative bias nearly nine times the detection limit!
Appendix A: Worked examples, and critical valves of test statistic A.1: Examples of applications The following subsections give an overview of selected applications, illustrating some critical issues arising in the estimation of detection and quantification limits, together with references to source literature. For the two examples, based on actual data for “simple” low-level counting and gamma-ray peak detection, respectively, expanded treatments showing numerical details (“worked examples”) are given in Sections A.3 and A.4. zeroes in the organic emissions database discussed at the beginning of this section. Also, it may partly underlie the significant current controversy concerning the meaning and use “MDL” (method detection limit) in connection with water pollution monitoring (Waterscience, 2005).
120
L.A. Currie
Table A1 Selected critical values of test statistics used in this chapter Degrees of freedom (ν)
t (0.95, ν)a
1 2 5 10 20 50 100 ∞
6.314 2.920 2.015 1.812 1.725 1.676 1.661 1.645
δ(0.95, 0.95, ν)
χ 2 (0.05, ν)
χ 2 (0.95, ν)
5.516 3.870 3.543 3.408 3.335
0.0039 0.1026 1.15 3.94 10.85 34.76 77.84
3.84 5.99 11.07 18.31 31.41 67.60 124.50
a z(0.95) = t (0.95, ∞).
Special note: Because of the tutorial, numerical intent of the examples in Sections A.3 and A.4, all calculated results are given to at least one decimal digit. The objective is to preserve computational (numerical) accuracy even though the convention concerning significant figures may be violated. By contrast, observed counts will always be integers because of the discrete nature of the Poisson process. In practice, of course, final results should always be reported to the correct number of significant figures. Selected critical values of test statistics used in this chapter are given in Table A1. A.2: Overview Low-level beta counting (39 Ar) Issues (Currie et al., 1998): Normal background (small RSD) Tests of Poisson hypothesis, randomness Heteroscedasticity, non-central-t Combined uncertainty (concentration domain) Two cases are presented using anticoincidence data (Set I) and coincidence data (Set II) from Currie et al. (1998). For Set I data the background variance is found to be consistent with that of a Poisson process and the observed counts are sufficiently large that the Poisson– normal approximation can be used for detection decisions and for the calculation of detection and quantification limits. The Set II data illustrate the contrasting situation where there is an extra variance component, requiring the use of the replication variance estimate s 2 and information drawn from the t (central and non-central) and χ 2 distributions. This is a common occurrence for non-anticoincidence, background limited, low-level counting where the primary background component derives from an external process such as cosmic ray interactions. For the Set II data, the background variance was primarily due to cosmic ray variations, which in the short term (intervals of a few days) are not even random, reflecting short term trends in barometric pressure. Peak detection (γ -ray spectrometry) Issues (Currie, 1997): Algorithm dependence Background, baseline, blank
Detection and quantification capabilities in nuclear analytical measurements
121
Null hypothesis dominance Multiple detection decisions Low-level data reporting Several of the above issues are illustrated in the numerical example given in Section A.4. Multiple detection decisions, mentioned but not covered in the example, require an adjustment of z or t to take into account the number of null hypothesis tests, because the overall probability for one or more false positives when k-tests are made is 1 − (1 − α)k ≈ kα
(α 1).
(A.1)
Thus, to maintain an overall false positive probability of 0.05, one must use 1-sided z(α ) or t (ν, α ), where α ≈ 0.05/k. For k = 5 independent tests, for example, z(0.01) = 2.326, and t (9, 0.01) = 2.821. In the latter case, it is also necessary to obtain an independent estimate of σo for each application of the t-test.1 Extensions The applications discussed above highlight some of the more common issues involving detection, quantification, and data reporting for low-level, nuclear-related measurements as found in the “simple” scalar, sample-background situation (see Section A.3) and simple peak detection, including multiple detection decisions and the treatment of background vs baseline vs blank (see Section A.4). Although the influence of interfering components is illustrated by the effect of an elevated baseline on isolated peaks, this latter example does not address more complicated cases such as overlapping peaks, where the impact of interference is amplified by a “variance inflation factor” due a poorly conditioned inverse matrix. Some discussion of this topic may be found in IUPAC (1995) and Currie (1997). Other noteworthy situations which will not be illustrated here include: (1) calibration-based detection limits, which capture the covariance between the estimated blank (as intercept) and sensitivity (as calibration factor or slope) as well as the variance function (dependence of σ 2 on concentration); and (2) more complicated functional and blank models, which arise in environmental isotope ratio measurements such as accelerator and stable isotope mass spectrometry, where non-linear blank correction functions are the rule, and multivariate and nonnormal blanks are often encountered. Some discussion of these advanced topics is given in Currie (1997, 2001). Numerical examples for calibration-based detection limits, based on real experimental data, may be found in ISO (1997) and Currie (1985b). A.3: Example 1: low-level beta counting (39 Ar) The first example illustrates the treatment of simple counting data for making detection decisions and the estimation of detection and quantification limits in the signal (counts) and radioactivity concentration (Bq/mol) domains. Two sets of actual background counting data, drawn from Currie et al. (1998), are used to illustrate the Poisson–normal case, where the standard deviation is (1) well represented by the square root of the mean number of counts (Set 1 data) and (2) based on replication, due to the presence of non-Poisson variance components (Set 2 data). Set 1 data are in fact the observed number of anticoincidence counts in ten, 1000 min counting periods; Set 2 are the corresponding cosmic ray induced coincidence counts.
122
L.A. Currie
A.3.1: The data (counts; t = 1000 min) Set 1: 265 305 277 277 263 312 310 318 286 270 Mean (B) = 288.3, √ Poisson-σB = 288.3 = 17.0, Replication-sB = 21.0. Using the χ 2 test, we find (s/σ )2 to be non-significant (p = 0.13), so the Poisson value for σB will be used. Set 2: 18730 19310 19100 19250 19350 19210 19490 19190 19640 18780 Mean (B) = 19205.0, √ Poisson-σB = 19205.0 = 138.6, Replication-sB = 283.2. Using the χ 2 test, we find (s/σ )2 to be very significant (p < 0.0001), so the replication value sB must be used as an estimate for σB . A.3.2: Equations for Poisson–normal variables (counts) ˆ Sˆ = y − B, ˆ = V (y) + V (B) ˆ = S + B + B/n = S + Bη, V (S) √ σo = (Bη) . . . OR . . . so = sB η [TEST: χ 2 ],
(A.2)
LC = zC σo
(A.5)
. . . OR . . .
LC = tC so ,
LD = LC + zσD = LC + z(z + σo ), . . . OR . . . LD ≈ δσo ≈ 2LC , where δ ≈ 2t 4ν/(4ν + 1) ≈ 2t, ν = n − 1, # $ LQ = kQ σQ = kQ (kQ /2) 1 + SQRT 1 + (2σo /kQ )2 , $ # LQ = 10σQ = 10 5 1 + SQRT 1 + (σo /5)2 .
(A.3) (A.4) (A.6a)
LD = z + 2zσo 2
(A.6b) (A.7a) (A.7b)
Note that the facts that σD > σo (Equation (A.6)) and σQ > σo (Equation (A.7)) derive from the variance function for Poisson counting data (Equation (A.3)). Also, when this is applied to actual counting data, a subtle but necessary approximation is made by calculating the Poisson variances from observed counts which serve as estimates for the (unobservable) population means, S and B. Parameter values for the current example are: n = 10,
ν = n − 1 = 9,
α = β = 0.05,
η = (1 + 1/n) = 11/10 = 1.10,
kQ = 10,
z = 1.645, t = 1.833, δ = 3.57, √ √ σo = σB η = σB 1.10,
√ √ so = sB η = sB 1.10.
Detection and quantification capabilities in nuclear analytical measurements
123
Note that LD and LQ require information on the dependence of the variance σ 2 on signal (or concentration) level (variance function). For the Set-1 data, this is given directly from Poisson “counting statistics” where the variance is equal to the (true) mean number of counts. For the Set-2 data, where we must use s 2 to estimate σ 2 , one really needs full replication data as a function of signal (or concentration) level. For the purpose of this example, however, we shall assume that the variance is approximately independent of concentration over the range of interest—i.e., σ ≈ σo for L LQ . In practice, such an assumption should always be tested. A.3.3: Signal domain (all units are counts) Set-1 data √ σB = 288.3 = 17.0, sB = 21.0 [χ 2 : NS], √ √ therefore: σo = σB η = 17.0 (11/10) = 17.8. SC = 1.645σo = 29.3, SD = 2.71 + 3.29σo = 61.3, SQ = 10σQ = 10(5) 1 + SQRT 1 + (σo /5)2 = 10(23.49) = 234.9. In the case of SQ , we see that the variance function for this particular Poisson process yields a ratio (σQ /σo ) of 1.32. New observation: y = 814; then Sˆ = 814 − 288.3 = 525.7. Sˆ exceeds SC , so “detected”. Report: Sˆ = 525.7, u = (814 + [288.3/10])1/2 = 29.0 [relative u = 5.5%]. Set-2 data
√ 19205.0 = 138.6, sB = 283.2 χ 2 : significant (p < 0.00001) , √ √ therefore: so = sB η = 283.2 (11/10) = 297.0 (σ assumed constant for S SQ ; homoscedastic). σB =
SC = 1.833so = 544.4, SD = δσo = 3.57σo , SˆD = 3.57so = 1060.3, SQ = 10σQ ≈ 10σo , SˆQ = 10so = 2970.0. The relative (standard) uncertainty for SˆD and SˆQ derive from that for so /σo . Rigorous bounds 2 for the latter can be obtained √ from the χ distribution, but a good approximation for ν > 5 is given by u(so /σo ) ≈ 1/ (2ν); for 9 degrees of freedom, this gives u = 23.6%. The minimum possible value for σo , of course, is the Poisson value (Currie, 1994). New observation: y = 19004; then Sˆ = 19004 − 19205.0 = −201.0. Sˆ does not exceed SC , so “not detected”. Report: Sˆ = −201.0, u = so = 297.0.
124
L.A. Currie
A.3.4: Concentration domain (Bq/mol)2 ˆ ˆ A, ˆ xˆ = (y − B)/A = S/ Aˆ = t Vˆ Eˆ = (60000 s)(0.000252 mol)(0.939) = 14.19 mol/Bq, where t is the counting time; V , the moles of Ar in the counting tube; and E, the counting efficiency for 39 Ar. # 2 2 $ uc (A)/A = SQRT u(V )/V + u(E)/E $ # = SQRT [0.52%]2 + [1.28%]2 = 1.4%. Set-1 data xˆD = 61.3/14.19 = 4.32 (uc = 1.4%), xˆQ = 234.9/14.19 = 16.55 (uc = 1.4%). New observation: xˆ = 525.7/14.19 = 37.0, # $ uc = SQRT [5.5%]2 + [1.4%]2 = 5.7% or 2.1 Bq/mol. Set-2 data xˆD = 1060.3/14.19 = 74.7, xˆQ = 2970.0/14.19 = 209.3. For the set-2 data, the relative uncertainties of xD , xQ√are dominated those SD , SQ , shown √ in Section A.3.3 to be approximately 1/ (2ν) = 1/ 18, or 23.6%. Combining this with the 1.4% relative uncertainty for A, as above, has little effect; the combined relative uncertainty for the concentration detection and quantification limits is still about 23.6%. See IAEA (2004) for a discussion of additional error components that must be considered for low-level β counting. New observation3 : x = −201.0/14.19 = −14.2, uc = 297.0/14.19 = 20.9. A.4: Example 2: gamma-ray peak detection The second example treats detection and quantification capabilities for isolated (“baseline resolved”) gamma ray peaks. The data were derived from a one hour measurement of the gamma ray spectrum of NIST Mixed Emission Source Gamma Ray Standard Reference Material (SRM) 4215F. The full spectrum for this SRM contains several major pure radionuclide peaks used for calibration of the energy scale. For the purposes of this example, however, we consider only the high energy region dominated by the Compton baseline plus 60 Co peaks at 1173 and 1332 keV and the 88 Y peak at 1836 keV. Within this region, shown in Figure A1,
Detection and quantification capabilities in nuclear analytical measurements
125
60 88 Fig. A1. SRM 4215F: γ -ray spectrum segment showing √ energy calibration ( Co [1173, 1332 keV], Y [1836 keV]) and impurity (40 K [1460 keV]) peaks. (Y -axis equals counts.)
there is a small impurity peak at 1460 keV due to 40 K, the subject of this example. The ordinate in Figure A1 uses the Poisson variance stabilizing transformation, counts1/2 , corresponding to a constant y-standard deviation of 0.50, which is 5% of the smallest y-axis interval (tic mark) shown. (In the figure, the y-axis ranges from 0 to 200, corresponding to a count range of 0 to 40,000 counts; and the x-axis ranges from channel 1100 to 2100, with the last datum in channel 2048.) Peak channel gross counts are 26184 (1173 keV), 20918 (1332 keV), 581 (1460 keV), and 8881 (1836 keV). The baseline in the region of the 40 K peak is ca. 265 counts/channel. The data shown in Figure A1 will be used to estimate detection and quantification limits for 40 K under the three “B” limiting conditions (background, baseline, blank), and to illustrate the reporting of low-level peak area data. To focus on these issues we limit the discussion here to the signal domain (peak area detection, quantification, and estimation). See IAEA (2004) for a discussion of additional error components that must be considered in radionuclide or element uncertainty estimation for gamma-ray spectrometry and neutron activation analysis. A.4.1: Background limiting This case is exactly analogous to the low-level counting example above, except that signal and background are here based on the summation of counts over a k-channel spectrum window. Given the approximate peak width of 3.5 channels (FWHM) in this region of the spectrum, we take a 6-channel window centered at the 40 K peak location for background and net peak area estimation. For a 60 min counting period, this gives a long-term average B = 3.6 counts for the 6-channel window. For the purposes of this example, we take this to be the “well-known” background case, where B has negligible uncertainty. Unlike the first example, the expected value of the background is so small that the normal approximation to the Poisson distribution is inadequate. Also, because of the discrete nature of the Poisson distribution, we must use the inequality relationship indicated in Equation (7.4), such that α 0.05. For a mean of 3.6 counts, this gives a critical (integer) value for the
126
L.A. Currie
gross counts of 7, for which α = 0.031. (For 6 counts, α = 0.073.) Using 7 counts for the gross counts critical value, Equation (7.5) gives 13.15 as the gross counts detection limit, for which β = 0.05. Subtracting the background of 3.6 gives the corresponding net signal (peak area) values for SC (3.4 counts) and SD (9.55 counts). Had the √ normal approximation to the Poisson distribution been used, we would have obtained 1.645 3.6 = 3.12 counts for SC , √ and 2.71 + 3.29 3.6 = 8.95 counts for SD . The net count quantification limit SQ is given by 50[1 + (1 + B/25)1/2 ] = 103.5 counts. In this case B is sufficiently small that it has little influence on the value of SQ . (See also Table 5, Section 7.4.) A.4.2: Baseline limiting Detection and Quantification capabilities under background limiting conditions tend to be overly optimistic, reflecting interference-free conditions. When the Compton baseline is nonnegligible, we get the situation depicted in Figure A1. Here, the limits are governed by both the baseline noise and the peak estimation algorithm and its estimated parameters. To illustrate, we take the simplest of estimation algorithms, the 3/6/3 square wave filter, for the estimation of the net peak area. This carries the assumption that the peak is located entirely within the central 6 channels, and that two 3-channel tails can serve as a reliable correction for a linear baseline. The channel count data for the 40 K peak in Figure A1 are: Baseline counts, Sum,
B = 1589.
Peak counts, Sum,
Bi = 266 270 266 [left tail], 279 258 250 [right tail];
Pi = 328 473 581 501 301 244;
P = 2428. Sˆ = P − B = 839;
Net peak area (counts), u(S) = (2428 + 1589)
1/2
= 63.4 (Poisson).
A.4.2.1: Detection, quantification limits (Poisson variance) SC , SD and SQ can be calculated using the equations given in Section A.3, using η = 2 for the paired signal-background case, since the baseline is estimated with the same number of channels as the gross peak. Thus, in units of counts, √ √ σB = 1589 = 39.9, σo = σB 2 = 56.4, SC = zσo = 1.645 · 56.4 = 92.8, SD = z2 + 2SC = 2.71 + 2(92.8) = 188.3, 1/2 SQ = 50 1 + 1 + (σo /5)2 = 616.2 net counts [ca. 10σo ]. Since Sˆ (839) exceeds SC (92.8), we conclude “detected”. A significance test can be applied also to Sˆ in comparison with SQ . Since the lower (0.05) tail of the 90% confidence interval, which equals 839 − 1.645 · 63.4 = 734.7, exceeds SQ , we conclude that the “true” (expected) value of S exceeds the quantification limit. Algorithm dependence. The above calculations apply to the simplest of all peak estimation algorithms, one that works reliably with isolated peaks supported by linear baselines. The same data, however, could be evaluated with far more sophisticated algorithms, based, for
Detection and quantification capabilities in nuclear analytical measurements
127
example, on a non-linear fit to a 3 or 4 parameter peak function (Gaussian–exponential, or Gaussian–Lorentzian). In such cases, the detection and quantification limits would be different, as they depend upon the entire measurement process, including data reduction algorithms. These performance characteristics would be further altered, of course, if the level or nature of the interference were greater or more complex, as with the case of overlapping peaks. Influence of the background. If the background spectrum is linear over the peak + baseline spectral region used for the above estimate, its influence is automatically taken into account by the baseline correction algorithm. If the background, itself, contains a significant 40 K peak, however, the magnitude and uncertainty of that peak must be considered in the estimation of S and its uncertainty, and in the calculation of SC,D,Q . It is not a problem for the present example, however, because of the relative magnitude of the Compton interference. That is, even if the entire 3.6 count background in the six channel peak window were 40 K, it would be equivalent to less than 0.5% of the baseline correction and less than 10% of the uncertainty for that correction. A.4.2.2: Detection, quantification limits (replication variance) Alternatively, the detection decisions and detection and quantification limits may be based on the replication variance s 2 . Comparison of the Poisson-σ to the replication-s is accomplished by taking the sum of the Poisson weighted squared residuals from fitting the baseline model. The sum should be distributed as chi-square if there is no extra (non-Poisson) variance component. The result of such a fit, using the 3 left and 3 right tail Bi ’s gives a non-significant slope (p = 0.45) and si = 10.28 counts per channel with 4 degrees of freedom. (Again, extra, non-significant digits are shown in the interest of numerical accuracy.) The same conclusion follows from calculating the difference between the means of the 3 left and 3 right Bi ’s compared to the Poisson-σ for that difference, which is 5.0 ± 23.0 counts. We can gain a degree of freedom, therefore, by computing s from the mean (per channel) background; this gives si = 9.97 counts per channel with 5 degrees of freedom. The estimated replication standard deviation √ 6 = 24.4 counts. Comparing this for the 6-channel background sum is therefore s = 9.97 √ with the Poisson-σ ( 1589) from Section A.4.2.1, we get (s/σ ) = (24.4/39.9) = 0.61, for which p(χ 2 ) = 0.87, consistent with Poisson variance only. That conclusion for this particular set of experimental observations could be the end of this example, but we continue because of the important tutorial value of illustrating the treatment of replication variance. When we choose to use the replication variance,4 which would be necessary if p(χ 2 ) < 0.05, we get the following results (units of counts): √ so = sB 2 = 34.4, ν = 5 degrees of freedom, α = β = 0.05, sB = 24.4, SC = t1−α,ν so = t0.95,5 · 34.4 = 2.015 · 34.4 = 69.3, SD = δσo = 3.87σo δ ≈ 2t 4ν/(4ν + 1) = 2t (20/21) = 1.905t = 3.84 , σo ≈ so ;
bounds are given by Chi square.
For 5 degrees of freedom, (s/σ )min = 0.48, (s/σ )max = 1.49 [90% interval]. Thus, SˆD = 3.87 · 34.4 = 132.0 net counts; 90% interval: 88.6 to 275.1 net counts. SˆQ = 344.0, SQ = 10σQ ≈ 10σo ,
90% interval: 230.9 to 716.7 counts.
128
L.A. Currie
The approximation assumes σ (S) to be constant. In practice the variance function should be determined. ˆ Conclusions. Since S(839) > SC (69.3), 40 K has been detected. Since the lower (0.05) limit for S (769.7 counts) exceeds the upper limit for SQ (716.7 counts), we conclude that S exceeds the quantification limit. A.4.3: Blank limiting5 This represents the most difficult case, where the very component of interest is present as a blank component of the reference spectrum. That, in fact, was the case for SRM 4215F, where the unintended 40 K peak was present as a blank contaminant. Calculation of SC , SD , and SQ then requires an estimate of σB for 40 K which must be obtained by replication of the blank. The distribution of the blank, and the magnitude of σB depend on the sources of the blank. For the particular SRM 4215F, it is believed that 40 K arose from background potassium in the detector system and a chemical impurity in the master solution for the standard, possibly derived from leaching of potassium from the glass container (Currie, 1997). In more complicated cases of sample preparation, trace levels of potassium can be derived from the sampling and sample handling environment. Since 40 K is a natural radionuclide having a half-life of ca. 1.2 × 109 years, it is therefore present in all potassium, and in favorable cases it can be used for the non-destructive analysis of potassium. In fact, such an application is relevant to research on atmospheric aerosol in remote regions, where it is of great practical importance to determine the minimum quantifiable amount of potassium by a non-destructive technique, such as the direct measurement of 40 K. Although σB can be determined only through blank replication of the actual measurement process, some observations can be offered in relation to the above mentioned sources of the blank. The detection system component, which likely represents a fixed contamination level of this long-lived radionuclide, is expected to exhibit a Poisson distribution, with a gross (peak + baseline) blank variance equal to the expected number of counts. If the blank derives from processing reagents and if its relative standard deviation is rather small (e.g., <10%), it is likely to exhibit normal behavior. On the other hand, if it comes primarily from low-level environmental contamination sources, experience shows that the relative standard deviation is frequently quite large (e.g., >20%), and the distribution, asymmetric and non-normal, since negative blank contributions cannot occur. To offer a brief numerical treatment, let us suppose that 10 replicates of the potassium blank gave an average net peak area of 620 √ counts with an estimated standard deviation sB of 75 counts. Since the ratio s/σ = 75/ 620 = 3.01 exceeds the 0.95 critical value ((s/σ )C = √ 1.88 = 1.37 for 9 df) from the chi-square distribution, we conclude that non-Poisson error components (chemical blank variations) are present; so we shall use the sB for estimation of detection and quantification limits. For this particular example, √ we take measurements of gross signal and blank to be paired.6 Thus, η = 2 and so = sB 2 = 106.1 counts. Values for SC , SD , and SQ (units of counts) are calculated as follows. t-Statistics For ν = 9 degrees of freedom, t (ν, α) = t (9, 0.05) = 1.833 non-central-t,
δ(ν, α, β) = δ(9, 0.05, 0.05) = 3.575,
Detection and quantification capabilities in nuclear analytical measurements
129
δ-approximation: 2t 4ν/(4ν + 1) = 1.946t = 3.567, SC = tso = 1.833(106.1) = 194.5 counts. Note that SC calculated in this way is a “single use” statistic. Repeated detection decisions require independent estimates of σo , as well as an adjustment of t as discussed in Section A.2. SD = δσo
provided that σ is constant between S = 0 and S = SD .
If σ (S) is not constant, the variance function must be determined. For the present illustration, we shall treat σ as constant (homoscedastic). Not knowing σo , we use so as an estimate, and set bounds from bounds for (s/σ ) from the χ 2 distribution. Thus, SˆD = δso = 3.575(106.1) = 379.3 counts, (s/σ )0.05 = 0.607,
(s/σ )0.95 = 1.37 (from χ 2 with ν = 9),
276.9 < SD < 624.9
(90% confidence interval),
SQ = kQ σQ = 10σo
provided that σ is constant between S = 0 and S = SQ .
(See the above note, concerning the variance function.) Thus, SˆQ = 10so = 1061.0 counts, 774.4 < SQ < 1747.9 (90% confidence interval). If σ is an increasing function of signal level, the above estimates will be biased low, but they can be viewed as lower bounds. The ultimate lower bounds, for counting data,√are given by the Poisson-based values for SD and SQ . For the present example, taking σB = 620 (negligible baseline and/or background case) these are √ SD (Poisson) = 2.71 + 3.29 2B = 118.6 counts, 1/2 = 405.7 counts. SQ (Poisson) = 50 1 + 1 + (35.2/5)2 A.4.4: Summary The comparative results for the 40 K detection and quantification capabilities as limited by the three B’s (background, baseline, blank), expressed in terms of the 1460 keV peak area counts are as follows: Background limiting: Baseline limiting: Blank limiting:
SD = 9.55, SD = 188.3, SˆD = 379.3,
SQ = 103.5. SQ = 616.2. SˆQ = 1061.0.
For the blank limiting case, estimated values for SD and SQ are given, because (non-Poisson) blank variability made it necessary to use a statistical estimate (so ) for σo . As shown in Section A.4.3, SD and SQ are consequently uncertain by about a factor of 1.5. Although this example was purposely limited to the signal (peak area) domain, it is interesting to consider the results in terms of radioactivity (Bq) for 40 K, and mass (mg) for naturally
130
L.A. Currie
radioactive potassium. Conversion factors (sensitivities), taking into account the 1460 keV gamma counting efficiency, 40 K gamma branching ratio, and K activity concentration (specific activity) are 42.2 peak area counts/Bq, and 1.26 counts/mg, respectively. Detection limits for 40 K for the three cases are thus 0.23 Bq (background limiting), 4.46 Bq (baseline limiting), and 8.99 Bq (blank limiting). The corresponding detection limit for potassium in the background limiting case is 7.58 mg. Uncertainties in the estimated sensitivities would introduce equivalent relative uncertainties in the estimated detection and quantification limits.5 Because of the applicability of 40 K counting to the direct, non-destructive assay of potassium, it is interesting also to compare the last result with what could be obtained by low-level β counting. For the measurement process discussed in Section A.3 the corresponding sensitivity is 1.62 counts/µg; since SD in that case was 61.3 counts, the mass detection limit for potassium would be (61.3/1.62) or 37.8 µg. The gain of more than a factor of one thousand in sensitivity came about because of major increases in branching ratio, counting efficiency, and counting time for the low-level β counting technique. Because β-particles are far more readily absorbed than gamma rays, this latter technique is most attractive for inherently thin samples (such as air particulate matter) or potassium that has been concentrated chemically, as opposed to the non-destructive analysis of large samples by gamma spectrometry. For selected applications, therefore, low-level β counting can be useful for the assay of natural levels of 40 K, as it is for 14 C and 3 H.
Notes 1. The use of tso for multiple detection decisions is valid only if a new estimate so is obtained for each decision. In the case of a large or unlimited number of null hypothesis tests, it is advantageous to substitute the tolerance interval factor K for Student’s-t. See Currie (1997) for an extended discussion of this issue. 2. The transformation from the signal domain (observed response, in the present case, counts) to the concentration domain is fraught with subtle “error propagation” difficulties, especially for calibration factors having large relative uncertainties. Following the discussion of Section 5 we take the simplest approach in this example, estimating the uncertainty in detection and quantification limits from the uncertainty in the calibration factor A. For small relative uncertainty in A, the result differs little from more sophisticated approaches employing “calibration-based” limits (Currie, 1997). When the relative uncertainty of A is large (e.g., >20%) there is a further complication due to increasing non-normality of the xˆ distribution. For the special case where the background and standard, and therefore x, are measured for each sample, the simplified expressions in Section A.3 can be applied directly to the observed series, with xi = Si /Ai = ((y − B)/A)i , provided that the relative standard deviation of the Ai is not so large that the normality assumption for the xi is seriously violated. 3. Set 2 data illustrate a special xˆ error propagation problem that arises for low-level data, when the net signal is at or near zero. The simplified form of the Taylor expansion does not work, because of (implicit) division by zero. The simplified form gives the relative variance of the estimated concentration xˆ as the sum of the relative variances of the net
Detection and quantification capabilities in nuclear analytical measurements
131
ˆ and the calibration factor (denominator, A): ˆ signal (numerator, S) (σx /x)2 = (σA /A)2 + (σS /S)2 . The S = 0 problem can be avoided by using the proper expression for the uncertainty of a ratio of random variables (Section 5.2, for independent S, A) 2 2 Vx = (1/A2 ) VS (1 + ϕA , ) + S 2 ϕA where ϕA is the relative standard deviation of A. 2 1, the result If S = 0, the last term in the brackets vanishes. If, in addition ϕA becomes σx ≈ σS /A, as with the final result for the Set-2 data. 4. The replication variance s 2 may always be selected as the variance estimator, but it is quite imprecise when ν is small. σ 2 (Poisson) is more precise and sets the lower bound for the true variance, but it underestimates the true variance if extra, non-Poisson components are present. A conservative but potentially biased approach, generally applied in low-level counting and accelerator mass spectrometry, is to use the larger of s 2 or σ 2 (Poisson) (Currie, 1994). 5. In keeping with the peak area (signal domain) treatment for the background and baseline parts of this example, we continue with the reference to the blank expressed as counts, and the symbols SC,D,Q . If the replications of the blank were expressed directly in x-units (radioactivity, Bq for 40 K; or mass, kg for K) the appropriate symbol set would be xC,D,Q . For a comprehensive treatment of all significant uncertainty components in the measurement of 40 K, see the worked example on Radionuclides in Marine Sediment Samples in the contribution of Dovlete and Povinec (2004) to IAEA (2004). 6. Paired measurements are always advisable, to offset possible errors from a drifting or changing blank. For a drifting blank, so may be calculated directly from paired blank observations; alternatively, sB may be calculated from “local” Bi replicates, or from the dispersion about a fitted trend line. Furthermore, if a non-normal blank distribution is likely, paired observations will force a symmetric distribution for Sˆnull (S = 0), but in such a case it is advantageous also to estimate the net signal from k-replicate pairs. That brings about approximate normality as a result of the central limit theorem. In the present example, the estimated relative standard deviation of the blank (75/620) is 0.12, so the assumption of normality may be considered acceptable.
References ASTM (1997). Standard Practice for 99%/95% Interlaboratory Detection Estimate (IDE) for Analytical Methods with Negligible Calibration Error. ASTM Standard D 6091-97, Washington, DC. ASTM (2000). Standard Practice for Interlaboratory Quantitation Estimate. ASTM Standard D 6512-00, Washington, DC. Bailey, K., Chen, C.Y., Du, X., Li, Y.M., Lu, Z.-T., O’Connor, T.P., Young, L. (2000). ATTA—a new method of ultrasensitive isotope trace analysis. Nucl. Instrum. Methods Phys. Res. B 172, 224–227. Boqué, R., Rius, F.X. (1996). Multivariate detection limits estimators [review]. Chemom. Intell. Lab. Syst. 32, 11–23. CELLAR (2004). Collaboration of European low-level underground laboratories: M. Hult, IRMM; G. Heusser, MaxPlanck-Institut für Kernphysik; M. Köhler, VKTA; J.-L. Reyss, LSCE; M. Laubenstein, INFN-LNGS; P. Povinec, IAEA-MEL; M. Schwaiger and T. Schröttner, ARCS; D. Arnold and S. Neumaier, PTB; P. Theodorsson, Science Institute (Iceland). CELLAR leaflet.
132
L.A. Currie
Chen, C.Y., Li, Y.M., Bailey, K., O’Connor, T.P., Young, L., Lu, Z.T. (1999). Ultrasensitive isotope trace analyses with a magneto-optical trap. Science 286, 1139–1141. Cohen, C.A. (1991). Truncated and Censored Samples. Marcel Dekker, New York. Collon, P. (1999). Development of cyclotron AMS for the detection of 81 Kr and first application to groundwater dating. Ph.D. thesis, Univ. of Vienna, Vienna, Austria. Collon, P., Kutschera, W., Lu, Z.-T. (2004). Tracing noble gas radionuclides in the environment. Annu. Rev. Nucl. Part. Sci. 54, 39. Cox, D.R., Lewis, P.A.W. (1966). The Statistical Analysis of a Series of Events. Methuen and Co., London. Csongor, E., Wilhelmova, L., Dvorak, Z., Povinec, P., Grgula, M. (1988). Appl. Radiat. Isotopes 39, 401. Currie, L.A. (1968). Limits for qualitative detection and quantitative determination, application to radiochemistry. Anal. Chem. 40, 586–593. Currie, L.A. (1972). The measurement of environmental levels of rare gas nuclides and the treatment of very low-level counting data. IEEE Trans. Nucl. Sci. NS-19, 119–126. Currie, L.A. (1978). Sources of error and the approach to accuracy in analytical chemistry. In: Kolthoff, I.M., Elving, P.J. (Eds.), Treatise on Analytical Chemistry, vol. I, Ch. 4. Currie, L.A. (1984). Lower Limit of Detection: Definition and Elaboration of a Proposed Position for Radiological Effluent and Environmental Measurements. NUREG/CR-4007, Nuclear Regulatory Commission, Washington, DC. Currie, L.A. (1985a). The Limitations of models and measurements as revealed through chemometric intercomparison. J. Res. NBS 90, 409. Currie, L.A. (1985b). The many dimensions of detection in chemical analysis, in chemometrics. In: Kurtz, D. (Ed.), Pesticide/Environmental Residue Analytical Determination. In: ACS Symp. Ser., vol. 284. American Chemical Society, Washington, DC, Ch. 5, pp. 49–81. See also Currie (1997). Currie, L.A. (1988). Detection: overview of historical, societal and technical issues. In: Currie, L.A. (Ed.), Detection in Analytical Chemistry: Importance, Theory, and Practice. In: ACS Symp. Ser., vol. 361. American Chemical Society, Washington, DC, Ch. 1. Currie, L.A. (1992). In pursuit of accuracy: Nomenclature, assumptions, and standards. Pure Appl. Chem. 64, 455– 472. Currie, L.A. (1994). Optimal estimation of uncertainty in reporting of accelerator mass spectrometry results. Nucl. Instrum. Methods Phys. Res. 92, 188–193. Currie, L.A. (1997). Detection: International update, and some emerging di-lemmas involving calibration, the blank, and multiple detection decisions. Chemom. Intell. Lab. Syst. 37, 151–181. Currie, L.A. (1999). International recommendations offered on analytical detection and quantification concepts and nomenclature. Anal. Chim. Acta 391, 103–134. Currie, L.A. (2000a). Evolution and multidisciplinary frontiers of 14 C aerosol science. Radiocarbon 42, 115–126. Currie, L.A. (2000b). Detection and quantification capabilities and the evaluation of low-level data: some international perspectives and continuing challenges. J. Radioanal. Nucl. Chem. 245, 145–156. Currie, L.A. (2001). Some case studies of skewed (and other ab-normal) data distributions arising in low-level environmental research. Fresenius J. Anal. Chem. 370, 705–718. Currie, L.A. (2004a). Detection and quantification limits: basic concepts, international harmonization, and outstanding (“low-level”) issues. Appl. Radiat. Isotopes 61, 145–149. Currie, L.A. (2004b). The remarkable metrological history of radiocarbon dating [II]. J. Res. Natl. Inst. Stand. Technol. 109, 185–217. Currie, L.A. (2004c). Uncertainty in measurements close to detection limits: detection and quantification capabilities. In: Parr, R., Povinec, P., Fajgelj, A., DeRegge, P. (Eds.), Quantifying Uncertainty in Nuclear Analytical Measurements. IAEA, Vienna, IAEA-TECDOC-1401, pp. 9–34. Currie, L.A. (2008). On the detection of rare and moderately rare nuclear events. J. Radioanal. Nucl. Chem. 276 [2], in press. Currie, L.A., Klouda, G.A. (2001). Detection and quantification capabilities for 85 Kr with the NIST low-level gas counting system; impacts of instrumental and environmental backgrounds. J. Radioanal. Nucl. Chem. 248, 239– 246. Currie, L.A., Parr, R.M. (1988). Perspectives on detection limits for nuclear measurements in selected national and international programs. In: Currie, L.A. (Ed.), Detection in Analytical Chemistry: Importance, Theory, and Practice. In: ACS Symp. Ser., vol. 361. American Chemical Society, Washington, DC, Ch. 9.
Detection and quantification capabilities in nuclear analytical measurements
133
Currie, L.A., Klouda, G.A., Klinedinst, D.B., Sheffield, A.E., Jull, A.J.T., Donahue, D.J., Connolly, M.V. (1994). Fossil- and bio-mass combustion: C-14 for source identification, chemical tracer development, and model validation. Nucl. Instrum. Methods Phys. Res. B 92, 404–409. Currie, L.A., Eijgenhuijsen, E.M., Klouda, G.A. (1998). On the validity of the Poisson Hypothesis for low-level counting; investigation of the distributional characteristics of background radiation with the NIST Individual Pulse Counting System. Radiocarbon 40, 113–127. Currie, L.A., Kessler, J.D., Marolf, J.V., McNichol, A.P., Stuart, D.R., Donoghue, J.C., Donahue, D.J., Burr, G.S., Biddulph, D. (2000). Low-level (submicromole) environmental 14 C metrology. Nucl. Instrum. Methods Phys. Res. B 172, 440–448. Currie, L.A., Kessler, J.D., Fletcher, R.A., Dibb, J.E. (2005). Long range transport of biomass aerosol to Greenland: multi-spectroscopic investigation of particles deposited in the snow. J. Radioanal. Nucl. Chem. 263, 399–411. CX/MAS (1984). CODEX Committee on Methods of Analysis and Sampling (CX/MAS 84/ et seq.) FAO/WHO Food Standards Program, Rome, Italy. See also: Currie, L.A., Horwitz, W. (1994). IUPAC recommendations for defining and measuring detection and quantification. Analysis 22, M24–M26. Davis, C.B. (1994). Environmental regulatory statistics. In: Patil, G.P., Rao, C.R. (Eds.), Handbook of Statistics, vol. 12. Elsevier, New York, Ch. 26. De Geer, L.-E. (2004). Currie detection limits in gamma-ray spectroscopy. Appl. Radiat. Isotopes 61, 151–160. Donahue, D.J., Jull, A.T.J., Linick, T.W., Hatheway, A., Toolin, L.J., Gore, B., Damon, P.E. (1987). Some results from the Arizona TAMS facility: AMS ages of athletic, artistic, and animal artifacts. Nucl. Instrum. Methods Phys. Res. B 29, 169–172. Dovlete, C., Povinec, P.P. (2004). Quantification of uncertainty in gamma-spectrometric analysis of environmental samples. In: Parr, R., Povinec, P., Fajgelj, A., DeRegge, P. (Eds.), Quantifying Uncertainty in Nuclear Analytical Measurements. IAEA, Vienna, IAEA-TECDOC-1401, pp. 103–126. Du, X., Purtschert, R., Bailey, K., Lehmann, B.E., Lorenzo, R., Lu, Z.-T., Mueller, P., O’Connor, T.P., Sturchio, N.C., Young, L. (2003). A new method of measuring 81 Kr and 85 Kr abundances in environmental samples. Geophys. Res. Lett. 30, 2068. Du, X., Bailey, K., Lu, Z.-T., Mueller, P., O’Connor, T.P., Young, L. (2004). An atom trap system for practical 81 Kr dating. Rev. Sci. Instrum. 70, 3224–3232. Ehrlich, G., Mai, H. (1966). Zur objektiven Bewertung des Nachweisvermögens in der HF-Funken-Massenspektrographie. Z. Anal. Chem. 218, 1–17. Eisenhart, C., Zelen, M. (1958). Elements of probability. In: Condon, E.U., Odishaw, H. (Eds.), Handbook of Physics. McGraw–Hill, New York, Ch. 12. Feigl, F. (1923). Tüpfel- und Farbreaktionen als mikrochemische Arbeitsmethoden. Mikrochemie 1, 4. Fukai, R., Ballestra, S., Murray, C.N. (1973). Intercalibration of methods for measuring fission products in seawater samples. Radioact. Sea 35, 3–27. Gibbons, R.D. (1994). Statistical Methods for Groundwater Monitoring. Wiley, New York. Gibbons, R.D. (1995). Some statistical and conceptual issues in the detection of low-level environmental pollutants. Environ. Ecol. Statist. 2, 125–167. Discussants: K. Campbell; W.A. Huber; C.E. White and H.D. Kahn; B.K. Sinha, W.P. Smith and H. Lacayo; W.A. Telliard; and C.B. Davis. Gibbons, R.D., Bhaumik, D.K. (2001). Weighted random-effects regression models with application to interlaboratory calibration. Technometrics 43, 192–198. Hoaglin, D.C., Mosteller, F., Tukey, J.W. (1983). Understanding Robust and Exploratory Data Analysis. Wiley, New York. Horwitz, W. (1990). IUPAC Commission on Analytical Nomenclature, “Nomenclature for Sampling in Analytical Chemistry”. Pure Appl. Chem. 62, 1193. Hubaux, A., Vos, G. (1970). Decision and detection limits for linear calibration curves. Anal. Chem. 42, 849–855. Hurtgen, C., Jerome, S., Woods, M. (2000). Revisiting Currie—how low can you go?. Appl. Radiat. Isotopes 53, 45–50. IAEA (2004). Parr, R., Povinec, P., Fajgelj, A., DeRegge, P. (Eds.), Quantifying Uncertainty in Nuclear Analytical Measurements. IAEA, Vienna, IAEA-TECDOC-1401. ICRM (2004). Proceedings of the Conference on Low-Level Radionuclide Measurement Techniques. Appl. Radiat. Isotopes 61 (2–3). ISO (1989a). Guide 35, Certification of Reference Materials—General and Statistical Principles (and references therein). International Organization for Standardization, Geneva, Switzerland.
134
L.A. Currie
ISO (1989b). Guide 33, Uses of Certified Reference Materials for the Assessment of a Measurement Procedure. International Organization for Standardization, Geneva, Switzerland. ISO (1993). Statistics—Vocabulary and Symbols, Standard ISO 3534. International Organization for Standardization, Geneva, Switzerland. ISO (1995). Guide to the Expression of Uncertainty in Measurement [GUM], issued by BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, and OIML. International Organization for Standardization, Geneva, Switzerland. ISO (1997). Capability of Detection, Standard ISO 11843-1 through-4. International Organization for Standardization, Geneva, Switzerland. ISO (2000). Determination of the Detection Limit and Decision Threshold for Ionizing Radiation Measurements, Standard ISO 11929-1 through-4. International Organization for Standardization, Geneva, Switzerland. ISO-IUPAC (1993). Deliberations on nomenclature related to limits of detection. In: Minutes of Nomenclature Harmonization Meeting; P.-Th. Wilrich (ISO), M. Parkany, chairman (IUPAC), L.A. Currie, rapporteur (IUPAC). Washington, DC. IUPAC (1987). Freiser, H., Nancollas, G.H. (Eds.), Compendium of Analytical Nomenclature, second ed. Blackwell Scientific Publ., Oxford, UK. IUPAC (1995). Nomenclature in evaluation of analytical methods, including detection and quantification capabilities (prepared by L.A. Currie). Pure Appl. Chem. 67, 1699–1723. IUPAC (1998). Inczédy, J., Ure, A.M., Lengyel, T., Gelencsér, A. (Eds.), Compendium of Analytical Nomenclature, third ed. IUPAC/Blackwell Science Ltd., Oxford, UK. Jackson, J.E. (1959). Technometrics 1, 359. Kaiser, H. (1947). Die Berechnung der Nachweisempfindlichkeit. Spectrochim. Acta 3, 40–67. Kaiser, H. (1966). Zur Definition der Nachweisgrenze, der Garantiegrenze und der dabei benutzten Begriffe. Z. Anal. Chem. 216, 80–94. Kaiser, H. (1972). Z. Anal. Chem. 260, 252. Kelly, W.R., Paulsen, P.J., Murphy, K.E., Vocke Jr., R.D., Chen, L.-T. (1994). Anal. Chem. 66, 2505–2513. Kurtz, D. (Ed.) (1985). Chemometrics in Pesticide/Environmental Residue Analytical Determination. In: ACS Symp. Ser., vol. 284. American Chemical Society, Washington, DC. See Currie, L.A., Ch. 5, Kurtz, D., Ch. 9. Kurtz, D.A., Taylor, J.K., Sturdivan, L., Crummett, W.B., Midkiff Jr., C.R., Watters Jr., R.L., Wood, L.J., Hanneman, W.W., Horwitz, W. (1988). Real-world limitations to detection. In: Currie, L.A. (Ed.), Detection in Analytical Chemistry: Importance, Theory, and Practice. In: ACS Symp. Ser., vol. 361. American Chemical Society, Washington, DC, Ch. 16. Lambert, D., Peterson, B., Terpenning, I. (1991). Nondetects, detection limits, and the probability of detection. J. Amer. Statist. Assoc. 86, 266–277. Laubenstein, M., Hult, M., Gasparro, J., Arnold, D., Neumaier, S., Heusser, G., Köhler, M., Povinec, P., Reyss, J.-L., Schwaiger, M., Theodórsson, P. (2004). Underground measurements of radioactivity. Appl. Radiat. Isotopes 61, 167–172. Liteanu, C., Riˇca, I. (1980). Statistical Theory and Methodology of Trace Analysis. Wiley, New York. Long, G.L., Winefordner, J.D. (1983). Limit of detection: a closer look at the IUPAC definition. Anal. Chem. 55 (7), 713A–724A. Loosli, H.H., Oeschger, H.O. (1969). Earth Planet. Sci. Lett. 7, 67. Mandel, J., Paule, R.C. (1970). Interlaboratory evaluation of a material with unequal numbers of replicates. Anal. Chem. 42, 1194. MARLAP (2004). Multi-Agency Radiological Laboratory Analytical Protocols Manual, vol. 20, Detection and Quantification Capabilities. MARLAP, http://www.epa.gov/radiation/marlap/manual.htm. Massart, D.L., Vandeginste, B.G.M., Deming, S.N., Michotte, Y., Kaufman, L. (1988). Chemometrics: A Textbook. Elsevier, Amsterdam. McCormack, M. (1988). Realistic detection limits and the political world. In: Currie, L.A. (Ed.), Detection in Analytical Chemistry: Importance, Theory, and Practice. In: ACS Symp. Ser., vol. 361. American Chemical Society, Washington, DC, Ch. 2. Méray, L. (1994). Deconvolution of nuclear spectra of low counts. Nucl. Instrum. Methods Phys. Res. A 353, 272– 275. Michel, R. (2000). Quality assurance of nuclear analytical techniques based on Bayesian characteristic limits. J. Radioanal. Nucl. Chem. 245, 137–144. Miller, R.G. (1981). Simultaneous Statistical Inference. Springer, New York.
Detection and quantification capabilities in nuclear analytical measurements
135
Nalimov, V.V., Nedler, V.V., Men’shova, N.P. (1961). Metrological evaluations in the detection of small concentrations by the method of emission spectrum analysis. Zavod. Lab. 27, 861–865. Natrella, M. (1963). Experimental Statistics. NBS Handbook 91. U.S. Government Printing Office, Washington, DC. Parr, R., Houtermans, H., Schaerf, K. (1979). In: Computers in Activation Analysis and Gamma-Ray Spectroscopy. U.S. Department of Energy, CONF-780421, p. 544. Patnaik, P.B. (1949). Biometrika 36, 202. Pearson, E.S., Hartley, H.O. (1951). Biometrika 38, 112. Povinec, P.P., Comanducci, J.F., Levy-Palomo, I. (2004). IAEA-MEL’s underground counting laboratory in Monaco—background characteristics of HPGe detectors with anti-cosmic shielding. Appl. Radiat. Isotopes 61, 85–93. Przyborowski, J., Wilenski, H. (1939). Homogeneity of results in testing samples from Poisson series. Biometrika 31, 313–323. Reeve, C.P. (1986). Paper N-86-4. Statistical Engineering Division, National Bureau of Standards. Roos, J.B. (1962). The limit of detection of analytical methods. Analyst 87, 832. Simgen, H., Heusser, G., Zuzel, G. (2004). Highly sensitive measurements of radioactive noble gas nuclides in the BOREXINO solar neutrino experiment. Appl. Radiat. Isotopes 61, 213–217. Steinkopff, T., Dyck, W., Frank, G., Frenzel, S., Salvamoser, J. (2004). The measurement of radioactive noble gases by DWD in the frame of the Global Atmospheric Watch Programme of WMO. In: Proc. Conf. on Low-Level Radionuclide Measurement Techniques. Appl. Radiat. Isotopes 61, 225–230. Sturchio, N.C., Du, X., Purtschert, R., Lehmann, B.E., Sultan, M., Patterson, L.J., Lu, Z.-T., Mueller, P., Bailey, K., O’Conner, T.P., Young, L., Lorenzo, R., Kennedy, B.M., van Soest, M., El Alfy, Z., El Kaliouby, B., Dawood, Y., Abdallah, A.M.A. (2004). One million year old groundwater in the Sahara revealed by krypton-81 and chlorine36. Geophys. Res. Lett. 31, L05503. Theodórsson, P. (1992). Quantifying background components of low-level gas proportional counters. Radiocarbon 34, 420–427. Tiku, M.L. (1972). Austral. J. Statist. 14, 37. Tukey, J.W. (1977). Exploratory Data Analysis. Addison-Wesley, Reading, MA. Tukey, J.W. (1984). Non-linear methods for smoothing data. In: Brillinger, D.R. (Ed.), The Collected Works of John W. Tukey, vol. II. Wadsworth, Monterey, CA, Ch. 22. U.S. Nuclear Regulatory Commission (1982). Standard Radiological Effluent Technical Specifications. NUREG0472. Velleman, P.F., Hoaglin, D.C. (1981). Applications, Basics, and Computing of Exploratory Data Analysis. PWS– Kent, Boston, MA. Waterscience (2005). Procedures for detection and quantitation, http://www.epa.gov/waterscience/methods/det/. Wilhelmova, L., Tomasek, M., Stukheil, K. (1994). Biol. Trace Elem. Res. 43, 725–736. Youden, W.J. (1969). Uncertainties in calibration. In: Ku, H.H. (Ed.), Precision Measurement and Calibration. NBS Spec. Publ. 300. U.S. Government Printing Office, Washington, DC, Ch. I-5.
This page intentionally left blank
137
Radiometric determination of anthropogenic radionuclides in seawater M. Aoyama∗ , K. Hirose Geochemical Research Department, Meteorological Research Institute (MRI), Tsukuba, 305-0052 Japan
Anthropogenic radionuclides in seawater have been concerned with their ecological effects and have been oceanographically used as a tracer. Current concentrations of anthropogenic radionuclides in the oceanic waters are generally extremely low. Determination of anthropogenic radionuclides in seawater has been traditionally performed with radiometric methods such as γ -spectrometry, β-counting and α-spectrometry. The radiometric method is still a useful tool to determine concentrations of anthropogenic radionuclides, although recently mass spectrometric methods have been developed extensively. Here we describe the radiometric methods to determine typical anthropogenic radionuclides such as 137 Cs, 90 Sr and plutonium in seawater, which includes recent development of the radiometric methods such as extremely low background γ -spectrometry at the Ogoya Underground Laboratory (OUL).
1. Introduction Huge amounts of anthropogenic radionuclides have been introduced into marine environments as global fallout from large-scale atmospheric nuclear-weapon testing, discharge from nuclear facilities and ocean dumping of nuclear wastes (UNSCEAR, 2000). The radiological and ecological effects of anthropogenic radionuclides are still of world concern. To assess the marine environmental effects of anthropogenic radionuclides, it is significant to clarify their behavior and fate in the marine environments. Therefore, concentrations of anthropogenic radionuclides in seawater are an important tool to evaluate the ecological effect of anthropogenic radionuclides. 137 Cs is one of the most important anthropogenic radionuclides in the field of environmental radioactivity because of its long physical half-life of 30.0 years. It is a major fission product (fission yield: 6–7%) from both plutonium and uranium (UNSCEAR, 2000). 137 Cs in the ocean has been mainly derived from global fallout (Reiter, 1978; Bowen et al., 1980; UNSCEAR, 2000; Livingston and Povinec, 2002), together with close-in fallout from the ∗ Corresponding author. E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11004-4
© 2008 Elsevier B.V. All rights reserved.
138
M. Aoyama and K. Hirose
Pacific Proving Ground nuclear explosions (Bowen et al., 1980; Livingston and Povinec, 2002), discharge of radioactive wastes from nuclear facilities and others (Sugiura et al., 1975; Pentreath, 1988; Hirose et al., 1999). 137 Cs in seawater has been determined since 1957 to elucidate the radioecological effects of anthropogenic radioactivity in the marine environment (Miyake and Sugiura, 1955; Rocco and Broecker, 1963; Shirasawa and Schuert, 1968; Saruhashi et al., 1975; Bowen et al., 1980; Folsom, 1980; Nagaya and Nakamura, 1987a, 1987b; Miyake et al., 1988; Hirose et al., 1992; Aoyama and Hirose, 1995; Hirose et al., 1999; Aoyama et al., 2000, 2001; Aoyama and Hirose, 2003; Hirose and Aoyama, 2003a, 2003b; Ito et al., 2003; Povinec et al., 2003, 2004; Hirose et al., 2005). Additionally, 137 Cs in seawater is a powerful chemical tracer of water mass motion on the time scale of several decades (Bowen et al., 1980; Folsom, 1980; Miyake et al., 1988; Miyao et al., 2000; Tsumune et al., 2001, 2003a, 2003b) because most of the 137 Cs in water columns is present in dissolved form. Another advantage of the use of 137 Cs as an oceanographic tracer is the quantity and accessibility of marine radioactivity during the past four decades in contrast with other chemical tracers such as CFCs (Warner et al., 1996). Another important fission product is 90 Sr, which is a β-emitter with a half-life of 28.5 years. It has been believed that the oceanic behavior of 90 Sr is very similar to that of 137 Cs because both 137 Cs and 90 Sr, a typical non-particle-reactive element, exist as ionic forms in seawater. In contrast to 137 Cs, measurements of 90 Sr in seawater are generally inconvenient because of complicated radioanalytical processes. However, the behavior of 90 Sr in the ocean differs from that of 137 Cs; for example, a significant amount of 90 Sr has been introduced to the ocean via river discharge in contrast to 137 Cs (Livingston, 1988), which is tightly retained in soil mineral surfaces. In fact, the 90 Sr/137 Cs ratios in oceanic waters have varied spatially and temporally. 90 Sr may be considered to be a tracer of the effect of river discharge to the ocean. These findings suggest that there is a need independently to determine 90 Sr in seawater. Plutonium in the ocean has been introduced by global fallout from large-scale atmospheric nuclear weapons testing programs, from which the major oceanic input occurred in the early 1960s (Harley, 1980; Perkins and Thomas, 1980; Hirose et al., 2001a). However, 239,240 Pu inventories in water columns in the North Pacific are significantly greater than estimates based on global fallout deposition (Bowen et al., 1980). This additional source of plutonium in the North Pacific has been attributed to inputs from close-in fallout from the U.S. nuclear explosions conducted at the Pacific Proving Grounds in the Marshall Islands in the 1950s (Bowen et al., 1980; Livingston et al., 2001). In order to have better understanding of the oceanic behavior of 239,240 Pu, it is important to elucidate the effect of close-in fallout of 239,240 Pu in the western North Pacific. Plutonium in the ocean is transported by physical and biogeochemical processes. The residence time of plutonium in surface waters of the open ocean ranges from 6 to 21 years and is generally shorter than that of the corresponding surface 137 Cs (Hirose et al., 1992, 2001b, 2003). In contrast to 137 Cs, plutonium is a typical particle-reactive radionuclide; plutonium in particulate matter of surface waters comprises from 1 to 10% of the total (Hirose et al., 2001b; Livingston et al., 1987), whereas particulate 137 Cs is less than 0.1% of the total (Aoyama and Hirose, 1995). Plutonium vertically moves with sinking biogenic particles (Fowler et al., 1983; Livingston and Anderson, 1983) and regenerates into soluble forms in deep waters as a result of microbial decomposition of particles. These biogeochemical processes typically produce vertical concentration profiles of plutonium which show a surface minimum, a mid-depth
Radiometric determination of anthropogenic radionuclides in seawater
139
maximum and, thereafter, a decrease with increasing water depth (Hirose, 1997; Tsumune et al., 2003a, 2003b). Although plutonium moves vertically by biogeochemical processes, most of the plutonium introduced into the Pacific Ocean still exists in the water column; plutonium inventories in sediments of the open ocean are generally less than 10% of the total inventories except in sea areas near Bikini and Enewetak (Livingston et al., 2001). However, plutonium behavior in oceanic water is further affected by physical processes such as advection and upwelling (Hirose et al., 2002). For example, the observation that the plutonium maximum layer in the mid-latitude region of the North Pacific has deepened with time (Livingston et al., 2001) is explained by a simple one-dimensional biogeochemical model (Hirose, 1997), but does not provide a mechanism for reducing the overall water-column inventory. Consequently, advection must play a significant role in determining the temporal trends in regional oceanic inventories of plutonium. The North Pacific deep waters (more than 2000 m depth) contain a significant amount of plutonium (Bowen et al., 1980), however, the input processes are still unknown. In this paper, we describe traditional radiometric methods of 137 Cs, 90 Sr and plutonium assay in seawater and exhibit typical recent results on the oceanic behaviors of 137 Cs, 90 Sr and plutonium. 2. Analytical method for 137 Cs analysis in seawater 2.1. Background 137 Cs
decays to 137 Ba by emitting β-rays (188 keV) and γ -rays (661.7 keV). Cs, that exists in ionic form in natural water, is one of the alkali metals and chemically shows less affinity with other chemicals. The concentration of stable Cs in the ocean is only 3 nM. Known adsorbents to collect Cs in seawater are limited, e.g., ammonium phosphomolybdate (AMP) and hexacyanoferrate compounds (Folsom and Sreekumaran, 1966; La Rosa et al., 2001). The AMP has been an effective ion exchanger of alkali metals (Van R. Smit et al., 1959). It has been known that AMP forms insoluble compound with Cs. AMP, therefore, has been used to separate other ions and concentrate Cs in environmental samples. In the late 1950s, determination of 137 Cs in seawater was carried out with β-counting because of underdevelopment of γ -spectrometry. Two to several ten mg of Cs carrier is usually added when the radiocesium is determined by β-counting because of the formation of a precipitate of Cs2 PtCl6 and calculation of chemical yields of cesium throughout the procedure (Yamagata and Yamagata, 1958; Rocco and Broecker, 1963). After the development of γ -spectrometry using Ge detectors, the AMP procedure with γ -spectrometry became a convenient concentration procedure for the determination in environmental samples. In Japan, the AMP method is recommended for radiocesium measurements in seawaters, in which Cs carrier is stated to be unnecessary (Science and Technology Agency, 1982). It, however, must be noted that large volumes of seawater samples (more than 100 liters) were required to determine 137 Cs because of the relatively low efficiency of Ge-detectors (around 10%). In the previous literature, the weight yield of AMP has not been used because the chemical yield of Cs could be obtained and the loss of a small amount of AMP during the treatment did not cause serious problems. Actually, the use of AMP reagent produced in the 1960s and
140
M. Aoyama and K. Hirose
Table 1 Performance of well-type Ge detectors operated in a ground-level lab (MRI) and in a underground lab (Ogoya) Institute
Type
Active volume (cm3 )
Absolute efficiencya (%)
Backgroundb (cpm/1 keV)
MRI
ORTEC 6 7 8 9 Canberra EYRISYS
280 80 280 600 199 315
20.5 10.8 16.5 23.7 14.5 20
0.092 0.033 0.109 0.074 0.0003 0.0016
Ogoya
a The absolute efficiencies of HPGE are calculated at the 660 keV photo-peak of 137 Cs. b The background values were calculated as a sum from 660 to 664 keV corresponding to the 662 keV photo-peak of 137 Cs.
in the mid 1980s gave a range from 70% to 90% as the weight yields of AMP without Cs carrier in laboratory experiments in 1996. These weight yields are in good agreement with the records of weight yields of AMP in our laboratory during the 1970s and 1980s. However, the weight yield of AMP without Cs carrier had been decreasing from the end of the 1980s and it sometimes became very low, less than 10%, in the mid 1990s. To improve 137 Cs determination in seawater, Aoyama et al. (2000) re-examined the AMP procedure. Their experiments revealed that stable Cs carrier of the same equivalent amount as AMP is required to form the insoluble Cs-AMP compound in an acidic solution (pH = 1.2 to 2.2). The improved method has achieved high chemical yields of more than 95% for sample volumes of less than 100 liters. Another improvement is to reduce the amount of AMP from several tens of grams to 4 g to adsorb 137 Cs from the seawater samples. As a result, the sample volume has been reduced from around 100 liters to less than 20 liters to be able to use high-efficiency well-type Ge-detectors (Table 1). This improvement for 137 Cs is favorable to its use as a tracer in the oceanographic field. However, there has been a serious problem regarding 137 Cs measurement; i.e., large-volume sampling of more than 100 liters has been required to determine 137 Cs concentrations in deep waters because of the very low concentrations of 137 Cs (less than 0.1 Bq m−3 ). A major problem not improving sensitivity is that high-efficiency well-type Ge-detectors result in higher backgrounds during γ -spectrometry in ground-level laboratories. Especially, it is difficult to determine accurate 137 Cs concentrations in deep waters (>1000 m) because of the difficulty of acquiring large, non-contaminated samples. However, recently, Komura (Komura, 2004; Komura and Hamajima, 2004) has established an underground facility (Ogoya Underground Laboratory: OUL) to achieve extremely low background γ -spectrometry using Ge detectors with high efficiency and low background materials. The OUL has been constructed in the tunnel of the former Ogoya copper mine (235 m height below sea level, Ishikawa prefecture) in 1995 by the Low Level Radioactivity Laboratory, Kanazawa University. The depth of the OUL is 270 m water equivalent and the contributions of muons and neutrons are more than two orders of magnitude lower than those at ground level. In order to achieve extremely low background γ -spectrometry, high efficiency well type Ge detectors specially designed for low level counting were shielded with extremely
Radiometric determination of anthropogenic radionuclides in seawater
141
low background lead prepared from the very old roof tiles of Kanazawa Castle. As a result, the background of γ -rays corresponding to the energy range of 137 Cs is two orders of magnitude lower than that in ground-level facilities as shown in Table 1. The detection limit of 137 Cs at the OUL is 0.18 mBq for a counting time of 10,000 min (Hirose et al., 2005). There is a residual problem of underground γ -spectrometry for 137 Cs measurements. AMP adsorbs trace amounts of potassium when Cs is extracted from seawater because K is a major component in seawater and radioactive potassium (40 K) contains 0.0118% of total K in natural materials. Trace amounts of 40 K cause elevation of the background corresponding to the energy range of 137 Cs due to Compton scattering of 40 K. If 40 K can be removed in AMP/Cs compound samples, the full performance of underground γ -spectrometry for 137 Cs measurements is established. To remove 40 K from the AMP/Cs compound, a precipitation method including the insoluble platinate salt of Cs was applied for purification of Cs. This method performed to remove trace amounts of 40 K from the AMP/Cs compound has a chemical yield of around 90% for 137 Cs (Hirose et al., 2006b). 2.2. Method 2.2.1. Sampling and materials Seawater samples are collected using a CTD-rosette sampler, which collects seawater of 12 liters volume at each of 24–36 different depth layers. All reagents used for 137 Cs, 90 Sr and Pu assay are special (G.R.) grade for analytical use. All experiments and sample treatments are carried out at ambient temperatures. It is very important to know the background γ activities of reagents. The 137 Cs activity in CsCl was less than 0.03 mBq g−1 using extremely low background γ -spectrometry. The 137 Cs activity in AMP was less than 0.008 mBq g−1 . There is no serious contamination of 137 Cs from other reagents. 2.2.2. γ -Spectrometry 137 Cs measurements were carried out by γ -spectrometry using well-type Ge detectors coupled with multi-channel pulse height analyzers. The performance of the well-type Ge detectors is summarized in Table 1. Detector energy calibration was done using IPL mixed γ -ray sources, while the geometry calibration was done using the “reference” material with the same tube and very similar density. 2.2.3. Recommended procedure We propose an improved AMP procedure with the ground-level γ -spectrometer as follows: (1) Measure the seawater volume (5–100 L) and put into a tank of appropriate size. (2) pH should be adjusted to 1.6–2.0 by adding concentrated HNO3 (addition of 40 mL conc. HNO3 for 20 L seawater sample makes pH of sample seawater about 1.6). (3) Add 0.26 g of CsCl to form an insoluble compound and stir at a rate of 25 L per minute for several minutes. (4) Weigh 4 g of AMP and pour it into a tank to disperse the AMP with seawater. (5) 1 h stirring at a rate of 25 L air per minute. (6) Settle until the supernate becomes clear. The settling time is usually 6 h to overnight, but no longer than 24 h.
142
M. Aoyama and K. Hirose
(7) Take an aliquot of 50 mL supernate to calculate the amount of the residual cesium in the supernate. (8) Loosen the AMP/Cs compound from the bottom of the tank and transfer into a 1–2 L beaker; if necessary, do additional step of decantation. (9) Collect the AMP/Cs compound onto 5B filter by filtration and wash the compound with 1 M HNO3 . (10) Dry up the AMP/Cs compound for several days at room temperature. (11) Weigh the AMP/Cs compound and determine weight yield. (12) Transfer the AMP/Cs compound into a Teflon tube of 4 mL volume and subject to γ -ray spectrometry. 2.2.4. Underground γ -spectrometry (1) The same procedure from step (1) to step (12). (2) Dissolve AMP/Cs compound by adding alkali solution. (3) pH should be adjusted to ca. 8.1 by adding 2 M HCl and adjust the volume of solution to ca. 70–100 mL. (4) Perform precipitation of Cs2 PtCl6 by adding chloroplatinic acid (1 g/5 mL D.W.) at pH = 8.1 and keep in refrigerator for a half-day. (5) Collect the Cs2 PtCl6 precipitate onto a filter and wash the compound with solution (pH = 8.1). (10) Dry the Cs2 PtCl6 precipitate for several days at room temperature. (11) Weigh the Cs2 PtCl6 precipitate and determine the weight yield. (12) Transfer the Cs2 PtCl6 precipitate into a Teflon tube of 4 mL volume and subject to underground γ -spectrometry. 2.3. Application to ocean samples 2.3.1. Database Marine radioactivity databases have been constructed in several institutes (e.g., IAEA, Marine Environmental Laboratory, Monaco (GLOMARD database, IAEA, 2000; Povinec et al., 2004, 2005), Meteorological Research Institute, Japan (HAM database, Aoyama and Hirose, 2004)). Aoyama and Hirose (2004) have been preparing a comprehensive database (HAM) including 137 Cs, 90 Sr and 239,240 Pu concentrations in seawater of the Pacific and marginal seas. The present total numbers of 137 Cs, 90 Sr and 239,240 Pu concentrations in the HAM database are 7737, 3972 and 2666 records, respectively. The numbers of surface 137 Cs, 90 Sr and 239,240 Pu concentrations in the HAM database are 4312, 1897 and 629 records, respectively. For data analysis, we carried out no special quality controls except removal of 239,240 Pu data in highly contaminated areas (the Pacific Proving). 2.3.2. Surface 137 Cs in the ocean The level of the 137 Cs concentrations in surface waters of the world ocean has been summarized in the IAEA-MEL research project (Livingston and Povinec, 2002; Povinec et al., 2005). The higher 137 Cs concentrations in the surface waters in 2000, ranging from 10 to 100 Bq m−3 , occurred in the Baltic Sea, Irish Sea, Black Sea and Arctic ocean, which reflects the effect of the Chernobyl fallout as does the discharge of radioactive wastes from the
Radiometric determination of anthropogenic radionuclides in seawater
143
Fig. 1. Schematic map of Pacific boxes.
Sellafield nuclear reprocessing plants. The level of surface 137 Cs in 2000 ranged from 1 to 10 Bq m−3 for most of the ocean areas in the world such as the Pacific and Indian Oceans, except the South Atlantic and Antarctic Oceans, whose 137 Cs concentrations was less than 1 Bq m−3 . The temporal change of chemical tracers can provide significant information on the physical and biogeochemical processes in the ocean. However, we have no complete data set of anthropogenic radionuclides temporally and spatially in the Pacific because only a few current data are present and the geographic distribution of sampling stations is heterogeneous. In order to elucidate the present levels of anthropogenic radionuclides in the Pacific, it is important to examine the temporal change of concentrations of anthropogenic radionuclides in surface waters. The temporal changes of 137 Cs and 239,240 Pu concentrations in surface waters differ for the various sea areas of the Pacific (Aoyama and Hirose, 1995; Hirose et al., 2001a). Therefore, the Pacific Ocean basin is divided into twelve boxes as shown in Figure 1, taking into account the latitudinal and longitudinal distributions of anthropogenic radionuclides (Hirose and Aoyama, 2003a, 2003b), the oceanographic knowledge and the latitudinal distribution of global fallout (UNSCEAR, 2000; Hirose et al., 1987, 2001b). Box 1 (north of 40◦ N) is the Subarctic Pacific Ocean, in which greater depositions of 137 Cs, 90 Sr and 239,240 Pu occurred in the 1960s (UNSCEAR, 2000; Hirose et al., 1987, 2001b). Boxes 2 and 3 (25◦ N– 40◦ N) are upstream and downstream of the Kuroshio extension, respectively, which correspond to the mixing area between the water masses of the Kuroshio and Oyashio currents. The greatest deposition of 137 Cs, 90 Sr and 239,240 Pu throughout the Pacific occurred in Box 2 in the 1960s. Boxes 4 and 5 (5◦ N–25◦ N) are downstream and upstream of the North Equatorial current, which is a typical oligotrophic ocean corresponding to the subtropical gyre. Box 5 includes the California current. These boxes include the highly radioactivitycontaminated islands (Bikini, Enewetak, Johnston islands and others) of the Pacific Proving
144
M. Aoyama and K. Hirose
Fig. 2. Calculated concentrations of 137 Cs in surface waters of the Pacific boxes. Small numbers show the observed values, which were obtained in the year indicated in parentheses.
ground nuclear weapons test sites, especially typical for plutonium (Livingston et al., 2001; Buesseler, 1997). The subtropical region of the western North Pacific was affected by close-in fallout compared with that in the eastern North Pacific. Boxes 6 and 7 (5◦ S–5◦ N) are downstream and upstream of the South Equatorial current, respectively. The boundary between Boxes 4, 5 and 6, 7 is approximately the Equatorial Counter current. Box 7 includes the equatorial upwelling region. Boxes 8 and 9 (5◦ S–25◦ S) are downstream and upstream of the weak South Equatorial current, respectively. Box 9 includes the French nuclear weapons test site (Mururoa). Box 10 (25◦ S–40◦ S) corresponds to the Tasman Sea. Box 11 (25◦ S–40◦ S) is the mid-latitude region of the South Pacific. Box 12 (40◦ S–60◦ S), including the Indian Ocean, is the Southern Ocean corresponding to the Antarctic circumpolar current. Temporal changes of 137 Cs concentrations in the surface waters of each box were examined using the HAM database. Surface 137 Cs decreased exponentially throughout the period from 1971 to 1998. Similar exponential decreases of surface 137 Cs occurred in other boxes of the North Pacific (Hirose and Aoyama, 2003a, 2003b). Although the observed data are scant in the South Pacific, it is expected that surface 137 Cs decreased exponentially after 1971 because the major input due to global fallout occurred in the mid 1960s and the major French nuclear weapons tests were conducted until 1971 (UNSCEAR, 2000). The concentrations of surface 137 Cs in 2000 can be extrapolated from the regression curves of the time-series data in each box. However, the confidence levels of the regression curves vary greatly between the boxes because the numbers of data differ significantly among the boxes. In the North Pacific, we obtain the estimated concentrations of anthropogenic radionuclides with higher quality. The calculated 137 Cs concentrations in surface waters of each box in 2000, together with the most recent observed data, are shown in Figure 2. The calculated 137 Cs concentrations in surface waters of each box in 2000 ranged from 1.2 to 2.8 Bq m−3 . The 137 Cs concentrations in the North Pacific surface waters are still higher than those in the South Pacific, although about four decades have passed since the major injection of 137 Cs in the Northern Hemisphere.
Radiometric determination of anthropogenic radionuclides in seawater
145
Fig. 3. Section of 137 Cs concentration (Bq m−3 ). Dots show positions of samples. Measurements of about sixty samples in the southern part in this section have not been completed yet.
Higher surface 137 Cs concentration occurs in the eastern Equatorial Pacific (Box 7). The lower values appear in the Tasman Sea (Box 10). It, however, is noteworthy that the surface 137 Cs concentration is homogeneous in the Pacific compared with those in the 1960s and 1970s. 2.3.3. Cross-section of 137 Cs along 165◦ E in the Pacific The cross-section of 137 Cs concentrations obtained from the Ryofu-Maru 2002 cruises along 165 deg. E is presented in Figure 3. A remarkable core with high 137 Cs concentrations in the 137 Cs cross-section which exceeded 2 Bq m−3 was found (Aoyama et al., 2004). This 137 Cs core occurred around 20◦ N, corresponding to the southern part of the subtropical gyre in the Northern Pacific. The depth of the 137 Cs core was between 200 and 500 m. The 137 Cs concentrations in the surface waters were 1.5 Bq m−3 or less. We also found a subsurface maximum at 100–150 m south of 15◦ N. The 137 Cs concentration in deep waters below 1500 m ranged around 0.01–0.1 Bq m−3 in the northern part of this section. In terms of density, the core of 137 Cs (200–500 m depth) corresponds to a sigma of 25–26.2 or 3. We can consider the gyre-scale subduction as a potential cause of the observed 137 Cs core. The subduction is a principal upper ocean ventilation process. The water masses related to subduction are characterized through the exchange of heat, moisture, and dissolved gases and transient tracers such as 137 Cs in our study. After the water masses were formed and transferred beneath the mixed layer, they are shielded by the upper water masses and subsequently their properties modified by mixing in the ocean interior. At present, two opportunities for the subduction in the North Pacific can be considered. One is a lighter variety of central mode water with a density of sigma around 26.0, whose outcrop occurs in the central North Pacific (35◦ –40◦ ) in winter. Another is a denser variety of central mode water with density of 26.4, whose outcrop occurs in the central northern North Pacific (>40◦ N). Overlaying the 137 Cs contour onto a potential vorticity section reveals that the 137 Cs core corresponds to a southern part of lower potential vorticity at density level of 26.0. The lower potential vorticity reflects the property of the origin of the water mass corresponding to a lighter variety of the central mode water. For the density corresponding to the denser variety of the central mode water, there was no marked signal in the 137 Cs concentrations. Comparing with previous data, the
146
M. Aoyama and K. Hirose
137 Cs concentrations in this core seem to be still increasing. Another typical feature is that the
subsurface 137 Cs maximum occurred in the 100–150 m depth layer around 20◦ N in the western North Pacific, which phenomenon was observed in previous papers (Bowen et al., 1980; Hirose and Aoyama, 2003a). The subsurface 137 Cs maximum corresponds to the North Pacific subtropical mode water (NPSTMW) with the 25.4 isopycnal, whose winter outcrop occurs around 35◦ N. The NPSTMW is transported south westward accompanied by subductions. Southward transport of the NPSTMW might control a longer half-residence time of the 137 Cs surface water concentrations and an increasing tendency of the 137 Cs water column inventory in the equatorial region (Aoyama and Hirose, 2003; Hirose and Aoyama, 2003a). The 137 Cs concentration in surface waters of the NW Pacific has decreased steadily to present levels between 2 and 4 Bq m−3 since the early 1960s. In a previous study (Saruhashi et al., 1975) a typical latitudinal distribution of 137 Cs was observed, which was high in midlatitudes (10–20 Bq m−3 ) and low in the Equatorial region (around 5 Bq m−3 ). The differences of the 137 Cs concentrations in the surface waters between the mid-latitudes and the Equatorial region have decreased over the past two decades. In the Equatorial region, the 137 Cs concentration in the surface waters showed no temporal change apart from radioactive decay (Aoyama and Hirose, 1995; Aoyama et al., 2001). Therefore, the current geographical distribution of surface 137 Cs in the NW Pacific reflects that of global fallout input re-distributed by surface/subsurface circulation in the North Pacific which includes subduction and equatorial upwelling processes. 3. Analytical method for 90 Sr 3.1. Background 90 Sr
decays to stable 90 Zr via 90 Y (half-life: 2.67 day; maximum energy of β-particles: 2.28 MeV). Strontium comprises about 0.025% of the Earth’s crust and the concentration of stable Sr in the ocean is 8.7 × 10−5 M. It is widely distributed with calcium. The chemistry of strontium is quite similar to that of calcium, of which the concentration in the ocean is 10−2 M. The biological behavior of strontium in the ocean is also very close to that of calcium, and thus the behavior of Sr in the ocean can be considered to be different from that of Cs. The only radiometric method for 90 Sr is β-counting. Therefore, radiochemical separation is required for determination of 90 Sr. An essential step in 90 Sr analytical methodologies is the separation and purification of the strontium, both to remove radionuclides which may interfere with subsequent β-counting and to free it from the large quantities of inactive substances typically present, i.e., calcium in seawater. In the 1950s, the oxalate technique was used to separate Sr and Ca (Miyake and Sugiura, 1955). They applied the fuming HNO3 method for purification of Sr. On the other hand, carbonate techniques (Sugihara et al., 1959) were used for 90 Sr determination for the Atlantic samples in the 1950s. Shirasawa and Schuert (1968) used similar procedures to those developed by Rocco and Broecker (1963) in which the oxalate technique was adapted. During the GEOSECS period, the oxalate technique was applied for 90 Sr determination in the world’s oceans (Bowen et al., 1980). Classical methods for the separation of strontium from calcium rely upon the greater solubility of calcium nitrate in
Radiometric determination of anthropogenic radionuclides in seawater
147
fuming nitric acid. Those procedures require numerous steps including repeated precipitation in strong nitric acid. Therefore, various alternative methods for separation have been proposed; precipitation methods of strontium sulfate and strontium rhodizonate (Weiss and Shipman, 1957), sorption of strontium on an ion-exchange resin from a solution of chelating agent such as CyDTA and EDTA (Noshkin and Mott, 1967). These methods, however, had not improved into shorter analytical procedures because the precipitation and extraction methods yield strontium fractions containing significant amounts of calcium. In the late 1970s, Kimura et al. (1979) proposed the use of macrocyclic polyethers for the separation of strontium and calcium. In the 1990s, extraction chromatography, using a solution of 4,4 (5 )-bis(tertbutylcyclohexano)-18-crown-6 in 1-octanol sorbed on an inert substrate, has been developed for the separation of strontium and calcium (Horwitz et al., 1990). Recently membrane filter coating crown ether was developed for separation of strontium from others (Lee et al., 2000; Miró et al., 2002). These modern techniques have contributed to down-sizing to small volumes of samples. On the other hand, large volumes of seawater have been used for determination of 90 Sr in seawater due to the low concentrations of 90 Sr. Therefore, the current practical method for separation and purification of 90 Sr in seawater still contains precipitation at the first step. There are some problems in the current practical methods using the carbonate technique; one is a lower recovery of Sr with a range from 30 to 60% and another is a long radiochemical separation. Since the radioactivity of 90 Sr in seawater is even lower in the surface water at present, improvement of the Sr recovery is one of the key issues for determination of 90 Sr activity in seawater. Another key point is that the Ca/Sr ratio in the carbonate precipitates is remarkably reduced from that in seawater. To improve 90 Sr determination in seawater, we re-examined the Sr separation technique for seawater samples. 3.2. Method 3.2.1. Sampling and materials Since 90 Sr concentrations in seawater are low, large volumes (50 to 100 L) of seawater are still required for assay of 90 Sr. Surface water samples of 100 L were collected with a pumping system on board. Deep water samples were collected with 100 L GoFlo type samplers. All of the water samples were filtered through a fine membrane filter (Millipore HA, 0.45 µm pore size) immediately after sampling. 90 Sr was assayed as 90 Y using β-counting following the radiochemical separation described in detail as follows: 3.2.2. Pre-concentration of 90 Sr The co-precipitation method has generally been used for extracting Sr from large volumes of seawater. Both the oxalate and carbonate techniques have been applied for pre-concentration of Sr from large volume water samples. The pre-concentration of 90 Sr with carbonate is performed by adding 500 g NH4 Cl and 500 g Na2 CO3 to 100 L seawater. To improve the recovery of 90 Sr using the carbonate technique, it is essential to remove Mg as hydroxide at pH = 12 from the sample seawater before performing the carbonate precipitation. 3.2.3. Radiochemical separation In the first step of the radiochemical separation, Sr is separated from calcium as an oxalate precipitate at pH = 4. After dissolution of the oxalate precipitate, Ra and Ba are removed
148
M. Aoyama and K. Hirose
with Ba chromate. After Ra and Ba are removed as precipitates, Sr is recovered as a carbonate precipitate. Further purification of Sr is carried out using fuming nitric acid to remove Ca. After 90 Sr–90 Y equilibrium has been attained, co-precipitation of 90 Y with ferric hydroxide is carried out and the solid is then mounted on a disk for counting. 3.2.4. β-Counting β-Counting of 90 Y is carried out by gas proportional counting with external solid samples. A typical efficiency of 90 Y counting by gas proportional counting with a thin window is ca. 40%. The detection limit of 90 Y is several mBq when the counting time is 360 min. 3.3.
90 Sr
concentration in ocean surface water
The distribution of 90 Sr concentrations in surface waters of the world ocean has been summarized in the IAEA-MEL research project (Livingston and Povinec, 2002; Povinec et al., 2005). The higher 90 Sr concentrations in the surface waters, ranging from 10 to 100 Bq m−3 , occurred in the Irish Sea, Black Sea and Baltic Sea, which reflects the effect of the Chernobyl fallout and the discharge of radioactive wastes from the Sellafield nuclear reprocessing plants. The level of surface 90 Sr ranged from 1 to 10 Bq m−3 for most of the ocean areas in the Northern Hemisphere, such as the North Pacific and North Atlantic Oceans, while for the oceans in the Southern Hemisphere such as the South Atlantic and Antarctic Oceans, 90 Sr concentrations were less than 1 Bq m−3 . The 90 Sr concentrations measured in surface waters in 1997 were in the range of 1.5– 2.2 Bq m−3 (Povinec et al., 2003). The highest 90 Sr surface concentration was observed in the mid-latitude region of the NW Pacific at St. 3. The geographical distribution of 90 Sr in the NW Pacific surface waters in 1997 is homogeneous, as for 137 Cs. However, still lower surface 90 Sr was observed in the Equatorial region near 10◦ N. The difference between maximum and minimum surface 90 Sr concentrations in the NW Pacific is smaller than for 137 Cs. 3.4.
90 Sr
concentrations in the water column in the Pacific Ocean
90 Sr profiles as well as 137 Cs in the water columns of the central NW Pacific Ocean at IAEA’97
(Figure 4), as well as GEOSECS and Hakuho-Maru stations are very similar to those of 137 Cs (Povinec et al., 2003). In the Kuroshio Extension and subtropical regions, compared with 90 Sr, the 137 Cs concentrations are higher by a factor of about 1.6, which represents a typical global fallout ratio for these radionuclides, whereas significantly low 137 Cs/90 Sr activity ratios (around 1) occurred in the Equatorial Pacific near 10◦ N, as more 90 Sr attached to CaCO3 particles from the atolls was deposited in close-in fallout than for 137 Cs. Low 137 Cs/90 Sr ratios were also observed in soil and sediment samples collected in Bikini, Enewetak and Rongelap atolls (Robison and Noshkin, 1999). 90 Sr in the surface water has been transported to deeper water layers than 137 Cs due to the co-precipitation with Ca in carbonate materials, which are later dissolved in the deep ocean. In the western North Pacific, there was also a remarkable decrease in surface and sub-surface concentrations between the GEOSECS and IAEA samplings-from about 4 to 1.5 Bq m−3 for 90 Sr and from about 6.5 to 2 Bq m−3 for 137 Cs, respectively. As we have seen in Figures 4 and 6, the distribution patterns of 137 Cs and 90 Sr in the water column were quite different from 239,240 Pu, because these elements
Radiometric determination of anthropogenic radionuclides in seawater
149
Fig. 4. 90 Sr profiles in the Pacific Ocean (using data of Povinec et al., 2003).
are soluble and conservative with much lower distribution coefficients (IAEA, 1985) than Pu. Therefore, most of the 137 Cs and 90 Sr deposited from global fallout into the ocean still remains in the water column with the highest concentrations in surface and sub-surface waters.
4. Analytical method for plutonium in seawater 4.1. Background Plutonium in the environments consists of a number of different isotopes: 238 Pu, 239 Pu, 240 Pu, and 241 Pu. 239 Pu and 240 Pu are the most abundant plutonium isotopes, and have radioactive half-lives of 24,110 and 6563 years, respectively. 238 Pu and 241 Pu, with radioactive half-lives of 87.7 and 14.1 years, respectively, are also important isotopes of plutonium despite having lower atom concentrations. All of the plutonium isotopes except 241 Pu are α-emitters; the αparticle energies of 238 Pu, 239 Pu and 240 Pu are 5.49, 5.10 and 5.15 MeV, respectively, while the maximum β-particle energy of 241 Pu is 5.2 keV. Therefore, α-spectrometry is effective for determining plutonium concentrations in seawater. However, the sum of 239 Pu and 240 Pu activities has generally been measured as the plutonium concentration because the separation between 239 Pu and 240 Pu is difficult using α-spectrometry due to the similar energy range of their α particles. On the other hand, measurements of 241 Pu were carried out by counting in situ growth 241 Am (half-life: 433 years). The activity of 241 Pu was calculated from the following equation # $−1 A(241 Pu) = A(241 Am){1 − T1/2,2 /T1/2,1 } exp(−λ1 t) − exp(−λ2 t) , (1) where A(241 Pu) and A(241 Am) are the activities of 241 Pu at separation time and ingrown 241 Am measured at re-count time, respectively, T 241 Pu and 1/2,1 and T1/2,2 the half-lives of 241 Am, respectively, and λ and λ the decay constants of 241 Pu and 241 Am, respectively. 1 2
150
M. Aoyama and K. Hirose
Recently, liquid scintillation spectrometers equipped with pulse height analysis and pulse shape analysis techniques have been developed. The low-level liquid scintillation counting system allows us to measure the spectra of both β- and α-emitters simultaneously, i.e., 241 Pu and 239,240 Pu, even at the low-level activity of environmental samples (Lee and Lee, 1999). However, the sensitivity of liquid scintillation counting is not enough to measure Pu isotopes in seawater due to the higher background of liquid scintillation counting (about 1 cpm). The chemistry of plutonium in natural waters is generally complicated because plutonium shows four valences (III, VI, V and VI) and forms complexes with inorganic and organic ligands (Coleman, 1965). Choppin and Morgenstern (2001) summarized the chemical behavior of plutonium in the aqueous environment. The most important chemical property is its oxidation state because many chemical processes, such as solubility, hydrolysis and others, are largely affected by valence. The oxidation state of plutonium affects its biogeochemical processes in the marine environment and analytical processes such as concentration and separation. The oxidation of plutonium depends on pH in solution; lower oxidation states are stabilized in acidic solution, whereas higher oxidation states are more stable forms in alkaline solution. The reactivity of Pu(IV) and Pu(III) is generally greater than that of Pu(V) and 2+ Pu(VI) because Pu(V) and Pu(VI) exist as PuO+ 2 and PuO2 , respectively. However, Pu(IV) in natural aquatic conditions is difficult to form complexes with organic ligands such as humic acids because the free Pu(IV) ion concentration is extremely low due to strong hydrolysis. Speciation studies (Nelson and Lovett, 1978; Nelson et al., 1984; Hirose and Sugimura, 1985; Fukai et al., 1987) revealed that the dominant species of Pu dissolved in seawater are oxidized forms (V and/or VI). Orlandini et al. (1986) and Choppin (1991) showed that Pu(V) dominates the oxidized fraction and demonstrated the instability of Pu(VI) in natural waters. In most natural systems the interesting oxidation states are Pu(IV) and Pu(V), while Pu(III) and Pu(VI) are generally insignificant. Taking into account hydrolysis and complexation with carbonate species, a dominant species of Pu(V) under the conditions of artificial seawater is PuO+ 2 whereas Pu(IV) exists as hydroxo complexes (Pu(OH)4 ) (Choppin and Morgenstern, 2001). On the other hand, the chemical form of Pu in marine particulate matter is an organic complex of Pu(IV) (Hirose and Aoyama, 2002). The 240 Pu/239 Pu atom ratio and 238 Pu/239,240 Pu and 241 Pu/239,240 Pu activity ratios in fallout, which depend on the specific nuclear weapons design and test yield, are variable (Koide et al., 1985). The global fallout average 240 Pu/239 Pu atom ratio is 0.18, based on aerosols, soil samples and ice core data (HASL, 1973; Krey et al., 1976; Muramatsu et al., 2001; Warneke et al., 2002). However, different nuclear test series can be characterized by either higher or lower ratios; i.e., fallout from Nagasaki, the Nevada test site and the Semipalatinsk test site is characterized by generally lower 240 Pu/239 Pu ratios, 0.042, 0.035 as an average value, and 0.036, respectively (Komura et al., 1984; Hicks and Barr, 1984; Buesseler and Scholkovitz, 1987; Yamamoto et al., 1996), whereas elevated 240 Pu/239 Pu ratios (0.21–0.36) have been measured in soil samples from the Bikini atoll (Hisamatsu and Sakanoue, 1978; Muramatsu et al., 2001). Plutonium isotope signatures, therefore, are a useful tool to identify the sources of plutonium and to have a better understanding of their behavior in the marine environments. Originally the plutonium isotope signatures in environmental sample were determined by thermal ionization mass spectrometry (Krey et al., 1976; Buesseler and Halverson, 1987). Recent development of Inductively Coupled Plasma Mass Spectrometry (ICP-MS) has provided more data on the 240 Pu/239 Pu atom ratio in marine sam-
Radiometric determination of anthropogenic radionuclides in seawater
151
Fig. 5. Calculated concentrations of 239,240 Pu in surface waters of the Pacific box. Small numbers show observed values, which were obtained in the year indicated in parentheses.
ples (Kim et al., 2002; Kim et al., 2003, 2004). By analyzing the 240 Pu/239 Pu atom ratios in seawater samples, Buesseler (1997) revealed that there is a significant contribution of close-in fallout from the Pacific Proving Ground nuclear testing to the North Pacific deep water. The Pu isotope ratios such as the 240 Pu/239 Pu atom ratio and 238 Pu/239,240 Pu activity ratio were used to estimate the contribution of close-in fallout Pu to total Pu water column inventory in the western North Pacific (Hirose et al., 2006a). 4.2. Method 4.2.1. Sampling Pu concentrations in seawater are extremely low (see Figure 5). Large volumes (50 to 500 L) of seawater are still required for assay of plutonium as for americium too. Surface water samples of 100 to 500 L were collected with a pumping system on board. Deep water samples were collected with 100 L GoFlo type samplers. All water samples were filtered through a fine membrane filter (Millipore HA, 0.45 µm pore size) immediately after sampling. Both dissolved and particulate plutonium were assayed using α-spectrometry following radiochemical separation using anion exchange resin, described in detail as follows: 4.2.2. Pre-concentration of Pu Co-precipitation methods have been generally used for extracting trace amounts of Pu from large volumes of seawater, which process is usually performed on-board. In the early days, co-precipitation of Pu with ferric hydroxide or bismuth phosphate was carried out (Bowen et al., 1971; Imai and Sakanoue, 1973). However, low and erratic recoveries of trace Pu in the environmental samples led Hodge et al. (1974) to propose a simple concentration method including the partial precipitation of (Mg, Ca) hydroxides/carbonates by the addition of small amounts of sodium hydroxide. Typically 34 mL of 50% NaOH solution was added to a 200 L
152
M. Aoyama and K. Hirose
seawater sample after addition of tracer into a rapidly stirred solution. The resulting milky mixture of (Mg, Ca) hydroxides/carbonates was stirred for 1 to 2 h and the fine precipitate allowed to settle overnight. After that, the clear water was removed from the container and the precipitate separated from the residual solution by centrifuging at 2500 rpm. A further pre-concentration method of Pu in seawater, co-precipitation with MnO2 has recently been proposed (La Rosa et al., 2001; Sidhu, 2003). After the seawater sample is acidified, 242 Pu is added as a yield tracer and KMnO4 is added to change the chemical forms of Pu to a dissolved form (Pu(VI)). After adding NaOH and MnCl2 solution to the solution, Pu co-precipitates with the bulky manganese dioxide at pH = 9. Finally the settled MnO2 precipitate is collected. Another pre-concentration method for Pu has been developed (Chen et al., 2001). A seawater sample of 200 L is acidified to pH = 2 with 12 M HCl (200 mL). After addition of ferric chloride (2 g), a known amount of tracer (242 Pu) and K2 S2 O5 (100 g), the solution is stirred for 1 h. In this stage, all of the Pu species in solution are reduced to Pu(III). Co-precipitation of Pu with ferric hydroxide occurs at pH = 10 by adding dilute NaOH solution (0.5–1 M). The ferric oxide precipitate formed contains small amounts of Ca(OH)2 and Mg(OH)2 . 4.2.3. Radiochemical separation Relatively large amounts of precipitates were obtained at the pre-concentration stage. Although a process to reduce precipitate is usually required, a simple separation method was proposed (Hodge et al., 1974; Hirose and Sugimura, 1985). Precipitates ((Mg, Ca) hydroxides/carbonates) were dissolved with 12 M HCl and added to bring the acid strength to 9 M (three times the dissolved materials). One drop of 30% H2 O2 was added for each 10 mL solution, and the solution was heated just below boiling for 1 h. After the solution had cooled, Pu was isolated by anion exchange techniques (Dowex 1-X2 resin, 100 mesh; a large column (15 mm of diameter and 250 mm long) was used). The sample solution was passed through the column and was washed with 50 mL 9 M HCl. At this stage, Pu, Fe and U were retained on the resin, whereas Am and Th eluted. Fe and U fractions were sequentially eluted with 8 M HNO3 . After elution of U, the column was washed with 5 mL 1.2 M HCl. Finally the Pu fraction was eluted with 1.2 M HCl (100 mL) containing 2 mL of 30% H2 O2 . The solution was dried on a hot plate. The chemical yield was around 70%. In order to remove the large amounts of MnO2 or Ca(OH)2 and Mg(OH)2 , additional coprecipitation techniques were performed (La Rosa et al., 2001). The MnO2 precipitates were separated and dissolved in 2 M HCl with excess NH2 OH HCl to reduce it to Mn(II). To scavenge Pu with ferric hydroxide, 50 mg of Fe(III) was added to the solution and oxidized from Pu(III) to Pu(IV) with NaNO2 . The Fe(OH)3 was formed by the addition of ammonium hydroxide to pH = 8–9. After briefly boiling, the pH in the solution was adjusted to 6–7, at which residual amounts of MnO2 changed to Mn(II). The Fe(OH)3 precipitate was separated from the supernatant solution by centrifugation. The precipitate was dissolved in 14 M HNO3 . Finally the acid concentration of the solution was adjusted to 8 M HNO3 . The solution was passed through the anion exchange column (Dowex 1 × 8, NO3 form; 10 mm diameter and 150 mm long), so that Pu and Th were adsorbed onto the resin, whereas Am, U and Fe remained in the effluent. The column was washed with 8 M HNO3 and the Th fraction was eluted with 10 M HCl. The Pu fraction was eluted with 9 M HCl solution containing 0.1 M NH4 I. The chemical recoveries were 40 to 60%.
Radiometric determination of anthropogenic radionuclides in seawater
153
The Fe(OH)3 precipitates including Ca(OH)2 and Mg(OH)2 were dissolved with 12 M HCl and the acid concentration of the solution was about 0.1 M HCl (Chen et al., 2001). The solution was adjusted to pH = 9–10 with 6% ammonium hydroxide. The precipitates were settled and the supernatant solution was discarded. The precipitates were collected onto GF/A filter paper by filtration. The Ca2+ and Mg2+ were removed by repeating this precipitation process. The precipitates were dissolved in 12 M HCl and the acid concentration was adjusted to about 6 M. U, Po and Fe in the solution were removed by extraction with 80 mL 10% TIOA/xylene. After separation of the organic phase, 10 mg of Fe(III) was added to the aqueous solution and formed precipitates with ammonium solution at pH = 10. The precipitates were dissolved in 2–3 mL 12 M HCl. 50 mg NaNO2 was added to the solution. The HNO3 concentration in the solution was adjusted to 8 M with 14 M HNO3 . The solution was passed through an anion exchange column (10 mm diameter and 130 mm long, 100–200 mesh AG 1 × 4, NO3 form). The column was washed with 20 mL 8 M HNO3 and 50 mL 12 M HCl. The Pu fraction was eluted with 80 mL 2 M HCl containing 2 g NH2 OH HCl and 0.1 g NaCl. After the addition of 4 mL 14 M HNO3 , the eluate was evaporated to dryness. The chemical yield was 70 to 85%. Recently new types of extraction chromatographic resins, instead of anion exchange resins, have been proposed for radiochemical separation and purification of the actinides (Horwitz et al., 1990). Kim et al. (2002) have developed analytical techniques for Pu determination in seawater using ICP-MS, incorporating a sequential injection system including a TEVA-Spec (Eichrom Industries) column (6.6 mm diameter and 25 mm long) for radiochemical separation of Pu. TRU Resin™ (comprised of a 0.75 M solution of the bifunctional organophosphorus extractant octylphenyl-N ,N -di-isobutyl carbamoylphosphine oxide in tri-n-butyl phosphate immobilized on an inert porous polymeric resin) is used for separation of the actinides (Horwitz et al., 1990; Grate and Egorov, 1998; La Rosa et al., 2001; Sidhu, 2003). 4.2.4. Radiochemical separation of plutonium in particulate matter Filtered samples (100–1000 L) were digested with 14 M HNO3 and 12 M HCl in an appropriate beaker on a hot plate. 242 Pu was added as a tracer of the chemical yield. The solution was evaporated to dryness on a hot plate. The residue was dissolved with 8 M HNO3 and NaNO2 added on a hot plate. After filtration, the solution was passed through an anion exchange column (8 mm diameter and 150 mm long: Dowex 1 × 8, NO3 form). The column was successively washed with 20 mL 8 M HNO3 and 50 mL 9 M HCl. The Pu fraction was eluted with 100 mL of 1.2 M HCl including 1 mL of 30% H2 O2 . The chemical yield was more than 90%. 4.2.5. Electrodeposition Pu samples for α counting were usually electroplated onto stainless steel disks. The diameter of the stainless steel disk depends on the active surface area of the detectors. The electrodeposition was performed using an electrolysis apparatus with an electrodeposition cell consisting of a Teflon cylinder, a cathode of platinum and the anode of the stainless steel disk. The electrodeposition of Pu has been carried out at two conditions, i.e., in aqueous sulfuric acid medium (Talvitie, 1972) and in ethanol medium (Hirose and Sugimura, 1985). To allow homogeneous plating onto the disk, the disk surface has to be carefully cleaned by appropriate washing processes.
154
M. Aoyama and K. Hirose
The purified Pu fraction was dissolved in 7 mL of 0.05 M H2 SO4 and transferred into an electrodeposition cell using 3 mL of 0.05 M H2 SO4 solution three times. Two drops of methyl blue and 25% ammonium solution were added to the solution. The pH of the solution was adjusted to 2.5. The Pu was electroplated onto a stainless steel disk at 0.7 A/m2 for 4–5 h. The purified Pu fraction was dissolved in 1 mL 2 M HCl and transferred into an electrodeposition cell using 20 mL ethanol. The Pu was electroplated onto a stainless steel disk (30 mm diameter) at 15 V and 250 mA for 2 h. 4.2.6. α-Spectrometry The α-spectrometer consists of several vacuum chambers including a solid-state detector, a pulse height analyzer and a computer system. The detector, which is of the silicon surface barrier type (PIPS, energy resolution: <25 keV (FWHM), counting efficiency: 15–25%), has an active surface area of 450–600 mm2 and a minimum depletion thickness of 100 µm. The vacuum in the chamber is less than 200 mTorr using a vacuum pump. The counting time is more than 80,000 s. 4.3. Application to ocean samples 4.3.1. Surface 239,240 Pu concentrations in the open ocean The 239,240 Pu concentrations in surface waters of the world ocean have been evaluated from the comprehensive data based on the IAEA-MEL research program on World-wide Marine Radioactivity Studies (Livingston and Povinec, 2002; Povinec et al., 2004, 2005). The highest 239,240 Pu concentration in the surface waters (>100 mBq m−3 ) occurred in the Irish Sea, in which the effect of the Sellafield discharge (Pentreath, 1988) has still continued. The relatively high 239,240 Pu concentrations in the surface waters (10–50 mBq m−3 ) have been observed in the northern North Atlantic, Barents Sea, North Sea, Mediterranean Sea and English Channel. The 239,240 Pu concentrations in the surface waters of the Pacific, Atlantic and other oceans ranged from 1 to 10 mBq m−3 except in the Antarctic Ocean, which is less than 0.5 mBq m−3 . The behavior of surface 239,240 Pu in the Pacific has been reexamined (Hirose et al., 1992; Hirose and Aoyama, 2003a, 2003b). The 239,240 Pu concentration in surface water has decreased exponentially since 1970 although the decrease rates depend on the sea area. The decrease rates for surface 239,240 Pu concentration are generally faster than those of 137 Cs, which can be explained by biogeochemical processes. The 239,240 Pu concentrations in surface waters of the Pacific in 2000 (Figure 5) are estimated to be in the range from 0.3 to 2.7 mBq m−3 , which spatial variation is larger than that of 137 Cs. Higher surface 239,240 Pu concentrations appear in the Subarctic Pacific (Box 1; see Figure 1) and the eastern equatorial Pacific (Box 7). Higher surface 239,240 Pu concentrations in the Subarctic Pacific can be explained by rapid recycling of scavenged 239,240 Pu due to large vertical mixing in the winter (Martin, 1985; Hirose et al., 1999), whereas surface 239,240 Pu in the eastern equatorial Pacific may be maintained at a higher level by equatorial upwelling (Quay et al., 1983). The lower values appear in the mid-latitude region of the eastern North Pacific (Box 3), which is consistent with the higher decrease rate of 137 Cs in the Box 3. Another point is that estimated surface 239,240 Pu concentrations in Boxes 2 and 4 of the western North Pacific are higher than those in Boxes 3 and 5 of the eastern North Pacific. Furthermore, there is only a small difference of surface 239,240 Pu concentrations between the Northern Hemisphere and Southern Hemisphere mid-latitude regions, although its causes are still
Radiometric determination of anthropogenic radionuclides in seawater
155
unknown. As a result, in addition to physical processes, biogeochemical processes are significant in controlling the current geographic distribution of 239,240 Pu concentrations in surface waters of the Pacific (Bowen et al., 1980; Livingston and Anderson, 1983; Hirose et al., 1992; Hirose, 1997; Hirose et al., 2001a), taking into account the relatively short residence time of surface 239,240 Pu (Hirose and Aoyama, 2003a, 2003b). 4.3.2. Vertical profiles of 239,240 Pu in the ocean Figure 6 shows vertical profiles of 239,240 Pu observed in the mid-latitude and subtropical regions of the western North Pacific in 1997 (Povinec et al., 2003). All vertical profiles of 239,240 Pu in the ocean are characterized by a typical distribution pattern with surface minimum, subsurface maximum and gradual decrease with increasing depth (Bowen et al., 1980; Nagaya and Nakamura, 1984, 1987a, 1987b; Livingston et al., 2001; Hirose et al., 2001a). Maximum 239,240 Pu concentrations in the water column appeared at around 700 m depth (Povinec et al., 2003), which is consistent with those observed in the mid-latitudes of the North Pacific in the 1980s (Hirose et al., 2001a) and the early 1990s (Livingston et al., 2001). The shape of the 239,240 Pu peak depends on the sea area, that is, a rather broad peak of 239,240 Pu maximum has been observed in the Sea of Japan (Yamada et al., 1996; Hirose et al., 1999, 2002; Ito et al., 2003), and a smaller 239,240 Pu maximum compared with that in the North Pacific occurred in the South Pacific (Chiappini et al., 1999). It is reported that the subsurface maximum of 239,240 Pu in mid-latitudes of the North Pacific has moved into deeper water during the past two decades (Livingston et al., 2001). This pattern of vertical 239,240 Pu profiles and their temporal change has been demonstrated by particle-scavenging processes, the removal of 239,240 Pu by sinking particles following regeneration of dissolved 239,240 Pu due to the biological degradation of particles (Livingston and Anderson, 1983; Hirose, 1997). The water column inventory of 239,240 Pu in the western North Pacific was 100–130 Bq m−2 (Hirose et al., 2001a; Povinec et al., 2003), which is the same order of magnitude as previous values observed in the western North Pacific (Bowen et al., 1980; Livingston et al., 2001). However, although the sediment inventories of 239,240 Pu were not taken into account, this water column 239,240 Pu inventory is about three times larger than that estimated from global fallout (Hirose et al., 1987, 2001b; UNSCEAR, 2000). This finding suggests that most of the 239,240 Pu inventory in the water column in the western North Pacific has originated from sources other than global fallout, such as close-in fallout, which is consistent with the fact that close-in fallout from testing at the Pacific Proving Grounds is the most important source of plutonium in the North Pacific according to a plutonium isotope study (Buesseler, 1997). It is probable that the present water column inventories of 239,240 Pu are controlled by horizontal advection and the bottom topography of the North Pacific instead of deposition due to global fallout and close-in fallout, because the major deposition of 239,240 Pu originating from atmospheric nuclear weapon testing occurred in the 1950s and early 1960s. There is a possibility that the apparent inventories of plutonium depend on depth. Therefore, 2300 m depth inventories of 239,240 Pu have been investigated to elucidate the characteristics of the spatial distribution of the 239,240 Pu water column inventories (Livingston et al., 2001; Ikeuchi et al., 1999). We calculated the 2300 m 239,240 Pu inventories at two stations (27◦ 48 N 130◦ 44 E: 66 Bq m−2 , 27◦ 45 N 130◦ 45 E: 73 Bq m−2 ) of the Shikoku Basin (Ikeuchi et al., 1999). The water column inventory of 239,240 Pu in the ocean is calculated from the 239,240 Pu
156
M. Aoyama and K. Hirose
Fig. 6. Depth distribution of 239,240 Pu concentrations (dots) and 238 Pu/239,240 Pu activity ratios (squares) in seawater of the NW Pacific Ocean (IAEA’97 expedition data).
Radiometric determination of anthropogenic radionuclides in seawater
157
vertical profile (Livingston et al., 2001). The higher 2300 m 239,240 Pu inventory was observed in the Sikoku Basin (Ikeuchi et al., 1999). The 2300 m 239,240 Pu inventories seem to be greater than those in mid-latitudes of the central North Pacific (35–53 Bq m−2 ) and the same order of magnitude as in the sea area near Bikini and Enewetak Atolls observed in 1997 (61 and 74 Bq m−2 ) (Livingston et al., 2001).
5. Sequential treatment of seawater samples It is cost-effective to use the same seawater sample for 137 Cs, Pu and 90 Sr determinations. In the first step, 137 Cs extraction by the AMP method described in Section 2.2 is performed. Then, Pu is extracted by the method described in Section 4.2. The last step is to perform carbonate precipitation to separate Sr from the seawater samples by the method described in Section 3.2.
6. Conclusion The analysis of anthropogenic radionuclides in seawater is still a challenging issue in the fields of oceanography and marine ecology because anthropogenic radionuclides are powerful tracers of ocean circulation and biogeochemical processes and are directly related to the effects on marine ecosystems. The analytical methods for anthropogenic radionuclides have been developed over the past four decades. In these studies, several kinds of analytical techniques have been proposed for each radionuclide; the major methods involve radiometric techniques such as γ -spectrometry, β-counting and α-spectrometry. Recently mass spectrometry such as ICP-MS and Accelerator Mass Spectrometry (AMS) and thermal ionization mass spectrometry (TIMS) have been developed for analysis of long-lived radionuclides such as plutonium (Buesseler and Halverson, 1987; Povinec et al., 2001). Pu analysis using sector field-ICPMS has succeeded in reducing sampling volumes to around 10 L of seawater (Kim et al., 2002). However, the radiometric methods are still more effective for 137 Cs and 90 Sr analyses in seawater. The analyses of anthropogenic radionuclides in seawater require large volumes of sample seawater and relatively long radiochemical separation and purification processes. Longer radiochemical processes cause lower and unstable chemical yields. The recent improvement of the 137 Cs analysis leads to higher and stable chemical yields (>90%) for a simple 137 Cs separation and concentration process. Therefore, the recommended procedure can be described for determination of 137 Cs in seawater. On the other hand, it is difficult to describe the recommended procedures for Pu and Sr analyses because of their lower and unstable chemical yields, although Pu analysis will be improved by using ICP-MS and small volumes of samples in future. Further studies to develop more convenient methodologies for Pu and Sr are needed. The comprehensive studies of radioactivity in the world-wide marine environment during the past decade have been achieved to have better understanding of the present levels of anthropogenic radionuclides in the world oceans and to establish a database of marine environmental radioactivity. These results reveal that anthropogenic radionuclides are a powerful tool to understand the decadal change of the oceans; that is, time-series and cross-section data
158
M. Aoyama and K. Hirose
have been used to verify the general ocean circulation model and biogeochemical modeling (Nakano and Povinec, 2003a, 2003b; Tsumune et al., 2003a, 2003b). Therefore, marine radioactivity studies should provide important information for understanding climate change and its resulting effects on the marine environment.
References Aoyama, M., Hirose, K. (1995). The temporal and spatial variation of 137 Cs concentrations in the western North Pacific and marginal seas during the period from 1979 to 1988. J. Environ. Radioact. 29, 57–74. Aoyama, M., Hirose, K. (2003). Temporal variation of 137 Cs inventory in the Pacific Ocean. J. Environ. Radioact. 69, 107–117. Aoyama, M., Hirose, K. (2004). Artificial radionuclides database in the Pacific Ocean: HAM database. Sci. World J. 4, 200–215. Aoyama, M., Hirose, K., Miyao, T., Igarashi, Y. (2000). Low level 137 Cs measurements in deep seawater samples. Appl. Radiat. Isotopes 53, 159–162. Aoyama, M., Hirose, K., Miyao, T., Igarashi, Y., Povinec, P.P. (2001). 137 Cs activity in surface water in the western North Pacific. J. Radioanal. Nucl. Chem. 248, 789–793. Aoyama, M., Hirose, K., Komura, K., Nemoto, K. (2004). Temporal variation of 137 Cs distribution and inventory along 165 deg. E in the North Pacific since 1960s to the present. In: International Conference on Isotopes in Environmental Studies—Aquatic Forum 2004. IAEA-CN-118, pp. 256–257. Bowen, V.T., Wong, K.M., Noshkin, V.E. (1971). Plutonium 239 in and over the Atlantic Ocean. J. Mar. Res. 29, 1–10. Bowen, V.T., Noshkin, V.E., Livingston, H.D., Volchok, H.L. (1980). Fallout radionuclides in the Pacific Ocean: vertical and horizontal distributions, largely from GEOSECS stations. Earth Planet. Sci. Lett. 49, 411–434. Buesseler, K.O. (1997). The Isotopic signature of fallout plutonium in the North Pacific. J. Environ. Radioact. 36, 69–83. Buesseler, K.O., Halverson, J.E. (1987). The mass spectrometric determination of fallout 239 Pu and 240 Pu in marine samples. J. Environ. Radioact. 5, 425–444. Buesseler, K.O., Scholkovitz, E.R. (1987). The geochemistry of fallout plutonium in the North Atlantic: II. 240 Pu/239 Pu ratios and their significance. Geochim. Cosmochim. Acta 51, 2623–2637. Chen, Q.J., Aarkrog, A., Nielsen, S.P., Dahlgaard, H., Lind, B., Kolstad, A.K., Yu, Y. (2001). Procedures for Determination of 239,240 Pu, 241 Am, 237 Np, 234,238 U, 228,230,232 Th, 99 Tc and 210 Pb–210 Po in Environmental Materials. Riso-R-1263 (EN), Riso National Laboratory, Roskilde, Denmark. Chiappini, R., Pointurier, F., Millies-Lacroix, J.C., Lepetit, G., Hemet, P. (1999). 240 Pu/239 Pu isotopic ratios and 239+240 Pu total measurements in surface and deep waters around Mururoa and Fangataufa atolls compared with Rangiroa atoll (French Polynesia). Sci. Total Environ. 237/238, 269–276. Choppin, G.R. (1991). Redox speciation of plutonium in natural water. J. Radioanal. Nucl. Chem. 147, 109–116. Choppin, G.R., Morgenstern, A. (2001). Distribution and movement of environmental plutonium. In: Kudo, A. (Ed.), Plutonium in the Environment. Elsevier Science, Amsterdam, pp. 91–105. Coleman, G.H. (1965). The Radiochemistry of Plutonium. University of California, Lawrence Radiation Laboratory, AEC Contract No. W-7405-erg-48. Folsom, T.R. (1980). Some measurements of global fallout suggesting characteristics the N. Pacific controlling dispersal rates of certain surface pollutants. In: Isotope Marine Chemistry. Uchida Rokakudo Publ. Co., Tokyo, pp. 51–114. Folsom, T.R., Sreekumaran, C. (1966). Some reference methods for determining radioactive and natural cesium for marine studies. In: Reference Methods for Marine Radioactivity Studies. IAEA, Vienna. Annex IV. Fowler, S.W., Ballestra, S., La Rosa, J., Fukai, R. (1983). Vertical transport of particulate-associated plutonium and americium in the upper water column of the Northeast Pacific. Deep-Sea Res. 30 (12A), 1221–1233. Fukai, R., Yamato, A., Thein, M., Bilinski, H. (1987). Oxidation states of fallout plutonium in Mediterranean rain and seawater. Geochem. J. 21, 51–57. Grate, J.W., Egorov, O.B. (1998). Investigation and optimization of on-column redox reactions in the sorbent extraction separation of americium and plutonium using flow injection analysis. Anal. Chem. 70, 3920–3929.
Radiometric determination of anthropogenic radionuclides in seawater
159
Harley, J.H. (1980). Plutonium in the environment—A review. J. Radiat. Res. 21 (1), 83–104. HASL (1973). Global atmospheric plutonium-239 and plutonium isotopic ratios for 1959–1970. In: HASL-237, Fallout Program Quarterly Summary Report. U.S. Department of Energy, Washington, DC, pp. III-2–III-28. Hicks, H.G., Barr, D.W. (1984). Nevada test site fallout atom ratios: 240 Pu/239 Pu and 241 Pu/239 Pu. UCRL-53499/1, Lawrence Livermore National Laboratory, Livermore, USA. Hirose, K. (1997). Complexation scavenging of plutonium in the ocean. In: Germain, P., Guary, J.C., Guéguéniat, P., Métivier, H. (Eds.), Radioprotection—Colloquies, vol. 32, pp. C2-225–C2-230. Hirose, K., Aoyama, M. (2002). Chemical speciation of plutonium in seawater. Anal. Bioanal. Chem. 372, 418–420. Hirose, K., Aoyama, M. (2003a). Analysis of 137 Cs and 239,240 Pu concentrations in surface waters of the Pacific Ocean. Deep-Sea Res. II 50, 2675–2700. Hirose, K., Aoyama, M. (2003b). Present background levels of 137 Cs and 239,240 Pu concentrations in the Pacific. J. Environ. Radioact. 69, 53–60. Hirose, K., Sugimura, Y. (1985). A new method of plutonium speciation in seawater. J. Radioanal. Nucl. Chem. 92, 363–369. Hirose, K., Aoyama, M., Katsuragi, K., Sugimura, Y. (1987). Annual deposition of Sr-90, Cs-137, Pu-239,240 from the 1961–1980 nuclear explosions: A simple model. J. Meteorol. Soc. Jpn. 65, 259–277. Hirose, K., Sugimura, Y., Aoyama, M. (1992). Plutonium and 137 Cs in the western North Pacific: Estimation or residence time of plutonium in surface waters. Appl. Radiat. Isotopes 43, 349–359. Hirose, K., Amano, H., Baxter, M.S., Chaykovskaya, E., Chumichev, V.B., Hong, G.H., Isogai, K., Kim, C.K., Kim, S.H., Miyao, T., Morimoto, T., Nikitin, A., Oda, K., Pettersson, H.B.L., Povinec, P.P., Seto, Y., Tkalin, A., Togawa, O., Veletova, N.K. (1999). Anthropogenic radionuclides in seawater in the East Sea/Japan Sea: Results of the first-stage Japanese–Korean–Russian expedition. J. Environ. Radioact. 43, 1–13. Hirose, K., Aoyama, M., Miyao, T., Igarashi, Y. (2001a). Plutonium in seawaters of the western North Pacific. J. Radioanal. Nucl. Chem. 248, 771–776. Hirose, K., Igarashi, Y., Aoyama, M., Miyao, T. (2001b). Long-term trends of plutonium fallout observed in Japan. In: Kudo, A. (Ed.), Plutonium in the Environment. Elsevier Science, Amsterdam, pp. 251–266. Hirose, K., Miyao, T., Aoyama, M., Igarashi, Y. (2002). Plutonium isotopes in the Sea of Japan. J. Radioanal. Nucl. Chem. 252, 293–299. Hirose, K., Aoyama, M., Povinec, P.P. (2003). Concentration of particulate plutonium in surface and deep-water samples collected during the IAEA’97 expedition. Deep-Sea Res. II 50, 2639–2647. Hirose, K., Aoyama, M., Igarashi, Y., Komura, K. (2005). Extremely low background measurements of 137 Cs in seawater samples using an underground facility (Ogoya). J. Radioanal. Nucl. Chem. 263, 349–353. Hirose, K., Aoyama, M., Igarashi, Y., Komura, K. (2006a). Oceanic 137 Cs: Improvement of 137 Cs analysis in smallvolume seawater samples using the underground facility (Ogoya). J. Radioanal. Nucl. Chem., doi:10.1007/s10967-008-0634-6. Hirose, K., Aoyama, M., Kim, C.S., Kim, C.K., Povinec, P.P. (2006b). Plutonium isotopes in seawater of the North Pacific: effect of close-in fallout. In: Povinec, P., Sanchez-Cabeza, J.A. (Eds.), International Conference on Isotopes and Environmental Studies. In: Radioactivity in the Environment, vol. 8. Elsevier, London. Hisamatsu, S., Sakanoue, M. (1978). Determination of transuranium elements in a so-called “Bikini Ash” sample and in marine sediment samples collected near Bikini atoll. Health Phys. 35, 301–307. Hodge, V.F., Hoffman, F.L., Foreman, R.L., Folsom, T.R. (1974). Simple recovery of plutonium, americium, uranium and polonium from large volumes of ocean water. Anal. Chem. 46, 1334–1335. Horwitz, E.P., Dietz, M.L., Diamond, H., LaRosa, J.J., Fairman, W.D. (1990). Concentration and separation of actinides from urine using a supported bifunctional organophosphorus extractant. Anal. Chim. Acta 238, 263–271. IAEA (1985). Sediment Kds and Concentration Factors for Radionuclides in the Marine Environment. Technical Report No. 247, International Atomic Energy Agency, Vienna, Austria. IAEA (2000). Global Marine Radioactivity Database (GLOMARD). IAEA-TECDOC-1146, IAEA, Vienna, Austria. Ikeuchi, Y., Amano, H., Aoyama, M., Berezhnov, V.I., Chaykovskaya, E., Chumichev, V.B., Chung, C.S., Gastaud, J., Hirose, K., Hong, G.H., Kim, C.K., Kim, S.H., Miyao, T., Morimoto, T., Nikitin, A., Oda, K., Pettersson, H.B.L., Povince, P.P., Tkalin, A., Togawa, O., Veletova, N.K. (1999). Anthropogenic radionuclides in seawater of the Far Eastern Seas. Sci. Total Environ. 237/238, 203–212. Imai, T., Sakanoue, M. (1973). Content of plutonium, thorium and protactinium in sea water and recent coral in the North Pacific. J. Oceanogr. Soc. Jpn. 29, 76–82.
160
M. Aoyama and K. Hirose
Ito, T., Povinec, P.P., Togawa, O., Hirose, K. (2003). Temporal and spatial variations of anthropogenic radionuclides in the Japan Sea water. Deep-Sea Res. II 50, 2701–2711. Kim, C.S., Kim, C.K., Lee, K.J. (2002). Determination of Pu isotopes in seawater by an on-line sequential injection technique with sector field inductively coupled plasma mass spectrometry. Anal. Chem. 74 (15), 3824–3832. Kim, C.K., Kim, C.S., Chang, B.U., Chung, C.S., Hong, G.H., Hirose, K., Pettersson, H.B.L. (2003). 240 Pu/239 Pu atom ratios insediments in the NW Pacific Ocean and its marginal seas. J. Radioanal. Nucl. Chem. 258, 265–268. Kim, C.K., Kim, C.S., Chang, B.U., Choi, S.W., Chung, C.S., Hong, G.H., Hirose, K., Igarashi, Y. (2004). Plutonium isotopes in seas around the Korean Peninsula. Sci. Total Environ. 318, 197–209. Kimura, T., Iwashima, K., Ishimori, T., Hamada, T. (1979). Separation of strontium-89 and -90 from calcium in milk with a macrocyclic ether. Anal. Chem. 51, 1113–1116. Koide, M., Bertine, K.K., Chow, T.J., Goldberg, E.D. (1985). The 240 Pu/239 Pu ratio, a potential geochronometer. Earth Planet. Sci. Lett. 72, 1–8. Komura, K. (2004). Background components of Ge detectors in Ogoya underground laboratory. Appl. Radiat. Isotopes 61, 179–183. Komura, K., Hamajima, Y. (2004). Ogoya Underground Laboratory for the measurement of extremely low levels of environmental radioactivity: Review of recent projects carried out at OUL. Appl. Radiat. Isotopes 61, 164–189. Komura, K., Sakanoue, M., Yamamoto, M. (1984). Determination of 240 Pu/239 Pu ratio in environmental samples based on the measurement of LX/X-ray activity ratio. Health Phys. 46, 1213–1219. Krey, P.W., Hardy, E.P., Pachucki, C., Rourke, F., Coluzza, J., Benson, W.K. (1976). Mass isotopic composition of global fall-out plutonium in soil. In: Transuranium Nuclides in the Environment. IAEA-SM-199/39, IAEA, Vienna, pp. 671–678. La Rosa, J.J., Burnett, W., Lee, S.H., Levy, I., Gastaud, J., Povinec, P.P. (2001). Separation of actinides, cesium and strontium from marine samples using extraction chromatography and sorbents. J. Radioanal. Nucl. Chem. 248 (3), 765–770. Lee, M.W., Lee, C.W. (1999). Determination of low level 241 Pu in environmental samples by liquid scintillation counting. Radiochim. Acta 84, 177–181. Lee, C.W., Hong, K.H., Lee, M.H., Cho, Y.H., Choi, G.S., Choi, Y.W., Moon, S.H. (2000). Separation and preconcentration of strontium from calcium in aqueous samples by supported liquid membrane containing crown ether. J. Radioanal. Nucl. Chem. 243, 767–773. Livingston, H.D. (1988). The use of Cs and Sr isotopes as tracers in the Arctic Mediterranean Seas. In: Charnock, H., Lovelock, J.E., Liss, P.S., Ehitfield, M. (Eds.), Tracers in the Ocean. Princeton Univ. Press, Princeton, NJ, pp. 161–176. Livingston, H.D., Anderson, R.F. (1983). Large particle transport of plutonium and other fallout radionuclides to the deep ocean. Nature 303, 228–230. Livingston, H.D., Povinec, P.P. (2002). A millennium perspective on the contribution of global fallout radionuclides to ocean science. Health Phys. 82 (5), 656–668. Livingston, H.D., Mann, D.R., Casso, S.A. (1987). Particle and solution phase depth distributions of transuranics and 55 Fe in the North Pacific. J. Environ. Radioact. 5 (1), 1–24. Livingston, H.D., Povinec, P.P., Ito, T., Togawa, O. (2001). The behavior of plutonium in the Pacific Ocean. In: Kudo, A. (Ed.), Plutonium in the Environment. Elsevier Science, Amsterdam, pp. 267–292. Martin, P.J. (1985). Simulation of the mixed layer at OWS November and Papa with several models. J. Geophys. Res. 90, 903–916. Miró, M., Gómez, E., Estela, J.M., Casas, M., Cerdà, V. (2002). Sequential injection 90 Sr determination in environmental samples using a wetting-film extraction method. Anal. Chem. 74, 826–833. Miyake, Y., Sugiura, Y. (1955). The Radiochemical Analysis of Radio-Nuclides in Sea Water Collected near Bikini Atoll. Report VI, No. 1, pp. 33–37. Miyake, Y., Saruhashi, K., Sugimura, Y., Kanazawa, T., Hirose, K. (1988). Contents of 137 Cs, plutonium and americium isotopes in the Southern Ocean waters. Pap. Meteorol. Geophys. 39, 95–113. Miyao, T., Hirose, K., Aoyama, M., Igarashi, Y. (2000). Trace of the recent deep water formation in the Japan Sea deduced from historical 137 Cs data. Geophys. Res. Lett. 27, 3731–3734. Muramatsu, Y., Hamilton, T., Uchida, S., Tagami, K., Yoshida, S., Robinson, W. (2001). Measurement of 240 Pu/239 Pu isotopic ratios in soils from the Marshall islands using ICP-MS. Sci. Total Environ. 278, 151–159. Nakano, M., Povinec, P.P. (2003a). Oceanic general circulation model for the assessment of the distribution of 137 Cs in the world ocean. Deep Sea Res. II 5, 2803–2816.
Radiometric determination of anthropogenic radionuclides in seawater
161
Nakano, M., Povinec, P.P. (2003b). Modelling the distribution of plutonium in the Pacific Ocean. J. Environ. Radioact. 69, 85–106. Nagaya, Y., Nakamura, K. (1984). Plutonium-239, plutonium-240, cesium-137, and strontium-90 in the central North Pacific. J. Oceanogr. Soc. Jpn. 40, 416–424. Nagaya, Y., Nakamura, K. (1987a). Artificial radionuclides in the western North Pacific. J. Oceanogr. Soc. Jpn. 32, 416–424. Nagaya, Y., Nakamura, K. (1987b). Artificial radionuclides in the western Northwest Pacific (II): 137 Cs and 239,240 Pu inventories in water and sediment columns observed from 1980 to 1986. J. Oceanogr. Soc. Jpn. 43, 345–355. Nelson, D.W., Lovett, M.B. (1978). Oxidation state of plutonium in the Irish Sea. Nature 276, 599–601. Nelson, D.M., Carey, A.E., Bowen, V.T. (1984). Plutonium oxidation state distributions in the Pacific Ocean during 1980–1981. Earth Planet. Sci. Lett. 68, 422–430. Noshkin, V.E., Mott, N.S. (1967). Separation of strontium from large amounts of calcium, with application to radiostrontium analysis. Talanta 14, 45–51. Orlandini, K.A., Penrose, W.R., Nelson, D.M. (1986). Pu(V) as the stable form of oxidized plutonium in natural waters. Mar. Chem. 18, 49–57. Pentreath, R.J. (1988). Sources of artificial radionuclides in the marine environment. In: Guary, J.C., Guegueniat, P., Pentreath, R.J. (Eds.), Radionuclides: A Tool for Oceanography. Elsevier, London, pp. 12–34. Perkins, R.W., Thomas, C.W. (1980). Transuranic Elements in the Environment. In: Hanson, W.C. (Ed.), Technical Information Center. U.S. Department of Energy, Washington, DC, pp. 53–82. Povinec, P.P., La Rosa, J.J., Lee, S.H., Mulsow, S., Osvath, I., Wyse, E. (2001). Recent developments in radiometric and mass spectrometry methods for marine radioactivity measurements. J. Environ. Radioact. 248 (3), 713–718. Povinec, P.P., Livingston, H.D., Shima, S., Aoyama, M., Gastaud, J., Goroncy, I., Hirose, K., Hynh-Ngoc, L., Ikeuchi, Y., Ito, T., LaRosa, J., Kwong, L.L.W., Lee, S.-H., Moriya, H., Mulsow, S., Oregioni, B., Pettersson, H., Togawa, O. (2003). IAEA’97 expedition to the NW Pacific Ocean—Results of oceanographic and radionuclide investigations of the water column. Deep-Sea Res. II 50, 2607–2637. Povinec, P.P., Hirose, K., Honda, T., Ito, T., Scott, E.M., Togawa, O. (2004). Spatial distribution of 3 H, 90 Sr, 137 Cs and 239,240 Pu in surface waters of the Pacific and Indian Ocean—GLOMARD database. J. Environ. Radioact. 76, 113–137. Povinec, P.P., Aarkrog, A., Buesseler, K.O., Delfanti, R., Hirose, K., Hong, G.H., Ito, T., Livingston, H.D., Nies, H., Noshkin, V.E., Shima, S., Togawa, O. (2005). 90 Sr, 137 Cs and 239,240 Pu concentration surface water time series in the Pacific and Indian Oceans—WOMARS results. J. Environ. Radioact. 81, 63–87. Quay, P.D., Stuiver, M., Broecker, W.S. (1983). Upwelling rates for the equatorial Pacific ocean derived from the bomb [14]C distribution. J. Mar. Res. 41 (4), 769–792. Reiter, E.R. (1978). Atmospheric Transport Processes. Part 4: Radioactive Tracers. Technical Information Center, U.S. Department of Energy, Washington, DC, pp. 1–605. Robison, W.L., Noshkin, V.E. (1999). Radionuclide characterization and associated dose from long-lived radionuclides in close-in fallout delivered to the marine environment at Bikini and Enewetak atoll. Sci. Total Environ. 237/238, 311–328. Rocco, G.G., Broecker, W.S. (1963). The vertical distribution of cesium 137 and strontium 90 in the oceans. J. Geophys. Res. 68, 4501–4512. Saruhashi, K., Katsuragi, Y., Kanazawa, T., Sugimura, Y., Miyake, Y. (1975). 90 Sr and 137 Cs in the Pacific waters. Rec. Oceanogr. Works Japan 13, 1–15. Science and Technology Agency, STA (1982). Procedures of Sample Treatment for Germanium Spectrometry, 20. STA, Tokyo (in Japanese). Shirasawa, T.H., Schuert, E.A. (1968). Fallout radioactivity in the north Pacific Ocean: Data compilation of Sr-90 and Cs-137 concentrations in seawater. In: Hardy, E.P., Rivera, J. (Eds.), Fallout Program Quarterly Summary Report HASL-197. Health and Safety Laboratory, U.S. Energy Research and Development Administration, Washington, DC, pp. I-66–I-94. Sidhu, R.S. (2003). A robust procedure for the determination of plutonium and americium in seawater. J. Radioanal. Nucl. Chem. 256 (3), 501–504. Sugihara, T.T., James, H.I., Troianello, E.J., Bowen, V.T. (1959). Radiochemical separations of fission products from large volumes of sea water. Anal. Chem. 31, 44–49. Sugiura, Y., Saruhashi, K., Miyake, Y. (1975). Evaluation on the disposal of radioactive wastes into the North Pacific. Pap. Meteorol. Geophys. 27, 81–87.
162
M. Aoyama and K. Hirose
Talvitie, N.A. (1972). Electrodeposition of actinides for alpha spectrometric determination. Anal. Chem. 44, 280–283. Tsumune, D., Aoyama, M., Hirose, K., Maruyama, K., Nakashiki, N. (2001). Calculation of artificial radionuclides in the ocean by an ocean general circulation model. J. Radioanal. Nucl. Chem. 248, 777–783. Tsumune, D., Aoyama, M., Hirose, K. (2003a). Behavior of 137 Cs concentrations in North Pacific in an ocean general circulation model. J. Geophys. Res. 108 (C8), 3262. Tsumune, D., Aoyama, M., Hirose, K. (2003b). Numerical simulation of 137 Cs and 239,240 Pu concentrations by an ocean general circulation model. J. Environ. Radioact. 69, 61–84. UNSCEAR (2000). Sources and Effects of Ionizing Radiation. United Nations, New York. Van R. Smit, J., Robb, W., Jacobs, J.J. (1959). AMP—Effective ion exchanger for treating fission waste. Nucleonics 17, 116–123. Warneke, T., Croudace, I.W., Warwick, P.E., Taylor, R.N. (2002). A new ground-level fallout record of uranium and plutonium isotopes for northern temperate latitudes. Earth Planet. Sci. Lett. 203, 1047–1057. Warner, M.J., Bullister, J.L., Wisegarver, D.P., Gammon, R.H., Weiss, R.F. (1996). Basin-wide distributions of chlorofluorocarbons CFC-11 and CFC-12 in the North Pacific: 1985–1989. J. Geophys. Res. C: Oceans 101 (C9), 20525–20542. Weiss, H.V., Shipman, W.H. (1957). Separation of strontium from calcium with potassium rhodizonate. Application to radiochemistry. Anal. Chem. 29, 1764–1766. Yamada, M., Aono, T., Hirano, S. (1996). 239+240 Pu and 137 Cs distributions in seawater from the Yamato Basin and the Tsushima Basin in the Japan Sea. J. Radioanal. Nucl. Chem. 210 (1), 129–136. Yamagata, N., Yamagata, T. (1958). Separation of radioactive caesium in biological materials. Bull. Chem. Soc. Jpn. 31, 1063–1068. Yamamoto, M., Tsumura, A., Katayama, Y., Tsukatani, T. (1996). Plutonium isotopic composition in soil from the former Semipalatinsk nuclear test site. Radiochim. Acta 72, 209–215.
163
Monte Carlo simulation of background characteristics of gamma-ray spectrometers— a comparison with experiment Pavel P. Povineca,b,∗ , Pavol Vojtylac , Jean-François Comanduccia a International Atomic Energy Agency, Marine Environment Laboratory, Monaco b Comenius University, Faculty of Mathematics, Physics and Informatics, Bratislava, Slovakia c European Organization for Nuclear Research (CERN), Geneva, Switzerland
1. Introduction Although the most effective way of increasing the factor of merit of a counting system is to increase counting efficiency and the amount of sample in a detection system, usually the only possible way is to decrease the detector background. Therefore in the past development of low level counting techniques investigations of detector background were the most frequent studies (Heusser, 1994, 1995; Povinec, 1994, 2004; Zvara et al., 1994). The background components of a typical low-level GE detector, not situated deep underground, are cosmic radiation (cosmic muons, neutrons and activation products of construction materials), radionuclides present in construction material (including the detector itself), radon and its progenies. For a present-day, carefully designed low-level HPGe spectrometer, the dominating background component is cosmic radiation, mainly cosmic muons (Heusser, 1995; Theodorsson, 1996; Vojtyla and Povinec, 2000; Laubenstein et al., 2004; Povinec, 2004). High energy cosmic particles are initiating a large number of complicated physical processes leading to background induction. In order to quantify the background, it is necessary to deal with rather complicated detector–shield assembles and physics. Obviously analytical solutions are not possible and the only way to solve this problem is Monte Carlo simulation of interaction processes. The development of a simulation code for background induction by cosmic rays is useful for optimization of a counting system in respect to its background characteristics. The prediction of detector background characteristics can be made before the system is built, that ∗ Corresponding author. Present address: Comenius University, Faculty of Mathematics, Physics and Informatics,
SK-84248 Bratislava, Slovakia. E-mail address:
[email protected] RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11005-6
© 2008 Elsevier B.V. All rights reserved.
164
P.P. Povinec et al.
can optimize the system characteristics, and it can also save money during the construction. Systematic investigations of the influence of various parameters on the detector background can be carried out as well (Vojtyla and Povinec, 2006). Monte Carlo simulations of background have been facilitated by the development of physics software and availability of powerful computers. As the processes leading to background induction are very weak processes that occur with low efficiencies, large numbers of events have to be processed to obtain results with acceptable statistical uncertainties. In a single Ge gamma-ray spectrometer there is no protection against cosmic muons, therefore a spectrometer with anticosmic shielding will greatly reduce background levels. The anticosmic shield can be made of gas or plastic scintillation detectors, which surround the lead shield housing the HPGe detector (Heusser, 1994; Vojtyla et al., 1994; Heusser, 1995; Semkow et al., 2002; Povinec et al., 2004; Schroettner et al., 2004). Another possibility is to use an anti-Compton spectrometer, which is a powerful tool for reducing the detector’s background as it combines both anticosmic and anti-Compton background suppression (Debertin and Helmer, 1998; Povinec et al., 2004). The most sophisticated spectrometric system is, however, the multi-dimensional anti-Compton gamma-ray spectrometer (Cooper and Perkins, 1971; Povinec, 1982) in which signals from the analyzing detectors (nowadays HPGe) create three-dimensional spectra (volumetric peaks) which can contain both coincidence and anticoincidence events.
2. Sources of detector background For better understanding of cosmic ray induced background and its effect on counting sensitivity of the detection system we need to review sources of the detector background. The main sources of background in low-level counting systems are: (i) (ii) (iii) (iv) (v)
cosmic radiation environmental radioactivity outside of the counting system radioactivity of construction materials of the shielding radioactivity of construction materials of the detector radon and its progenies.
2.1. Cosmic radiation Cosmic ray induced background is the most important component of the detector background, especially in surface and medium depth underground laboratories. Primary cosmic ray particles are consisting mainly of protons (90%), alpha particles (9%) and heavier nuclei that bombard the upper atmosphere with a flux of about 1000 particles per m2 per second. During their interactions with atmospheric nuclei (O, N, Ar, Ne, Kr, Xe) they produce secondary particles (neutrons, protons, pions, electrons, positrons, photons, muons and neutrinos), as well as cosmogenic radionuclides (e.g., 3 H, 7 Be, 14 C, etc.). The relative intensities of charged particles at sea level are as follows:
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
165
1 for muons 0.34 for neutrons 0.24 for electrons 0.009 for protons and 0.0007 for pions. The muon flux at sea level is about 1.9 cm−2 s−1 (Gaiser, 1990). The fluxes of cosmic ray particles depend on the geomagnetic latitude and on variations of solar activity during the eleven year solar cycle. From the variety of cosmic ray particles observed at sea level only muons, neutrons, electrons and photons are important for background induction. The soft component of cosmic rays (electrons and photons) can be easily shielded by heavy materials such as lead, concrete, steel, copper, etc. Protons from the nucleonic component of cosmic radiation have low intensity compared to neutrons and, moreover, they are converted to neutrons in nuclear reactions with shielding materials of the building and a metallic shield. Therefore only muons and neutrons (the hard component of cosmic rays) are important for background induction, as especially in the case of muons they can easily penetrate the shielding materials surrounding the detector. Buildings and other shielding structures absorb secondary neutrons (200 g cm−2 ) better than muons (2000 g cm−2 at sea level). Therefore, the relative composition of hard cosmic radiation varies according to the distance below the surface. At low depths (∼10 m w.e. (water equivalent is a unit frequently used in underground physics to normalize the depth for different rock overburden)), muons are by far the prevailing component of cosmic radiation. The muon energy spectrum is shifting to higher energies with increasing depth as the softer spectrum is gradually filtered out. Secondary neutrons (actually we should call them tertiary, to distinguish them from the neutrons produced in the upper atmosphere) are produced in the shielding (building materials, metallic shielding) by muon capture and fast muon nuclear interactions. At moderate depths (∼30 m w.e.), the negative muon capture prevails as the muon energy spectrum is rich in low energy muons that are likely to be stopped. On the other hand, fast muon nuclear interactions dominate deep underground (>100 m w.e.). However, not only particle flux magnitudes, but also mechanisms and efficiencies of background induction processes are important. Muons deposit tens of MeV energy in a HPGe detector by direct ionization. This is outside of interest for most of low-level counting applications. The main processes by which cosmic muons induce detector background are electromagnetic processes which are taking place in the metallic shield. Muons generate delta electrons that radiate bremsstrahlung initiating electromagnetic particle showers. Photons from showers are then forming a background radiation as they can penetrate through thicker layers of material. For muons of higher energies there is also a contribution from direct electron– positron pair formation and muon bremsstrahlung. For muons of lowest energies, muon decay positrons and electrons with energies up to 52 MeV, and negative muon nuclear capture, contribute as well. Processes including a production of nuclei are even more complicated to study. During muon capture, the atomic number Z of a nucleus decreases by one, and the nucleus radiates gamma-rays and fast neutrons when re-arranging its structure. These processes lead to activation of shielding and detector materials. Heusser (1995) estimated the flux of neutrons produced by muon capture to be 1.1×10−4 cm−2 s−1 for the muon flux of 8×10−3 cm−2 s−1 .
166
P.P. Povinec et al.
A part of high energy muons (>100 Ge) also interacts with nuclei via virtual photons, causing their spallation accompanied by emission of numerous hadrons. These reactions are another source of tertiary neutrons which are important in deep underground laboratories. A muon-induced background gamma-spectrum of a Ge detector is characterized by a dominant 511 keV annihilation peak and a continuum peaking at about 200 keV. The integral background count rates are within 0.6–1.6 s−1 at sea level, depending on the size of the detector, the detector and shield materials and the shield lining. Annihilation peak count rates are usually ∼0.01 s−1 . The comparison of background spectra measured with various shielding material shows that the continuum decreases with the increase in the atomic number Z of the shielding material and its maximum shifts to higher energies. The reason is in a stronger self-absorption of induced radiation in higher Z materials. Thin layers (mm) of lining have a significant influence on the shape and background counting rate (Vojtyla et al., 1994). An efficient way to eliminate the muon induced background is the use of anticosmic (guard) detectors surrounding the main analyzing detector (Heusser, 1994, 1995; Vojtyla et al., 1994; Heusser, 1995; Semkow et al., 2002; Povinec et al., 2004; Schroettner et al., 2004). These detectors register cosmic muons before they generate secondary particles in the shield, and anticoincidence electronics rejects such background counts. Gas proportional chambers placed inside the shield have a small efficiency for gamma-rays, and they can introduce additional radioactive contamination into the shield (Heusser, 1994, 1995; Vojtyla et al., 1994). NaI(Tl) detectors used as anticosmic shielding have much better efficiency for gamma-rays, however, due to contamination of crystals and glass of photomultipliers by 40 K, and a lower efficiency for cascade gamma-quanta, it is not advisable to use them inside a shield, unless they are part of an anti-Compton spectrometer (Povinec, 1982; Povinec et al., 2004). At present mostly plastic scintillators or large proportional chambers placed outside of the shield are used. The advantage of proportional chambers is their small count rate (∼200 s−1 ), and hence negligible corrections for dead time. Plastic scintillation detectors, although having higher counting rates, they can detect neutrons. The basic mechanisms of background induction by neutrons in Ge detectors are their capture and nuclear excitation by inelastic scattering. A neutron induced gamma-ray spectrum of a Ge detector contains a number of lines that can be attributed to nuclei in a detector working medium (Ge isotopes), and to nuclei of materials in the vicinity of the detector. Their intensities depend on the size of the detector, the cryostat and shielding materials, and most importantly on the shielding depth below the surface. Heusser (1993) compiled a large number of gamma-lines observed so far in background gamma-ray spectra. A characteristic feature of neutron induced background spectra are saw-tooth shaped peaks with energies of 595.8 and 691.0 keV corresponding to inelastic scattering of fast neutrons on 74 Ge and 72 Ge. The Ge nuclei de-excite almost immediately compared to the time of charge collection and the de-excitation energy is summed with the continuously distributed recoil energy of the nucleus. Therefore, these peaks have sharp edges from a lower energy side and tails towards higher energies. The summation does not take place during de-excitation of long-lived isomer states of Ge isotopes that de-excite after collection of the charge from the primary process. The most intensive lines of this kind are those of 75m Ge and 71m Ge (T1/2 = 47.7 s and 21.9 ms, respectively) with energies of 139.7 and 198.3 keV and count rates of up to 2 × 10−3 s−1 . Further, lines coming from the shielding materials can be identi-
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
167
fied in neutron induced background spectra, e.g., 206 Pb, 207 Pb, 63 Cu and 65 Cu (Heusser, 1993; Vojtyla et al., 1994). Neutrons also induce a background continuum and the annihilation peak, probably via hard gamma-rays emitted from highly excited nuclei. Hard gamma-rays loose their energy by Compton scattering, and initiate electromagnetic showers during their passage through shield materials. The shaped of the continuum is ruled by Z of the material and lining used, similarly as in the case of muon induced background. The neutron induced background can be minimized with antineutron shields consisting of layers made of low Z materials for their slowing down (e.g., polyethylene, paraffin) with admixture of boron or lithium for their capturing. However, their usefulness depends on shielding layers situated above the laboratory. If neutrons are mostly tertiary, i.e. produced in the shield, the neutron absorbing layer should be the innermost layer. This leads, however, to large shield dimensions and hence higher shield costs, as the neutron absorbing layer should be tens of cm thick to be efficient enough. For shields situated deeper underground (>10 m w.e.) where tertiary neutrons produced by cosmic muons prevail, a better solution would be to use an anticosmic shielding place outside of the shield. Although the anticosmic shielding is very effective way to decrease the detector background, it cannot be 100% effective in eliminating the effects of cosmic rays. However, a substantial overburden by placing the detection system deep underground can have profound effects on the background reduction. Down to a few tens of m w.e. cosmic rays still prevails, and an anticosmic shielding would be still useful. In the transition region to a few hundreds of m w.e., cosmic rays are still of importance, however, radioactivity free materials for the construction of detectors and shields become more and more important. Below about a thousand m w.e. the cosmic ray background is already negligible. The hadronic component of cosmic rays is efficiently removed by the first 10 m w.e. (Heusser, 1995), and comic muon induced particle fluxes (tertiary neutrons, photons) are ruled by the penetrating muon component. There are two other background components that were not covered earlier as they are negligible for surface laboratories, but important for specific applications carried out deep underground: neutrons can also be created in uranium fission and (α, n) reactions in surrounding materials. The estimated flux of about 4.5 × 10−5 cm−2 s−1 corresponding to the average U and Th concentrations in the Earth’s crust (Heusser, 1992) is negligible compared to that generated by cosmic rays in shallow depths, but it is important in experiments carried deep underground. These neutrons can also penetrate thick shields and induce background either directly or via a long-term activation of the construction material. The simplest way of eliminating this background source is to shield the U and Th containing walls with a neutron shielding material of sufficient thickness (at least 20 cm). The secondary cosmic ray particles can induce background not only in one event by interactions in the detector and surrounding materials that are shorter in time than the resolution time of electronics, but also by production of radioactive nuclei in materials intended to be used underground but processed and stored at sites with higher cosmic ray fluxes. The cosmic activation component is caused by short-lived radionuclides (e.g., 56–58 Co, 60 Co) produced primarily in other target elements such as copper or iron. The activated contamination can be sometimes more important than the residual primordial contamination. The most important mechanisms of the activation are secondary neutrons (at shallow depths or higher altitudes), and by negative muon capture and fast muon interactions (at underground laboratories). Many
168
P.P. Povinec et al.
short-lived radionuclides can be produced during the time of an intercontinental airplane flight due to high fluxes of secondary cosmic ray particles (mostly neutrons) at altitudes of about 10 km. At shallow and deep shielding depths, after filtering out secondary neutrons, fast muon interactions and negative muon capture activate materials either directly by disintegration and changing nuclei, or through tertiary neutrons. Compared to the direct background induction, the fast muon interactions at shallow shielding depths are more important than the muon capture. 65 Zn, 57 Co, 58 Co and 54 Mn are produced at saturated production rates of tens to hundreds of µBq kg−1 by the activation of germanium at sea level. 54 Mn is also produced at a high saturated production rate of 4.5 mBq kg−1 in iron used for the construction of shieldings (Brodzinski et al., 1990). The magnitude of background due to cosmic ray activation is, e.g., in the case of saturated 65 Zn activity of 0.44 mBq kg−1 in germanium at sea level in a 1 kg Ge detector, which gives an integral background rate above 50 keV of only 11 counts per day (Brodzinski et al., 1988). 2.2. Environmental radioactivity outside of the counting system The sources of background radiation are decay product of nuclei in U and Th decay series and 40 K. Generated gamma-radiation is critical and therefore a massive metallic shielding is required to protect detectors. A typical sea level gamma-ray flux above 50 keV at 1 m above ground is 10 photons cm−2 s−1 , representing the most intensive photons flux to which an unshielded detector is exposed. The cosmic photon flux is less than 1% of the environmental flux. A contribution from all other primordial radionuclides to the detector background is negligible. The average concentrations of primordial radionuclides in the continental upper crust are 850 Bq kg−1 of 40 K, 100 Bq kg−1 of 87 Rb, 44 Bq kg−1 of 232 Th and 36 Bq kg−1 of 238 U. Except for 232 Th, the concentrations are about twice as small in soils, but large variations can occur. The highest concentrations are found in granites and pegmatites. Secular equilibrium in decay series is rarely achieved in surface and near surface geological environments because the radionuclides in the chain migrate and take part in various chemical and physical processes. A special case is radon which breaks each series and can, due to its inert nature, migrate far from the rock or soil surface. Therefore radon will be treated separately in this chapter later. The impact of cosmogenic radionuclides produced in the atmosphere on the background of Ge detectors is usually negligible. Large amounts of radionuclides were released to the environment during the Chernobyl accident. Relatively long-lived 137 Cs is still present as a surface contamination of construction materials, therefore it should be checked and removed if present, especially in materials used for the detector construction. 2.3. Radioactivity of construction materials of the shielding A detector should be placed in a massive shield, best made of high-Z material, to minimize the environmental background. An anticoincidence detector with high efficiency for gammaradiation (e.g., NaI(Tl), BGO—bismuth germanate) can be used to further decrease the detector background, and usually also scattered photons, when working as an anti-Compton
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
169
gamma-ray spectrometer (Debertin and Helmer, 1998). A comparison of background gammaray spectra with and without massive shield can be found, e.g., in Sýkora et al. (1992). A background spectrum is usually dominated by 40 K (1461 keV), 214 Bi (609.3 and 1120.3 keV) and 208 Tl (583.1 and 2614.7 keV). The integral background of an unshielded Ge gamma detector of usual size is ∼100 s−1 . In the energy region above 200 keV and under the Compton edge of the 208 Tl peak, the background is reduced by two orders of magnitude by placing the detector inside the shield. Except for the 511 keV annihilation peak, attenuation factors for lines of the radionuclides in question are in agreement with the assumed absorption properties of the shielding layers for gamma-radiation of corresponding energy. On the other hand, the continuum in the spectrum is attenuated almost independently of the energy. It could be expected that the photons of lower energies would be absorbed more efficiently than those with higher energies. The reason that this does not occur is that the lower energy photons are continuously replenished by Compton-scattered photons of higher energies when passing through the thick layer of material. In this way, the continuous spectrum reaches an equilibrium shape with a general attenuation coefficient. Smith and Levin (1990) using Monte Carlo simulation found a mass attenuation coefficient of 0.045–0.050 cm2 g−1 for both high- and low-Z shielding materials. It is not worth increasing the thickness of the shielding over about 150 g cm−2 because other background components begin to dominate the total background. Muon-induced background together with a usually smaller neutron component rule the background beyond this thickness. Moreover, increasing the thickness of a high-Z material like lead enhances the production of tertiary neutrons by negative muon capture (mostly at low altitudes below sea level) and fast muon interactions (deeper underground). Paradoxically, the background increases slightly due to interactions of tertiary neutrons when increasing the shield thickness over 150 g cm−2 . Lead has proven to be the best material for bulk shields as it is cheap, has good mechanical properties, a high atomic number, low neutron capture cross-section, small neutron gamma yields and a low interaction probability for cosmic rays including the forming of radionuclides by activation. However, its intrinsic radioactivity can often be disturbing. Many authors investigated the origin of lead contamination in the past (Kolb, 1988; Eisenbud, 1973; Heusser, 1993). When using gamma spectrometry to check lead contamination, 210 Pb and its progenies 210 Bi and 210 Po are hidden contaminants as they cannot be clearly identified by gamma lines. Due to the long half-life of 210 Pb (22 years), the abundance of these radionuclides in lead can be much higher than expected from secular equilibrium. 210 Pb itself does not contribute to the background of Ge gamma-spectrometers as its very soft beta (Emax of 16.5 and 63 keV) and gamma (46.5 keV) radiations cannot escape self-absorption. The background is induced by the progeny 210 Bi emitting hard beta radiation with end-point energy of 1161 keV. Beta-radiation generates bremsstrahlung in the high-Z material, and also lead Xrays with energies of 72.8, 75.0, 84.9 and 87.4 keV. The continuum peaks at around 170 keV. As the end-point energy of 210 Bi is not sufficient for efficient pair production, the annihilation peak cannot be observed in the spectrum. Concentrations of 210 Pb reported in the literature range from 0.001 to 2.5 kBq kg−1 (Povinec et al., 2004). Ordinary contemporary lead has a 210 Pb massic activity of ∼100 Bq kg−1 . Lead with 210 Pb massic activity of about 1 Bq kg−1 is commercially available, lead with a lower activity is very expensive. Once 210 Pb is present in lead it cannot be removed by any chemical procedure and the only way to eliminate it is to wait several half-lives for its de-
170
P.P. Povinec et al.
cay. Old lead is an alternative to modern radioactive pure lead, however, its supply from water pipes, sunken shiploads or the ballast of sailing ships are limited. Moreover, old lead may not always be radioactive pure. If uranium was not removed, 210 Pb will have been supplied from the uranium decay series from the beginning of lead production. Another possibility is to use old iron plates or rods for the detector shielding. A recent iron may be dangerous because of 137 Cs contamination from Chernobyl fallout, or due to a contamination during its production. Therefore iron should be always tested before purchasing. Electrolytic copper is a better material, however, it is much more expensive than iron or ordinary lead. Also in situ production of cosmogenic radionuclides in copper may be disadvantageous when operating at shallow depths. 2.4. Radioactivity of construction materials of the detector The most sophisticated techniques for reducing the background will be unsuccessful if the detector is made of contaminated materials. Selecting radio-pure materials for the detector is important as these materials are close to the detector sensitive volume and radiation from them is detected with high efficiency. Most materials are contaminated with the primordial radionuclides, 40 K, 238 U and 232 Th. Complex production processes and contacts with various reagents increase the probability of contamination with primordial radionuclides. Usually, each production step must be checked to be sure that the produced material will be free of contaminants to the desired level. Physical and chemical procedures to which to processed material is subjected, result in isolated impurities (e.g., 210 Pb, 226 Ra, 228 Tl) as elements of the decay series have different properties and the chain breaks down. In fabricating metals, the redox potential is generally an important factor influencing the degree of radio-purity that can be achieved. That is probably the reason that copper can be produced very cleanly and is widely used in low-level counting applications. Its redox potential of 0.337 V is high compared to those of K, U and Th and their progenies. Cooper is routinely purified after melting by electrolytic dissolution and subsequent redeposition in solution even in large scale production. The contribution of radioactivity of detector materials to the background depends significantly on the detector type. It is usually lowest for semiconductor detectors as in this case a detection medium has to be very pure merely to enable the proper function of a semiconductor diode (a charge collection). The material is cleaned from primordial radionuclides and also from 137 Cs as a by-product through zone refining and crystal pulling. The background originates in parts placed around the crystal, and closer a part is the cleaner it should be. Materials used for crystal holders, end-caps and signal pins have to be selected most carefully. To obtain the best results the crystal holders should be made of electrolytic copper (<0.1 mBq kg−1 of 226 Ra, 208 Tl and 40 K), but this can reduce the transparency for low-energy gamma-rays. Low-Z aluminum is widely used, but it is usually contaminated with U and Th. However, aluminum devoid of these radionuclides to the 100 ppt level is available and is now used in low-level versions of detectors. Insulating materials (Teflon, PTFE, etc.) are usually radio-pure (<1 ppm K). All materials used must be cleaned of surface contamination using acid or lye or be electropolished.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
171
The care in selecting materials for detector encapsulation may be unnecessary if two other sources of background, molecular sieve and electronic circuits, are not removed. A molecular sieve is used to maintain the vacuum in the cryostat and can contain up to 100 Bq kg−1 of 226 Ra, 208 Tl and 40 K. Specially prepared activated charcoal with massic activities below 1 Bq kg−1 of 226 Ra and 20 Bq kg−1 is an alternative. Vacuum maintaining substances should be placed far from the crystal and behind the shield. If the cooled FET assembly is close to the crystal, the loss in energy resolution is negligible. For HPGe detectors, the background from contamination can be as low as a few counts per day in peaks of radionuclides in question, and therefore negligible compared to other components in systems operating in moderate shielding depths. 2.5. Radon and its progenies Background induced by radon is part of the environmental radioactivity background, but because of its specific nature it will be covered separately. Two radon isotopes—222 Rn and 220 Rn—belong to the 238 U and 232 Th series, respectively. We already discussed that these primordial radioelements are present in all construction materials. As radon is an inert gas, it diffuses after its formation from the walls to the vicinity of the detector. The radon concentration can vary between 0.1–100 Bq m−3 in living and working areas (Beláˇn et al., 1992). From the point of view of gamma background, the 222 Rn (T1/2 = 3.82 day) progenies— 214 Pb (242.0, 295.2 and 351.9 keV) and 214 Bi (609.3 and 1120.3 keV)—are more important than the 220 Rn (T1/2 = 55.6 s) progenies—-212 Pb (238.6 keV) and 208 Tl (583.1 and 2614.7 keV), as the 220 Rn half-life is much shorter. The easiest method of eliminating the radon background is to flush the cavity around a detector with nitrogen evaporating from the Dewar (Brodzinski, 1991; Sýkora et al., 1992). A flow rate of about 2 L/min is sufficient. A small overpressure that builds up expels radon from the detector cavity. More sophisticated systems use shield cavities that can be evacuated and filled with radon-free gas, e.g., nitrogen or old air (Heusser, 1991). Plastic materials should not be placed inside the shield as their surface draws radon progenies by electrostatic attraction. If these measures are not taken, the radon contribution to the detector background can be several counts per hour, even in well ventilated laboratories (Vojtyla et al., 1994). A comparison of ranges of various background components in a typical well-designed 1 kg low-level HPGe detector inside a lead shield with good ventilation is presented in Figure 1.
3. Fluxes of cosmic muons We have seen that the dominating background component of counting systems that are not placed deep underground is the flux of cosmic muons. We suppose that the detector itself and the shield have been constructed from selected radioactivity free materials. Usually the detection systems are placed deep enough, where a contribution to the background from secondary neutrons is negligible. Therefore, cosmic muons and their interactions with the counting system are the most important for the background induction. The first thing for assembling a code for simulation of background by cosmic muons we need to know their actual fluxes on the Earth. As cosmic muons are the most intensive and
172
P.P. Povinec et al.
Fig. 1. Ranges of background components in a typical well-designed 1 kg low-level HPGe detector inside a lead shield.
most penetrating particles at the sea level (we do not consider neutrinos), there are many experimental data available (Allkofer and Grieder, 1984). An ideal solution how to represent a muon flux would be to derive a function that would take into account all the parameters on which the muon flux depends. However, this would be too complicated approach as the flux depends on too many parameters, which change with time. The task is therefore to derive a simple standard description of the muon flux used as input to background simulation codes while being aware of magnitude and sense of possible deviations for a simulated system. We shall present only the basic parameters on which the cosmic muon flux depends, and the experimental basis for developing the code. 3.1. Geography The muon flux depends strongly on altitude, therefore we need to consider the real arrangements. We shall consider in our case that the detection system is placed almost at the sea level, representing differences in air mass overburden of less than 40 g cm−2 which can be neglected. The muon flux is higher at higher latitudes and lower near to the equator, as the Earth’s magnetic field deflects primary cosmic rays at equatorial latitudes more intensively due to the stronger magnetic field. A comparison of integral muon fluxes measured in various latitudes at sea level shows that within the latitude range 40◦ –60◦ (most of the industrially developed world), the muon flux varies by about 6% (Allkofer and Grieder, 1984)—therefore we can neglect these differences. 3.2. Muon momentum The differential muon flux depends strongly on the muon momentum. Allkofer and Grieder (1984) compiled available data and constructed the muon spectrum presented in Figure 2
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
173
Fig. 2. Spectrum of cosmic muons of both charges in vertical direction at sea level (a fit is based on data of Allkofer et al., 1971).
which has been used in our simulations. We see that the prevailing part of muons have momenta between 0.5–10 GeV/c, and the differential flux decreases rapidly with higher momenta. 3.3. Muon charge As cosmic muons are produced mostly by primary protons which prevail in cosmic radiation, positive muons will also prevail in the muon flux. This positive charge surplus depends on the multiplicity of particles produced during spallation reactions in the upper atmosphere, with a weak dependence of the muon charge ratio μ+ /μ− on the muon momentum. The differences in behavior of positive and negative muons in respect to background induction are important only for muon decay and muon capture, which are minor contributors to the total background. Therefore a momentum independent muon charge ratio μ+ /μ− = 1.28 can be accepted to simplify the model (Allkofer, 1975). 3.4. Polar direction The differential cosmic muon flux also depends on the polar angles, ϑ (the zenith angle) and ϕ (the azimuth angle). The major dependence is on the zenith angle. The muon flux in slanted directions is influenced by atmospheric absorption, muon decay and Earth’s magnetic field. Muons, which are produced mostly in higher altitudes, must travel a longer distance to the
174
P.P. Povinec et al.
Earth’s surface if they come from slanted direction, so that some of them, depending on their energy, lose all energy by ionization or decay. In general, the larger is the zenith angle ϑ, the smaller is the muon flux, and the spectrum is relatively depleted in lower energy muons. The Earth’s magnetic field bends muon trajectories depending on the muon charge sign. The trajectories of positive muons are different from those of negative muons when observed in the East–West plane. Different trajectory lengths cause different surviving probabilities, so that the differential muon flux depends also on the azimuth angle ϕ. This dependence is stronger for muons of lower momenta (below 5 GeV/c). In the eastern direction, a trajectory of a negative muon is shorter than that of a positive one produced at the same altitude, therefore the spectrum is relatively enriched in negative muons. In the western direction, it is contrary. There is a deviation from the general behavior for directions close to the horizontal for muons of the highest energies (above about 500 GeV/c) when the differential flux increases with the zenith angle ϑ (Allkofer and Grieder, 1984). This is explained by the following: there are two possible fates for pions produced in the upper atmosphere. Either they decay to muons or they initiate a nuclear reaction. A pion moving in almost horizontal direction spends most of its time in thin air so the chance for decay is higher compared to the concurring nuclear reaction. However, we do not need to include this effect to the model as it applies only to muons with the highest energies that have negligible fluxes (see Figure 2). Three parameters are required to describe the differential muon flux at a fixed point: the muon momentum p, the zenith angle ϑ and the azimuth angle ϕ. In principle, a fit over these three parameters could be done provided a sufficiently consistent and broad data would be available. The dependence of the differential muon flux j (ϑ, p) on the zenith angle ϑ can be expressed as (Allkofer, 1975) j (ϑ, p) = j (0, p) cosn(p) ϑ, where the exponent n(p) depends on the muon momentum p, differently for the eastern and western directions. The exponent decreases from p ≈ 3.3 with increasing muon momentum, and reaches zero for p > 30 GeV/c, when the muon flux is considered to be isotropic. In the case when a simulated detection system has axial symmetry around the vertical axis, the variations in azimuth angle ϕ are averaged out and ϕ-independent values of n(p) can be used. 3.5. Temporal variations When assessing the overall accuracy of any cosmic muon background simulation, we have to take into account that the muon flux varies also with time. Long-term variations are attributed to the eleven year solar cycle, diurnal and 27 day cyclic variations have also been observed. The solar wind, which is intensive during periods of high solar activity, declines the primary cosmic rays from their way to the Earth. The effect is most pronounced for secondary neutrons, but can also cause muon flux variations, which are usually within 5%. Short-term variations (from 2 h to 2 days) due to magnetic storms can occur as well. There are also seasonal variations of cosmic muon flux which can be observed mainly in middle latitudes. They are caused by variations of the tropopause height which rises from about 9 km in winter to about 15 km in summer. Therefore the height in which muons are mostly produced varies with the tropopause height. In summer, when muons are produced higher, most of them do not survive their longer way to the sea level and they decay. Therefore
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
175
the cosmic muon flux is less intensive in summer (by about 7% from the average) than in winter, when it is by about 6% greater than the average flux. Short-term variations known as barometric effect originate in variations of the atmospheric air pressure. At low pressure, the atmospheric depth through which muons have to penetrate is thinner, and more muons succeed to reach the Earth surface. Deviations by more than a few percent occur due to this effect, therefore background of a detection system situated above the surface without a proper anticosmic shielding should be corrected for the barometric effect. We have seen that we are dealing with too many parameters which affect the actual muon flux, and we cannot include them all into the simulation. The variations are usually smaller than 10%, therefore we cannot expect a better agreement of simulated results with reality. However, such accuracy √ is acceptable as the background enters the factor of merit in the square root (F = ε/ B, where ε is the efficiency of the detector and B is its background).
4. Physics of the simulation 4.1. Processes with muons Before covering the simulation codes a brief list of processes with muons that might be important for background induction will be discussed. How these processes are treated with GEANT will also be discussed so that the accuracy of simulation can be assessed. The details of the methods used in GEANT are given in the CERN (1990) report. 4.1.1. Ionization and delta electron production Muons are charged particles heavier then electrons that do not interact in strong interactions, and therefore they loose and deposit energy in matter mostly due to electromagnetic interactions. Because cosmic muons have high energies, the ionization process which is most important for massive charged particles, has some peculiarities, the most important of which is the production of energetic delta electrons. In collisions with atomic electrons, which are mediated by Coulomb interactions, atoms may be excited or electrons can get enough energy to leave atoms. The energy spectrum of liberated electrons is very wide, from almost zero energy up to the maximum energy (Tmax ) that can be transferred to an (almost) unbound electron given by the kinematics law Tmax =
2mc2 (γ 2 − 1) m 2 , m 1 + 2γ M + M
where m is the electron mass, M is the muon mass, γ and c have their usual meaning. For a muon with the momentum of 10 GeV/c, the maximum kinetic energy transferable to an electron is as much as 4.8 GeV. Clearly, electrons with such high energies can generate other particles mostly due to bremsstrahlung. The standard Bethe–Bloch formula for ionization losses of non-relativistic massive charged particles cannot be applied for ultra-relativistic cosmic muons. The presence of high energy delta electrons has two main consequences: (i) the energy lost by a particle in a thin layer of material has larger statistical fluctuations than expected from a simple Gaussian distribution,
176
P.P. Povinec et al.
(ii) there is a difference between the energy lost and the energy deposited in thin material layer as the energy transferred to an energetic delta electron may escape from the layer with the electron and its secondary radiation. A more realistic GEANT option is the explicit generation of delta rays with the calculation of quasi-continuous energy loss using the restricted Bethe–Bloch formula. The definition of delta electrons is quite vague and depends on the regions of energies of interest. If the electron kinetic energy cut, Tcut , is introduced, below which the electron is considered to deposit all its energy in one point, and not to transport it in any form to other points in the set-up, the restricted Bethe–Bloch formula for Tmax > Tcut has the form (CERN, 1992) √ 1 dE Tmax Tcut z2 Z β2 Tcut C δ = D 2 ln − 1+ , − − ρ dx 1 2 Tmax 2 Z Aβ where D = 4πNa re2 mc2 = 0.307 MeV cm2 g−1 . A, Z and ρ are the mass number, atomic number and density of the material, respectively; z is the particle charge, and I = 16Z 0.9 eV is the effective ionization potential. The factor δ is a correction for the density effect, and C is the shell correction term. For Tmax < Tcut the formula has a different form (CERN, 1992). It has been known for many years that most of the muon induced background in Ge gammaspectrometers is caused by bremsstrahlung radiation of delta electrons. As the energy region of interest is above a few tens of keV, the electron kinetic energy cut above which delta electrons should be explicitly generated should also be a few tens of keV, at least in the inner layers of the shield and in the detector itself. Therefore Tmax will always be greater than Tcut for cosmic muons. The differential cross-section of delta electron production in interactions of muons (spin 1/2) can be written as (CERN, 1992) T2 dσ 2 1 1 2 T + = 2πre m 1−β , dT β T2 Tmax 2E 2 where T is the electron kinetic energy, E is the muon energy, and the other symbols have the usual meaning. Far below the kinematic limit, Tmax , the delta electron spectrum is inversely proportional to the square of its kinetic energy, and the number of electrons produced above a given threshold is inversely proportional to the threshold energy. There are well established facts on delta electron production. The lower the delta electron cut, Tcut , is, the more realistic the simulation of the ionization process is achieved. However, the process requires more computing time as more particles have to be processed. 4.1.2. Direct electron–positron pair production When a muon of energy E moves in the field of a nucleus of charge Z, it can create an electron–positron pair through a fourth order QED process with a differential cross-section (CERN, 1985), 2 2 1−ν m ∂ 2σ φμ , = α 4 (Zλ- 2 ) φe + ∂v∂ρ 3π ν M where E+ − E− E+ + E− and ρ = . ν= E E
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
177
α is the fine structure constant, λ- is the Compton wavelength, E + and E − are the total energies of the positron and the electron, respectively; the terms φe and φμ can be found in the CERN (1985) report. There are kinematic ranges for the fraction of muon energy transferred to pair ν: 4m 3 √ M 1/3 e Z = νmin ν νmax = 1 − E 4 E and 0 = ρmin
% % % % ρ(ν) ρmax (ν) = 1 −
6m E 2 (ν − 1)
1−
4m . νE
There is an energy threshold of 2.04 MeV (4mc2 ) for this process. The dominant contribution comes from the low-ν region. In GEANT, this process is simulated for the muon energy range 1–104 GeV using parametrized expressions for cross-sections with an inaccuracy of no more than 10% (CERN, 1990). It is possible to set an energy cut for the resulting pair higher than the kinematic limit. In this case, the energy expended by the muon for the production of a pair is treated as a quasi-continuous loss, and the muon is absorbed at the interaction point. There is no reason to use higher cuts in simulations of background induction as the direct pair production is an important process, and an acceleration of computation is not significant. 4.1.3. Muon bremsstrahlung The muon bremsstrahlung is not often discussed, although muons are massive particles, and the process takes a place in the Coulomb field of the nucleus. We shall not present here any formulae for calculating a differential cross-section of this process as they are too complicated (CERN, 1985). Muon bremsstrahlung events are quite rare but very powerful so that a muon can lose a large fraction of its energy in one interaction. Therefore, it contributes significantly to mean energy losses of high energy muons (about 3 MeV/(g cm−2 )) for 1000 GeV muons in iron (CERN, 1985). The hard photon produced generates an electromagnetic shower in the material. Due to its high energy, a shower can contain up to 105 particles with energies above 50 keV. GEANT simulates this process in the muon energy range 1–104 GeV with accuracy of parametrized cross-sections not worse than 12% (CERN, 1990). As for the direct pair production, it is possible to set an energy cut below which the muon bremsstrahlung is treated as a quasi-continuous energy loss. However, this process is rare and the computation acceleration is negligible. 4.1.4. Multiple scattering To complete the list of electromagnetic processes with muons, multiple scattering of muons on atoms in the material should be covered as well. This process is characterized by numerous elastic collisions with the Coulomb field of nuclei, and to a lesser degree with the electron field, which leads to a statistical declination of a muon from its original direction. It is important only for muons below about 1 GeV/c, and it gains importance as the muon approaches zero kinetic energy. However, the expressions are too complex to be presented in this review. The first three electromagnetic processes with muons are compared in Figure 3, where macroscopic cross-sections in lead for muons with momenta up to 100 GeV/c are shown. The
178
P.P. Povinec et al.
Fig. 3. Macroscopic cross-sections of electromagnetic processes with muons in lead as used by the GEANT code.
cross-section of delta electron production is given for electrons with kinetic energies above 10 MeV. At this energy the ionization and radiation energy losses of electrons are equal. The cut for muon bremsstrahlung of 50 keV, and electron–positron pair production just above the kinematic limit are considered. The dominance of production of delta electrons for muons momenta below 100 GeV/c is dominant. 4.1.5. Muon decay Muon decay is a well-known weak process that occurs mostly when a muon has been stopped within the set-up. The cause is a relatively long mean life-time of 2.197 µs, which is prolonged due to the relativistic time dilatation for energetic cosmic muons. A positive (negative) muon decays to a positron (electron) and neutrinos escaping from the set-up according to the decay schemes μ+ → e+ + νe + ν μ
and
μ− → e− + ν e + νμ ,
respectively. A decay accompanied by a photon emission occurs in 1.4 ± 0.4% cases. There is a kinematic limit for the positron (electron) momentum in the muon rest system given as 1 m2 p0 = Mc 1 − 2 = 52.8 MeV/c 2 M if we consider mass-less neutrinos. The standard V –A weak interaction model gives for the positron (electron) momentum, pe , spectrum, neglecting terms of m/E (Morita, 1973) M 5 gμ2 c4 2pe 2 p(x) = , (1 − x) + ρ (4x − 3) x 2 with x = 7 3 9 Mc 16π h¯ which is in good agreement with experiment for ρ = 3/4.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
179
Most of the positrons (electrons) have high energies and are likely to generate electromagnetic showers if a muon decays in a high-Z material like lead. GEANT treats decay of particles in a general way so that the four-momenta of the decay products are generated with isotropic angular distribution in the center-of-mass system, and then transformed into the laboratory system. The consequence of such a general approach is the simplification of the momentum spectra of the decay products. The decay accompanied by gamma-ray emission is not simulated. However, it was found in a detailed simulation that the muon decay is only a minor process in background induction by cosmic muons. 4.1.6. Negative muon capture A concurrent process to negative muon decay in a material is muon capture. A negative muon is attracted by the nucleus, and if it does not succeed in decaying, it finally falls to the 1s level of a muonic atom. Due to its 207 times bigger mass compared to electron, a large part of the muon probability distribution function in the 1s state rests within the nucleus, especially in heavy nuclei. The relatively long life-time of muons allows the basic weak process μ− + p → n + νμ to take place. Simple kinematic calculations show that a free neutron could gain only 537 MeV of kinetic energy while the rest escapes with the muon neutrino. However, in a nucleus, a higher energy is transferred to the nucleus (typically 10–20 MeV) because of the non-zero proton momentum and nuclear effects. The binding energy of negative muons in heavy nuclei can be quite high, e.g., 10.66 MeV in lead. On the bases of experimental data (Suzuki et al., 1987) it can be calculated that muon capture without muon decay occurs in natural Pb, Cd, Cu, Fe and Al in 97%, 96%, 93%, 90% and 61%, respectively. As more than 99% of muons in a counting system equipped with a lead shield stop in lead due to its high mass fraction in the whole system (calculated in a simulation), practically all the stopped negative muons are captured without electron emission. The muon capture is therefore the most intensive source of tertiary neutrons in laboratories placed in moderate shielding depths. After the negative muon capture the excited nucleus de-excites by the emission of several neutrons. Emission of charged particles in heavy elements is suppressed due to the Coulomb barrier at these energies. In lead, 1.64 ± 0.16 neutrons are emitted in average (Morita, 1973). The probabilities of multiplicities of the emitted neutrons are 0.6% for no neutron, 59.1% for one neutron, 23.6% for two, 5.1% for three and 12.4% for four neutrons. The neutron energy spectrum for lead was measured by Schröder et al. (1974). It contains an evaporation part expressed as dN(E) ∝ E 5/11 e−E/θ dE in the energy region of 1 to 4 MeV. The effective nuclear temperature θ is 1.22 MeV. The part of the neutron spectrum above 4.5 MeV decreases exponentially (dN/dE ∝ exp(−E/Ed )), and is attributed to direct pre-compound emission. The decrement energy Ed is 8 ± 1 MeV, and the fraction of the high-energy distribution integrated from 4.5 to 20 MeV accounts for 9.7–1.0% of the total spectrum. Later Kozlowski et al. (1985) extended the measurements of the neutron spectrum up to 50 MeV and found a decrement energy Ed in the energy region
180
P.P. Povinec et al.
10–50 MeV of 8.6 ± 0.5 MeV, which is in accordance with that measured by Schröder et al. (1974). The de-excitation of a nucleus may be completed by emission of several gamma-rays, however, there are no reliable data on gamma-ray energies and intensities for the materials in question. Also X-rays with correspondingly higher energies can be emitted from muon-atomic levels during the descent of the muon to the 1s level. Although such X-rays are used for material analysis, they have never been identified in background spectra, and their contribution is considered to be negligible. 4.1.7. Fast muon nuclear interactions These interactions, also called photonuclear muon interactions, occur at high muon energies, and they are mediated via a virtual photon, interacting with nucleons. The total cross-section per nucleon can be expressed for muons above 30 GeV as (CERN, 1990) σ = 0.3(E/30)0.25 . It is considered to be constant (0.3 µb/nucleon) for muons below 60 GeV. The macroscopic cross-section for muons below 30 GeV in lead is only 2.0 × 10−6 cm−1 and increases only slowly for higher muon energies. As in muon bremsstrahlung, the fast muon interaction is characterized by rare but very hard events. Therefore, despite its small cross-section, it contributes to mean energy losses of high-energy muons. However, the contribution is still only about 5% (<0.5 MeV g−1 cm2 ) of the total energy losses of high energy muons in iron, and even less for heavier materials (RPP, 1994). GEANT treats fast muon nuclear interactions in the same way as the hadronic shower simulation code GEISHA (Fesefeldt, 1985). The energy and angle of a final state muon is generated according to the free quark parton model. The hadrons are generated in an approximate way. The virtual photon replaced by the real pion of random charge with the same kinetic energy and inelastic scattering of the pion on the nucleus is simulated. While the final state generated in this way gives a good approximation for calorimetric purposes, the kinematics of the final state may be a rather poor approximation (CERN, 1990). 4.2. Processes with electrons, positrons and photons Except for the negative muon capture and fast muon nuclear interactions, the muon interactions in a shield result in energetic electrons, positrons and photons. Therefore, processes with these particles will also be discussed as they mediate the interaction of a muon with a detector medium. 4.2.1. Ionization and delta electron production There are some important differences between interactions of electrons and positrons due to the quantum effect of the impossibility of distinguishing between two electrons. The first manifestation is ionization and delta electron production by electrons and positrons. The maximum energy transferable to a free atomic electron by a positron is as much as the initial positron kinetic energy, while for the electron, only half of the initial electron kinetic energy can be transferred. This leads to different formulae for continuous energy losses by ionization (Berger–Seltzer formulae; Sternheimer, 1971), as well as different formulae for delta electron
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
181
production by electrons (e− e− Möller scattering) and by positrons (e+ e− Bhabha scattering; CERN, 1992). GEANT uses explicit theoretical formulae for sampling these processes. 4.2.2. Bremsstrahlung Bremsstrahlung of electrons was theoretically investigated by Seltzer and Berger (1985) who calculated photon spectra for elements with Z = 6, 13, 29, 47, 74 and 92 in the electron kinetic range 1 keV–10 GeV, and they reported good (within a few percent) agreement with experiments. The parametrization introduces errors of bremsstrahlung cross-sections of less than 15% for electron kinetic energies below 1 MeV, and less than 6% for electron kinetic energies from 1 MeV to 10 TeV. The errors of photon spectrum parametrization are smaller than 12%, typically a few percent (CERN, 1990). Energy losses due to soft bremsstrahlung (photon below an energy cut than can be set by the user) are added to the quasi-continuous energy losses, and no photons are explicitly generated. There are important differences between bremsstrahlung of positrons and electrons at low energies. They can be taken into account using a correction factor which is a function of T /Z 2 (Kim et al., 1986). For example, the total electron bremsstrahlung cross-section for production of photons above 10 keV in lead is greater by almost 50% than for positrons, if the electron and positron kinetic energies are 1 MeV (CERN, 1990). This correction is included in GEANT. 4.2.3. Positron annihilation The positron annihilation can take place with a free or loosely bound atomic electron, during which two photons are emitted, or with a strongly bound electron (mostly in the K-shell), during which only one photon is emitted, as the excessive momentum is taken up by the atom. The cross-sections of the two-photon and one-photon annihilations are given by the Heitler formulae and sampled in GEANT. One-photon annihilation, with electrons only in the K-shell is considered. More complicated annihilation schemes are not simulated as their contribution is negligible. The contribution of the one-photon annihilation is largest for heavy materials, as much as 20% for ≈400 keV positrons in lead, and it decreases for lower and higher positron energies. 4.2.4. Photoelectric effect The cross-sections for the photoelectric effect, as well as the K, LI and LII shell energies are parametrized in GEANT. The relative errors of the cross-sections are less than 25% near the shell-energy edges, and less than 10% elsewhere. Beginning with the GEANT version 3.16 also X-ray and Auger electron emissions are simulated. 4.2.5. Compton scattering The total cross-sections for Compton-scattering are parametrized introducing relative errors not exceeding 6% in the photon energy region from 20 keV up to 100 GeV. However, the Klein–Nishina formula is used for the differential cross-sections, i.e. the Compton-scattering is treated as taking place on free electrons. Neglecting the binding energies and momenta of atomic electrons has crucial influence on the resulting electron energy spectrum near the Compton edge. The Compton edge is blurred in reality and not so sharp as calculated using the
182
P.P. Povinec et al.
Klein–Nishina formula. Therefore a response function of a Ge detector can never be computed with sufficient accuracy around the Compton edge using the present versions of GEANT. 4.2.6. Electron–positron pairs and Rayleigh scattering The cross-sections of the electron–positron pair production are parametrized in GEANT with relative errors less than 5%. Relative errors of the parametrization of the cross-sections for the Rayleigh scattering (coherent scattering on atoms without changing the photon energy, but changing the photon direction) are less than 10% for photon energies below 1 MeV. 4.3. Processes with hadrons GEANT is also capable to simulate hadron transport, hadronic interactions and hadronic showers. It utilizes translations of hadron transport codes GEISHA (Fesefeldt, 1985) and FLUKA (Aarnio et al., 1990), which can be operationally switched-on or switched-off by the user. However, these codes are mostly intended for hadronic interactions at high energies, and they do not simulate processes important for the background induction properly. Fortunately, except for the negative muon capture, processes with muons which leads to hadronic showers (fast muon nuclear interactions, nuclear photoeffect with high energy photons, and photo-fission) are rare and unimportant for the total resulting background. In particular, neutrons are tracked down to 10 keV, and then simply absorbed without generation of particles emitted during the de-excitation of a compound nucleus. Therefore, even if an additional routine would be written for the negative muon capture in lead including neutron and photon emissions, the resulting neutrons could not be transported correctly with the plain GEANT. Other neutron transport codes, better suited for these processes must be employed. One of such codes is MCNP (LANL, 1986), frequently used for calculation of production rates of radionuclides, however, simulations of the background induced by neutrons would need some improvements.
5. Monte Carlo simulation codes 5.1. Surface simulations As was already mentioned the GEANT code system developed at CERN for high energy physics simulations was adopted for the simulation of the cosmic muon induced background. The code is too complicated to be described in detail in this chapter, the reader should see the GEANT documents for more information on the code system and its environment (CERN, 1990). We shall describe first a simulation of background of HPGe detectors placed in a lead shield at the sea level. Later we shall go underground, and we shall also include an antiCompton shielding. The GEANT code has already been validated and results compared for HPGe spectrometers operated in surface laboratories (Vojtyla, 1995, 1996; Vojtyla et al., 1994; Vojtyla and Povinec, 2000). For a simulation of irradiation of a shield with detector in free space (undisturbed by overburden materials), we can consider the muon flux to be homogeneous, though not isotropic. This means that the function describing the differential muon flux j (p, ϑ, ϕ) is constant in
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
183
every point of space. The volume of a shield is convex in most cases, therefore every muon, which hits the shield, enters its volume only once. We neglect improbable events when muons are scattered back to the shield, for example, from the floor. The time T corresponding to the number of N simulated events is calculated using a simple scaling equation N , dN/dt where dN/dt is the number of muons that hit the shield in a time t. The number of muons that pass through a unit area during a unit time, and their angle and momentum distributions depend on whether the area is horizontal or vertical. In general, the number of muons that hit a unit area with a normal vector σ in unit time is given as " " dN j (p, ϑ, ϕ) σω dΩ, = dp dS dt Ω T =
where ω = (cos ϕ sin ϑ, sin ϕ sin σ, cos σ ) is a unit vector in the direction of the space angle dΩ. It should be integrated throughout all space from which muons hit the area from outside. For horizontal areas we have " " 2π " π/2 dN = dp dϕ dϑ j (p, ϑ, ϕ) sin ϑ cos ϑ (1) dS dt h 0 0 because σ = (0, 0, 1) (sin ϑ stems from dΩ). For vertical areas " " Φ+π/2 " π/2 dN = dp dϕ dϑ j (p, ϑ, ϕ) sin2 ϑ cos(Φ − ϕ) dS dt ν Φ−π/2 0
(2)
as σ = (cos Φ, sin Φ, 0), where Φ is the azimuth angle of the vector normal to the area dS. The muon impact rate can be calculated integrating Equations (1) and (2) over the whole shield surface except the basis (there is no muon flux from the bottom). The definition of angles and vectors is illustrated in Figure 4. For a simple model of the muon flux, not depending on the azimuth angle ϕ, the integrals in Equations (1) and (2) equal to " 2π dN = j (0, p) dp dS dt h n(p) + 2 and
dN dS dt
" =
2I (n)j (0, p) dp,
ν
where I (n) is an integral of the type " 1 I (n) = x n 1 − x 2 dx. 0
The total muon impact rate at a rectangular shield with a height H and a basis A × B is given as " " dN 2 2π = πr j (0, p) dp + 4πH RI (n)j (0, p) dp. (3) dt n+2
184
P.P. Povinec et al.
Fig. 4. Definition of angles and vectors for muon flux calculations.
The total muon impact rate a rectangular shield with a height H and a basis A × B is given as " " dN 2π (4) = AB j (0, p) + 4H (A+)I (n)j (0, p) dp. dt n+2 The first integral represents muons falling on the horizontal area from the top, while the second one corresponds to muons hitting the shield from sides. As the angular and momentum distributions are different for vertical and horizontal areas, each case is treated separately. The probability that a muon hits a horizontal or a vertical area is proportional to the first and the second integral in Equation (3) or (4), respectively. The probability that the absolute value of the muon momentum will be p is proportional to the function standing in the respective integral in Equation (3) or (4). The probabilities of hitting horizontal and vertical areas, as well as probabilities of all considered muon momentum absolute values are calculated during the initialization phase of the simulation code. The range of muon momenta 0.2–100 GeV/c (divided into 999 bins with 0.1 GeV/c) was found sufficient for background simulations. The muon impact rates both from the top and from the sides, and their branching ratios were calculated as well. Figure 5 shows a typical picture of a muon passing and interacting with the shield and the Ge detector. The defined differential muon flux for the horizontal area " " Jh = dp j (p, ϑ, ϕ) cos ϑ dΩ and the omnidirectional muon flux " " J0 = dp j (p, ϑ, ϕ) dΩ were found to be in agreement with experimental results (Allkofer and Grieder, 1984).
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
185
Fig. 5. Interaction of 50 GeV/c muon with a Ge-detector.
5.2. Underground simulations The muon generator in codes for simulating muon-induced background under the overburden is similar to that at sea level, but muon rejection and momentum reduction algorithms are applied. After calculating the kinematics of the muon, the range of the muon in rock is calculated and compared to the slanted depth of the counting system X = D/ cos ϑ, where D is the vertical depth and ϑ is the muon zenith angle. If the range is shorter than the slanted depth, the event is rejected. Otherwise, the original muon momentum p0 is reduced according to the formula pred = p0 − R −1 (X), where R −1 (X) is the inverse function of a function describing the range of muons in dependence on the muon momentum R(p). It is assumed that the rock has a standard rock composition with a density if 2.65 g cm−3 , the average mass number A = 22 and the average atomic number Z = 11 (Heusser, 1995). The transport of muons through the standard rock was simulated using GEANT and data on mean ranges for various muon momenta are presented in Figure 6. 5.3. Anti-Compton shielding An anti-Compton shielding made of NaI(Tl) annulus of 11 cm in the inner diameter, 30 cm of the outer diameter and the height of 40 cm was considered in calculations. An NaI(Tl)
186
P.P. Povinec et al.
Fig. 6. Mean range of muons in standard rock calculated using the GEANT code.
Fig. 7. Anti-Compton arrangement of HPGe and NaI(Tl) detectors.
stopcock of 7.6 cm in diameter and 7.6 cm long closed the annulus from one side. An n-type coaxial HPGe detector of 100% relative efficiency was inserted inside the NaI(Tl) annulus. The whole detector set-up was placed inside a shield made of 10 cm thick lead, dimensions of 70 × 70 cm, and length of 90 cm. There are two possibilities how to position the detectors in the shield—horizontally or vertically. A draw of the horizontal installation is shown in Figure 7. Simulations were carried out separately for both installations. The code for the horizontal installation has a more complicated muon generator than the standard codes described earlier. Because the basis of the box is not a square, bombardment of sides of different areas has to be correctly sampled so that impact rates are proportional to the respective areas.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
187
Fig. 8. A comparison of measured (top) and simulated (bottom) background gamma-ray spectra of a HPGe detector.
6. Simulation results and discussion 6.1. Single gamma-ray spectrometer The developed code was tested for an existing set-up so the obtained results can be compared with measurements. Figure 8 shows a comparison of the experimental background gammaray spectrum with the calculated spectrum. The agreement of the background continua is good throughout all the energy and count rate ranges, except for a low energy part below 100 keV, where the simulated counting rates are lower by 10–20%. The neutron-induced saw-tooth peaks and the primordial peaks are not simulated. The simulation overestimates the count rate in the annihilation peak by 38% which can be partly explained by complete charge collection considered in the simulation which overestimates peak count rates. To introduce some systematics into the large number of possible set-ups we used the following nomenclature to denote a particular set-up: • Detector type—1 = coaxial detector with 100% relative efficiency; 2 = well detector with 150% relative efficiency. • Shield—DZ = 15 cm Pb + 1 mm Cd + 2 mm Cu (descending Z lining); PBO = 15 cm Pb only; PC05 = 15 cm Pb + 5 mm Cu; PC1O = 15 cm Pb + 10 mm Cu. • Size—S = small, diameter 42 cm, height 51 cm; M = medium, diameter 60 cm, height 70 cm; L = large, diameter 100 cm, height 120 cm. • Shape—C = cylindrical; R = rectangular. • Overburden—sea level; U = 30 m w.e. underground. For example, 2DZMR means the detector 2 placed in a rectangular descending-Z shield of medium size at sea level. Three sizes of shields were used: the small (S) one is close to the detector and only a minimum of space is available inside; the medium (M) size shield is intended for 1 L samples; and the large (L) shield can take even larger samples. A thickness of 15 cm of lead has been found to be the optimum shielding thickness for large volume HPGe detectors situated at sea level or at shallow depths underground. If thicker shielding was used in simulations, the background was higher due to interactions of muons with the shield.
188
P.P. Povinec et al.
Fig. 9. Background gamma-ray spectra for the HPGe well detector of 150% relative efficiency in various shields and shield linings.
The resulting background spectra for detector 2 in the energy region 0–1500 keV are shown in Figure 9 for the lead shield without any lining, with the descending-Z lining and with 5 mm of copper. It can be seen by analyzing these results that the background levels depend
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
189
mainly on dimensions of the shield, and on the shield lining. The observation that the smallest background is obtained in the smallest shield has been a surprised as in previous lowlevel counting studies opposite recommendations could be found (e.g., Watt and Ramsden, 1964). It seems that an equilibration of radiation takes place for very large shields. If the detector dimensions are small compared to the inner dimensions of a shield, a larger inner surface area radiating secondary particles generated by cosmic muons is compensated by a smaller place angle. The small background obtained for shields close to the detector (S-size) can be partly explained by a kind of self-anticoincidence. The cosmic muons inducing the highest probable background must pass through material layers close to the inner surface of the shield. If the shield is tight, these muons also pass close to the sensitive volume of the Ge crystal and therefore they are likely to hit it. However, if this happens, tens of MeV of energy is deposited in the sensitive volume, rejecting the given event from the lower-energy part of the background spectrum. The background spectra of small shields are relatively depleted in the region below 500 keV and the effect is more pronounced for the annihilation peak. This implies that this effect indeed takes place. However, the special simulation with direct deposition by muons in a Ge detector switched-off showed that it accounts for only about 30–40% of the total effect. The rest is probably caused by a lesser amount of high-Z material radiating secondary particles in small shields. It is interesting that the background depends only slightly on the shape of the shield if the inner shield dimensions are preserved. Rectangular shields provide only a few percent greater backgrounds than cylindrical ones as documented by the comparison of the corresponding spectra (Figure 9). The second important observation is that thick layers of low-Z lining increase the background, as can be seen in Figure 10 where integral count rates for detectors 1 and 2 are presented for different set-ups. The effect is most pronounced around 200 keV. For example, the maximum in the continuum for the shield 1PC10LR (lining with 1.0 cm of Cu) is 2.2 times greater than that of the corresponding lead-only shield 1PBOLR. This effect was also observed experimentally (Heusser, 1993) and explained as a consequence of smaller selfabsorption coefficients for lower-Z materials. On the other hand, shields with lower-Z lining have smaller count rates in the annihilation peaks and above about 500 keV. However, the decrease is not important for this size of HPGe detector. The smallest backgrounds are obtained with the smallest shields without lining, i.e. only with lead walls. The shield dimensions should be kept as small as possible depending on the largest sample size intended for analysis. For example, it is not worth leaving an extra space in the shield of the well-type detector if samples are only analyzed inside the detector well. If the lead X-rays are disturbing, descending-Z lining is superior. It is advisable to design the shield so that the lining can be easily removed. Thick lower-Z lining should be completely avoided. The spectra above about 20 MeV are almost identical independently of the shield because this is the region where direct ionization by cosmic muons prevails. The dominant peak obtained for 61 MeV (Figure 11) corresponds to the track length of muon in the Ge crystal (the ionization loses of muons in Ge are 7.3 MeV/cm). The presence of the overburden of 30 m w.e. reduces the background spectra by a scaling factor of 4. Larger reductions seem to occur for shields with thick lower-Z linings.
190
P.P. Povinec et al.
Fig. 10. Integral background counting rates for the HPGe detector of 100% (top) and 150% (bottom) relative efficiency in various set-ups.
Fig. 11. Background gamma-ray spectrum of HPGe detector (100% relative efficiency) in higher energy region.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
191
Fig. 12. Lower-energy background spectra of the anti-Compton gamma-ray spectrometer in the horizontal and vertical positions at sea level (top) and at 30 m w.e. underground (bottom).
6.2. Anti-Compton spectrometer Obviously, the anti-Compton shield reduces not only the Compton-continuum but also the cosmic muon induced background. The reduction is very efficient due to the high detection efficiency of NaI(Tl) crystals for secondary cosmic-ray particles and gamma-rays. The results for the vertical and horizontal anti-Compton shield arrangements at sea level and underground are summarized in Figure 12. The simulated background spectra at 30 m w.e. underground show the same features as those at sea level, but the background is reduced by a factor of ∼4 that was observed also in the passive set-ups. A striking feature of the background spectra is that the difference between the horizontal and vertical arrangements is small in the lower energy region, however, it is visible. Therefore, below the energies of 1000 keV the vertical arrangement may be slightly better than the horizontal one (Vojtyla and Povinec, 2006). The spectra in higher energy regions (presented in Figure 13) differ significantly in the region above ∼10 MeV, where the contribution of the secondary processes in the shields is negligible, and the direct ionization by muons prevails.
192
P.P. Povinec et al.
Fig. 13. Higher-energy background gamma-ray spectra of the anti-Compton spectrometer in the horizontal and vertical positions at the sea level (top) and at 30 m w.e. underground (bottom).
The background reduction factor by the anti-Compton rejection depends on the energy in the spectrum. Figure 14 shows the background reduction factors calculated for various energies up to 3000 keV. It reaches values of about 200 for the lowest energies (50–100 keV) and decreases down to about 40 at 1500 keV. A cosmic-muon rejection factor of at least 40 (at around 1 MeV) can be reached when the anti-Compton rejection is operational. In such situation, the cosmic-muon background is reduced to such a level that other background components prevail, like those from the residual contamination of the detector and anti-Compton shielding materials, or from radon, especially for the underground facilities. However, such high cosmic-muon rejection factors can be reached only if construction materials with negligible radionuclide contamination have been used for the construction of the HPGe detector, its cryostat, NaI(Tl) detectors and the passive shielding. Figure 15 shows the energy deposition spectra in the annular and stopcock NaT(Tl) detectors placed at the surface or 30 m w.e. for the horizontal and vertical positions. The spectra confirm high efficiency of the anticoincidence detectors. The peak observed in the spectrum represents the track length of muon in the NaI(Tl) crystal.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
193
Fig. 14. Dependence of the background reduction factors of the anti-Compton suppression on the energy for the horizontal and vertical positions at sea level.
Fig. 15. Energy deposition spectra in the annular and stopcock NaI(Tl) crystals for the horizontal and vertical positions in the surface (left two graphs) and the underground laboratory operating at 30 m w.e. (right two graphs).
7. Comparison of modeling and experimental results The most important recent development in the radiometric methods of analysis of radionuclides in the environment has been the availability of large volume HPGe detectors (about
194
P.P. Povinec et al.
200% relative efficiency compared to a 76×76 mm NaI(Tl) detector). However, because of the large volume and mass of an HPGe detector, interactions between cosmic rays and the detector affect considerably its background characteristics. An anticosmic shielding and/or underground operation of detectors has become therefore important for their optimum utilization for analysis of environmental radionuclides (Heusser, 1994; Povinec, 1994; Vojtyla et al., 1994; Heusser, 1995; Reyss et al., 1995; Niese et al., 1998; Neder et al., 2000; Neumaier et al., 2000; Vojtyla and Povinec, 2000; Semkow et al., 2002; Povinec, 2004; Povinec et al., 2004; Schroettner et al., 2004). 7.1. IAEA-MEL’s underground laboratory The underground laboratory (Counting lAboratory for enVironmEntal radionuclides, CAVE), designed following the Monte Carlo simulation was constructed at the IAEA-MEL Monaco (Povinec et al., 2004, 2006; Povinec, 2005). In this chapter we present the main characteristics of HPGe gamma-ray spectrometers and compare them with modeling predictions. The CAVE laboratory is situated in an underground cellar in a car parking area at a depth of 35 m w.e. The laboratory is equipped with: (i) A common lead shield housing four large volume HPGe detectors (Figure 16). (ii) An anticosmic veto shielding made of plastic scintillation detectors surrounding the lead shield which protect the HPGe detectors against cosmic radiation (Figure 16). Such a novel design, supported by Monte Carlo simulations, when several HPGe detectors are shielded against cosmic rays by a common guard detection system has been used for the first time in low-level gamma-ray spectrometry. (iii) An anti-Compton gamma-ray spectrometer, comprising an n-type HPGe detector and NaI(Tl) anticoincidence shielding placed in a lead shield. 7.1.1. Gamma-ray spectrometers with anticosmic shielding HPGe detectors are placed in the lead shield made of two layers. The outer layer is made of an ordinary lead 7.5 cm thick. The second, the internal layer which is 5 cm thick, is made of very low activity lead, which was specially ordered for the underground laboratory (210 Pb activity is below 0.1 Bq kg−1 ). The construction features of the lead shielding with HPGe detectors are shown in Figure 16. The lead shield is surrounded on all sides and from the top by plastic scintillation sheets 7 cm thick, which are viewed by 5 cm diameter photomultipliers (Figure 16). The plastic scintillation detectors are connected with HPGe detectors in anticoincidence, so they work as a guard for HPGe detectors, eliminating effects of nucleonic component on their background, and partially also the effects of the hard component of cosmic rays that is composed mainly of muons. There are also secondary particles (mostly electrons, positrons and photons) produced by interaction of cosmic ray particles with the lead shield which affect the detector background. HPGe detectors (coaxial p-type) were specifically designed for low-level gamma-ray spectrometry in an underground laboratory. They are of U-type with preamplifiers housed outside of the lead shield, however, the FET is mounted on a Cu plate connected to the cooling finger. Only materials with minimum radionuclide contamination were used for the detector construction.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
195
Fig. 16. HPGe detectors in the lead shield with anticosmic veto detectors made of plastic scintillation detectors.
Four types of HPGe detectors are housed in the common lead shield surrounded by the anticosmic guard: (i) A 100% relative efficiency coaxial detector (CANBERRA) with cryostat made of electrolytic copper with carbon window. The detector resolution at 1.33 MeV 60 C line is 2.09 keV, the peak/Compton ratio is 80. (ii) A 150% relative efficiency well type detector (CANBERRA) with cryostat made of electrolytic copper (well dimensions—32.5 mm diameter, depth 73 mm), well walls made of thin copper. The detector resolution at 1.33 MeV 60 C line is 2.48 keV, the peak/Compton ratio is 51. (iii) A 170% relative efficiency coaxial detector (EURISYS) with cryostat and window made of pure aluminum. The detector resolution at 1.33 MeV 60 C line is 2.08 keV, the peak/Compton ratio is 101. (iv) A 200% relative efficiency well type detector (CANBERRA) with cryostat made of electrolytic copper (well dimensions—25 mm diameter, depth 60 mm), well walls made of thin copper. The detector resolution at 1.33 MeV 60 C line is 2.31 keV, the peak/Compton ratio is 120. ORTEC NIM modular electronics have been used for signal processing and data acquisition. The block schema of the electronics used for the HPGe detectors with anticosmic shield-
196
P.P. Povinec et al.
ing is presented in Figure 17. The signals after amplification are fed separately for timing and energy evaluation. Each HPGe detector is connected in anticoincidence with the plastic scintillation guard detector, therefore events caused by penetrating cosmic muons are only partially registered in gamma-ray spectra. Time amplitude converters have been used for time analysis of signals from HPGe detectors and the plastic scintillation detectors. Using a proper delay (1 µs) of pulses from a HPGe detector and a width (40 µs) of the gate signal in the multichannel analyzer it has been possible to reach high efficiency of the guard plastic scintillation detector with negligible losses in the counting rate of HPGe detectors. During all measurements radon is expelled from the detector chambers by the evaporation of nitrogen from the detector’s Dewar containers, thus keeping stable background during measurements (Sýkora et al., 1992). It should be stressed that this detection system is used only for analysis of low activity samples. 7.1.2. Coincidence–anticoincidence (anti-Compton) gamma-ray spectrometer The coincidence-anticoincidence gamma-ray spectrometer (Figure 18) has as the main gamma-ray detector an n-type HPGe detector (ORTEC) of 100% relative efficiency (the resolution at 1.33 MeV 60 C line is 2.35 keV, the peak/Compton ratio is 64). The detector arrangement is of U-type with a preamplifier situated outside of the lead shield, however the FET is mounted on Cu plate connected with cooling finger. The detector cryostat is made of electrolytic copper, the window is made of high purity aluminum. The HPGe detector is surrounded by NaI(Tl) shielding (annulus of 30 cm in diameter and 40 cm long; top NaI(Tl) detector is of 76 mm in diameter and 76 mm long). A double side Si beta-ray detector (ORTEC), located between the HPGe detector and the top NaI(Tl) detector has been used in beta– gamma coincidence studies as well. All detectors are housed in a shield of 10 cm thick, made of ordinary lead previously used for shielding of gas proportional detectors. The coincidence– anticoincidence gamma-ray spectrometer is a versatile detection system that can operate in several detection modes: (i) A single HPGe or NaI(Tl) gamma-ray spectrometer. (ii) An anti-Compton gamma-ray spectrometer when the HPGe detector is connected in anticoincidence with the NaI(Tl) detectors, thus suppressing cosmic ray events registered by the NaI(Tl) detectors, as well as Compton scattered gamma-quanta registered by both the HPGe and NaI(Tl) detectors. (iii) A double coincidence gamma-ray spectrometer—gamma–gamma coincidences between the HPGe detector and the top NaI(Tl) detector; the same with the NaI(Tl) annulus either in coincidence or anticoincidence (e.g., in the case of analysis of 60 Co) (Staníˇcek and Povinec, 1986; Povinec et al., 2006). (iv) A triple coincidence gamma-ray spectrometer—all three detectors in coincidence (e.g., in the case of analysis of positron emitters (22 Na and others) when the annihilation quanta are registered by the HPGe detector and the top NaI(Tl) detector, and the characteristic gamma-rays by the NaI(Tl) annulus) (e.g., Sýkora and Povinec, 1986). (v) A sum coincidence gamma-ray spectrometer—coincidence summing of cascade (or annihilation) gamma-rays, enabling to “clean” the gamma-ray spectrum by the registration of the full energy absorption peaks only. Because of higher efficiency and lower background, free of any interferences, this arrangement helps to reach best detection limits
197
Fig. 17. Block schema of electronics for HPGe detectors with anticosmic shielding.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
198
P.P. Povinec et al.
Fig. 18. Anti-Compton gamma-ray spectrometer with HPGe detector (100% relative efficiency) and NaI(Tl) annulus with copper lining.
for emitters with cascade or annihilation photons in large volume samples (Povinec, 1982; Sýkora and Povinec, 1990). (vi) A beta–gamma coincidence spectrometer—coincidences between the beta-ray (Si or a gas detector) and HPGe/NaI(Tl) detectors are registered. This spectrometer is suitable for analysis of small volume samples, which are mounted on both sides of the beta-ray detector (Hlinka et al., 1977). (vii) A beta–gamma–gamma coincidence spectrometer—coincidences between the Si detector, the HPGe detector and the top NaI(Tl) detector. This arrangement, because of extremely low background can reach best detection limits for beta–gamma (–gamma) emitters in small volume samples (Zvara et al., 1994). The electronics (ORTEC) used for the anti-Compton spectrometer and various coincidence– anticoincidence modes is very complex (Figure 19).
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
199
Fig. 19. Block schema of the coincidence–anticoincidence electronics.
7.2. Background characteristics of HPGe detectors with anticosmic shielding The basic characteristics of the HPGe detectors, including the total background are compared in Table 1. As the volumes of the detectors differ significantly, it is necessary to compare their background characteristics per kg of Ge. It can be seen that the lowest background has been reached with the 150% detector. On the contrary, the detector with “pure” aluminum cap has a relatively highest background due to its contamination by natural radionuclides. It is necessary therefore that construction materials should be carefully tested by the manufacturing companies for the presence of radionuclides before the detector construction, especially for the detector’s cryostat and its window. However, it has also been a surprise that three detectors with copper cryostats ordered as low background detectors for an underground laboratory from the same company had very different background characteristics. It looks like that the companies cannot guarantee similar detector background characteristics even when using similar construction materials; the final product may differ. It is interesting to notice that the total detector background per kg of Ge, in the energy window 40–2000 keV, is decreasing with increasing detector volume, i.e. from 290 h−1 kg−1
200
P.P. Povinec et al.
Table 1 Characteristics of HPGe detectors in the IAEA-MEL’s underground laboratory Detector
100%
150%
170%
200%
Type Cryostat End-cap material FWHM at 1.33 MeV (keV) Ge mass (kg) Background at 40–2000 keV (h−1 kg−1 ) Background at 40–2000 keV with anticosmic shielding (h−1 kg−1 ) Reduction factors with anticosmic shielding
Coaxial Cu C 2.09 2.15 290 40
Well Cu Cu 2.48 2.50 239 27
Coaxial Al Al 2.08 3.29 230 59
Well Cu Cu 2.31 4.18 205 33
7.2
8.9
3.9
6.2
Fig. 20. Background gamma-ray spectrum of 200% relative efficiency HPGe detector operating without (top) and with anticosmic shielding (bottom).
Ge for the 100% efficiency detector to 205 h−1 kg−1 Ge for the 200% efficiency detector (background spectra of this detector without and with the veto anticosmic shielding are shown in Figure 20). However, the background with the anticosmic shielding does not follow this rule, but clearly shows a larger contribution of radioactive contamination of the detector construction materials to its background. The lowest total background with the anticosmic shielding was obtained for the 150% efficiency detector (27 h−1 kg−1 Ge), the highest for the 170% efficiency detector (59 h−1 kg−1 Ge). The 170% efficiency detector have the cryostat made of “pure” aluminum,
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
201
however, this has been clearly contaminated by U and Th (and their decay products), as well as by 40 K. Generally aluminum is from the point of view of radioactive contamination not suitable construction material. When we compare the reduction factors of the anticosmic shielding for different detectors (Table 1) we see that the highest background reduction was obtained for the 150% efficiency detector (9), and the lowest for the 170% efficiency detector (4). The 100% and 200% efficiency detectors have the reduction factors within these limits (7 and 6, respectively). The background reduction factors are considerably lower than we would expect from Monte Carlo simulations (Figure 14). We know from gas counting systems (Povinec, 1972, 1980; Povinec et al., 1979) that an anticosmic shielding may decrease the proportional detector background even by a factor of 100, in agreement with Monte Carlo simulations. The worse background characteristics of HPGe detectors may be due to several reasons: (i) Higher efficiencies of HPGe detectors to scattered gamma-quanta which is both of cosmic ray origin and of radioactive contamination of detector construction materials. (ii) Leakage of muons through the anticosmic shielding, as the shielding geometry is far from an optimum 4π geometry. (iii) Cosmic ray secondaries produced by muons passing the lead shield which were not discriminated by the anticoincidence circuit, as it activation time is usually too short (40 µs in our case) compared with the development time of a cosmic ray shower in the lead shield (200 µs). However, because of dead time problems of HPGe spectrometers with anticosmic shielding, it is usually difficult to increase the width of the gate signal in the anticoincidence circuit above 50 µs. We know that hard cosmic ray particles will produce more secondaries in thick lead shield, therefore the lead thickness should be optimized for a given depth below the earth surface (150 mm for surface laboratories). Too thick lead shield may be therefore contraproductive for obtaining good background characteristics. As for the operation of a semiconductor detector the diode material, in our case Ge, must be very clean to keep loses of charge collection on the detector electrodes as low as possible, the most probably the detector contamination is caused by a cryostat material, which is usually (and preferably) electrolytic copper or aluminum. The window may also be important, if a combination of copper, aluminum, beryllium or carbon materials are used. If the threshold energy need not be very low, it looks like that the best material is electrolytic copper. If a low energy window is required, a carbon fiber window may be the best choice. There are other construction materials that could affect the detector background. Preamplifier is usually situated outside of the lead shield, however the FET transistor is connected directly with the diode, as well as a copper cooling finger. These parts, together with soldering contacts, may be therefore crucial for obtaining a low detector background. The annihilation peak at 511 keV is still dominant in all background spectra. This peak is produced by annihilation of electrons and positrons in the shield and in the detector, which are products of the interaction of secondary cosmic rays with materials surrounding the detector. The second important finding is a large background continuum, which has a maximum at around 200 keV. This background continuum is produced by the interactions of cosmic muons with the lead shield, as we already discussed. Although the measuring time of the spectrum presented in Figure 20 was not long enough to see well the peaks of natural and cosmic ray
202
P.P. Povinec et al.
Fig. 21. Background of HPGe detectors (from 40 to 2700 keV) as a function of the operating depth of CELLAR underground laboratories. The solid line shows the muon fluence rate in arbitrary units normalized to the background counting rate above ground. All detectors have only passive shielding, except the IAEA-MEL detector (after Laubenstein et al., 2004).
activated radionuclides, some peaks can be identified in the background spectra, e.g., 40 K (1.46 MeV) and 214 Bi (609.3 keV). When comparing the integral backgrounds of HPGe detectors divided by the mass of the Ge crystal, operating in different underground laboratories (Laubenstein et al., 2004; Povinec et al., 2004), it can be seen (Figures 21, 22) that the CAVE underground laboratory operating at 35 m w.e. with anticosmic shielding has the background comparable with underground laboratories operating with a passive shielding only at 250 m w.e. depth (e.g., Pagava et al., 1992; Niese et al., 1998; Hult et al., 2000; Neumaier et al., 2000; Komura and Hamajima, 2004). Therefore, an anticosmic shielding in an underground laboratory operating at a shallow depth is extremely important for reducing the detector background and should be widely used. A further decrease in the detector background can be obtained only in laboratories operating deep underground. However, it looks like that for the present state of the art of HPGe low background detectors the optimum depth would be around 1000 m w.e. Although in deeper laboratories the cosmic ray muon flux is much weaker, it does not improve anymore background characteristics of HPGe detectors (Figure 21). We believe that a further reduction in the detector background would be possible only with a new generation of HPGe detectors, specially designed (and produced!) for deep underground laboratories. Special arrangements must be also made how to decrease a radon contribution to the detector background, especially if frequent changes of samples in the detection system are required (Heusser et al., 2006).
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
203
Fig. 22. Comparison of underground background spectra of HPGe detectors of the CELLAR laboratories. All detectors have only passive shielding, except the IAEA-MEL detector (after Laubenstein et al., 2004).
Fig. 23. Background spectrum of the 100% HPGe detector with Compton suppression (counting time 60,000 s).
7.3. Background characteristics of the anti-Compton gamma-ray spectrometer Figure 23 shows the Compton suppression of the background of the 100% HPGe detector. The background in the energy interval 30–2500 keV has decreased from 0.5 to 0.1 s−1 with anti-
204
P.P. Povinec et al.
Compton shielding. However, when comparing the background spectra of the HPGe detectors with anticosmic shielding (Figure 20) with the background spectrum of the anti-Compton spectrometer (Figure 23) it can be noticed that the 100% HPGe detector has in the background spectrum several gamma-lines from natural radionuclides, probably due to a contamination of the ordinary lead shield (an internal low activity lead shielding has not been installed yet). It may be a surprise that the background of the HPGe detector with the anti-Compton shielding is bigger than that of single detectors shielded by plastic guard detectors. The shielding geometry of the anti-Compton spectrometer is usually much better than that of a detector with anticosmic shielding only. However, radioactive contamination of construction materials play a crucial role in a low-level anti-Compton spectrometer. As HPGe detectors have similar background characteristics, the radioactive contamination may originate in the NaI(Tl) annulus and stopcock detectors, and in photomultipliers. The dominant contaminant in this case is 40 K, which is very difficult to remove from the NaI(Tl) detectors. Another possibility would be to use a BGO scintillation detector, which because of higher detection efficiency could have smaller dimensions. However, the radioactive contamination in this case is even worse due to presence of radioactive bismuth and lead in the material. Because of contamination problems anti-Compton gamma-ray spectrometers will not reach a lowest possible background. They are, however, useful detection systems when a nondestructive analysis of samples containing a wide range of gamma-ray emitters should be carried out, e.g., analysis of 137 Cs (662 keV) in sediments where Compton scattered gammarays from 40 K (1.46 MeV) have negative impact on detection limits. A typical gamma-ray spectrum of marine fish sample measured without and with the Compton suppression is presented in Figure 24.
8. Conclusions There are techniques available at present capable to model background characteristics of HPGe detectors and thus to optimize the design of low-level gamma-ray spectrometric installations. The Monte Carlo simulation tools, such as GEANT developed at CERN, can deal with the simulation of particle passage through the detectors and shields. A computing power is not anymore problem, as desk top computers of sufficient speed and power are available. Therefore the influence of various parameters on the detector background can be studied well in advance, and the cosmic muon induced background can be estimated before a low-level detector system is constructed. Simulations of characteristics of HPGe detectors placed inside a lead shield without and with anti-Compton shield in surface or shallow underground laboratories revealed useful trends in design of such systems. Given the detector set-up, its background spectra induced by cosmic ray muons can be scaled down by a factor corresponding to the shielding depth. Generally, a radioactive contamination of construction parts of the shield and HPGe detectors itself is still dominating factor as the obtained background was always higher than predicted by Monte Carlo simulations. No important differences were observed when the anti-Compton system was positioned horizontally or vertically. An NaI(Tl) anti-Compton gamma-ray spectrometer inside a passive shield has a detrimental effect on the cosmic muon induced background if the anti-Compton
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
205
Fig. 24. Compton suppression (bottom) of the gamma-ray spectrum (IAEA-414 reference material—Irish Sea and North Sea fish).
rejection is switched off because of a thick layer of a relatively low-Z material, compared with standard shielding materials like lead. On the other hand, a cosmic-muon rejection factor of at least 40 (at around 1 MeV) is predicted when the anti-Compton rejection is operational. In such situation, the cosmic-muon background is reduced to such a level that other background components should prevail, like those from the residual contamination of the detector and anti-Compton shield materials, or from radon, especially in underground facilities. However, such high cosmic-muon rejection factors can be reached only if construction materials with negligible radionuclide contamination have been used for the construction of the HPGe detector, its cryostat, the surrounding NaI(Tl) detectors and the passive shield, and the detector operates in a place with low radon concentration. The CAVE underground counting laboratory with anticosmic shielding of detectors has permitted to effectively decrease the background of large volume HPGe detectors, especially when operating in anticosmic, anti-Compton or coincidence modes. The anticosmic shielding has proved to be a valuable investment as the background of HPGe detectors operating at shallow depths is comparable with underground laboratories operating at moderate depths of about 250 m w.e. For example, in the case of analysis of 137 Cs in seawater samples it has been possible to decrease the sample volume by about a factor of 10, which greatly reduces sampling time, and the same seawater volumes could be used for gamma-ray spectrometry as well as for mass spectrometry measurements. It is also advantageous to analyze sediment and biota samples in the underground laboratory as smaller volumes can be measured (especially
206
P.P. Povinec et al.
if well-type detectors are used), and in the case of sediments, high resolution core profiles can be studied (Povinec, 2004).
Acknowledgements The authors acknowledge assistance of Dr. I. Levy-Palomo during measurements in the IAEAMEL underground laboratory. The IAEA is grateful for the support provided to its Marine Environment Laboratories by the Government of the Principality of Monaco.
References Aarnio, P.A., et al. (1990). Fluka User’s Guide. Technical Report TIS-RP-190, CERN, Geneva. Allkofer, O.C. (1975). Introduction to Cosmic Radiation. Thieming, Munich. Allkofer, O.C., Grieder, P.K.F. (1984). Cosmic Rays on Earth. Physics Data Report No. 25-1, University of Kiel. Allkofer, O.C., Karstensen, K., Dau, W.D. (1971). The absolute cosmic ray muon spectrum at sea level. Phys. Lett. B 36, 425–427. ˇ Beláˇn, T., Chudý, M., Durana, L., Grgula, M., Holý, K., Levaiova, D., Povinec, P., Richtáriková, M., Šivo, A. (1992). Investigation of radionuclide variations in the Bratislava air. In: Povinec, P. (Ed.), Rare Nuclear Processes. World Scientific, Singapore, pp. 345–366. Brodzinski, R.L. (1991). Low-level gamma-ray spectrometry. J. Phys. G Nucl. Part. Phys. 17, S403–S414. Brodzinski, R.L., Reeves, J.H., Avignone, F.T., Miley, M.S. (1988). J. Radioanal. Nucl. Chem. 124, 513. Brodzinski, R.L., Reeves, J.H., Avignone, F.T., Miley, M.S. (1990). Further reduction of radioactive backgrounds in ultra sensitive germanium spectrometers. Nucl. Instrum. Methods A 292, 337–342. CERN (1985). Yellow Report 85-03. CERN, Geneva. CERN (1990). GEANT Detector Description and Simulation Tool. CERN Program Library Office, CERN, Geneva. CERN (1992). Particle Data Group. CERN, Geneva. Cooper, J.A., Perkins, R.W. (1971). An anticoincidence shielded dual Ge(Li) gamma-ray spectrometer for environmental and biological samples. Nucl. Instrum. Methods 99, 125–146. Debertin, K., Helmer, R.G. (1998). Gamma and X-Ray Spectrometry with Semiconductor Detectors. Elsevier, Amsterdam. Eisenbud, M. (1973). Environmental Radioactivity. Academic Press, New York. Fesefeldt, H.C. (1985). Simulations of Hadronic Showers, Physics and Applications. Technical Report PITHA 85-02, RWTH, Aachen. Gaiser, T. (1990). Cosmic Rays and Particle Physics. Cambridge Univ. Press, Cambridge. Heusser, G. (1991). Studies of γ -ray background with a low level germanium spectrometer. Nucl. Instrum. Methods B 58, 79–84. Heusser, G. (1992). Progress in Ge-spectrometry at the Max-Planck-Institut für Kernphysik Heidelber. In: Povinec, P. (Ed.), Rare Nuclear Processes. World Scientific, Singapore, pp. 247–255. Heusser, G. (1993). Cosmic ray-induced background in Ge-spectrometry. Nucl. Instrum. Methods B 83, 223–228. Heusser, G. (1994). Background in ionising radiation detection. In: Garcia-Leon, M., Garcia-Tenorio, R. (Eds.), Low-Level Measurements of Radioactivity in the Environment. World Scientific, Singapore, pp. 69–112. Heusser, G. (1995). Low-radioactivity background techniques. Ann. Rev. Nucl. Part. Sci. 45, 543–590. Heusser, G., Laubenstein, M., Neder, H. (2006). Low-level germanium gamma-ray spectrometry at the µBq/kg level and future developments towards higher sensitivity. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 495–510. Hlinka, V., Usacev, S., Povinec, P., Chudý, M. (1977). Die Untersuchung der Koinzidenzmethoden bei der Identifikation einiger Radioisotopen. Acta Univ. Comen. Phys. 18, 109–116. Hult, M., Martinez Canet, M.J., Koehler, M., Das Neves, J., Johnston, P.N. (2000). Recent developments in ultra low-level γ -ray spectrometry at IRMM. Appl. Radiat. Isotopes 53, 225–230.
Monte Carlo simulation of background characteristics of gamma-ray spectrometers
207
Kim, L., Pratt, R.H., Seltzer, S.M., Berger, M.J. (1986). Phys. Rev. A 33, 3002. Kolb, W.A. (1988). Background reduction of a semi-planar Ge(HP) detector. Environ. Int. 14, 367–370. Komura, K., Hamajima, Y. (2004). Y. Ogoya underground laboratory for the measurement of extremely low levels of environmental radioactivity: review of recent projects carried out at OUL. Appl. Radiat. Isotopes 61, 185–190. Kozlowski, T., Bertl, W., Povel, H.P., Sennhauser, U., Walter, H.K., Zglinski, A., Engfer, R., Grab, C.H., Hermes, E.A., Isaak, H.P., Van Der Schaaf, A., Van Der Pluym, J., Hesselink, W.H.A. (1985). Energy spectra and asymmetries of neutrons emitted after muon capture. Nucl. Phys. A 436, 717–732. Laubenstein, M., Hult, M., Gasparro, J., Arnold, D., Neumaier, S., Heusser, G., Kohler, M., Povinec, P.P., Reyss, J.-L., Schwaiger, M., Theodorsson, P. (2004). Underground measurements of radioactivity. Appl. Radiat. Isotopes 61, 167–172. Los Alamos National Laboratory, LANL (1986). MCNP Code. LANL, Los Alamos. Morita, M. (1973). Beta Decay and Muon Capture. Benjamin, Reading. Neder, H., Heusser, G., Laubenstein, M. (2000). Low-level gamma-ray germanium spectrometer to measure very low primordial radionuclide concentrations. Appl. Radiat. Isotopes 53, 191–195. Neumaier, S., Arnold, D., Böhm, J., Funck, E. (2000). The PTB underground laboratory for dosimetry and spectrometry. Appl. Radiat. Isotopes 53, 173–178. Niese, S., Köhler, M., Gleisberg, B. (1998). Low-level counting techniques in the underground laboratory Felsenkeller, Dresden. J. Radioanal. Nucl. Chem. 233, 167–172. Pagava, S., Burchuladze, A., Robakidze, T., Rusetski, L., Tsintsadze, D., Povinec, P., Chudý, M., Staníˇcek, J. (1992). Tbilisi underground laboratory. In: Povinec, P. (Ed.), Rare Nuclear Processes. World Scientific, Singapore, pp. 300–304. Povinec, P. (1972). Very low background proportional counter for tritium dating. Nucl. Instrum. Methods 101, 613– 614. Povinec, P. (1980). Proportional chambers for low level counting. Nucl. Instrum. Methods 176, 111–115. Povinec, P.P. (1982). Dual parameter gamma-ray spectrometer for low-level counting. Isotopenpraxis 18, 423–428. Povinec, P.P. (1994). Underground low-level counting. In: Garcia-Leon, M., Garcia-Tenorio, M. (Eds.), Low-Level Measurements of Radioactivity in the Environment. World Scientific, Singapore, pp. 113–140. Povinec, P.P. (2004). Developments in analytical technologies for marine radionuclide studies. In: Livingston, H.D. (Ed.), Marine Radioactivity. Elsevier, Amsterdam, pp. 237–294. Povinec, P.P. (2005). Ultra sensitive radionuclide spectrometry: Radiometrics and mass spectrometry synergy. J. Radioanal. Nucl. Chem. 261, 413–417. Povinec, P., Szarka, J., Usaˇcev, S. (1979). Multielement proportional counters for low-level 3 H and 14 C counting. Nucl. Instrum. Methods 163, 363–369. Povinec, P.P., Comanducci, J.F., Levy-Palomo, I. (2004). IAEA-MEL’s underground counting laboratory in Monaco—background characteristics of HPGe detectors with anti-cosmic shielding. Appl. Radiat. Isotopes 61, 85–93. Povinec, P.P., Comanducci, J.F., Levy-Palomo, I. (2006). IAEA-MEL’s underground counting laboratory—The design and main characteristics. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 538–553. Review of Particle Properties, RPP (1994). Phys. Rev. D 50, 1173. Reyss, J.-L., Schmidt, S., Legeleux, F., Bonté, P. (1995). Large, low background well-type detectors for measurements of environmental radioactivity. Nucl. Instrum. Methods A 357, 391–397. Schröder, W.U., Jahnke, U., Lindenberger, K.H., Röschert, G., Engfer, R., Walter, H.K. (1974). Z. Phys. 268, 57. Schroettner, T., Schwaigerm, M., Kindl, P. (2004). Optimization of an active cosmic veto shielding. Appl. Radiat. Isotopes 61, 133–138. Seltzer, S.M., Berger, M.J. (1985). Nucl. Instrum. Methods B 12, 95. Semkow, T.M., Parekh, P.P., Schwenker, C.D., Khan, A.J., Bari, A., Colaresi, J.F., Tench, O.K., David, G., Guryn, W. (2002). Low-background gamma spectrometry for environmental radioactivity. Appl. Radiat. Isotopes 57, 213–223. Smith, P.F., Levin, D.J. (1990). Dark matter detection. Phys. Rep. 187, 203–280. Staníˇcek, J., Povinec, P.P. (1986). Internal pair production in alpha-decaying nuclei and gamma-ray intensities of 241 Am. Nucl. Instrum. Methods B 17, 462–466. Sternheimer, R.M. (1971). Phys. Rev. B 3, 3681. Suzuki, T., Maesday, D.F., Roalsvik, J.P. (1987). Phys. Rev. C 35, 2212.
208
P.P. Povinec et al.
Sýkora, I., Povinec, P.P. (1986). Measurement of electron capture to positron emission ratios in light and medium nuclides. Nucl. Instrum. Methods B 17, 467–471. Sýkora, I., Povinec, P.P. (1990). Low-level gamma-ray spectrometry. Acta Phys. Univ. Comen. 31, 83–120. Sýkora, I., Durˇcík, M., Staníˇcek, J., Povinec, P.P. (1992). Radon problem in low-level gamma-ray spectrometry. In: Povinec, P.P. (Ed.), Rare Nuclear Processes. World Scientific, Singapore, pp. 321–326. Theodorsson, P. (1996). Measurement of Weak Radioactivity. World Scientific, Singapore, 333 pp. Vojtyla, P. (1995). A computer simulation of the cosmic-muon background induction in a Ge γ -spectrometer using GEANT. Nucl. Instrum. Methods B 100, 87–96. Vojtyla, P. (1996). Influence of shield parameters on cosmic-muon induced backgrounds of Ge γ -spectrometers. Nucl. Instrum. Methods B 111, 163–170. Vojtyla, P., Povinec, P.P. (2000). A Monte Carlo simulation of background characteristics of low-level HPGe detectors. Appl. Radiat. Isotopes 53, 185–190. Vojtyla, P., Povinec, P.P. (2006). Monte Carlo simulation of the muon-induced background of an anti-Compton gamma-ray spectrometer placed in a surface and underground laboratory. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 529–537. Vojtyla, P., Beer, J., Stavina, P. (1994). Experimental and simulated cosmic muon induced background of a Ge spectrometer equipped with a top side anticoincidence proportional chamber. Nucl. Instrum. Methods B 86, 380– 386. Watt, D.E., Ramsden, D. (1964). High Sensitivity Counting Techniques. Pergamon, Oxford. Zvara, I., Povinec, P.P., Sýkora, I. (1994). Determination of very low levels of radioactivity. Pure Appl. Chem. 66 (12), 2537–2586.
209
Underground laboratories for low-level radioactivity measurements Siegfried Niese∗ Nuclear Engineering and Analytics Inc., Dresden-Rossendorf, Germany
1. Introduction If we want to improve the detection limits in radioactivity measurements we need to increase the efficiency of the detector and to reduce its background. The sources of background are external radioactivity, the radioactivity of the detector itself, and cosmic rays, mainly muons and neutrons and their secondaries produced in the shield. Because of the fact that the range of high-energy nucleons and muons exceed the range of the radiation originating during radioactive decay of a nucleus by one and two orders of magnitude, respectively, we need for its reduction thick shielding. To reduce the contribution of cosmic rays to the detector background we have therefore to install the detector in an underground counting room. An improvement of the detection limit in an underground installation takes place in all cases when cosmic rays contribute significantly to the detector background located above ground. While the shielding against external radiation needs only a few meters of a dense material, the cosmic radiation has a low attenuation coefficient, and needs therefore thick layers of rock, or specially designed anticoincidence detectors. For this reason we first introduce in this chapter the components and properties of cosmic rays. We already mentioned that neutrons and muons are the most important for contributing to the background in low-level counting installations. The installation of the counting equipment can take place in a deep cellar, in an old mine, in a tunnel used for traffic, in a natural cave, in a cave formerly used for storage, or in a cave which is constructed especially for the counting equipment. There is a broad variation of depths, applications and type of construction of underground laboratories over the world. In this chapter we focus on medium deep underground laboratories designed for counting of samples. Some underground laboratories are located very deep, but they are mostly used for investigations of fundamental problems in nuclear and particle physics. This research needs an extremely low background as very rare events are studied. Special investigations have been carried out for selection of construction materials and shielding with very low radioactivity ∗ E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11006-8
© 2008 Elsevier B.V. All rights reserved.
210
S. Niese
contamination. These experiences have also been very useful for development of counting systems used for sample measurements. In 1898 Elster and Geitel (1898) from Wolfsburg made the first underground measurements at depths of 300 and 800 m in the German Harz Mountains. They compared the intensity of radiation from a radium source situated above ground and underground with the aim of finding out whether the source of energy inducing the radiation from radium is in the atmosphere or not. Fourteen years later, when it was well known that radioactivity represents a process of the decay of atoms, and rocks are radioactive because they contain potassium, thorium and uranium, Viktor Moritz Hess took a measuring instrument with him on a balloon flight, assuming that the intensity of radiation will be reduced with the height because of the growing distance from the earth’s surface. A surprisingly contrary result led him on August 7, 1912 during a balloon flight from Aussig (now Usti nad Labem, Czech Republic) to Pieskow, near Berlin, when he reached a height of 5350 m, to the discovery of cosmic rays (Hess, 1912; Smekal, 1968). In the following years intense studies of comic rays led to the discovery of new elementary particles. It was soon obvious that the main contribution to the background of gas filled counters comes from cosmic rays. To reduce their contribution, the counters were shielded not only by lead, but also by an active shielding made of anticoincidence counters. The cosmic rays penetrate thick layers of detector shielding material, react with the material and produce secondary radiation in the shielding material itself. Therefore the contribution of cosmic rays to the background became more and more interesting. After the background reduction with coincidence–anticoincidence techniques, several investigations were carried out in underground installations with reduced intensities of cosmic rays (Glower and Watt, 1957). It is usual to express the thickness of the shielding material (e.g., rock), which is traversed by cosmic rays in meters of water equivalent (m w.e.). The thickness of the atmosphere at sea level amounts to 10.2 m w.e. Tanaka et al. (1967) described the installation of the first underground laboratory at a depth of about 100 m w.e. for coincidence measurements of the positron emitter 26 Al in meteorites. Later in this laboratory a betagamma coincidence and an anticoincidence spectrometer were installed as well (Yamakoshi and Nogami, 1976). Later a systematic study of the background in a water dam was done by Kaye et al. (1972). In the following years a lot of background measurements were done in deep tunnels and mines and in some other locations of different type and depth followed by the construction of special underground laboratories for studies of elementary particles and the measurement of low radioactivity (Povinec, 1994). Oeschger and Loosli (1975) described the first measurements of samples for nuclear dating with proportional counters installed in a special underground laboratory. Later they built an underground laboratory at 70 m w.e. depth under their institute and compared results with those obtained within the St. Gotthard tunnel at 3000 m w.e. (Oeschger et al., 1981). The first application of an underground laboratory for the measurement of neutron-activated samples was done in our “Felsenkeller” underground laboratory in Dresden (Helbig et al., 1984). In the following years a number of underground laboratories for radioactivity measurements at moderate depth and for fundamental studies at greater depth were installed. At present we have a lot of experience in shielding of detectors against radiation and in the selection and purification of low-radioactive materials for the construction of low background detectors. The physics of active shielding against muons using anticoincidence detectors, and
Underground laboratories for low-level radioactivity measurements
211
of passive shielding with thick rock covering the underground laboratory is well understood. Very complicated is the production and shielding of neutrons and their reactions with the construction materials of the detectors. This is a reason for a more detailed description of this topic. 2. Cosmic rays and the influence of depth on the background 2.1. Cosmic rays in the atmosphere To understand the behavior of cosmic rays in rocks and inside underground laboratories it is necessary to understand their sources and processes of their interactions in the atmosphere. In cosmic ray physics, the concept of depth of the atmosphere is used instead of the height. The units used are either g/cm2 or m w.e.; 1 m w.e. = 100 g/cm2 . Sea level is then at a depth of 1020 g/cm2 , corresponding to 10.2 m w.e. The processes in the lower atmosphere near the earth’s surface are similar to the processes in the upper layers of the earth crust. They are described in more detail by Schopper et al. (1967), Theodorsson (1992), Heusser (1995) and Ziegler (1998). The primary galactic cosmic rays entering the atmosphere are nuclei of light atoms with very high energy, reaching 1020 eV. The cosmic ray flux is about 1000 particles m−2 s−1 with the following relative composition: Protons 86% Helium nuclei 12.7% Heavier nuclei 1.3% Their energy distribution per s, sr, and GeV above 10 GeV is described by the formula n(E) = 0.3/E 2.5±0.2 . In space, far from the earth, galactic cosmic rays are isotropic. Coming closer to the earth their path is influenced by the large dipole of the magnetic field of the earth, diverting less energetic particles back to space. At the equator a proton moving in a vertical direction must have an energy of at least 15 GeV to reach the earth. At middle latitudes the minimum energy is only 1–4 GeV, and at the magnetic poles the vertical fluxes of cosmic-ray particles are not influenced. Passing through the atmosphere, nuclear reactions with nuclei of the atmosphere are taking place, which are changing the composition of the cosmic rays. At sea level, outside of a laboratory, cosmic rays consist of the following secondary components, measured in particles m−2 s−1 (NCRP, 1987): Muons Neutrons Electrons Protons Pions
190 64 46 1.4 0.13
This shows a reduction of the total flux of protons by a factor of about 700. The reduction factor of the atmosphere for protons of 0.2 GeV amounts to 2 × 103 , and for protons of 10 GeV it is 5 × 103 .
212
S. Niese
At high altitudes outside of the atmosphere the count rate measured with a Geiger counter placed in a rocket is constant. At a height of about 50 km, where the primary cosmic-ray particles have traversed an atmospheric layer of 2 g/cm2 , the count rate begins to rise and reaches a maximum at an altitude of 13.7 km, when they have traversed about 150 g/cm2 . The count rate rise in the upper atmosphere is due to the production of secondary particles by collisions of the primary particles with nuclei of the atoms of the atmosphere. At the altitude of maximum count rate, production of secondary particles is equal to the loss of particles that have lost all their energy, mainly through nuclear collisions, but partly also through ionization. Below this region, the loss of particles is higher than the production of new ones, and the counting rate decreases rapidly. 2.2. Variation of the intensity of cosmic rays There is a variation in the cosmic-ray intensity with a period of 11 years by the magnetic modulation effect caused by solar plasma. This phenomenon is related to solar eruptions and is inversely correlated to the 11 year period of sunspots. A further decrease in galactic cosmicray intensity of up to 10% occurs after a solar flare. This may last a few days, but sometimes full recovery takes several weeks. Temporal variations of the intensity of cosmic rays in space and in the vicinity of the earth are caused by solar winds and by modulation of the atmospheric pressure. This will also be the result of seasonal temperature deviations. The influence of the geomagnetic latitude on cosmic rays is described by Ziegler (1998). 2.3. Nuclear reactions in the atmosphere Traversing the atmosphere the primary protons decrease both in number and energy. The mean interaction length liP of energetic protons in the atmosphere amounts to about 80 g/cm2 , and it is almost independent of its energy. On average, at each collision they lose about half of their energy. In the atmosphere the high energy primary particles cause violent explosionlike events when they collide with atomic nuclei, primarily of nitrogen and oxygen. This first phase is a collision with a single nucleon inside a nucleus. These reactions are similar for all nuclei and lead to multiple collision, both elastic and inelastic with pion production inside the nucleus. Protons, neutrons and pions with energies from 200 to 500 MeV are emitted mainly in the direction of the incident proton. After this violent nuclear cascade the remaining nucleus is left in a highly excited state and then additional protons and neutrons evaporate isotropically, having a broad energy spectrum with a maximum at about 1 MeV, and extending somewhat above 10 MeV. When the energy of the incident proton is below 1 GeV, the emitted particles are mainly protons and neutrons, but above 1 GeV the number of pions is usually larger. Energetic neutrons and pions can initiate the same kind of collisions as protons, and their interaction length is similar. After a collision, the energetic protons have a mean attenuation length la in the atmosphere of 120 g/cm2 , considerably longer than their interaction length of 80 g/cm2 . Because of the fact that the total depth in the atmosphere amounts to about 13 interaction lengths of protons, none of the primary protons will reach the earth’s surface without passing through a nuclear collision. The interaction length li , the mean length of a particle until the next interaction, is easier to understand than the attenuation length la , which describes the total process. It consists of
Underground laboratories for low-level radioactivity measurements
213
some processes yielding an enhancement as well as a reduction of the number of particles. For the description of processes with cosmic rays the attenuation length is more used than its reciprocal, the attenuation coefficient μa = 1/ la , which is preferred for the attenuation of the radiation emitted in a radioactive decay. The interaction length corresponds with a value calculated from the geometric cross-section σg of the reaction particles, σg = π(rT + rCR )2 , with the radii rCR of the cosmic-ray particle (CR), and rT of the target nucleus, which can be calculated as 1/3
rT = r0 AT , with the radius of a nucleon r0 = 1.37 × 10−13 cm and the number of nucleons in the target AT (Sitte, 1961; Webber, 1967). A charged pion decays to a muon and a neutrino π + → μ+ + ν
(2.6 × 10−8 s),
and a neutral pion decays into two gamma quanta γ and one neutrino π 0 → 2γ + ν
(1.8 × 10−16 s).
The charged pions interact strongly with nuclei and give rise to further nuclear transformations. In air at sea level, at the highest gas density, a pion with energy of 1 GeV traverses a mass thickness of 7 g/cm2 in its mean path length. This is only 8% of its interaction length, so the majority of pions will decay in the atmosphere without colliding with gas nuclei. Because of its short half-life a neutral pion π0 only travels in the order of micrometers before its decay. The created photons give rise to electromagnetic cascades which represent the main source of the soft component of cosmic rays in the atmosphere. 2.4. Production of neutrons In the atmosphere protons and pions are absorbed exponentially with depth with an attenuation length of about 120 g/cm2 . The protons are producing continuously secondary particles at a rate, which is nearly proportional to their flux. The pion–proton flux ratio increases slowly with depth and amounts to about 0.08 at sea level. Each primary nucleon produces a large number of neutrons with a broad energy spectrum, containing knock-on neutrons from the high-energy cascade, and evaporated neutrons. Their energy spectrum is expressed by P (E) dE = Ee−E/E0 dE, where the temperature of the excited nucleus E0 is about 1 MeV. The path of neutrons in the atmosphere is longer than that of protons because they lose no energy through ionization. The neutrons are slowed down by elastic and inelastic collisions with atmospheric nuclei, until they are finally captured. The root mean square value of their range, the attenuation length, is
214
S. Niese
about 150 g/cm2 . As the production and slowing down processes are the same at each depth, the shape of the neutron energy spectrum does not change below 200 g/cm2 , only its intensity. The energy spectrum at sea level extends from thermal energy to above 1 GeV. The small bump at 1 MeV is due to evaporation neutrons. Some more than 10% of neutrons have energy above 10 MeV. The energy spectrum of these neutrons has a high-energy bump at 80 MeV. Around 6.2 neutrons cm−2 s−1 are produced in the vertical column of the atmosphere, about 80% by evaporation and 20% by knock-on processes. Each primary proton produces on average 20 neutrons. After thermalization, 64% of the thermal neutrons are captured by nitrogen nuclei in a 14 N(n, p)14 C-reaction producing 14 C. Other processes capture 20% of the neutrons and 16% of neutrons leak out from the atmosphere. The ratio of the neutron/proton flux increases from about 22 at the upper atmosphere to 38 at sea level. Because of the fact that neutrons do not lose energy by ionization the neutron path length is longer and the flux of neutrons is therefore higher than that of protons. The measured values for the neutron flux (in particles m−2 s−1 ) at sea level show a considerable variation due to variations of primary protons: 40 (Cocconi and Tongiorgi, 1951) and 62 (Yamashita et al., 1966). 2.5. Muons Muons are elementary particles with the same charge as electrons, but with a 207-fold rest mass of that of the electron. They are decay products of charged pions. Their direction of propagation is nearly the same as that for their parent pions, and on average they receive nearly 80% of their parent’s energy. They are very penetrating as their nuclear interaction crosssection is only about 2×10−33 m2 . They lose energy practically only through electromagnetic interaction. The predominant processes are ionization and excitation. Knock-on collisions with electrons, and, to a lesser degree, bremsstrahlung and pair production take place too. The muon energy spectrum at the place of production is therefore nearly the same as that of the pions. However, passing through the atmosphere the spectrum is modified due to ionization losses, decay and capture of negative muons. Traveling through matter the muons lose energy like electrons. The loss amounts to about 2 MeV/g. In air a considerable number of muons decay to electrons, which receive about 1/3 of the kinetic energy of the muons while neutrinos carry away the rest of the energy. In this case the electrons and neutrinos share the muon rest mass energy of 106 MeV. The maximum energy of the electrons amounts to 50 MeV and their mean energy to about 40 MeV. In dense material most of the muons lose their energy through ionization before their decay. At a height of 10 km (a depth of 270 g/cm2 ) about 90% of muons produced in the atmosphere have already been formed. Although a large part of the muons decay in the atmosphere a substantial part will reach sea level. Because of the high penetration power of the muons, their flux decreases more slowly with atmospheric depth than that of its primary source, the protons. The stopped negative muons μ− are captured by atoms. After capture muonic X-rays are emitted. Because of the higher mass of the muon in comparison with the electron its orbits are closer to the nuclei of the atoms and the energies of the muonic X-rays are higher than electronic ones. A second process competes with the decay of muons: the capture by a proton in the nucleus, whereby a neutron n and a neutrino ν are formed,
Underground laboratories for low-level radioactivity measurements
215
μ− + p → n + ν. Decay is more probable in materials of low atomic number, but capture is dominating for nuclei with higher atomic number Z. For Z = 20, about 80% of the negative muons are captured and at Z = 50 this value amounts to 95%. The rest mass energy of a muon (106 MeV) is released and the nucleus is left in a highly excited state. It decays by emitting one or more neutrons. The number of neutrons emitted depends on the mass number of the capturing nucleus. In lead it amounts to 1.6 neutrons per muon. 2.6. Electrons (positrons) and photons Cosmic ray electrons with energy E0 lose energy mainly by the emission of photons by bremsstrahlung with a rather flat spectrum from zero to E0 . The photons receive on average about 40% of E0 . The mean length l0 traversed by an electron before radiating a photon is called the radiation length, measured in g/cm2 . It is practically independent of E0 , but approximately inversely proportional to the atomic number of the penetrated material. The most probable photon interaction process is electron–positron pair production. The average length traversed by the photon is nearly the same as the radiation length of the electron. The repeated combination of these two processes gives rise to electromagnetic showers or electron–positron–photon cascades which are called the soft component of the cosmic rays. Showers can also be initiated by energetic photons, emitted in the decay of neutral pions, which are dominating at middle and high altitudes or by an electron that has received a high energy from the decay of a muon. The photons and electrons in a shower are emitted mainly in the same direction as their parent particles and therefore travel inside a rather narrow cone. The production of a shower starts with an electron with the energy E0 . It loses half of its energy through a bremsstrahlung photon after traversing one radiation length. We assume that the photon after a second radiation length creates an electron–positron pair, each electron receiving half of the photon energy (E0 /4). After two radiation lengths we will have, in addition to the original electron, a photon, an electron and a positron, all with the energy E0 /4. This multiplying process is repeated until the critical energy Ec is reached where electrons lose more energy by ionization than radiation, and the photons lose more energy through Compton scattering than through pair production. The critical energy depends on the atomic number and it is for air, concrete, iron and lead 81, 50, 21 and 7.4 MeV, respectively. At the end of an electromagnetic shower in air, the mean energy of the electrons and photons amounts to 81 MeV. Than the electrons lose energy mainly by collisions, but still produce a few photons. They have a maximum range of about 55 g/cm2 . The photons gradually lose energy, mainly through Compton scattering, giving rise to a number of photons of decreasing energy. Their attenuation length in air is 55 g/cm2 (about 500 m at sea level) in the energy range of 10–80 MeV, but decreases to 20 g/cm2 at 2 MeV. A significant number of photons have a range far beyond that of the electrons. At sea level we have photons and electrons with energy of tens of MeV produced in showers initiated at low altitude. The low energy electrons are being slowed down and a number of Compton energy degraded photons are produced for each photon close to the critical energy. Electrons follow the same flux pattern as that of protons, but their flux decreases more slowly close to sea level. The ratio of electrons to protons is nearly constant, but at a depth of about 700 m it begins to rise. In the upper atmosphere, electrons are produced predominantly
216
S. Niese
Table 1 Mean concentrations of radioactive elements in different rock types Rock type
K, %
40 K, g/t
Th, g/t
U, g/t
226 Ra, Bq/kg
Upper crust, mean Ultra basite Basalt Granite Schist Sandstone Limestone
2.5 0.03 0.83 3.3 2.3 1.1 0.27
3.0 0.035 1.0 3.9 2.7 1.3 0.32
13 0.005 3 18 11 1.7 1.7
2.5 0.003 0.5 3.5 3.2 0.45 2.2
31 0.04 6 44 40 5.6 28
by energetic photons emitted in the decay of neutral pions. Below 700 m the ratio begins to rise, mainly through muon decay electrons, when the contribution of muons becomes significant, as their flux decreases much more slowly than that of protons.
3. Natural radioactivity 3.1. Radioactivity of rocks and building materials Radioactivity of rocks and building materials, which are produced from rocks, are the main sources of background of unshielded gamma-spectrometers situated above ground and underground. To take full advantage of the reduction of cosmic rays in an underground laboratory we have to reduce the contribution of the environmental radioactivity and the intrinsic activity of the detectors too. We can directly measure the radioactivity of rocks or estimate it from chemical analysis. 1 g of K, Th or U in 1000 kg of rock represents a massic activity of 0.0031, 4.0 or 12.5 Bq/kg, respectively. It is the aim of metallic shields of the detectors to reduce the contribution of the radioactivity of the surrounding rock or building material to the background. The mean concentration of radioactive elements in the upper earth’s crust amounts to 2.5% of potassium, 2.5 ppm of uranium and 11 ppm of thorium, which correspond to 772 Bq 40 K, 31 Bq 238 U and 44 Bq 232 Th, respectively, in 1 kg of rock. The radioactive elements belong to the large ion lithophile elements and are enriched in rocks and minerals with enhanced alkali element content like granites, and depleted in basic and ultra basic rocks. Table 1 gives mean concentrations of the radioactive elements in different rock types together with the concentration of 40 K and the massic activity of 226 Ra. Building materials are produced from rocks, with the corresponding massic activities: chalk and cement from limestone, concrete from cement and rock material, bricks from clay. We consider the most important lines in each decay chain only. The part of the spectrum with lower energy as a result of scattering process is not taken into account. The shape of the spectrum does not change after some decimeters of rock material because of the fact that the quanta with lower energy have higher absorption coefficients. The gamma-ray flux in the rocks is proportional to the concentration of radioactive elements. Examples of the gamma-ray flux in rocks containing 1 ppm of a radioactive element are given in Table 2.
Underground laboratories for low-level radioactivity measurements
217
Table 2 Gamma-ray flux in rocks containing 1 ppm of the radioactive element Long-lived nuclide a, Bq/mg of element Daughter nuclide with high gamma energy b, branching in the decay chain Eγ , MeV wγ , transition probability t, specific transition rate, s−1 mg−1 μr , absorption coefficient in rock, cm2 g−1 Φγ , gamma flux of the energy mentioned, cm−2 s−1 Mean element concentration in the upper crust, ppm Φγ , mean gamma flux in the crust, cm−2 s−1 μPb , absorption coefficient in lead, cm2 g−1
40 K
232 Th
238 U
0.0309 40 K 1.0 1.460 0.107 3.3 × 10−4 0.052 6.4 × 10−6 30,000 0.19 0.056
4.0
12.6 214 Bi 1.0 1.764 0.16 2.02 0.05 0.0404 2.5 0.101 0.047
208 Tl
0.36 2.614 1.00 1.44 0.039 0.0369 11 0.406 0.042
Table 3 Concentrations of K, Th and U in selected rocks Rock type
Location
K, %
Th, g/t
U, g/t
Dunite Serpentinite Hornblende monzonite Rock salt Limestone Granite
Twin Sisters (CA) Zöblitz (Saxony) Dresden, Plauenscher Grund Asse near Brunswick, 925 m depth Gran Sasso laboratory Mont Blanc tunnel
0.0001 0.01 3.0 <0.1 0.02–0.17 3.3–6.6
0.01 0.08 45 <0.005 0.06 22.5
0.004 0.1 10 <0.008 0.35 2.3–23
As a product of the activity a of a radioactive element, the branching ratio b of the decay chain of interest, and the transition probability ω of the level with the gamma energy of interest, we obtain for the specific transition rate t, t = abω, and with the absorption coefficient μr in the rock we obtain the specific gamma flux Φγ of the corresponding energy for the unit of weight used. From its concentration in the rock we obtain the gamma flux of the gamma-line with corresponding energy, Φγ = t/μr . The gamma flux will be reduced drastically by a lead shield. A shield of 17 cm lead reduces the flux of the gamma radiation with photon energy of 2.61 MeV about 6600-fold. The rocks where the counting equipment is located and from which the building materials are produced show large variations in chemical composition. Table 3 lists some typical values. 3.2. Measures for the reduction of the intrinsic detector radioactivity Care must be taken to avoid possible adsorption of radon decay products during the fabrication of any detector parts (Hubert et al., 1986). There is also a need to select materials with
218
S. Niese
Table 4 Massic activities of radionuclides in various materials used for fabrication of radiation detectors (Bq kg−1 ) Materials
228 Tl
214 Bi
40 K
Aluminum Beryllium Epoxy Molecular sieve Mylar aluminized Printed circuit board Quartz Reflector materials Rubber sponge Silicone foam
0.1–3.5 0.15 1–70 8 2 30 0.1–1 <0.01–1.5 1–3 0.3
<0.1–0.35 10 1–1000 40 3 70 <0.3–15 <0.1–3 1–20 0.8
0.5–15 <20 <20–1200 150 <30 70 <3 <0.1–5 <10–30 <3
lowest radioactive contamination, both for detectors and other construction parts such as photomultipliers and preamplifiers. Aluminum for the cryostat of the germanium crystal should be substituted by electrolytic copper or magnesium. Indium which is usually used as a contact element between the germanium crystal and the copper cooling may be substituted by pure lead. Examples of radionuclide concentrations in some construction materials are given in Table 4 (Brodzinski et al., 1985). For the extremely low detector background which is needed in fundamental research the materials that are used for the fabrication of the detector and its parts should be chemically purified and stored underground to decrease the production of cosmic-ray induced radionuclides in the material (Brodzinski et al., 1988, 1990). 3.3. Selection of materials for buildings, shieldings and installations Before constructing a low-level measurement laboratory we must analyze each of the materials used for the building, shielding, construction, installation and the detector. The massic activity of the bricks used for buildings outside a metallic shielded chamber in the “Felsenkeller” laboratory was not allowed to exceed the values for the surrounding rock. The metallic chamber in the “Felsenkeller” laboratory in Dresden (Niese, 1996) was constructed from steel layers of different massic activity of 60 Co. The outside sheet (9.3 g/cm2 ) of new steel contains 1.2 Bq/g, followed by 135 g/cm2 hard steel granulate with 0.2 Bq/kg and the inner 71 g/cm2 thick layers of the old steel and lead contains <0.1 Bq/kg of 60 Co. 3.3.1. Selection of paint for the protection of the inner surfaces of the shielded chamber and of installation materials In four commercial painting materials for corrosion protection of steel we found mean values of 10 and 5 Bq kg−1 of 232 Th and 226 Ra, respectively. Assuming a usage of 0.2 kg m−2 for the paint we can calculate that these values exceed the allowed massic activity. The apparent massic activity of the painting material for corrosion protection should not exceed 10% of the corresponding value for the shielding material in the same layer. As solid components the paint contains iron oxide, aluminum powder for the anti-corrosion potential and barium sulfate. For low activity paint it is advantageous to substitute barium sulfate with talcum, and aluminum powder with zinc powder and select the best iron oxides (Table 5).
Underground laboratories for low-level radioactivity measurements
219
Table 5 Massic activities of pigments used for corrosion protection (Bq kg−1 ) Sample
238 U
226 Ra
Commercial paint, a.m. Iron oxide Talcum Zinc powder
2.0 0.7 <1
5 1.3 0.6 <0.1
210 Pb
0.9
232 Th
40 K
10 0.5 0.12 <0.02
0.4 2.7 <0.004
Table 6 Activities in commercially available installation materials (Bq) Material
Quantity
238 U
226 Ra
232 Th
40 K
Temperature control Fluorescence lamp Cable Installation material Socket Distributor Switch
1 piece 1 piece 30 m 7 kg 5 pieces 1 piece 1 piece
0.3 1.8 7.5 11 0.5 <0.2 <0.05
0.35 1.9 22 18 0.2 0.4 <0.03
0.08 1.5 15 1.5 0.5 <0.2 <0.04
1.2 72 18 7.4 <2.5 <0.6 <0.3
Installation materials may also contain significant amounts of radioactive elements. The main sources of radiation are inorganic solid minerals in plastics. Table 6 shows as an example activity values in some commercially available installation materials. 3.3.2. Radon in the air To reduce the radon concentration inside the metallic shield the air is usually flushed with nitrogen, which evaporates from the Dewar. Inside the cave, the mine or other underground places depending on the geological situation we often find enhanced concentrations of radon. The circulation of fresh air from the outside would reduce the radon concentration in the underground laboratory. In the “Felsenkeller” laboratory we found mean activity concentrations of 222 Rn of about 200 Bq m−3 . After installation of an air circulation system, the activity decreased to 27 Bq m−3 , which corresponds to the activity concentration of radon outside of the cave. 3.3.3. Shielding of detectors Because of the high cost of lead with low 210 Pb content the extremely low 210 Pb activity lead (0.5–2 Bq/kg) is only necessary for the inner shielding layers (2–5 cm). The outer shielding layers may be constructed with higher 210 Pb activity lead. The lead can enhance the background not only by the bremsstrahlung from the beta rays of the daughter nuclide 210 Bi, but also from gamma-rays of 210 Pb decay products. Also decay products of the short-lived 220 Rn may be absorbed on solid surfaces, especially when they are electrically charged.
220
S. Niese
4. Cosmic rays in surface and underground laboratories 4.1. High energy neutrons and protons The overburden varies considerably from one laboratory to another. The counting system may be in a room where there is only a thin roof above, in a basement of a large building having 6 floor plates or more above the counting room, in a specially constructed room below a few meters of soil or rocks, in a cave, a tunnel used for traffic or in an old mine. In rocks and building materials the attenuation thickness of high-energy neutrons or protons is higher than in air and amounts to about 1.6 m w.e. (160 g cm−2 ). When ΦP and ΦP0 are the fluxes of high-energy neutrons or protons under an overburden of mass m in m w.e., the attenuation factor fP is given by the formula fP = ΦP /ΦP0 = e−m/1.6 . The attenuation factor is valid for neutrons too: fN = fP . 4.2. Muons Muons are very penetrating, therefore they are known as the hard component of cosmic rays; they penetrate hundreds of meters below the surface of the earth. The attenuation of muons can be approximated by the formula Φμ /Φμ0 = 10−1.32 log d−0.26(log d)
2
with d = 1 + m/10, where Φμ0 is the intensity at zero overburden (190 muons m−2 s−1 ; NCRP, 1987) and Φμ below m meters of w.e. The parameter d can be interpreted as the total overburden including the atmosphere. The equation describes the muon flux with an accuracy of about 5% up to a depth of 100 m w.e. and within 10% up to 1000 m w.e. The muon attenuation in thick lead and iron absorbers shows only small differences to the absorption in rocks (Theodorsson, 1992). The cosmic-ray muons are the main source of the background of lead shielded detectors. They ionize the matter of the detector yielding a signal proportional to the number of produced ion pairs, which itself is proportional the length of traversing the detector. Because of the fact that the ranges of the muons are an order of magnitude greater than the dimensions of detectors, the absorbed energy in the detector and its corresponding signal represents the dimension of the detector and not the energy of the muons. In germanium the muons lose about 6–7 MeV per traversed cm. The background spectrum of a 2000 mm2 × 20 mm thick LEGe detector is rather steep between 100 and 1500 keV according to the Compton scattering of gamma-rays and bremsstrahlung from cosmic rays. A broad bump around 13 MeV is attributed to the energy loss of muons traversing the detector (Verplancke, 1992). An anticoincidence detector surrounding the main detector used for radionuclide analysis is highly recommended for shielding the muons. As the muons traverse both detectors, they will be registered in both detectors within a small time interval, and therefore such a signal may be rejected.
Underground laboratories for low-level radioactivity measurements
221
In a cylindrical germanium detector with a length of 5.7 cm, a diameter of 4.86 cm, a borehole of 4.8 cm and a diameter of 1.2 cm (a relative efficiency of 20%) cosmic rays cause a background contribution up to 80 MeV with a significant maximum at about 35 MeV and a width (FWHM) of about 18 MeV (Müller et al., 1990). 4.3. Neutrons At the air/concrete interface the thermal neutron flux first rises to a maximum below an overburden of about 0.4 m w.e. as a result of the higher neutron yield of concrete in comparison with air, because of its higher Z. The conversion of high energy neutrons and protons is therefore higher under the first floor plate than in the air above it. At an overburden of about 4 m w.e. both components are equal. Some m w.e. below this depth the hadron component becomes insignificant (Theodorsson, 1992). Dep et al. (1994) showed the results of measurements and Monte Carlo calculations of the relative neutron flux under a layer of concrete of varying thickness. At 17 m w.e. Da Silva et al. (1995) found 0.081 neutrons m−2 s−1 . The neutron flux will be reduced together with the flux of muons with depth. The cosmic ray produced neutrons in rocks are the result of interactions of fast muons with the nuclei of the rock. The production rate rn of neutrons per g rock is given as rn = 1.7 × 10−3 Φμ , where Φμ is the muon flux. Taking into account the rock density ρ, with the attenuation coefficient for the neutrons from natural radioactivity kn = 1.53 × 10m−1 , we obtain the neutron flux caused by muons Φn = rn ρ/kn , Φn = 0.11ρΦμ . Under high rock cover the neutrons are mainly produced by reactions of alpha particles from the natural decay chains of thorium and uranium with light elements of the rock, and to a smaller extend by spontaneous fission of uranium. At an overburden of 3400 m w.e. in the Gran Sasso Laboratory the neutrons are originating only from natural radioactivity sources. Alpha particles of the nuclides of the Th- and U-series in the rock react with light elements and produce neutrons. In the three energy regions— thermal, 0.05 eV–1 keV and >2.5 MeV—neutron fluxes of 0.02, 0.013 and 0.026 neutron m−2 s−1 were measured by Rindi et al. (1989). Rocks as natural neutron sources In depths below 50 m w.e. the production of neutrons by (α, n) reactions with alpha particles from U and Th decay products and light nuclei like 9 Be, 17 O, 23 Na, 25 Mg, 27 Al, 29 Si became more important. Neutron production rates rN,p in oxide components of rocks are listed in Table 7 (Hashemi-Nezhad and Peak, 1995). Under steady state conditions the mass specific neutron production rate rN,p is equal to the absorption rate rN,a . Multiplying the mean neutron production rate rN,p of the rock with its density δ, and dividing by the neutron absorption
222
S. Niese
Table 7 Neutron production rates in oxide components of rocks Oxide
Weight % of the oxide in the rock
rN,p (g−1 y−1 ) per 1 ppm Th and 1 g of the oxide
rN,p (g−1 y−1 ) per 1 ppm U and 1 g of the oxide
SiO2 Al2 O3 FeO K2 O MgO H2 O TiO2 CaO Fe2 O3 Na2 O
57.6 23.8 7.52 4.21 1.78 1.16 0.97 0.91 0.89 0.81
0.184 0.89 0.157 0.23 1.15 0.096 0.45 0.048 0.137 3.70
0.42 0.48 0.046 0.078 0.62 0.046 0.25 0.023 0.049 3.00
coefficient μa,N yields the neutron flux Φ = rN,p δ/μa,N . A typical granite rock, found in the mine in Brook Hills, Australia, containing 3.3 ppm of U and 12.9 ppm of Th, yielded a neutron intensity of 19.5 cm−3 y−1 from (α, n) reactions, and together with the contribution of the spontaneous fission of 23.8 cm−3 y−1 (Hashemi-Nezhad and Peak, 1995). With the absorption coefficient μa,N = 1.53 × 10−2 cm−1 and a production rate of 7.6 × 10−7 cm−2 s−1 , they obtained a neutron flux of Φ = 0.49 m−2 s−1 for this rock. 4.4. Reaction of cosmic rays with the shielding material Inside a 10 cm thick lead shield the flux of external gamma rays from primordial radioactivity in the laboratory materials, and cosmic-ray showers formed outside the shield are suppressed. Energetic neutrons, protons, and muons penetrate the 10 cm thick lead layer, the high-energy neutrons and protons suffering considerable absorption, but muons very little. Background pulses due to protons and muons traversing the detector can be eliminated effectively, either by pulse-height discrimination, or through the use of anti-cosmic counters. The most important background component comes from the secondary radiation: neutrons, electrons and photons, which are produced in the lead. Lead has a low cross-section for thermal neutrons. Without moderating or absorbing material inside the shield, it will be transparent to thermal neutrons. The flux and spectrum of neutrons produced outside the shield will then be almost the same inside and outside the shield. Additional fast neutrons are produced in the lead itself. Lead, exposed to high-energy protons, is a very effective neutron source. The interaction length of high-energy neutrons and protons in lead is about 170 g/cm2 (1.7 m w.e. corresponds to 15 cm of lead). About 50% of the high-energy nucleons incident on the shield will therefore experience a collision, with a high probability of producing neutrons. Arthur et al. (1988) have made a very useful study of thermal and fast neutrons inside a lead shield of varying thickness. They measured the neutrons with a BF3 proportional counter, sensitive only to thermal neutrons. In order to determine the flux of fast neutrons, a 5 cm thick layer of paraffin surrounded the counter. The net
Underground laboratories for low-level radioactivity measurements
223
count rate of fast neutrons inside the lead shields increases with thickness. The background of the detector was determined by covering it with a 0.05 mm foil of cadmium that absorbs the external neutrons. The measurements demonstrate the production of fast neutrons in the lead. But the lead has no measurable effect on the thermal flux inside the shield, as measured with the bare counter. In lead the neutrons produce photons, which contribute to the background continuum in the detector. Further sources of the continuum in the background of germanium detectors are bremsstrahlung formed by inelastic scattering of muons, elastic scattering of neutrons in germanium, the formation of radionuclides in germanium, which emit beta rays, and the Compton continuum of the high energy gamma rays from neutron capture in germanium. We measured the flux of the fast neutrons within an energy range of the fission and found these fluxes—above ground 28 m−2 s−1 , under 125 m w.e. in hornblende monzonite 2.6 m−2 s−1 , in serpentinite 0.26 m−2 s−1 , and within a 5 cm thick lead shield surrounded by serpentinite 3.9 m−2 s−1 (Niese, 2007), demonstrating the influence of rock covering, natural radioactivity and lead shielding. Lindström et al. (1990) calculated above ground inside a 15 cm thick lead shield from the 596 and 693 keV peaks of inelastic scattering of 74 Ge and 72 Ge using a cross-section of 80 mb a fast neutron flux of 200 neutrons m−2 s−1 . A Marinelli beaker inside the lead shield enhances the thermal neutron flux; the authors found with a 1 l beaker a thermal flux of 200 neutrons m−2 s−1 . Hashemi-Nezhad and Peak (1995) calculated the neutron flux in a rock containing 0.81% sodium oxide, 1 ppm of uranium and 1 ppm of thorium. They obtained 0.5 m−2 s−1 , whereby half of the neutrons are formed by the (α, n) reaction with sodium. A lot of knowledge about the reaction of cosmic rays with lead was obtained using neutron monitors proposed by Simpson et al. (1953). BF3 -counters were wrapped with about 2 cm of polyethylene and 13.5 cm of lead and distributed over the earth’s surface for the continuous measurement of secondary cosmic-ray particles. Bercovitch et al. (1960) show that each of the secondary particles with energies of 200 to 300 MeV and more produces in lead at each interaction about 8 evaporation neutrons with a mean energy of about 2 MeV. The elastic neutron cross-section in lead in the energy range from a fraction of some eV to 5 MeV amounts to 5–10 barn, corresponding to a mean free path of 3–6 cm. Neutrons lose on average only 1% of their kinetic energy in each elastic collision with the lead nuclei. We can therefore conclude that the fast neutrons produced in the lead do not contribute measurably to the thermal flux inside the lead shield as long as there is no moderating material there. In the laboratory placed above ground, most of the neutrons come from interactions of high energy neutrons and protons. They can cause spallation reactions. Therefore we can see in the gamma spectrum of a shielded germanium detector, spallation products of germanium, e.g., 57 Co. In a lead shield, neutrons are produced by high energy neutrons and protons, their production by muons becoming important with an overburden of some meters. The neutron yield is growing with increasing atomic number Z of the material. Fast neutrons from lead produce in germanium 71 Ge, 72 Ga, and 76m Ge by (n, 2n), (n, p), and (n, n ) reactions. As a product of the (n, γ ) reaction with thermal neutrons we observe 71 Ge too. In deeper underground, where the muon flux is reduced, the formation of neutrons by alpha particles from natural radionuclides dominates. The energy of the fast neutrons is lower and a part of them are thermalized by light elements in the rock. Than we only observe products of (n, n ) and (n, γ ) reactions.
224
S. Niese
Table 8 Activation of germanium with an isotopic neutron source outside a lead shield Eγ , keV
Counts, h
Counts, h (+ acryl)
Assignment
53.4 66.7 68.7 139.6 159.5 174.9 198.3 595.6
13.4 ± 5.4 4.5 ± 5.0 16.6 ± 8.0 7.4 ± 4.5 7.1 ± 5.6
72 Ge(n, γ )73m Ge, 74 Ge(n, 2n)73m Ge
6.7 ± 4.2 6.5 ± 5.0
19.1 ± 9.5 20.0 ± 8.1 4.9 ± 4.9 25.5 ± 6.6 49.9 ± 8.0 14.6 ± 7.8 21.8 ± 5.2 15.8 ± 5.4
691.3
10.6 ± 5.3
17.8 ± 7.5
72 Ge(n, n γ )72m Ge, asymmetric and broadened
73m Ge, sum: 13.3 + 53.4 keV
73 Ge(n, n γ )73m Ge, asymmetric and broadened 74 Ge(n, γ )75m Ge, 76 Ge(n, 2n)75m Ge 76 Ge(n, γ )77m Ge 70 Ge(n, γ )71m Ge, 72 Ge(n, 2n)71m Ge 70 Ge(n, γ )71m Ge 72 Ge(n, 2n)71m Ge
74 Ge(n, n γ )74m Ge, asymmetric and broadened, 73 Ge(n, γ )74m Ge
Table 9 Background counting rate (r in 24 h) of a Ge-detector with different shielding. Within the shield a plastic box may be filled with 4.8 L of water. Pb means 10 cm of additional lead outside of the shielding; Cd means 0.5 mm of Cd-foil around the detector; 6 Li = 1% solution of 6 Li-acetate and 10 B = 1% solution of sodium borate Eγ , keV
Nuclide
r
AC
H2 O
AC + H2 O
AC + H2 O + Cd
AC + H2 O + 6 Li
H2 O + 10 B
Pb + AC
53.6 66.7 139.5 159.7 198.4 596.6 691.3 569.6 803.3 1063
Ge73m Ge73m Ge75m Ge77m Ge71m Ge74∗ Ge72∗ Pb207∗ Pb206∗ Pb207∗
64 <50 89 48 121 59 90 41 <23 18
68 43 45 17 46 6.3 9 10 8 7
270 240 404 327 407 151 69 <17 <16 <13
220 240 320 71 266 81 14 <7 <8 <7
37 29 37 14 38 16 <3 <4 28 <3
88 68 53 <6 62 4.7 <2 <3 <2 <2
54 56 44 <6 58 3.3 <2 2.5 3.5 4.6
41 75 51 5.1 55 9.1 3.9 5.6 <1 3.6
∗ means excited level.
It is not simple to assign the peaks in the background spectrum to corresponding neutron reactions. Table 8 shows the results of experiments where a fast neutron source of 3.7 GBq of 241 Am surrounded by Be emitting about 103 neutrons s−1 was put on a 10 cm thick lead shield. The total distance between the neutron source and detector amounts to about 20 cm. Measurements were done with and without 1 cm additional acrylic lining inside a 2000 mm2 × 20 mm Ge detector (Verplancke, 1992). In both cases we have a mixture of fast and thermal neutrons. Arthur and Reeves (1992) investigated the influence of different measures on the formation of peaks in the background of a Ge-detector. Counting rates of a Ge-detector with 31.5% relative efficiency, shielded with 15 cm lead and outside with a 10 cm plastic scintillator (AC = anticoincidence on) placed above ground, are presented in Table 9. Water samples inside the lead shield enhances the flux of thermal neutrons and the yield of (n, γ ) reactions. The absorber for thermal neutrons Cd, 6 Li, and 10 B reduce the thermal flux
Underground laboratories for low-level radioactivity measurements
225
Table 10 Calculation of neutron fluxes from gamma lines in the background of a Ge detector (NGe = 7E24) Eγ , keV
Nuclide
h
σ , barn
ω
ε
r, s−1
Φ, m−2 s−1
53.6 139.5 159.7 596.6 691.3
Ge73m Ge75m Ge77m Ge74∗ Ge72∗
0.2737 0.3674 0.0767 0.3674 0.2737
1.0 0.16 0.9 0.080 0.080
0.105 0.34 0.109 1.0 1.0
0.9 0.9 0.9 0.7 0.7
0.74E−3 1.03E−3 0.56E−3 0.68E−4 1.04E−3
th: 41 th: 82 th: 117 f: 330 f: 680
Table 11 Counting rates from neutron reactions in the background of Ge-detectors above ground and underground Eγ , keV
FK
PTB
Reactions
FK/PTB
140.5 197.9 278 595.8 A.M.
5E−5 3E−5 4.5E−5 3E−5
6.3E−4 6E−4 5.9E−4 4E−4
76 Ge(n, 2n)75m Ge; 74 Ge(n, γ )75m Ge
0.08 0.05 0.08 0.07 0.075
70 Ge(n, γ )71m Ge; 72 Ge(n, 2n)71m Ge 63 Cu(n, γ )64 Cu; 65 Cu(n, 2n)64 Cu 74 Ge(n, n ); 74 Ge(n, p)74 Ga
and the corresponding reactions. The additional lead does not enhance the flux of the neutrons. It seems that the enhancement by the higher mass of lead is compensated by the attenuation of high-energy hadrons. Using the available nuclear data for the cross-sections σ , transition probabilities ω, and the estimated efficiencies ε, the fluxes of thermal and fast neutrons can be calculated from the detector counting rates (Table 10). The counting rates under germanium peaks obtained with a Ge borehole detector, 30% (150 cm3 , 0.80 kg) located in a metallic chamber (20 cm of iron and 3 cm of lead) at 125 m w.e. with additional shielding of 17 cm of lead located in the “Felsenkeller” underground laboratory are compared with the values obtained with a 32% Ge-detector (160 cm3 , 0.856 kg) located in a counting room in the ground floor of a building above ground of the PhysikalischeTechnische Bundesanstalt Braunschweig (PTB), with walls of 10 cm lead and additional detector shields of 10 cm lead are compared in Table 11. The counting rates in the detector of the PTB (Kolb, 1992) are lower by a factor of about 1.5 than the values obtained by Arthur and Reeves (1992)—this may be explained by some differences in the equipment and in the geomagnetic position. In the “Felsenkeller” underground laboratory high-energy hadrons are reduced by orders of magnitude and muons by a factor of about 50. The observed low reduction by a factor of about 13 may be explained by the natural radioactivity of the rock. 4.5. Electrons and photons In concrete shielding, the radiation length of high-energy cosmic-ray photons and electrons is 22 g/cm2 and their critical energy 38 MeV. The maximum electron density in a shower initiated by a 2.2 GeV electron will occur at a depth of 145 g/cm2 below its start. This is equivalent to the mass in three concrete floor plates. The electron–photon cascade inside the
226
S. Niese
laboratory can therefore come from both the outside and from building materials. Because the critical energy is lower in concrete than in air, the gamma spectrum inside the laboratory is shifted to lower energies compared to outside air. Electromagnetic showers, the soft component of cosmic rays that are formed outside the shield, cannot penetrate the 10 cm thick lead layer. A new generation of showers, however, is formed in the lead and this will give rise to a flux of electrons and photons inside the shield. These showers are secondary protons formed from pions and muons, mainly through knockon electrons. With a small overburden the muons produce about 2/3 and the protons about 1/3 of the electron flux. It can be assumed that the ratio to the gamma flux will be similar. The muon component decreases only slowly but the component of high-energy neutrons and protons exponentially. While at 5 m w.e. the nucleonic component has fallen to an insignificant value, the muon flux falls only to 55% of its surface value. This trend can be assumed to be similar in lead. 4.6. Influence of the depth on the detector background The neutron flux enhances within the first meter water equivalent of a rock layer or concrete. Within the following ten meters the neutron flux is reduced mainly by the reduction of the high-energy hadron component. Within some hundreds of meters w.e. the background is reduced due to a lower muon flux. Below a depth of 100 m w.e., beside the intrinsic background of the counting equipment and the reduced flux of muons, the contribution of neutrons to the background is determined mainly by the radioactivity of rocks. Study of the contribution of non-cosmic-ray components to the detector background in deep underground laboratories is important for investigation of rare nuclear decays and processes. These results are also very valuable for design and operation of medium deep underground laboratories used for the measurement of low radioactivity. 5. Selected underground laboratories A list of underground laboratories in order of their depth is given in Table 12. Depths of most of the places is taken from a table given by Povinec (1994). Table 12 Survey of some underground laboratories Institute
Location
Type
MPI für Kernphysika
Heidelberg, Germany Monaco Tucson, Arizona Houston, Texas Bern, Switzerland
Cave
IAEAb University of Arizona NASA University of Bern
Depth, m
Cave Underground building Shaft
12 10
Shaft
25
m w.e.
Application
15
γ -Spectrometry
35
Low-level counting LSC
20
NAA 70
β-Counting (continued on next page)
Underground laboratories for low-level radioactivity measurements
227
Table 12 (continued) Institute
Location
Type
University of Tokyo
Nokogiri-yama, Japan Dresden, Germany Ogoya, Japan Freiberg, Germany Geel, Belgium Solotvina, Ukraine Asse, Germany Soudan Canfranc, Spain Kamioka, Japan Broken Hills, Australia Gran-Sasso, Italy Baksan Valley, Russia Homestake, SD, USA Frejus, France
Cave
VKTAc Kanazawa University Mining Academy IRMMd INRe Ukrain. Acad. Sci. PTBf University of Minnesota University of Zaragoza University of Tokyo University of Sydney INFNg INIh , Russ. Academy PNL-USCi LSMj SMSRk , Italy University of Kingston TIFRl
Mont Blanc Sudbury, Ontario Kolar Field, India
Depth, m
m w.e.
Application
180
Low-level counting
Cave
47
125
Low-level counting
Copper mine Silver mine
135 147
270 390
γ -Spectrometry
Test mine
223
500
γ -Spectrometry
Salt mine
430
1000
Physics
Salt mine
750
1750
Mine Highway tunnel
1800 2100
γ -Spectrometry and dosimetry Physics Physics
Mine
2700
Neutrino research
3500
Neutrino research
Cave
4400
Gold mine
4400
Neutrino research, 2β-decay 2β-Decay
Ore mine
1230
Highway tunnel
1400
Highway tunnel
1780
4800
Highway tunnel Mine
2200
5000 6200
Gold mine
8500
Physics, low-level counting Neutrino research Neutrino observatory Neutrino physics
a Max-Planck-Institut für Kernphysik. b International Atomic Energy Agency. c Verein für Kernverfahrenstechnik und Analytik Rossendorf. d Institute for Reference Materials and Measurements. e Institute of Nuclear Research of the Ukrainian Academy of Sciences. f Physikalische-Technische Bundesanstalt, Braunschweig. g Istituto Nazionale Fisica Nucleare, Gran Sasso. h Institute of Nuclear Research of the Russian Academy of Sciences, Moscow. i Pacific Northwest Laboratory and University of South Carolina. j Laboratorie Souterrain de Modane (LSM) operated by the Centre National de la Recherche Scientifique (CNRS). k Service Mixte de Securité Radiologique, Monthery. l Tata Institute of Fundamental Research, Mumbay.
228
S. Niese
5.1. Underground laboratories near the surface In laboratories located at depths of 10–40 m w.e., the hadron component of cosmic rays is reduced drastically. The contribution of muons to the detector background can be effectively reduced with the help of active anticoincidence shielding (Heusser, 1995). Max-Planck-Institut für Kernphysik, Heidelberg A counting room is installed within a cave near the Institute. Germanium detectors, which are produced from radioactivity-free materials are shielded by lead which is surrounded by an anticoincidence guard made of gas counters with gold plated tungsten wires (Heusser, 1995). Physics Institute, Garching near Munich A special room was built with concrete walls and a 5 m thick layer of soil over it. In this lab besides other installations a germanium spectrometer was installed in a lead shield, together with anticoincidence shielding consisting of plastic scintillators. University of Arizona, Tucson A dating laboratory was constructed under the building at 10 m depth, housing “Quantulus” liquid scintillation spectrometers made by Wallac. Their background for 14 C measurements is reduced by a factor of 2 (Kalin and Long, 1989). NASA-JSC laboratory The Radiation Counting laboratory (RCL) was built 20 m underground for the measurement of extraterrestrial samples at the NASA Johnson Space Center (JSC), Houston, Texas (Lindström et al., 1990). The laboratory has been mostly used for analysis of cosmogenic radionuclides in lunar samples and meteorites. IAEA-MEL laboratory The underground laboratory is situated in an underground cellar at a depth of 30 m w.e. The laboratory is equipped with ventilation and an air conditioning system maintains overpressure in the laboratory, stable humidity and temperature. A common lead shielding houses four large volume HPGe detectors. Anti-cosmic plastic scintillation detectors surround the lead shielding and protect all the detectors against cosmic radiation. An anti-Compton gammaspectrometer, comprising an n-type HPGe detector and NaI(Tl) shielding, as well as a Quantulus low-background liquid scintillation spectrometer have also been installed in the underground laboratory. The lead shield is comprised of two layers. The outer layer is made of low activity lead 7.5 cm thick, and the second, the internal layer, is made of very low activity lead (210 Pb below 0.1 Bq kg−1 ) 5 cm thick. Monte Carlo modeling of background characteristics showed that the background continuum in the gamma-ray spectrum is produced by interactions of cosmic-ray muons with the lead shield (Vojtyla and Povinec, 2000). The anticoincidence background gamma-ray spectra of the HPGe detectors showed that they are similar to those measured at 200 m w.e. (Povinec et al., 2004). This clearly indicates that the anticoincidence shielding is very important for detectors placed at shallow depths (Povinec et al., 2004, 2005).
Underground laboratories for low-level radioactivity measurements
229
5.2. Medium depth underground laboratories Physical Institute of the University of Bern In 1975 a low level laboratory was installed at 25 m depth (70 m w.e.) in a cellar under the Physical Institute of the University of Bern (Oeschger et al., 1981), reachable by an elevator. 0.4 m thick walls of low radioactive concrete made of serpentinite and Danish cement should reduce the background from the surrounding environment. This material reduced the gamma flux by a factor of ten, the muon flux was reduced by a factor of ten by the rock overburden, and the neutron flux by a factor of 4. For a comparison the authors measured the muon flux in a side-tunnel of the St. Gotthard-Strassentunnel with 1100 m granite overburden (3000 m w.e.), where it was reduced by a factor of 106 . The underground laboratory is housing proportional counters (measurement of 14 C, 37 Ar, 39 Ar), liquid scintillation counters and HPGe detectors. “Felsenkeller” underground laboratory in Dresden The underground laboratory is installed at a depth of 47 m (125 m w.e.) within a cave in the Weisseritz valley at normal traffic level. An old brewery formerly used the cave. After preliminary measurements with NaI(Tl) and Ge(Li) detectors (Helbig et al., 1982) a special counting chamber with a 0.7 m thick shield of the low activity ultramafic rock serpentinite from Zöblitz (Saxony) was installed in the cave (Helbig et al., 1984). It shields the radiation from the surrounding rock. The rock in this region consists of hornblende monzonite (syenite) and contains 5% of K, 50 ppm of Th and 10 ppm of U. This corresponds to 1300, 72 and 120 Bq/kg of 40 K, 208 Tl and 214 Bi, respectively. 46% of cosmic muons are shielded by the rock cover. The river Weißeritz flows upstream through a basin (“Döhlener Becken”) where in the past uranium-rich charcoal was mined and stockpiles with enhanced uranium content are located. This yields in an enhanced radon activity of about 30 Bq m−3 in the air of the valley. Despite the enhanced radioactivity in the environment, the laboratory has been installed in this cave because of its easy access and the vicinity of a number of institutes. In 1995 the laboratory was reconstructed and an additional counting chamber with walls of steel and lead was installed. The old chamber was incorporated in the new laboratory. The length, width and height of the new chamber are 6, 3 and 2.2 m, respectively. The shielding consists from outside to inside of 10 mm of a new steel, 270 mm of granulate hard steel, 36 mm of an old steel, 30 mm of lead and 12 mm old steel. The total surface thickness is 210 g cm−2 . For thermal isolation a 10 m thick layer of polystyrene foam (styropore) is used. The detectors themselves are shielded separately with 10 to 17 cm of lead (Niese, 1996). The activity of 222 Rn in the cave amounts to 200 Bq/m3 . Heated fresh air is introduced from outside into the laboratory. The evaporated nitrogen from the cryostat is used to replace the radon-containing air within the shielding of the detector. Most of the remaining peaks in the background spectra are caused by reaction products of neutrons. According to Heusser (1995) the neutron flux density from the cosmic rays at sea level is about 80 m−2 s−1 and under 125 m w.e. 0.5 m−2 s−1 . Most of the neutrons above ground are produced by muons in the lead shielding. The source of the excess of neutrons in this underground laboratory is the natural radioactivity in the hornblende monzonite. Hashemi-Nezhad and Peak (1995) calculated the neutron flux in a rock containing 0.81% of sodium oxide, 1 ppm of uranium and 1 ppm of thorium. They obtained 0.5 m−2 s−1 , whereby about one half of the neutrons are produced by (α, n) reactions with sodium. The hornblende
230
S. Niese
monzonite, which contains 10 ppm of U, 50 ppm of Th and 6% of Na2 O, should yield a neutron flux of 50 m−2 s−1 , which explains the high flux of neutrons observed in the laboratory. Grand Coulee Dam The Batelle Memorial Institute, Richland, Washington has installed counting devices in a gallery of the Grand Coulee Dam 110 m under the top of the dam. The solid concrete provides 78 m of horizontal shielding downstream and 13 m upstream. They used this place for the investigation of the sources of background in NaI(Tl)-spectrometers (Kaye et al., 1972). Ogoya underground laboratory The laboratory was constructed by the Low-Level Radioactivity Laboratory of the Kanazawa University, Japan, in a tunnel of a former copper mine with an overburden of 135 m of rock. The tuff breccias have a density of 2.0 g cm−3 , that yields a depth of 270 m w.e. (Komura, 1997). More than 10 HPGe detectors have been installed in the laboratory. IRRM Geel The laboratory was constructed in a special mine designed for investigations of the behavior of nuclear wastes in clay. A place at 225 m depth was reserved for the low-level counting laboratory (Wordel et al., 1993). The laboratory is equipped with various HPGe detectors. 5.3. Deep underground laboratories for nuclear physics experiments Baksan neutrino laboratory Experiments for low radioactivity measurements were performed in a horizontal cave in the Baksan valley at 660 m w.e. To reduce the background from the environmental shale rock a two-layer wall has been installed. The outer one consists of concrete from low-activity dunite and the inner one from dunite without cement (Kovalchuk et al., 1975). Later a new laboratory was installed in a cave at a depth of 4400 m w.e. where rare decay experiments were carried out (Pomansky, 1986). Solotvina underground laboratory The laboratory was constructed in 1984 by the Institute for Nuclear Research of the Ukrainian National Academy of Sciences. It is situated on the west of Ukraine, in Solotvina, a small town near the border with Romania. The town is known for its salt mines, a large Allergy Hospital with an underground department is located there as well, and salt lakes with medical mud. The laboratory is built in the salt (NaCl) mine 430 m underground (146 m below sea level). It consists of one big and few smaller halls with full area near 1000 m2 . The natural temperature is 22–24 ◦ C. At the depth of 1000 m w.e. the cosmic ray flux is reduced by a factor of 104 (to the value of 1.7 × 10−2 m−2 s−1 ). The neutron flux is below 2.7 × 10−2 m−2 s−1 . The radon concentration in the air is below 33 Bq m−3 . Due to the low-radioactive contamination of salt, the natural background is 10–100 times lower than in most other underground laboratories. The principal scientific goal of the Laboratory is the search for rare processes in nuclear and particle physics (e.g., double beta decay of atomic nuclei) (Zdesenko et al., 1988).
Underground laboratories for low-level radioactivity measurements
231
PTB underground laboratory In 1991 the Physikalische-Technische Bundesanstalt (PTB) established an underground laboratory for dosimetry and spectrometry (UDO) in the Asse salt mine, near Braunschweig. Due to the depth of 925 m (2100 m w.e.) the cosmic muon flux is reduced by 5 orders of magnitude. There is also an advantage of a low activity of the surrounding rock salt and low concentration of radon in the air. A 88% relative efficiency HPGe-detector located in a shield of 20 cm of low activity lead and 1 cm of electrolytic copper had a background integrated over 40 to 2750 keV of 0.012 s−1 kg−1 (Neumaier et al., 2000). Homestake gold mine laboratory Raymond Davis of Brookhaven National Laboratory installed in 1965 in the Homestake gold mine the first neutrino experiment studying the reaction 37 Cl(νe , e− )37 Ar, which yielded him the Nobel Prize for physics in 2002. Miniature gas counters were used for measuring the 37 Ar activity. The method of reducing the background of Ge-detectors was studied by Brodzinski et al. (1988). A lot of interesting experiments in nuclear physics were carried out in this mine, including efficient active and passive shielding, selection of materials and their purification from cosmic-ray-induced nuclides, and studies of pulse-shape discrimination for obtaining very low background (Brodzinski, 2005). The Soudan laboratory The laboratory, operated by the School of Physics and Astronomy of the University of Minnesota, is located in the Soudan Underground Mine State Park in the oldest iron mine in Minnesota. The first physics experiments at Soudan began in 1980. The current laboratory was constructed between 1984 and 1986. At a depth of 1800 m w.e. the number of cosmic-ray particles is reduced by a factor of 100,000. Modane underground laboratory The laboratory is located in the middle of the tunnel of Fréjus which links Modane to Bardonecchia through the France–Italy border, at 1260 m altitude. It was created in 1982, with the construction of the tunnel. It’s made of a principal hall (31 × 10 × 11 m) and several small rooms. The big hall has been occupied by the NEMO (Arnold et al., 2005) and EDELWEISS (Martineau et al., 2004) experiments. Numerous germanium detectors are installed in “white rooms” to measure low-activity samples for oceanography, material selection, archeology (14 C counters) and radioecology (Torres and Hubert, 1996). Gran Sasso National Laboratory This is the largest underground laboratory in the world, dedicated for experiments in nuclear and particle physics and nuclear astrophysics. It is located between the towns of L’Aquila and Teramo, about 120 km from Rome at the side of the 10 km long freeway tunnel crossing the Gran Sasso Mountain. The laboratory consists of three large experimental halls, each about 100 m long, 20 m wide and 18 m high and service tunnels, with a total volume of about 180,000 m3 . The 1400 m rock cover (3300 m w.e.) gives a reduction factor of the cosmic ray flux of about one million; moreover, the neutron flux is a thousand times less than on the surface, thanks to the low uranium and thorium content of the dolomite rocks of the mountain. Using a spectrometer with 102% efficiency, a background of 0.0025 s−1 was measured in
232
S. Niese
the energy range from 50 to 2750 keV (Neder et al., 2000). The aim of the laboratory is to host experiments that require a low background environment. The research topics include neutrino physics with neutrinos produced naturally in the Sun and in Supernova explosions, and neutrino oscillations with a beam from CERN, the search for neutrino mass, double beta decay, dark matter, and other nuclear studies of astrophysical interest (Arpesella, 1996). Canfranc underground laboratory The laboratory is located in the international railway tunnel of Somport at Canfranc in the Spanish Pyrenees. The first installation (1986) consisted of two small halls (about 10 m2 each) located 780 m from the Spanish entrance below a rock overburden of 675 m w.e. In 1988, a prefabricated cabin, specially reinforced (of about 15 m2 ), was installed next to the two halls. In 1991 a new prefabricated cabin (about 27 m2 ), was added to the one already in operation and both were moved to a new location 1200 m from the Spanish entrance (below an overburden of 1380 m w.e.). In 1994, during the excavation works for a new motorway, a new experimental hall, 118 m2 , 2520 m from the Spanish entrance and below an overburden of 2450 m w.e. was constructed. Zaragoza University, in collaboration with many international institutes, carried out in the Canfranc laboratory experiments on rare nuclear decays. Sudbury Neutrino Observatory The laboratory is operated by the Queens University, Kingston, Ontario, Canada. In the ˇ 6200 m w.e. deep Creighton mine a 1000 t D2 O Cerenkov detector was installed to look for cosmic neutrinos. 6. Detectors and detection modes 6.1. Comparison of background of Ge detectors The comparison of the backgrounds of different detectors in different locations is helpful for studies of the sources of background. We can compare the counting rates under different peaks or in different energy regions. A very useful approach is to measure the background between 40 and 2750 keV and normalizing it to 1 kg of germanium with the assumption that the background is nearly proportional to the mass of the detector. Table 13 gives some background values measured by the CELLAR (Collaboration of European Low-level Underground Laboratories) group (Laubenstein et al., 2004) without and with an anticoincidence (AC) shielding. We see that the reduction of the background with depth corresponds with the muon flux until a depth of about 500 m w.e. At deeper depth the muon flux is reduced more than the background, which indicates that other sources of background become more important. Table 13 demonstrates that the background is reduced by more than three orders of magnitude when Ge-detectors are installed in deep underground laboratories. Also anticoincidence shielding is valuable in shallow depth laboratories. 6.2. Gas counters in underground laboratories Loosli and Oeschger (1982) demonstrated that the background of a gas proportional counter is nearly proportional to its volume. They showed that, even at a depth of 70 m w.e., where the
Underground laboratories for low-level radioactivity measurements
233
Table 13 Comparison of the background of Ge-spectrometers between 40 and 2750 keV in CELLAR laboratories (normalized to 1 kg of Ge) (after Laubenstein et al., 2004) Laboratory
Location
Depth, m w.e.
Background, s−1 kg−1
Background with AC, s−1 kg−1
ARCa MPI IfK IAEA-MEL VKTA Kanazawa Univ. IRMM PTB LNGS LSCE
Seibersdorf Heidelberg Monaco Dresden Ogoya Geel Asse Gran Sasso Frejus
1 15 35 125 270 500 2100 3800 4800
0.58 0.2 0.08 0.042 0.009 0.0030 0.0032 0.0010 0.00035
0.095 0.024 0.010
a Austrian Research Center.
nucleonic component of the cosmic rays is reduced by 9 orders of magnitude, and the flux of muons by a factor 11, the background was only reduced by a factor 2–4.5. This demonstrates that the main source of background is radioactivity in the detector itself and in its surrounding. A further reduction of the background was achieved using miniature proportional counters, as demonstrated in the GALLEX experiment (Heusser, 1994). For large volume counters a system with a multi-element structure can be used in which a single wire counter is replaced by a chamber consisting of seven or more counters of the same dimension arranged, e.g. hexagonally, separated from each other by cathode wires only. This enables the use of external as well as internal anticoincidence (e.g., in the case of 3 H analysis; Povinec, 1972; Povinec et al., 1979). The chamber works simultaneously as a track detector (Povinec, 1980; Povinec et al., 1990). This principle was successfully applied for the study of double beta decay of Xe in the Gran Sasso and Baksan laboratories (Bellotti et al., 1992). 6.3. Liquid scintillation counting While a background of commercially available HPGe-detectors may be close to theoretical limits, liquid scintillation counters (LSC) are far from their limits. About 50% of the background of commercial low-level LSC is due to the equipment itself, which contains materials not free from radioactivity. Even in the Gran Sasso underground laboratory a LSC of the Quantulus type shows in the radiocarbon region only a reduction of 65% in comparison with the measurements at the surface (Plastino et al., 2001). This is due to the activity of the equipment itself, e.g., the potassium content of the photo multiplier tubes. Installation of LSC in an underground laboratory does not yield such an advantage as in the case of a Ge-detector. However, the work in fundamental physics demonstrates the principal possibility to reduce the background of liquid scintillation apparatus using the experiences demonstrated in the BOREXINO experiment (Heusser et al., 2006).
234
S. Niese
6.4. Coincidence and anticoincidence systems and pulse shape discrimination The background of a coincidence system is further reduced underground. In a low-background chamber under 100 m w.e. sandstone near Mt. Nokokiriyama, 100 km south of Tokyo, Tanaka et al. (1967) have arranged a coincidence system consisting of two NaI(Tl) detectors in a 20 cm thick iron cell on which a plastic scintillator was placed as an anticoincidence guard. Beta-gamma coincidence methods applied underground may also reduce the background by more than one order of magnitude, without a drastic reduction of the efficiency (Niese et al., 1989; Goldbrunner et al., 1997). An anticoincidence detector arrangement with veto detectors reduces the background, which is caused by charged particles from outside the detector system. This principle is very effective above ground, in low and medium depth underground laboratories, and if extreme pure materials for the detector construction are used, in deep underground laboratories too. The technique is advantageous in all cases when the charged particles from outside contribute a significant part to the background. Pulse shape discrimination, which is used in most of the commercial liquid scintillation counters, can also be used in low-level gamma spectrometry for analysis of ultra-low radioactivity samples (Brodzinski, 2005).
7. Applications In a number of underground laboratories a broad variety of samples for studying different problems have been measured. In this review we shall present only some of the applications. 7.1. Dating and atmospheric physics There were many applications in radiocarbon dating and in using radionuclides in the atmosphere as tracers. The first underground radiocarbon measurements with gaseous samples started in Bern (Oeschger et al., 1981), where also investigations of argon isotopes in the atmosphere and hydrosphere have been carried out. Liquid scintillation spectrometers have been installed underground in several laboratories (e.g., Kalin and Long, 1989; Plastino et al., 2001; Povinec et al., 2005). Gamma-spectrometry has been widely used for investigation of cosmogenic and radiogenic radionuclides in the atmosphere and rain water (e.g., Komura et al., 2006). 7.2. Investigation of meteorites In several laboratories the radioactivity of meteorites, especially cosmogenic radionuclides, have been frequently measured. However, after the development of accelerator mass spectrometry (AMS), many long-lived radionuclides (e.g., 10 Be, 14 C, 26 Al, 53 Mn) were measured by AMS, and radiometric methods concentrated on short-lived radionuclides [e.g., the meteorites Hohenlangenbeck (Niese et al., 1990), Trebbin (Niese and Helbig, 1990), Kobe (CK4) (Komura et al., 2002)]. For long-lived nuclides like 26 Al radiometric measurements are used for large samples to obtain a mean value (Arnold et al., 2002).
Underground laboratories for low-level radioactivity measurements
235
7.3. Neutron activation analysis Measurement of neutron-activated samples in an underground laboratory was used by Niese and Helbig (1986), Niese et al. (1993) and Lindström et al. (1990) with the aim of improving the detection limit for ultra pure materials or to analyze small samples. 7.4. Radioactivity in the environment The largest number of samples comes from investigations and controls of anthropogenic and natural radionuclides in the environment. Radioactivity in the environment was investigated not only for radiation protection but also for the investigation of transport processes in nature. A large number of samples of mine water, river water, sediments and plants came from the remediation of former uranium mines (Niese, 1996). 7.5. Marine radioactivity Low concentrations of radionuclides in seawater require the use of either large volume samples or highly sensitive detection systems (Reyss et al., 1995). For example, the analysis of 137 Cs using a gamma-spectrometer placed in an underground laboratory allows a reduction in the amount of seawater at least by a factor of 10 (Povinec, 2004, 2005; Povinec et al., 2001; Hirose et al., 2005). 7.6. Nuclear weapons explosions 60 years after the atomic bomb explosion in Hiroshima, it was possible by measuring 152 Eu in various samples to determine the doses due to neutrons at different places. It was even possible to measure the shorter-lived 60 Co in steel samples from Hiroshima (Hult et al., 2004). 7.7. Nuclear accidents After an accident in the Tokaimura (Japan) fuel plant, neutron doses at different places in the vicinity of the facility were investigated. Jewels from gold and various metals were analyzed in the underground laboratory (Komura, 1997). 7.8. Intrinsic radioactivity of detectors An underground laboratory is a very useful tool for the determination of the intrinsic radioactivity of the detector itself. If the background of the equipment in the underground laboratory is not reduced to the same extent as the intensity of cosmic radiation then we can determine the intrinsic background of the equipment itself. While the intrinsic background of commercial low-level gamma spectrometers is now relatively low as the result of improved selection of construction materials, other commercially available devices often show higher intrinsic background. For example, the background of the best commercially available LSC, the “Quantulus” placed in the Gran Sasso underground laboratory, was reduced only by a factor 2, which shows that half of the background is due to the material of the equipment. A similar situation is observed with automatic “low-level” proportional counters.
236
S. Niese
The PTB underground laboratory has also been used for the calibration of radiation dosimeters (Neumaier, 2000).
8. Conclusions Underground laboratories have proved to be important prerequisites for low-level analysis of radionuclides in different matrices and for applications of radionuclide tracers in the environment. The concept of shielding with different layers, and the arrangement of part of the shielding as the wall of a counting room, is advantageous to apply in medium depth underground laboratories. The remaining background inside the metallic shielding is caused by cosmic muons and neutrons and neutrons produced by muons, spontaneous fission and (α, n) reactions within the rock. The contribution of neutrons of both sources may be reduced by different kinds of shielding, which must be arranged between the two parts of the metallic shield. In surface laboratories the lead shield is necessary to reduce the soft component of cosmic rays, the influence of the radioactivity of the building, and the influence of radionuclide sources handled in the laboratory itself. There are a number of materials, e.g., ultra basic rocks, marble, steel, plastics, which can be used instead of normal concrete or bricks. The thickness of the high-density metallic shielding can then be reduced, and as a result, the production of secondary neutrons in the shield will be lower in comparison with a thick metallic shield. Low radioactivity materials for the construction of buildings are necessary, especially for the measurement of large samples, e.g., for whole body counting, which are too large for conventional lead shielding. In shallow and medium depth underground laboratories anticoincidence shielding can further reduce the background from cosmic muons. A lot of experience has been gathered both in construction of detectors and their operation in underground laboratories, including coincidence–anticoincidence techniques and software, suitable for the investigation of single events.
Acknowledgements The construction of the “Felsenkeller” underground laboratory in Dresden has given me the opportunity to collect a lot of experience in low-level counting. During the work I had the assistance of Wolfgang Helbig, Dieter Birnstein, Matthias Köhler and Michael Saupe. In the nineties Nuclear Engineering and Analytics Inc. reconstructed the lab, which was founded in 1982 by the Nuclear Research Centre Rossendorf, in a generous way and has given me the possibility to continue with underground experiments after my retirement.
References Arnold, D., Neumaier, S., Sima, O. (2002). Deep underground gamma spectrometric measurement of 26 Al in meteorite samples. Appl. Radiat. Isotopes 56, 405–408. Arnold, J., et al. (2005). Technical description and performance of the NEMO3 detector. Nucl. Instrum. Methods A 536, 79–122.
Underground laboratories for low-level radioactivity measurements
237
Arpesella, C. (1996). A low-background counting facility at Laboratori Nazionali del Gran Sasso. Appl. Radiat. Isotopes 47, 991–996. Arthur, R.J., Reeves, J.H. (1992). Methods for achieving ultra-low backgrounds in above-ground germanium detector systems. J. Radioanal. Nucl. Chem. 160, 297–304. Arthur, R.J., Reeves, J.H., Miley, H.S. (1988). Use of low-background germanium detectors to pre-select high purity materials intended for construction of advanced ultra low-level detectors. IEEE Trans. Nucl. Sci., 582–585. Bellotti, E., Cremonesi, O., Fiorini, E., Gervasio, G., Povinec, P., Ragazzi, S., Rossi, L., Sverzelatti, P.P., Szarka, J., Tabarelli de Fatis, T., Zanotti, L. (1992). Multielement proportional chamber—A study of background in the Gran Sasso underground laboratory. Nucl. Instrum. Methods A 323, 125–134. Bercovitch, M., Carmichael, H., Hanna, G.C., Hincks, E.P. (1960). Phys. Rev. 119, 412–431. Brodzinski, R.L. (2005). Next generation germanium spectrometer background reduction techniques at 2 MeV. J. Radioanal. Nucl. Chem. 264, 139–143. Brodzinski, R., et al. (1985). Nucl. Instrum. Methods A 239, 207. Brodzinski, R.L., Reeves, J.H., Avignone III, T., Miley, M.S. (1988). Achieving ultra low background in a germanium spectrometer. J. Radioanal. Nucl. Chem. 124, 513–521. Brodzinski, R.L., Miley, M.S., Reeves, J.H., Avignone III, T. (1990). Further reduction of radioactive backgrounds in ultra sensitive germanium spectrometers. Nucl. Instrum. Methods A 292, 337–342. Cocconi, G., Tongiorgi, V. (1951). Nuclear disintegrations induced by µ-mesons. Phys. Rev. 84, 29–36. Da Silva, A., et al. (1995). Nucl. Instrum. Methods Phys. Res. A 353, 553. Dep, L., Elmore, D., Fabryka-Martin, J., Masarik, J., Reedy, R.C. (1994). Production rate systematics on in-situ cosmogenic nuclides in terrestrial rocks, Monte Carlo approach of investigating 35 Cl(n, γ )36 Cl. Nucl. Instrum. Methods Phys. Res. B 92, 321–325. Elster, J., Geitel, F. (1898). Versuche an Becquerelstrahlen. Ann. Phys. Chem. N.F. 66, 735–740. Glower, R.N., Watt, D.E. (1957). Philos. Mag. 2, 49. Goldbrunner, T., et al. (1997). Neutron activation analysis of detector components for the solar neutrino experiment BOREXINO. J. Radioanal. Nucl. Chem. 219, 293–297. Hashemi-Nezhad, S.R., Peak, L.S. (1995). Nucl. Instrum. Methods A 357, 524–534. Helbig, W., Görner, W., Niese, S. (1982). Background spectroscopy measurements using an underground laboratory. In: Povinec, P. (Ed.), Low-Level Counting. Proceedings of the Second Int. Conf. on Low Radioactivities ’80. VEDA, Bratislava, pp. 277–284. Helbig, W., Niese, S., Birnstein, D. (1984). Untergrundlaboratorium für die Messung geringer Radioaktivität. Isotopenpraxis 20, 60–64. Hess, V.M. (1912). Über Beobachtungen der durchdringenden Strahlung bei sieben Ballonflügen. Phys. Z. 13, 1084– 1091. Heusser, G. (1994). Background in ionizing radiation detection—illustrated by Ge-spectrometry. In: Garcia-Leon, M., Garcia-Tenorio, R. (Eds.), Proc. 3rd Int. Summer School on Low-Level Measurements of Radioactivity in the Environment: Techniques and Applications, Huelva, Spain, 1993. World Scientific, Singapore. Heusser, G. (1995). Low radioactivity background techniques. Annu. Rev. Nucl. Part. Sci. 45, 543–590. Heusser, G., Laubenstein, M., Neder, H. (2006). Low-level germanium gamma-ray spectrometry at the µBq/kg level and future developments towards higher sensitivity. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 495–510. Hirose, K., Aoyama, M., Igarashi, Y., Komura, K. (2005). Extremely low background measurements of 137 Cs in seawater samples using an underground facility (Ogoya). J. Radioanal. Nucl. Chem. 263, 349–353. Hubert, Ph., Dassie, D., Larrieu, P., Leccia, F., Mennrath, P., Chevallier, A., Henck, R., Gutknecht, D., Morales, R., Nunez-Lagoz, R., Morales, J., Villar, J.A. (1986). α-rays inducted background in ultra low level counting with Ge-spectrometers. Nucl. Instrum. Methods A 252, 87–90. Hult, M., Gasparro, J., Vasselli, R., Shizuma, K., Hoshi, M., Arnold, D., Neumaier, S. (2004). Deep underground measurements of 60 Co in steel exposed to the Hiroshima atomic bomb explosion. Appl. Radiat. Isotopes 61, 173–177. Kalin, R.M., Long, A. (1989). Radiocarbon dating with the Quantulus in an underground counting laboratory: Performance and background sources. Radiocarbon 31, 359–367. Kaye, J.H., Brauer, F.P., Conally, R.E., Rieck, H.G. (1972). Background reductions obtained with gamma detectors by use of massive cosmic-ray shielding. Nucl. Instrum. Methods 100, 333–348.
238
S. Niese
Kolb, W. (1992). Meßanleitung Umweltradioaktivität, 1. Lieferung: Gammasp/Nulleff . Gustav-Fischer-Verlag, Stuttgart–Jena. Komura, K. (1997). Challenge to detection limit of environmental radioactivity. In: Tsujimoto, T., Ogawa, Y. (Eds.), Proc. Int. Symp. on Environmental Radiation, Tsuruga, Japan, pp. 56–75. Komura, K., Inoue, M., Nakamura, N. (2002). Cosmogenic radionuclides in the recently fallen Kobe (CK4) meteorite. Geochem. J. 36, 333–340. Komura, K., Muguntha Manikandan, N., Yamaguchi, Y., Inoue, M., Abe, T., Murata, Y. (2006). Levels of airborne radionuclides at Hegura Island, Japan. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 554–561. Kovalchuk, E.L., Kuzminov, V.V., Pomansky, A.A., Zatsepin, G.T. (1975). Deep underground laboratory for lowradioactivity measurements. In: Povinec, P., Usacev, S. (Eds.), Low Radioactivities and Applications. Proc. Int. Conf. on Low Radioactivities ’75. SPN, Bratislava, pp. 23–27. Laubenstein, M., et al. (2004). Underground measurement of radioactivity. Appl. Radiat. Isotopes 61, 167–172. Lindström, R.M., Lindström, D.J., Slaback, L.A., Langland, J.K. (1990). A low-background gamma-ray assay laboratory for activation analysis. Nucl. Instrum. Methods A 299, 425–429. Loosli, H.H., Oeschger, H. (1982). Low-level gas proportional counting in an underground laboratory. In: Povinec, P. (Ed.), Low-Level Counting. Proceedings of the Second Int. Conf. on Low Radioactivities ’80. VEDA, Bratislava, pp. 117–125. Martineau, O., et al. (2004). Calibration of the EDELWEISS cryogenic heat-and-ionisation germanium detector for dark matter search. Nucl. Instrum. Methods A 530, 426–439. Müller, G., Wissmann, F., Schröder, F., Mondry, G., Brinkmann, H.J., Smend, F., Schumacher, M., Fettweis, B., Carchon, R. (1990). Low-background counting using Ge(Li) detectors with anti-muon shields. Nucl. Instrum. Methods A 295, 133–139. NCRP (1987). Report of the National Council of Radiation Protection and Measurement No. 94, Bethesda, MD, USA. Neder, H., Heusser, G., Laubenstein, M. (2000). Low-level γ -ray germanium spectrometer to measure very low primordial radionuclide concentrations. Appl. Radiat. Isotopes 53, 191–195. Neumaier, S., Arnold, D., Böhm, J., Funck, E. (2000). The PTB underground laboratory for dosimetry and spectrometry. Appl. Radiat. Isotopes 53, 173–178. Niese, S. (1996). Application of low-level counting techniques for the investigation of the impact of uranium mining as well as remediation on the environment. Appl. Radiat. Isotopes 47, 109–112. Niese, S. (2007). Measurement of fast neutrons above ground and in 47 m depth with a 3 He proportional counter. J. Radioanal. Nucl. Chem. 272, 173–176. Niese, S., Helbig, W. (1986). Detection limits in activation analysis using Ge(Li)-detectors installed in an underground laboratory. J. Radioanal. Nucl. Chem. 100, 155–163. Niese, S., Helbig, W. (1990). Cosmic induced radioactivity in the meteorite Trebbin. Freiberger Forschungshefte C 441, 90–93. Niese, S., Helbig, W., Kleeberg, H. (1989). Multi-sample beta–gamma coincidence spectrometry in an underground laboratory. J. Radioanal. Nucl. Chem. 129, 387–391. Niese, S., Loos, G., Gleisberg, B., Helbig, W. (1990). Determination of trace elements and cosmogenic radionuclides in meteorites by NAA and gamma spectrometry. Isotopenpraxis 26, 138–141. Niese, S., Krogner, K., Werner, C.D., Gleisberg, B. (1993). Radiochemical neutron activation analysis of ultramafic rocks from the Saxonian Granulite Massif. J. Radioanal. Nucl. Chem. 169, 81–92. Oeschger, H., Loosli, H.H. (1975). New developments in sampling and low level counting of natural radioactivity. In: Povinec, P., Usacev, S. (Eds.), Low Radioactivities and Applications. Proc. Int. Conf. in Low Radioactivities ’75. SPN, Bratislava, pp. 13–22. Oeschger, H., Beer, J., Loosli, H.H., Schotterer, U. (1981). Low level counting systems in deep underground laboratories. In: Methods of Low-Level Counting and Spectrometry, IAEA-SM 252/13. IAEA, Vienna, pp. 459–474. Plastino, W., Kaihola, L., Bartomolei, P., Bella, F. (2001). Cosmic background reduction in the radiocarbon measurements by liquid scintillation spectrometry at the underground laboratory of Gran Sasso. Radiocarbon 43, 157–161. Pomansky, A.A. (1986). Underground low background laboratories of the Baksan Neutrino Observatory. Nucl. Instrum. Methods B 17, 406–410. Povinec, P. (1972). Very low background proportional counter for tritium dating. Nucl. Instrum. Methods 101, 613.
Underground laboratories for low-level radioactivity measurements
239
Povinec, P. (1980). Proportional chambers for low level counting. Nucl. Instrum. Methods 176, 111–115. Povinec, P.P. (1994). Underground low level counting. In: Garcio-Leon, M., Garcia-Tenorio, R. (Eds.), Low-Level Measurements of Radioactivity in the Environment. Proc. Third Int. Summer School. World Scientific, Singapore. Povinec, P.P. (2004). Developments in analytical technologies for marine radionuclide studies. In: Livingston, H.D. (Ed.), Marine Radioactivity. Elsevier, Amsterdam, pp. 237–294. Povinec, P.P. (2005). Ultra-sensitive radionuclide spectrometry: Radiometrics and mass spectrometry synergy. J. Radioanal. Nucl. Chem. 263, 413–417. Povinec, P., Szarka, J., Usacev, S. (1979). Multielement proportional counters for low-level 3 H and 14 C counting. Nucl. Instrum. Methods 163, 363–369. Povinec, P., Sitár, B., Kladiva, E., Seman, M., Budagov, Yu.A., Semenov, A.A. (1990). A time projection chamber for double-beta-decay experiments. Nucl. Instrum. Methods A 293, 562–568. Povinec, P.P., la Rosa, J.J., Lee, S.H., Osvath, I., Wyse, E. (2001). Recent developments in radiometric and mass spectrometry methods for marine radioactivity measurements. J. Radioanal. Nucl. Chem. 248 (3), 713–718. Povinec, P.P., Comanducci, J.-F., Levy-Palomo, I. (2004). IAEA-MEL’s underground counting laboratory in Monaco —background characteristics of HPGe detectors with anti-cosmic shielding. Appl. Radiat. Isotopes 61, 85–94. Povinec, P.P., Comanducci, J.F., Levy-Palomo, I. (2005). IAEA-MEL’s underground counting laboratory (CAVE) for the analysis of radionuclides in the environment at very low-levels. J. Radioanal. Nucl. Chem. 263, 441–445. Reyss, J.-L., Schmidt, S., Legeleux, F., Bonte, P. (1995). Low background well-type detectors for measurements of environmental radioactivity. Nucl. Instrum. Methods A 357, 391–397. Rindi, A., et al. (1989). Nucl. Instrum. Methods Phys. Res. A 267, 529. Schopper, E., Lohrmann, E., Mauck, G. (1967). Nukleonen in der Atmosphäre. In: Flügge, S. (Ed.), Handbuch der Physik, Bd. 46/2. Simpson, J.A., Fonger, W., Freimann, S.B. (1953). Phys. Rev. 90, 934–950. Sitte, K. (1961). Penetrating showers. In: Flügge, S. (Ed.), Handbuch der Physik, Bd.46/1, Kosmische Strahlung. Springer, Berlin, pp. 181–194. Smekal, F.G. (1968). Österreichs Nobelpreisträger. Frick, Vienna. Tanaka, S., Sakamoto, K., Takagi, J. (1967). An extremely low-level gamma ray spectrometer. Nucl. Instrum. Methods 56, 319–324. Theodorsson, P. (1992). Measurement of Week Radioactivity. World Scientific, Singapore. Torres, R., Hubert, Ph. (1996). LSM, Bull Inform. Utilisateures No. 40. Verplancke, J. (1992). Low level gamma spectroscopy: low, lower, lowest. Nucl. Instrum. Methods A 312, 174–182. Vojtyla, P., Povinec, P.P. (2000). A Monte Carlo simulation of background characteristics of low-level HPGe detectors. Appl. Radiat. Isotopes 53, 185–190. Webber, W.R. (1967). The spectrum and change composition. In: Flügge, S. (Ed.), Handbuch der Physik, Bd.46/2, Kosmische Strahlung. Springer, Berlin, pp. 181–194. Wordel, R., Mouchel, D., Bonne, A., Meynendonckx, P., Vanmarcke, H. (1993). Low level gamma-ray measurements in a 225 m deep underground laboratory. In: Garcio-Leon, M., Garcia-Tenorio, R. (Eds.), Low-Level Measurements of Radioactivity in the Environment. Proc. Third Int. Summer School. World Scientific, Singapore. Yamakoshi, K., Nogami, K. (1976). An extremely low-background beta–gamma coincidence spectrometer. Nucl. Instrum. Methods 134, 519–524. Yamashita, M., Stephens, L.D., Patterson, H.W. (1966). Geophys. Res. 71, 3817. Zdesenko, Yu.G., et al. (1988). In: Underground Physics ’87: Proc. of the 2nd Int. Symp. Nauka, Moscow, p. 291. Ziegler, J. (1998). Terrestrial cosmic ray intensities. IBM J. Res. Dev. 42 (1).
This page intentionally left blank
241
Accelerator mass spectrometry of long-lived light radionuclides A.J. Timothy Jull∗ , George S. Burr, J. Warren Beck, Gregory W.L. Hodgins, Dana L. Biddulph, Lanny R. McHargue, Todd E. Lange NSF Arizona AMS Laboratory, the University of Arizona, Tucson, AZ 85721-0081, USA Abstract Many different kinds of paleoclimatic, geological and archaeological records can be characterized by measuring their radionuclide concentrations using accelerator mass spectrometry (AMS). The purpose of this paper is to highlight some applications of AMS, using studies conducted at the Arizona AMS Facility as examples. These include studies of 14 C, 10 Be, 26 Al, and 129 I. The work can be generally divided into two types: (1) methodological studies designed to refine and improve the capabilities of AMS, and (2) studies which utilize radiogenic isotopes as geochronometers or as geochemical tracers. Studies of the first type include the development of our 26 Al measurement capabilities, the construction on an automated sample preparation line and the construction of a plasma oxidation line. Studies of the latter type include 14 C dating of corals, speleothems and bones; new records of 10 Be from marine sediments and extraterrestrial materials; and 129 I studies of the pathways of this isotope in the surface ocean.
1. Introduction Since the first work on radiocarbon dating by decay counting (Arnold and Libby, 1949; Libby, 1955), the field of radionuclide studies has expanded widely both as a dating tool and as a tracer. A rapid expansion of applications to much smaller samples became feasible with the development of accelerator mass spectrometry in the late 1970s (Tuniz et al., 1998). Since then, AMS machines have evolved from the large accelerators (∼10 MV) and cyclotrons in early 1980 to much smaller accelerators (less than 1 MV) recently designed specifically for AMS. Commonly measured radionuclides by AMS in the above research fields are 3 H, 14 C, 10 Be, 26 Al, 36 Cl, 41 Ca, and 129 I. Radionuclide measurement using AMS differs from earlier decay-counting methods in that the amount of the radionuclide in the sample is measured directly, rather than waiting for individual radioactive decay events to occur (Elmore and Phillips, 1987; Jull, 2007). For 14 C, this means greatly enhanced sensitivity. For longer-lived radionuclides, such as 26 Al, 10 Be, 36 Cl, and 129 I, the analytical advantages of AMS are even greater. The sensitivity is achieved by ionizing sample atoms efficiently, accelerating them to high energies ∗ Corresponding author. E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11007-X
© 2008 Published by Elsevier B.V.
242
A.J.T. Jull et al.
Fig. 1. A diagram of 3 MV Pelletron accelerator mass spectrometer at the University of Arizona.
and counting individual isotopes using particle detection techniques developed for nuclear physics. Longer-lived radionuclides which were very difficult to measure using decay counting techniques are now routinely measured with AMS (Tuniz et al., 1998; Fifield, 1999; Jull and Burr, 2006). A diagram of an AMS instrument, the 3 MV AMS machine at the University of Arizona, is shown in Figure 1. 1.1. Radiocarbon dating by accelerator mass spectrometry For 14 C the AMS technique is 1000 to 10,000 times more sensitive than decay counting techniques. Today, an external precision of about ±0.35% in 14 C content, or ±30 years in uncalibrated radiocarbon age is possible on a single 0.5-mg-sized sample target in 20 min of measurement time. Samples as small as 100 µg or less have been successfully dated to about ±80 years BP and even smaller samples (∼10–20 µg) have been measured for special experiments (e.g., Jull et al., 1998). With longer counting times or when multiple targets are measured, we can reduce the single target error to about 0.2%, or better than ±20 years in radiocarbon age (McNichol et al., 2001; Donahue et al., 1990a, 1990b). This increased sensitivity allows us to reduce sample size so as to preserve invaluable archaeological samples or for biomedical tracer studies.
Accelerator mass spectrometry of long-lived light radionuclides
243
1.2. Measurement of radiocarbon In AMS radiocarbon studies, the material of interest is converted into graphite after cleaning the sample, combustion to carbon dioxide, and catalytic conversion to graphite, as described later in this paper. The graphite powder can then be pressed into a sample target holder, which is then loaded into the AMS machine’s ion source, shown on the left-hand side of Figure 1. The graphite is bombarded with Cs+ ions under vacuum, and the sputtered ions include many C− ions. We extract the C− ions from the ion source to ground by placing the ion source at a negative potential of about −50 kV. The accelerated C− ions are then deflected in a 90◦ magnet, which allows us the first separation of ions of different mass. Because 14 C has a modern natural abundance of 10−12 , further separation is needed. Hence, we accelerate C− ions to 2.5 MV in the accelerator (at the center of Figure 1) and these ions then pass through a gas canal (often called the “stripper canal”), where they lose electrons and become positively charged. The C3+ ions are then accelerated out of the machine and pass through electrostatic filters (to select ions of one energy (E)/charge (q) ratio) and magnetic elements (to select ions of the same ME/q 2 ). The separated ions of 14 C3+ can be completely distinguished from any other ions in the detector, in part because 14 N− ions are unstable (Elmore and Phillips, 1987). 1.3. Recent developments in AMS A recent trend in the AMS field has been the development of increasingly smaller AMS instruments (Hughey et al., 1997, 2000; Suter et al., 1997, 2000). Currently, a “small” AMS is one which has a maximum terminal voltage of about 1 MV or less. These machines require less space and resources than larger AMS instruments, and are adequate for standard radiocarbon measurements. The principal disadvantage of small AMS machines is that their relatively low terminal voltages prohibit the use of stripping to high charge states (> ∼2+), where filtering techniques used in nuclear physics are most efficient. A number of 0.5 MV machines have now been built and are operational at the University of Poznan, University of Georgia and University of California–Irvine, with additional machines under construction. Recent efforts to go to even lower terminal voltages, perhaps as low as 200 kV, are also underway. If successful, these studies could mean the elimination of costly accelerators. A prototype design of such a machine is under construction in Zürich (Synal et al., 2006). In general, as the operating terminal voltage used for AMS decreases, the technical engineering difficulties increase. At present, the lower limit for the terminal voltage of AMS machines manufactured commercially remains at 0.5 MV. There have been a number of studies highlighting the potential use of these small AMS machines for applications to 10 Be (Grajcar et al., 2004), 129 I (Synal et al., 2006), and Pu (Fifield et al., 2004) measurements.
2. The NSF Arizona Accelerator Mass Spectrometry Laboratory The Arizona facility was established 26 years ago as one of the first purpose-built laboratories for radiocarbon AMS. The laboratory began measuring 10 Be soon after its inception, and now produces 26 Al and 129 I measurements as well. Jull and Burr (2006) and Jull et al. (2006)
244
A.J.T. Jull et al.
Fig. 2. Fields of study for AMS analyses made at the laboratory in 2005. Data compiled according to the affiliation of the submitter.
describe the history, current status, and future direction of the Arizona AMS research program and for the field as a whole. 2.1. AMS instrumentation At the Arizona laboratory, we operate 2 AMS instruments. Our first machine, a 2.5 MV General Ionex Tandetron, was installed in 1981, after the first dedicated AMS in Oxford. The second, a 3 MV NEC Pelletron AMS, was installed in 2000 (see Figure 1). Recently, we have completely overhauled the General Ionex Tandetron, including installation of new HVEE accelerator tubes. The General Ionex instrument is used exclusively to measure 14 C. The NEC machine is used for 14 C, 10 Be, 26 Al, and 129 I. This AMS instrument was designed to measure a variety of isotopes, and features a large high energy analysis magnet, sufficient to bend any element on the periodic table, with multiple detectors (surface barrier, gas ionization chamber, time of flight) to assist with filtering a variety of interfering isobars. 2.2. Service analysis program Since 2001, the Arizona AMS facility has provided between 4500 and 6000 sample measurements per year, surpassing the previous 5-year total by more than 25%. In 2005 we produced 5563 radiocarbon measurements. These results were generated in support of 265 individuals from over 100 universities, 27 government laboratories and a dozen museums in the U.S. and abroad. Figure 2 summarizes the many fields which use radiocarbon in their research. This list reflects the emphasis on Earth Sciences and Archeology, and the variety of other fields which rely on 14 C.
3. Sample pretreatment and processing Many AMS laboratories are pursuing active research programs to improve and refine sample pretreatment and processing techniques. In the case of radiocarbon dating, the goal of this work is to isolate material which best reflects the time of formation, free of contaminants and altered material. These techniques usually involve physical and chemical cleaning, and may include sophisticated chemical extraction techniques. A second avenue of research is focused
Accelerator mass spectrometry of long-lived light radionuclides
245
on automation and miniaturization, especially useful to tracer studies. Some examples from the Arizona laboratory are given below. 3.1. Radiocarbon sample processing techniques Some current procedures for AMS radiocarbon sample processing are summarized below (after Jull et al., 2003, 2004): (1) Acid–base–acid (ABA) method for charcoal, wood, cellulose, plant material, animal tissue: After physical inspection, samples are cleaned with 1 N HCl acid, 0.1% NaOH and 1 N HCl, washed with distilled water, dried, and combusted at 1000 ◦ C with CuO. Hatté et al. (2001) discussed some modifications which can be employed to overcome potential problems with the acid–base–acid method. (2) Selective dissolution for carbonates: In general, samples are partially dissolved with 100% H3 PO4 to remove surface carbonate. More than 50% of the total carbonate is typically removed, under a vacuum. The remaining carbonate is then acidified with H3 PO4 to collect the remaining CO2 , as discussed by Burr et al. (1992). (3) Selective combustion for sediments: After cleaning in 1 N HCl and drying, the sample is combusted at 400 ◦ C in ∼0.3 atm oxygen gas (McGeehin et al., 2001, 2004). (4) Oxidative acid cleaning for old charcoal: We have constructed a new line to clean charcoal samples using the oxidative acid method of Bird et al. (1999). This is of particular interest for charcoal samples >20 kyr (Bird et al., 2003; Bird, 2007). (5) Solvent extraction for textiles, parchment, canvas, art works and artifacts: The samples are given the ABA pretreatment and after washing and drying, they are Soxhlet extracted with hexane, then ethanol and finally methanol. After washing in distilled water, and drying, they are combusted at 900 ◦ C with CuO. Bruhn et al. (2001) proposed a more complex, 6-stage cleaning procedure for art works. After cleaning and combustion, samples are converted to graphite using Fe catalyst and Zn as the reducing agent for the reaction CO2 + 2Zn → 2ZnO + C (on iron). The progress of this reaction is monitored by measuring the gas pressure of the reaction chamber. A zero gas pressure indicates the quantitative conversion of CO2 to graphite, and insures that no isotopic fractionation has occurred. As mentioned previously, the graphite produced from the sample is pressed into a target holder for analysis. The measurement of 14 C follows the procedures and calculations described in detail by Donahue et al. (1990a, 1990b) and McNichol et al. (2001). 3.1.1. Plasma oxidation sample preparation system A new development in our laboratory is a refinement of the plasma oxidation method first proposed and tested by Marvin Rowe and colleagues at Texas A&M University. The method uses a low temperature oxygen plasma to selectively oxidize paint and pigments from rock surfaces (Russ et al., 1992). We have completed construction of a plasma oxidation chamber for carbonaceous samples (Jones et al., 2005). The method utilizes a capacitively coupled RF power source to produce excited oxygen species within a low-temperature (50 ◦ C) plasma (Figure 3). The potential advantage of the technique is the capability to oxidize thin films or selective carbon species from the sample surface. Possible applications include: the oxidation
246
A.J.T. Jull et al.
Fig. 3. Argon plasma generated by the plasma oxidation unit.
Fig. 4. Preparing bone samples for automated chemical treatment.
of soil organic matter, selective oxidation of soot, and oxidation of surface layers in a sample.
14 C
profiles generated by successive
3.1.2. Automated sample processing Due to the high throughput of samples, there is great pressure to automate repetitive tasks in the laboratory (Figure 4). The Arizona laboratory has constructed a continuous flow automated CO2 trapping system capable of processing up to 40 samples per day. The device is computer controlled, and interfaces with a mass spectrometer, CHN elemental analyzer and Gilson robotic carbonate hydrolysis device. The device not only allows for automated sample processing, but adds the capability of immediate δ 13 C analysis and CHN analysis. These data can be used to assess the degree of sample preservation or to help interpret sample results.
Accelerator mass spectrometry of long-lived light radionuclides
247
4. Radiocarbon studies 4.1.
14 C
calibration
The radiocarbon age is defined from the amount of 14 C relative to “modern carbon”, defined as 1950AD wood, where the industrial effect of reduced 14 C has been removed. Effectively this is 1850AD wood age-corrected to 1950AD. Using these assumptions, the radiocarbon age may be computed as Radiocarbon age = −8033 ln(14 Csample /14 Cmodern ). This equation assumes a “Libby” half-life of 5568 years, which is known to be incorrect. We also recognize that the production rate of radiocarbon in the atmosphere changes with time, and that temporal changes in atmospheric radiocarbon are associated with climatic factors. These effects cause radiocarbon ages to diverge from calendar ages. Such variations are all accounted for by calibrating to samples of known age. Libby (1955) was the first to note that radiocarbon ages sometimes diverged from the “true” calendar age with his famous “Curve of Knowns”. Since about 1980, a vast amount of time and energy has been devoted by scientists in the radiocarbon field to calibration of the radiocarbon time-scale (Klein et al., 1982). This eliminated a growing discord of different methodologies for calibration of radiocarbon dates (Stuiver and Kra, 1986; Stuiver et al., 1993; Stuiver and van der Plicht, 1998; Reimer, 2004). 4.1.1. Tree ring chronologies The radiocarbon “calibration curve” was originally established by studying the changes in 14 C content of known-age tree rings. The first trees studied were Bristlecone Pines from the White Mountains of California (Figure 5). Tree ring 14 C measurements from living and dead wood from these long-lived trees were cross-correlated to establish a 14 C chronology longer than the life of a single tree. Dendrochronology, the science of correlating and cross-referencing variations in the widths of tree-rings has been used to produce even longer chronologies using German and Irish Oak tree records, incorporating sub-fossil wood recovered from buried logs in river sediments and peat bogs. Currently, the continuous tree ring sequence extends back 12,400 years. 4.1.2. Extension of the tree ring record There have been continuing improvements in the length of the radiocarbon time-scale. The improvement in the radiocarbon calibration curve over the last 26,000 yr has allowed us to cross-correlate fluctuations in the 14 C curve directly with those in the ice-core record (Reimer et al., 2004). This capability has improved attempts to cross-correlate different climatic events observed in one record with other proxy records. This extension of the calibration curve uses tree rings to about 12,400 calibrated years and beyond that used corals and varved marine sediments. There are also newer but as yet less well-established records which should take us back to the limits of radiocarbon dating, using lake sediments and speleothem records (van der Plicht et al., 2004). The fluctuations in the calibration curve demonstrate that changes in the 14 C production rate and/or changes in the terrestrial carbon cycle have occurred. Intriguingly, it has often
248
A.J.T. Jull et al.
Fig. 5. A Bristlecone Pine tree growing in the White Mountains of California.
been noted that reversals (negative excursions to younger apparent age) are associated with cold climatic events and lowered carbon content of the atmosphere. This suggests a possibility of linkages to climate driven by carbon-cycle changes, such as might be produced by ocean circulation changes. They could also, of course, be due to a higher production rate. Excursions to higher 14 C age are likely due to reductions in production rate or increases in the ventilation of 14 C-depleted carbon from the deep ocean. Conversely, excursions to lower 14 C age are likely due to reductions in production rate or attenuated ocean ventilation rates. Periods of constant 14 C over an extended period of time can also be explained by reductions in production rate. 4.1.3. INTCAL98 and INTCAL04 An extended tree ring curve was first published in 1998, after the success of the initial curves published in 1986 and 1993. This curve included U–Th dated coral records from the Atlantic and Pacific, and varved marine sediment records from the Cariaco Basin (Stuiver and van der Plicht, 1998). The International Calibration team (known as IntCal) established rigid protocols for data integrity for inclusion in the calibration database. In 2004 the IntCal04 calibration dataset was introduced, and is currently incorporated into a number of computer programs, such as Calib 5.1 (http://www.calib.org), for dates up 26,000 calendar years BP (cal. BP). Because of discrepancies between various records beyond that time, as will be discussed below, it was decided at the 2003 Radiocarbon Conference business meeting in Wellington, New Zealand, to officially recognize a record up to 26,000 cal. yr BP until these questions were
Accelerator mass spectrometry of long-lived light radionuclides
249
Table 1 Records included in the INTCAL database Record
Time period (cal yr BP)
Reference
German Oak and Pine Irish Oak U.S. west coast conifers Corals
0–12,410 120–7160 0–2099 12,400–26,000
Cariaco Basin varved sediments
10,500–14,700
Reimer et al. (2004) Reimer et al. (2004) Stuiver and van der Plicht (1998) Bard et al. (1990) Edwards et al. (1993) Bard et al. (1998) Burr et al. (1998) Burr et al. (2004) Cutler et al. (2004) Fairbanks et al. (2005) Hughen et al. (2004)
resolved (see Table 1). The records eventually included in the most recent INTCAL04 volume (Radiocarbon, vol. 46, No. 3) include tree ring data, coral data (e.g., Bard et al., 1990, 1993, 1998; Burr et al., 1998, 2004) and varved sediments from the Cariaco Basin (Hughen et al., 1998). The decision to include the Cariaco sediment record in the INTCAL98 calibration was explained by Stuiver and van der Plicht (1998) as follows: “In addition, as an exception to the rule, it was decided to include Late Glacial marine varves because this newly developed data set strengthened the coral/tree ring link considerably”. An example of a section of the calibration curve is given in Figure 6. Coral records appear to have wide acceptance as proxy records of atmospheric 14 C at a given time, but even this might be a problem. Since corals sample bicarbonate in the surface ocean, not the atmosphere directly, one needs to make a reservoir correction, often assumed to be around 400 yr (e.g., Fairbanks, 1989)—some more recent studies such as Fairbanks et al. (2005) and the compilation of Hughen et al. (2004) make site-specific reservoir corrections. However, an unsolved problem is whether even site-specific corrections can be assumed to be constant in the Pleistocene, and, if not, how much error this might create. Recent discussion at the 10th International Conference on Accelerator Mass Spectrometry (Berkeley, CA) in 2005 and the 19th International Radiocarbon Conference in Oxford, April 2006 demonstrate that reservoir variations are especially significant for intermediate and deep waters, especially at high latitudes. Coral records offer the highest possible resolution for radiocarbon calibration beyond the limit of the tree-ring chronologies, because of their relatively fast growth rate. However, the existing radiocarbon calibration contains few continuously banded examples from corals. Two exceptions are a Diploastrea coral head from Vanuatu which lived during the Younger Dryas (Burr et al., 1998) and a Goniastrea coral from Papua New Guinea which lived approximately 13,000 years ago (Burr et al., 2004), according to U–Th dates. These two records have resolution and precision comparable to tree ring records. 4.1.4. Other calibration records Many other proxy records have been studied, with the intent of extending the radiocarbon calibration curves. These included studies of marine sediment records (e.g., Voelker et al.,
250
A.J.T. Jull et al.
Fig. 6. Data for the last 50,000 yr, excluding the new results of Fairbanks et al. (2005). The radiocarbon calibration up to 26,000 yr was agreed at the 18th International Radiocarbon Conference in Wellington, in 2003. Hughen et al. (2004) and Fairbanks et al. (2005) discuss various possible extensions to this curve.
2000), lake sediment records (Kitagawa and van der Plicht, 1998), speleothems (Beck et al., 2001, 2005, 2006) Lake Lisan sediments (Schramm et al., 2000), and records of Atlantic foraminifera (Bard et al., 2004). However, these records are generally not considered to be good enough for calibration and are referred to as “comparison/calibration records” by INTCAL (see van der Plicht et al., 2004). Recently, Fairbanks et al. (2005) have produced a new coral record, which has the potential of pushing back the calibration to 50,000 yr (Figure 7). This new record will be updated frequently by these authors and may provide the basis of an improved calibration back to 50,000 yr in future. However, some work still needs to be done to integrate the various datasets beyond 26,000 yr. Speleothems offer similar advantages to coral, with a time resolution of decades or better. One such record, obtained from a stalagmite recovered from a cave in the Bahamas, provides a nearly continuous record of atmospheric 14 C from 45 to 11 ka (Beck et al., 2001; Richards et al., 2006). This record (Figure 8) derived using TIMS U, Th, and Pa measurements and AMS 14 C ages reveals highly elevated and extremely variable 14 C between 45 and 33 ka BP which appear to be correlative with peaks in cosmogenic 36 Cl and 10 Be isotopes (Raisbeck et al., 1990; Baumgartner et al., 1998) observed in polar ice cores.
Accelerator mass spectrometry of long-lived light radionuclides
251
Fig. 7. New comparison curve of radiocarbon versus U–Th ages from Bahamas and other corals, from Fairbanks et al. (2005), showing the good agreement between corals and tree-ring dates in the period up to 15,000 yr.
4.2. Paleoclimate studies In order to correlate distinct climatic features, it is important to be able to correlate independently dated events. The improvement in the radiocarbon calibration curve over the last 26,400 yr (and possibly to 50,000 yr) allows to cross-correlate fluctuations in the 14 C curve with climatic fluctuations in such things as ice-core records (Reimer et al., 2004; Fairbanks et al., 2005), as we have already discussed. We should also note that the same event might have a phase lag in different regions, or a different expression. This is particularly true during the Glacial/Interglacial transition, which is of great interest due to the scale of climatic change at that time (e.g., Markgraf, 2001). During the Holocene, we also observe appreciable climatic fluctuations. These are less well understood, but may be associated with solar forcing (Damon and Sonnett, 1991). Other periodicities in the Holocene climatic record can often be related to solar fluctuations, the most obvious being in the medieval warm period and the Maunder minimum, periods associated with colder weather in Europe. There is a variety of literature on this subject. Millennial-scale periodicities (e.g., Bond et al., 1997; Alley et al., 1999, 2001) have been recognized in a number of records, including varved lake sediments, loess deposits, marine records and forest-fire records. In this paper, we will highlight some climatic signals, which can be well dated using the small-sample capabilities of accelerator mass spectrometry
252
A.J.T. Jull et al.
Fig. 8. Speleothem results of 14 C age versus calendar or U–Th age for a variety of reservoirs (adapted from Beck et al., 2001).
(AMS). These signals can be seen not only in the climate record but affect the extinction of mega fauna as well as archaeological events. 4.3. Forest fire studies The study of forest fires is potentially a good proxy for changes in climate. We are currently studying a sedimentary fan at Bear Flat, British Columbia that potentially contains a forest fire record of the last 36,000 years. Dates from charcoal collected at the site indicate periodicity of large fires through the Holocene, with an average of four fires per thousand years. We conclude that the observed fire frequency is related to large-scale climatic conditions over a millennial time scale (Jull and Geertsema, 2006; Sanborn et al., 2006). There is a considerable record of forest-fire history from different regions of the world (see, for example, Meyer et al., 1995, 2001; Hallett and Walker, 2000; Hallett et al., 2003; Turcq et al., 1998; Lertzman et al., 2002; Pierce et al., 2004). A characteristic of many of these studies is evidence for marked periodicities, especially on century and millennial time scales of fire frequency. Meyer et al. (1995) demonstrated a spectacular record of forest fires from the Yellowstone National Park, which is now cited in all subsequent papers. In this record, Meyer et al. (1995) noted millennial-scale forcing and proposed that “Bond” cycles might be a forcing mecha-
Accelerator mass spectrometry of long-lived light radionuclides
253
nism. Later studies showed similar periodicities, but a different phase relation at a location in southern Idaho (Pierce et al., 2004). Recently, Jull and Geertsema (2006) reported on radiocarbon dating of charcoal from paleosols and buried charcoal horizons in a unique sequence which potentially records the last 36,000 14 C years from a fan at Bear Flat, British Columbia. This site included evidence for forest fire charcoal is found over the last ca. 13,500 radiocarbon yr before present (yr BP) or 16,250 ± 700 calibrated yr BP. The latest evidence of fire is during the Medieval Warm Period. Over 50 discrete fire-related horizons have been observed. These charcoal ages show a periodicity in large fires on a millennial scale through the Holocene—an average of four fires per thousand years and fire frequency appears to be related to climate. Alley et al. (2001) noted that forest fire frequency increases during North Atlantic cold events. Other studies in British Columbia have been undertaken with colleagues from the University of Northern British Columbia from sites in a wetter region of central British Columbia (e.g., Sanborn et al., 2006). 4.4. Studies of extinct megafauna Radiocarbon dates were obtained by accelerator mass spectrometry on bones of extinct large mammals recovered from tar pits. Results on some samples of Glyptodon and Holmesina (extinct large mammals similar to armadillos) yielded results of >25 and >21 ka, respectively. We also studied the radiocarbon ages of 3 different samples of bones from the extinct Cuban ground sloth, which appear to have survived on some islands of the West Indies (Caribbean Sea) until about 4400 yr ago (Steadman et al., 2005). 14 C dating of a tooth from a Madagascar Hippopotamus confirms that this member of the extinct “prehistoric” megafauna actually survived until well after European colonization. Gelatinized collagen subsamples independently dated by two AMS facilities yielded ages of 99±36 and 214±40 yr BP. Calibrated at 2σ , these dates give a range of 1639–1950 cal yr BP, placing a member of this extinct megafauna securely in colonial times. Dates on other extinct taxa show that many other prehistoric megafauna broadly overlapped the human presence on the island and may have survived until colonial times as well. A summary of all the available dates has been discussed by Burney et al. (2004). Apart from some isolated evidence for earlier settlement with good associations, the predominant view is that early man arrived in the western hemisphere close to the end of the last Glacial (see Nuñez et al., 1994; Haynes, 1984, 1991, 1992; Martin and Klein, 1984; Meltzer et al., 1997). The conventional model assumes that early humans arrived in the new world via a Bering land bridge. We note that since the Bering Strait is only 30 m deep at its shallowest point and we can assume that the last sea-level rise would have closed off this route between Asia and the Americas. The Bering land bridge should have remained intact until ∼10,000 radiocarbon years BP (11,000 calendar years). We can understand this in terms of the sea-level rise history, which occurred in two stages of ∼60 and ∼50 m, as shown by Bard et al. (1990) and Edwards et al. (1993). The rapid expansion of early man into central North America does not appear to have occurred until about 12,000 radiocarbon years BP, although the dating of the Monte Verde site in Chile (∼12,500 14 C yr BP; Meltzer et al., 1997) suggests a slightly older time.
254
A.J.T. Jull et al.
This period appears to be during the “Older Dryas”, which is observed in many records ∼12,200 14 C yr BP (e.g., Goslar et al., 2000). Dyke et al. (2001) have summarized the available radiocarbon records for the margins of the Laurentide ice sheet, which indicate that regions east of the Rocky Mountains were ice free at that time. In addition, the results of Jull and Geertsema (2006) and others suggest that parts of this region may have been ice-free even earlier. The extinction of many megafauna at the end of the late Pleistocene is well known (Martin and Klein, 1984). Hence, the exact time of these extinctions, and whether they are caused by climate alone, or by a combination of factors, such as the expansion of humans into previously unoccupied areas is a matter of great interest. Indeed, neither of these factors alone seems to explain all the observations. Taylor (2000) has summarized the earliest radiocarbon dates on human bones from the new world. In all cases, these dates are close to the dates for the last evidence for mammoths and other large megafauna (e.g., Martin and Klein, 1984; Steadman et al., 2005). About one half of all species of large land mammals of North America disappeared at the close of the Pleistocene (Martin and Klein, 1984). 4.5. Mass mortality of coral reefs Recognition of mass mortality on ancient coral reefs provides a context for evaluating the frequency, magnitude, and causes of mass mortality on living reefs. We are working in collaboration with John Pandolfi (Queensland University, Australia) and others to establish a detailed chronology for several regional mass mortality events which occurred during the early to mid Holocene in Papua New Guinea. The most striking of the mortality events, extending over 16 km along the ancient coastline, is dated at ∼9100–9400 cal year BP, and is associated with a volcanic ash horizon. The subsequent recolonization of the reef surface is recorded in numerous exposed reef sections along the Huon peninsula. This situation provides an exceptional opportunity to assess the relative importance of mass mortality disturbance events in shaping the benthic community structure on coral reefs. 4.6. In situ-produced cosmogenic radionuclides A new methodology incorporating AMS measurements is the use of cosmogenic radionuclides produced in situ at the surface of the Earth by interactions of cosmic radiation with the silicate in surface rocks (Gosse and Phillips, 2001). Measurements of radionuclides produced in situ in the surfaces of rocks, soils can be used to estimate exposure age and erosion rates. This is particularly true for the nuclides 10 Be, 14 C, 26 Al, and 36 Cl. These methods can be applied to landscape evolution variability, such as weathering, sediment transport and soil development, retreat and advance of glaciers, tectonics, volcanic flows, meteorite impacts and other phenomena. This method relies on the time of exposure of a sample near the surface of the Earth, where it will be exposed to significant cosmic radiation. Higher-altitude samples receive more exposure. A limitation of these methods is the need to be concerned about various spatial corrections, that is, the location of the sample as a function of latitude, altitude and partial shielding by surrounding geological features can affect the results. These applications have been summarized by Gosse and Phillips (2001), as well as by Cockburn and Summerfield (2004).
Accelerator mass spectrometry of long-lived light radionuclides
255
Lifton et al. (2001) have been successful is extracting in situ 14 C produced in rock surfaces and applying the same methodology to these studies. 14 C is particularly useful for evaluating the amount of erosion on surfaces, and potentially for archaeological applications. International collaboration on cosmogenic radionuclides Recently, a project has been funded by both the U. S. National Science Foundation and the European Union to produce the “baseline” information necessary to obtain precision dates using these methods. This program, called CRONUS, anticipates that if we can define production rates, scaling parameters and other components of the calculations, such as cross sections, then we have the possibility of obtaining ±5% measurements of age for in situ cosmogenic radionuclides. These methods will then select primary geologic calibration sites, to provide the “ground truth” studies for these measurements. An important task in the in situ field is the improvement of standard sites and reference locations. The development of in situ 14 C is most important to future understanding of earthsurface processes. In a detailed paper, Lifton et al. (2001) showed that in situ 14 C could be used to date the ages of the Bonneville shorelines in Utah and also showed the method was consistent with other dating estimates of these surfaces. It is also important to recognize the importance of muon reactions, which can produce cosmogenic products at a considerable depth in the Earth’s surface (e.g., Braucher et al., 2003; Heisinger et al., 2002). The Lake Bonneville shorelines will be the focus of the international cosmic-ray produced nuclide systematics on Earth (CRONUS) program and the first “intercomparison” site. We have completed field work in which we collected samples from surfaces at secular equilibrium along mid- and low-latitude altitude transects to assess the altitudinal and latitudinal dependence of integrated late Quaternary in situ 14 C production rates. We also have received samples from other investigators from Namibia, Antarctica, Australia and New Zealand to better constrain in situ 14 C production rate scaling globally. Results to date from these transects (ranging from sea level to nearly 4 km altitude) confirm the viability of using in situ 14 C in saturated surfaces to constrain in situ cosmogenic nuclide production rate scaling models (Lifton et al., 2002; Pigati and Lifton, 2004). Future CRONUS field sites, both for the US and EU programs, include the Isle of Skye— for a wide range of different lithologies—glacially polished surfaces in the northeastern USA, lava flows from Hawaii and Mediterranean and Atlantic islands.
5.
10 Be
10 Be
studies
is produced by spallation reactions between atmospheric oxygen and nitrogen and cosmic radiation. The intensity of the cosmic ray flux depends on galactic and solar sources, and modulation by the heliomagnetic and geomagnetic fields (Raisbeck et al., 1987, 1990). After formation, 10 Be is quickly removed from the atmosphere by precipitation and deposited onto the surface of the Earth, where it is transported throughout the ocean and is eventually sequestered within marine sediments. A record of the cosmic ray flux, modified by marine processes, may be interpreted from marine sediment cores and provides a valuable record of past geomagnetic and cosmic-ray phenomena.
256
A.J.T. Jull et al.
Fig. 9. Correlation of 10 Be and δ 18 O observed in marine sediment from the Blake Outer Ridge (DSDP site 72) showing a climatic effect on the 10 Be record in addition to cosmogenic effects.
Our previous work documented the 10 Be record in the authigenic fraction of marine sediments from the Gulf of California, Leg 64, site 480 (McHargue et al., 1995), and the Blake Ridge, CH88-10P (McHargue et al., 2000). Recent work was summarized by McHargue and Donahue (2005). 10 Be was found to correlate well with the paleomagnetic field as measured from the same sediments of the Blake Outer Ridge at CH88-10P (Schwartz et al., 1998). A longer 10 Be record measured from a nearby core demonstrated that climatic effects can influence 10 Be deposition (Figure 9). The climatic interference on 10 Be concentrations in marine sediments can be minimized by normalization of 10 Be to the mass of the authigenic fraction, as opposed to the total mass, or to 9 Be (McHargue and Donahue, 2005). 10 Be
in lake sediments Some new projects highlight progress to measure 10 Be in lake sediments. These are from Lake Malawi, and Lake Bosumtwi, in Africa and Lake Qinghai (China). The possibility that 10 Be can be used to date to lake sediments arose from the observation that in marine sediments the minimum concentrations of 10 Be in the overall data set are similar due to comparable attenuation of its production in the atmosphere when the intensity of the geomagnetic field is near maximum.
Accelerator mass spectrometry of long-lived light radionuclides
6.
26 Al
257
and 129 I studies
At our laboratory, we have developed 26 Al-measurement capability using our 3 MV NEC instrument. We use injection of the atomic beam and an analysis setup similar to 14 C, except that the 27 Al Faraday cup is located on the outside of the beam line. These studies have relatively low currents (compared to other nuclides) of ∼100–170 nA. We have already obtained interesting results relevant to geomorphological studies as well as lunar samples. Recent developments will focus on archaeological studies. Since the installation of the new NEC 3 MV Pelletron accelerator we have measured more than 700 129 I samples. Our results show we can now achieve a machine random error of 4.3% for iodine samples. Repeated analysis of the low-level Woodward iodine standard has yielded results in the low 10−14 range for 129 I/127 I ratios. Chemical extraction techniques for a variety of environmental samples including seawater, corals, stalagmites and seaweed have been developed. With these techniques we can measure 129 I/127 I ratios in samples with as little as 5 µg total iodine content. We have produced an 84-year record showing anthropogenic iodine invasion into corals in the South Pacific, and a 40,000-year record of cosmogenic iodine in a stalagmite from the Bahamas islands (Biddulph, 2004; Biddulph et al., 2005). 129 I appears to be an excellent tracer for surface ocean circulation. Two high resolution coral records of 20th century 129 I and 14 C were obtained from sites in the Solomon Islands and Easter Island, suggesting differences in oceanic and atmospheric 129 I transport at the two sites. A recent 129 I study in stalagmites and corals that there is some preliminary data suggesting that 129 I may track changes in the geomagnetic field and the cosmic ray flux. Data from a stalagmite (Biddulph et al., 2005) seems to reproduce magnetic excursions that have been seen in other radioisotope records such as 10 Be and 36 Cl. The relatively long half-life of 129 I (15.7 million years) would enable us to analyze records that go back nearly 100 million years. 7. Conclusions We have presented some applications of accelerator mass spectrometry which use the Arizona AMS laboratory. There are many varied uses of AMS and this paper only highlights a few chosen examples. The original focus of the NSF Arizona AMS Laboratory was AMS radiocarbon dating. We have expanded our capability of AMS measurements and research fields have expanded to other radionuclides such as 10 Be, 129 I, and 26 Al. We believe that the service to the scientific community and our active internal research programs make the Arizona laboratory one of best integrated AMS science laboratories in the world. In a recent paper, we posed the question (Jull and Burr, 2006): “Is the future of AMS bigger or smaller?” It is certainly bigger and perhaps almost unlimited for the scientist who considers the wide application of AMS to many areas of the geological and ocean sciences, and beyond. Acknowledgements The work at our laboratory is supported in part by grant EAR0448461 from the U.S. National Science Foundation.
258
A.J.T. Jull et al.
References Alley, R.A., Clark, P.U., Keigwin, L.D., Webb, R.S. (1999). Making sense of millennial-scale climate change. In: Clark, P.U., Webb, R.S., Keigwin, L.G. (Eds.), Mechanisms of Global Climate Change at Millennial Time Scales. In: Geophysics Monographs, vol. 112. American Geophysical Union, Washington, DC, pp. 385–394. Alley, R.B., Anandakrishnan, S., Jung, P., Clough, A. (2001). Stochastic resonance in the North Atlantic: Further insights. In: Seidov, D., Haupt, B.J., Maslin, M. (Eds.), The Oceans and Rapid Climate Change: Past, Present and Future. American Geophysical Union, Washington, DC, pp. 57–68. Arnold, J.R., Libby, W.F. (1949). Age determination by radiocarbon content: checks with samples of known age. Science 110, 678–680. Bard, E., Hamelin, B., Fairbanks, R.G., Zindler, A. (1990). Calibration of the 14 C timescale over the past 30,000 years using mass spectrometric U–Th ages from Barbados corals. Nature 345, 405–410. Bard, E., Arnold, M., Fairbanks, R.G., Hamelin, B. (1993). 230 Th/234 U and 14 C ages obtained by mass spectrometry on corals. Radiocarbon 35, 191–199. Bard, E., Arnold, M., Hamelin, B., Tisnerat-Laborde, N., Cabioch, G. (1998). Radiocarbon calibration by means of mass-spectrometric 230 Th/234 U and 14 C ages of corals: An updated database including samples from Barbados, Mururoa and Tahiti. Radiocarbon 40, 1085–1092. Bard, E., Ménot-Combes, G., Rostek, F. (2004). Present status of radiocarbon calibration and comparison records on Polynesian corals and Iberian margin sediments. Radiocarbon 46, 1189–1202. Baumgartner, S., Beer, J., Masarik, J., Wagner, G., Meynadier, L., Synal, H.A. (1998). Geomagnetic modulation of the 36 Cl flux in the GRIP ice core, Greenland. Science 279, 1330–1332. Beck, J.W., Richards, D.A., Edwards, R.L., Silverman, B.W., Smart, P.L., Donahue, D.J., Herrera-Osterheld, S., Burr, G.S., Calsoyas, L., Jull, A.J.T., Biddulph, D. (2001). Extremely large variations of atmospheric 14 C concentration during the last glacial period. Science 292, 2453–2458. Beck, J.W., Hoffmann, D.L., Richards, D.A., Smart, P.L., Paterson, B.A., Mattey, D.P., Kretchmark, T., Hawkesworth, C.J. (2005). High resolution records of 14 C variations in two Bahamian speleothems. In: Proceedings of the 10th International Conference on Accelerator Mass Spectrometry, Berkeley, CA, Sept. 3–7, 2005. Beck, J.W., Hoffmann, D.L., Richards, D.A., Smart, P.L., Paterson, B.A., Mattey, D.P., Kretchmark, T., Hawkesworth, C.J. (2006). Comparison of 14 C variations in two Bahamian speleothems (abstract). In: Proceedings of the 19th International Radiocarbon Conference, Oxford, England, April 3–7, 2006. Biddulph, D.L. (2004). Development and Applications of the NSF-Arizona AMS 129 I Program. Ph.D. Thesis, University of Arizona. Biddulph, D.L., Beck, J.W., Burr, G.S., Donahue, D.J. (2005). Two 60-year records of 129 I from coral skeletons in the Southern Pacific Ocean. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 592–598. Bird, M.I. (2007). Radiocarbon dating: charcoal. In: Elias, S. (Ed.), Encyclopedia of Quaternary Sciences. Elsevier, Oxford, pp. 2950–2958. Bird, M.I., Ayliffe, L.K., Fifield, L.K., Turney, C.S.M., Cresswell, R.G., Barrows, T.T., David, B. (1999). Radiocarbon dating of “old” charcoal using a wet oxidation, stepped-combustion procedure. Radiocarbon 41, 127–140. Bird, M.I., Fifield, L.K., Santos, G.M., Beaumont, P.B., Zhou, Y., di Tada, M.I., Hausladen, P.A. (2003). Radiocarbon dating from 40 to 60 ka BP at Border Cave, South Africa. Quatern. Sci. Rev. 22, 943–947. Bond, G., Showers, W., Cheseby, M., Lotti, R., Almasi, P., deMenocal, P., Priore, P., Cullen, H., Hajdas, I., Bonani, G. (1997). A pervasive millennial-scale cycle in North Atlantic Holocene and glacial climates. Science 278, 1257–1266. Braucher, R., Brown, E.T., Bourles, D.L., Colin, F. (2003). In situ produced 10 Be measurements at great depths: implications for production rates by fast muons. Earth Planet. Sci. Lett. 211 (3–4), 251–258. Bruhn, F., Duhr, A., Grootes, P.M., Mintrop, A., Nadeau, M.J. (2001). Chemical removal of conservation substances by ‘soxhlet’-type extraction. Radiocarbon 43 (2A), 229–237. Burney, D.A., Burney, L.P., Godfrey, L.R., Jungers, W.L., Goodman, S.M., Wright, H.T., Jull, A.J.T. (2004). A chronology for late prehistoric Madagascar. J. Human Evol. 47, 25–63. Burr, G.S., Edwards, R.L., Donahue, D.J., Druffel, E.R.M., Taylor, F.W. (1992). Mass spectrometric 14 C and U–Th measurements in coral. Radiocarbon 34, 611–618. Burr, G.S., Beck, J.W., Taylor, F.W., Récy, J., Edwards, R.L., Cabioch, G., Corrège, T., Donahue, D.J., O’Malley, J.M. (1998). A coral-based radiocarbon calibration between 11,700 and 12,400 calendar years BP derived from 230 Th ages of corals from Espiritu Santo Island, Vanuatu. Radiocarbon 40 (3), 1093–1105.
Accelerator mass spectrometry of long-lived light radionuclides
259
Burr, G.S., Galang, C., Taylor, F.W., Gallup, C., Edwards, R.L., Cutler, K., Quirk, W. (2004). Radiocarbon results from a 13 ka BP coral from the Huon Peninsula, Papua New Guinea. Radiocarbon 46 (3), 1211–1224. Cockburn, H.A.P., Summerfield, M.A. (2004). Geomorphological applications of cosmogenic isotope analysis. Prog. Phys. Geogr. 28, 1–42. Cutler, K.B., Gray, S.C., Burr, G.S., Edwards, R.L., Tayor, F.W., Cabioch, G., Beck, J.W., Cheng, H., Moore, J. (2004). Radiocarbon calibration and comparison to 50 kyr BP with paired 14 C and 230 Th dating of corals from Vanuatu and Papua New Guinea. Radiocarbon 46, 1127–1160. Damon, P.E., Sonnett, C.P. (1991). Solar and terrestrial components of the atmospheric 14 C variation spectrum. In: Sonnett, C.P., Giampapa, M.S., Matthews, M.S. (Eds.), The Sun in Time. Univ. of Arizona Press, Tucson, pp. 361– 388. Donahue, D.J., Jull, A.J.T., Linick, T.W., Toolin, L.J. (1990a). Radiocarbon measurements at the University of Arizona AMS Facility. Nucl. Instrum. Methods Phys. Res. Sect. B 52, 224–228. Donahue, D.J., Linick, T.W., Jull, A.J.T. (1990b). Isotope-ratio and background corrections for accelerator mass spectrometry radiocarbon measurements. Radiocarbon 32, 135–142. Dyke, A.S., Andrews, J.T., Clark, P.U., England, J., Miller, G.H., Shaw, J., Veillette, J.J. (2001). Radiocarbon dates pertinent to defining the last glacial maximum for the Laurentide and Innuitian ice sheets. Open File 4120, Geological Survey of Canada, Toronto, Canada, 54 pp. Edwards, R.L., Beck, J.W., Burr, G.S., Donahue, D.J., Chappell, J.M.A., Bloom, A.L., Druffel, E.R.M., Taylor, F.W. (1993). A large drop in atmospheric 14 C/12 C and reduced melting in the Younger Dryas documented with 230 Th ages of corals. Science 260, 962–968. Elmore, D., Phillips, F.M. (1987). Accelerator mass spectrometry for measurement of long-lived radioisotopes. Science 236, 543–550. Fairbanks, R.G. (1989). A 17,000-year glacio-eustatic sea level record: influence of glacial melting rates on the Younger Dryas event and deep-ocean circulation. Nature 342, 637–642. Fairbanks, R.G., Mortlock, R.A., Chiu, T.-C., Cao, L., Kaplan, A., Guilderson, T.P., Fairbanks, T.W., Bloom, A.L., Grootes, P.M., Nadeau, M.-J. (2005). Radiocarbon calibration curve spanning 0 to 50,000 years BP based on paired 230 Th/234 U/238 U and 14 C dates on pristine corals. Quatern. Sci. Rev. 24, 1781–1796. Fifield, L.K. (1999). Accelerator mass spectrometry and its applications. Rep. Prog. Phys. 62, 1223–1274. Fifield, L.K., Synal, H.A., Suter, M. (2004). Accelerator mass spectrometry of plutonium at 300 kV. Nucl. Instrum. Methods Phys. Res. Sect. B 223–224, 802–806. Goslar, T., Arnold, M., Tisnérnat-Laborde, N., Hatté, C., Paterne, M., Ralska-Jasiewiczowa, M. (2000). Radiocarbon calibration by means of varves versus 14 C ages of terrestrial macrofossils from Lake Go´sciaz and Lake Perespilno, Poland. Radiocarbon 42, 403–414. Gosse, J., Phillips, F.M. (2001). Terrestrial in-situ cosmogenic nuclides: theory and application. Quatern. Sci. Rev. 20, 1475–1560. Grajcar, M., Döbeli, M., Kubik, P.W., Maden, C., Suter, M., Synal, H.A. (2004). 10 Be measurements with terminal voltages below 1 MV. Nucl. Instrum. Methods Phys. Res. Sect. B 223–224, 190–194. Hallett, D.J., Walker, R.C. (2000). Paleoecology and its application to fire and vegetation management in Kootenay National Park, British Columbia. J. Paleolimnol. 24, 401–414. Hallett, D.J., Lepofksy, D.S., Mathewes, R.W., Lertzman, K.P. (2003). 11,000 years of fire history and climate in the mountain hemlock rain forests of southwestern British Columbia based on sedimentary charcoal. Can. J. Forest Res. 33, 292–312. Hatté, C., Morvan, J., Noury, C., Paterne, M. (2001). Is classical acid–alkali–acid treatment responsible for contamination? An alternative proposition. Radiocarbon 43, 177–182. Haynes, C.V. (1984). Stratigraphy and Late Pleistocene extinctions in the United States. In: Martin, P.S., Klein, R.G. (Eds.), Quaternary Extinctions. Univ. of Arizona Press, Tucson. Haynes, C.V. (1991). Geoarchaeological and paleohydrological evidence for a Clovis-age drought in North America and its bearing on extinction. Quatern. Res. 35, 438–450. Haynes, C.V. (1992). Contributions of radiocarbon dating to the geochronology of the peopling of the New World. In: Taylor, R.E., Long, A., Kra, R.S. (Eds.), Radiocarbon after Four Decades. Springer-Verlag, New York. Heisinger, B., Lal, D., Jull, A.J.T., Kubik, P.W., Ivy-Ochs, S., Neumaier, S., Knie, K., Lazarev, V., Nolte, E. (2002). Production of selected cosmogenic radionuclides by muons. Geochim. Cosmochim. Acta 66, A558.
260
A.J.T. Jull et al.
Hughen, K.A., Overpeck, J.T., Lehman, S.J., Kashgarian, M., Southon, J., Peterson, L.C., Alley, R., Sigman, D.M. (1998). Deglacial changes in ocean circulation from an extended radiocarbon calibration. Nature 391, 65–68. Hughen, K.A., Baillie, M.G.L., Bard, E., Bayliss, A., Beck, J.W., Blackwell, P.G., Buck, C.E., Burr, G.S., Cutler, K.B., Damon, P.E., Edwards, R.L., Fairbanks, R.G., Friedrich, M., Guilderson, T.P., Herring, C., Kromer, B., McCormac, F.G., Ramsey, S.W., Reimer, P.J., Reimer, R.W., Remmele, S., Southon, J.R., Stuiver, M., Talamo, S., Taylor, F.W., van der Plicht, J., Weyhenmeyer, C.E. (2004). Marine radiocarbon age calibration, 20–0 ka BP. Radiocarbon 46, 1059–1086. Hughey, B.J., Shefer, R.E., Klinkowstein, R.E., Zhao, X.L., Kieser, W.E., Litherland, A.E. (1997). Experimental investigation of the destruction of 1 MeV 12 CH2 molecules in single and double stripping foils. Nucl. Instrum. Methods Phys. Res. B 123, 186–192. Hughey, B., Skipper, P.L., Klinkowstein, R.E., Shefer, R.E., Wishnook, J.S., Tannenbaum, S.R. (2000). Low-energy biomedical GC-AMS system for 14 C and 3 H detection. Nucl. Instrum. Methods Phys. Res. 172, 40–46. Jones, K.B., Hodgins, G.W.L., Lange, T.E. (2005). A new plasma oxidation apparatus for radiocarbon dating (abstract). In: Proceedings of the 10th International Conference on Accelerator Mass Spectrometry, Berkeley, CA, Sept. 5–10, 2005. Jull, A.J.T. (2007). Radiocarbon dating: AMS method. In: Elias, S. (Ed.), Encyclopedia of Quaternary Sciences. Elsevier, Oxford, pp. 2911–2918. Jull, A.J.T., Burr, G.S. (2006). Accelerator mass spectrometry: Is the future bigger or smaller? Earth Planet. Sci. Lett. 243, 305–325. Jull, A.J.T., Geertsema, M. (2006). Forest fire frequency over the last 16,500 years in northern British Columbia determined from AMS radiocarbon dating of charcoal and sediments. Radiocarbon 48, 435–456. Jull, A.J.T., Courtney, C., Jeffrey, D.A., Beck, J.W. (1998). Isotopic evidence for a terrestrial source of organic compounds found in Martian meteorites ALH 84001 abd EETA 79001. Science 279, 366–369. Jull, A.J.T., Burr, G.S., Beck, J.W., Donahue, D.J., Biddulph, D., Hatheway, A.L., Lange, T.E., McHargue, L.R. (2003). Accelerator mass spectrometry at Arizona: Geochronology of the climate record and connections with the ocean. J. Environ. Radioact. 69 (1–2), 3–19. Jull, A.J.T., Burr, G.S., McHargue, L.R., Lange, T.E., Lifton, N.A., Beck, J.W., Donahue, D.J. (2004). New frontiers in dating of geological, paleoclimatic and anthropological applications using accelerator mass spectrometric measurements of 14 C and 10 Be in diverse samples. Global Planet. Change 41, 309–323. Jull, A.J.T., Burr, G.S., Beck, J.W., Hodgins, G.W.L., Biddulph, D.L., Gann, J., Hatheway, A.L., Lange, T.E., Lifton, N.A. (2006). Application of accelerator mass spectrometry to environmental and paleoclimate studies at the University of Arizona. Radioact. Environ. 8, 3–23. Kitagawa, H., van der Plicht, J. (1998). Atmospheric radiocarbon calibration to 45,000 yr B.P.: Late Glacial fluctuations and cosmogenic isotope production. Science 279, 1187–1190. Lertzman, K., Gavin, D., Hallett, D., Brubaker, L., Lepofsky, D., Mathewes, R. (2002). Long-term fire regime estimated from soil charcoal in coastal temperate rain forests. Conserv. Ecol. 6 (2). Paper 5 (on-line journal). Libby, W.F. (1955). Radiocarbon Dating. Univ. of Chicago Press, Chicago. Klein, J., Lerman, J.C., Damon, P.E., Ralph, E.K. (1982). Calibration of radiocarbon dates—Tables based on the consensus data of the Workshop on Calibrating the Radiocarbon Time Scale. Radiocarbon 24, 103–150. Lifton, N.A., Jull, A.J.T., Quade, J. (2001). A new extraction technique and production rate estimate for in situ cosmogenic 14 C in quartz. Geochim. Cosmochim. Acta 65, 1953–1969. Lifton, N.A., Pigati, J., Jull, A.J.T., Quade, J. (2002). Altitudinal variation of in situ cosmogenic 14 C production rates: Preliminary results from the Southwestern US. Geochim. Cosmochim. Acta 66 (15A), A457. Markgraf, V. (2001). Interhemispheric Climate Linkages. Academic Press, New York. Martin, P.S., Klein, R.G. (1984). Quaternary Extinctions: A Prehistoric Revolution. Univ. of Arizona Press, Tucson. McGeehin, J., Burr, G.S., Jull, A.J.T., Reines, D., Gosse, J., Davis, P.T., Muhs, D., Southon, J. (2001). Steppedcombustion 14 C dating of sediment. Radiocarbon 43 (2A), 255–262. McGeehin, J., Burr, G.S., Hodgins, G., Bennett, S.J., Robbins, J.A., Morehead, N., Markewich, H. (2004). Steppedcombustion 14 C dating of bomb carbon in lake sediment. Radiocarbon 46 (1), 893–900. McHargue, L.R., Donahue, D.J. (2005). Effects of climate and the cosmic-ray flux on the 10 Be content of marine sediments. Earth Planet. Sci. Lett. 232, 193–207. McHargue, L.R., Damon, P.E., Donahue, D.J. (1995). Enhanced cosmic-ray production of 10 Be coincident with the Mono Lake and Laschamp geomagnetic excursions. Geophys. Res. Lett. 22, 659–662.
Accelerator mass spectrometry of long-lived light radionuclides
261
McHargue, L.R., Donahue, D., Damon, P.E., Biddulph, D., Sonett, C.P., Burr, G. (2000). Geomagnetic modulation of the late Pleistocene cosmic-ray flux as determined by 10 Be from Blake Outer Ridge marine sediments. Nucl. Instrum. Methods Phys. Res. Sect. B 172, 555–561. McNichol, A.P., Jull, A.J.T., Burr, G.S. (2001). Converting AMS data to radiocarbon values: considerations and conventions. Radiocarbon 43, 313–320. Meltzer, D.J., Grayson, D.K., Ardila, G., Barker, A.W., Dincauze, D.F., Haynes, D.V., Mena, F., Nuñez, L.A., Stanford, D.J. (1997). On the Pleistocene antiquity of Monte Verde, southern Chile. Amer. Antiquity 62, 659–663. Meyer, G.A., Wells, S.G., Jull, A.J.T. (1995). Fire and alluvial chronology in Yellowstone National Park: Climatic and intrinsic controls on Holocene geomorphic processes. Geol. Soc. Am. Bull. 107, 1211–1230. Meyer, G.A., Pierce, J.L., Wood, S.H., Jull, A.J.T. (2001). Fire, storms, and erosional events in the Idaho batholith. Hydrol. Processes 15, 3025–3038. Nuñez, L.A., Varela, J., Casamiquela, R., Villagrán, C. (1994). Reconstrucción multidisciplinaria de la occupación prehistórica de Quereo, Centro de Chile. Latin Amer. Antiquity 5, 99–118. Pierce, J.L., Meyer, G.A., Jull, A.J.T. (2004). Fire-induced erosion and millennial scale climate change in northern ponderosa pine forests. Nature 432, 87–90. Pigati, J.S., Lifton, N.A. (2004). Geomagnetic effects on time—integrated cosmogenic nuclide production with emphasis on in situ 14 C and 10 Be. Earth Planet. Sci. Lett. 226 (1–2), 193–205. Raisbeck, G.M., Yiou, F., Bourlès, J., Lestringuez, J., Deboffle, D. (1987). Measurements of 10 Be and 26 Al with a tandetron AMS facility. Nucl. Instrum. Methods Phys. Res. B 29, 22–26. Raisbeck, G.M., Yiou, F., Jouzel, J., Petit, J.R. (1990). 10 Be and 2 H in polar ice cores as a probe of the solar variability influence on climate. Philos. Trans. R. Soc. London Ser. A 330, 463–470. Reimer, P.J. (2004). In: IntCal04 Calibration Issue. Radiocarbon 46, 1029–1304. Reimer, P.J., Baillie, M.G.L., Bard, E., Bayliss, A., Beck, J.W., Bertrand, C.J.H., Blackwell, P.G., Buck, C.E., Burr, G.S., Cutler, K.B., Damon, P.E., Edwards, R.L., Fairbanks, R.G., Friedrich, M., Guilderson, T.P., Hogg, A.G., Hughen, K.A., Kromer, B., McCormac, G., Manning, S., Bronk Ramsey, C., Reimer, R.W., Remmele, S., Southon, J.R., Stuiver, M., Talamo, S., Taylor, F.W., van der Plicht, J., Weyhenmeyer, C.E. (2004). INTCAL04 terrestrial radiocarbon age calibration, 0–26 kyr BP. Radiocarbon 46, 1029–1058. Richards, D.A., Hofmann, D.L., Beck, J.W., Smart, P.L., Paterson, B.A., Mattey, D.P. (2006). Exploring the potential causes of atmospheric 14 C variation using multi-proxy evidence from Bahamian speleothems 34–45 ka. In: Radiocarbon Conference Abstracts, Oxford, England, April 3–7, 2006. Russ, J., Hyman, M., Rowe, M. (1992). Direct radiocarbon dating of rock art. Radiocarbon 34 (3), 867–872. Sanborn, P., Geertsema, M., Jull, A.J.T., Hawkes, B. (2006). Soil charcoal evidence for Holocene fire-related sedimentation in an inland temperature rainforest, east–central British Columbia, Canada. Holocene 16, 1–13. Schramm, A., Stein, M., Goldstein, S.I. (2000). Calibration of the 14 C time scale to >40 ka by 234 U/230 Th dating of Lake Lisan sediments. Earth Planet. Sci. Lett. 175, 27–40. Schwartz, M., Lund, S.P., Johnson, T.C. (1998). Geomagnetic field intensity from 71 to 12 ka as record in deep-sea sediments of the Blake Outer Ridge, North Atlantic Ocean. J. Geophys. Res. 103, 443–452. Steadman, D.W., Martin, P.S., McPhee, R.D.E., Jull, A.J.T., McDonald, H.G., Woods, C.A., Gann, J., Hodgins, G.W. (2005). Young radiocarbon dates from Caribbean ground sloths. Proc. Natl. Acad. Sci. USA 102, 11763–11768. Stuiver, M., Kra, R. (1986). In: Calibration Issue. Radiocarbon 28, 808–1030. Stuiver, M., van der Plicht, J. (1998). In: INTCAL98: Calibration Issue. Radiocarbon 40, 1041–1159. Stuiver, M., Long, A., Kra, R. (1993). In: Calibration 1993. Radiocarbon 35, 1–244. Suter, M., Jacob, S., Synal, H.A. (1997). AMS of 14 C at low energies. Nucl. Instrum. Methods Phys. Res. Sect. B 123, 153–158. Suter, M., Jacob, S.W.A., Synal, H.A. (2000). Tandem AMS at sub-MeV energies. Nucl. Instrum. Methods Phys. Res. Sect. B 172, 40–46. Synal, H.A., Stocker, M., Suter, M. (2006). MICADAS: a new compact radiocarbon AMS system (abstract). In: Proc. 10th International Conference on Accelerator Mass Spectrometry, Berkeley, CA, p. 61. Taylor, R.E. (2000). The contribution of radiocarbon dating to New World archaeology. Radiocarbon 42, 1–21. Tuniz, C., Bird, J.R., Fink, D., Herzog, G.F. (1998). Accelerator Mass Spectrometry: Ultrasensitive Analysis for Global Science. CRC Press, Boca Raton, 371 pp. Turcq, B., Siffedine, A., Martin, L., Absy, M.L., Soubies, F., Suguio, K., Volkmer-Ribeiro, C. (1998). Amazonian rainforest fires: A lacustrine record of 7000 years. Ambio 27, 139–142.
262
A.J.T. Jull et al.
van der Plicht, J., Beck, J.W., Bard, E., Baillie, M.G.L., Blackwell, P.G., Buck, C.E., Friedrich, M., Guilderson, T.P., Hughen, K.A., Kromer, B., McCormac, F.G., Bronk Ramsey, C., Reimer, P.J., Reimer, R.W., Remmele, S., Richards, D.A., Southon, J.R., Stuiver, M., Weyhenmeyer, C.E. (2004). NOTCAL04-comparison calibration 14 C records 26–50 cal kyr BP. Radiocarbon 46, 1225–1238. Voelker, A.H.L., Grootes, P.M., Nadeau, M.J., Sarntheim, M. (2000). Radiocarbon levels in the Iceland Sea from 25–53 kyr and their link to the earth’s magnetic field intensity. Radiocarbon 42, 437–452.
263
Accelerator mass spectrometry of long-lived heavy radionuclides L.K. Fifield∗ Department of Nuclear Physics, Research School of Physical Sciences and Engineering, The Australian National University, Canberra, ACT 0200, Australia Abstract Accelerator mass spectrometry (AMS) is presently the most sensitive technique for the environmental measurement of radionuclides with half-lives greater than about 5000 years. Traditionally, it has been used principally for light elements, of which the most familiar example is 14 C. With some modifications, however, it may also be applied to elements as heavy as the actinides, and in particular to 239,240,244 Pu and 236 U. Here, the principles and methodology of heavy-element AMS are described, and the ways in which these have been implemented in various laboratories around the world are detailed. Although the emphasis is on plutonium and uranium, other isotopes such as 237 Np and 226,228 Ra are also considered. For completeness, the measurement of the long-lived fission products 99 Tc and 129 I by AMS is also discussed briefly. Actual or potential applications of the method in the areas of environmental science, biomedicine, and nuclear safeguards are reviewed.
1. Introduction Anthropogenic α-particle emitting nuclides with half-lives that are long relative to the human life-span have been released into the environment by nuclear testing, nuclear accidents and reprocessing operations. Among the most significant are 239,240 Pu, 236 U and 237 Np. Quantifying the releases, and tracing their subsequent dispersal has traditionally been the task of α-particle counting or, more recently, of thermal ionization or inductively coupled plasma (ICP) mass spectrometers. Although these are mature methodologies, each has its limitations. These limitations are largely surmounted by the relatively new technique of accelerator mass spectrometry (AMS). For realistic counting times, α-particle counting is limited in sensitivity to ∼50 µBq (O’Donnell et al., 1997). Counting times required for sensible results at this level are ∼4 weeks. For 239 Pu, the limit corresponds to 20 fg. In addition, α-particle counting is unable to resolve the two most important plutonium isotopes, 239 Pu and 240 Pu, because their α-particle energies differ by only 11 keV in 5.25 MeV. Hence, information on the 240 Pu/239 Pu ratio, which is a useful indicator of the source of the plutonium, is not available. ∗ E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11008-1
© 2008 Elsevier B.V. All rights reserved.
264
L.K. Fifield
Fig. 1. Essential features of an AMS system for the measurement of 236 U.
Mass-spectrometric methods do give information on the 240 Pu/239 Pu ratio, and potentially have higher sensitivity than α-particle counting, with values as low as ∼1 fg having been reported, but are sensitive to molecular interferences (Wyse et al., 2001). For example, 238 UH, 208 Pb31 P, etc. could interfere with measurement of 239 Pu. Accelerator mass spectrometry combines the high-sensitivity of the mass-spectrometric methods with a high level of discrimination against molecular interferences. The means of achieving this are described in detail below. Applications of the method are presented in subsequent sections. Although the principal emphasis of this chapter is on the α-particle emitting actinides, two particularly long-lived fission products, 99 Tc and 129 I, may also be measured by AMS with sensitivities that are considerably higher than competing techniques. These are also considered briefly below.
2. Principles of AMS as applied to heavy radionuclides 2.1. Uranium-236 It is convenient to introduce the concepts of the AMS method using 236 U as an illustrative example. Figure 1 shows the basic elements of an AMS system configured for 236 U. It consists essentially of two spectrometers separated by an accelerator that serves as a molecular dissociator. Its basic features are: 1. A low-energy mass spectrometer consisting of a negative ion source and a mass-analyzing magnet. The UO− molecular ion is selected because the U− atomic ion is produced only weakly by negative ion sources of the type employed for AMS.
Accelerator mass spectrometry of long-lived heavy radionuclides
265
2. Acceleration of the negative ions to the positive high-voltage terminal of a tandem electrostatic accelerator. 3. Dissociation of the molecules and removal of electrons to produce positive uranium ions. 4. Acceleration of the positive ions back to ground potential, followed by a second magnetic analysis that selects a single charge state with a well-defined energy. 5. An additional analysis by an electrostatic analyzer or velocity filter. The combination of magnetic and electrostatic analyses eliminates many of the potential sources of background. 6. A detection system, derived from the types of detectors used in basic nuclear physics research for detecting energetic charged particles, allows positive identification of each individual 236 U ion and the rejection of any remaining background ions. AMS measures an isotope ratio, which would be 236 U/238 U in our illustrative example. The 236 U numerator is derived from the counting rate of 236 U ions in the detector. The 238 U denominator is derived by periodically switching the system to transmit 238 U ions through the full AMS system and measuring their flux as an electric current in a Faraday cup. 2.2. Plutonium isotopes AMS of 236 U is more or less conventional in that the 236 U is measured by ion counting while the essentially “stable” isotope, 238 U, can be readily measured as an electric current. In the case of plutonium, there is no stable isotope. Hence, it is necessary to add a “spike” of a known amount of one of the long-lived isotopes, generally 242 Pu, to the sample in order to quantify the concentrations of the isotopes of interest, which are usually 239,240 Pu or 244 Pu. Both cost and radiological issues dictate that the amount of spike added is in the range 1– 10 pg, in marked contrast to a typical uranium AMS sample that will contain a few milligrams of uranium. Consequently, both the isotopes of interest and the 242 Pu spike must be measured by ion counting (Fifield et al., 1997, 1996). 2.3. Neptunium-237 Neptunium-237 would be very similar to 239 Pu if pure 236 Np were readily available to use as a spike. Unfortunately, it is not, and measurements of 237 Np at both the ANU and Livermore have therefore been made relative to a 242 Pu spike (Brown et al., 2004; Fifield et al., 1997; Keith-Roach et al., 2001). This has two drawbacks. First, the 242 Pu must be added after the neptunium extraction chemistry, and hence does not compensate for any losses of yield in the extraction process. Secondly, the negative-ion formation probabilities of NpO− and PuO− are not the same. A value of 0.74 for the relative formation probability has been measured using a sample containing known amounts of 237 Np and 242 Pu (Keith-Roach et al., 2001). Both factors are additional sources of uncertainty. In principle, the first handicap may be overcome by using 2.4-day 239 Np as a chemical yield monitor that can be monitored via the 277 and 228 keV γ -rays emitted in its decay. The 239 Np may be milked from 243 Am retained on an ion-exchange column. Since the 243 Am stock contains some 241 Am, which is the parent of 237 Np, care must be taken to ensure that 237 Np has not had the opportunity to build up before the 239 Np is milked (Keith-Roach et al., 2001).
266
L.K. Fifield
2.4. Radium Both the 238 U and 232 Th decay series pass through radium isotopes. The 238 U series decays through 226 Ra, which is an α-particle emitter with a half life of 1600 years, long enough to make AMS detection an attractive alternative to α-decay counting. The 232 Th series decays through 228 Ra, which is a β-particle emitter with a half-life of only 5 years. Normally, AMS would not offer any advantages over decay counting for a lifetime as short as this. In order to identify 228 Ra uniquely by a decay-counting technique, however, it is necessary to detect the characteristic α-particles from the decay of its 228 Rn granddaughter, which has a half-life of 1.9 years. This necessitates a wait of 6 months or more for sufficient 228 Rn activity to grow in after separation of the radium from the sample. Hence a mass-spectrometric method such as AMS offers the benefit of faster turn-around time. Natural background isotopic concentrations typically found in soils are 1 pg/g for 226 Ra and 5 fg/g for 228 Ra, and hence the very high sensitivity of AMS offers significant advantages in terms of both sample size and a reduction in the complexity of sample preparation compared to α-particle spectroscopy. The interest in radium stems from the need to monitor releases into the environment from uranium mining activities. If uranium is escaping into the environment, the spatial distribution of uranium will be more variable than that of thorium. Accordingly, measurements of the 226 Ra/228 Ra ratio may provide a probe with which to assess variations in the amount of uranium-process derived 226 Ra. Furthermore, for contaminated or rehabilitated areas where the 226 Ra/228 Ra ratio is anomalous, measurements of the transport of material away from the site via the ratio could provide information on the local erosion rate. In contrast to plutonium or uranium, radium does not readily form the oxide molecular negative ion. Instead, the most prolific negative ion discovered to date is RaC− 2 . In order to prepare samples in a form that will produce a good yield of this negative ion, the radium is first separated with an anion exchange resin. Graphite powder is then added to the acid solution containing the radium and the solution evaporated to dryness before pressing into sample holders. The AMS method (Tims et al., 2004) closely follows that for plutonium. AMS measures only the 228 Ra/226 Ra ratio. Since there are no other long-lived isotopes to use as a spike, an isotope dilution method must be employed in order to determine the concentrations. This involves the preparation of two samples, with and without the addition of a known amount of pure 226 Ra, and measuring the 228 Ra/226 Ra ratio for both. 2.5. Iodine-129 Iodine-129 is a prolific fission product, and has been released into the environment in considerable quantity by nuclear weapons testing, by nuclear accidents, Chernobyl and Three Mile Island in particular, and by controlled releases from nuclear-fuel reprocessing plants. A total of 2.4 tons of 129 I was released from the La Hague and Sellafield plants up to 1997, and this was increasing at the rate of ∼300 kg/year (Raisbeck and Yiou, 1999). Because the discharges from both La Hague and Sellafield are carried into the North Atlantic by the Gulf Stream, 129 I is proving very useful as an oceanographic tracer in an area of the world’s oceans that plays a vital role in the global thermohaline circulation (Gascard et al., 2004a, 2004b; Raisbeck and Yiou, 1999; Raisbeck et al., 1995; Yiou et al., 2004, 2002, 1994). AMS offers the advantages of very high sensitivity, and therefore small sample sizes. A liter of North
Accelerator mass spectrometry of long-lived heavy radionuclides
267
Atlantic water is sufficient for a measurement. Other applications have included post facto dosimetry of populations exposed to the Chernobyl accident, monitoring releases from nuclear facilities (Kilius et al., 1994; Rucklidge et al., 1994), and dating brines associated with oil fields, gas hydrates, or coalbed methane (Fehn et al., 2000, 2003; Fehn and Snyder, 2005; Moran et al., 1995; Snyder and Fehn, 2004, 2002; Snyder et al., 2003). In the Chernobyl application, long-lived 129 I has been used as a proxy for the much shorter-lived 131 I (Mironov et al., 2002). The latter is radiologically significant because it accumulates in the thyroid, but with a half-life of only 8 days, most of it had decayed away before it could be measured. Since the relative amounts of 131 I and 129 I in the Chernobyl fallout were well known, a later measurement of 129 I allows the exposure to 131 I to be determined. Because the stable isobar, 129 Xe, does not form negative ions, 129 I may be measured almost equally well on small AMS systems operating with accelerating voltages as low as 0.5 MV as on larger systems operating at >5 MV (Kilius et al., 1990). Since iodine readily forms negative ions, 129 I− is the negative ion of choice for injection into the accelerator. An electrostatic analyzer between ion source and mass-analyzing magnet is crucial in small-accelerator systems to eliminate the low-energy tail of the intense 127 I− beam. In order to avoid interference from M = 129 molecular ions, charge state 3+ or higher is selected for analysis after acceleration. 2.6. Technetium-99 Technetium-99 is also a prolific fission product, and has been released principally from the Sellafield plant. The largest releases took place between 1994 and 2003 following the commissioning of the Enhanced Actinide Recovery Plant, and averaged 150 kg/yr. Since 2003, a new recovery process has substantially reduced the amount released to the Irish Sea. Technetium99 is a β-emitter with a half life of 212,000 years. Concentrations in seawater or seaweed have traditionally been measured using liquid-scintillation or gas-proportional counting of the βdecay after chemical extraction of the Tc to reduce other activities as much as possible, or by ICP-MS (Leonard et al., 2004; McCartney et al., 1999). Sensitivity of both the counting and mass-spectrometric techniques is at the level of pg of the isotope. The principal application has been in monitoring the dispersal of the Sellafield discharges, which are readily detectable along the Scandinavian coast across the North Sea. In a similar way to 129 I, it would be possible to use 99 Tc as an oceanographic tracer of North Atlantic circulation. Since 99 Tc was mostly sourced from Sellafield, whereas 129 I has been discharged from both La Hague and Sellafield, the two isotopes provide complementary information (Raisbeck and Yiou, 1999). Realization of the potential of 99 Tc, however, requires higher sensitivity than is available from radiometric or ICP-MS techniques. AMS potentially offers higher sensitivity provided that the contribution from the 99 Ru isobar can be kept low. In practice this requires good chemistry to minimize any Ru in the sample to keep 99 Ru rates low in the detector, and a detector that can discriminate between 99 Tc and 99 Ru. Effective suppression of Ru can be achieved with a technetiumspecific ion-exchange resin (Eichrom TEVA-Spec™), and this is now routinely employed for both decay-counting (to reduce 106 Ru) and ICP-MS (Leonard et al., 2004; McCartney et al., 1999). Discrimination between 99 Tc and 99 Ru requires an ionization detector that makes multiple measurements of the energy loss of the ions as they slow down in gas,
268
L.K. Fifield
but this is only effective if the ion energy is at least 120 MeV (Bergquist et al., 2000; Fifield et al., 2000). Hence a large accelerator operating at >10 MV is required. Sample preparation is again similar to that for plutonium. At the ANU, 10 pg of rhodium is added for normalization purposes after separation and purification of the Tc by ion exchange techniques, and the resulting mixture is dispersed in iron oxide. At Livermore, the Tc is dispersed in a niobium oxide matrix, and the 93 Nb beam current is used for normalization (Bergquist et al., 2000). The TcO− negative ion is selected for analysis. Although the Tc− ion is equally prolific, the choice of the oxide ion provides some suppression against Ru because the Ru− /RuO− ratio is greater than the Tc− /TcO− ratio (Fifield et al., 2000). The AMS method developed at the ANU (Fifield et al., 2000) measures sequentially 103 Rh, 101 Ru, and 99 Tc in the 13+ charge state at energies of ∼200 MeV in order to determine the 99 Tc/103 Rh ratio. All three isotopes are measured by ion-counting in the ionization chamber. Measurement of 101 Ru allows subtraction of the 99 Ru contribution to the observed counting rate at mass-99 as an alternative to separating the 99 Tc and 99 Ru with the detector. This works well for samples where the bulk of the count rate is due to 99 Tc, but is insufficiently precise for samples with low 99 Tc concentrations. At Livermore, the ion energy is 125 MeV, and the 93 Nb beam current is measured in an off-axis Faraday cup after the analyzing magnet (Bergquist et al., 2000).
3. Advantages of AMS relative to conventional mass spectrometry Both 236 U and plutonium isotopes have been measured by conventional mass spectrometry, using either thermal-ionization (TIMS) or inductively coupled plasma (ICP-MS) positive ion sources. For uranium, outputs from TIMS and ICP ion sources are in the range of 10–100 pA of 238 U (∼108 –109 ions/s). Hence, at a 236 U/238 U ratio of 10−10 , which is at the upper end of the ratios found naturally in high grade uranium ores, the flux of 236 U ions would be only 0.01–0.1 s−1 . On the other hand, ratios can be 10−7 or higher in environments contaminated by nuclear accidents, reprocessing operations or depleted uranium projectiles, and 236 U fluxes are correspondingly higher. At these low counting rates it is necessary to use a secondary electron multiplier to count the 236 U ions. Background under the 236 U peak arises from the 235 UH molecule or other molecules of mass 236, and from the tail of the 238 U beam. The best reported UH+ /U+ ratios from ICP sources using ultrasonic nebulizers and desolvators are 2–3 × 10−6 (Ketterer et al., 2003). Hence the 235 UH contribution at mass-236 corresponds to a 235 UH/238 U ratio of 2×10−8 . Further, the low-energy tail of the intense 238 U beam is ∼10−6 of the main peak (Ketterer et al., 2003). Variability in both the molecular and tail contributions limits the sensitivity of ICP-MS to 236 U/238 U ratios of ∼10−7 . TIMS ion sources, on the other hand, produce much lower molecular beams, and a decelerating lens after magnetic analysis can reduce the 238 U tail to a negligible level. Background is then dominated by the dark noise of the electron multiplier and by scattered lighter-mass ions. Richter et al. (1999) report background counting rates of 10−2 s−1 and a sensitivity of 1 × 10−10 in the 236 U/238 U ratio. In contrast to the lower currents available from TIMS and ICP sources, negative ion outputs from sputter sources are ∼100 nA of 238 UO− , and even given typical transmission of 3%, still represent 2 × 1010 238 U ions/s after acceleration. Hence, sufficient counts of 236 U can
Accelerator mass spectrometry of long-lived heavy radionuclides
269
be acquired to allow measurement of ratios as low as ∼10−12 provided that backgrounds are suitably low. At a ratio of 10−12 , 236 U ions would be detected at a rate of about 1 count per minute. The 235 U17 O or 238 U14 N molecular interferences, as well as other more exotic molecular interferences involving lighter elements such as 133 Cs107 Ag12 C, are broken up in the stripper of the AMS system, and any backgrounds arising from the very small fraction of fragments that pass the subsequent analysis system are readily discriminated by the detection system at the high energies pertaining after acceleration. For plutonium isotopes, abundance sensitivity is not a problem for conventional mass spectrometry due to the absence of a relatively intense beam of similar mass. Molecular interferences such as 238 UH, 208 Pb31 P, etc. can still be a problem, however. These can be at least partially overcome with a high-resolution magnetic sector instrument, and sensitivities at the fg level have been reported for modern high-resolution ICP-MS systems (Wyse et al., 2001). Nevertheless, AMS offers advantages in terms of immunity to molecular interferences which might go unrecognized in a conventional system, and sensitivity can approach 0.1 fg (Brown et al., 2004; Fifield et al., 1997, 1996).
4. Advantages of AMS relative to α-spectrometry Alpha-particle spectroscopy has been widely used for measuring plutonium concentrations, particularly for tracing releases from nuclear-fuel reprocessing facilities. The limit to its sensitivity is imposed by realistic counting times relative to the half-lives of the isotopes. Assuming a maximum counting time of one month, and a typical 240 Pu/239 Pu ratio of 20%, only 2×10−6 of the atoms would decay in this time. Since the (geometric) efficiency of detection is ∼0.3, fewer than one in a million of the Pu atoms in the sample will be detected. Realistically, the sensitivity limit is ∼50 µBq (O’Donnell et al., 1997) or 5 × 107 atoms. AMS is at least two orders of magnitude more sensitive. The 20 Ma half-life of 236 U limits the utility of α-particle spectroscopy for this isotope. A realistic sensitivity limit of ∼50 µBq corresponds to 5 × 1010 atoms. Limited use of the technique has been made, for example, to search for the presence of depleted uranium resulting from the use of depleted uranium weaponry in Kosovo (Desideri et al., 2002). An additional limitation of α-particle spectroscopy is that the measurement yields only the sum of the 240 Pu and 239 Pu activities, but not the 240 Pu/239 Pu ratio. Since the α-particle energies of the decays of the two isotopes differ by only 11 keV in 5.2 MeV, the two peaks are not resolved at typical detector energy resolutions of 25–40 keV. It is possible to deduce some information from centroid shifts, but since centroids are sensitive to details of sample deposition on the stainless steel discs employed, this method is imprecise at best.
5. Detailed description of AMS methodology 5.1. Negative ion source The negative ion sources used in AMS are almost universally Cs sputter sources. These employ a focused beam of positive Cs ions to sputter material from the sample, and Cs vapor to
270
L.K. Fifield
donate electrons to the sputtered atoms or molecules. The Cs beam is produced by surface ionization of some of the Cs vapor on a hot tantalum “ionizer”. A voltage difference of 5–10 kV maintained between the ionizer and sample serves three functions: to focus the Cs ions on to the sample, to give the Cs ions sufficient energy to sputter effectively, and to extract the negative ions through a hole in the middle of the ionizer (Fifield, 1999). Typical sample masses are 1–10 mg, and the material is pressed into a hole in the sample holder that is typically 1 mm in diameter. The Cs beam is focused to a spot size on the sample of ∼0.5 mm. Samples for 236 U analyses are in the form of uranium oxide which is mixed with aluminum or silver powder to ensure electrical and thermal conduction. Atomic ratios of metal powder to uranium are in the range 1:1 to 4:1. In some applications, particularly in the safeguards area, only very small amounts of uranium are available. In these cases, the sample may be bulked with iron oxide (Hotchkis et al., 2000b; Marsden et al., 2001). In the case of plutonium, macroscopic quantities of the element are not available, and it is necessary to disperse the atoms of plutonium in an iron oxide matrix. Again, aluminum or silver powder is mixed with the iron oxide to make the sample conducting. Intensities of the molecular UO− or PuO− ions are two orders of magnitude greater than the atomic U− or Pu− ions, and hence the former are the ions of choice. Beam currents of ∼100 nA of UO− ions (6 × 1011 ions/s) may be extracted from a uranium oxide sample. The question of efficiency of negative ion formation and extraction in these sources is taken up below under the discussion of sensitivity. 5.2. Low energy mass analysis After extraction from the ion source, the negative ions are pre-accelerated to energies of 20– 160 keV and then mass-analyzed by a 90◦ magnet. Relatively large magnets with radii between 30 and 100 cm are required to bend the very heavy ions. Further, the apertures at the object and image points of the magnet must be chosen to achieve a resolving power M/M ∼ 1000, because it is important to prevent neighboring masses from entering the accelerator. Switching between isotopes, 239 Pu, 240 Pu and 242 Pu for example, may be accomplished either by changing the magnetic field, or by changing the energy of the ions in the magnetic field while keeping the magnetic field fixed. The latter is achieved by applying a voltage to the vacuum box of the magnet, which clearly must be insulated for the purpose. At PuO− energies of ∼50 keV, the voltages required are ∼0.4 and ∼1.2 kV to switch from 239 PuO− to 240 PuO− and 242 PuO− , respectively. 5.3. Acceleration and stripping Accelerators operating at 0.3 to 11 MV have been variously employed at different laboratories. Following mass analysis, the negative ions are accelerated to the high voltage terminal of a tandem accelerator where the molecular ions are dissociated and electrons removed from the resulting atoms to create multiply charged positive ions. This dissociation and stripping process generally takes place in low-pressure oxygen or argon gas, although very thin carbon foils are employed at the Munich laboratory. The now-positive ions, which are distributed across a range of charge states, are further accelerated back to ground potential. A single
Accelerator mass spectrometry of long-lived heavy radionuclides
271
charge state with a well-defined energy is then selected by another large magnet or electrostatic analyzer. At those laboratories employing magnetic analysis, the accelerating voltage is invariably limited by the combination of the bending power of the magnet and the charge state probability distribution. At the ANU 14UD Pelletron accelerator, for example, the analyzing magnet can bend ions with ME/q 2 up to 220 MeV-amu. Here M is the mass of the ion in amu, E the energy in MeV, and q the charge on the ion in units of the electronic charge. In this case, the optimal compromise between transmission and stability is to analyze Pu5+ or U5+ ions. The maximum energy is then ∼23 MeV which is achieved at an accelerating voltage of only ∼4 MV (Fifield et al., 1996). This is well below the 11–14 MV at which the accelerator runs best, and requires that ∼60% of the accelerator be shorted out. The considerations that determine this choice of charge state and hence of accelerating voltage are as follows: – The stripping yield drops off rapidly with increasing charge state. For example, the yield in the 6+ charge state is a factor of two less than in the 5+ , even allowing for the fact that it is possible to use a higher accelerating voltage of ∼4.8 MV at the higher charge state. – Although the stripping yield to the 4+ charge state is higher than that of the 5+ , it would be necessary to operate the accelerator at ∼3 MV in order that the ions could pass around the analyzing magnet. Because transmission through the accelerator is lower at this lower voltage, there is no net gain in using the lower charge state. There would, however, be advantages to using the lower charge states at smaller accelerators. Equivalent considerations apply to the other laboratories where actinide AMS is practiced, but play out differently depending on the available hardware. Laboratory-specific details will be taken up later. 5.4. Final analysis and detection Following the analyzing magnet, it is usual to have an additional analysis that involves an electric field. This may be either an electrostatic analyzer (ESA) or a velocity filter. The latter, also known as a Wien filter, employs crossed electric and magnetic fields to allow only ions with a definite velocity to pass undeflected. Since the analyzing magnet selects ions of constant ME/q 2 , and an ESA or Wien filter selects ions of constant E/q or velocity (E/M) respectively, the combination of the two selects only ions which have the same M/q. An ESA with a 90◦ deflection angle has higher resolution than a Wien filter, but the latter has the advantage that it is simpler to align and can be turned off, to tune the beam for example. Detectors fall into different categories, depending on the species to be detected. For Pu, backgrounds from uranium tend to be low (see the discussion of backgrounds below) because uranium concentrations in the sample can be reduced to very low levels with appropriate chemistry. If uranium background is negligible, then the detector has only to discriminate between Pu ions and lower-energy ions in lower charge states. To take a specific example, if Pu5+ ions are selected, the only background ions that reach the detector are 4+ , 3+ , 2+ and 1+ ions that have the same M/q as the plutonium. Since these have 4/5, 3/5, 2/5 and 1/5 of the plutonium energy, respectively, a simple energy measurement is usually all that is required to discriminate them from the Pu5+ ions. An ionization chamber is the detector of choice on the basis of robustness and adequate energy resolution, although silicon detectors have also been employed.
272
L.K. Fifield
For 236 U, on the other hand, the sample contains unavoidably high levels of 238 U and 235 U, which are potent sources of background, even in an AMS system. The origins of these backgrounds will be discussed in more detail below, but it is sufficient to acknowledge at this point that even with a high-resolution ESA or Wien filter, there are inescapable backgrounds of 238 U and 235 U ions that limit the sensitivity of the 236 U/238 U ratio to ∼10−8 to 10−9 if they are not resolved from 236 U. An ionization chamber does not have sufficient energy resolution to separate the three uranium isotopes since the energies of the 238 U and 235 U ions differ by only −0.8% and 0.4%, respectively, from the 236 U ions. In order to achieve the required resolution, time-of-flight systems with a flight path of ∼2 m have therefore been employed. An energy measurement is still required to separate the 236 U ions from the lower energy ions with the same M/q, since these have the same velocity and hence the same flight time as the 236 U ions. An interesting recent development is the use of calorimetric low-temperature detectors, although these are still some way from routine application. These detectors measure the very small temperature changes produced by the heavy ion as it stops in a sapphire substrate. Temperature changes are manifested as changes in resistance of a thin-film superconducting aluminum strip thermometer operated close to its transition temperature of T ≈ 1.5 K. Such detectors offer the advantage of considerably higher energy resolution compared to the more conventional ionization or silicon detectors. Kraft et al. (2004) reported relative energy resolutions, E/E, in the range 0.4–0.9% for 17 MeV uranium ions. This resolution approaches that, required to separate the different isotopes of uranium, and such detectors may in future be a viable alternative to time-of-flight detectors for 236 U measurements. Each of these detection systems is described in more detail below. 5.4.1. Ionization chambers Ionization chambers are simple and robust, and offer modest resolution that is generally sufficient to resolve lighter-mass ions of different charge q but approximately the same mass/charge ratio (m/q) as the ions of interest. Only a total energy measurement is required. At the ANU, a resolution E/E = 3% is achieved for 23 MeV plutonium ions. Propane is employed as the detector gas at a pressure of 55 Torr, and the ions pass into the detector through a 0.7 µm mylar window with a diameter of 13 mm, losing ∼3 MeV or 13% of their energy in the process. Typical energy spectra are shown in Figure 2. The origin of the nonplutonium peaks in these spectra are considered in the discussion of backgrounds below. This detector is also used for 237 Np and 226,228 Ra. At the other extreme, an ionization detector is also used by the Zurich group (Wacker et al., 2005), but to detect plutonium ions with an energy of only 1.2 MeV. A silicon nitride membrane with a thickness of 40 nm and an area of 3 × 3 mm2 retains the isobutane gas at a typical pressure of 15 mbar. The realization that these ultra-thin silicon nitride membranes made excellent detector windows was crucial to this application since 1.2 MeV plutonium ions would stop in the thinnest available mylar (0.5 µm) but lose only 25% of their energy in the silicon nitride. Considerable attention has also been paid to reducing the electronic noise in this detector by minimizing capacitance and using cooled FET pre-amplifiers (Suter et al., 2007). The noise contribution to the resolution is equivalent to only 12 keV in ion energy. Typical spectra which contrast the performance of a silicon detector with a low-noise ion chamber with two different thicknesses of silicon nitride window are shown in Figure 3
Accelerator mass spectrometry of long-lived heavy radionuclides
273
Fig. 2. Energy spectra from the ionization chamber for (a) 239 Pu, (b) 240 Pu, and (c) 242 Pu from a “typical” environmental sample where the Pu was derived from global fallout. A spike of 4 pg of 242 Pu was added to the sample for normalization purposes. Acquisition times were 2, 3, and 1 min, respectively. The pulser served to monitor dead-time in the electronics and data acquisition system. Note that the 240 Pu spectrum shows peaks at lower energy that arise from fragments of molecules that were dissociated in the gas stripper and subsequently passed the high-energy analysis. These are identified by atomic species, isotope and charge state. Events that fall between the intense 48 Ti peak and the 96 Mo peak are due to pile-up of two 48 Ti ions that arrive within the resolving time of the detector electronics. Careful inspection of the Pu peaks reveals the expected small centroid shifts from one isotope to the next.
(Wacker et al., 2005). Resolution of the ion chamber with a 40 nm silicon nitride window is ∼20% of the signal height. Note, however, that the magnitude of the signal does not represent the full 1.2 MeV of the incoming ions but is reduced by the twin factors of energy loss in the window and pulse height defect. Plutonium ions at these low energies lose a substantial fraction of their energy in the gas, not in collisions with electrons to produce ionization, but by so-called “nuclear scattering”, which is actually an ion-atom scattering process in which the moving plutonium ion transfers energy to a gas atom as a whole. According to the program
274
L.K. Fifield
Fig. 3. The spectra obtained from (a) a surface barrier detector, (b) a gas ionization detector with a 100 nm Si–N window, and (c) a gas ionization detector with a 40 nm Si–N window are shown, for 1.2 MeV 240 Pu3+ (black line) or 0.8 MeV 160 Dy2+ (gray line) ions. The dotted line marks the low level discriminator to cut off virtually all 2+ ions. [From Wacker et al. (2005).]
Accelerator mass spectrometry of long-lived heavy radionuclides
275
TRIM of Ziegler and Biersack (2003), only about 30% of the plutonium energy is transferred directly to ionization. Fortunately, the recoiling gas atoms also create ionization as they slow down in turn, with the net result that ∼70% of the available energy ultimately appears as ionization. Given that the ions lose ∼0.4 MeV in the silicon nitride window, and that ∼70% of what is left is converted to ionization, the signal height is equivalent to the deposition of only about half of the original 1.2 MeV in the detector gas. 5.4.2. Time-of-flight systems Time-of-flight systems (TOF) for AMS consist of “start” and “stop” detectors typically separated by 2 m, and also incorporate a measurement of the ion’s energy. Start detectors must transmit the ions, and hence are all based on the collection of electrons liberated from a thin carbon foil as the ions pass through it. The electrons are accelerated to typically 1 keV, and transported isochronously to a micro-channel plate (MCP) assembly. MCPs may be operated at very high gain of ∼107 , and produce very fast output pulses with rise times of ∼1 ns that are ideal for fast-timing applications. In order to minimize losses of ions due to scattering, the carbon foils should be as thin as possible. Diamond-like carbon foils as thin as 0.5 µg/cm2 (Liechtenstein et al., 2004) are now widely used, although they have the disadvantage that they must be supported on a copper mesh with a transparency of only 75%. Isochronous transport of the electrons has been accomplished in one of two ways: (i) Orienting the carbon foil normal to the incoming ions, and reflecting the electrons through 90◦ with an electrostatic mirror to transport them to the MCP which is out of the ion path. (ii) Orienting the carbon foil at 45◦ to the incoming ions. The MCP can then be mounted parallel to the foil but out of the ion path, and views the upstream side of the foil so that it does not see scattered ions. This system has the advantage that it requires two fewer grids than the mirror device, and one less high-voltage power supply, but has the disadvantage that there is a geometric contribution to the timing resolution due to the 45◦ tilt of the foil. For example, if the beam has a full width at half maximum (fwhm) of 2 mm at the foil, there will be a variation of 2 mm in flight path. This translates into a contribution of 460 ps to the resolution of the TOF system for 23 MeV uranium ions. The stop detector may be either a silicon detector, in which case it also provides the measurement of the ion’s energy, or another MCP detector. In order to collect as many as possible of the ions that passed through the start detector, active areas of at least 25 mm in diameter are advantageous. Because of the large area, only the mirror design of MCP detector is suitable. If the MCP option is adopted, then it must be followed by either an ionization chamber or silicon detector to provide the energy information. Although intrinsic resolutions of 0.3–0.4 ns have been reported for such TOF systems (Steier et al., 2002), in practical situations the observed resolution is generally about 1 ns. As can be seen from Table 1, this resolution is sufficient to separate 236 U from 235 U and 238 U ions of the same magnetic rigidity provided that count rates of the latter are low. This is illustrated by Figure 4, which shows a two-dimensional plot of energy vs TOF and the one-dimensional projection on to the TOF axis for a sample with 236 U/238 U = 6 × 10−11 . A drawback of a TOF system is that its efficiency is typically only 30–50% due both to losses on the multiple grids and foil-supporting meshes of the start and stop detectors, and to losses due to beam divergence and scattering in the foil of the start detector.
276
L.K. Fifield
Table 1 Flight times over a 2 m flight path for 5+ uranium ions of different masses but the same magnetic rigidity. Energies are typical of those used at the ANU Ion
Energy (MeV)
Flight time, T (ns)
T (ns)
235 U
23.995 23.893 23.692
518.3 520.5 525.0
−2.2
236 U 238 U
4.4
Fig. 4. (a) Two-dimensional spectrum of energy deposited in the Si detector vs TOF for a uranium ore sample with a 236 U/238 U ratio of 6 × 10−11 . (b) One-dimensional projection of (a) on to the TOF axis.
6. Implementation at various laboratories 6.1. ANU At the ANU, the accelerating voltage of ∼4 MV and the choice of the 5+ charge state are dictated by the maximum mass–energy product (220 MeV-amu) of the 90◦ analyzing magnet. A Wien filter is employed to remove backgrounds which have the same ME/q 2 as the ions
Accelerator mass spectrometry of long-lived heavy radionuclides
277
Fig. 5. AMS system for the measurement of actinides at Lawrence Livermore National Laboratory. [After Brown et al. (2004).]
of interest but different velocities. Switching between different isotopes is accomplished by changing (i) the field in the injection magnet, (ii) the terminal voltage of the accelerator, and (iii) the electric field of the Wien filter. Switching times in this slow-cycling procedure are 15 s. For plutonium, measurement times are generally 1 min at 242 Pu, 3 min at 240 Pu and 2 min at 239 Pu. This sequence is repeated as many times as necessary, with 3 loops being typical. For 236 U, each loop consists of integration of the 238 U beam current for 10 s, and counting of 236 U ions for 5 min. Detection systems are an ionization chamber for 239 Pu, 237 Np or 226,228 Ra, and a TOF system for 236 U. Only the total energy signal from the ionization detector is recorded, since it is necessary to distinguish only between the 5+ Pu ions with energy of ∼23 MeV, and ions in lower charge states with 4/5, 3/5, 2/5 and 1/5 of this energy. 6.2. Lawrence Livermore National Laboratory At Livermore, an analyzing magnet with a deflection angle of only 30◦ followed by a 45◦ ESA allows the use of a higher terminal voltage of 6.5 MV. This arrangement is depicted in Figure 5. Again, gas-stripping is employed and the 5+ charge state is selected. The detection system for plutonium, 237 Np, and 236 U is a two-anode longitudinal-field ion chamber (Brown et al., 2004; McAninch et al., 2000). Switching between isotopes is accomplished with a fast-switching system. Different voltages are applied to the vacuum box of the injection magnet in order to inject the various isotopes into the accelerator. An electrostatic steerer after the 30◦ analyzing magnet directs the different isotopes through the image slits of the magnet and into the ESA. The all-electric switching can be fast, ∼10 µs. Measurement times are 0.1 s for 242 Pu and 0.4 s for 239 Pu and 240 Pu. Cycling between the 242 Pu–239 Pu pair continues for 10 s before changing to the 242 Pu–240 Pu pair for the next 10 s.
278
L.K. Fifield
Fig. 6. High-energy actinide beam-line at the ANTARES facility at the Australian Nuclear Science and Technology Organisation (ANSTO). [Reprinted with permission from Hotchkis et al. (2000a).]
6.3. ANSTO The operating conditions at ANSTO are essentially the same as at the ANU, i.e. the accelerating voltage is 4 MV and the 5+ charge state is selected. There the similarities end, however. A heavy-element beam-line consisting of a high-resolution, 2.5 m radius, 90◦ ESA followed by a large (2.0 m radius) 90◦ magnet (Figure 6) has been constructed to analyze the ∼24 MeV heavy ions after acceleration (Hotchkis et al., 2000a, 2000b). The advantage of having the magnet as the final element in the system is that, for a given charge state, it disperses in mass following the E/q selection of the ESA. Hence the focal plane of the magnet can be instrumented at specific locations with various detection systems tailored to the intensity of the isotopes to be measured. An ionization chamber is employed for Pu measurements, which is preceded by a TOF system, comprised of two micro-channel plate detectors for 236 U measurements. The more intense 238 U, 235 U or 234 U beams can be measured by either Faraday cups or an electron multiplier, as shown in Figure 7. 6.4. VERA Despite the lower terminal voltage of 3 MV of the VERA accelerator, the 5+ charge state is selected because this is the minimum charge that the 90◦ analyzing magnet can bend (Steier
Accelerator mass spectrometry of long-lived heavy radionuclides
279
Fig. 7. Detector instrumentation at the focal plane of the high-energy actinide beam-line at ANSTO as in 2000. [Reprinted with permission from Hotchkis et al. (2000a).] Note that this has subsequently evolved.
et al., 2002). The yield for stripping 238 U16 O− to 238 U5+ was measured to be 6%. A highresolution 90◦ ESA after the analyzing magnet strongly suppresses ions with different M/q ratios. Any residual background ions are identified by a TOF system consisting of two microchannel plate detectors. This is backed by an ionization chamber to discriminate against lower charge state ions of the same M/q. The overall efficiency of the detection system is 20–30%. A slow-cycling procedure is employed for plutonium analyses. Switching times between isotopes are 15 s and measurement times range between 20 and 100 s (Hrnecek et al., 2005; Winkler et al., 2004). For 236 U, a quasi-fast switching procedure has been adopted. Ion source fluctuations are monitored by fast switching of the 238 U16 O− beam into the low-energy offaxis Faraday cup during post-acceleration measurement of both the 238 U beam current and 236 U ions. Due to the small fractional mass difference between 238 U and 236 U, it is, however, necessary to change the accelerator voltage when switching between isotopes at the high-energy side. This takes ∼15 s. Hence, the 238 U5+ beam current is measured only every 5 min. This frequency is quite adequate to monitor drifts in stripping yield (Steier et al., 2002; Vockenhuber et al., 2003). 6.5. Munich The Munich facility is blessed with a very large analyzing magnet with a mass–energy product of 350 MeV-amu. This makes the use of foil stripping at a higher terminal voltage a viable option. An accelerating voltage of 12.5 MV is employed, and the 11+ charge state selected. Ions are detected by a TOF system consisting of a MCP “start” detector and a “stop” detector which combines an ionization chamber to provide a E signal and a silicon detector to provide both the timing signal and a residual-energy signal. Details may be found in Wallner et al. (2000). 6.6. ETH, Zurich At Zurich, it has been shown that AMS of plutonium can be performed almost equally well on a small accelerator operating at only 0.3 MV (Fifield et al., 2004; Wacker et al., 2005). The
280
L.K. Fifield
Table 2 Relative negative-ion formation probabilities for oxide ions of the actinides, normalized to Pu Negative ion
Relative formation probabilitya
ThO− UO− NpO− PuO−
0.15 ± 0.01 0.43 ± 0.04 0.77 ± 0.02 1.00
a Absolute negative ion yields can be derived from this table using the observation that the yield for uranium oxide is
0.3%.
1.2 MeV Pu ions are analyzed by a combination of 90◦ magnetic and electric analyzers, and detected in a gas ionization detector with a 40 or 50 nm thick silicon nitride window. Since the Pu ions lose only 300–400 keV in this window, the resolution of the detector is substantially better than a silicon detector, and is sufficient to resolve the 3+ Pu ions from 2+ and 1+ ions with the same M/q, as shown in Figure 3. It appears that this system may even be suitable for high-sensitivity 236 U measurements (Wacker et al., 2005). Perhaps paradoxically, this small system has the highest transmission of any AMS system for actinides. Stripping yield to the 3+ charge state is surprisingly high at 20% and this, combined with minimal losses after stripping allow the Zurich system to achieve transmission as high as 15% from ion source to detector. By contrast, the corresponding figure for systems using the 5+ charge state is ∼3%. 6.7. Weizmann Institute Techniques for analyzing both 236 U and plutonium isotopes have been developed by Paul et al. at the 14UD accelerator of the Weizmann Institute (Berkovits et al., 2000; Paul et al., 2001, 2003). Foil stripping to the 9+ or 11+ charge state at a terminal voltage of 7.1 MV was employed. Transmission of 0.1–0.2% is significantly less than at facilities where gas stripping is used. The detection system provides both time-of-flight and energy information for each ion. 6.8. New facilities New AMS facilities based on 3 and 1 MV tandems have recently been installed in Naples and Seville, respectively. In both cases, the injection and analyzing magnets have been specified with future actinide measurements in mind. 7. Efficiency Efficiency is the product of negative ion yield and transmission. Negative ion yield has been measured to be 0.3% for uranium by monitoring the current of 238 U16 O− ions from a sample containing a known amount of uranium as it was run to exhaustion (Fifield et al., 1996). Relative yields for Th, U, Np and Pu are listed in Table 2 and were determined using samples containing these elements in known proportions (Fifield et al., 1997).
Accelerator mass spectrometry of long-lived heavy radionuclides
281
Transmission from ion source to detector varies from system to system. At ANU, transmission in the 5+ charge state is ∼3%. From Table 2, the overall efficiency for detecting a uranium ion is ∼10−4 . For Pu, it is about a factor of two higher. Since backgrounds are very low, a signal of 10 counts is readily observable. Hence, sensitivities of ∼105 atoms of 236 U, 237 Np or Pu are achievable with AMS.
8. Backgrounds 8.1. Plutonium Although AMS of plutonium has intrinsically very low background levels, ions other than those of interest are also detected. These fall into two categories: 1. Other 5+ ions with very similar energies. Of these, the most serious is 238 U, although 235 U or 232 Th are also possibilities. After selection by the analyzing magnet, 238 U5+ and 239 Pu5+ differ in energy by only 0.4%. This is well below the 3% resolution of the ionization detector. Those laboratories with a high-resolution ESA may be able to prevent the 238 U ions from reaching the detector, but a Wien filter will not in general deflect the 238 U5+ ions sufficiently to allow the interception of all of the 238 U ions before the detector. Hence, it is important to understand the origin of the 238 U ions and the likely extent of the problem. Uranium-containing negative molecular ions of mass 255, of which the most important is 238 U17 O− , are injected into the accelerator along with the 239 Pu16 O− ions. Similarly, 238 U18 O− ions are injected with 240 Pu16 O− . After molecular dissociation and stripping in the high-voltage terminal and subsequent acceleration, 238 U5+ and 239 Pu5+ ions differ in momentum by 0.25%. These particular 238 U5+ ions are not a problem because they are readily eliminated by the analyzing magnet, which is typically operated at a resolution of 0.1% or better in momentum. The problematic 238 U5+ ions are rather those that have the same momentum as the 239 Pu5+ ions and that therefore pass around the analyzing magnet. To satisfy this condition, the 238 U5+ must gain an extra 0.4% of energy. Some of the ions that are stripped to 6+ in the terminal can undergo charge-changing collisions and change to 5+ . If these collisions occur ∼3% of the way down the high-energy tube, the 238 U ions will have the correct energy to pass around the analyzing magnet. The probability of this process is critically dependent on the vacuum in the high-energy tube—the better the vacuum, the lower the probability. Since gas-stripping is employed, a recirculating gas stripper with additional differential pumping to minimize the gas entering the high-energy tube is crucial. Child et al. (2005) have explored this 238 U background as a function of the amount of uranium added to a blank sample, and found that even 1 ng of uranium (i.e. about 1 ppm in a 1 mg sample) leads to an apparent 239 Pu signal of 10 fg. A sample to which no uranium was added yielded an apparent 239 Pu signal of ∼3 fg due to residual uranium in the ion source or intrinsic to the sample. Their results are shown in Figure 8. A similar (unpublished) study performed at the ANU showed a sensitivity to uranium concentration that was two orders of magnitude lower than reported by Child et al. (2005), i.e. 100 ng of uranium was
282
L.K. Fifield
Fig. 8. Apparent Pu in fg due to background 238 U ions that are indistinguishable from 239 Pu ions in the detector, as a function of uranium concentration in the sample. [Reprinted with permission from Child et al. (2005).]
required to produce an apparent 239 Pu signal of 10 fg. This is borne out by regular measurements at the ANU on process blanks which routinely exhibit an apparent 239 Pu signal of <0.2 fg. This lower sensitivity to uranium highlights the importance of the vacuum in the high-energy acceleration tube close to the high-voltage terminal. In the ANU’s 14UD accelerator, this is typically better than 10−8 Torr due to the excellent pumping of the gas stripper and the intrinsically high vacuum of the NEC tube design. Uranium backgrounds at 240 Pu can be higher than at 239 Pu because the 238 U18 O− ion is five times more prolific than the 238 U17 O− ion. If these oxide molecules, rather than the 238 U16 OH− ion or the high-energy tail of the 238 U16 O− ion distribution, are the principal source of 238 U ions in the accelerator, then the 238 U background at 240 Pu will be approximately five times higher than at 239 Pu. There is some evidence that this is the case, at least for the ANU system (Fifield et al., 1996). 2. The second category of non-Pu ions that may arrive at the detector is ions with lower charge states but approximately the same M/q. To consider a specific example, any 191 Pt4+ ions with the same ME/q 2 as 239 Pu5+ ions will be only slightly deflected by a Wien filter (or will pass around an ESA) and will therefore reach the ionization detector when the AMS system is set to transmit 239 Pu. Because they have only 4/5 of the energy of the 239 Pu ions, they are readily distinguished by the ionization chamber, and hence are only a problem if counting rates are high. They have their origin in injection of molecular negative ions of mass 255. Possibilities involving common elements would be 191 Pt16 O4 , 191 Pt48 Ti16 O, or 191 Pt32 S16 O . Because M/q for 191 Pt4+ ions is not exactly the same as for 239 Pu5+ ions, 2 a charge-changing collision from 5+ to 4+ in the high-energy tube is necessary to give the 191 Pt the correct energy to traverse the analyzing magnet. Taken together, the low natural abundance of platinum, the complexity of the molecules, and the low probability of the charge-changing collision result in extremely low counting rates of 191 Pt4+ ions at the detector. This is equally true of the lower charge state ions 143 Sm3+ , 96 Mo2+ and 48 Ti+ .
Accelerator mass spectrometry of long-lived heavy radionuclides
283
Counting rates of these background ions are therefore low, and typical spectra at the 239 Pu settings are very clean, as can be seen in Figure 2a. Similar considerations apply to 242 Pu, for which a typical spectrum is shown in Figure 2c. The situation is, however, rather different for 240 Pu5+ . In this case, M/q is an integer. Ignoring differences in binding energies for the moment, it follows that there is no need to invoke a charge-changing collision in the high-energy tube. Atoms stripped to 48 Ti+ , 96 Mo2+ , 144 Sm3+ , and 192 Pt4+ in the gas stripper alone will have the correct energy to pass around the analyzing magnet. Hence, counting rates of these ions can be orders of magnitude higher at the 240 Pu settings than at 239 Pu or 242 Pu settings, as illustrated by Figure 2b. Nevertheless, counting rates are generally well within the capabilities of the detector. Note, however, that 240 Pu is rather less tightly bound than the lighter interferences, i.e. M/A for 240 Pu is 0.12% greater than 96 Mo for example. As a result, 240 Pu5+ and 96 Mo2+ do not follow exactly the same average trajectories in the analyzing magnet. With appropriate apertures, the counting rates of the M/q interferences can be reduced substantially, especially when the analyzing magnet is followed by a high-resolution ESA. 8.2. Uranium-236 AMS of 236 U must contend with the fact that uranium is a major constituent of the sample. Backgrounds from 238 U and 235 U, which arise in essentially the same way as the 238 U background in the plutonium case described above, will therefore be a more serious problem than was the case for plutonium. At the 236 U settings, the mass-252 236 U16 O− ion is selected for injection into the accelerator. Other ions of the same mass, but with either 238 U or 235 U as a constituent, are 238 U14 N− , 238 U12 CH− , 235 U17 O− , and 235 U16 OH− . Of these, 235 U17 O− is unavoidable and will be 2 injected with a beam current of ∼0.3 pA (∼2 × 106 ions/s) when the 238 U16 O− output is 100 nA. Experience with plutonium suggests that the U16 OH− beam is generally weaker than the U17 O− beam. Measurements on test samples at the ANU showed that, although the 238 U14 N− ion can be a significant source of background if the sample is in the form of uranium nitrate, the normal procedure of baking the samples at 800 ◦ C is very effective at eliminating nitrogen. Hence the 238 U14 N− ion does not contribute materially to the 238 U background for “real” samples. Further, the addition of 1 mg of graphite to a uranium oxide sample did not increase the 238 U background, which implies that the 238 U12 CH− 2 ion is not an important source of background either. In fact, the principal source of background appears to arise from the low-energy tail of the 238 U16 O− beam. Specifically, if a 238 U16 O− ion has 0.8% less energy than normal, it will be accepted by the injector magnet and injected into the accelerator. The most likely mechanism for producing such lower-energy ions is the formation of “hot” molecules which break up during the extraction process in the ion source (Litherland, 1987). Similarly, 235 U16 O− ions with 0.4% more energy than normal will also be accepted. Some ions can gain extra energy from the backscattered Cs ions during the sputtering process. A detailed discussion of the various mechanisms is given by Litherland (1987) in the context of carbon beams. Although the relative contributions of the different sources of background are difficult to quantify, it is
284
L.K. Fifield
believed that these energy-degraded 238 U16 O− ions are the dominant source of background in the ANU system. Some AMS systems have an energy-analyzing ESA between the ion source and the injector magnet. Provided that this has sufficiently high resolution, the energy-degraded 238 U16 O− or energy-enhanced 235 U16 O− ions can be eliminated before the injector magnet. A resolution E/E > 250 would be required to eliminate both the 238 U16 O− and 235 U16 O− ions.
9. Sample preparation Sample preparation methods have been adapted from techniques developed for alpha-particle counting. We shall briefly discuss only methods for plutonium. There are three stages: (i) Initial scavenging of the plutonium from the original sample. If a 242 Pu spike is required, it is added at this stage. Four possibilities can be distinguished: (a) Soils and sediments. The plutonium will generally reside on the surfaces of the grains, and hence can be liberated by leaching in hot 8M nitric acid (Tims et al., 2004). (b) Water, including ice. Plutonium is co-precipitated with either Fe(OH)3 or MnO2 by adding a suitable amount of Fe or Mn and then making the solution alkaline with NaOH. (c) Biological samples, e.g., blood, faeces or urine. The Pu is generally co-precipitated with either calcium phosphate or calcium/magnesium phosphate. (d) Rocks, and in particular uranium ores (for measurement of Pu produced in situ by neutron absorption by 238 U). After crushing and sieving, the <250 µm fraction is dissolved in a 40:60 mixture of concentrated HF and HNO3 . This effectively dissolves any uranium-bearing minerals, including silicates. Often, a residue of graphite and sulfides remains after dissolution, but these contain very little of the uranium, or by implication, of the plutonium. Insoluble fluorides are also precipitated. Since PuF4 is insoluble, it is necessary to take the fluorides back into solution with HNO3 and to drive off the fluorine as HF by repeatedly drying down from nitric acid solution. (ii) Separation of the actinides from everything else. After taking the scavenged sample back into solution, this separation is usually accomplished with U-TEVA™ (uranium tetravalent) ion-exchange resin (Eichrom Industries, Chicago). The sample is loaded on to the column in 3M nitric acid, and only uranium and other tetravalent actinides are retained on the column as nitrate complexes. Plutonium and thorium can then be selectively eluted by a mixture of 5M hydrochloric and 0.05M oxalic acids. Uranium is subsequently eluted with dilute 0.01M hydrochloric acid. (iii) Plutonium may be separated from any residual uranium, and less importantly from thorium and other actinides, on an anion-exchange column, Bio-Rad AG 1-X8 for example. The plutonium is loaded on to the column in 8M nitric acid and any residual uranium washed through with additional acid. Thorium can then be eluted with 12M hydrochloric acid. Finally, plutonium is eluted from the resin with a freshly made, warm (40 ◦ C) mixture of 12M HCl and 0.1M NH4 I. The iodide ion reduces Pu to its trivalent state, breaking the anionic complex and releasing the Pu from the column.
Accelerator mass spectrometry of long-lived heavy radionuclides
285
At the end of this procedure, the plutonium and very little else is in solution. In order to produce a sample for the ion source, considerably more bulk than would be provided by a few pg of Pu is, however, required. Almost invariably, this has been provided by iron oxide. Typically 1–2 mg of iron as Fe(NO3 )3 is added to the plutonium-containing solution. The solution is then either simply evaporated to dryness, or iron is precipitated as Fe(OH)3 by making the solution alkaline. The resulting solid is converted to Fe2 O3 by baking at 800 ◦ C, mixed with a conductor that has variously been aluminum, silver, copper or niobium, and pressed into a sample holder. In contrast to plutonium, where efficiency is paramount and where it is important to ensure valence equilibrium between the 242 Pu spike and Pu in the sample, the chemistry for extracting uranium is more forgiving in the sense that the principal requirement is only that enough uranium be extracted for a measurement, Since no spike is added, efficiency of extraction and equilibration are not serious issues.
10. Applications 10.1. Tracing discharges from plutonium-processing and fuel reprocessing plants In a series of papers, Oughton and co-workers have employed AMS to measure plutonium concentrations and plutonium and uranium isotopic ratios in regions affected by the nuclearweapons production complex of the former Soviet Union (Figure 9). Specifically (i) In the vicinity of the Mayak production and processing facility in the Urals, significant releases of plutonium have contaminated the local environment. In the 1950’s, waste containing weapons grade plutonium with low 240 Pu/239 Pu ratios (<5%) was released directly to the Techa River. The Techa is a tributary of the Ob river which is one of the two major rivers draining Siberia to the Arctic Ocean. Subsequently, containment dams were built to intercept waste, and the plant reprocessed an increasing proportion of waste from civil nuclear power reactors. Hence 240 Pu/239 Pu and 236 U/238 U ratios change over time. In addition, an explosion in a high-level waste tank in 1963 sent a plume of radioactive material to the northeast of the plant. Plutonium measurements have been carried out in the swamp just downstream of the containment dams, which would have received some of the earliest releases, in water and sediments from the dams themselves, in soils and sediments down the Techa River and from areas affected by the Khyshtym accident, and in present-day water from the river (Borretzen et al., 2005; Oughton et al., 2000). In addition, 236 U/238 U ratios have been measured in both the swamp and the dams (Borretzen et al., 2005). Taken together, these give a rather complete picture of the quantities and origins of plutonium in the environs of the plant, and the extent to which it has dispersed. In particular, plutonium from the plant’s operations is barely detectable in the river at distances greater than 200 km from the plant. (ii) In contrast, measurements in sediments and water from the other major Siberian river, the Yenisey, point to an influence of the Krasnoyarsk processing plant. Located 1000 km upstream from the estuary, its influence is evidenced by low 240 Pu/239 Pu ratios in both water and sediment from the upper reaches of the estuary (Lind et al., 2006; Skipperud
286
L.K. Fifield
Fig. 9. Map of the Ob and Yenisey rivers showing sources of weapons’ related contamination. The squares represent plutonium production and reprocessing facilities, and the stars nuclear weapon test sites. [From Skipperud et al. (2004).]
et al., 2004). This influence extends out into the Kara Sea, and back into the adjacent estuary of the Ob River. (iii) Dumping of nuclear reactors from decommissioned submarines and other nuclear waste has occurred near Novaya Zemlya in the Russian Arctic. Ratios of 240 Pu/239 Pu have been employed to monitor possible leakage from these dumping sites (Oughton et al., 2004).
Accelerator mass spectrometry of long-lived heavy radionuclides
287
AMS has also been employed to study the dispersal of 237 Np and 236 U discharged from the Sellafield reprocessing plant in Cumbria, UK. Keith-Roach et al. (2000) measured the seasonal variation of 237 Np in pore water in a nearby salt marsh, and compared it with Pu and 241 Am in order to investigate the influence of microbial activity in mobilizing these isotopes. Marsden et al. (2001, 2006) measured 236 U/238 U ratios as a function of depth in sediments from the same marsh to complement radiometric measurements of concentrations of Pu isotopes, 241 Am and 244 Cm. Comparison with the documented history of releases from the Sellafield plant yields important information on the mechanisms that disperse the discharges into the environment. 10.2. Human biochemistry of plutonium A major benefit of the high sensitivity of AMS is that it allows biomedical studies on human subjects because doses of long-lived radionuclides can be kept very low. Plutonium is a particular case in point, since its isotopes are α-particle emitters and the element is chemically toxic. Nature has been kind in that the longest-lived plutonium isotope, 244 Pu, has a half-life of 80 Ma and hence a very low specific activity. Further, although it can be extracted from spent reactor fuel, it was produced in only negligible quantities by nuclear weapons testing. Hence, the combination of low specific activity and absence from the natural environment make it a useful tracer of plutonium uptake, retention and excretion in human subjects. Two studies in which AMS has been used to detect the 244 Pu in blood, urine and fecal samples have been reported (Etherington et al., 2003; Newton et al., 2005; Stradling et al., 2002). In the first (Newton et al., 2005), a group of six women were given a dose of mixed 237 Pu and 244 Pu by intravenous injection in 1995. It was necessary to purify the 244 Pu by massseparation so that only traces of the other plutonium isotopes remained, in order to minimize the radiological burden to the subjects. The 237 Pu, with its half-life of 37 days, allowed the short term behavior of the dose to be followed by γ -ray counting of the 100 keV X-rays that are emitted following its β-decay to 237 Np. Subsequently, it has been possible to follow the decline in plutonium in the subjects out to 8 years by measuring 244 Pu in blood and urine samples by AMS. Even after 8 years, 244 Pu is still detectable in a 20 ml blood sample. Results are shown in Figure 10, where they are compared with an extensively modified revision (Leggett et al., 2005) of the ICRP’s current biokinetic model for systemic plutonium (ICRP, 1993). The second study determined the uptake of plutonium from the lungs when the plutonium was attached to aerosols (Etherington et al., 2003; Stradling et al., 2002). This was designed to mimic the most probable pathway for uptake by workers in nuclear processing plants. 10.3. Safeguards A potential application of AMS in the nuclear safeguards area is the identification of clandestine nuclear activities by measuring 236 U in small particles. These may have been collected as aerosols or from the leaves of plants, and elevated 236 U/238 U ratios would constitute unambiguous evidence of material that had passed through a nuclear reactor (Hotchkis et al., 2000a, 2000b). A closely related application is the identification of areas contaminated by depleted uranium weaponry. Armour-piercing shells made of depleted uranium were used in the recent Iraq and
288
L.K. Fifield
Fig. 10. (a) Concentrations (% of injection kg−1 ) of plutonium in whole blood from six human female subjects as a function of time after injection with a mix of 237 Pu and 244 Pu. The line indicates the central tendency of the data. Data out to 60 days were obtained from γ -ray counting of 237 Pu decays. Subsequent data points, which extend out to 8 years after injection, were derived from AMS measurements of 244 Pu. (b) Concentrations from a revised version (Leggett et al., 2005) of the ICRP’s current biokinetic model for systemic plutonium (ICRP, 1993), compared with the estimated ±95% range in adult females. [From Newton et al. (2005).]
Kosovo wars, and there are concerns about possible health effects on both military personnel and the civilian population of the affected areas. Depleted uranium is a by-product of the uranium-enrichment process, and the feedstock often contains some recycled reactor fuel with high levels of 236 U. Hence, some 236 U finds its way into the depleted uranium and can serve as a very sensitive fingerprint of the presence of depleted uranium in an area (Danesi et al., 2003). Again, the high sensitivity of AMS is required for its detection. 10.4. Plutonium as a tracer of soil erosion and sediment transport Plutonium has advantages in terms of sample size, sensitivity and precision over the widely used 137 Cs as a tracer of soil erosion and sediment transport over the past 50 years. Both were
Accelerator mass spectrometry of long-lived heavy radionuclides
289
produced by nuclear weapons testing in the 1950s and 60s, and subsequently dispersed globally. Both bind tightly to soil and sediment particles, especially in terrestrial environments, and hence may be used to track the erosion of soil particles and their subsequent transport by river systems to the sea. Approximately 5 times as many atoms of Pu as of 137 Cs were produced. Decay of 137 Cs in the intervening ∼45 years has increased this ratio to ∼14. In addition, there is some evidence that, in contrast to Pu, Cs desorbs from sediments in a saline environment. Whereas 137 Cs is readily detected via the 662 keV γ -ray emitted following its β-decay to 137 Ba, AMS is the method of choice for the detection of Pu in sediments, although some work has been done with α-particle counting. Sample sizes for AMS are typically 4 g of soil or sediment, and the same number of counts can be achieved for 239 Pu in 7 min of counting as is achieved for 137 Cs in 2 days of counting of a 100 g sample. Further, since the 137 Cs γ -ray peak sits on a background which must be subtracted, the statistical precision of the essentially backgroundfree Pu measurements is higher. AMS counting times of ∼1 h are quite feasible, which allows even higher precision to be achieved if desired. Sample preparation is somewhat more complex for Pu than for 137 Cs, but is not especially difficult. Surface-bound plutonium is leached from the grains with hot 8M nitric acid, and the Pu is subsequently separated from uranium and other elements on an anion-exchange column and then dispersed in Fe2 O3 as described above. A spike of typically 4 pg of 242 Pu is added before leaching for normalization purposes. A pilot study is underway in the 10,000 km2 Herbert River catchment in NE Queensland, Australia to determine the relative merits of Pu and 137 Cs. This catchment covers grazing land in its upper reaches, undisturbed forest in the middle reaches, and intensive sugar-cane agriculture on the coastal plains. Soils and sediments collected from each of these land-use types show a very good correlation between Pu and 137 Cs, as shown in Figure 11. 10.5. Ores Although 239 Pu and 236 U are generally thought of as exclusively man-made isotopes, they do in fact occur naturally in uranium ores, albeit at very low levels. Indeed, Seborg first reported the detection of natural plutonium in an ore from Cigar Lake in Saskatchewan as early as 1948. The processes that produce these isotopes are the same as in a reactor, i.e. neutron capture by 238 U and 235 U, respectively. The neutrons are generated principally by spontaneous fission of 238 U and by (α, n) reactions on light elements such as Na, Mg and Al. In a typical ore-bearing rock, these two contributions are approximately equal. Potential applications include neutron flux monitoring (Purser et al., 1996), fingerprinting individual ores, and exploration for new uranium deposits. Neutron fluxes in an ore body are determined not only by the uranium content, but also by the water content, the concentrations of major neutron-producing elements such as Na, Mg and Al, and by the concentrations of trace neutron absorbers such as B, Gd and Sm. Since 239 Pu is produced by absorption of epithermal neutrons, whereas 236 U is produced largely by capture of thermal neutrons, it is possible to deduce both the thermal and epithermal neutron fluxes in the ore over the past 50 ka. A particular concern of uranium-producing countries is that their uranium might find its way into nuclear weapons. Hence a fingerprint of a particular ore-body, or even of a particular
290
L.K. Fifield
Fig. 11. Plot of 137 Cs concentration vs 239 Pu concentration for terrestrial samples from a range of land-use areas in the Herbert river catchment, N. Queensland, Australia. [From Everett et al. (2007).]
shipment of ore, would allow that specific uranium to be tracked. Any processing of the ore is not going to change the uranium isotopic composition, and hence the 236 U/238 U ratio might serve as such a fingerprint. The presence of detectable amounts of 236 U in high-uranium groundwater would be a strong indicator that the uranium was derived from an ore-body. Hence, the 236 U/238 U ratio may prove to be a useful exploration tool in the search for new uranium deposits.
11. Summary Accelerator mass spectrometry is proving to be an exquisitely sensitive tool for studying longlived actinides in the environment, and an increasing number of AMS laboratories are adding an actinide capability to their repertoire. Applications range across a broad spectrum. Isotopes of plutonium are finding application in tracing the dispersal of releases from nuclear accidents and reprocessing operations, in studies of the biokinetics of the element in humans, and as a tracer of soil loss and sediment transport. Uranium-236 has also been used to track nuclear releases, but also has a role to play in nuclear safeguards and in determining the extent of environmental contamination in modern theaters of war due to the use of depleted uranium weaponry.
Accelerator mass spectrometry of long-lived heavy radionuclides
291
References Bergquist, B.A., Marchetti, A.A., Martinelli, R.E., McAninch, J.E., Nimz, G.J., Proctor, I.D., Southon, J.R., Vogel, J.S. (2000). Technetium measurements by accelerator mass spectrometry at LLNL. Nucl. Instrum. Methods B 172, 328–332. Berkovits, D., Feldstein, H., Ghelberg, S., Hershkowitz, A., Navon, E., Paul, M. (2000). 236 U in uranium minerals and standards. Nucl. Instrum. Methods B 172, 372–376. Borretzen, P., Standring, W.J.F., Oughton, D.H., Dowdall, M., Fifield, L.K. (2005). Plutonium and uranium atom ratios and concentration factors in Reservoir 11 and Asanov Swamp, Mayak PA: An application of accelerator mass spectrometry. Environ. Sci. Technol. 39, 92–97. Brown, T.A., Marchetti, A.A., Martinelli, R.E., Cox, C.C., Knezovich, J.P., Hamilton, T.F. (2004). Actinide measurements by accelerator mass spectrometry at Lawrence Livermore National Laboratory. Nucl. Instrum. Methods B 223–224, 788–795. Child, D.P., Hotchkis, M.A.C., Williams, M.L. (2005). Improvements in actinide and fission product analysis by AMS—Raising the bar and lowering detection limits in heavy element AMS. In: Proceedings of the 46th Annual Meeting of the Institute for Nuclear Materials Management, Arizona, USA, July 10–14, 2005. Danesi, P.R., Bleise, A., Burkart, W., Cabianca, T., Campbell, M.J., Makarewicz, M., Moreno, J., Tuniz, C., Hotchkis, M. (2003). Isotopic composition and origin of uranium and plutonium in selected soil samples collected in Kosovo. J. Environ. Radioact. 64, 121–131. Desideri, D., Meli, M.A., Roselli, C., Testa, C., Boulyga, S.F., Becker, J.S. (2002). Determination of 236 U and transuranium elements in depleted uranium ammunition by α-spectrometry and ICP-MS. Anal. Bioanal. Chem. 374, 1091–1095. Etherington, G., Stradling, G.N., Hodgson, A., Fifield, L.K. (2003). Anomalously high excretion of Pu in urine following inhalation of plutonium nitrate. Radiat. Protect. Dosim. 105, 321–324. Everett, S.J., Tims, S.G., Hancock, G.J., Fifield, L.K. (2007). Comparison of Pu and 137 Cs as tracers of sediment transport in terrestrial environments. J. Environ. Radioact., submitted for publication. Fehn, U., Snyder, G.T. (2005). Residence times and source ages of deep crustal fluids: interpretation of 129 I and 36 Cl results from the KTB-VB drill site, Germany. Geofluids 5, 42–51. Fehn, U., Snyder, G., Egeberg, P.K. (2000). Dating of pore waters with 129 I: Relevance for the origin of marine gas hydrates. Science 289, 2332–2335. Fehn, U., Snyder, G.T., Matsumoto, R., Muramatsu, Y., Tomaru, H. (2003). Iodine dating of pore waters associated with gas hydrates in the Nankai area, Japan. Geology 31, 521–524. Fifield, L.K. (1999). Accelerator mass spectrometry. Rep. Prog. Phys. 38, 136–186. Fifield, L.K., Cresswell, R.G., di Tada, M.L., Ophel, T.R., Day, J.P., Clacher, A.P., King, S.J., Priest, N.D. (1996). Accelerator mass spectrometry of plutonium isotopes. Nucl. Instrum. Methods B 117, 295–303. Fifield, L.K., Clacher, A.P., Morris, K., King, S.J., Cresswell, R.G., Day, J.P., Livens, F.R. (1997). Accelerator mass spectrometry of the planetary elements. Nucl. Instrum. Methods B 123, 400–404. Fifield, L.K., Carling, R.S., Cresswell, R.G., Hausladen, P.A., di Tada, M.L., Day, J.P. (2000). Accelerator mass spectrometry of 99 Tc. Nucl. Instrum. Methods B 168, 427–436. Fifield, L.K., Synal, H.-A., Suter, M. (2004). Accelerator mass spectrometry of plutonium at 300 kV. Nucl. Instrum. Methods B 223–224, 802–806. Gascard, J.C., Raisbeck, G., Sequeira, S., Yiou, F., Mork, K.A. (2004a). Correction to “The Norwegian Atlantic Current in the Lofoten basin inferred from hydrological and tracer data (I-129) and its interaction with the Norwegian Coastal Current”. Geophys. Res. Lett. 31, L08302. Gascard, J.C., Raisbeck, G., Sequeira, S., Yiou, F., Mork, K.A. (2004b). The Norwegian Atlantic Current in the Lofoten basin inferred from hydrological and tracer data (I-129) and its interaction with the Norwegian Coastal Current. Geophys. Res. Lett. 31, L01308. Hotchkis, M., Fink, D., Tuniz, C., Vogt, S. (2000a). Accelerator mass spectrometry analyses of environmental radionuclides: sensitivity, precision and standardisation. Appl. Radiat. Isotopes 53 (1–2), 31–37. Hotchkis, M.A.C., Child, D., Fink, D., Jacobsen, G.E., Lee, P.J., Mino, N., Smith, A.M., Tuniz, C. (2000b). Measurement of 236 U in environmental media. Nucl. Instrum. Methods B 172, 659–665. Hrnecek, E., Steier, P., Wallner, A. (2005). Determination of plutonium in environmental samples by AMS and alpha spectrometry. Appl. Radiat. Isotopes 63, 633–638.
292
L.K. Fifield
ICRP (1993). Age-dependent doses to members of the public from intake of radionuclides, Part 2: Ingestion dose coefficients. In: International Commission on Radiological Protection, Publication 67. Ann. ICRP 23 (3/4). Keith-Roach, M.J., Day, J.P., Fifield, L.K., Bryan, N.D., Livens, F.R. (2000). Seasonal variations in interstitial water transuranium concentrations. Environ. Sci. Technol. 34, 4273–4277. Keith-Roach, M.J., Day, J.P., Fifield, L.K., Livens, F.R. (2001). Measurement of 237 Np in environmental water samples by accelerator mass spectrometry. Analyst 126, 58–61. Ketterer, M.E., Hafer, K.M., Link, C.L., Royden, C.S., Hartsock, W.J. (2003). Anthropogenic 236 U at Rocky Flats, Ashtabula River harbor, and Mersey estuary: three case studies by sector inductively coupled plasma mass spectrometry. J. Environ. Radioact. 67, 191–206. Kilius, L.R., Baba, N., Garwan, M.A., Litherland, A.E., Nadeau, M.-J., Rucklidge, J.C., Wilson, G.C., Zhao, X.-L. (1990). AMS of heavy ions with small accelerators. Nucl. Instrum. Methods B 52, 357–365. Kilius, L.R., Rucklidge, J.C., Soto, C. (1994). The dispersal of 129 I from the Columbia River estuary. Nucl. Instrum. Methods B 92, 393–397. Kraft, S., Adrianov, V., Bleile, A., Egelhof, P., Golser, R., Kiseleva, A., Kiselev, O., Kutschera, W., Meier, J.P., Priller, A., Shrivistava, A., Steier, P., Vockenhuber, C. (2004). First application of calorimetric low-temperature detectors in accelerator mass spectrometry. Nucl. Instrum. Methods Phys. Res. A 520, 63–66. Leggett, R.W., Eckerman, K.F., Khokhryakov, V.F., Suslova, K.G., Krahenbuhl, M.P., Millerc, S.C. (2005). Mayak worker study: An improved biokinetic model for reconstructing doses from internally deposited plutonium. Radiat. Res. 164, 111–122. Leonard, K.S., McCubbin, D., McDonald, P., Service, M., Bonfield, R., Conney, S. (2004). Accumulation of technetium-99 in the Irish Sea? Sci. Total Environ. 322, 255–270. Liechtenstein, V.K., Ivkova, T.M., Olshanski, E.D., Golser, R., Kutschera, W., Steier, P., Vockenhuber, C., Repnow, R., von Hahn, R., Friedrich, M., Kreissig, U. (2004). Recent investigations and applications of thin diamond-like carbon (DLC) foils. Nucl. Instrum. Methods A 521, 197–202. Lind, O.C., Oughton, D.H., Salbu, B., Skipperud, L., Sickel, M.A., Brown, E., Fifield, L.K., Tims, S.G. (2006). Transport of low 240 Pu/239 Pu atom ratio plutonium-species in the Ob and Yenisey Rivers to the Kara Sea. Earth Planet. Sci. Lett. 251, 33–43. Litherland, A.E. (1987). Fundamentals of accelerator mass spectrometry. Philos. Trans. R. Soc. London A 323, 5–21. Marsden, O.J., Livens, F.R., Day, J.P., Fifield, L.K., Goodal, P.S. (2001). Determination of U-236 in sediment samples by accelerator mass spectrometry. Analyst 126, 633–636. Marsden, O.J., Abrahamsen, L., Bryan, N.D., Day, J.P., Fifield, L.K., Gent, C., Goodal, P.S., Morris, K., Livens, F.R. (2006). Transport and accumulation of actinide elements in the near-shore environment: field and modelling studies. Sedimentology 53, 237–248. McAninch, J.E., Hamilton, T.F., Brown, T.A., Jokela, T.A., Knezovich, J.P., Ognibene, T.J., Proctor, I.D., Roberts, M.L., Sideras-Haddad, E., Southon, J.R., Vogel, J.S. (2000). Plutonium measurements by accelerator mass spectrometry at LLNL. Nucl. Instrum. Methods B 172, 711–716. McCartney, M., Rajendran, K., Olive, V., Busby, R.G., McDonald, P. (1999). Development of a novel method for the determination of 99 Tc in environmental samples by ICP-MS. J. Anal. At. Spectrom. 14, 1849–1852. Mironov, V., Kudrjashov, V., Yiou, F., Raisbeck, G.M. (2002). Use of 129 I and 137 Cs in soils for the estimation of 131 I deposition in Belarus as a result of the Chernobyl accident. J. Environ. Radioact. 59, 293–307. Moran, J.E., Fehn, U., Hanor, J.S. (1995). Determination of source ages and migration patterns of brines from the U.S. Gulf Coast basin using 129 I. Geochim. Cosmochim. Acta 59, 5055–5069. Newton, D., Talbot, R.J., Pich, G.M., Fifield, L.K., Priest, N.D. (2005). Long-term behavior of injected plutonium in healthy women. In: Oeh, U., Roth, P., Paretzke, H.G. (Eds.), Proceedings of 9th International Conference on Health Effects of Incorporated Radionuclides, Neuherberg, Germany, Nov 29–Dec 1, 2004. Available at http:// www.gsf.de/heir/ProceedingsHEIR2004_ebook.pdf, pp. 311–317. O’Donnell, R.G., Mitchell, P.I., Priest, N.D., Strange, L., Fox, A., Henshaw, D.L., Long, S.C. (1997). Variations in the concentration of plutonium, strontium-90 and total alpha-emitters in human teeth collected within the British Isles. Sci. Total Environ. 201, 235–243. Oughton, D.H., Fifield, L.K., Day, J.P., Cresswell, R.G., Skipperud, L., Di Tada, M.L., Salbu, B., Strand, P., Drozcho, E., Mokrov, Y. (2000). Plutonium from Mayak: Measurement of isotope ratios and activities using accelerator mass spectrometry. Environ. Sci. Technol. 34, 1938–1945. Oughton, D.H., Skipperud, L., Fifield, L.K., Cresswell, R.G., Salbu, B., Day, P. (2004). Accelerator mass spectrometry measurement of 240 Pu/239 Pu isotope ratios in Novaya Zemlya and Kara Sea sediments. Appl. Radiat. Isotopes 61, 249–253.
Accelerator mass spectrometry of long-lived heavy radionuclides
293
Paul, M., Valenta, A., Ahmad, I., Berkovits, D., Bordeanu, C., Ghelberg, S., Hashimoto, Y., Hershkowitz, A., Jiang, S., Nakanishi, T., Sakamoto, K. (2001). Experimental limit to interstellar 244 Pu abundance. Astrophys. J. 558, L133–L135. Paul, M., Valenta, A., Ahmad, I., Berkovits, D., Bordeanu, C., Ghelberg, S., Hershkowitz, Y.H., Jiang, S., Nakanishi, T., Sakamoto, K. (2003). A window on nucleosynthesis through detection of short-lived radionuclides. Nucl. Phys. A 719, C29–C39. Purser, K.H., Kilius, L.R., Litherland, A.E., Zhao, X.-L. (1996). Detection of 236 U: a possible 100-million year neutron flux integrator. Nucl. Instrum. Methods B 113, 445–452. Raisbeck, G.M., Yiou, F. (1999). 129 I in the oceans: origins and applications. Sci. Total Environ. 238, 31–41. Raisbeck, G.M., Yiou, F., Zhou, Z.Q., Kilius, L.R. (1995). 129 I from nuclear fuel reprocessing facilities at Sellafield (U.K.) and La Hague (France); potential as an oceanographic tracer. J. Mar. Syst. 6, 561–570. Richter, S., Alonso, A., Bolle, W.D., Wellum, R., Taylor, P.D.P. (1999). Isotopic “fingerprints” for natural uranium ore samples. Int. J. Mass Spectrom. 193, 9–14. Rucklidge, J., Kilius, L., Fuge, R. (1994). 129 I in moss down-wind from the Sellafield nuclear fuel reprocessing plant. Nucl. Instrum. Methods B 92, 417–420. Skipperud, L., Oughton, D.H., Fifield, L.K., Lind, O.C., Tims, S., Brown, J., Sickel, M. (2004). Plutonium isotope ratios in the Yenisey and Ob estuaries. Appl. Radiat. Isotopes 60, 589–593. Snyder, G.T., Fehn, U. (2002). Origin of iodine in volcanic fluids: 129 I results from the Central American Volcanic Arc. Geochim. Cosmochim. Acta 66, 3827–3838. Snyder, G., Fehn, U. (2004). Global distribution of 129 I in rivers and lakes: implications for iodine cycling in surface reservoirs. Nucl. Instrum. Methods Phys. Res. B 223–224, 579–586. Snyder, G.T., Riese, W.C., Franks, S., Fehn, U., Pelzmann, W.L., Gorody, A.W., Moran, J.E. (2003). Origin and history of waters associated with coalbed methane: 129 I, 36 Cl, and stable isotope results from the Fruitland Formation, CO and NM. Geochim. Cosmochim. Acta 67, 4529–4544. Steier, P., Golser, R., Kutschera, W., Liechtenstien, V., Priller, A., Valenta, A., Vockenhuber, C. (2002). Heavy ion AMS with a “small” accelerator. Nucl. Instrum. Methods B 188, 283–287. Stradling, N., Etherington, G., Hodgson, A., Bailey, M.R., Hodgson, S., Pellow, P., Shutt, A.L., Birchall, A., Rance, E., Newton, D., Fifield, K. (2002). Comparison between biokinetics of inhaled plutonium nitrate and gadolinium oxide in humans and animals. J. Radioanal. Nucl. Chem. 252, 315–325. Suter, M., Döbeli, M., Grajcar, M., Müller, A., Stocker, M., Sun, G., Synal, H.-A., Wacker, L. (2007). Advances in particle identification in AMS at low energies. Nucl. Instrum. Methods B 259, 165–172. Tims, S.G., Hancock, G.J., Wacker, L., Fifield, L.K. (2004). Measurements of Pu and Ra isotopes in soils and sediments by AMS. Nucl. Instrum. Methods B 223–224, 796–801. Vockenhuber, C., Ahmad, I., Golser, R., Kutschera, W., Liechtenstein, V., Priller, A., Steier, P., Winkler, S. (2003). Accelerator mass spectrometry of heavy long-lived radionuclides. Int. J. Mass Spectrom. 223–224, 713–732. Wacker, L., Chamizo, E., Fifield, L.K., Stocker, M., Suter, M., Synal, H.A. (2005). Measurement of actinides on a compact AMS system working at 300 kV. Nucl. Instrum. Methods B 240, 452–457. Wallner, C., Faestermann, T., Gerstmann, U., Hillebrandt, W., Knie, K., Korschinek, G., Lierse, C., Pomar, C., Rugel, G. (2000). Development of a very sensitive AMS method for the detection of supernova-produced long living actinide nuclei in terrestrial archives. Nucl. Instrum. Methods B 172, 333–337. Winkler, S., Ahmad, I., Golser, R., Kutschera, W., Orlandini, K.A., Paul, M., Priller, A., Steier, P., Valenta, A., Vockenhuber, C. (2004). Developing a detection method of environmental 244 Pu. Nucl. Instrum. Methods B 223– 224, 817–822. Wyse, E.J., Lee, S.H., Rosa, J.L., Povinec, P., de Mora, S.J. (2001). ICP-sector field mass spectrometry analysis of plutonium isotopes: recognizing and resolving potential interferences. J. Anal. At. Spectrom. 16, 1107–1111. Yiou, F., Raisbeck, G.M., Zhou, Z.Q., Kilius, L.R. (1994). 129 I from nuclear fuel reprocessing; potential as an oceanographic tracer. Nucl. Instrum. Methods B 92, 436–439. Yiou, F., Raisbeck, G.M., Christensen, G.C., Holm, E. (2002). 129 I/127 I, 129 I/137 Cs and 129 I/99 Tc in the Norwegian coastal current from 1980 to 1998. J. Environ. Radioact. 60, 61–71. Yiou, F., Raisbeck, G., Imbaud, H. (2004). Extraction and AMS measurement of carrier free 129 I/127 I from seawater. Nucl. Instrum. Methods B 223–224, 412–415. Ziegler, J.F., Biersack, J.P. (2003). SRIM—The Stopping and Range of Ions in Matter. http://www.srim.org.
This page intentionally left blank
295
Analysis of radionuclides using ICP-MS Per Roos∗ Risø National Laboratory, DK-4000, Roskilde, Denmark
1. Introduction The inductively coupled plasma (ICP) ion source has successfully been interfaced to most mass analyzers such as the linear quadrupole, time of flight (TOF), Fourier transform ion cyclotron resonance (FTICR) and the electric/magnetic sector. Of these the linear quadrupole is at present the most common choice. Quadrupole mass analyzers are widely used because of their good sensitivity, wide dynamic range, robustness and relatively low cost. The drawback with quadrupole based instruments is that they have a limited mass resolution and a relatively poor spectral abundance sensitivity which together with the numerous isobaric interferences, appearing in connection with the ICP source, creates some problems. Even though elemental ICP-MS is a robust and easy-to-use technique one of the most troublesome factors affecting both accuracy and sensitivity are the ion interferences. The introduction of double focusing sector field (ICP-SFMS) instruments in the late 1980’s made higher mass resolution possible and several spectral interferences could be resolved and corrected for in the analysis. The sector-field instruments also provided lower background and higher ion transmission and consequently lower detection limits. The analysis of radioisotopes using ICP-MS often concerns trace or ultra trace measurements or high precision isotope ratio measurements (e.g., U and Th isotopes). This means that all kind of interferences, which might influence the signal at the masses of interest, is of great concern. A partial review of how to overcome spectral overlaps in elemental ICP-MS was presented by Vanhaecke and Moens (2004). These interferences arise from at least three sources: the plasma support gas alone or in combination with the sample/solvent or the introduction system (e.g., ArH+ , ArC+ , Ar+ 2, etc.), from the sample/solvent (e.g., UH+ ) or from the instrument (e.g., Ni+ ). The different ways in which these interferences have been minimized has been by using high-resolution (magnetic sector) instruments, matrix separation techniques, cold plasma and more recently by collision and/or reaction cells. Below a short description of these techniques is given with a focus of application to radioisotope measurements. ∗ E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11009-3
© 2008 Elsevier B.V. All rights reserved.
296
P. Roos
While increasing the resolution by using a double focusing sector field instrument may help in many cases the maximum resolution achievable is around 10,000 which is not sufficient to resolve a number of interferences (including the well known UH+ from 239 Pu). Furthermore, the transmission of ions decreases with increasing resolution, which for the analysis of trace/ultratrace radioisotopes may be unacceptable. Also the change in peak shape (going from flat-topped to rather triangular) when increasing the resolution means less accuracy in isotope ratio measurements. The use of matrix separation in connection with radiometric measurement techniques requiring alpha- or beta counting is well developed within the radioanalytical sciences. Still difficulties appear due to the very different priorities in element removal between the radiometric and the mass spectrometric methods and for many elements matrix removal alone may not be sufficient to avoid interfering signals at the analyte mass. The use of separation techniques will only be of help in removing interferences originating from the sample atoms. The use of cold plasma (low rf-power, high nebulizer/cool gas flow, increased sampling depth and a shielded torch) conditions in order to minimize the ionization of elements with a high first ionization potential was first reported by Jiang et al. (1988). Using ‘cold plasma’ conditions they managed to suppress argon ionization (ionization potential 15.76 eV) and thereby reduce the Ar+ and ArH+ signals to the degree where it became possible to analyze potassium isotope ratios at ppm concentrations (masses 39 and 41). The need to shield the torch in order to be able to operate in cold plasma mode stems from the fact that the secondary discharge between the plasma and the sampling orifice is stimulated under cold plasma conditions which therefore counteract the effect. There are several ways of suppressing the secondary discharge, shielding the torch by inserting a grounded metal shield between the load coil and the torch (Gray, 1986) being one of them. Sakata and Kawabata (1994) observed that suppressing the secondary discharge was of itself insufficient to attenuate the argide species, operation under cooler plasma conditions was also required. Two sources of argide interferences could therefore be identified: one being the plasma itself and the other the first stage of the vacuum interface owing to the secondary discharge. The use of cold plasma conditions has almost exclusively been used to analyze elements where the Ar-based ions (Ar+ , ArH+ , ArO+ , ArN+ , ArOH+ , Ar+ 2 , etc.) interfere. These are mainly the isotopes of K, Ca, Fe. Also light elements like lithium has benefits from cold plasma conditions due to the reduced mass discrimination. Although the Ar+ 2 dimer interferes with selenium the high ionization potential of selenium (9.75 eV) prohibits efficient analysis of this element during cold plasma conditions. The analysis of the long lived 79 Se (1.1 My half-life), disturbed by tailing from the 40 Ar+ 2 peak, thus has little to gain by using this technique. In general the method is not applicable to elements with a ionization potential above about 7–8 eV due to the sensitivity drop. For elements like selenium the difference in sensitivity between hot and cold plasma conditions is 3–4 orders of magnitude. It should also be emphasized that although the argide interferences are suppressed the significant increase of oxide ions (MO+ ) may limit applications of this technique either because of the formation of new interferences at the mass of interest or because of the analyte forms stable oxygen ions. The method has also been reported to lead to larger variations in mass discrimination (Weyer and Schwieters, 2003). A characterization of the plasma and its analytical possibilities during cold plasma conditions is given in Tanner (1995). The use of cold plasma could theoretically be attractive in the analysis of iodine (129 I) in order to reduce the influence of the 129 Xe+ interferences (xenon ionization
Analysis of radionuclides using ICP-MS
297
potential 12.13 eV) but the relatively high ionization potential of iodine (10.46 eV) would result in a significant transmission drop and the technique has therefore not been applied to this element. The only published application of cold plasma to the analysis of radioisotopes has so far been the analysis of 90 Sr where the 90 Zr isobar was suppressed (Vonderheide et al., 2004; Zoriy et al., 2005). With the introduction of a collision/reaction cell in the ICP-MS instruments a significant analytical improvement was achieved. Applications of gas filled cavities acting as collision cells are well known in organic mass spectrometry in order to study dissociation of organic molecules. Collision cells are also used in AMS for the purpose of destroying polyatomic species and also, more importantly, to strip the analyte ions of electrons to convert the negative ion into a positive one. By placing a multipole cell (acting as a ion guide to focus the ions to the axis of the multipole—not a mass filter, multipoles such as hexapoles or octapoles are more effective ion guides than quadrupoles, particularly at the low mass end) filled with low pressure non-reactive gas (such as helium) in front of the quadrupole mass analyzer a reduction of molecular interferences like ArO+ could be achieved by collision induced dissociation if the ion energy was sufficiently high to break the molecule bond. In order to for the polyatomic ions to have sufficient kinetic energy a pure collision cell is typically operated under a relatively low pressure (less than 1 mTorr). In practice, the efficiency of collision cells has not been too good. Douglas (1989), in an attempt to reduce the intensity of CeO+ , derived from the ICP source, by collisional fragmentation observed that the scattering losses were greater to or equal than the fragmentation yield. Removal of the analyte therefore more or less exceeded that of the interfering ion which therefore prevented the use of the collision cell alone as an effective tool to remove interfering polyatomic ions. Also species like Ar+ would not be removed to any great extent by collisions with a gas like helium. Instead of using the collision gas alone Rowan and Houk (1989) showed that introducing gases such as CH4 in the collision cell could eliminate argide ions more rapidly. Instead of using the gas filled volume only as a collision cell it was now also acting as a chemical reaction cell. In spite of these findings the introduction of collision/reaction cells in commercial instruments was not stimulated, partly perhaps of the almost simultaneous introduction of high resolution ICP-MS sector field instruments which could to some extent solve the problems of interfering ions but also because of lack of knowledge in how to use these reaction gas cells in atomic mass spectrometry. What was needed to convince the user community was a reaction gas with high affinity for the interfering ions (primary argon species) and very low affinity for the analyte ions. The reaction gas also needed to be of low molecular weight in order to minimize scattering losses by the analyte ions. The reactions of Ar+ with H2 had been studied extensively (e.g., Liao et al., 1990) and by applying this knowledge to elemental mass spectrometry Eiden et al. (1996) were the first to show that it was possible to completely neutralize Ar+ and thus remove it from the mass spectrum by introduction of H2 as a reaction gas in the collision cell. It was found that the reaction of H2 with Ar+ was very fast (∼106 times faster than with most other atomic ions thus assuring minimal losses of the analyte ions). Neutralization of Ar+ means that ArN+ , ArO+ , ArC+ , ArCl+ , and Ar+ 2 are removed as well enabling significant improvement of isotopes of Ca, K, Fe, Se.
298
P. Roos
Reactions leading to neutralization of Ar+ are Ar+ + H2 → ArH+ + H (hydrogen atom transfer), (proton transfer), ArH+ + H2 → H+ 3 + Ar + + Ar + H2 → H2 + Ar (charge transfer). The removal of Ar+ (and thus also many argon based polyatomic ions) is not only of importance due to as an isobaric disturbance, the elimination in Ar+ ion flux results in a less space-charge limited effects at the ion optics which favors the lighter elements like Li (mentioned above). The great advantage with H2 is that it shows a minimum of reactivity with other ions and basically only reacts with Ar+ , N+ , O+ , Si+ , Kr+ , and Cl+ . Since the ionmolecule interaction between the interfering ion and the reaction gas (H2 , CH4 , NH3 , etc.) best takes place when the ions are thermalized (allowing more reaction time) the use of the non-reactive collision gas is still important in a collision/reaction cell although the removal of the interfering ions is due to chemical reactions with the reaction gas. Usually helium is used as the collision gas since the fractional kinetic energy loss (or gain) of the plasma ions due to elastic collisions with the gas atoms is smallest when the mass difference between the ions and the collision gas atoms are large. This leads to many small losses of kinetic energy finally resulting in an ion in thermal equilibrium with the surrounding gas atoms. The ions are then said to be ‘thermalized’. A review of the use of reaction- and collision cells for ICP-MS has been given by Tanner et al. (2002). Due to the enormous specificity of the reaction gases with certain ions and the large number of different reaction gases combined with a comprehensive knowledge on these reactions the use of reaction cells to remove most kind of interfering ions seems at a first glimpse very attractive. A problem with collision/reaction cells, as already noted by Rowan and Houk (1989), was the occurrence of new ions produced in the reaction cell. If not taken in to account the removal of the ‘old’ interferences would be at the cost of new ones. The way this has been solved in commercial instruments has either been through kinetic energy discrimination or by mass discrimination. Since the incoming analyte ions will have a kinetic energy somewhat higher than the ions produced in the reaction cell (unless the incoming analyte ions have been completely thermalized) an electric potential could thus allow for some discrimination against the new interfering ions, that is kinetic energy discrimination. The success of this approach depends on the difference in energy distribution between two group of ions. This means that the method is most effective when the pressure is sufficiently low not to thermalize the analyte ions. This however counteracts the reaction rate in the cell (reduction of the ion energy facilitates the use of thermal ion-molecule chemistry). A compromise is thus required between the efficiency of reactive removal of isobaric interferences derived from the plasma and kinetic energy discrimination against new interferences produced in the reaction cell. Obviously, these compromises become more severe the more reactive the reaction gas is. Illustrative examples of how well analyte and cluster ions formed in the collision/reaction cell are removed for various cell bias is shown in Feldmann et al. (1999a). An alternative way of discriminating unwanted products is by mass discrimination. Since hexapoles or higher multipoles are ineffective as mass sorting devices a quadrupole is required. By using this inside the collision/reaction cell only ions with a selected mass are allowed to pass. Efficient kinetic energy discrimination can be used to reject both interfering ions created in the reaction/collision cell (because they will have a lower kinetic energy) as well as poly-
Analysis of radionuclides using ICP-MS
299
atomic entering the gas cell. This is however only possible if the ions (analyte as well as polyatomic interferences) entering the gas cell possess similar energies and, importantly, have a narrow energy distribution. In order to minimize energy spread the use of grounded or shielded plasma is necessary since this stabilizes the plasma and thus the energy transfer to the ions. Since the polyatomic ions posses a greater cross-sectional area than the monoatomic analyte ions they will loose more energy by frequent collisions in the gas cell compared with the analyte ions. By using a potential offset (barrier) at the end of the cell which the ions need to pass in order to reach into the mass to charge sorting quadrupole low energy ions will be rejected. An important aspect of rf-multipole ion guides is the shape and depth of the confining potential well. The quadrupole has the steepest potential gradient from center of the device to the rods while hexa- and octapoles have progressively flatter potential well minima. The higher order of the multipole, the greater acceptance for ion passage, and thus a higher transmittance. The wider potential well in hexa- and octapoles means that the ions occupy a larger volume and thus travels farther in passing through the cell. They therefore experience an increased number of collisions. Since the extent of chemical reaction in the gas cell depends exponentially on the number of collisions the increased path length through the cell has a large effect on the reaction efficiency. The general advantage with higher order multipoles is that they accept a wider range of ion masses simultaneously transmitted and with higher efficiency than quadrupole based gas cells. The advantage with quadrupole based cells is that they can be used in mass selective rejection with a narrow band width. Available commercial instruments can roughly be classified into the basis of the cell design with Perkin-Elmer SCIEX instruments based on quadrupoles while Thermo Instruments and GV-instruments uses hexapoles and Agilent Corporation finally a octapole. Varian has launched an instrument with a unique solution to the reaction cell in that it is situated between the sampler and skimmer cone. In many cases only H2 has been used due to its good ability to remove argon based polyatomic ions while at the same time being almost completely unreactive with ions of most other elements. Unfortunately, one of the elements which shows somewhat increased hydride abundance is uranium as was noted by Feldmann et al. (1999b). The use of collision/reaction cells also has a more or less significant influence on analytical parameters such as ion transmission, mass bias, abundance sensitivity and resolution. The thermalization of plasma ions due to collisions with the gas molecules results not only in lower kinetic energy but also a more narrow energy distribution that becomes easier for the quadrupole to sort by mass to charge (m/z) without loosing ions due to having different kinetic energy. This also greatly improves abundance sensitivity and especially the extreme tailing which normally reaches several masses away from the peak. In normal quadrupole MS the abundance sensitivity is limited by the number of ions with high enough kinetic energy to overcome the filtering rf-field (Blaum et al., 1998). Improvement in the abundance sensitivity can be done by introduction of additional ion-optics (e.g., electrostatic lenses or more or longer quadrupoles). Thermalizing the ions has a twofold influence on abundance sensitivity in that the kinetic ion energy spread is reduced and that the time spent in the mass filtering rffield of the quadrupole is increased. Since both ion transmission and abundance sensitivity are two of the most important parameters in using ICP-MS for the analysis of radioisotopes the use of collision/reaction cells has significantly improved the capabilities of this technique. The need for a good abundance sensitivity occurs both during analysis of low intensity peaks in the
300
P. Roos
presence of extreme high abundance isotopes of the same element (e.g., 129 I in the presence of 127 I or 236 U in the presence of 235&238 U) and due to poor separation of interfering elements in the chemical purification process (e.g., the disturbance of 238 U on 239 Pu). Boulyga and Becker (2002) showed that the peak tail was reduced by up to three orders of magnitude when helium was introduced in the collision cell of a Platform ICP-CC-QMS (GV-instruments, former Micromass). They also observed that increasing the instrument resolution only had a minimum effect on the abundance sensitivity. As was first reported by Douglas and French (1992) for two-dimensional rf-quadrupoles, the collision of heavy analyte ions with the collision cell gas atoms caused, contrary to what was expected, an increase in analyte ion transmission with increasing gas pressure. The signal increase was attributed to the collisional dampening of kinetic energy that confined the ions closer to the axial centerline of the multipole and the process was therefore called collisional focusing. The collisional focusing and the thermalizing of ion energies greatly improve the transmission through the mass sorting quadrupole. Since the thermalizing effect is best for heavy ions the increase in signal intensity is most significant for the heavy elements. For the lighter elements scattering in the gas cell as well as space charge effects due to the high density of positive ions from the plasma, which are effectively focused and moderated in the rf-multipole, causes the signal to be reduced. Boulyga et al. (2001) observed a 5–6-fold increase of the 238 U+ signal when increasing the He-flow from zero to 10 ml min−1 in a Platform (Micromass) hexapole collision cell. For the same conditions 7 Li+ was reduced by the same magnitude. Also the cell bias, used for the kinetic energy discrimination, is of major importance for the transmission of analyte ions as well as for the removal of unwanted ions. Due to the thermalizing effect heavier ions are removed more effectively than lighter ones. Thus, in the study above uranium was removed by more than 70% when applying the 1.5– 2 V hexapole bias necessary to efficiently remove argon based interferences. At the same time 7 Li transmission increased by about 20% due to reduced space charge and scattering. In general the ion transmission in the pressurized collision/reaction cell depends on sample matrix, element mass, mass of collision gas atoms, gas pressure and composition as well as on cell bias. This shows the complexity of the use of collision/reaction cell devices in multielement analysis. The effect on mass discrimination is thus obvious, and tends to increase when the collision/reaction cell is in use relative to when it is not pressurized. Although mass discrimination increases with the use of collision/reaction cells the precision in isotope ratio determination has been reported to improve (Bandura et al., 2000). The reason for this is probably due to the collisional induced broadening of the ion package along the axis of the pressurized collision/reaction cell. This means that high frequency signal noise is averaged out which then causes less time variation in the measured isotope ratios. The fluctuations in signal is normally ascribed to be related to instability in the plasma torch itself, variations in the ion extraction process through the sampling/skimmer cones as well as inhomogeneous spatial distribution of analyte ions along the central channel of the plasma. All these fluctuations occur at various frequencies ranging from below Hz to several kHz. The signal integration times (dwell times) allowed in most commercial ICP-MS may reach down to 0.1 ms so usually the low frequency noise is of limited concern whereas the high frequency noise has to be averaged out in some way. Precise isotope ratio measurements are thus to be performed quickly enough to avoid variations in the low frequency variations but long enough to allow averaging the high frequency noise. Bandura et al. (2000) showed that they
Analysis of radionuclides using ICP-MS
301
could broaden a 0.2 ms 107 Ag+ ion package to last for 2–4 ms (FWHM) and concluded that when pressurizing the collision/reaction cell the high and intermediate frequency ion signal fluctuations were damped so that averaging using long dwell times was no longer necessary. Using Ag and Pb solutions of 40 ppb they obtained isotope ratio RSDs of 0.02–0.03% at a counting statistics error of 0.01–0.02%.
2. Analysis of Pu-isotopes The interest in determining plutonium isotopes in environmental samples may be caused by dosimetric reasons in the case of accidents or releases, by interest in its biogeochemical behavior in the environment or by tracing the source of the plutonium using the isotopic composition as a fingerprint. There are several different techniques available for the determination of trace amounts of plutonium. Of these the most widely used for environmental samples has traditionally been alpha spectrometry even though complementary techniques like thermal ionization mass spectrometry (TIMS) also has been used for high precision isotopic analysis and low-level analysis (Buesseler and Halversen, 1987; Cooper et al., 2000). To a somewhat less extent low-energy characteristic X-ray measurements (Pu LX -rays) and fission track analysis (FTA) (Johansson and Holm, 1996) has been used. Its limited application may be because of lack of sensitivity (Pu LX -ray measurements) as well as its inability to provide isotopic information (FTA only measures the 239 Pu isotope). The use of ICP-MS has during the last decade gained increased interest as an alternative to alpha spectrometry because of the good sensitivity, short analysis time and 239 Pu–240 Pu–242 Pu isotopic information otherwise difficult to obtain through alpha spectrometry. The technique should be seen as a complement rather than a replacement to ordinary alpha spectrometry since the latter technique still is much better suited for the analysis of the 238 Pu/239+240 Pu ratio, which in some cases is more informative than the 240 Pu/239 Pu ratio obtained by mass spectrometry. The 238 Pu isotope is very difficult to analyze in environmental materials by ICP-MS due to the much more abundant 238 U. Mass spectrometric techniques are also usually insufficient to accurately analyze the short-lived 241 Pu (14.4 y) in normal environmental samples unless the 241 Pu/239 Pu ratio is large like in the Chernobyl accident, or the sample size is large enough to contain in the order of 100 mBq or more 239 Pu (which for fallout samples means a 241 Pu content of around 100 fg). The use of the ingrowths method (241 Am) or analysis by low-level liquid scintillation counting is more sensitive with respect to 241 Pu but may necessarily not be more accurate due to problems with cross-calibration of tracers (242 Pu–243 Am), imprecise detector efficiency or unknown blank contribution and calibration problems (LSC). Analyzing Pu-isotopes in the mBq to Bq range (i.e., pg to ng of 239 Pu) usually poses few problems with ICP-MS. Even with relatively poor radiochemistry and an ICP-MS not optimized for the purpose, determination of the Pu-isotopes is fairly straightforward. Due to the high sensitivity of many ICP-MS instruments it is however also possible to analyze Puisotopes at a far lower level. Traditionally, sample collection and preparation based on alpha spectrometry has typically meant hundreds of liter of water and several grams of sediment or soil. Due to the fact that the instrumental sensitivity is good enough for the analysis down to a few femtograms (some µBq) sample size may similarly be reduced to a few liters of seawater
302
P. Roos
or milligrams of sediment. This, theoretically, opens up completely new types of investigations of the biogeochemical behavior of Pu, e.g., interstitial waters in soils and sediments, mineral surfaces or biologically incorporated Pu. However, very few of these kinds of studies have been published. The type of studies were ICP-MS has been used deviates, scientifically, very little compared to traditionally studies were alpha spectrometry has been used. Furthermore, sample sizes, even though they may be reduced, still are of the same magnitude as has been used for alpha spectrometry. The major difference in the use of ICP-MS compared to alpha spectrometry for Pu-analysis has thus not been studies in a lower concentration range but rather to use the ICP-MS to provide 239 Pu/240 Pu isotopic ratios. Even though the reason for this may partly be due to lack of ideas and funding a major obstacle is the ability to reliably measure Pu-isotopes at very low and ultra low levels. This may especially be a problem at laboratories where the instrumentation is shared between many research groups and the possibility to elaborate on the equipment and associated items is very limited. There are a number of problems to address and deal with before being able to trust the weak signals provided by the instrument. Some of the problems when measuring ultra low levels of Pu-isotopes with commercial ICP-MS instruments are the relatively low abundance sensitivity, the risk of interferences from polyatomic species, blanks, background and sensitivity. In order to minimize the problems, both instrumental settings as well as sample conditions may be optimized. Tailing from interfering peaks may be restricted somewhat by attempting to change the peak shape by hardware settings (mainly effective in collision/reaction cell instruments), polyatomic species may be partly removed by desolvating equipment and proper chemical separations or by using reaction gases in collision/reaction cell instruments. Sensitivity may be improved by a suitable sample inlet system. 2.1. Abundance sensitivity and UH+ The relatively poor abundance sensitivity of ICP-MS instruments in general is a problem of particular importance when using ICP-MS for ultra low-level measurements of any element. The origin of this problem is rather complicated but is partly due to the rather large spread of the energy of ions produced in the plasma (10–20 eV). Other important factors contributing to the peak tails are collisions between ions and residual gas molecules in the flight tube (depending on the vacuum and the length of the flight path), scattering of slits and on other material. The energy spread produced by the plasma instability may be reduced by grounding the load coil or by inserting a metal shield (guard electrode) between plasma and load coil to reduce the capacitive coupling between plasma and load coil. The energy spread of the ions may also be reduced by using a collision cell where ions are thermalized or, more commonly, by using an electrostatic analyzer (ESA) which is standard equipment in nearly all sector field ICP-MS instruments. Even so, the abundance sensitivity rarely goes below about 10 ppm in single detector SF-ICP-MS although Ketterer et al. (2004) reported a 5 ppm abundance sensitivity for a VG Axiom instrument at low (R = 410) resolution. With TIMS, this problem is less important due to the much lower energy spread by the thermal-generated ion source (around 0.2 eV). Adding energy filters (ESA, Retarding Potential Quadrupole (RPQ) or Wide Aperture Retarding Potential (WARP; Chen et al., 1992; Cheng et al., 2000; Rubin, 2001) to TIMS instruments may result in abundance sensitivity far below 1 ppm. Such
Analysis of radionuclides using ICP-MS
303
Fig. 1. Peak shape of at mass 238 using a PlasmaTrace 2 (Micromass, UK) sector field ICP-MS in two low resolutions modes. Note that resolution normally is measured at the 10% peak height, in this case at a peak height of 500,000 counts. Spectra acquired using a 5 ppb CRM 112a solution obtained with fully open slits (rounded peak) and with source slit partly closed (flat top peak).
restrictions in peak tailing are necessary when performing high precision 230 Th/232 Th ratio measurements where atom ratios are around 10−5 –10−6 . Some of these filters (e.g., WARP) is designed to eliminate scattered ions which have lower energy, and the filter would therefore only reduce tailing on the low mass side of the peaks. The improvement on the high mass side (e.g., for 239 Pu relative 238 U) would be limited. Even better abundance sensitivities than provided by TIMS are required when measuring isotopes like 36 Cl, 26 Al or 129 I where ratios 36 Cl/35 Cl, 26 Al/27 Al and 129 I/127 I are 10−10 or less which therefore requires accelerator mass spectrometry (AMS). This technique has also been used for the analysis of low levels of Puisotopes (Oughton et al., 2004) with the great advantage in the ultra low abundance sensitivity which means that uranium removal need not to be performed at the same level as for ICP-MS. Apart from the peak tailing, which mainly is of instrumental origin, the UH+ interference at mass 239 also must be considered. The mass resolution needed to resolve this peak from 239 Pu is in the order of 40,000 which therefore not is practically possible. During ultra lowlevel measurements it is also not desirable to increases resolution due to loss in transmission. Since the UH+ interference depends on the amount of hydrogen (water) present in the plasma the magnitude of the peak depends on several factors such as nebulizer flow rate, sample uptake rate and sample introduction methods. Among these the sample introduction method is a very important factor in affecting the hydride generation. The importance of the abundance sensitivity on neighboring masses may be seen in Figure 1.
304
P. Roos
Table 1 Contribution from 238 U to masses 237, 239 and 240 calculated on an integrated peak area basis and on peak top position basis. Measurements were performed at fully open slits and partly closed source slit respectively using a PlasmaTrace 2 HR-ICP-MS 237/238
239/238
240/238
85 ppm 176 ppm
11 ppm 26 ppm
Narrow source slit, R = 440, relative transmission = 47% Abundance sensitivity based on peak top measurements 19 ppm 70 ppm Abundance sensitivity based on peak area measurements 48 ppm 140 ppm
3.5 ppm 14 ppm
Open slits, R = 320, relative transmission = 100% Abundance sensitivity based on peak top measurements 35 ppm Abundance sensitivity based on peak area measurements 86 ppm
The significant tailing from the large 238 U peak stretches out over a large m/z range. It is especially pronounced at the high mass side, which is due to the UH+ (and to a much less de238 U peak is given to mass 237, 239 and 240 gree UH+ 2 ). In Table 1 the contribution from the calculated either on an integrated peak area basis or on peak top position basis where only the top counts has been used. Since no real peak was present at mass 239 the position of the 239 peak was determined from the mass calibration. From the results in the table it is clear that the abundance sensitivity depends both on resolution and on the way it is calculated. Normally abundance sensitivity should consider the mass 237 to 238 signal ratio in order to avoid the influence of UH+ on mass 239 provided the peaks can be assumed to be symmetrical. However, the ways in which values found in the literature are defined are rarely presented and it is obvious that the 239 to 238 signal ratio sometimes is used instead, producing a mixed tail + UH+ contribution. If peak top position values are used rather than integrated peak areas the results are of course generally better. The reason for the higher abundance sensitivity when using the whole peak area is due to a more significant contribution of the 238 U tail being included than when using the more distant peak position. The better abundance sensitivity when increasing the resolution is mostly due to the more restricted path ions have to travel to reach the detector but also the better vacuum may play a minor role. Correction for tailing from adjacent peaks onto a given mass is commonly done either by subtracting values interpolated from signals measured at half-mass positions from the peak to be corrected (Chen et al., 1986) or by applying a predetermined correction value. Disadvantage with the first method is that the peak is subtracted for its own tailing and that the tailing is assumed to be linear which it is not. Also with the second method care needs to be taken as how to apply the obtained value. For instance, if it is a value obtained over the integrated peak it is important to use the same mass interval during the correction as during the separate measurement of the abundance sensitivity. Thus, it is difficult to accurately perform corrections for peak tailing which is necessary at ultra low-level measurements unless uranium concentrations may be held sufficiently low. Once the abundance sensitivity has been determined it is important that the peak tail profile is fairly constant over the narrow mass range of interest (e.g., between mass 238 and 242 for instance) and that it is independent of the ion beam intensity. Deschamps et al. (2003) investigated this and found that it was reasonably true. The problem of poor abundance sensitivity and how to correct for it has been addressed in several papers (e.g., Deschamps et al., 2003; Baglan et al., 2004; Thirlwall, 2001). As seen from Table 1, one possible way to improve the
Analysis of radionuclides using ICP-MS
305
Table 2 Comparing the UH+ /U ratio for two different sample introduction systems. A pure uranium solution (NBL U112a) and an instrument resolution of about 1500 were used in the study Sample introduction system
UH+ /U
Concentric nebulizer + cooled cyclonic spray chamber CETAC USN 5000+
63 ppm 23 ppm
abundance sensitivity in ICP-MS instruments is by increasing the resolution of the system. This however also reduces the sensitivity. Improving resolution by decreasing the source slit width however also improves the peak shape since flat-topped peaks are obtained with the PlasmaTrace 2 HR-ICP-MS instrument used in this case. This is an advantage in the analysis of low concentrations simply because peak hopping means less time spent on low intensity parts of the spectrum and jumping between the flat regions of the peaks is safer than a similar maneuver for rounded peaks (as in quadrupole ICP-MS) if drift in mass calibration occurs with time. A comparison between the amount of UH+ measured using a concentric nebulizer and a CETAC USN 5000+ shows (Table 2) the importance of removing the hydrogen source (in this case water). The UH+ is significantly reduced when the water loading is lowered, as is the case with the USN. Comparing with values in Table 1 the magnitude of interference from peak tailing and UH+ are about of equal size. Previously reported values of the UH+ /U+ ratio has ranged from high 130 ppm (Zoriy et al., 2004a) using DIHEN and 100 ppm (Becker et al., 2004) using a microconcentric nebulizer to 5.5 ppm (Taylor et al., 2001) using a MCN-6000. There is some tendency that UH+ /U+ ratios reported for quadrupole ICP-MS are somewhat lower than for sector field ICP-MS using similar nebulizers (e.g., Becker and Dietze, 1999; Pointurier et al., 2004). Improvement in the UH+ /U+ ratio is most of all achieved by selecting an appropriate nebulizer. Kim et al. (2000), using a PlasmaTrace 2, reported a five-fold reduction in the ratio when changing from an ordinary pneumatic nebulizer to a microconcentric nebulizer with desolvation (MCN-6000). They also used the same type of USN as was included here and reported a UH+ /U+ ratio of 40 ppm. A value identical to what was reported for this instrument earlier, using the same USN as was used now (Sturup et al., 1998). There are also other ways of improving the UH+ /U+ ratio. Boulyga et al. (2002) reported an improvement by a factor of two in the UH+ /U+ ratio by optimizing gas flow and RF-power. An interesting way of reducing the UH+ interference for the mass 239 is to use heavy water, D2 O. By entirely replacing the sample H2 O with D2 O, Vais et al. (2004a, 2004b) succeeded in reducing the UH+ /U from about 13 to 0.3 ppm. This may be an attractive option if only 239 Pu is of interest. The method will of course mean more problems for mass 240 (UD+ ) but this may be of less importance in many studies concerning the environmental behavior of plutonium. 2.2. Polyatomic interferences There are a number of potential interferences apart from the UH+ that may be of importance at trace level determination of plutonium isotopes. These may originate from combinations of
306
P. Roos
Fig. 2. Mass spectrum of the range 239–243 m/z for a 1 ppm Pb + 5% HCl solution obtained in a similar manner as during a standard Pu-analysis procedure.
elements present in the plasma gas, sample and solute atoms. Of major concern are combinations of Pb, Tl, and Hg isotopes with isotopes of Cl and Ar since they all may contribute to the mass range 239–242. Even though these interferences may not be of great importance when analyzing gram amounts of soil and sediments, where total Pu levels are in the pg to ng range, care must be taken either when ultra low levels are analyzed and/or when high precision isotopic ratio information is of importance. Metals such as Pb, Hg and Tl are also frequently found in multielement tuning solutions used for ICP-MS calibrations. Since it is difficult to remove elements originating from the plasma gas (Ar and trace rare gases) focus is therefore mainly directed to remove constituents in the sample and solute. Considering the range of concentrations found in environmental samples and the impact of polyatomic interferences on the minor Pu-isotopes (239 Pu, 240 Pu and 241 Pu) analyzed in ICP-MS, lead would be the most important element to remove. Figure 2 illustrates the background due to polyatomic species likely to be originating from Pb, Cl and Ar isotopes during a typical Pu-analysis sequence with a sample 1 ppm in lead (environmental concentrations are typically in the range of 1–200 ppm for uncontaminated soils or sediments) and 5% HCl. There are also a range of other potential interferences that may form depending on sample type and processing, e.g., rare earth elements (Nygren et al., 2003) or complex organic matter such as urine (Ting et al., 2003) and this necessitates careful analysis of the risks given during any sample and sample processing procedure including blank samples to be analyzed in an identical manner as real samples. Another alternative in identifying the influence of polyatomic interferences in a sample is to use a higher resolution or to mass calibrate very carefully and observe potential peak shifts (Wyse et al., 2001) but this is seldom possible with ultra trace Pu analysis since the peak
Analysis of radionuclides using ICP-MS
307
shape does not permit accurate mass determination. Using higher resolution is usually not a viable alternative due to the loss in transmission. 2.3. Sample introduction system Due to the often very limited possibilities of making changes to the instrument hardware the most common way of improving the instrument sensitivity is by using a suitable sample introduction system. Conventional nebulizer-spray chamber arrangement suffers from high sample consumption (1–2 ml/min) and relatively low analyte transport to the plasma (a few percent). The use of micronebulizers such as the direct injection high efficiency nebulizer (DIHEN) may provide an almost 100% sample transport to the plasma at low flow rates (1– 100 µl/min). Even with such low flow rates the solvent load to the plasma is considerably greater than with conventional nebulizers. This results in increased occurrence of oxides and increased capacitive coupling between load coil and plasma, which in turn increases the secondary discharge strength and therefore broadens the ion kinetic energy distribution impairing the ion transmission through the mass analyzer. The use of a grounded conductive shield between the load coil and the plasma is therefore a need when using the DIHEN. The use of shielding torch electrode or interlaced load coils reducing or eliminating the plasma potential may even during normal conditions using standard nebulizers provide a sensitivity enhancement of around a factor of ten due to the improved ion transmission through the mass analyzer. In modern ICP-MS instruments either solution (shielded torch electrode or interlaced coils) is nowadays standard equipment. It is also a necessary tool when operating the instrument in a ‘cold’ plasma mode. While the DIHEN not necessarily may increase the sensitivity dramatically when defined as count rate per sample concentration (MHz per ppm) the gain in absolute sensitivity (counts per atoms consumed from the sample) is more pronounced. This is an important consideration when the total amount of sample is limited as is usually the case when analyzing environmental ultra low levels of radioisotopes. Since the analysis of very low levels of plutonium often means having limited amount of sample it is important to make the best use of what is available. Routine use of ICP-MS instruments require optimization of torch position, gas flows and some optical settings to gain maximum sensitivity with respect to counts per second per concentration unit in tuning solution, irrespectively of sample consumption rate. Less frequent does the tuning protocol address the sample introduction efficiency, which is more important when a limited amount of sample is available. Normally when liquid samples are used about 90% of the sample volumes are wasted. The concentration of uranium found in the drain solution from a cyclonic spray chamber is approximately the same (±10%) as the feed solution, thus 90% of the sample 238 U atoms will be wasted. Bulk drain from a ultrasonic nebulizer (drain from non-nebulized liquid, condensed solution from heater and cool steps) also contained the same concentration as the feed solution although it may be expected that the heater/cooler drains may differ in concentrations from the feed solution (being lower). The main volume of the drain is however made up by the non-nebulized liquid from the spray chamber. As judged from Figure 3 the optimum sensitivity of a concentric nebulizer is obtained by using a pump flow resulting in an uptake rate of about 0.4–0.5 ml per minute. If also considering the fraction of sample actually reaching the torch (Figure 4) it is clear that the higher the feed rate of the solution the smaller fraction of the sample is actually being used.
308
P. Roos
Fig. 3. Measured 238 U count rate as a function of sample uptake rate for a concentric nebulizer and cooled cyclonic spray chamber.
Fig. 4. Fraction of sample 238 U reaching the torch as a function of sample uptake rate. The fraction reaching the plasma was calculated as net mass of solution to plasma (mass of feed solution minus mass of drain solution) divided by feed solution. Concentration of uranium was found to be about the same in feed and drain solutions.
This of course is important when it comes to optimum use of a limited amount of sample and from Figure 4 it is obvious that the lower the flow rate of sample to the nebulizer the better it is used as long as instrument background is not preventing longer counting times. The better instrument response at lower uptake rates (Figure 5) is probably a combination of better aerosol generation (as seen from Figure 4) and transport through the spray chamber
Analysis of radionuclides using ICP-MS
309
Fig. 5. Instrument response versus analyte (238 U) feed rate reaching the plasma. The analyte feed rate was calculated from feed minus drain multiplied by concentration of 238 U in feed solution.
as well as more efficient ionization of uranium in the plasma (less energy is spent per time unit on the water entering the plasma). Another effect reducing the efficiency at higher feed rates (although of less importance) is the increased rate of oxide formation, which probably is due to lowering of the plasma temperature at increased feed rates. The comparison between the concentric nebulizer and the USN (Table 3) show that the USN produced a signal roughly 15–20 times more intense than the concentric nebulizer for a similar sample feed rate. This is partly due to a better transport of sample to the plasma (50% better) but mainly due to the better use of the injected material in the plasma (about a factor of 10), which is due to less plasma energy being wasted on evaporating water when comparatively more dry aerosols are injected. If the concentric nebulizer had been used with a sample uptake rate based on its maximum sensitivity (cps ppb−1 ) of around 0.5 g min−1 , rather than based on maximum use of fraction of sample reaching the torch, the difference between the two nebulizers would have been even greater. The count rate per unit mass of uranium reaching the plasma at 0.5 g min−1 was 1.5 × 106 cps per pg U instead of 2.1 × 106 at an uptake rate of 0.14 g min−1 used in the comparison example. Although not investigated here, the analyte transport efficiency of the two nebulizers is also affected by parameters such as nitric acid concentrations (e.g., Stewart and Olesik, 1998) and salt load but the transport efficiency may not necessarily change in a similar manner with differences in concentration for the two nebulizers due to different physical aerosol generation processes. We may finally conclude that very little is gained in
310
P. Roos
Table 3 Sample introduction characteristics for two different sample introduction systems using a 1 ppb (0.2 ppb for the USN) 238 U solution
Uptake rate (g min−1 ) Drain rate (g min−1 ) Used liquid (g min−1 ) Fraction of feed used Feed rate of U to plasma (pg s−1 ) Cps per ppb consumed of sample Cps per pg U to plasma UO+ /U
Concentric nebulizer cyclonic spray chamber
CETAC U-5000AT+
0.14 0.126 0.0133 9.5% 0.19 380,000 2.1 × 106 0.0038
0.14 0.119 0.0203 14.5% 0.07 7,500,000 22.2 × 106 0.00085
trying to reduce the UO+ formation for the concentric nebulizer since this constitute a very small fraction for both sample introduction systems used. Recently the use of gas-phase chemistry in reaction cells has been explored as a mean to separate several elements from each other. A comprehensive literature database on cation reactions in gas phase with different molecules (Ancich, 2003) may provide suggestions to which reaction gases that could be of use when aiming of removing unwanted interferences in ICP-MS. Based on known reactions of actinides with gases like ammonia, carbon dioxide and ethylene Vais et al. (2004a, 2004b) and Tanner et al. (2004) showed that it was possible to almost completely remove uranium from plutonium in an Elan DRC II ICP-MS using either of these gases. While plutonium was essentially unreactive uranium reacted fast with these gases and the ion intensity of the 238 U+ was reduced by 104 –105 . Although suggested to use as a substitute to ordinary analytical separation chemistry the application to low level measurements of plutonium is less reliable due to increased risk of new interferences, as + pointed out in both reports, such as PbO+ 2 and PbN2 H4 . Also the risk of increased formation 238 + of UH may limit the use of reaction gases in trace Pu analysis without any prior separation chemistry. Even if the relative UH+ abundance during use of the gases tested above did not show any increase relative to standard conditions (UH+ /U about 10−5 –10−6 ) the much larger uranium concentrations present in non-separated samples would set a limit. 3. Analysis of 237 Np Although neptunium (as 239 Np) was the first transuranic element discovered (McMillan and Abelson, 1940) the number of studies conducted on occurrence and behavior of neptunium in the environment has been scarce. Most certainly this is due to the difficulties in analyzing this element using radiometric methods. Being an alpha emitter with a long half-life (2.14 million years) and with a relatively high mobility in the environment has caused it to be one of the most hazardous radioisotopes in spent nuclear fuel after some ten to hundred thousand years of storage. 237 Np present in the environment has been produced both by the 238 U(n, 2n)237 U → 237 Np reaction through fast neutrons in nuclear bomb testing and reactors and due to the
Analysis of radionuclides using ICP-MS
311
→ 237 Np reaction in nuclear reactors followed by releases from spent fuel reprocessing. It is also produced as a consequence of the decay of 241 Am. As with most other long-lived artificial radioisotopes in the environment the concentrations of 237 Np are so low that direct determination by ICP-MS is practically impossible which therefore has necessitated a number of proposed separation procedures to purify neptunium from nearly any type of matrix. Examples of such developed procedures are Yamamoto et al. (1994), Baglan et al. (2002), Ayranov et al. (2005), Egorov et al. (2001), Ji et al. (2001) and Vance et al. (1998). In general the separation chemistry used to analyze neptunium by solid state alpha spectrometry and INAA can also be applied to samples for ICP-MS since in all cases the major problem is to completely remove the sample matrix and to minimize the remaining uranium in the sample before measurement. In solid state alpha spectrometry the electrodeposited source must be almost devoid of uranium due to the same alpha energy of 234 U and 237 Np of 4.78 MeV. In INAA the uranium should be minimized especially in cases where 239 Np is used as tracer due to the production of 239 U → 239 Np which otherwise leads to an overestimate of the chemical separation yield after irradiation. The difference in using ICP-MS is that detection limits for 237 Np are lower than for solid state alpha spectrometry and INAA which therefore means more thorough removal of uranium. Use of other mass spectrometric techniques such as glow-discharge MS, AMS and TIMS may in some respect require none (glow-discharge MS) or less demands on the separation from uranium due excellent abundance sensitivity (AMS and TIMS). Alternative instrumental ways of improve the separation of neptunium from uranium include capillary electrophoresis-ICP-MS (Kuczewski et al., 2003) and flow-injection technique based chemical separations (e.g., Egorov et al., 2001). One of the main problems in analyzing 237 Np, either by radiometric or mass spectrometric methods, is the lack of a suitable tracer. Both the gamma emitting short-lived 239 Np (2.35 days half-life, obtained from 243 Am decay) and 235 Np (half-life of 396 days) may be used to trace the separation chemistry (e.g., Holm et al., 1987). If alpha spectrometry is to be used for the analysis of the purified 237 Np a well-defined geometry factor (counts per decay) may be used. For ICP-MS an internal standard is required which should be close in mass and ionization potential to the analyte. La Rosa et al. (2005) used very pure 236 U (99.97 atom%) as internal standard. Advantages with this choice is that the ionization potential of uranium is closer to neptunium than any other element and that the wash-out of uranium using weak nitric acid usually is much more effective than for instance for plutonium. The 236 U, however, has to be added with caution in order to match the 237 Np concentrations in order to avoid disturbance caused by UH+ and poor abundance sensitivity. For most laboratories 242 Pu is a common choice, perhaps because it is a common isotope in those laboratories already performing actinide analysis. The 242 Pu isotope is relatively close to 237 Np with respect to mass and ionization potential, can be obtained very pure (essentially free from 237 Np) and may be added to the sample without risk of disturbing the analysis of 237 Np either through peak tailing or the creation of some polyatomic species. Also, plutonium and neptunium chemistry is similar. Although 236 Np (half-life of 154,000 y and produced by deuterium bombardment of pure 235 U) would be a suitable tracer for the ICP-MS analysis of 237 Np it is not commercially available. It has been used in a few studies, e.g., Kenna (2002) using multicollector sector field ICP-MS and in work by Beasley et al. (1998), Kelley et al. (1999) and Cooper et al. (2000) using TIMS. 235 U(n, γ )236 U(n, γ )237 U
312
P. Roos
Since the chemical behavior of neptunium and plutonium is very similar when both exist in the IV valence state, attempts has been made to use 242 Pu as a tracer for neptunium in the complete analysis step (separation chemistry + as internal standard in the ICP-MS measurement). Chen et al. (2002) developed a robust method based on ordinary anion exchange and TIOA solvent extraction in which Pu acts as a tracer for neptunium in water as well as sediment and soil samples up to about 10 g. Similar work was done by Kim et al. (2004) using Eichrom TEVA extraction chromatography resins and valid for sediment/soil samples up to about 4 g. Although several potential traps exist in both methods (e.g., ensuring that neptunium is actually in the IV valence state) careful work by a trained chemist will produce reliable results. The possibility of performing simultaneous analysis of both elements should not be underestimated when working with samples where the risk of having inhomogeneous composition is large, e.g., in Irish Sea sediments where activity partly is associated to hot particles and the Np/Pu ratio is known to have varied over the years (Assinder, 1999). The use of 242 Pu as a tracer for 237 Np requires that the relative response of the elements in the ICP-MS is known. The use of standard solutions with known Np/Pu ratios should therefore be inserted between samples when running the analysis. It has been observed in the author’s laboratory that the ratio not necessarily remains constant over a day when analyzing for Np and most certainly may change over several days by up to 10% depending on tuning conditions. Moreover, the memory effect of Pu and Np is somewhat different in typical wash solutions (2–5% HNO3 ). The main interference of 237 Np is the tailing from the 238 U. The separation chemistry should therefore aim at trying to reduce the U concentrations to less than a factor 105 –106 of the 237 Np concentration given an abundance sensitivity of around 10−6 for quadrupoles (and somewhat worse for sector field instrument). For sector field instruments the possibility of increasing the resolution may be a potential solution when analyzing samples with sufficiently high 237 Np concentrations but the abundance sensitivity does not improve at the same rate as the resolution which limits this alternative. When performing ultra trace measurement of 237 Np the situation becomes, as for most elements, more complicated. Several polyatomic interferences are possible and it will be necessary to remove a number of stable elements so that they are present preferably in sub ppm concentrations or less. Examples of polyatomic interferences are 197 Ag40 Ar, 181 Ta40 Ar16 O, 183 W40 Ar14 N and combinations of thallium and mercury isotopes with sulfur and chlorine isotopes as well as palladium isotopes with xenon and 153 Eu84 Kr. Some of these polyatomic interferences can be resolved from 237 Np using HRICP-MS instruments at higher resolution (e.g., 181 Ta40 Ar16 O at a resolution of about 2000) but with trace quantities of 237 Np first priority is to obtain maximum transmission instead. To date no attempts to remove the polyatomic interferences using collision cell technology has been reported but certainly the argide based interferences have the potential to be reduced using this technique. When using 242 Pu as a tracer and/or internal standard lead, mercury and thallium concentrations in the final sample should also be considered due to the risk of forming chloride and argide interferences at mass 242. The internal standard is however usually added in such amounts that interferences are a negligible problem. Detection limits for 237 Np in current ICP-MS system designed for ultra trace detection are, like for 239&240 Pu, more dependent on blank levels (uranium and polyatomic interferences) than on instrument sensitivity. Sector field instruments as well as some recent quadrupole instruments equipped with guard electrode (or other techniques eliminating the plasma
Analysis of radionuclides using ICP-MS
313
potential) using high-efficiency nebulizers may be tuned to have signal intensities of 10– 20 MHz ppb−1 while still maintaining instrument background count rates at mass 237 below 1 cps. Detection limits under such conditions may be taken to the sub fg per sample provided the blank can be kept low enough. The development of better systems is going fast and it is interesting to take a look at the ‘early days’ of using the ICP-MS in analyzing 237 Np. In an intercomparison study of 237 Np in artificial urine conducted in early 1990 (Lee et al., 1995) it was concluded that ICP-MS at the present time was not yet as well established for the measurement of 237 Np as alpha spectrometry and that at the levels analyzed using alpha spectrometry had a better accuracy than ICP-MS. The concentrations used in the study were 3 and 50 mBq L−1 (0.1 and 1.7 ng L−1 ) of 237 Np and up to 160 mBq L−1 of 238 U. Problems in the analysis of 237 Np were lack of Np-tracer, low sensitivity and high blanks. Reported detection limits for ICP-MS (around 0.1 ng L−1 ) were not better than detection limits for alpha spectrometry. Today, where detection limits of 237 Np fairly easily may reach sub ppt levels on nearly all commercial ICP-MS systems and the separation chemistry has been developed to remove uranium, corresponding activity levels measurable using ICP-MS is in the range 0.01–1 µBq, levels that is impossible to detect even with the best radiometric systems available. 4. Analysis of 99 Tc Of the technetium isotopes present in the environment only 99 Tc is relevant for studies using mass spectrometric techniques (the occurrence of 97,98 Tc is too low). Although naturally present in very small amounts due to spontaneous fission of 238 U and slow neutron induced fission of 235 U in uranium ores, its occurrence in nature is almost exclusively from the nuclear fuel cycle and nuclear bomb tests. The total amount introduced into the environment is around 1.6 PBq or 2500 kg of which the major part (around 90%) has been to the marine environment due to releases from reprocessing activities. 99 Tc is a pure beta emitter with maximum beta particle energy of 294 keV. The traditional ways of analyzing this isotope has been through beta particle counting using gas flow GM counters or by liquid scintillation. Neither of these methods is ensuring that the signal recorded is from 99 Tc which therefore requires thorough radiochemical separation techniques to be used in order to remove all other possible contributors to the beta activity counted. The determination limit when using radiometric methods is in the order of 1–10 mBq. With a half-life of 211,000 years 99 Tc has a relatively low specific activity (1 mBq = 1.6 pg) and could therefore be more suited for analysis by mass spectrometric methods. The first reported trace-level analysis of 99 Tc using mass spectrometry was by Anderson and Walker (1980) while the first paper describing 99 Tc analysis by ICP-MS in environmental samples was presented by Kim et al. (1989). Other mass-spectrometric techniques used for analyzing 99 Tc include RIMS (e.g., Downey et al., 1984) and AMS (e.g., Wacker et al., 2004). Basically there are two problems to deal with when 99 Tc is to be analyzed at environmental levels requiring chemical separation techniques. The first is a suitable tracer when performing isotope dilution MS, the second are ways of dealing with interferences from Mo and Ru. Traditionally 99m Tc, obtained from a 99 Mo–99m Tc generator, has been widely employed as a yield determinant when performing radiochemical measurements. For mass spectrometric
314
P. Roos
measurements at mass 99 the 6 h half-life 99m Tc can obviously not be used. In principle three different methods of handling the tracer problems can be found in the literature. The first (e.g., Mas et al., 2004a or Uschida et al., 2000) is by using a short-lived Tc-isotope 99m Tc or 95m Tc as a monitor for the chemical isolation of Tc and then using a stable isotope, e.g., 103 Rh or 209 Bi as an internal standard. In the second method the chemical analogue rhenium is used (e.g., Mas et al., 2004b) throughout the whole analysis (separation chemistry + ICP-run) and finally a method using 97 Tc as a tracer was reported by Beals (1995). As internal standard an element not disturbing the Tc-signal or itself being disturbed by the sample matrix should be used. It should have a mass in the same region as the analyte (in order to minimize mass bias) and should have a similar ionization potential. It should have a minimal memory effect during analysis and be able to be removed by the selected separation chemistry if present in the sample. Rhodium (monoisotopic, 103 Rh) fulfills most of these demands but may be difficult to keep in weak (1–5%) HNO3 solutions for prolonged periods. Indium is a generally popular internal standard and was also suggested to be useful for Tc by MacCartney et al. (1999). They pointed out the potential risk of disturbing the mass 115 by 99 Tc16 O but this risk is small. Perhaps the larger risk with using Indium as an internal standard for Tc is the relatively large difference in ionization potential (7.28 eV for Tc and 5.79 eV for In). At first glance the choice of 97 Tc as a tracer would be an excellent alternative. The tracer is however very rare and whenever encountered the 99 Tc/97 Tc ratio should be investigated since it depends on the way of production of the 97 Tc, that is the purity of the target used. Beals (1995) used a 97 Tc tracer produced by neutron activation of enriched 96 Ru. The 99 Tc/97 Tc ratio in a blank sample was after correction from interferences calculated to be 0.0025. The m/z 97 is contrary to m/z 99 not free from isobars but have contribution from 97 Mo (natural abundance 9.55%) and even if Mo is sufficiently reduced in the chemical isolation of Tc the instrument background is, from the authors experience, considerably higher than that of m/z 99. The origin of this background is unclear but several candidates present in air and/or plasma gas are possible such as 81 Br16 O, 83 Kr14 N, 80 Kr17 OH, Ar2 16 OH, 84 Kr13 C and/or 79 Br18 O but may also have contribution from species like 40 Ar57 Fe. To overcome the relative high background more of the 97 Tc has to be added to the sample, an amount that has to be compromised with the amount of 99 Tc present in the tracer. The use of the short-lived 99m Tc obtained from a 99 Mo generator is widely used due to the relatively low cost, acceptable contribution of 99 Tc and ease of measurement of the 140 keV gamma-rays emitted. The source of the 99 Mo is either from 235 U fission or from neutron activation of 98 Mo. The former method is required when manufacturing the high activity generators needed in clinical use and is also the most common type of generator used to obtain 99m Tc for tracing the separation chemistry of 99 Tc. The drawback of these generators in radioanalytical work is the risk of radioactive contaminants not removed in the separation of 99 Mo from 235 U fission and activation products. This is a well-recognized problem, which has been investigated by several authors (e.g., Wasserman and Fourie, 1981). The problem with these contaminants is mainly when performing radiometric measurements and has little influence on mass-spectrometric measurements. With the introduction of ICP-MS one of the goals has been to achieve lower detection limits than what has been possible using radiometric methods. The risk of 99 Tc in the 99m Tc eluate from the 99 Mo–99m Tc generator has therefore again been highlightened. The problem is not new and has been well documented (e.g., Mattsson, 1978) but using ICP-MS lower levels of Tc has been possible to determine. Provided the gen-
Analysis of radionuclides using ICP-MS
315
erator in use is extensively eluted before taking into use (to avoid excess build-up of 99 Tc from decaying 99m Tc) and using a short ingrowths period (e.g., 2 h instead of 24 h used when maximum 99m Tc activity is obtained) and using moderate amounts of 99m Tc the risk of contamination with 99 Tc can be kept to insignificant levels. Hill et al. (2000) studied the presence of 99 Tc in 99 Mo–99m Tc generators and concluded that the amount of 99 Tc found in the eluted 99m Tc well matched the theoretical amount, that is the number of 99 Tc atoms produced both from the eluted 99m Tc and from the decaying 99m Tc on the generator aluminum oxide column during the ingrowths period. Mas et al. (2000) concluded that the detection limit increased marginally, from 0.4 to 0.5 ppt, when using 99m Tc up to 2 MBq from a clinical 99 Mo–99m Tc generator. The problem of 99 Tc contamination may be more severe when using 95m Tc as tracer. Tagami and Uchida (1996) reported on considerable amounts of 99 Tc in their 95m Tc tracer produced through a 96 Mo(p, 2n)95m Tc reaction. Clearly the presence of other Mo-isotopes would mean a risk of producing other isotopes, including 99 Tc. The same authors (Sekine et al., 1999) reported on the production of 95m Tc through the reaction 93 Nb(α, 2n)95m Tc which provided a tracer almost free from 99 Tc. The use of rhenium as a chemical tracer for 99 Tc analysis by radiometric methods has been used for several years in various laboratories. A detailed description of the method used at the fisheries laboratories (CEFAS) in UK was reported by Harvey et al. (1992). Although rhenium and technetium have similar chemical behavior their redox potentials differ and the ion-exchange behavior on anion resins is not exactly identical. The risk of over- or underestimating the Tc yield is therefore not negligible. Mas et al. (2004b) reported on their results of using rhenium (as 97% enriched 185 Re) and 95m Tc + indium (as an internal standard) as yield monitors in the same sample in order to compare the two tracers. They could conclude that using the separation method described by them (Eichrom, TEVA column) the agreement between 99 Tc concentrations determined by 95m Tc or 185 Re was almost identical. Kim et al. (2002), using rhenium as tracer and TEVA columns in an on-line flow injection setup, also concluded that the 99 Tc/Re ratio remained constant during the separation. By choice of impropriate acid concentrations the Re–Tc behavior may however differ considerably on the TEVA column. In 2 M HNO3 (used to wash the column in Tc-analysis) Tc is strongly retained but Re is gradually eluted (Tagami and Uchida, 2000). The use of reference material (either commercial ones or produced in the laboratory) cannot be underestimated when new methods are designed and especially when non-isotopic tracers like rhenium for Tc are used. The use of rhenium as a tracer for Tc has the advantage that it may be added to the sample in sufficient amount to dominate over any kind of interference and the risk of Tc contamination in the tracer is negligible. If, for some reason, the rhenium tracer is added in low amounts the chemical separation procedure should be designed in such a way that potential interfering elements is removed as far as possible in the process. In the case of rhenium this mainly deals with tungsten due to the hydrides 184 WH and 186 WH and to a lesser extent the rare earth elements which may form oxides, carbides, hydroxides and argides in this mass range. The drawbacks of using rhenium as tracer are the risk of incomplete tracing of Tc and to a less extent a need to perform a mass bias correction due to the difference in masses between m/z 99 and 185/187. Also the risk of different memory effects and differences in ionization potential, 7.28 for Tc versus 7.88 for Re, should also be taken into account when using Re as tracer.
316
P. Roos
Table 4 Examples of interfering species on mass 99 Isobar or concomitant
Oxide
Hydride
Argide
Chloride
Doubly charged ion
98 Mo (24.1%)
83 Kr16 O (11.5%)
98 MoH (23.8%)
59 Co40 Ar (100%)
62 Ni37 Cl (3.6%)
198 Hg (10%)
98 Ru (1.1%)
81 Br18 O (49.5%)
98 RuH (1.9%)
63 Cu36 Ar
64 Zn35 Cl (48.6%)
199 Hg (16.8%)
99 Ru (12.7%)
67 Zn16 O (4.1%) 2 51 V16 O (99.8%) 3
100 Mo (9.6%) 100 Ru (12.6%)
43 Ca16 O40 Ar
198 Pt (7.2%)
40 Ca18 OH40 Ar
4.1. Interferences The general background situation at mid masses is normally much higher than at the high mass end due to various combinations of the high abundance lighter elements. Interferences can appear from isobaric overlapping, polyatomic ion formation, doubly charged ions and refractory oxides. The peak background at mass 99 may thus be expected to be considerably higher than for example plutonium at mass 239. Some examples of possible contributions to the 99 peak background is given in Table 4. The abundance of the various interferences depends on a number of parameters where the sample matrix, RF power, sample introduction methodology, plasma gas flows, cone conditions and sampling depth in the plasma is of major concern. For instance, the use of electrothermal vaporization (ETV) as sample introduction may cause a significantly increase in carbide species such as 87 Rb12 C and 87 Sr12 C. On the other hand, using ETV means that the water vapor transported to the plasma is much less compared to when the sample is in liquid state and transported to the plasma through (e.g., a Meinhardt type of nebulizer). The ETV thus reduces the formation of oxides and hydrides. The ETV may also actively be used to remove the interfering ruthenium due to its lower boiling point (e.g., Song and Probst, 2000). Examples of potential interferences originating from trace impurities in the plasma argon gas could be 84 Kr15 N, 86 Kr13 N, 82 Kr17 OH and 40 Ar2 18 OH. Combinations with carbon and nitrogen continuously present in the air or the sample may create species like 85 Rb14 N, 87 Rb12 C, 87 Sr12 C or hydroxides like 82 Se17 OH. In practice, however, the instrument background at mass 99 is relatively low in most cases, indicating that species formed exclusively due to the plasma gas and air molecules is of little concern. Mas et al. (2002) investigated the importance of a number of potential interferences on mass 99 and concluded that the only major concern when analyzing environmental samples was that of 99 Ru which exist in nature at an abundance of 12.7% and which have some chemical similarities with technetium. In soils and seawater ruthenium concentrations are in the range of 1 ppb and 1 ppt, respectively, which means that the decontamination factor for ruthenium in the chemical isolation process needs to be better than 106 –107 when Tc concentrations in a soil sample only contaminated by atmospheric fallout is around 1–10 mBq kg−1 . The correction from remaining ruthenium in the sample may be done by observing intensities of the other Ru isotopes, preferably at mass 101, which is free from other isobars, but may not necessarily be free from other interfering species. It should be emphasized that it may not always be the sample itself, which is the main source of ruthenium (or any other stable contaminant). The laboratory chemicals and glassware/Teflon
Analysis of radionuclides using ICP-MS
317
used should be controlled and thereafter set aside to be used for the particular analysis. KeithRoach et al. (2002) found high levels of ruthenium in the BioRad ion exchange resin used for the initial Tc separation and thus included a separate rinsing step to remove the Ru from the resin before using it. Even though molybdenum has no stable isotope at mass 99 its great abundance in nature (1 ppm in soils and 10 ppb in seawater) may cause problems due to the formation of 98 Mo1 H and from poor abundance sensitivity which causes peak tailing from the 98 Mo and 100 Mo peaks. Again, corrections from remaining Mo in the sample may be done by determining the MoH/Mo ratio (usually in the order of 10−5 –10−6 ) for the specific setup and sample matrix and by observing mass 95 and/or 97 which is free from isobaric interferences. The background at these masses due to other interfering species may, however, be significant and thus the use of these masses may be of limited use. Normally the decontamination of molybdenum present in the sample itself is sufficient for the analysis of 99 Tc and rather the diffuse Mo source in glassware, chemicals, etc. is the major problem. This means avoiding glassware (with high Mo concentrations) and minimizing all chemical reagents. The source of Mo in some laboratories, notably those where radiochemical separation methods are taking place, may also be from the long-time use of the inorganic ion exchanger ammonium molybdophosphate (AMP), used as a powder and therefore airborne, to concentrate radiocesium from water samples. Mas et al. (2002) observed that the contribution from 98 Mo to mass 99 was mainly due to abundance sensitivity and to a less extent due to formation of the MoH. The relative contributions from MoH and abundance sensitivity is however dependent on sample introduction methodology and instrument operation conditions. From their work it could be concluded that as long as the Mo concentration in the purified sample was below around 10 ppb the contribution to the m/z 99 was small in comparison to the instrument background. With the introduction of ICP-MS many hoped that the often complicated radiochemical purification techniques needed for radiometric measurements in alpha and beta counting would be only but a memory and that measurements performed with ICP-MS could be done with much less demand on separation chemistry. It has however become evident that it frequently is the other way around and that purification techniques used for mass spectrometric measurements needs to be even more stringent than similar techniques for radiometric measurements. This is perhaps easy to understand when realizing that the abundance of the stable elements relative to the radioactive analyte often is far much greater than the relative abundance of the different radioisotopes involved in radiochemical separation methods. The separation chemistry in ICP-MS is necessary in order to both remove interfering elements and to minimize the matrix induced signal suppression. Normally, both radiometric and mass spectrometric methods have difficulties in performing accurate measurements of 99 Tc close to the detection limits but the difference between the methods are that for radiometric measurements the procedure blank can fairly easy be kept the same as the instrument blank and thus the detection limit is governed by the instrumentation background. For ICP-MS the limit is more often set by the blank contribution from the chemical processing rather than the instrument blank alone. In this respect radiometric techniques have a better defined detection limit than the ICP-MS technique which are more prone to vary from batch to batch of samples processed. This requires that the ambition of ‘above blank’ signal should be kept higher in ICP-MS than in for instance gas flow GM counting even though the latter only provides a non-spectrometric signal. Without considering the laboratory
318
P. Roos
blank detection limits for 99 Tc by ICP-MS fairly easily reach down to sub ppt levels (1 ppt 99 Tc equals 0.6 mBq/ml) and sector field instruments equipped with ultrasonic nebulizers may reach down to ppq levels (µBq/ml) as reported by Takaku (2005). 5. Analysis of 129 I Iodine-129 is a long-lived radioisotope with a half-life of 16 My. It is naturally produced in the atmosphere due to spallation reactions with Xe atoms and in the Earth crust by fission of 238 U and 235 U as well as neutron capture reactions of 128 Te and 130 Te. The inventory of naturally produced 129 I is about 260 kg but due to releases from nuclear activities (mainly reprocessing of spent nuclear fuel) the current inventory amounts to more than 4000 kg (26 PBq) of which the La Hague reprocessing plant accounts for the dominating part, around 75%. Iodine plays an important part in the metabolism of mammals and is therefore of great interest in health science. The high volatility of iodine and the large amount of 131 I (8 days half-life) produced in nuclear fission reactors have also created an interest in the field of radioecology and emergency preparedness. Although the 131 I isotope is of major concern only in the early phase of a nuclear incident the 129 I isotope is considerably more interesting in the long perspective. The releases of gaseous iodine or attached to aerosols from the La Hague and Sellafield reprocessing plants has enabled its use as a tracer for atmospheric transport (e.g., Santos et al., 2005; Kieser et al., 2005). The conservative behavior in marine oxic waters combined with the well-known release pattern from the major sources, the nuclear fuel reprocessing plants at La Hague and Sellafield, has warranted its use as a tracer for water currents. An overview of its role as an environmental tracer is given by Hou (2004). Studies of 129 I may also aid in understanding of the behavior of stable iodine which is monoisotopic (127 I). Stable iodine are commonly measured by ion chromatography (IC) which also is particularly useful in separating the different species iodide and iodine in the solution. The detection limit using IC is typically in the order of some ppb which is sufficient for stable iodine but usually not for 129 I. The low specific radioactivity (6.5 MBq g−1 ) of this isotope has made direct radiometric measurements difficult. The most used method has been INAA (e.g., Hou et al., 1999; Hou, 2004; Muramatsu and Yoshida, 1995) in which the 130 I (12.4 h) isotope is produced which is easily measured by gamma spectrometry. By far the most useful method for the analysis of 129 I is by accelerator mass spectrometry, AMS (e.g., Kieser et al., 2005). Depending on the degree of anthropogenic influence of 129 I on environmental material the 129 I/127 I atom ratio varies from about 10−14 to 10−6 and only by using AMS can such ratios be measured, especially in seawater (Povinec et al., 2000). A review of the analysis of 129 I has been presented by Schmidt et al. (1998). Problems associated with analyzing 129 I with ICP-MS is low sensitivity (high ionization potential, 10.45 V, and therefore low ionization in the plasma), elemental and polyatomic interferences from 129 Xe, 89 Y40 Ar, 115 In14 N, 113 Cd16 O, etc., memory effects due to the adsorption of iodine on tubing, etc., problems with low abundance sensitivity in ICP-MS creating tailing from the 127 I peak into the mass 129 window as well as formation of 127 I1 H2 . Interferences due to the metal oxides can often be disregarded. Haldiman et al. (2000) measured the contribution at masses 127 and 129 at 100 ppm Cd concentrations using a Cinnabar
Analysis of radionuclides using ICP-MS
319
mini-cyclonic spray chamber and a Micromist nebulizer which corresponded to equivalent contribution at mass 127 and 129 of about 0.4 ppb. The problems associated with the tailing and hydrides of 127 I sets a lower limit of the 129 I/127 I ratio which can be analyzed by ICP-MS. The much better abundance sensitivity, the break-up of polyatomic ions in the stripper of AMS instruments and the negligible negative ion formation of xenon (129 Xe) is the reason why only these may be used for low 129 I/127 I ratios. The loss of iodine in the sample introduction system is a well recognized fact when analyzing aqueous samples. The wash-out is normally considerably longer when elemental iodine and iodide is present. With iodate (IO− 3 ) the memory effect is usually less. Poor stability and memory effects may occur at low pH in the presence of iodide and is probably attributed to the formation of volatile species such as HI and I2 , therefore alternative ways of introducing the iodine to the plasma is preferred. Kerls et al. (1996) experimented on volatilizing the iodine from the water phase by first oxidize iodide to iodine and by producing small droplets in a Meinhard nebulizer. Iodine was transferred from the aqueous to the gas phase where it was further carried to the plasma by an additional argon gas flow. By using this sample introduction system they managed to increase the detection power by a factor of 50 thereby reaching a detection limit of 127 I of 0.1 ppb using a Thermo Finnigan Element 1 sector field instrument. Due to the 129 Xe problem early measurements of 129 I using ICP-MS were focused on performing the analysis on a short as possible time to prevent the accumulated 129 Xe signal. Farmer et al. (1998) analyzed 129 I in air via a flow injection method where they investigated the possibility of thermal decomposition of a PdI2 precipitate collected on a filter. In order to avoid the iodine sorption on a large carbon surface they avoided the use of an ETV and instead constructed a thermal decomposition cell made of quartz. The quartz tube, heated to 700 ◦ C in about 40 s in order to decompose the PdI2 into Pd metal and I2 , was directly connected to the ICP-torch and was swept with argon that had passed a USN. The USN was set to nebulize distilled water which provided the otherwise dry argon with sufficient moisture to significantly reduce sorption tailing of the evolved iodine. Although with a significant 129 Xe background the effect of a short iodine pulse and by correcting for the 129 Xe contribution by monitoring 131 Xe they claimed a detection limit of 30 fg of 129 I. Bienvenu et al. (2004) used a Perkin-Elmer Elan 6000 to analyze 129 I in environmental samples collected around the La Hague. By correction of the 129 Xe contribution via 131 Xe (21.2% abundance) they reported a detection limit of 15 ppt in standard solutions. In synthetic solutions, similar to their sample solutions, containing 0.3% Na and 0.1% stable iodine they obtained detection limits around 50 ppt. Apart from the matrix suppression also the presence of 127 I1 H2 made analysis less reliable, especially since the formation of 127 I1 H2 could be expected to vary from sample to sample and changing instrument conditions. In spite of the relatively poor detection limits ICP-MS were considered significantly better than what could be obtained using radiometric methods which was about 0.1 Bq (15 ng) for a 15 h counting time using X-ray spectrometry. Even when using short acquisition times the contribution from 129 Xe is frequently too high. 129 Xe has a natural abundance of 26.4% and is present at about 20 ppb in the atmosphere. Although the argon used in the ICP analysis is of high purity, xenon is present also here. A significant step forward in reducing this troublesome interference has been through the in-
320
P. Roos
troduction of collision/reaction cell instruments. Specifically oxygen (O2 ) introduced in small quantities in the reaction cell has been found to significantly reduce xenon ions by charge transfer while at the same time reactions with iodide ions is only about 10%. Eiden et al. (1997) showed that the reaction rate of 129 Xe+ with O2 was 104 times faster than with 129 I+ . Further improvement could be achieved if the abundance sensitivity, or peak tailing, could be reduced so that the influence of 127 I on mass 129 could be minimal. The peak tailing is mostly due to the energy spread of the ions. By pressurizing the collision cell with an unreactive gas of low mass atoms, such as helium, the ions in the ion beam would be ‘thermalized’ and acquire a more homogeneous ion energy, thereby allowing for a better ion transmission and a lower tailing as was observed by, e.g., Boulyga and Becker (2002). Izmer et al. (2003) investigated the ion-transmission of 127 I and background on mass 129 (from 129 Xe and tailing from 127 I as well as contribution from 127 IH2 ) while introducing helium and oxygen in the collision cell. By introducing oxygen in the collision/reaction cell they managed to reduce the 129 m/z background to the instrument blank background and a mixture of oxygen/helium resulted in a reduced detection limit of 2 orders of magnitude. The detection limit reached in a later paper (Izmer et al., 2004) was 0.4 ppt with a 129 I/127 I ratio down to 10−7 . Brown et al. (2005) used a 1% tertiary amine carrier solution (Spectrasol CFA-C) to prevent buildup/memory of iodine in the ICP-MS introduction system and explored the use of a dynamic reaction cell (DRC) to remove 129 Xe by letting it react with O2 . They reported a dramatic reduction in the background at mass 129 by using the DRC (Perkin-Elmer Elan DRC II) but an interference of unknown origin appeared when injecting the O2 reaction gas which therefore prevented quantitative analysis of 129 I at low levels. They reported an instrument detection limit of about 10 ppt for 129 I in instrument standard analytical mode. 6. Analysis of 90 Sr, 135 Cs and 137 Cs The analysis of 90 Sr and 137 Cs has for some 50 years been conducted by standard radiometric methods. Strontium usually by radiochemical separation of the 90 Y daughter followed by beta counting, and 137 Cs by gamma spectrometry even though it earlier also was determined by beta counting after isolation. Could there possibly be any significant advantages of using ICP-MS when determining these relatively short-lived (30 years) radioisotopes? The use of ICP-MS is in no way a reasonable alternative when analyzing for 137 Cs since high-resolution gamma spectrometry can be used in both environmental and nuclear accident situations with much better identification power and sensitivity than ICP-MS. For 90 Sr the use of ICP-MS at environmental levels is not of interest due to inferior detection limits compared to radiometric methods but may become of interest in emergency events related to nuclear accidents when levels are higher, and a great number of short-lived radioisotopes complicate radiometric measurements based on traditional methods. Also the need for fast answers is a demand, which further warrants the use of ICP-MS. In routine measurements at nuclear power plants the use of ICP-MS may also be of interest for the same reasons that the number of short-lived radioisotopes may complicate radiometric measurements of 90 Sr. The main interference in the measurements of 90 Sr by ICP-MS is 90 Zr which has a natural abundance of 51% and which is also produced in the nuclear fission of 235 U and 239 Pu and
Analysis of radionuclides using ICP-MS
321
therefore not only present in environmental samples. Berryman and Probst (1997) studied the possibility of analyzing 90 Sr using a ETV system in order to discriminate against zirconium which have a very high boiling point. By selecting an appropriate temperature they managed to improve the relative Sr/Zr ratio by a factor of 50 as compared to conventional liquid nebulization. In spite of the improvements the detection limit reported for 90 Sr was still around 10 Bq/ml (equivalent to 2 pg/ml). Vonderheide et al. (2004) further investigated the possibilities of analyzing 90 Sr using sector-field ICP-MS (ThermoFinnigan Element) and collision-cell equipped ICP-MS (Micromass Platform). Although the introduction of oxygen in the collision cell allowed a removal of zirconium by about a factor of ten (as also showed earlier by Eiden et al., 1997) the detection limits for the Platform CC-ICP-MS was limited by the Daly detector noise to about 10 Bq/ml. Using the SF-ICP-MS in cold plasma mode (reduced rf-power and increased nebulizer gas flow) tuned to minimize the Zr signal (ionization potential 6.8 versus 5.7 V for Sr) resulted in significant reduction of the Zr signal. In this way detection limits of around 0.1 Bq/ml was achievable in pure water samples. In real samples, however, the levels of stable strontium may turn out to be the limiting parameter since the abundance sensitivity is not sufficiently good. In the work of Vonderheide et al. (2004) the detection limits in urine was about a factor of 20 higher than in pure water due to the presence of stable strontium. In a later work conducted by the same group Zoriy et al. (2005) showed that the determination of 90 Sr using the cold plasma method agreed well with 90 Sr determinations done on ground water samples in the 0.05–0.2 Bq/ml range. While the analysis of 137 Cs alone by mass spectrometric methods by no means would be of interest the analysis of 135 Cs certainly would. Being a pure beta emitter with a long half-life (2 million years) and a production rate in thermal nuclear fission reactors almost the same as for 137 Cs it would be of great interest to be able to determine this radioisotope by mass spectrometric methods. Since the ratio 135 Cs/137 Cs could be an interesting source term indicator the simultaneous mass spectrometric measurement of 137 Cs is partly motivated. Alternatively the 135 Cs could be normalized to stable cesium, which is monoisotopic at mass 133, and have no isobaric interference. The main problem in analyzing 135 Cs by ICP-MS is the isobar 135 Ba (natural abundance 6.6%) and in absence of stable barium the tailing from the 134 Xe and 136 Xe isotopes with natural abundance of 10.4 and 8.9%, respectively. Chemical removal of barium is thus necessary before measurement. Due to the great abundance of barium in nearly all material in normal laboratories sufficient removal is difficult. Further discrimination against barium may be done by using electrothermal vaporization of the sample. Song et al. (2001) tried using ETV to improve the Cs/Ba ratio by elaborating with temperature and different chemical modifiers and managed to achieve an absolute detection limit of 0.2 µBq (0.4 fg) for 135 Cs. Using KSCN as chemical modifier they enhanced the Cs signal around 60 times without significantly enhancing the barium signal. By further using a vaporisation temperature of 1100 ◦ C most of the cesium was vaporised while only some 0.03 per mille of the barium was vaporised. Apart from the possibility of discriminating against barium by using the ETV the short duration of the signal helps in the signal to background ratio with respect to the 134 Xe and 136 Xe isotopes present either in air or as impurities in the argon gas.
322
P. Roos
7. Analysis of 241 Am The analysis of 241 Am (or Am-isotopes) using ICP-MS has certain advantages when considering the analytical challenges in using conventional alpha spectrometry at environmental levels. In alpha spectrometry the sample amount for Am-isotopes is limited by the total amount of REE in the sample, which is difficult to completely remove, and which degrades the alpha spectrum making separation of the 243 Am and 241 Am peaks difficult. In mass spectrometry the presence of REE would at first not be considered a problem in the analysis of Am but also here they may introduce problems by causing interferences in combination with, e.g., phosphates (Nygren et al., 2005). Very few papers on Am measurements in environmental samples by mass spectrometry have been published. Agarande et al. (2001) analyzed 241 Am in environmental soils and reference material up to 100 g with good comparison to alpha spectrometry. More common are reports on the analysis of americium in spent nuclear fuel using ICP-MS (e.g., Chartier et al., 1999 or Wolf et al., 2005). 8. Analysis of 210 Pb Usually present in sufficient amount in surface soils and sediments 210 Pb is easily determined by gamma spectrometry using the weak 46 keV line even though the emission probability is only around 4% per decay. For most other environmental samples concentrations of 210 Pb is lower and the analysis is therefore made more sensitive by using the daughter isotopes 210 Bi (beta counting) or 210 Po (alpha counting). Direct beta counting of 210 Pb is difficult due to being a weak beta emitter (maximum energy 64 keV) and the lack of suitable yield determinants. Problems using the daughter products include long ingrowth time (210 Po) and lack of a suitable yield determinant (210 Bi). This makes it interesting to try alternative ways of determining 210 Pb at low concentrations. Larivière et al. (2005a, 2005b) investigated the possibility of using ICP-MS and evaluated different interferences using three different ICPMS instruments (sector field, quadrupole based and DRC equipped quadrupole ICP-MS). By converting lead into the volatile tetraethyl lead they could vaporize and transport the lead dry to the plasma thereby minimizing the contribution from hydride species like 208 PbH2 and 209 BiH. The tailing from 209 Bi and the stable lead isotopes, however, still constituted a major problem due to the relative poor abundance sensitivity of ICP-MS systems, especially for the sector field instruments. Due to that ethylated species of bismuth also are volatile this element is a major problem, chemical removal of this element prior to analysis is therefore imperative. In spite of the improved sample introduction technique and a thorough chemical clean-up of the samples the detection limits for a 1 L water sample was not better than 90 mBq/L (10 pg/L) after pre-concentration, mainly because of the tailing from the stable lead isotopes. Contrary to direct determination of 210 Pb by gamma spectrometry, the severe influence of the stable lead isotopes on the analysis makes it meaningless to increase the sample weight from which pre-concentration is made. This is not the case when using gamma spectrometry. It can therefore be concluded that as long as the sample amount available more or less is unlimited direct gamma spectrometry is a better alternative that ICP-MS. Future improvements in ICPMS technology may however change this. The introduction of collision cell technology may not remove the stable lead but the thermalization of ions improve abundance sensitivity and the removal of many interferences is possible (e.g., Epov et al., 2003).
Analysis of radionuclides using ICP-MS
323
9. Analysis of 226 Ra There are 4 isotopes of radium which are of interest in environmental radioactivity studies, (11 days), 224 Ra (3.6 days), 226 Ra (1600 y) and 228 Ra (5.8 y). Of these only 226 Ra is practically measurable using mass spectrometric methods. Even though 226 Ra is most easily determined using LSC (detection limit of around 1 mBq per sample) or gamma spectrometry, the use of mass spectrometric techniques is not without interest because of the very low instrumental background at higher masses and the rapid measurements. The analysis of 226 Ra in low-level LSC and gamma spectrometry frequently means measuring the daughter products (222 Rn and/or short-lived alpha and beta emitting radon daughters 218 Po, 214 Po and 214 Bi) after a sufficient build-up time (equilibrium reached after about 3 weeks), which would be avoided if instead measurements could be performed directly using mass spectrometry. Radium-226 has previously been determined by thermal ionization mass spectrometry, TIMS (e.g., Cohen and O’Nions, 1991) down to femtogram levels (1 fg equals 37 µBq of 226 Ra). Similarly as when using radiometric methods chemical isolation of radium is necessary when analyzing 226 Ra by ICP-MS even though there are no isobaric interferences and in fact the presence of other isotopes between masses 209 to 232 are minimal. The reason for separating radium before analysis is both due to the risk of polyatomic interferences and signal suppression due to a high salt load to the plasma. Since radium is an alkaline earth element high concentrations of barium, strontium and calcium may be expected to follow in the separation and polyatomic species like 88 Sr138 Ba may appear. Joannon and Pin (2001) observed severe signal suppression (<10% signal) when using conventional cation exchange to separate Ba+Ra from the large amounts of Sr and Ca present in an ordinary ground water sample. Separation methods sufficiently good for radiometric measurements using LSC will usually not be enough for ICP-MS measurements because of the incomplete matrix removal. Benkhedda et al. (2005), using a sector-field ICP-MS instrument (ELEMENT-2) combined with flow injection technique and a APEX-Q nebulizer (Elemental Scientific, USA), investigated the magnitude of interference at m/z 226 by adding different concentrations of strontium and barium to a sample matrix typical of their chemically separated radium fractions. They showed that the presence of even small amount like 0.6 µg/ml of strontium and 0.2 µg/ml barium combine to generate a signal at m/z 226 equivalent to 6 pg/L (0.22 Bq/L). Other potential interferences include 208 Pb18 O and 146 Nd40 Ar2 as well as several combinations of molybdenum isotopes with xenon isotopes. Due to a very sensitive instrument set up and thorough chemical cleanup of the sample, detection limits of about 0.5 mBq could be reached. This is comparable or somewhat better than what is achievable using LSC but the major advantage is that the instrumental analysis time is very short compared to radiometric methods. Larivière et al. (2005b), also using a ELEMENT-2 equipped with a APEX-Q nebulizer, reported detection limits for 226 Ra of only 0.1 mBq. Again, collision cell technology (CCT) may improve the possibilities of analyzing 226 Ra. Epov et al. (2003) reported on a 50-fold better sensitivity when using the CCT on a Platform ICP-QMS compared to when it was not used. One of the problems in analyzing radium by either mass spectrometry or using radiometric methods is the lack of suitable tracers. In mass spectrometry the only choice is the relatively short-lived 228 Ra (5.8 years half-life) which has been used frequently in both TIMS (e.g., Cohen and O’Nions, 1991) and multi-collector ICP-MS analysis (e.g., Foster et al., 2004) of 226 Ra. Foster et al. (2004) measured radium and barium in sea water using a Nu multicollector 223 Ra
324
P. Roos
instrument with three ion counting detectors and an Aridus Microconcentric nebulizer. The high sensitivity of the set-up made it possible to determine 226 Ra in 100 ml samples with a two sigma uncertainty of less than 10%. They, however, had to take precautions against an unknown polyatomic interference appearing both at mass 226 and 228. In general, even though non-radiometric techniques for determining radioisotopes have increased dramatically with the introduction of ICP-MS it is not only to the good. There is a certain curse in the development of more sensitive techniques in that it drives the collection of samples to successively smaller sample masses even in cases where much larges samples are readily available. Examples are marine waters collected by bottles holding several tens of liters but where only some hundred milliliters or less are used for the analysis. The gain in simplifying the separation chemistry for the small samples seems to be quickly lost by increasing demand in controlling blank levels and non-sample related interferences. If the main focus is to provide reliable data it would perhaps be intelligent to not only use the recent improvements in instrumentation but also to combine this with the vast knowledge in how to handle much larger samples than what is currently the standard.
10. Analysis of U and Th isotopes There are numerous examples of publications concerning mass spectrometry measurements of the natural composition of U and Th isotopes in the geological literature. TIMS was for many years the only reliable tool to measure 234 U/238 U and particularly 230 Th/232 Th ratios due mainly to the very good abundance sensitivity needed. For the measurement of 230 Th/232 Th ratios abundance sensitivity better than 10−6 to 10−7 is needed. Although quadrupole ICP-MS instruments have abundance sensitivities of around 10−6 or better, early instruments could not provide enough sensitivity. Later introduction of sector field instruments made it possible to achieve similar or even better sensitivities than TIMS instruments but abundance sensitivity in these instruments are usually poor (10−5 ). Increasing the resolution may improve the abundance sensitivity but at the cost of sensitivity. Modern quadrupole ICP-MS instruments as well as multi-collector sector field ICP-MS instruments provided with an extra energy filter have both the needed sensitivity and abundance sensitivity to handle 230 Th/232 Th measurements (e.g., Luo et al., 1997). The need for good abundance sensitivity is also the main requirement when measuring 236 U which otherwise will be disturbed both by 235 U and 238 U tailing as well as the 235 UH+ . Even though measurements of 236 U in environmental samples previously have been very scarce the use of depleted (recycled) uranium ammunition during the Balkan conflict and the Kuwait war as well as the presence of fuel fragments in the vicinity of the Chernobyl reactor have created some interest in measurement techniques for this isotope. Due to the difficulties in analyzing the 236 U by alpha spectrometry (alpha energies partly overlap with 235 U) mass spectrometry is a better alternative. Several attempts to analyze 236 U by ICP-MS have shown that it is possible at 236 U/238 U levels down to 10−7 using quadrupole instruments or sector field instruments used at medium or high resolution (e.g., Boulyga et al., 2002; Boulyga and Becker, 2001; Desideri et al., 2002). In an attempt to minimize the isobaric interference from 235 UH+ Zoriy et al. (2004a, 2004b) used D2 O as solvent instead of ordinary water and thereby managed to reduce the hydride interference at mass 236 to 10−6 (one magnitude lower than when ordinary
Analysis of radionuclides using ICP-MS
325
water was used) when also using a Aridus dry sample introduction system. By further using a medium resolution (m/m = 4450) applied to their ELEMENT SF-ICP-MS they managed to keep the abundance sensitivity down to around 10−6 both with respect to 235 U and 238 U.
References Agarande, M., Benzoubir, S., Bouisset, P., Calmet, D. (2001). Determination of 241 Am in sediments by isotope dilution high resolution inductively coupled plasma mass spectrometry (ID HR ICP-MS). Appl. Radiat. Isotopes 55, 161–165. Ancich, V.G. (2003). An index of the literature for biomolecular gas-phase cation–molecule reaction kinetics. JPL Publication 03-19, Jet Propulsion Laboratory, Pasadena, CA. Anderson, T.J., Walker, R.L. (1980). Determination of pictogram amounts of Tc-99 by resin bead mass-spectrometric isotope-dilution. Anal. Chem. 52, 709–713. Assinder, D.J. (1999). A review of the occurrence and behavior of neptunium in the Irish Sea. J. Environ. Radioact. 44, 335–347. Ayranov, M., Krähenbühl, U., Sahli, H., Röllin, S., Burger, M. (2005). Determination of neptunium in soil by ICPMS. Radiochim. Acta 93, 631–635. Baglan, N., Bouvier-Capely, C., Cossonnet, C. (2002). Determination of 237 Np at trace level: evaluation of various analytical procedures. Radiochim. Acta 90, 267–272. Baglan, N., Hemet, P., Pointurier, F., Chiappini, R. (2004). Evaluation of a single collector, double focusing sector field inductively coupled plasma mass spectrometer for the determination of U and Pu concentrations and isotopic compositions at trace level. J. Radioanal. Nucl. Chem. 261 (3), 609–617. Bandura, D.R., Baranov, V.I., Tanner, S.D. (2000). Effect of collisional dampening and reactions in a dynamic reaction cell on the precision of isotope ratio measurements. J. Anal. At. Spectrom. 15, 921–928. Beals, D.M. (1995). Determination of technetium-99 in aqueous samples by isotope dilution inductively coupled plasma mass spectrometry. J. Radioanal. Nucl. Chem. 204 (2), 253–263. Beasley, T.M., Kelley, J.M., Maiti, T.C., Bond, L.A. (1998). 237 Np/239 Pu atom ratios in integrated global fallout: a reassessment of the production of 237 Np. J. Environ. Radioact. 38, 133–146. Becker, J.S., Dietze, H.-J. (1999). Precise isotope ratio measurements of uranium, thorium and plutonium by quadrupole-based inductively coupled plasma mass spectrometry. Fresenius J. Anal. Chem. 364, 482–488. Becker, J.S., Zoriy, M., Halicz, L., Teplyakov, N., Muller, C., Segal, I., Pickhardt, C., Platzner, I.T. (2004). Environmental monitoring of plutonium at ultratrace level in natural water (Sea of Galilee—Israel) by ICP-SFMS and MC-ICP-MS. J. Anal. At. Spectrom. 19, 1257–1261. Benkhedda, K., Larivière, D., Scott, S., Evans, D. (2005). Hyphenation of flow injection on-line preconcentration and ICP-MS for the rapid determination of 226 Ra in natural waters. J. Anal. At. Spectrom. 20, 523–528. Berryman, N., Probst, T. (1997). Rapid determination of 90 Sr by electrothermal vaporization–inductively coupled plasma mass spectrometry (ETV-ICP-MS). Radiochim. Acta 76, 191–195. Bienvenu, P., Brochard, E., Excoffier, E., Piccione, M. (2004). Determination of iodine-129 by ICP-QMS in environmental samples. Can. J. Anal. Sci. Spectrosc. 49 (6), 423–428. Blaum, K., Geppert, Ch., Müller, P., Nörtershäuser, W., Otten, E.W., Schmitt, A., Trautmann, N., Wendt, K., Bushaw, B.A. (1998). Properties and performance of a quadrupole mass filter used for resonance ionization mass spectrometry. Int. J. Mass Spectrom. 181, 67–87. Boulyga, S.F., Becker, J.S. (2001). Determination of uranium isotopic composition and 236 U content of soil samples and hot particles using inductively coupled plasma mass spectrometry. Fresenius J. Anal. Chem. 370, 612–617. Boulyga, S.B., Becker, J.S. (2002). Improvement of abundance sensitivity in a quadrupole-based ICP-MS instrument with a hexapole collision cell. J. Anal. At. Spectrom. 17, 1202–1206. Boulyga, S.B., Ditze, H.-J., Becker, J.S. (2001). Performance of ICP-MS with hexapole collision cell and application for determination of trace elements in bio-assays. Mikrochim. Acta 137, 93–103. Boulyga, S.F., Matusevich, J.L., Mironov, V.P., Kudrjasov, V.P., Halicz, L., Segal, I., McLean, J.A., Montaser, A., Becker, J.S. (2002). Determination of 236 U/238 U isotopic ratio in contaminated environmental samples using different ICP-MS instruments. J. Anal. At. Spectrom. 17, 958–964.
326
P. Roos
Brown, C.F., Geiszler, K.N., Vickerman, T.S. (2005). Extraction and quantitative analysis of iodine in solid and solution matrixes. Anal. Chem. 77 (21), 7062–7066. Buesseler, K.O., Halversen, J. (1987). The mass spectrometric analysis of fallout 239 Pu and 240 Pu in marine samples. J. Environ. Radioact. 5 (6), 425–444. Chartier, F., Aubert, M., Pilier, M. (1999). Determination of Am and Cm in spent nuclear fuels by isotope dilution inductively coupled plasma mass spectrometry and isotope dilution thermal ionization mass spectrometry after separation by high-performance liquid chromatography. Fresenius J. Anal. Chem. 364, 320–327. Chen, J.H., Edwards, R.L., Wasserburg, G.J. (1986). 238 U, 234 U and 232 Th in seawater. Earth Planet. Sci. Lett. 80 (3–4), 241–251. Chen, J.H., Edwards, R.L., Wasserburg, G.J. (1992). Mass spectrometry and applications to uranium-series disequilibrium. In: Ivanovich, M., Harmon, R.S. (Eds.), Uranium-Series Disequilibrium, Applications to Earth, Marine, and Environmental Sciences. Clarendon Press, Oxford (Chapter 6). Chen, Q.J., Dahlgaard, H., Nielsen, S.P., Aarkrog, A. (2002). Pu-242 as tracer for simultaneous determination of Np-237 and Pu-239, Pu-240 in environmental samples. J. Radioanal. Nucl. Chem. 253 (3), 451–458. Cheng, H., Edwards, R.L., Hoff, J., Gallup, C.D., Richards, D.A., Asmerom, Y. (2000). The half-lives of uranium-234 and thorium-230. Chem. Geol. 169 (1–2), 17–33. Cohen, A.S., O’Nions, R.K. (1991). Precise determination of femtogram quantities of radium by thermal ionization mass-spectrometry. Anal. Chem. 63 (23), 2705–2708. Cooper, L.W., Kelley, J.M., Bond, L.A., Orlandini, K.A., Grebmeier, J.M. (2000). Sources of the transuranic elements plutonium and neptunium in arctic marine sediments. Mar. Chem. 69 (3–4), 253–276. Deschamps, P., Doucelance, R., Ghaleb, B., Michelot, J.-L. (2003). Further investigations on optimized tail correction and high-precision measurement of uranium isotopic ratios using multi-collector ICP-MS. Chem. Geol. 201, 141– 160. Douglas, D.J. (1989). Some current perspectives on ICP-MS. Can. J. Spectrosc. 34 (2), 38–49. Douglas, D.J., French, J.B. (1992). Collisional focusing effects in radio frequency quadrupoles. J. Am. Soc. Mass Spectrom. 3, 398–408. Downey, S.W., Nogar, N.S., Miller, C.M. (1984). Resonance ionization mass-spectrometry of technetium. Int. J. Mass Spectrom. Ion Processes 61 (3), 337–345. Desideri, D., Meli, M.A., Roselli, C., Testa, C., Boulyga, S.F., Becker, J.S. (2002). Determination of 236 U and transuranium elements in depleted uranium ammunition by α-spectrometry and ICP-MS. Anal. Bioanal. Chem. 374, 1091–1095. Egorov, O.B., O’Hara, M.J., Farmer III, O.T., Grate, J.W. (2001). Extraction chromatographic separation and analysis of actinides using sequential injection techniques with on-line inductively coupled plasma mass spectrometry (ICP-MS) detection. Analyst 126, 1594–1601. Eiden, G.C., Barinaga, C.J., Koppenaal, D.W. (1996). Selective removal of plasma matrix ions in plasma source mass spectrometry. J. Anal. At. Spectrom. 11 (4), 317–322. Eiden, G.C., Barinaga, C.J., Koppenaal, D.W. (1997). Beneficial ion/molecule reactions in elemental mass spectrometry. Rapid Commun. Mass Spectrom. 11 (1), 37–42. Epov, V.N., Taylor, V., Larivière, D., Evans, R.D., Cornett, R.J. (2003). Collision cell chemistry for the analysis of radioisotopes by inductively coupled plasma mass spectrometry. J. Radioanal. Nucl. Chem. 258 (3), 473–482. Farmer, O.T., Barinaga, C.J., Koppenaal, D.W. (1998). Determination of 129 I in ambient air by inductively coupled plasma mass spectrometry (ICP/MS). J. Radioanal. Nucl. Chem. 234 (1–2), 153–157. Feldmann, I., Jakubowski, N., Stuewer, D. (1999a). Application of a hexapole collision and reaction cell in ICP-MS. Part I: Instrumental aspects and operational optimization. Fresenius J. Anal. Chem. 365, 415–421. Feldmann, I., Jakubowski, N., Thomas, C., Stuewer, D. (1999b). Application of a hexapole collision and reaction cell in ICP-MS. Part II: Analytical figures of merit and first applications. Fresenius J. Anal. Chem. 365, 422–428. Foster, D.A., Staubwasser, M., Henderson, G.M. (2004). 226 Ra and Ba concentrations in the Ross Sea measured with multicollector ICP mass spectrometry. Mar. Chem. 87, 59–71. Gray, A.L. (1986). Influence of load-coil geometry on oxide and doubly charged ion response in inductively coupled plasma source–mass spectrometry. J. Anal. At. Spectrom. 1 (3), 247–249. Haldiman, M., Eastgate, A., Zimmerli, B. (2000). Improved measurement of iodine in food using inductively coupled plasma isotope dilution mass spectrometry. Analyst 125, 1977–1982. Harvey, B.R., Williams, K.J., Lovett, M.B., Ibbett, R.D. (1992). Determination of Tc-99 in environmental material with rhenium as a yield monitor. J. Radioanal. Nucl. Chem. 158 (2), 417–436.
Analysis of radionuclides using ICP-MS
327
Hill, D.M., Barnes, R.K., Wong, H.K.Y., Zawadzki, A.W. (2000). The quantification of technetium in generatordriven pertechnetate using ICP-MS. Appl. Radiat. Isotopes 53 (3), 415–419. Holm, E., Aarkrog, A., Ballestra, S. (1987). Determination of Np-237 in large volume of sea-water by a radiochemical procedure. J. Radioanal. Nucl. Chem. 115 (1), 5–11. Hou, X. (2004). Application of 129 I as an environmental tracer. J. Radioanal. Nucl. Chem. 262 (1), 67–75. Hou, X.L., Dahlgaard, H., Rietz, B., Jacobsen, U., Nielsen, S.P., Aarkrog, A. (1999). Determination of I-129 in seawater and some environmental materials by neutron activation analysis. Analyst 124 (7), 1109–1114. Izmer, A.V., Boulyga, S.F., Becker, J.S. (2003). Determination of 129 I/127 I isotope ratio in liquid solutions and environmental soil samples by ICP-MS with hexapole collision cell. J. Anal. At. Spectrom. 18, 1339–1345. Izmer, A.V., Boulyga, S.F., Zoriy, M.V., Becker, J.S. (2004). Improvement of the detection limit for determination of 129 I in sediments by quadrupole inductively coupled plasma mass spectrometry with collision cell. J. Anal. At. Spectrom. 19, 1278–1280. Ji, Y.-Q., Li, J.-Y., Luo, S.-G., Wu, T., Liu, J.-L. (2001). Determination of traces of 237 Np in environmental samples by ICP-MS after separation using TOA extraction chromatography. Fresenius J. Anal. Chem. 371, 49–53. Jiang, S.-J., Houk, R.S., Stevens, M.A. (1988). Alleviation of overlap interferences for determination of potassium isotope ratios by inductively coupled plasma mass spectrometry. Anal. Chem. 60, 1217–1221. Joannon, S., Pin, C. (2001). Ultra-trace determination of 226 Ra in thermal waters by high sensitivity quadrupole ICP–mass spectrometry following selective extraction and concentration using radium-specific membrane discs. J. Anal. At. Spectrom. 16, 32–37. Johansson, L., Holm, E. (1996). Determination of trace-amounts plutonium using fission track analysis. Nucl. Instrum. Methods A 376, 242–247. Keith-Roach, M., Sturup, S., Oughton, D.H., Dahlgaard, H. (2002). Comparison of two ICP-MS set-ups for measuring 99 Tc in large volume water samples. Analyst 127, 70–75. Kelley, J.M., Bond, L.A., Beasley, T.M. (1999). Global distribution of Pu isotopes and 237 Np. Sci. Total Environ. 237/238, 483–500. Kenna, T.C. (2002). Determination of plutonium isotopes and neptunium-237 in environmental samples by inductively coupled plasma mass spectrometry with total dissolution. J. Anal. At. Spectrom. 17 (11), 1471–1479. Kerls, W., Becker, J.S., Dietze, H.J., Dannecker, W. (1996). Determination of iodine using a special sample introduction system coupled to a double-focusing sector field inductively coupled plasma mass spectrometer. J. Anal. At. Spectrom. 11, 723–726. Ketterer, M.E., Hafer, K.M., Link, C.L., Kolwaite, D., Wilson, J., Mietelski, J.W. (2004). Resolving global versus local/regional Pu sources in the environment using sector ICP-MS. J. Anal. At. Spectrom. 19, 241–245. Kieser, W.E., Zhao, X.L., Soto, C.Y., Tracy, B. (2005). Accelerator mass spectrometry of 129 I: Technique and applications. J. Radioanal. Nucl. Chem. 263 (2), 375–379. Kim, C.K., Otsuji, M., Takaku, Y., Kawamura, H., Shiraishi, K., Igarashi, Y., Igarashi, S., Ikeda, N. (1989). Application of inductively coupled plasma mass spectrometry to the determination of 99 Tc in soil samples. Radioisotopes 38, 151–152. Kim, C.-S., Kim, C.-K., Lee, J.-I., Lee, K.-J. (2000). Rapid determination of Pu isotopes and atom ratios in small amounts of environmental samples by an on-line sample pre-treatment system and isotope dilution high resolution inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom. 15, 247–255. Kim, C.K., Kim, C.S., Rho, B.H., Lee, J.I. (2002). Rapid determination of 99 Tc in environmental samples by high resolution ICP-MS coupled with on-line flow injection. J. Radioanal. Nucl. Chem. 252 (2), 421–427. Kim, C.S., Kim, C.K., Lee, K.J. (2004). Simultaneous analysis of 237 Np and Pu isotopes in environmental samples by ICP-SF-MS coupled with automated sequential injection system. J. Anal. At. Spectrom. 19, 743–750. Kuczewski, B., Marquardt, C.M., Seibert, A., Geckeis, H., Kratz, J.V., Trautmann, N. (2003). Separation of plutonium and neptunium species by capillary electrophoresis–inductively coupled plasma–mass spectrometry and application to natural groundwater samples. Anal. Chem. 75, 6769–6774. La Rosa, J., Gastaud, J., Lagan, L., Lee, S.-H., Levy-Palomo, I., Povinec, P.P., Wyse, E. (2005). Recent developments in the analysis of transuranics (Np, Pu, Am) in seawater. J. Radioanal. Nucl. Chem. 263 (2), 427–436. Larivière, D., Reiber, K.M., Evans, R.D., Cornett, R.J. (2005a). Determination of 210 Pb at ultra-trace levels in water by ICP-MS. Anal. Chim. Acta 549, 188–196. Larivière, D., Epov, V.N., Reiber, K.M., Cornett, R.J., Evans, R.D. (2005b). Micro-extraction procedures for the determination of Ra-226 in well waters by SF-ICP-MS. Anal. Chim. Acta 528, 175–182.
328
P. Roos
Lee, S.C., Hutchinson, J.M.R., Inn, K.G.W., Thein, M. (1995). An intercomparison study of 237 Np determination in artificial urine samples. Health Phys. 68 (3), 350–358. Liao, C.-L., Xu, R., Nourbakhsh, G.D., Baer, M., Ng, C.Y. (1990). Experimental and theoretical total state-selected and state-to-state absolute cross-sections. II. The Ar+ (2 P3/2,1/2 ) + H2 reaction. J. Chem. Phys. 93 (7), 4832– 4844. Luo, X., Rehkämper, M., Lee, D.-C., Halliday, A.N. (1997). High precision 230 Th/232 Th and 234 U/238 U measurements using energy-filtered ICP magnetic sector multiple collector mass spectrometry. Int. J. Mass Spectrom. Ion Processes 171, 105–117. MacCartney, M., Rajendran, K., Olive, V., Busby, R.G., McDonald, P. (1999). Development of a novel method for the determination of 99Tc in environmental samples by ICP-MS. J. Anal. At. Spectrom. 14, 1849–1852. Mas, J.L., Garcia-Leon, M., Bolivar, J.P., Sanchez-Angulo, C. (2000). The use of 99m Tc as a tracer in the determination of 99 Tc by ICP–mass spectrometry. J. Anal. At. Spectrom. 15, 1369–1373. Mas, J.L., Garcia-Leon, M., Bolivar, J.P. (2002). 99 Tc atom counting by quadrupole ICP-MS. Optimisation of the instrumental response. Nucl. Instrum. Methods Phys. Res. Sect. A 484, 660–667. Mas, J.L., Garcia-Leon, M., Bolivar, J.P. (2004a). 99 Tc detection in water samples by ICP-MS. Radiochim. Acta 92, 39–46. Mas, J.L., Tagami, K., Uschida, S. (2004b). Method for the detection of Tc in seaweed samples coupling the use of Re as a chemical tracer and isotope dilution inductively coupled plasma mass spectrometry. Anal. Chim. Acta 509, 83–88. Mattsson, S. (1978). Tc-99 in instant pertechnetate Tc-99m. J. Radioanal. Chem. 45 (2), 341–347. McMillan, E.M., Abelson, P.H. (1940). Radioactive element 93. Phys. Rev. 57, 1185. Muramatsu, Y., Yoshida, S. (1995). Determination of 129 I and 127 I in environmental samples by neutron activation analysis (NAA) and inductively coupled plasma mass spectrometry (ICP-MS). J. Radioanal. Nucl. Chem. 197, 149–159. Nygren, U., Rodushkin, I., Nilsson, C., Baxter, D. (2003). Separation of plutonium from soil and sediment prior to determination by inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom. 18, 1426–1434. Nygren, U., Ramebäck, H., Baxter, D.C., Nilsson, C. (2005). Lanthanide phosphate interferences in actinide determination using inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom. 20, 529–534. Oughton, D.H., Skipperud, L., Fifield, L.K., Cresswell, R.G., Salbu, B., Day, P. (2004). Accelerator mass spectrometry measurement of 240 Pu/239 Pu isotope ratios in Novaya Zemlya and Kara Sea sediments. Appl. Radiat. Isotopes 61, 249–253. Pointurier, F., Baglan, N., Hemet, P. (2004). Ultra low-level measurements of actinides by sector field ICP-MS. Appl. Radiat. Isotopes 60, 561–566. Povinec, P.P., Oregioni, B., Jull, A.J.T., Kieser, W.E., Zhao, X.-L. (2000). AMS measurements of 14 C and 129 I in seawater around radioactive waste dump sites. Nucl. Instrum. Methods Phys. Res. Sect. B 172, 672–678. Rowan, J.T., Houk, R.S. (1989). Attenuation of polyatomic ion interferences in inductively coupled plasma mass spectrometry by gas-phase collision. Appl. Spectrosc. 43, 976–980. Rubin, K.H. (2001). Analysis of 232 Th/230 Th in volcanic rocks: a comparison of thermal ionization mass spectrometry and other methodologies. Chem. Geol. 175 (3–4), 723–750. Sakata, K., Kawabata, K. (1994). Reduction of fundamental polyatomic ions in inductively coupled plasma–mass spectrometry. Spectrochim. Acta B 49, 1027. Santos, F.J., López-Gutiérrez, J.M., Garcia-León, M., Suter, M., Synal, H.-A. (2005). Determination of 129 I/127 I in aerosol samples in Seville (Spain). J. Environ. Radioact. 84, 103–109. Schmidt, A., Schnabel, Ch., Handl, J., Jakob, D., Michel, R., Synal, H.-A., Lopez, J.M., Suter, M. (1998). On the analysis of iodine-129 and iodine-127 in environmental materials by accelerator mass spectrometry and ion chromatography. Sci. Total Environ. 223 (2–3), 131–156. Sekine, T., Konishi, M., Kudo, H., Tagami, K., Uchida, S. (1999). Separation of carrier free Tc-95m from niobium targets irradiated with alpha-particles. J. Radioanal. Nucl. Chem. 239 (3), 483–487. Song, M., Probst, T.U. (2000). Rapid determination of technetium-99 by electrothermal vaporization–inductively coupled plasma–mass spectrometry with sodium chlorate and nitric acid as modifiers. Anal. Chim. Acta 413, 207–215. Song, M., Probst, T.U., Berryman, N.G. (2001). Rapid and sensitive determination of radiocesium (Cs-135, Cs-137) in the presence of excess barium by electrothermal vaporization–inductively coupled plasma–mass spectrometry (ETV-ICP-MS) with potassium thiocyanate as modifier. Fresenius J. Anal. Chem. 370, 744–751.
Analysis of radionuclides using ICP-MS
329
Stewart, I.I., Olesik, J.W. (1998). The effect of nitric acid concentration and nebulizer gas flow rates on aerosol properties and transport rates in inductively coupled plasma sample introduction. J. Anal. At. Spectrom. 13, 1249– 1256. Sturup, S., Dahlgaard, H., Nielsen, S.C. (1998). High resolution inductively coupled plasma mass spectrometry for the trace determination of plutonium isotopes and isotope ratios in environmental samples. J. Anal. At. Spectrom. 13, 1321–1326. Tagami, K., Uchida, S. (1996). Analysis of technetium-99 in soil and deposition samples by inductively coupled plasma mass spectrometry. Appl. Radiat. Isotopes 47 (9–10), 1057–1060. Tagami, K., Uchida, S. (2000). Separation of rhenium by an extraction chromatographic resin for determination by inductively coupled plasma–mass spectrometry. Anal. Chim. Acta 405, 227–229. Takaku, Y. (2005). Measurement of environmental and biological radionuclides. In: Nelms, S.M. (Ed.), ICP Mass Spectrometry Handbook. Blackwell Publishing, CRC Press, Boca Raton, Florida, USA. Tanner, S. (1995). Characterization of ionization and matrix suppression in inductively coupled ‘cold’ plasma mass spectrometry. J. Anal. At. Spectrom. 10, 905–921. Tanner, S.D., Baranov, V.I., Bandura, D.R. (2002). Reaction cells and collision cells for ICP-MS: a tutorial review. Spectrochim. Acta B 57, 1361–1452. Tanner, S.D., Li, C., Vais, V., Baranov, V.I., Bandura, D.R. (2004). Chemical resolution of Pu+ from U+ and Am+ using a band-pass reaction cell inductively coupled plasma mass spectrometer. Anal. Chem. 76, 3042–3048. Taylor, R.N., Warneke, T., Milton, J.A., Croudace, I.W., Warwick, P.E., Nesbitt, R.W. (2001). Plutonium isotope ratio analysis at femtogram to nanogram levels by multicollector ICP-MS. J. Anal. At. Spectrom. 16, 279–284. Thirlwall, M.F. (2001). Inappropriate tail corrections can cause large inaccuracy in isotope ratio determination by MC-ICP-MS. J. Anal. At. Spectrom. 16, 1121–1125. Ting, B.G., Pappas, R.S., Paschal, D.C. (2003). Rapid analysis for plutonium-239 in urine by magnetic sector inductively coupled plasma–mass spectrometry using Aridus desolvation introduction system. J. Anal. At. Spectrom. 18, 795–797. Uschida, S., Tagami, K., Ruhm, W., Steiner, M., Wirth, E. (2000). Separation of Tc-99 in soil and plant samples collected around the Chernobyl reactor using a Tc-selective chromatographic resin and determination of the nuclide by ICP-MS. Appl. Radiat. Isotopes 53, 69–73. Vais, V., Li, C., Cornett, J. (2004a). Preventing uranium hydride formation in standard uranium samples for determination of 239 Pu by ICP-MS. J. Anal. At. Spectrom. 19, 1281–1283. Vais, V., Li, C., Cornett, J. (2004b). Separation of plutonium from uranium using reactive chemistry in a bandpass reaction cell of an inductively coupled plasma mass spectrometer. Anal. Bioanal. Chem. 380, 235–239. Vance, D.E., Belt, V.F., Oatts, T.J., Mann, D.K. (1998). Neptunium determination by inductively coupled plasma mass spectrometry (ICP-MS. J. Radioanal. Nucl. Chem. 234 (1–2), 143–146. Vanhaecke, F., Moens, L. (2004). Overcoming spectral overlap in isotopic analysis via single- and multi-collector ICP–mass spectrometry. Anal. Bioanal. Chem. 378, 232–240. Vonderheide, A.P., Myroslav, V.Z., Izmer, A.V., Pickhardt, C., Caruso, J.A., Ostapczuk, P., Hille, R., Becker, J.S. (2004). Determination of 90 Sr at ultratrace levels in urine by ICP-MS. J. Anal. At. Spectrom. 19, 675–680. Wacker, L., Fifield, L.K., Tims, S.G. (2004). Developments in AMS of Tc-99. Nucl. Instrum. Methods Phys. Res. Sect. B 223, 185–189. Wasserman, H.J., Fourie, P.J. (1981). Excessive Co-60 in Mo-99–Tc-99(m) generator columns. Phys. Med. Biol. 26 (2), 319–320. Weyer, S., Schwieters, J. (2003). High precision Fe isotope measurements with high mass resolution MC-ICPMS. Int. J. Mass Spectrom. 226 (3), 355–368. Wolf, S.F., Bowers, D.L., Cunnane, J.C. (2005). Analysis of high burnup spent nuclear fuel by ICP-MS. J. Radioanal. Nucl. Chem. 263 (3), 581–586. Wyse, E.J., Lee, S.H., La Rosa, J., Povinec, P., de Mora, S.J. (2001). ICP-sector field mass spectrometry analysis of plutonium isotopes: recognizing and resolving potential interferences. J. Anal. At. Spectrom. 16, 1107–1111. Yamamoto, M., Kofuji, H., Tsumura, A., Yamasaki, S., Yuita, K., Komamura, M., Komura, K., Ueno, K. (1994). Temporal feature of global fallout 237 Np deposition in paddy field through the measurement of low-level 237 Np by high resolution ICP-MS. Radiochim. Acta 64, 217–224. Zoriy, M.V., Halicz, L., Ketterer, M.E., Pickhardt, C., Ostapczuk, P., Becker, J.S. (2004a). Reduction of UH+ formation for 236 U/238 U isotope ratio measurements at ultratrace level in double focusing sector field ICP-MS using D2 O as solvent. J. Anal. At. Spectrom. 19, 362–367.
330
P. Roos
Zoriy, M.V., Pickhardt, C., Ostapczuk, P., Hille, R., Becker, J.S. (2004b). Determination of Pu in urine at ultratrace level by sector field inductively coupled plasma mass spectrometry. Int. J. Mass Spectrom. 232, 217–224. Zoriy, M.V., Ostapczuk, P., Halicz, L., Hille, R., Becker, J.S. (2005). Determination of 90 Sr and Pu isotopes in contaminated groundwater samples by inductively coupled plasma mass spectrometry. Int. J. Mass Spectrom. 242, 203–209.
331
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides N. Erdmanna,b,∗ , G. Passlerc , N. Trautmanna , K. Wendtc a Institute of Nuclear Chemistry, Johannes Gutenberg-University Mainz, Fritz-Strassmann-Weg 2,
D-55128 Mainz, Germany b European Commission Joint Research Centre, Institute for Transuranium Elements, P.O. Box 2340,
D-76125 Karlsruhe, Germany c Institute of Physics, Johannes Gutenberg-University Mainz, Staudinger Weg 7, D-55128 Mainz, Germany
Abstract Resonance ionization mass spectrometry (RIMS) is a sensitive and selective method for the determination of extremely low abundances of long-lived radionuclides. The detection limits are about 106 atoms per sample and an isotopic selectivity up to 1013 has been achieved. The potential of RIMS using different experimental arrangements is outlined for the determination of isotope ratios and lowest abundances of long-lived radioisotopes of interest like 238–244 Pu, 90 Sr, and 41 Ca. Recent developments in improving detection limits and the spatial resolution of this technique are briefly discussed.
1. Introduction Trace analysis of long-lived radionuclides is required for nuclear safety assessment, in environmental science, chemistry and physics and other scientific fields. These radionuclides have various sources of origin (Tykva and Sabol, 1995), and one can distinguish between natural radioactivity including primordial and cosmogenic species, and anthropogenic radioactivity, among others produced by nuclear explosions, releases from nuclear power stations, medical applications, etc. Radionuclides have been dispersed into the environment with consequences that their concentrations are rather low and thus very sensitive methods are needed for their detection. Usually radiometric techniques are used for the determination of the content of radioactive isotopes. However, they have some disadvantages, especially for the long-lived ones, due to long measuring times, insufficient selectivity and background problems. For example, alphaspectroscopic determination of plutonium is not an isotope selective analysis because it is difficult to distinguish between 239 Pu and 240 Pu due to their very similar α-energies (239 Pu: 5.157 MeV; 240 Pu: 5.168 MeV). Furthermore, the detection limit and the measuring time ∗ Corresponding author. Address: European Commission Joint Research Centre, Institute for Transuranium Elements, P.O. Box 2340, D-76125 Karlsruhe, Germany. E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11010-X
© 2008 Elsevier B.V. All rights reserved.
332
N. Erdmann et al.
depend on the half-life of the isotope under investigation, e.g., a detection limit of 4 × 108 atoms for 239 Pu (T1/2 = 2.41 × 104 a) in 1000 min counting time has been obtained (Peuser et al., 1981). Therefore, a direct counting of radionuclides is more favored even though it can be hampered by a large surplus of neighboring stable isotopes as well as elemental or molecular isobars. The abundances of long-lived radionuclides can vary between 10−8 –10−12 , or even lower of the total sample or the neighboring isotope content. This implies that only detection methods with a good elemental and isotopic selectivity, a high efficiency and effective background suppression are applicable. A moderate isotopic selectivity, combined—as a rule—with lowest detection limits is necessary in the case when the element consists only of radionuclides, and no stable isotopes are present, as for the actinides with the exception of uranium and thorium. Conventional mass spectrometric methods like inductively coupled plasma mass spectrometry (ICP-MS) (Becker, 2003; Becker and Dietze, 1999a, 1999b, 2000; Boulyga et al., 2001; Huber et al., 2003; Ketterer et al., 2004; Lariviere et al., 2006; Wyse et al., 2001; Zoriy et al., 2004) or thermal ionization mass spectrometry (TIMS) (Dai et al., 2001; Wallenius and Mayer, 2000; Wallenius et al., 2000) are independent of the isotope half-life, but the analysis may suffer from isobaric interferences due to the non-selective ionization process. However, very recently isotopic ratio measurements with ICP-MS yielding a detection limit of less than 107 atoms of plutonium using a multi-collector sector-field ICP-MS system have been reported (Boulyga et al., 2003; Zoriy et al., 2004). For applications requiring an extremely high isotopic selectivity and good isobaric suppression, accelerator mass spectrometry (AMS) can be used. This technique provides isotopic selectivity up to 1015 and a detection limit of ∼104 atoms (Fifield et al., 1996; Tuniz et al., 1998; Oughton et al., 2004) by combining different selection stages. With a new generation of AMS operated at low terminal voltages (Wacker et al., 2005), sensitivities approaching 106 atoms for plutonium isotopes were achieved (Fifield et al., 2004). The development of lasers as strong and quasi-monochromatic light sources with a good tuneability has brought a significant progress in ultratrace analysis. The optical excitation of atoms by resonant absorption of laser light in a step-wise process followed by the photoionization of the excited atom can nowadays be performed in a very efficient way. Combined with selective and nearly background free detection of the resulting photo-ions in mass spectrometers, this technique is the basis of resonance ionization mass spectrometry (RIMS) (Hurst and Payne, 1988; Letokhov, 1987). The RIMS-method has been applied in the last years for the isotopically selective ultratrace analysis of long-lived radionuclides with detection limits down to 106 atoms and isotopic selectivity of up to 1013 (Huber et al., 2003; Trautmann et al., 2004; Wendt and Trautmann, 2005). In the present paper, the general concept of RIMS is presented, experimental arrangements for the determination of long-lived radionuclides are described and some applications of the RIMS method are discussed. 2. Principles of resonance ionization mass spectrometry As outlined, atom counting is very often more effective in ultratrace determination of longlived radionuclides than decay counting. The standard mass spectrometric techniques have
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
333
some limitations, mainly caused by isobaric interferences or when very high isotopic selectivity (1010 ) is required. In 1972, RIMS was proposed (Ambartzumian and Letokhov, 1972) as a technique to circumvent these disadvantages. The conceptual idea of RIMS is the application of selective resonant optical excitation and ionization processes, leading to an extreme element selectivity and thus suppression of isobaric interferences. RIMS offers several outstanding properties: (i) Almost complete suppression of atomic or molecular isobaric interferences due to the element selective optical excitation and ionization processes. (ii) High overall efficiency resulting in low detection limits in the fg-range (∼106 atoms) due to the high cross-sections for optical excitations. (iii) Good isotopic resolution, based on the selective ion detection in a mass spectrometer. (iv) Feasibility of ultra-high isotopic selectivity by taking advantage of the isotope shift in the atomic transitions using narrow-band lasers, in addition to the abundance sensitivity of the mass spectrometer. The individual components necessary for the RIMS procedure are briefly described below. 2.1. Atom sources RIMS generally requires neutral atoms in the gas phase in order to realize the selective optical excitation and ionization steps of the element under investigation. Thus, after a chemical separation, the resulting analyte must be efficiently vaporized and atomized within the RIMS apparatus. A number of different atomization techniques can be used, depending basically on the particular task to be accomplished. A simple way to produce an atomic beam is thermal evaporation from a filament or a small vessel by resistive heating. This process is associated with a rather large spatial distribution and velocity spread within the resulting atomic cloud. High laser power and a large overlap area are needed, and the spectral line width of the laser beams must cover the full Doppler broadening within the atomic cloud to ensure a good efficiency. As the sample is usually evaporated continuously, cw-lasers or pulsed laser systems with repetition rates in the kHz range must be used to avoid significant losses in efficiency. In order to obtain a better collimation and localization of the atomic beam, the chemically separated sample is evaporated inside a cylindrical oven with a length l and a bore diameter d (d l). The resulting aperture angle α of the atomic beam can be estimated to be α ∼ arctan 2d/ l (Ramsey, 1956). In this way, the interaction volume between laser beams and atoms is smaller, with the consequence that less laser power is needed. The collimation also has the advantage that the velocity spread perpendicular to the effusing atomic beam is significantly reduced, resulting in an improved optical selectivity by suppression of the Doppler shift, in particular when narrow-bandwidth continuous wave lasers are used. The smaller d/ l is chosen, the narrower is the atomic beam distribution, and the better is the optical resolution. The choice of oven dimensions for a given application results as a compromise between optimum collimation and reasonable time for complete evaporation of the sample, which is usually necessary for ultratrace analysis. The efficiency of the atomic beam production depends on the element under investigation and its chemical form. Very often the investigated element is present as an oxide, and the reduction of the oxides requires suitable reducing agents (Eichler et al., 1997). These must
334
N. Erdmann et al.
be added onto the filament or to the sample before inserting it into an oven. For example, the plutonium evaporation is performed in such a way that the plutonium hydroxide is deposited electrolytically on a tantalum backing and subsequently covered with a ∼1 µm thick titanium layer by a sputtering technique. When this “sandwich filament” is heated, plutonium hydroxide is converted into plutonium oxide and evaporated from the backing. During the diffusion process through the titanium layer it is reduced to the atomic state and released in the atomic form into the high vacuum of the RIMS apparatus. There are different metals, such as titanium, zirconium, and thorium, which can be used, and the choice of the reducing agent depends on the solubility and diffusion constant of the analyte in the metal layer as well as on its reduction capability and adsorption enthalpy. The information on the spatial distribution of an analyte in a sample is lost as soon as chemical procedures are applied before RIMS is used for the analysis. For analytical applications involving spatial resolution, e.g., hot particles sampled on filters, dedicated direct techniques like laser desorption or ion sputtering should be considered. In secondary ion mass spectrometry (SIMS) or laser ablation microprobe mass analysis (LAMMA), these techniques offer good spatial resolution—down to micrometer size. Significant amounts of neutral species are present during the desorption and in the sputtering plume (Bogaerts et al., 2003), enabling the combination with a post-ionization step by means of RIMS for enhancement of efficiency and selectivity. The broad distribution of individual atomic states populated in the desorption process has to be considered, because for RIMS applications, the optical excitation process needs to start from a well defined and well populated atomic level for an efficient detection (Bastiaansen et al., 2004; Erdmann et al., 2003; Lievens et al., 1991; Maul et al., 2004, 2005a, 2005b; Vandeweert et al., 1997, 2001). 2.2. Resonant optical excitation The key feature of RIMS is the resonant excitation of atoms in several steps by properly tuned laser light followed by the ionization step. Two to three step excitation/ionization processes with ultraviolet, visible or infrared light are used. The principle of the resonant ionization process of RIMS is shown in Figure 1. Starting from the ground state or a low-lying state, the atoms are step-wise excited to a high-lying state by absorption of one (a) or two (b) resonant photons. If two resonant steps are used, either a real or a virtual intermediate state is involved; the latter usually leads to a reduction of the excitation efficiency. The highly excited atom is finally ionized by absorption of an additional photon, which either non-resonantly raises the electron energy beyond the ionization limit to the continuum (c), or, alternatively, resonantly populates an autoionizing state (d), i.e., a bound state lying energetically above the first ionization potential. This autoionizing state immediately decays through emission of an electron and formation of a residual positive ion. As a third alternative, high-lying Rydberg states can be populated resonantly (e) and subsequently ionized, e.g., by application of an electric field (f), by far infrared photons (g), or by collisions with buffer gas atoms. Due to rather high cross-sections for the photon–atom interactions, an efficiency of the optical excitation and ionization near 100% can be reached with well tuned and spectrally optimized laser light. Total photon fluxes of 1014 –1018 photons, either per second from cwlasers, or within a single pulse of typically about ∼10 ns duration, from pulsed laser systems, are easily achievable. The cross-section for resonant optical excitation from the ground state
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
335
Fig. 1. Principle of two- or three-step resonant excitation and ionization of an atom by laser light. The various excitation paths and ionization mechanisms (a–g) are discussed in the text.
is of the order of λ2 /2π, i.e., ∼10−10 cm2 (∼1014 barn). In many cases, the bottleneck of the resonance ionization (RIS) process is the non-resonant ionizing step into the continuum with typical cross-sections of only 10−17 –10−19 cm2 which limit the efficiency. Via an autoionizing state, the efficiency is increased by two to three orders of magnitude compared to non-resonant ionization. Nearly the same enhancement is obtained, if a Rydberg state is involved as the final step of excitation. Power and spectral bandwidth of the lasers are crucial factors in the attempt to achieve highest possible ionization efficiency. The bandwidth of the lasers relative to the experimental line width of the transition determines the laser power needed for saturation as well as the optical selectivity. The optical selectivity of a laser transition is given (Letokhov, 1987; Letokhov and Chebotayev, 1977) as S = I1 (νL − ν1 )/I2 (νL − ν2 ) − 1, where ν1 and ν2 are the center frequencies of the transitions for two interfering species, e.g., two isotopes of the same element. I1 (νL − ν1 ) and I2 (νL − ν2 ) are the normalized excitation intensities of these transitions as a function of the laser frequency νL . As the laser itself exhibits a spectral profile, the excitation intensities Ii (νL − νi ) are not simply described by the line shape of the optical transition, but are convolutions of the optical resonance profile with the spectral profile of the laser radiation. As can be extracted from this relation, the selectivity
336
N. Erdmann et al.
is the better the larger the difference of the excitation frequencies of the wanted (ν1 ) and the unwanted species (ν2 ) is, and the narrower each of the line shapes and the bandwidth of the lasers are. The most outstanding property of RIMS is its extremely high elemental selectivity, i.e., the almost complete suppression of isobars and neighboring masses stemming from other elements or molecules. This is due to the fact that the density of levels which are accessible via electrical dipole transitions in atoms is of the order of ∼1/eV for low-lying levels, and ∼100/eV for higher ones around principal quantum number n ≈ 20 (1 eV ≈ 2.4 × 1014 Hz), while the typical natural line width is around 7×10−8 eV (≈16 MHz for ∼10 ns lifetime). The line width of a common type, pulsed tuneable laser is below 10−4 eV, and only ∼5 × 10−9 eV for cw-lasers. Thus, the probability of accidentally matching a transition of an unwanted atomic species is absolutely negligible, especially in the case of a two- or three-step excitation. As long as a coherence induced by the laser radiation in the individual excitation steps does not significantly contribute, the overall selectivity of a more-than-one-step resonant excitation may be estimated by simply multiplying the selectivity obtained in each individual step. Contributions from other elements or molecules in the RIMS signal stem mainly from thermal ionization in the atomic beam source, or from non-resonant photo-ionization. Particularly the latter plays a considerable role, if high power ultraviolet light is used in the RIS process. The efficient selection of an individual isotope with RIMS requires more sophisticated steps than the suppression of isobaric interferences. In a number of cases, the abundance sensitivity of the available mass spectrometer does not match the requirements. For example, the sensitive detection of a long-lived radioisotope of an element with omnipresent stable isotopes might easily require isotopic selectivity of the order of 1010 and above, which is inaccessible by conventional mass spectrometry (Heumann et al., 1995). Here, the RIMS approach can take advantage of the natural isotope shift in optical transitions. However, this isotope shift is generally small—of the order of 10−4 to 10−6 of the transition frequency—and differs for each element and electronic transition. In particular, elements in the mass range around 100 amu tend to have small isotope shifts and therefore a high isotopic selectivity is hard to achieve. High optical selectivity requires narrow transitions and a small laser bandwidth. Thus, for applications involving highest isotopic selectivity, cw-lasers stabilized to a bandwidth of <1 MHz must replace the powerful pulsed lasers, which usually have spectral profiles in the order of some GHz. Even though the cw-lasers deliver much lower output power, this is often sufficient for saturation under proper experimental conditions. Here, the arrangement must be optimized in such a way, that the spectral laser line width matches the natural absorption profile of the atoms. This can be realized by a perpendicular overlap of well collimated atomic and laser beams. Since the excitation by cw-lasers avoids any duty cycle losses, the detection efficiency is enhanced for a continuously evaporated analyte. High-resolution multi-step excitation of an atom using cw-lasers creates a coherent coupling of the individual excitation steps. Thus, a simple estimate of excitation efficiency and overall selectivity, using rate equation as in the case of the pulsed laser RIMS, is no longer feasible. A correct treatment by means of the density matrix formalism is indispensable for reliable estimations on multidimensional line profiles, achievable isotopic selectivity, and overall efficiencies (Bushaw et al., 2000). Recently, an optical isotopic selectivity as high as 108 has been realized, resulting
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
337
in an overall isotopic selectivity of ∼1012 , in combination with a detection limit of less than 106 atoms (Müller et al., 2001). 2.3. Laser systems Numerous types of tuneable lasers have been used for resonant excitation in RIMS applications, and only a general overview on typical RIMS laser systems can be given here. For a non-resonant ionization step (i.e., according to scheme (c) or (g) in Figure 1), easy-to-use powerful fixed frequency lasers are favorable. Typical RIMS applications, primarily addressing isobaric suppression with only moderate demands on isotope selectivity, for which the resolution of the mass spectrometer is sufficient, involve pulsed laser systems, like high repetition rate copper vapor pumped dye lasers which can cover the complete range of visible and IR light. Recently, solid state laser systems, e.g., Nd:YAG pumped titanium:sapphire lasers, came on the market. These lasers cover a limited spectral bandwidth of ∼700 to ∼1000 nm, and therefore very often efficient frequency conversion processes in nonlinear optical media must be used to generate blue to ultraviolet light. The pulsed laser systems can easily be combined with a time-of-flight mass spectrometer, where the short laser pulse delivers the start signal and the incoming ions the stop signal. For a good overall efficiency, the temporal and spatial overlap of the continuously evaporated atoms from the sample with the laser pulse structure must be considered; i.e., high repetition rate pulsed lasers, which deliver about 5000–25000 pulses per second, e.g., Cu-vapor, specialized Nd:YAG or Nd:YLF lasers, are required (Grüning et al., 2004; Passler et al., 1997). Standard low repetition rate pulsed lasers in the 10 to 100 Hz regime (e.g., flash-lamp pumped Nd:YAG, or excimer laser) can be used efficiently, when the sample is volatilized by a synchronized pulsed laser desorption or ablation, respectively (Krönert et al., 1991; Sauvage et al., 2000) or when the atoms are slowed down and stored in a gas cell for a sufficiently long interaction time (Kudryavtsev et al., 2001, 2002; Moore et al., 2005). Tuneable cw-lasers for the high resolution mode use liquid dyes or solid state active media (e.g., titanium:sapphire). Also diode lasers, involving extended cavities for tuning (ECDLs), are in use. Typically, the accuracy of the laser stabilization is of the order of some 100 kHz, combined with a long term stability of a few MHz or even less. 2.4. Mass analyzers and ion detection Mass selective ion detection is an inherent part of the RIMS procedure. Almost the whole spectrum of available mass spectrometer types has been applied in RIMS. In general, mass dispersive systems can be classified into two categories: either operating continuously or pulsed. Continuous analyzers include radiofrequency quadrupole (RFQ) and magnetic sector field analyzers. Both of these filter types are working on a specific, pre-selected mass-to-charge ratio m/Z, which they filter either spatially or by spatial resonance oscillations, before the ions reach the detector. These systems are best combined with continuous optical excitation/ionization. Mass spectra are obtained by scanning the transmitted m/Z range over the region of interest. The most common type of such a continuous analyzer is the RFQ mass filter, which is compact in size, has a high transmission efficiency, fast scan rate and can be optimized to abundance sensitivities as low as 10−9 . Magnetic sector field selectors may have
338
N. Erdmann et al.
higher mass resolution, but suffer from slow speed and hysteresis effects during scanning as well as from strong mass peak tailing, resulting in lower abundance sensitivities typically not better than 10−6 . For highest mass resolution, magnetic sectors are combined in double focusing systems with additional electric sector field deflectors in order to reduce the kinetic energy spread stemming from the ion source. Pulsed mass spectrometers include time-of-flight mass spectrometers (TOF-MS) as well as quadrupole ion traps and ion cyclotron resonance (ICR) in Penning traps. For RIMS with pulsed lasers, only TOF-MS is practically in use, combining a rather simple experimental construction with high transmission. A full mass spectrum for each laser pulse can be obtained here by analyzing the detector signal as a function of arrival time of the ions produced by the resonant pulse laser excitation. Quantitative ion detection is rather simple and efficiently carried out with single channel secondary electron multipliers (channeltrons), channelplates or discrete multi-stage secondary electron multipliers (SEV’s).
3. Applications of RIMS 3.1. High sensitivity, moderate resolution pulsed laser RIMS RIMS with pulsed laser excitation/ionization and TOF-MS is nowadays routinely applied for the determination of smallest amounts of actinides, mainly plutonium, and technetium in environmental and biological samples (Grüning et al., 2004; Passler et al., 1997; Trautmann et al., 2004). Typical isotopic ratios for plutonium of different origin published in the literature are presented in Table 1. One can see that the isotopic fingerprint for plutonium of different origin is quite distinct. For the determination of plutonium by RIMS in different samples, several steps are necessary: As a first step, plutonium is chemically separated from the sample. Prior to this, Table 1 Isotopic composition of plutonium of different origin Plutonium source
238 Pu
239 Pu
240 Pu
241 Pu
242 Pu
Fallout (Hanson, 1980) Weapons (Nunnemann et al., 1998) Isotope battery (Hanson, 1980) LWR1 (Benedict, 1981)a LWR2 (Benedict, 1981)a Chernobyl (Nunnemann et al., 1998)
–
83.5%
15.0%
1.2%
0.3%
–
97%
3%
–
–
80%
17%
3%
–
–
2.4%
58.4%
24.0%
11.4%
3.9%
4.3%
37.9%
27.7%
18.3%
11.2%
0.3%
66.2%
26.1%
5.5%
1.8%
a LWR light water reactor spent fuel with differing fuel composition.
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
339
known amounts of 236 Pu (optional) and 244 Pu are added as tracers to determine the chemical yield (α-spectroscopy of 236 Pu) and the overall efficiency via the count rate of 244 Pu in the RIMS apparatus. The chemical separation procedure varies according to the sample matrix. In all cases it includes a co-precipitation step of plutonium with Fe(OH)3 and an extraction chromatography with a TEVA Spec SPS resin (Aliquat 336 N). Afterwards, the plutonium fraction is electrolytically deposited as plutonium hydroxide on a tantalum backing and covered with a thin titanium layer (∼1 µm) by sputtering. In the RIMS apparatus, this sandwich filament is heated to about 1000 ◦ C under high vacuum conditions. The atoms are ionized near the filament surface by a three-step, three-color laser excitation. High repetition rate lasers are used to reach a good temporal overlap between the continuously evaporating atomic beam and the pulsed laser beams. A commercially available, Q-switched and intracavity-doubled (532 nm) Nd:YAG pump laser (Clark-MXR ORC-1000) with a repetition rate of 1–25 kHz, a power of up to 40 W at 7 kHz and a pulse length of ∼300 ns is applied to pump simultaneously three titanium:sapphire (Ti:Sa) lasers. The three lasers are synchronized with intracavity Pockels cells used as Q-switches. The Ti:Sa laser light can be frequency doubled in an external, single pass set-up using a BBO-crystal. For elements like plutonium with a first ionization potential of ∼6 eV one frequency doubled step and two red photons are sufficient for ionization. The laser light is transported through an optical fiber and focused into the interaction region with the atomic beam in the mass spectrometer. The photoions are mass separated in a reflectron TOF-MS and counted with a multi-channel plate detector. The mass resolution of the TOF-MS is R ≈ 600. Figure 2 shows the experimental setup with the Nd:YAG pump laser, three titanium:sapphire lasers, a frequency doubling unit, and the TOF apparatus. An efficient three step ionization scheme for plutonium has been found with λ1 = 420.76 nm, λ2 = 847.28 nm, and λ3 = 767.53 nm, the latter populating a high lying Rydberg state which is ionized by an electric field. The isotope shifts in this excitation scheme have been measured for the plutonium isotopes 238, 239, 240, 241, 242, and 244, and are taken into account for exact isotope ratio measurements. With this arrangement, an efficiency of ∼3 × 10−5 has been reached for a single isotope, leading to a detection limit of 1 × 106 atoms (0.4 fg) with a 3σ -confidence level, which is more than two orders of magnitude better than the one for conventional α-spectroscopy of 239 Pu (T1/2 = 2.41 × 104 a). The correctness of the isotope ratio measurements for plutonium with the RIMS technique was determined using the NIST standard SRM 996. The measured ratios obtained with RIMS and normalized to 244 Pu are in good agreement with the certified ratios (Grüning et al., 2004). Examples for isotope selective determination of plutonium in environmental samples of different origin are given in Figure 3. The tracer isotope 244 Pu is present in all spectra. Figure 3(a) shows the RIMS-TOF spectrum of sediment from the French nuclear weapons test site on the Mururoa Atoll (reference material IAEA-368). The huge excess of 239 Pu compared to 240 Pu is typical for nuclear weapons plutonium. Here, an overall (239 Pu + 240 Pu) radioactivity of 31 mBq/g was detected in good agreement with the certified value. In Figure 3(b), the spectrum of a dust sample from the northern part of Germany is displayed. The typical isotope ratio of fallout plutonium (240 Pu/239 Pu ∼ = 0.16), can be seen. Figure 3(c) shows a mass spectrum of a soil sample contaminated with plutonium from the Chernobyl accident. The isotope pattern is characteristic for reactor grade plutonium. Pulsed laser RIMS has also been used for the determination of the first ionization potentials of the actinides up to einsteinium (Erdmann et al., 1998; Waldek et al., 2001; Worden et al.,
340
N. Erdmann et al.
Fig. 2. Schematic view of a setup for RIMS with a pulsed laser system—three tuneable titanium:sapphire lasers pumped by a high repetition rate Nd:YAG laser—and a reflectron time-of-flight mass spectrometer.
2006). With this technique, the first ionization potentials of Ac, Am, Cm, Bk, Cf, and Es were measured for the first time. 3.2. Pulsed laser RIMS in a laser ion source The temporal overlap of the continuously evaporated atomic beam with pulsed laser beams can be improved by means of a “laser ion source” where the sample atoms are confined inside a heated, cylindrical chamber and therein resonantly ionized by the laser beams, which enter
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
341
Fig. 3. RIMS spectra of plutonium from environmental samples of different origin: (a) Sediment from the Mururoa Atoll, (b) dust sample from northern Germany, (c) soil sample contaminated by the Chernobyl accident.
through a small hole. The ions are extracted by an electric field through the same or a second hole while the neutral atoms are undergoing numerous wall collisions and are thus stored inside the chamber over a period of several laser shots. A schematic view of such a laser ion source, with the laser beams entering from one side, the ions being extracted through a second hole, is shown in Figure 4. In this way, a ionization efficiency of the order of a few percent can be realized. Such a laser ion source is well suited for combination with a magnetic mass spectrometer and has been used, e.g., for ultratrace analysis of technetium (Passler et al., 1997). Very recently, an effective excitation/ionization scheme for Tc using three Ti:Sa laser beams has been investi-
342
N. Erdmann et al.
Fig. 4. Principle of a laser ion source. The atoms are confined inside a heated chamber, where they can undergo several wall collisions, and resonantly ionized by the laser beams. Ions are extracted by the electric field between acceleration voltage, Uacc , and extraction electrode.
gated (Wies, 2006). Due to the high ionization potential of Tc (7.28 eV), an excitation ladder of λ1 = 429.83 nm, λ2 = 395.15 nm, and λ3 = 841.72 nm, populating an autoionizing state is used, which implies two frequency doubled steps. The experiments have been performed with 99 Tc (T1/2 = 2.1 × 105 a). Laser ion sources have also found applications at on-line radioisotope production facilities for the investigation of man-made short-lived exotic nuclei (Alkhazov et al., 1991; Barzakh et al., 1997; Köster, 2002). Here, the suppression of a strong surplus of isobars is of highest importance which can be further improved by combination with ion trapping and bunching techniques (Wies et al., 2005). 3.3. High resolution RIMS for ultra high isotopic selectivity A special challenge in elemental trace analysis is the determination of natural or anthropogenic ultratrace isotopes with relative abundances well below 10−9 with respect to a neighboring isotope of the same element. As shown in Figure 5, there are numerous long lived radioisotopes existing in the natural environment with abundances of 10−9 down to 10−18 and even beyond. Possible sources for these isotopes range from anthropogenic radioactive contaminations—which are due to global fallout from nuclear explosions, controlled and accidental releases from processes within the nuclear fuel cycle or the medical use of radioisotopes—to the omnipresent natural radioisotopes, which are formed primordial, cosmogenic, or by other sources. Most of these isotopes may serve as sensitive probes for studies in geological, environmental, cosmo-chemical or bio-medical research (Kutschera, 1990; Lu and Wendt, 2003). High resolution RIMS (HR-RIMS) using frequency stabilized narrowbandwidth cw-lasers has been explored during the last decades as one method to get access to the range of isotopic abundances below 10−9 (Wendt et al., 2000). Alternatively, novel quantum optical approaches using atom manipulation and trapping by laser light are also presently under development for such applications (Chen et al., 1999).
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
343
Fig. 5. Abundance of long-lived radionuclides. The individual origin is indicated by the different symbols.
3.3.1. Noble gas determination using pre-enrichment and pulsed laser RIMS The interest in noble gas ultratrace determination ranges from radiodating of groundwater, oceanic tracer studies up to surveillance of releases from nuclear reactors. Major ultratrace isotopes of interest are 3 He, 39 Ar and 81,85 Kr. The study of the isotopes 85 Kr and 81 Kr has been the pioneering work in RIMS application to ultra rare isotopes in its early stages, leading to the lowest detection limits and highest selectivity reported ever up to now; about 500 atoms in samples of altogether 1023 atoms and with suppression of 15 orders of magnitude of other Kr-isotopes (Payne et al., 1994). The apparatus used for this investigation consisted of two conventional mass spectrometric pre-enrichment steps, applying successively a velocity filter and a quadrupole mass spectrometer, yielding a suppression of 9 orders of magnitude of the interfering isotopes. Finally a pulsed laser, three-step RIMS at an RFQ filter installed in a closed vacuum system was used to obtain the outstanding specifications with regard to selectivity and sensitivity. Apart from the large number of experimental steps, a major disadvantage of this method was the inconvenient transition of the first optical excitation from the atomic ground state at 116 nm, which can only be produced by an extensive nonlinear frequency mixing process of strong pulsed laser radiation. This step can be overcome in fast atomic beam laser RIMS, where the excitation can start from an elevated metastable atomic state, which is populated in a charge exchange process applied for the neutralization of the accelerated positive ions. From there, resonant excitation/ionization with visible light is possible. This collinear RIMS approach was first developed for the determination of the trace isotope 3 He in environmental samples (Kudryavtsev and Letokhov, 1982; Kudryavtsev et al., 1988), and soon after applied for the study of the
344
N. Erdmann et al.
isotope shift and hyperfine structure of short-lived Yb-isotopes (Schulz et al., 1991). The competing technique of photon burst detection can reach an extremely high isotopic selectivity, but requires efficient photon collection from the beam of fast moving atoms (LaBelle et al., 1996). 3.3.2. Determination of radiostrontium by collinear RIMS In cases where naturally occurring isotope shifts are too small to reach the required selectivity, collinear RIMS can be applied, making use of the strong mass dependent Doppler shift of optical resonance transitions. With kinetic energies of the atoms in the range of several 10 keV, large artificial isotope shifts (many GHz) are introduced by the Doppler effect as a consequence of the different masses of the isotopes. The conservation of the longitudinal energy spread during acceleration additionally leads to a reduction in the velocity spread dependent Doppler broadening of the excitation width. The determination of the release of the radiotoxic isotopes 89,90 Sr into the environment requires an isotopic selectivity of ∼1010 for the suppression of the dominant stable isotope 88 Sr. Collinear resonance ionization on a fast atomic beam was successfully used for this task. The strontium sample is placed inside a conventional ion source and undergoes surface ionization. Then, acceleration is applied to form an ion beam of typically 10 to 30 keV energy. After mass separation in a sector field magnet, the ion beam is neutralized in a charge exchange process via collisions with cesium vapor (102 mbar). The neutral Sr atoms are predominantly formed in metastable states, which can be excited in a single step into a high lying Rydberg state for field ionization in a strong electric field. Apart from the high artificial isotope shift and the possibility to start optical excitation from elevated metastable states, collinear laser spectroscopy offers the advantage, that powerful fixed frequency lasers can be used. Frequency tuning into resonance for the optical excitation can be performed through the Doppler effect by adjusting the ion acceleration voltage properly. An isotopic selectivity of ∼1011 and an efficiency of 10−5 for the determination of 90 Sr were achieved, corresponding to a detection limit of 2 × 106 atoms of 90 Sr (Wendt et al., 1997). For comparison, AMS specifications of a slightly better selectivity of 3 × 1013 but a somewhat worse detection limit of 5×107 atoms were reported for 90 Sr determination (Paul et al., 1997). 3.3.3. Multi-step RIMS on thermal atomic beams—a universal approach The technique of multi-step narrow-band cw-laser RIMS on a thermal atomic beam combined with quadrupole mass spectrometry has provided excellent results on a variety of elements. This method has been pioneered by Bushaw with the primary goal of ultra selective trace analysis of lead (Bushaw, 1989) and 41 Ca, for which in both cases highest isotopic selectivities above 1010 are mandatory. To meet this requirement, up to three coherent narrow bandwidth excitation steps and a final ionization step are used. In combination with the abundance sensitivity of an optimized quadrupole mass spectrometer, an extremely high overall selectivity and a good overall efficiency of about 10−6 were realized. A major advantage of this method is its limited experimental expenditure. A schematic diagram of such an apparatus is shown in Figure 6. It consists of a commercial mass spectrometer, up to three tuneable cwlaser sources and, additionally, a fixed frequency ionization laser in the visible or far infrared spectral range. Compact and reliable extended cavity diode lasers (ECDL) can often be used, as long as the required wavelengths are accessible. They are spectrally controlled by a computerized frequency stabilization system with a precision of about 1 MHz. For cases where
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
345
Fig. 6. Schematic view of a HR-RIMS apparatus using three tuneable cw-lasers and one ionization laser. The laser beams interact in perpendicular geometry with atoms released from a collimated beam source. The ions are mass separated by a quadrupole mass spectrometer.
a wide spectral tuning range is required, commercial cw Ti:Sa or dye lasers can be added. However, this increases the complexity of the RIMS system. As an example, the determination of 41 Ca is described. After chemical pre-separation, the calcium sample is introduced as nitrate into a collimated atomic beam source made of pure pyrolyzed graphite. A typical sample volume is 10 µl, containing a predetermined amount of a few micrograms of stable Ca with the main isotope 40 Ca. By adding titanium as reducing agent, a rather steady and well collimated atomic beam is generated, which can be overlapped by the four laser beams in a perpendicular geometry behind the exit of the oven. Counter propagation of the three laser beams for resonant excitation strongly suppresses the residual Doppler broadening, which results from the opening angle of the atomic beam, and leads to a highest optical selectivity. The laser–atom–beam interaction region is located within the interior of a standard cross beam ionizer source of a commercial quadrupole mass spectrometer. Due to the low energy spread of the created laser ions, extremely high suppression of neighboring masses is achieved in the mass spectrometer with values of about 108 (Blaum et al., 1998). For the precise determination of 41 Ca in an analytical sample, an isotope ratio measurement to a reference isotope (43 Ca) with a predetermined known number of atoms is carried out involving background correction. During this procedure, the tuneable lasers as well as the mass spectrometer are repetitively set to the position of the isotope of interest, the reference isotope and a background measurement. For addressing 41 C analysis two or three step excitation schemes with the optical transitions 4s 2 1 S0 → 4s4p 1 P1 → 4s4d 1 D2 → 4s15f 1 F , with λ = 422.8 nm, λ = 732.8 nm, λ = 868.5 nm were used. Two step excitation 3 1 2 3
346
N. Erdmann et al.
Fig. 7. (Left) Two- and three-step excitation schemes for 41 Ca determination. (Right) Analytical results for 41 Ca determination in nuclear reactor concrete. Sample 1: reactor wall; sample 2: reactor floor; sample 3: blank. Two-step excitation: data points (open diamonds) and 3σ detection limit (dotted line); three-step excitation: data points (open circles) and 3σ detection limit (dashed line); calibration sample: data points (black closed circles) and absolute values (short lines).
(2 + 1 ) implies ionization already from the 4s4d 1 D2 level with visible light from an Arion laser. This reduces the experimental expenditure but results in limited optical selectivity. Highest specifications are obtained with three step excitation and final non-resonant ionization (3 + 1 ) with 10.6 µm laser light from a powerful CO2 laser. An intercomparison of the results from these two approaches is given in Figure 7. The 41 Ca inventory in the concrete of the biological shield of a nuclear research reactor was determined, alternatively using both excitation schemes. The first, more simple approach (2 + 1 ) gives a 3σ detection limit of 5 × 10−10 in the isotopic abundance (indicated by the dotted line) and leads only to upper limits of the 41 Ca inventory (data points: open diamonds). The extension of the technique to three resonant excitation steps enhances the sensitivity down to a 3σ detection limit of 6 × 10−11 for these samples (dashed line) and yields precise results for the 41 Ca content in the concrete from the reactor wall (sample 1), the reactor floor (sample 2) together with a value for the blank (sample 3), as given by the data points (open circles). For calibration of the technique, aliquots of all samples have additionally been spiked with a known amount of 41 Ca in the 10−8 range (short lines), which were perfectly reproduced by the measurements (data points: black closed circles) (Müller et al., 2001). Additional application of this technique to 41 Ca determination in biomedical samples has shown its high versatility (Denk et al., 2006). For elements with a complex spectrum like lanthanides or actinides, optical three step resonant excitation leading directly into a narrow autoionizing level of the atom just above the ionization threshold can be applied. Such a HR-RIMS scheme is presently under development
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
347
Fig. 8. Autoionizing spectra of uranium using a three step excitation through second excited states with different total angular momentum J .
for the selective determination of 236 U, which might serve as an indicator for the determination of anthropogenic uranium in the environment with isotopic abundances in the range of 10−8 down to 10−11 . Even though uranium is one of the best studied elements with respect to its atomic structure and spectroscopy, high resolution spectroscopic data are very scarce and must be investigated in detail for the application of the HR-RIMS technique (Schumann et al., 2005). Figure 8 shows a section of the autoionizing spectrum of uranium which was accessed by a three step resonant excitation from the atomic ground state. The three spectra involve second excited levels with different total angular momentum J = 6, 7, or 8. Especially, the very narrow peaks with resonance enhancements of up to 106 in the ion count rate, observed around 49,956 or 49,972 cm−1 in the J = 8 spectrum, are promising candidates for highest efficiency and selectivity in the laser based ultra trace determination of 236 U. 3.4. Future prospects of RIMS In recent years, novel RIMS instruments have been developed, with the goal to further improve the detection limit and to gain information about the lateral distribution of the investigated nuclides within the sample. With an ion guide laser ion source (Backe et al., 1997; Sewtz et al., 2003), a better detection limit might be reached. The atoms are stored in a buffer gas cell at an
348
N. Erdmann et al.
argon pressure of ∼35 mbar for 40 ms within the interaction volume illuminated by the laser light. A repetition rate of 200 Hz as delivered by an excimer laser-pumped dye laser system is sufficient for an efficient resonant ionization. The ions are guided to a nozzle by electric fields, flushed out of the cell by a gas jet, separated from the buffer gas by skimmers, mass separated in a quadrupole mass spectrometer, and finally counted by a channeltron detector. The overall efficiency obtained so far is about 10−4 . Often radiotoxic material is released in form of or attached to small particles with typical dimensions between several tens of nanometers up to several micrometers as “hot particles”. These particles have attracted high interest for risk assessment of contaminated areas as well as for nuclear forensic analyses (Betti et al., 1999; Bunzl, 1997; Pöllänen et al., 1999; Salbu, 2000; Salbu et al., 1998, 2004). They are analyzed as part of newly implemented safeguards programs by international governmental organizations for the control of nuclear material, such as the International Atomic Energy Agency (IAEA) and the Euratom Safeguards Office (ESO), in order to strengthen the control compliance of nuclear facilities declarations and detect undeclared nuclear activities (Donohue, 1998, 2002; Donohue and Zeisler, 1993; Tamborini and Betti, 2000). The transport mechanism of particle bound radioactivity in the environment is correlated to the mobility of the particle. It is of great importance to characterize single particles in order to detect particle-to-particle variations. Of special concern for a risk assessment is the characterization of micro-particles containing high concentrations of actinides. In recent years, secondary ion mass spectrometry, SIMS, or aerosol mass spectrometry using laser desorption, have been applied for particle analysis (Betti et al., 1999; Tamborini and Betti, 2000; Tamborini et al., 2002; Tamborini, 2004). In the case of actinide mixtures, the interpretation of the results is hampered by isobaric interferences (238 U/238 Pu, 241 Pu/241 Am). These can be excluded by coupling resonance ionization mass spectrometry (RIMS) with atomization techniques that allow high lateral resolution, such as laser desorption or ion beam sputtering. By combining a commercially available MALDI-TOF mass spectrometer with a pulsed laser system for post-ionization of the desorbed neutrals from surfaces or grains, a lateral resolution of ≈30 µm (Maul et al., 2004) is obtained. For laser desorption, a nitrogen laser (λ = 337 nm, pulse length ≈3–4 ns, pulse energy variable up to 300 µJ) has been applied. Ions resulting from the desorption process are suppressed by means of a pulsed repelling potential, synchronized to the laser pulse. Resonant ionization is performed with Nd:YAG laser pumped dye lasers. First experiments have been carried out using gadolinium as a model element. A mass resolution of m/m > 2200 was obtained and an overall detection efficiency of 1.5 × 10−4 could be reached with a double-resonant excitation scheme with subsequent infrared ionization yielding a resonance enhancement of two orders of magnitude (Maul et al., 2005b). Furthermore, the coupling of resonant post-ionization of neutrals produced by ion sputtering was investigated, as this method allows even higher lateral resolution, down to the submicrometer range, and hence allows the study of only one particle at a time in the investigated volume. To evaluate such a set-up, experiments were performed on uranium as a proof-ofprinciple (Erdmann et al., 2003). Secondary neutrals mass spectrometry (SNMS), using a Ga ion source with sub-micrometer spot-size and non-resonant laser ionization was applied to uranium particles and it could be shown that non-resonant laser post-ionization enhances the sensitivity for the detection of uranium by two orders of magnitude compared to SIMS.
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
349
Fig. 9. Schematic view of the experimental arrangement for the coupling of ion beam sputtering by a 69 Ga+ ion gun and laser post-ionization of the sputtered neutrals. The ions are accelerated into a TOF-MS by a pulsed extraction voltage, Uextr .
Taking these results into account, a new instrument is currently being set up, using a commercial TOF-SIMS instrument (Ion-Tof, Münster, Germany) with a pulsed gallium liquid metal ion gun that allows a sub-micrometer focus. The ion beam can be rastered over the sample surface (up to 500 × 500 µm2 ) and spatially resolved images can be obtained for selected masses (Berg et al., 2006). This apparatus will be coupled to a high repetition rate Nd:YAG pumped Ti:sapphire laser system for 3-color, 3-step resonant ionization of the sputtered neutrals yielding high elemental selectivity. The high laser repetition rate enables an optimum coupling to the pulse repetition rate of the ion gun. To synchronize the pulsed ion beam with the laser system, the TOF-SIMS has been modified. After the primary 69 Ga ion pulse, a pulsed repelling voltage suppresses the released secondary ions, before the pulsed laser beams are focused, parallel to the sample surface, into the plume of secondary neutral particles, as shown in Figure 9. The laser-ions are extracted into the reflectron TOF-MS by means of a pulsed voltage. The sample preparation procedure has to be optimized to obtain a high emission of actinide atoms during ion beam sputtering (Erdmann et al., 2006). With this arrangement, it is planned to determine traces of actinides in micro-particles without isobaric interferences.
4. Conclusion and outlook RIMS is well suited for trace analysis of long-lived radioactive isotopes due to its low detection limit in the fg range, its extreme element and isotope selectivity and the short measuring time of about one hour. In the last years, reliable solid state and diode laser systems have become available for applications of RIMS in routine analysis. In the near future, RIMS will mainly be applied for more specific problems and not as a standard method for trace analysis
350
N. Erdmann et al.
of almost all elements. Using different experimental arrangements, the capability of RIMS has been illustrated by the measurements of isotope ratios and abundances of the plutonium isotopes, 238 Pu–244 Pu, 90 Sr, and 41 Ca in various samples yielding a detection limit of ∼106 atoms and isotope selectivities up to 5 × 1012 . Further developments will concentrate on the improvement of the detection limit, e.g., by the storage of atoms in a buffer gas cell with subsequent resonant laser ionization, and on particle analysis without isobaric interferences, especially in the actinide region. This can be realized by coupling high lateral resolution atomization techniques like laser desorption or ion beam sputtering with RIMS.
References Alkhazov, G.D., Batist, L.K., Bykov, A.A., Vitman, V.D., Letokhov, V.S., Mishin, V.I., Panteleyev, V.N., Sekatsky, S.K., Fedoseyev, V.N. (1991). Application of a high efficiency selective laser ion-source at the IRIS facility. Nucl. Instrum. Methods Phys. Res. Sect. A 306, 400–402. Ambartzumian, R.V., Letokhov, V.S. (1972). Selective two-step (STS) photoionization of atoms and photodissociation of molecules by laser radiation. Appl. Opt. 11, 354–358. Backe, H., Eberhardt, K., Feldmann, R., Hies, M., Kunz, H., Lauth, W., Martin, R., Schöpe, H., Schwamb, P., Sewtz, M., Thörle, P., Trautmann, N., Zauner, S. (1997). A compact apparatus for mass selective resonance ionization spectroscopy in a buffer gas cell. Nucl. Instrum. Methods Phys. Res. Sect. B 126, 406–410. Barzakh, A.E., Denisov, V.P., Fedorov, D.V., Orlov, S.Y., Seliverstov, M.D. (1997). A mass-separator laser ion source. Nucl. Instrum. Methods Phys. Res. Sect. B 126, 85–87. Bastiaansen, J., Philipsen, V., Lievens, P., Silverans, R.E., Vandeweert, E. (2004). Influence of the atomic structure on the quantum state of sputtered Ir atoms. Phys. Rev. A 70, 052902-(1–6). Becker, J.S. (2003). Mass spectrometry of long-lived radionuclides. Spectrochim. Acta B 58, 1757–1784. Becker, J.S., Dietze, H.-J. (1999a). Precise isotope ratio measurements for uranium, thorium and plutonium by quadrupole-based inductively coupled plasma mass spectrometry. Fresenius J. Anal. Chem. 364, 482–488. Becker, J.S., Dietze, H.-J. (1999b). Application of double-focusing sector field ICP mass spectrometry with shielded torch using different nebulisers for ultratrace and precise isotope analysis of long-lived radionuclides. J. Anal. At. Spectrom. 14, 1493–1500. Becker, J.S., Dietze, H.-J. (2000). Mass spectrometry of long-lived radionuclides. In: Meyer, R.A. (Ed.), Encyclopedia of Analytical Chemistry: Applications, Theory and Instrumentation. John Wiley and Sons, Chichester, NY, pp. 12947–12961. Benedict, M. (1981). Nuclear Chemical Engineering. McGraw–Hill, New York. Berg, T., Maul, J., Erdmann, N., Bernhard, P., Schuppler, S., Nagel, P., Sudek, C., Ott, U., Schönhense, G. (2006). Coupling of imaging NEXAFS with secondary ion mass spectrometry for the chemical and isotopic analysis of presolar cosmic grains. Anal. Bioanal. Chem. 386, 119–124. Betti, M., Tamborini, G., Koch, L. (1999). Use of secondary ion mass spectrometry in nuclear forensic analysis for the characterization of plutonium and highly enriched uranium particles. Anal. Chem. 71, 2616–2622. Blaum, K., Geppert, C., Muller, P., Noertershaeuser, W., Otten, E.W., Schmitt, A., Trautmann, N., Wendt, K., Bushaw, B.A. (1998). Properties and performance of a quadrupole mass filter used for resonance ionization mass spectrometry. Int. J. Mass Spectrom. 181, 67–87. Bogaerts, A., Chen, Z., Gijbels, R., Vertes, A. (2003). Laser ablation for analytical sampling: what can we learn from modeling? Spectrochim. Acta B 58, 1867–1893. Boulyga, S.F., Testa, C., Desideri, D., Becker, J.S. (2001). Optimisation and application of ICP-MS and alphaspectrometry for determination of isotopic ratios of depleted uranium and plutonium in samples collected in Kosovo. J. Anal. At. Spectrom. 16, 1283–1289. Boulyga, S.F., Desideri, D., Meli, M.A., Testa, C., Becker, J.S. (2003). Plutonium and americium determination in mosses by laser ablation ICP-MS combined with isotope dilution technique. Int. J. Mass Spectrom. 226, 329–339. Bushaw, B.A. (1989). High-resolution laser-induced ionization spectroscopy. Prog. Anal. Spectrom. 12, 247–276. Bushaw, B.A., Nörtershäuser, W., Wendt, K. (2000). Lineshapes and optical selectivity in high-resolution doubleresonance ionization mass spectrometry. Spectrochim. Acta B 54, 321–332.
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
351
Bunzl, K. (1997). Probability for detecting hot particles in environmental samples by sample splitting. Analyst 122, 653–656. Chen, C.Y., Li, Y.M., Bailey, K., O’Connor, T.P., Young, L., Lu, Z.-T. (1999). Ultrasensitive isotope trace analyses with a magneto-optical trap. Science 286, 1139–1141. Dai, M.H., Buesseler, K.O., Kelley, J.M., Andrews, J.E., Pike, S., Wacker, J.F. (2001). Size-fractionated plutonium isotopes in a coastal environment. J. Environ. Radioact. 53, 9–25. Denk, E., Hillegonds, D., Vogel, J., Synal, A., Geppert, C., Wendt, K., Fattinger, K., Hennessy, C., Berglund, M., Hurrell, R.F., Walczyk, T. (2006). Labelling the human skeleton with 41 Ca to assess changes in bone calcium metabolism. Anal. Bioanal. Chem. 386, 1587–1602. Donohue, D.L. (1998). Strengthening IAEA safeguards through environmental sampling and analysis. J. Alloys Compd. 271–273, 11–18. Donohue, D.L. (2002). Strengthened nuclear safeguards. Anal. Chem. 74, 28A–35A. Donohue, D.L., Zeisler, R. (1993). Analytical chemistry in the aftermath of the Gulf War. Anal. Chem. 65, 359A– 368A. Eichler, B., Hübener, S., Erdmann, N., Eberhardt, K., Funk, H., Herrmann, G., Köhler, S., Trautmann, N., Passler, G., Urban, F.-J. (1997). An atomic beam source for actinide elements: Concept and realization. Radiochim. Acta 79, 221–233. Erdmann, N., Nunnemann, M., Eberhardt, K., Herrmann, G., Huber, G., Köhler, S., Kratz, J.V., Passler, G., Peterson, J.R., Trautmann, N., Waldek, A. (1998). Determination of the first ionization potential of nine actinide elements by resonance ionization mass spectroscopy (RIMS). J. Alloys Compd. 271–273, 837–840. Erdmann, N., Betti, M., Kollmer, F., Benninghoven, A., Grüning, C., Philipsen, V., Lievens, P., Silverans, R.E., Vandeweert, E. (2003). Resonance and nonresonant laser ionization of sputtered uranium atoms from thin films and single microparticles: Evaluation of a combined system for particle trace analysis. Anal. Chem. 75, 3175– 3181. Erdmann, N., Kratz, J.V., Trautmann, N., Passler, G., Betti, M., Lievens, P., Silverans, R.E., Vandeweert, E. (2006). Resonance ionization mass spectrometry for element- and isotope-selective trace analysis of actinides in environmental micro-particles. In: Alvarez, R., Bryan, N.D., May, I. (Eds.), Recent Advances in Actinide Science, ACTINIDES 2005 Conference Proceedings. RSC Publishing. Special Publication No. 305, ISBN: 978-0-85404678-2, pp. 122–124. Fifield, L.K., Cresswell, R.G., Di Tada, M.L., Ophel, T.R., Day, J.P., Clacher, A.P., King, S.J., Priest, N.D. (1996). Accelerator mass spectrometry of plutonium isotopes. Nucl. Instrum. Methods Phys. Res. Sect. B 117, 295–303. Fifield, L.K., Synal, H.-A., Suter, M. (2004). Accelerator mass spectrometry of plutonium at 300 kV. Nucl. Instrum. Methods Phys. Res. Sect. B 223–224, 802–806. Grüning, C., Huber, G., Klopp, P., Kratz, J.V., Kunz, P., Passler, G., Trautmann, N., Waldek, A., Wendt, K. (2004). Resonance ionization mass spectrometry for ultratrace analysis of plutonium with a new solid state laser system. Int. J. Mass Spectrom. 235, 171–178. Hanson, W.C. (1980). Transuranic Elements in the Environment. DOE/TIC-22800, Technical Information Center/U.S. Department of Energy, Washington, DC. Heumann, K.G., Eisenhut, S., Gallus, S., Hebeda, E.H., Nusko, R., Vengosh, A., Walczyk, T. (1995). Recent developments in thermal ionization mass-spectrometric techniques for isotope analysis—a review. Analyst 120, 1291–1299. Huber, G., Passler, G., Wendt, K., Kratz, J.V., Trautmann, N. (2003). In: L’Annunziata, M.F. (Ed.), Handbook of Radioactivity Analysis, second ed. Academic Press, San Diego, pp. 799–843. Hurst, G.S., Payne, M.G. (1988). Principles and Applications of Resonance Ionization Spectroscopy. Hilger Publications, Bristol. Ketterer, M.E., Hafer, K.M., Link, C.L., Kolwaite, D., Wilson, J., Mietelski, J.W. (2004). Resolving global versus local/regional Pu sources in the environment using sector ICP-MS. J. Anal. At. Spectrom. 19, 241–245. Köster, U. (2002). Resonance ionization laser ion sources. Nucl. Phys. A 701, 441C–451C. Krönert, U., Becker, St., Bollen, G., Gerber, M., Hilberath, Th., Kluge, H.-J., Passler, G., and the ISOLDE Collaboration (1991). On-line laser spectroscopy by resonance ionization of laser-desorbed, refractory elements. Nucl. Instrum. Methods Phys. Res. Sect. A 300, 522–537. Kudryavtsev, Y.A., Letokhov, V.S. (1982). Laser method of highly selective detection of rare radioactive isotopes through multistep photoionization of accelerated atoms. Appl. Phys. B 29, 219–221.
352
N. Erdmann et al.
Kudryavtsev, Y.A., Letokhov, V.S., Petrunin, V.V. (1988). Collinear laser photo-ionization of helium isotopes in a fast atomic beam. Opt. Commun. 68 (1), 25–28. Kudryavtsev, Y., Bruyneel, B., Huyse, M., Gentens, J., Van den Bergh, P., Van Duppen, P., Vermeeren, L. (2001). A gas cell for thermalizing, storing and transporting radioactive ions and atoms. Part I: Off-line studies with a laser ion source. Nucl. Instrum. Methods Phys. Res. Sect. B 179, 412–435. Kudryavtsev, Y., Bruyneel, B., Franchoo, S., Huyse, M., Gentens, J., Kruglov, K., Mueller, W.F., Prasad, N.V.S.V., Raabe, R., Reusen, I., Van den Bergh, P., Van Duppen, P., Van Roosbroeck, J., Vermeeren, L., Weissman, L. (2002). The Leuven isotope separator on-line laser ion source. Nucl. Phys. A 701, 465C–469C. Kutschera, W. (1990). Accelerator mass spectrometry: A versatile tool for research. Nucl. Instrum. Methods Phys. Res. B 50, 252–261. LaBelle, R.D., Hansen, C.S., Mankowski, M.M., Fairbank, W.M. (1996). Isotopically selective atom counting using photon burst detection. Phys. Rev. A 54, 4461–4463. Lariviere, D., Taylor, V.F., Evans, R.D., Cornett, R.J. (2006). Radionuclide determination in environmental samples by inductively coupled plasma mass spectrometry. Spectrochim. Acta B 61, 877–904. Letokhov, V.S. (1987). Laser Photoionization Spectroscopy. Academic Press, Orlando. Letokhov, V.S., Chebotayev, V.P. (1977). Nonlinear Laser Spectroscopy. Springer, Berlin–Heidelberg–New York. Lievens, P., Philipsen, V., Vandeweert, E., Silverans, R.E. (1991). Emission of ground and metastable state Ni and Co atoms during ion-beam sputtering: Quantitative measurements of population and kinetic energy distributions. Nucl. Instrum. Methods Phys. Res. Sect. B 135, 471–476. Lu, Z.-T., Wendt, K.D.A. (2003). Laser-based methods for ultra selective trace-isotope analyses. Rev. Sci. Instrum. 74, 1169–1179. Maul, J., Berg, T., Eberhardt, K., Hoog, I., Huber, G., Karpuk, S., Passler, G., Strachnov, I., Trautmann, N., Wendt, K. (2004). A laser desorption/resonance enhanced photoionisation TOF-system for the spatially resolved trace analysis of elements. Nucl. Instrum. Methods Phys. Res. Sect. B 226, 644–650. Maul, J., Karpuk, S., Huber, G. (2005a). Bimodal velocity distribution of atoms released from nanosecond ultraviolet laser ablation. Phys. Rev. B 71, 045428-(1–5). Maul, J., Eberhardt, K., Huber, G., Karpuk, S., Passler, G., Roca-Saiz, M.C., Strachnov, I., Trautmann, N., Wendt, K. (2005b). Multi-color resonance ionization of laser ablated gadolinium at high laser power. Opt. Commun. 256, 364–372. Moore, I.D., Nieminen, A., Billowes, J., Campbell, P., Geppert, Ch., Jokinen, A., Kessler, T., Marsh, B., Penttilä, H., Rinta-Antila, S., Tordoff, B., Wendt, K., Äystö, J. (2005). Development of a laser ion source at IGISOL. J. Phys. Rev. G: Nucl. Part. Phys. 31, S1499–S1502. Müller, P., Bushaw, B.A., Blaum, K., Diel, S., Geppert, C., Nähler, A., Trautmann, N., Nörtershauser, W., Wendt, K. (2001). 41 Ca ultratrace determination with isotopic selectivity >1012 by diode-laser-based RIMS. Fresenius J. Anal. Chem. 370, 508–512. Nunnemann, M., Erdmann, N., Hasse, H.-U., Huber, G., Kratz, J.V., Kunz, P., Mansel, A., Passler, G., Stetzer, O., Trautmann, N., Waldek, A. (1998). Trace analysis of plutonium in environmental samples by resonance ionization mass spectrometry (RIMS). J. Alloys Compd. 271–273, 45–48. Oughton, D.H., Skipperud, L., Fifield, L.K., Cresswell, R.G., Salbu, B., Day, P. (2004). Accelerator mass spectrometry measurement of 240 Pu/239 Pu isotope ratios in Novaya Zemlya and Kara sediments. Appl. Radiat. Isotopes 61, 249–253. Passler, G., Erdmann, N., Hasse, H.-U., Herrmann, G., Huber, G., Köhler, S., Kratz, J.V., Mansel, A., Nunnemann, M., Trautmann, N., Waldek, A. (1997). Application of laser mass spectrometry for trace analysis of plutonium and technetium. Kerntechnik 62, 85–90. Paul, M., Berkovits, D., Cecil, L.D., Feldstein, H., Hershkowitz, A., Kashiv, Y., Vogt, S. (1997). Environmental 90 Sr measurements. Nucl. Instrum. Methods Phys. Res. B 123, 394–399. Payne, M.G., Deng, L., Thonnard, N. (1994). Applications of resonance ionization mass spectrometry. Rev. Sci. Instrum. 65, 2433–2459. Peuser, P., Gabelmann, H., Lerch, M., Sohnius, B., Trautmann, N., Weber, M., Herrmann, G., Denschlag, H.O., Ruster, W., Bonn, J. (1981). Detection Methods for Trace Amounts of Plutonium. IAEA-SM-252/40, IAEA, Vienna, pp. 257–262. Pöllänen, R., Ikäheimonen, T.K., Klemola, S., Juhanoja, J. (1999). Identification and analysis of a radioactive particle in a marine sediment sample. J. Environ. Radioact. 45, 149–160. Ramsey, N.F. (1956). Molecular Beams. Oxford Univ. Press, New York.
Resonance ionization mass spectrometry for trace analysis of long-lived radionuclides
353
Salbu, B. (2000). Source-related characteristics of radioactive particles: A review. Radiat. Prot. Dosim. 92, 49–54. Salbu, B., Krekling, T., Oughton, D.H. (1998). Characterisation of radioactive particles in the environment. Analyst 123, 843–849. Salbu, B., Lind, O.C., Skipperud, L. (2004). Radionuclide speciation and its relevance in environmental impact assessments. J. Environ. Radioact. 74, 233–242. Sauvage, J., Boos, N., Cabaret, L., Crawford, J.E., Duong, H.T., Genevey, J., Girod, M., Huber, G., Ibrahim, F., Krieg, M., Le Blanc, F., Lee, J.K.P., Libert, J., Lunney, D., Obert, J., Oms, J., Peru, S., Pinard, J., Putaux, J.C., Roussière, B., Sebastian, V., Verney, D., Zemlyanoi, S., Arianer, J., Barre, N., Ducourtieux, M., Forkel-Wirth, D., Le Scornet, G., Lettry, J., Richard-Serre, C., Veron, C. (2000). COMPLIS experiments: Collaboration for spectroscopy measurements using a pulsed laser ion source. Hyperfine Interact. 129, 303–317. Schulz, C., Arnold, E., Borchers, W., Neu, W., Neugart, R., Neuroth, M., Otten, E.W., Scherf, M., Wendt, K., Lievens, P., Kudryavtsev, Y.A., Letokhov, V.S., Mishin, V.I., Petrunin, V.V., and the ISOLDE Collaboration (1991). Resonance ionization spectroscopy on a fast atomic ytterbium beam. J. Phys. B 24, 4831–4844. Schumann, P.G., Wendt, K.D.A., Bushaw, B.A. (2005). High-resolution triple-resonance autoionization of uranium isotopes. Spectrochim. Acta 60, 1402–1411. Sewtz, M., Backe, H., Dretzke, A., Kube, G., Lauth, W., Schwamb, P., Eberhardt, K., Grüning, C., Thörle, P., Trautmann, N., Kunz, P., Lassen, J., Passler, G., Dong, C.Z., Fritzsche, S., Haire, R.G. (2003). First observation of atomic levels for the element fermium (Z = 100). Phys. Rev. Lett. 90. 163002. Tamborini, G. (2004). SIMS analysis of uranium and actinides in microparticles of different origin. Mikrochim. Acta 145, 237–242. Tamborini, G., Betti, M. (2000). Characterisation of radioactive particles by SIMS. Mikrochim. Acta 132, 411–417. Tamborini, G., Wallenius, M., Bildstein, O., Pajo, L., Betti, M. (2002). Development of a SIMS method for isotopic measurements in nuclear forensic applications. Mikrochim. Acta 139, 185–188. Trautmann, N., Passler, G., Wendt, K.D.A. (2004). Ultratrace analysis and isotope ratio measurements of long-lived radioisotopes by resonance ionization mass spectrometry (RIMS). Anal. Bioanal. Chem. 378, 348–355. Tuniz, C., Bird, J.R., Fink, D., Herzog, G.F. (Eds.) (1998). Accelerator Mass Spectrometry. CRC Press, Boca Raton. Tykva, R., Sabol, J. (1995). Low-Level Environmental Radioactivity. Technomic Publ., Lancaster. Vandeweert, E., Philipsen, V., Bouwen, W., Thoen, P., Weidele, H., Silverans, R.E., Lievens, P. (1997). Anomalous sputtering behavior observed by quantitative measurements of the population partition of metastable Ni atoms. Phys. Rev. Lett. 78, 138–141. Vandeweert, E., Lievens, P., Philipsen, V., Bastiaansen, J., Silverans, R.E. (2001). Measurements of population partitions and state-selected flight-time distributions of keV ion-beam-sputtered metastable atoms. Phys. Rev. B 64, 195417-(1–12). Wacker, L., Chamizo, E., Fifield, L.K., Stocker, M., Suter, M., Synal, H.A. (2005). Measurement on actinides on a compact AMS system working at 300 kV. Nucl. Instrum. Methods Phys. Res. Sect. B 240, 452–457. Waldek, A., Erdmann, N., Grüning, C., Huber, G., Kunz, P., Kratz, J.V., Lassen, J., Passler, G., Trautmann, N. (2001). RIMS measurements for the determination of the first ionization potential of the actinides Actinium up to Einsteinium. In: Parks, J.E., Young, J.P. (Eds.), Resonance Ionization Spectroscopy 2000. In: AIP Conference Proceedings, vol. 584. AIP Press, Melville, NY, pp. 219–224. Wallenius, M., Mayer, K. (2000). Age determination of plutonium material in nuclear forensics by thermal ionization mass spectrometry. Fresenius J. Anal. Chem. 366, 234–238. Wallenius, M., Peerani, P., Koch, L. (2000). Origin determination of plutonium material in nuclear forensics. J. Radioanal. Nucl. Chem. 246, 317–321. Wendt, K., Trautmann, N. (2005). Recent developments in isotope ratio measurements by resonance ionization mass spectrometry. Int. J. Mass Spectrom. 242, 161–168. Wendt, K., Bhowmick, G.K., Bushaw, B.A., Herrmann, G., Kratz, J.V., Lantzsch, J., Müller, P., Nörtershäuser, W., Otten, E.W., Schwalbach, R., Seibert, U.-A., Trautmann, N., Waldek, A. (1997). Rapid trace analysis of 89,90 Sr in environmental samples by collinear laser resonance ionization mass spectrometry. Radiochim. Acta 79, 183–190. Wendt, K., Trautmann, N., Bushaw, B.A. (2000). Resonant laser ionization mass spectrometry: an alternative to AMS? Nucl. Instrum. Methods Phys. Res. Sect. B 172, 162–169. Wies, K. (2006). Dissertation. Institute of Physics, Johannes Gutenberg-University Mainz. Wies, K., Geppert, C., Blaum, K., Brück, K., Kluge, H.-J., Schwarz, S., Wendt, K. (2005). Development towards a laser ion source trap for the production of exotic species. Hyperfine Interact. 162, 29–38.
354
N. Erdmann et al.
Worden, E.F., Blaise, J., Fred, M., Trautmann, N., Wyart, J.-F. (2006). Spectra and electronic structures of free actinide atoms and ions. In: Morss, L.R., Edelstein, N.M., Fuger, J., Katz, J.J. (Eds.), The Chemistry of the Actinide and Transactinide Elements, third ed. Springer, Dordrecht, pp. 1836–1892. Wyse, E.J., Lee, S.H., Rosa, J.L., Povinec, P., Mora, S.J.d. (2001). ICP-sector field mass spectrometry analysis of plutonium isotopes: recognizing and resolving potential interferences. J. Anal. At. Spectrom. 16, 1107–1111. Zoriy, M.V., Pickhardt, C., Ostapczuk, P., Hille, R., Becker, J.S. (2004). Determination of Pu in urine at ultratrace level by sector field inductively coupled plasma mass spectrometry. Int. J. Mass Spectrom. 232, 217–224.
355
Environmental radioactive particles: A new challenge for modern analytical instrumental techniques in support of radioecology Maria Bettia,∗ , Mats Erikssonb , Jussi Jernströmc,1 , Gabriele Tamborinia a European Commission, Joint Research Centre, Institute for Transuranium Elements, P.O. Box 2340,
76137 Karlsruhe, Germany b International Atomic Energy Agency, Marine Environmental Laboratories, MC 98000, Monaco c University of Helsinki, Department of Chemistry, Laboratory of Radiochemistry, 00014 Helsinki, Finland
Abstract Environmental radioactive particles have assumed in recent years a predominant role for the understanding of the transport mechanisms in the environment as well as for the availability of radionuclides to humans. Their characterization by non-destructive instrumental methods can be very useful as applied to radioecological studies. Three specific cases of radioactive environmental particles stemming from different release scenarios are discussed in terms of the information available when characterized by non-destructive spectroscopic methods. In particular, application of SEM-EDX, SIMS and synchrotron-based techniques like µ-XRF, 3D-µ tomography and µ-XANES are considered. The complementarity of the methods is also highlighted.
1. Introduction Particles bearing radioactivity can be introduced into the environment as a consequence of several natural and anthropogenic processes. Human activities involving nuclear weapons and the nuclear fuel cycle may lead to the significant creation and release of radioactivity (Betti, 2000). Modern nuclear weapons produce radionuclides from fission, fusion and neutron activation products (Appleby and Luttrell, 1993). The type and composition of a nuclear device markedly affect the kinds of radioactivity produced, while the location and size of the detonation determine the quantity of the radioactivity released to the biosphere. Atmospheric nuclear weapons testing, conducted from 1945 until 1980, have introduced the majority of the deposited fraction of actinides onto the earth’s surface (Betti, 2000). The nuclear fuel cycle includes mining, milling, fuel enrichment and fabrication, reactor operation, spent fuel storage, reprocessing facilities and waste storage. Every single process ∗ Corresponding author. E-mail address:
[email protected] 1 Present address: Risø National Laboratory – Technical University of Denmark, P.O. 49, DK-4000 Roskilde, Den-
mark. RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11011-1
© 2008 Elsevier B.V. All rights reserved.
356
M. Betti et al.
may lead to the release of different radionuclides in different amounts to the environment and also to production of by-products that can be later dispersed. For instance, during the enrichment process of uranium hexafluoride, depleted uranium is obtained as a by-product (Betti, 2003). Nuclear accidents are other important sources of non-routine release of radioactivity into the environment. Accidents include an explosion in a waste storage tank, at Kyshtym, former USSR (1957), and those involving nuclear reactors at Windscale, UK (1957), Three Mile Island, USA (1978), Chernobyl, former USSR (1986). Accidental damage to nuclear weapons has resulted in the release of fissile material: at Palomares, Spain (1966) and Thule, Greenland (1968). Additional releases have occurred upon re-entry of satellites powered by nuclear sources [629 TBq 238 Pu source] such as SNAP (1964) and Cosmos 954 (1978) (Salbu, 2001). The ensuring contamination of the environment by radionuclides following a tropospheric nuclear explosion results from the spread of the explosion’s radioactive products in the atmosphere and their subsequent deposition on the surrounding areas. The radioactive products are formed, either as a result of the nuclear fission or fusion of the fuel or by the interaction between neutrons generated by the explosion and the device material (including the fuel) and the environment. They are also composed of radionuclides that were originally contained in the device itself before the explosion, as well as their transformation products. The distribution and fixing of radioactive products take place in the inert environmental material located in the immediate area of the detonation. This can give rise to the possibility of further spread of the different radionuclides within the environment either as separate atoms/molecules or combined with particle-carriers generated from inert material (Izrael, 2002). The radionuclide fractionation starts when the evaporated matter condenses, resulting in the selective capture of isotopes of certain elements by the liquid phases at the time of radioactive particle generation. As explained by Izrael (2002), this arises because the different products or their oxides have different saturated vapor pressures at the same temperature and nuclides belonging to the same mass chain at different stages of particle generation can as a consequence occur in the form of different chemical elements with different properties as a result of radioactive transformation. The most refractory elements are condensed within liquid particles of the ground or any other local material and are distributed within these particles, while the more volatile elements condense later, often after solidification of these particles. It would appear reasonable that the large particles that leave the radioactive cloud earlier than the small ones would be enriched in refractory nuclides, whereas the smaller particles would be enriched in volatiles or by those having gaseous or volatile precursors in a transformation chain. In this connection, it results that radioactive contamination of the environment will be largely controlled by the nuclear and physico-chemical processes occurring in the fireball or explosion cavity, as well as by the amount and properties of the material in the surrounding zone. In all serious events with significant releases of refractory elements, hot particles should be expected. Unless the source term is well characterized, model predictions of the dispersion, ecosystem transport and dose assessment will suffer from unacceptably large uncertainties. Hot particles represent point sources of radiological significance; particles may contribute to localized skin doses or particles inhaled or ingested can be retained within animals and humans for an unexpectedly long time. To assess the impact of hot particle contamination of ecosystems and to implement cost-efficient measures, information is needed on particle characteristics and on the behavior of particles and associated radionuclides in the ecosystem.
Environmental radioactive particles: A new challenge
357
The radioactive particles formed determine the transport mechanisms of radionuclides in the environment as well as their bioavailability. Their relatively large size carries the majority of the released activity (Eriksson, 2002; Kashkarov et al., 2003), therefore, their characterization as to elemental and isotopic composition as well as morphology, size distribution and chemical composition, is of great relevance to studying their environmental impact in terms of mobility, weathering, and corrosion rates (Salbu et al., 2004). From this point of view, the identification of the source term is necessary. On the other hand, source term identification is also essential in nuclear forensic investigations. Elemental mapping of the major constituents as well as the content and distribution of trace elements need to be determined. From the structure of the particle, in terms of the distribution of the chemical elements, the origin and the process involved in its production can be revealed. Moreover, information on the history of the material in terms of preferential leachability and, therefore, transfer to the food chain can be obtained. In relation to this, the elemental speciation plays one of the most important roles. In fact, chemical speciation as the determinant of reactivity is critical towards understanding and predicting the fate and transport in the environment of radionuclides like uranium, plutonium, neptunium, americium, strontium, which are the most common related to the above mentioned release scenarios. During the last few years, advances have been made in the exploitation of instrumental analytical techniques for characterizing radioactive environmental particles. Secondary ion mass spectrometry (SIMS) has been employed to develop precise and accurate methods for the determination of the isotopic composition of individual radioactive particles (Betti et al., 1999; Betti, 2005). Moreover, it has been demonstrated that synchrotron-based techniques are very powerful as applied to particle characterization for revealing their chemical composition in terms of element distribution as well as the oxidation states of matrix elements (Jernström et al., 2004; Török et al., 2004; Eriksson et al., 2005; Salbu et al., 2001, 2003, 2004, 2005). In this contribution some recent applications of these techniques to samples stemming from real nuclear release scenarios will be considered with the scope of illustrating how the modern instrumental analytical chemistry can contribute to radioecological studies. To demonstrate this, three specific cases, on which high expertise in our laboratory has been acquired, will be considered: the characterization of depleted uranium particles, of uranium-bearing particles stemming from a nuclear fuel reprocessing plant, and of U/Pu-containing particles dispersed in a marine environment after the accidental damage of a nuclear device. 2. Characterization of depleted uranium particles Depleted uranium (DU) is the by-product in the production of enriched uranium for civil uses (nuclear fuels). If, as feed material for the enrichment, reprocessed uranium is employed, DU can contain traces of minor actinides and fission products (WISE, 2003). Due to its physical properties (e.g., high density that is about twice that of lead) it finds applications for civil and military uses. For instance, it is employed in counterweights or ballast in aircraft; as radiation shield in medical equipment; as containers for the transport of radioactive material and as a chemical catalyst (Betti, 2003). Depleted uranium ammunitions have been used during the Gulf war and the conflicts in Bosnia–Herzegovina as well as in Kosovo. Radiological assessments on the effects of dispersion of uranium into the environment as well as investigations of the chemical and radio-toxicity of uranium have been carrying out (JERN, 2003).
358
M. Betti et al.
Different authors have characterized by instrumental analytical techniques depleted uranium particles collected in soil samples in Kosovo and Kuwait (Danesi et al., 2003; Török et al., 2004; Salbu et al., 2005). The common strategy consisted in separating the particles from the soil samples and using X-rays for the determination of their chemical composition coupled, in some cases, with scanning electron microscopy. µ-XRF as well as SEM-EDXRF can be employed to obtain information on the size distribution of the particles. Measurements performed independently by two different groups found that most of the identified depleted uranium particles in the samples were in the range of some µm in diameter (less than 10 µm). Exceptional larger particles, up to 40 µm (Danesi et al., 2003), were also detected. In order to detect the isotopic composition of the depleted uranium particles, SIMS has been the most appropriate technique as it has already been extensively used for this purpose for particles of different origin (Tamborini, 1998; Tamborini et al., 1998, 2002; Tamborini and Betti, 2000; Betti et al., 1999; Erdmann et al., 2000; Tamborini et al., 2004). Soil particle samples collected from the Kosovo area a few years after the war showed the presence of fine particles with depleted uranium as a major component which were easily identified by SIMS (Török et al., 2004). The ultrafine uranium particles were often attached to larger soil particles and contained Ti and Al being typical components of the penetrator and its cladding. The particles localized by the automated SEM with gunshot residue program were then analyzed by SIMS as for uranium isotopic ratios. The results obtained clearly revealed the presence of DU and traces of 236 U were also detected. SIMS isotopic images for 238 U and 235 U are reported in Figure 1. Clearly due to the depletion of uranium the signal of the 235 cannot be revealed. The results obtained from the measurements of the particles are reported in Table 1. As for 234 U and 236 U, due to the low signal intensity the error is often larger. In general, the total uncertainty on the minor isotopes is decreasing (better counting statistics) when the enrichment increases. The uncertainty on the 236 U is significantly higher than that on the 234 U because the 236 intensity is obtained after correction for hydride contribution and the respective errors are propagated into the 236 uncertainty. With such a particle diameter (∼1 µm), the precision that can be achieved is considered the detection limit of the method in these conditions. The signal on mass 239 (due only to the formation of 238 UH+ molecules) is used to correct the signal on mass 236 for the contribution of 235 UH+ . As for 235 U, in this case the uncertainty is not strongly dependent on the intensity of the signal but on other parameters, e.g., the analysis time. Török et al. (2004) also investigated the chemical composition of the DU particles by µXRF and EMPA as well as the uranium oxidation states by µ-XANES. The results showed that mostly uranium is present in the particles in the U(IV) form, the maximum ratio of U(VI) to the total U content being obtained as 24% indicating that the particle mainly contained the less mobile form of U. Salbu et al. (2003) also analyzed DU particles from Kosovo using micro-XANES. Based on the results for 12 particles, the authors claimed that approximately 50% of the DU particles were in the form of UO, while the remaining DU particles were U3 O8 or a mixture of oxidized forms. Due to the fact that in both studies the number of particles analyzed was not statistically significant, we can conclude that the results are in agreement showing the potential of the technique in providing valuable information on the percentage of the element distributed in the different oxidation states. Salbu et al. (2005) also studied the oxidation states of uranium in DU particles stemming from Kuwait. They found that, compared to well-defined standards, all investigated particles
Environmental radioactive particles: A new challenge
359
Fig. 1. SIMS images of 235 U and 238 U in depleted uranium particles separated from soil samples stemming from Kosovo sites.
were oxidized. Furthermore, in this investigation, the authors demonstrated that the DU originated from reprocessed uranium. In fact, from the analysis of the particles by accelerator mass spectrometry, the presence of 236 U was revealed.
3. Characterization of uranium-containing particles originating from a nuclear fuel reprocessing plant The British Nuclear Fuels plc (BNFL) establishments at Sellafield have been releasing, according to authorization, discharges of low-level radioactive waste into the Irish Sea since 1952. The peak period for the releases occurred in the 1970s, since when the level of discharges has been reduced (Betti et al., 2004). The most important part of the discharged waste originates both from the cooling ponds in which the spent Magnox fuel elements are being stored prior to re-processing, and from the ‘sea tanks’ where low-level liquid wastes are
360
Table 1 Isotopic composition obtained by SIMS for particles of DU separated from a sample collected at 5–10 cm depth (after Török et al., 2004) 234 U/238 U
235 U/238 U
Ratio
1σ
Ratio
1σ
Ratio
1σ
wt%
1σ
wt%
1σ
wt%
1σ
wt%
1σ
1 2 3 4 5 6 7 8 9 10 11 12
1.09E−05 1.46E−05 7.49E−06 2.46E−05 2.21E−05 2.70E−05 4.20E−05 1.57E−05 5.57E−05 1.37E−05 1.80E−05 2.03E−05
6.30E−06 6.98E−06 8.13E−06 8.26E−06 2.21E−05 1.26E−05 4.13E−05 1.05E−05 3.77E−05 1.46E−05 8.53E−06 1.46E−05
2.03E−03 2.04E−03 2.04E−03 2.06E−03 1.92E−03 2.03E−03 2.06E−03 2.05E−03 2.16E−03 2.10E−03 2.03E−03 2.08E−03
1.06E−04 1.51E−04 2.01E−04 1.52E−04 1.93E−04 1.16E−04 4.74E−04 1.26E−04 2.51E−04 1.84E−04 1.49E−04 2.23E−04
3.60E−05 3.92E−05 3.38E−05 4.60E−05 3.58E−05 6.05E−05 6.47E−05 4.56E−05 1.04E−04 3.40E−05 5.27E−05 6.45E−05
1.22E−05 1.22E−05 2.50E−05 2.05E−05 2.29E−05 1.54E−05 5.59E−05 1.82E−05 5.60E−05 1.97E−05 2.27E−05 3.73E−05
0.001 0.001 0.001 0.002 0.002 0.003 0.004 0.002 0.005 0.001 0.002 0.002
0.001 0.001 0.001 0.001 0.002 0.001 0.004 0.001 0.004 0.001 0.001 0.001
0.200 0.201 0.201 0.203 0.189 0.200 0.203 0.202 0.213 0.207 0.200 0.205
0.010 0.015 0.020 0.015 0.019 0.011 0.047 0.012 0.025 0.018 0.015 0.022
0.004 0.004 0.003 0.005 0.004 0.006 0.006 0.005 0.010 0.003 0.005 0.006
0.001 0.001 0.002 0.002 0.002 0.002 0.006 0.002 0.006 0.002 0.002 0.004
99.795 99.794 99.795 99.790 99.805 99.791 99.787 99.792 99.771 99.788 99.793 99.787
0.011 0.015 0.020 0.015 0.019 0.012 0.048 0.013 0.026 0.018 0.015 0.023
Average
2.27E−05
2.05E−03
236 U/238 U
5.14E−05
234 U
0.002
235 U
0.202
236 U
0.005
238 U
99.791
M. Betti et al.
Particle No.
Environmental radioactive particles: A new challenge
361
(A)
(B) Fig. 2. Alpha track images of two radioactive particles separated from the sediment. Exposure time for particle A was 43 d, for particle B 8 d. (Reproduced with permission of the Royal Society of Chemistry; Jernström et al., 2004.)
collected and neutralized before being discharged. Actinides, namely Np, Pu, Am and Cm, detected in both cooling pond water and ‘sea tank’ water have been found to be mainly attached to particulate matter. Bulk profiles of a core sediment have been studied for Np, Pu, Am and major fission products (Perna et al., 2005). The highest level of activity was associated with hot particles, typically less than 20 × 50 µm in size, originating from spent nuclear fuel debris from the BNFL plant. These most radioactive particles usually contained uranium together with plutonium, americium and curium as the major alpha emitters. Jernström et al. (2004) separated radioactive particles from an Irish Sea sediment core after they had been localized using solid state nuclear track detection (SSNTD). Figure 2 shows the alpha track images of two radioactive particles separated from the sediments. Then the particles were characterized by electron microscope and different X-ray microanalysis methods including micro X-ray fluorescence spectroscopy (µ-XRF) and micro X-ray absorption near-edge struc-
362
M. Betti et al.
ture spectroscopy (µ-XANES). After these investigations the particles remained intact and were available for further studies. Due to the fact that the bulk analysis revealed the presence of Pu and Am, the isolated particles were also investigated as for these radioisotopes. Only in one particle was it possible to detect a weak signal of Pu by µ-XRF. Using the same technique, elemental maps were constructed in order to find out the position of the highest U intensity, and any existence of elemental correlation. In particular, only for two particles trace amounts of thorium were found in the same region as the high U intensity. A comparison was made with the results obtained from a particle originating from the same sediment core but one layer below. The analysis of this particle revealed a high intensity of zirconium located with thorium, uranium, yttrium and hafnium. During the scanning of the U-rich particles additional uranium signals with lower intensity were detected. A common factor for these other few particles is a slightly higher Th intensity compared to U intensity, the Th/U ratio being approximately from 1 to 4. This strongly differs from the Th/U intensity ratio as measured for other particles. Oxidation states were also investigated by µ-XANES. Compared to the µ-XANES spectra of the U(IV), U3 O8 and U(VI) standards, the shift in the inflection point energy in the spectra of all three analyzed particles indicates an increase in the uranium oxidation state. Uranium is present in the particles mostly in the U(IV) form, the ratio of U(VI) to the total U content being between 34 ± 6% and 40 ± 6%. The uranium in the particles was found to be in two oxidation states at different percentages: 60–66% as U(IV) and 34–40% as U(VI).
4. Characterization of U/Pu particles from an inadvertent destruction of nuclear weapons In our laboratory particular attention has been focused on the source identification of particles released as a consequence of the accident which occurred in the Thule area (NW-Greenland) in January 1968. Details of this accident and the results obtained have been published elsewhere (Eriksson, 2002; USAF, 1970; Aarkrog, 1971; Aarkrog et al., 1984; McMahon et al., 2000; Moring et al., 2001). Moring et al. (2001) identified particles containing U and Pu originating form this accident in sea sediment samples. The environmental radioactive particles, such as those analyzed in our laboratory, are generally embedded in a bulk matrix and therefore it is necessary to isolate them and perform specific single grain analysis to provide unbiased results. The analytical approach has been to perform non-destructive measurements in order to reveal their chemical composition (Eriksson et al., 2005). One of the advantages of using non-destructive methods is the direct analysis of the samples, thus avoiding problems of cross-contamination, which is particularly important for U at the ng levels present in the particles here studied. Characteristic L X-rays and gamma-rays from the radioactive disintegrations were measured. This provides information on the specific radionuclide composition of the radioactive particles. Before analyzing the particles by microscopic X-ray fluorescence (µ-XRF), they have been examined by scanning electron microscopy (SEM) combined with energydispersive X-ray (EDX) or wavelength-dispersive X-ray (WDX) spectrometry. This is of particular interest when the particles are supposed to contain a mixture of U and Pu, revealing the surface and to some extent subsurface composition. Utilizing µ-XRF at a synchrotron
Environmental radioactive particles: A new challenge
363
radiation (SR) facility, information on the elemental composition of the particles has been obtained. In particular, by using a beam size of 2.5 µm, the fine structure of the material was observed. The elemental distributions of U and Pu in a particle were also studied with the µXRF tomography technique. Microscopic X-ray absorption near-edge structure (µ-XANES) spectroscopy has been exploited to study the oxidation states of uranium and plutonium in U/Pu mixed particles. In order to obtain single grain hot particles from bulk sediment samples, a sample splitting technique was used. The technique is based on the decay of the plutonium isotope 241 Pu (T1/2 = 14.35 y), which is present in the weapons material, to the gamma-emitting daughter 241 Am (T 1/2 = 432.2 y, gamma line: 59.54 keV, intensity: 35.9%), measurable by gammaspectrometry. When 241 Am activity is identified, the sample is split into two equal halves and measured again. The part with the activity is again divided and measured. After about 20 sample splittings, starting with about 150 g of bulk material, only a few grains (∼µg) remain. This technique is explained in detail elsewhere (Eriksson et al., 2002). Only one single radioactive particle remains after a complete sample splitting of the bulk material. The longest working time spent on the sample splitting (the particle with the lowest activity, ∼2.5 Bq) was around 4 h. The single radioactive particle was attached to an adhesive carbon tape. To identify and locate the particle position on the tape, the Beta Camera was used. Since the Beta Camera is also a light-sensitive device, and there is need for a position reference system on the tape, five holes were punched out in a special pattern. The holes were superimposed by light on the image obtained from the acquisition of the hot particle. For all the particles in the present study, the acquisition time on the Beta Camera was less than 30 min. This is a very short time compared to the more common autoradiograph film technique, which usually takes weeks before an image is produced. All the particles were analyzed by SEM with an EDX detector attached. The morphology, size and the surface elemental composition of the particles were studied. The secondary (SE) and back-scattered electron (BE) images also served as guiding maps when relocation of the particles was done at the SR-facilities ANKA and HASYLAB. The SEM used has a maximum acceleration voltage of 20 kV, which did not allow the efficient excitation of U and Pu L shells. In the EDX spectra the M X-ray peaks of U and Pu were not resolved due to the strong overlap between the Mα peaks of Pu and the Mβ peaks of U. Absolute determination of the Pu/U elemental ratio could not be done, as standard mixed Pu/U particles would be required to calibrate the pulse height spectra in combination with deconvolution of the spectra or a Monte Carlo simulation of the source-sample-detector system (Ro et al., 2003). From the BE images with the EDX spectra it resulted that both U and Pu are present on the surface of all the particles. The particle sizes ranged from 15 to 42 µm and some of the particles are embedded or to some extent coated. These coatings were analyzed with EDX and contained elements typical in sediment material, i.e. Si, Al, Mg and Fe. Several spots on the high atomic number areas of the particles were analyzed with EDX. All the analyses showed that U and Pu co-exist on the whole high atomic number surface. From the U and Pu M X-ray peak intensity variations between different particles were observed, however the magnitude of the variations could not be determined with this system. In addition, one particle (Thu79-3) was measured with a SEM equipped with a WDX analyzer. The difference between the two systems is that EDX uses a Si(Li) detector with a multichannel analyzer (MCA), while WDX is based on the diffraction of the produced charac-
364
M. Betti et al.
teristic X-rays in a crystal lattice and the X-ray quanta are detected by a proportional counter. WDX measures one-energy only (wavelength) at a time, leading to a better resolution but a lower efficiency. However, the low efficiency leading to low count-rate can be compensated to some extent by the use of a higher beam current from the filament. With the WDX system used it was possible to determine the Pu/U elemental ratios from the M X-ray lines without doing deconvolution on the spectra. This was done by correcting the area for the U Mα and Pu Mβ peaks (non-overlapping peaks) with the emission probabilities. This particle is constituted of many small fragments/inclusions. Carbon is the main constituent of the particle, indicating that it was formed during the accident, probably from burning of rubber or similar materials. When the weapons were disintegrated by conventional explosion, fragments from them were trapped in this piece of burning material. Five spots/inclusions of this particle were analyzed for the Pu/U elemental ratio. The determined Pu/U elemental ratios by SEM-EDX were 0.09 to 0.43 in these five spots. The large variations in the ratio are probably due to the fact that the inclusions originated from different parts of the weapons rather than to an effect of their preferential leaching. The SEM back-scattered image (top) of one Thule hot particle with corresponding EDXspectrum and the uranium and plutonium intensity maps as obtained by µ-XRF are shown in Figure 3. By performing µ-XRF, it was found that U and Pu are homogeneously distributed in the particles. The Pu/U intensity ratios vary between 0.22 and 0.36. This variation may reflect several phenomena. Among these we can consider the preferential leaching of one element relative to the other; the origin of the particles from different parts of one of the nuclear weapons involved; or that the particles stem from a different source term. Detailed investigation on the Pu isotopic ratios by mass-spectrometry, which are on going at the present in our laboratory, complementary to those described below such as micro-spatial elemental distribution and micro-tomography, can help in concluding which of the above-mentioned causes is the most probable. The same particle analyzed with the SEM-WDX system was analyzed using a 2.5 µm µXRF beam at ANKA. The particle, with a size of 350 × 250 µm2 , was scanned over an area of 130 × 100 µm2 with step sizes of 1.25 µm in both z and y directions and 3 s per point as measuring time. It was observed that the U and Pu Lα intensity maps correspond fairly well; however, in the U/Pu intensity ratio there are variations. Some regions have a U/Pu intensity ratio as high as 20 (Pu/U 0.05). This means that the inclusions in the particle originate from different parts of the weapons or from different weapons, rather than being due to an effect of leaching, as the inclusions have been more or less isolated from the aquatic environment in this large particle. µ-XRF measurements with a 20 µm beam would never have revealed the fine structure of the particle due to the “low” spatial resolution. When analyzing individual Pu/U particles with respect to source-term identification and leachability, the use of the larger (20 µm) and smaller (2.5 µm) spot sizes is a complementary requirement. The large beam provides the average elemental ratio in the particles and the small beam the possible heterogeneity in the elemental distribution. The elemental distributions in two particles were investigated: a U/Pu-rich particle recovered from the sea sediment near Thule, NW Greenland, and a Pu-rich particle embedded in a coral matrix, collected from the surface soil in Mururoa Atoll. The particles were scanned in pencil-beam geometry. For each particle, 64 area projections, separated by 5.625 degrees,
Environmental radioactive particles: A new challenge 365
Fig. 3. SEM back-scattered image (top) of one Thule hot particle with corresponding EDX-spectrum. Uranium and plutonium intensity maps as obtained by µ-XRF.
366
M. Betti et al.
Fig. 4. Reconstructed distribution of elements in U/Pu-rich particles recovered from sea sediment, Thule, Greenland. The plutonium (blue) and uranium (green) are coated with iron (red) rich sediment matrix.
consisting of 64 (horizontal) by 10 (vertical) pixels, were collected simultaneously in X-ray fluorescence and X-ray absorption modes. The horizontal beam spacing was 19 and 11 µm for the Mururoa and the Thule particles, respectively. The corresponding vertical beam spacing was 42 and 20 µm. The 3D distributions of Pu, U and the matrix elements (Sr for the Mururoa particle and Fe for the Thule particle) were reconstructed. The individual slices of the tomographic reconstruction were obtained using a filtered back-projection algorithm. A preliminary method of absorption effect correction was applied. In this method the data collected by the X-ray fluorescence detector, the XRF sinograms, were corrected for absorption effects using the X-ray absorption sinograms collected with the SDD placed behind the analyzed sample. The reconstructed slices were processed by a 3D rendering software AMIRA to obtain 3D element distribution models. From the reconstructed data it can be noticed that the Thule particle, recovered from the sea sediment, has been partially coated with a Fe-rich sediment matrix during its long residence (29 y) in the seabed. Uranium and plutonium elemental distribution are not even in the particles. The Mururoa particle, recovered from the surface of the coral atoll, was found to be attached to the coral matrix only from one side. This suggests that the merging of plutonium with the matrix material occurred during the nuclear test explosion, the source term of this particle. The reconstruction of the U and Pu distribution in the particle stemming from Thule sea sediments as obtained by tomography measurements at the synchrotron is reproduced in Figure 4. In order to study the reactivity of the particles in terms of mobility, leachability, weathering and transport/transfer of radionuclides into the environment and from the environment to humans, the determination of oxidation states is essential. XANES is an established technique for the determination of oxidation states and local chemical environment, even at relatively low concentrations. As already asserted, studies have systematically been performed for actinide aqueous ions (Conradson et al., 1998, 2004) and it has been re-
Environmental radioactive particles: A new challenge
367
cently exploited also for U at particle level (Salbu et al., 2001, 2003; Török et al., 2004; Jernström et al., 2004). The particles separated from the Thule sediments were investigated by µ-XANES for U and Pu oxidation states. For the first time a systematic study of Pu oxidation states in environmental released particles has been performed. In order to study uranium oxidation states in individual particles U-L3 µ-XANES spectra were collected from the center as well as from the edge region of each particle. The U XANES spectra collected at the center of four U/Pu particles sampled at Thule (Thu68-3, Thu68-5, Thu68-6, 975374-5) were compared with the spectra of the standards. The zero of the energy scale was set to the inflection point of the U(IV) standard. All spectra were processed using the least-squares fitting method described above. No significant difference was encountered between the results obtained for spectra collected from the center and from the edge of the same particles. The error of the fitting method is within 5% (absolute). Uranium is present in the particles mostly in the U(IV) form, the maximum ratio of U(VI) to the total U content was obtained as 25% indicating that the particles mainly contained the less mobile form U(IV). Plutonium oxidation states were studied as for uranium, i.e. through recording and fitting Pu-L3 µ-XANES spectra from different regions of each individual particle. The Pu-L3 XANES spectra collected at the center of four U/Pu particles (Thu68-3, Thu68-5, Thu68-6, 975374-5) were compared to the standard spectra. The zero of the energy scale was set to the inflection point of the Pu(IV) standard. The plutonium oxidation state was also not significantly different in different regions of the same particle. Two kinds of particles could be distinguished. The particles Thu68-3 and Thu68-5 contain Pu mostly (more than 90%) in the Pu(IV) form. In the particles Thu68-6 and 975374-5, however, the Pu(VI) form is dominant (67–75%). From the results obtained it is clear that the reactivity of the particles in the environment as well as in human fluids will be different and therefore it would be worthwhile to study the extent to which the preponderance of one or other oxidation state will influence Pu fate and transport. In order to obtain this information, leaching of the particles in environmental media and human body fluids needs to be carried out. Lind et al. (2005) have recently examined particles stemming from the Thule environment by µ-XANES. Their measurements showed that U and Pu are present in the particles as mixed oxides. U was found in the oxidation state IV whereas Pu apparently is a mixture of Pu(III) and Pu(IV).
5. Conclusions In the three specific cases examined here, it is clearly shown that the combination of nondestructive instrumental analytical techniques constitutes a powerful methodology for the characterization of radioactive environmental particles independently of the release scenario. In the case of the investigation of DU particles, the automatic SEM is a powerful tool to localize, identify and characterize particles containing uranium present in soil samples in terms of size distribution and elemental chemical composition. In fact, for most of the particles found in the soil samples stemming from Kosovo, an average size of about 1 µm was measured, which is in agreement with the value reported by other authors. The elemental composition revealed the presence of Ti and Al, which are peculiar components of the penetrator
368
M. Betti et al.
and its cladding, respectively. SIMS has been particularly useful to determine the uranium isotopic composition and it indicates clearly that the particles contained depleted uranium. Moreover, the presence of traces of 236 U was evident, as also indicated by AMS in other studies. By SIMS, however, exact quantification of this isotope is hindered by a high measurement uncertainty. In the case of the particles from Irish Sea sediment, localized and characterized using solid state nuclear track detection, electron microscopy and X-ray spectrometry, methods based on synchrotron radiation light are powerful for the identification of their source term. Two of the particles were found to be possibly stemming from nuclear fuel debris, while one of them was of natural origin. Moreover, the techniques are not destructive, and after revealing the chemical composition and the oxidation states, the particles are still available for isotopic composition examination and further studies of leaching properties. As for the U/Pu mixed particles from the Thule environment, non-destructive synchrotron radiation based techniques, combined with SEM-EDX-WDX, gamma and X-ray spectrometry, can provide information on the source terms of discrete environmental radioactive particles. In particular, the chemical mapping as well as fine structure composition can be achieved by µ-XRF and the 3D elemental distribution by tomography. The results for the four particles investigated by µ-XRF showed that U and Pu are homogeneously distributed. However, in the particles investigated with µ-tomography and the 2.5 µm µ-XRF beam, heterogeneities of the U/Pu ratio distribution were observed. XANES providing information on the oxidation states can give an indication of the reactivity of the particles and consequently of their transport and fate in the environment and human body fluids. The combination of all these non-destructive techniques is a first step in providing an entire picture of the origin, route and fate of nuclear material released as particles. From all these reflections and conclusions, it appears that an analytical strategy based on a combination of all the techniques discussed in this contribution is the future tool to be exploited widely in radioecological studies.
Acknowledgements The IAEA is grateful for the support provided to its Marine Environment Laboratories by the Government of the Principality of Monaco.
References Aarkrog, A. (1971). Radioecological investigations of plutonium in an arctic marine environment. Health Phys. 20, 31–47. Aarkrog, A., Dahlgaard, H., Nilsson, K. (1984). Further studies of plutonium and americium at Thule, Greenland. Health Phys. 46, 29–44. Appleby, L.J., Luttrell, S.P. (1993). Case-studies of significant radioactive releases. In: Warner, F., Harrison, R.M. (Eds.), Radioecology after Chernobyl. John Wiley & Sons, Chichester, UK, pp. 33–53. Betti, M. (2000). Environmental monitoring of radioisotopes by mass spectrometry and radiochemical methods in urban areas. Microchem. J. 67, 363–373. Betti, M. (2003). Civil use of depleted uranium. J. Environ. Radioact. 64 (2), 113–119.
Environmental radioactive particles: A new challenge
369
Betti, M. (2005). Isotope ratio measurements by secondary ion mass spectrometry (SIMS) and flow discharge mass spectrometry (GDMS). Int. J. Mass Spectrom. 242, 169–182. Betti, M., Tamborini, G., Koch, L. (1999). Use of secondary ion mass spectrometry in nuclear forensic analysis for the characterization of plutonium and highly enriched uranium particles. Anal. Chem. 71, 2616–2622. Betti, M., Aldave de las Heras, L., Janssens, A., Henrich, E., Hunter, G., Gerchikov, M., Dutton, M., van Weers, A.W., Nielsen, S., Simmonds, J., Bexon, A., Sazykina, T. (2004). J. Environ. Radioact. 74, 243–254. Conradson, S.D., Al Mahamid, I., Clark, D.L., Hess, N.J., Hudson, E.A., Neu, M.P., Palmer, Ph.D., Runde, W.H., Tait, C.D. (1998). Oxidation state determination of plutonium aqua ions using X-ray absorption spectroscopy. Polyhedron 17, 599–602. Conradson, S.D., Abney, K.D., Begg, B.D., Drady, E.D., Clark, D.L., den Auwer, Ch., Ding, M., Dorhout, P.K., Espinosa-Faller, F.J., Gordon, P.L., Haire, R.G., Hess, N.J., Hess, R.F., Keogh, D.W., Lander, G.H., Lupinetti, A.J., Morales, L.A., Neu, M.P., Palmer, Ph.D., Paviet-Hartmann, P., Reilly, S.D., Runde, W.H., Tait, C.D., Veirs, D.K., Wastin, F. (2004). Higher order speciation effects on plutonium L3 X-ray absorption near edge spectra. Inorg. Chem. 43, 116–131. Danesi, P.R., Markowicz, A., Chine-Cano, E., Burkarrt, W., Salbu, B., Donohue, D., Ruedenauer, F., Hedberg, M., Vogt, S., Zahradnik, P., Ciurapinski, A. (2003). Depleted uranium particles in selected Kosovo samples. J. Environ. Radioact. 64 (2), 143–154. Erdmann, N., Betti, M., Stetzer, O., Tamborini, G., Kratz, J.V., Trautmann, N., van Geel, J. (2000). Production of monodisperse uranium oxide particles and their characterization by scanning electron microscopy and secondary ion mass spectrometry. Spectrochim. Acta Part B 55, 1565–1575. Eriksson, M. (2002). On Weapons Plutonium in the Arctic Environment (Thule, Greenland). Risø-R-1321(EN), Risø National Laboratory, Roskilde, Denmark. Eriksson, M., Ljunggren, K., Hindorf, C. (2002). Plutonium hot particle separation techniques using real-time digital image systems. Nucl. Instrum. Methods Phys. Res. Sect. A 488, 375–380. Eriksson, M., Osán, J., Jernström, J., Wegrzynek, D., Simon, R., Chinea-Cano, E., Markowicz, S., Bamford, A., Tamborini, G., Török, S., Falkenberg, G., Alsecz, A., Dahlgaard, H., Wobrauschek, P., Streli, C., Zoeger, N., Betti, M. (2005). Source term identification of environmental radioactive Pu/U particles by their characterisation with non-destructive spectrochemical analytical techniques. Spectrochim. Acta Part B 60, 455–469. Izrael, Yu.A. (2002). In: Baxter, M.S. (Ed.), Radioactive Fallout after Nuclear Explosions and Accidents. Elsevier, Oxford, UK. JERN (2003). Special Issue on DU. J. Environ. Radioact. 64, 87–259. Jernström, J., Eriksson, M., Osán, J., Tamborini, G., Török, S., Simon, R., Falkenberg, G., Alsecz, A., Betti, M. (2004). Non-destructive characterization of low radioactive particles from Irish Sea sediment by micro X-ray synchrotron radiation techniques: micro X-ray fluorescence (µ-XRF) and micro X-ray absorption near edge spectroscopy (µ-XANES). J. Anal. At. Spectrom. 19, 1428–1433. Kashkarov, L.L., Kalinina, G.V., Perelygin, V.P. (2003). α-Particle track investigation of the Chernobyl Nuclear Power Plant accident region soil samples. Radiat. Meas. 36, 529–532. Lind, O.C., Salbu, B., Janssens, K., Proost, K., Dahlgaard, H. (2005). J. Environ. Radioact. 81, 21–32. McMahon, C.A., Vintró, L.L., Mitchell, P.I., Dahlgaard, H. (2000). Oxidation state distribution of plutonium in surface and subsurface waters at Thule, northwest Greenland. Appl. Radiat. Isotopes 52, 697–703. Moring, M., Ikäheimonen, T.K., Pöllänen, R., Ilus, E., Klemola, S., Juhanoja, J., Eriksson, M. (2001). Uranium and plutonium containing particles in a sea sediment sample from Thule, Greenland. J. Radioanal. Nucl. Chem. 248, 623–627. Perna, L., Jernström, J., Aldave de las Heras, L., Hrnecek, E., de Pablo, J., Belti, M. (2005). Characterization of an Irish Sea radioactively contaminated marine sediment core by radiometric and mass spectrometric techniques. J. Radioanal. Nucl. Chem. 263, 367–373. Ro, C.-U., Osán, J., Szalóki, I., de Hoog, J., Worobiec, A., Van Grieken, R. (2003). A Monte Carlo program for quantitative electron-induced X-ray analysis of individual particles. Anal. Chem. 75, 851–859. Salbu, B. (2001). Actinides associated with particles. In: Kudo, A. (Ed.), Plutonium in the Environment. Elsevier, Oxford, UK, pp. 121–138. Salbu, B., Krekling, T., Lind, O.C., Oughton, D.H., Drakopoulos, M., Simionovici, A., Snigireva, I., Snigirev, A., Weitkamp, T., Adams, F., Janssens, K., Kashparov, V.A. (2001). High energy X-ray microscopy for characterisation of fuel particles. Nucl. Instrum. Methods Phys. Res. Sect. A 467–468, 1249–1252.
370
M. Betti et al.
Salbu, B., Janssens, K., Lind, O.C., Proost, K., Danesi, P.R. (2003). Oxidation states of uranium in DU particles from Kosovo. J. Environ. Radioact. 64 (2), 167–173. Salbu, B., Lind, O.C., Skipperud, L. (2004). Radionuclide speciation and its relevance in environmental impact assessments. J. Environ. Radioact. 74, 233–242. Salbu, B., Janssens, K., Lind, O.C., Proost, K., Gijsels, L., Danesi, P.R. (2005). Oxidation states of uranium in depleted uranium particles from Kuweit. J. Environ. Radioact. 78, 125–135. Tamborini, G. (1998). The development of the SIMS technique for the analysis of radionuclide in microparticles from environmental materials. University of Paris-Sud, Orsay, France. Tamborini, G., Betti, M. (2000). Characterisation of radioactive particles by SIMS. Mikrochim. Acta 132, 411–417. Tamborini, G., Betti, M., Forcina, V., Hiernaut, T., Koch, L. (1998). Application of secondary ion mass spectrometry to the identification of single particles of uranium and their isotopic measurement. Spectrochim. Acta Part B 53 (9), 1289–1302. Tamborini, G., Phinney, D., Bildstein, O., Betti, M. (2002). Oxygen isotopic measurements by secondary ion mass spectrometry in uranium oxide microparticles: A nuclear forensic diagnostic. Anal. Chem. 74, 6098–6101. Tamborini, G., Donohue, D.L., Rudenauer, F.G., Betti, M. (2004). Evaluation of practical sensitivity and useful ion yield for uranium detection by secondary ion mass spectrometry. J. Anal. At. Spectrom. 19, 203–208. Török, S., Osán, J., Vincze, L., Kurunczi, S., Tamborini, G., Betti, M. (2004). Characterization and speciation of depleted uranium in individual soil particles using microanalytical methods. Spectrochim. Acta Part B 59, 689– 699. USAF Nuclear Safety (1970). Project Crested Ice, AFRP 122-1, Jan/Feb/Mar 1970, No. 1, p. 65. WISE Uranium Project (2003). http://www.antenna.nl/wise/uranium.
371
Activation analysis for the determination of long-lived radionuclides Xiaolin Hou∗ Risø National Laboratory, NUK-202, Technical University of Denmark, DK-4000, Denmark Abstract Activation analysis, as a sensitive analytical technique for the determination of isotopes, is widely used for the determination of radionuclides, particularly long-lived radionuclides, in environmental samples. The principle, methodology and procedure of activation analysis are presented in this article. The applications of activation analytical methods for the determination of 99 Tc, 135 Cs, 129 I, 232 Th, 235 U, 238 U, and 237 Np in environmental samples are described and compared with other competitive methods.
1. Introduction In the study of the environmental radioactivity and radioecology and the assessment of environmental effects of nuclear power plants and disposal of nuclear waste, a number of radionuclides have to be determined. In general, determination of radionuclides is performed by direct activity measurement, i.e. radiometric analysis. According to the particles emitted by target radionuclides, α-, β-, and γ -spectrometry are used in the radiometric analysis. These methods are sensitive for short- and medium-lived radionuclides, however, for long-lived radionuclides, these methods are not sensitive enough for the analysis of environmental samples. Mass spectrometry is a typical method for isotope analysis, in which thermal ionization mass spectrometry (TIMS) and accelerator mass spectrometry (AMS) have a high sensitivity for the determination of many radionuclides (e.g., Povinec, 2004), and recently inductively coupled plasma mass spectrometry (ICP-MS) is widely used for the determination of some radionuclides, especially long-lived ones (e.g., Wyse et al., 2001). Activation analysis is well known as a sensitive and reliable method for the determination of trace elements, but it is actually an analytical method for isotopes, because it is based on the nuclear characteristics and nuclear reactions of the isotope of an element. It can therefore be used for the determination of radionuclides, especially long-lived radionuclides. According to the bombarding particles, activation analysis is divided into neutron, photon and charged particle activation analysis. Of them, the most used is neutron activation analysis (NAA) using a nuclear reactor as neutron source because of its high sensitivity and accuracy. ∗ E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11012-3
© 2008 Elsevier B.V. All rights reserved.
372
X. Hou
The principle of activation analysis for long-lived radionuclides is that of conversion of the radionuclide to be determined by activation to another nuclide having a much shorter half-life and more favorable measurement characteristics preferably for γ -spectrometry. As a result, activation analysis is favorable only for some radionuclides, of which 99 Tc, 126 Sn, 129 I, 135 Cs, 226 Ra, 230 Th, 232 Th, 235 U, 238 U, 237 Np, 231 Pa and 242 Pu are among the more interesting from the view point of environmental radioactivity and nuclear waste disposal (Rosenberg, 1993; Byrne and Benedik, 1999). Compared to other techniques, activation analysis has the following advantages for the determination of long-lived radionuclides: (1) The detection limits for some nuclides are lower than those obtained by radiometric methods and most kinds of mass spectrometry; (2) High resolution γ -spectrometry allows the qualitative identification of the short-lived radionuclides produced by activation of a long-lived radionuclide present and also their quantitative determination; (3) The low density of γ -lines allows their separation easily and their higher energy reduces the matrix errors; (4) Purely instrumental methods can be used for the determination of some radionuclides, such as 232 Th and 238 U, which avoids contamination in the preparation of samples. In this article, the principle of activation analysis and its application for the determination of some long-lived radionuclides are presented.
2. Principle of activation analysis In activation analysis, the target nuclide is exposed to a flux of bombarding particles, such as neutrons, photons and charged particles. Of these, neutron activation analysis (NAA) is the most widely used, because it is the most sensitive. Neutron activation is the reaction of a nucleus of an isotope with neutrons to produce another radioactive species. Neutrons used in NAA can be produced by a nuclear reactor, a radioisotope or an accelerator, of which the nuclear reactor is the most common neutron source due to its high neutron flux density and suitable neutron energy. Neutrons can exhibit a wide range of energies from thermal neutrons with an average energy of 0.025 eV in a thermal nuclear reactor to fast neutrons of 14 MeV in an accelerator neutron generator. By bombardment of an element with neutrons, a neutron is absorbed by the target nucleus to produce a highly energetic state of the resulting nucleus containing an additional neutron, and some excess energy is immediately lost, usually by emission of a gamma ray, a proton or an alpha particle. In a sample exposed to neutrons, the type of nuclear reactions depends upon the energy of the neutrons and on the isotopes of the elements presented. The main reaction occurring with thermal neutron is the (n, γ ) reaction. In this case the highly energetic level of the produced nucleus is de-excited by emission of a gamma ray. The (n, p) and (n, α) reactions are induced by fast neutrons. In conventional NAA, the element or a radionuclide is determined by measurement of the radionuclides formed by de-excitation of the produced nucleus. However, the element can also be determined by measurement of the gamma rays emitted during the de-excitation of the produced nucleus, which is so called prompt gamma activation analysis due to the very fast de-excitation process (10−12 s). The probability of a particular reaction is expressed by the activation cross-section, σ , with units of the barn (10−28 m2 ); an isotope of element has its specific activation crosssection, and the cross-section of a particular nucleus depends on the energy of the neutron.
Activation analysis for the determination of long-lived radionuclides
373
The (n, γ ) reaction normally has a higher activation cross-section than the (n, p) and (n, α) reactions. The rate of formation of a radionuclide in neutron activation is expressed as dN ∗ (1) = σ φN, dt m where N is the number of the target nuclei, N = M NA , here m is the mass of the target isotope (in grams), M is the atomic mass, and NA is Avogadro’s number; N ∗ is the number of activated nuclei at time t, φ is the neutron flux density (neutron m−2 s−1 ), which is used to express the neutron numbers passing through unit area in unit time, which can also be considered as a product of the velocity of the neutrons and the concentration. If the nuclide formed is radioactive, it will decay with time, the decay rate of the radionuclide being dN ∗ = −λN ∗ ; dt
(2)
here λ is the decay constant of the activated nuclide, λ = Tln1/22 , T1/2 is the half-life of the activated nuclide. The production rate of the activated nuclide is expressed as dN ∗ = (σ φN ) − (λN ∗ ). dt The activity or disintegration rate (A) at the end of irradiation time ti is then A = λN ∗ = σ φN(1 − e−λti ).
(3)
(4)
The saturation activity of the activated nuclide (Am), i.e. the activity when the production of activity is equal to the decay of the activity, can be calculated as m Am = σ φN = (5) NA θ σ φ. M In the reactor, besides thermal neutrons and fast neutrons, some part of the neutrons has energy between thermal and fast neutrons, which is called the epithermal neutrons, with energy extending from 0.5 eV to 1 MeV. The neutron activation cross-section for a particular nucleus depends on the energy of the neutrons. Many nuclei absorb thermal neutrons with a crosssection which decreases linearly with increasing neutron energy (known as 1/ν absorbers). It is usual to refer to thermal cross-sections. However, not all target nuclei are 1/ν absorbers, and there are many examples of nuclei which preferentially absorb epithermal neutrons. At these higher energies the neutron cross-section is referred to as the resonance integral crosssection (I ). In this case it is important to include the resonance integral term in the calculation of the activation rate: A = N(φth σth + φepi I )(1 − e−λti );
(6)
here, φth is the thermal neutron flux density, φepi is the epithermal neutron flux density, σth is the thermal neutron cross-section, and I is the resonance integral cross-section. Taking account of the decay of activated radionuclides during the decay period (td ) and counting (tc ), the measured activity of radionuclides is calculated by A = N(φth σth + φepi I )(1 − e−λti )e−λtd (1 − e−λtc )/λ.
(7)
374
X. Hou
The actual number of events recorded by a detector for a particular radionuclide is only a fraction f of the number of decays calculated from Equation (7), because not every decay can emit a characteristic γ ray, and once a gamma ray is emitted it may not reach the detector. Taking all this into account, the simplest and most accurate way to determine an element by NAA is to irradiate and measure a comparator standard with an exactly known content of the element together with the sample. In this case, the ratio of the element content in the sample As ms = m . (ms ) to that in the comparator standard (mc ) is equal to the ratio of their activities, A c c Therefore the content of target element in the sample can be calculated by ms =
As mc . Ac
(8)
In addition, from such as single comparator, it is also possible to calculate the sensitivity of other elements by means of k-factors (deCorte and Simnits, 2003), which are experimentally determined ratios of saturation specific activities expressed in counts: k=
f θσ M∗ , f ∗θ ∗σ ∗ M
(9)
where the asterisk refers to the element of the single comparator. θ is the abundance of the isotope. The factor f is a combination of the emission probability of η and detection efficiency ε, and if the relative efficiency function of the detector is known, calibration may be based on k0 values (deCorte and Simnits, 2003): ε k = k0 ∗ , (10) ε where k0 =
ηθ σ η∗ θ ∗ σ ∗
M∗ . M
(11)
These k0 values are fundamental constants and may be found in tabulation; methods for taking into account the influence on k0 values of a difference in neutron flux spectrum have been developed. The analytical sensitivity of NAA for various elements can be predicted from Equation (7) combined with the emission probability of the gamma-rays of the radionuclide and the counting efficiency of the characteristic energy of the radionuclide. In actual analysis, the activity from other activated elements will interfere with the detection of the target element by increasing the baseline counts under its peaks, so the detection limit will be poorer than estimated. However, in the sample with known composition, the practical detection limit may be predicted in advance (Guinn et al., 1978).
3. Methodology and equipment Optimal utilization of the special qualities of NAA requires an appropriate methodology, which is adapted to the analysis of environmental samples for different elements or radionuclides. Figure 1 shows a scheme for NAA, which divides NAA into several steps, some of which are indispensable, while others are supplementary for specific purposes.
Activation analysis for the determination of long-lived radionuclides
375
Fig. 1. Procedure for neutron activation analysis.
3.1. Sampling Sampling is the first step for any analysis, but it is the most important step in any meaningful investigation (Parr, 1986). There is increasing awareness that the quality of an analytical result may often be influenced more by sample collection than by final measurements. One of the main considerations during sampling should be the representativeness of the analyzed sample with regard to the object under study. Another main consideration should be to avoid any contamination during sampling; suitable storage of the sample until analysis should also be considered to prevent losses of the elements of interest and also contamination. The selection and collection of a representative sample may need consideration of two aspects, the type of sample and the size of the selected sample. Contamination during sampling mainly comes from the tools and the container used and the sampling environment. The best materials for tools from the point of view of eliminating contamination are polyethylene, Teflon, or other plastic materials. For containers, polyethylene, Teflon and synthetic quartz are the best materials in view of contamination and absorption of trace elements on the walls of the container. The problems of sampling and sample handling have been given more attention in environmental radioactivity studies. For the determination of radionuclides, contamination from the tools may be less important than in the determination of trace elements. However, cross-contamination among samples is more important, because the difference of radionuclide concentration in different samples may be a few orders of magnitude, such as for the radionuclides in different layers of a soil or sediment core or in samples from different locations. In this case, the tool used for sampling may need to be completely cleaned before sampling a new sample. In addition, the representativeness of the sample is more important for radionuclides than for trace elements, because the distribution of radionuclides in the samples may be seriously inhomogeneous, especially for samples in which “hot particles” exist. In this case,
376
X. Hou
a large sample or special sampling and data statistics techniques are required to get an accurate result (Eriksson, 2002). In the analysis of environmental samples for radionuclides, a large sample normally needs to be collected and transported to the laboratory with or without pre-concentration. The storage of the sample before analysis is also important, as unsuitable storage may cause loss of radionuclides. The storage method is mainly specific to the radionuclide determined. For example, a water sample is normally acidified with nitric acid to pH 2 after collection, but this is not suitable for the determination of 129 I, because iodide and iodate can react to form elemental iodine, which is volatile and can be easily lost from the sample. A number of sampling protocols have been recommended, the detailed discussion on which issue can be found in many articles and books (Isaksson, 2000; Markert, 1994; Nielsen, 2000; Parr, 1986; Conklin, 2004). 3.2. Pre-separation Due to the very low concentrations of most radionuclides in environmental samples, preseparation and pre-concentration of the radionuclide of interest from the matrix and main interfering elements is normally necessary for the improvement of the detection limit. The simplest pre-concentration is lyophilization and ashing of the samples, which can be directly carried out in the irradiation container such as a quartz ampoule and aluminum foil to eliminate problems associated with sample transfer. However, for most radionuclides in environmental samples, a chemical separation has to be carried out; this is for two reasons, (1) the radionuclide of interest has to be separated from the large amount of matrix elements to reduce the radiation dose after irradiation; (2) a large sample (such as 10–100 g soil or sediment, 50–100 l seawater) cannot be directly irradiated in the irradiation channel of the nuclear reactor. The chemical separation procedure is different for different radionuclides; the individual separation procedure will be discussed in Section 5. Since NAA cannot be directly used for chemical speciation, various chemical species of the radionuclides have to be separated before the irradiation for study of the chemical speciation of radionuclides. In recent years, many methods have been developed for chemical speciation analysis of radionuclides. All these methods are based on a pre-separation using various chemical and biochemical separation techniques, such as co-precipitation, extraction, ion exchange, gel chromatography, high performance liquid chromatography, supercritical fluid extraction, electrophoretic and fractionation techniques. The chemical speciation analysis of some radionuclides is discussed in Section 5. 3.3. Sample preparation Before neutron irradiation, the sample must be transferred to an irradiation container for transport to and from the irradiation position. These containers should not contain the element of interest, nor should they add significantly to the total amount of radioactivity produced during activation. Suitable materials are low-density polyethylene for moderate irradiation and high purity aluminum and synthetic quartz for prolonged irradiation. For short-term irradiation, the sample can be wrapped in a thin polyethylene film and inserted in a polyethylene irradiation capsule; after irradiation, the sample with the polyethylene foil is directly measured as it has no significant contribution of elements and radioactivity from the polyethylene
Activation analysis for the determination of long-lived radionuclides
377
film and as there would be difficulties in quantitative removal of the sample from the irradiated film. For long-term irradiation of a sample in a reactor, it is normally required that the sample has to be dried to avoid any problem of explosion due to the high pressure produced in the container by the radiolysis of water in the samples. In addition, the removal of the water in the sample will improve the determination of the radionuclide of interest because an increased amount of sample can be irradiated and the counting geometry can be improved. Lyophilization techniques are normally used for the removal of water, although evaporation by heating can be used. The dried sample is normally wrapped in high-purity aluminum foil, which is then inserted in an aluminum container for irradiation; this is because only a short-lived radioisotope of aluminum, 28 Al (T1/2 = 2.24 min), is formed in the neutron irradiation of Al, which quickly disappears by decay, so does not contribute too much radiation dose to the operator after a few days decay. For some radionuclides such as 129 I, a synthetic quartz ampoule has to be used due to its volatilization during the irradiation and the very low impurities in quartz materials and the lower radioactivity from activated SiO2 . 3.4. Irradiation Different neutron sources are available for the activation analysis of samples. Isotopic neutron sources, such as Ra–Be and Am–Be source, which rely on the (α, n) reaction, and the 252 Cf source that is based on spontaneous fission, yield neutron spectra with a limited range of neutron energies. Accelerators can also be used to produce a particular energy of neutrons, such as 14 MeV neutrons produced by the (D, T ) reaction and 2.5 MeV neutrons by the (D, D) reaction. A miniature, sealed-tube neutron generator with a yield of 108 –1011 neutron per second has been widely used as a neutron source for NAA (Reijonen et al., 2004; Chichester and Simpson, 2004). However, due to the low neutron flux produced in these kinds of neutron sources, they are very seldom used for the NAA of radionuclides. Nuclear reactors are much more suitable for NAA of long-lived radionuclides, because they provide much higher neutron flux density (1011 –1014 n cm−2 s−1 ) with correspondingly higher sensitivities. In addition, neutrons in the reactor can be very well moderated to thermal energies, which is very suitable for NAA because of the high thermal neutron activation cross-sections of radionuclides and less interference from fast neutron reactions. For some radionuclides such as 135 Cs, 238 U, 232 Th, 231 Pa, and 237 Np, which are preferably determined by epithermal NAA, an epithermal neutron irradiation channel can be set up in some reactors by shielding thermal neutrons with cadmium (Hou et al., 1996a, 1996b). Almost all research reactors installed pneumatic transfer system for easy and automatic transportation of samples to and from the irradiation position; some of these can also quickly transfer the irradiated sample to the counting position above the detector for the determination of elements by counting short-lived radionuclides. The detection limit and analytical precision of NAA can be improved by repeated irradiation and counting of samples, so-called cyclic NAA. This is particularly useful for the elements determined by counting short-lived radionuclides (such as determination of 99 Tc by counting 100 Tc with its 15.8 s half-life). Cyclic NAA systems have been installed in some reactors (Hou, 2000).
378
X. Hou
3.5. Measurement Long-lived radionuclides are determined by measurement of the activity of the short-lived radionuclides formed during neutron irradiation, which can be carried out by counting the number of events taking place in the detector. The choice of the detector used for NAA depends on the type of decay and energies of the radiation emitted by the radionuclides formed. Some radionuclides produced by neutron activation of uranium and thorium decay by emission of neutrons, which are counted by means of a neutron detector with a filling of BF3 or 3 He gas. These analyses require that the detector be installed at the reactor site because of the short half-lives of these radionuclides and it is consequently not commonly used for radionuclides. The most common radionuclides measured in NAA decay by emission of characteristic γ rays, and the most widely used detector is a gamma semiconductor detector, most of high purity germanium (HpGe). This detector has a high energy resolution; it may separate gamma rays with energy difference of only 2–3 keV. According to the energies of the radionuclides of interest, different types of germanium detectors can be used. The planar germanium detector and coaxial HpGe with thin window are used for the measurement of low-energy gamma rays down to 10–20 keV and have the best energy resolution; in addition, a silicon (lithium) detector can also be used for the measurement of low energy gamma rays and X-rays. A normal coaxial HpGe detector is used for most radionuclides with energies of gamma rays higher than 60 keV, and a well-type HpGe detector can be used for improvement of the counting efficiency if very low-level radioactivity needs to be measured; in this kind of detector the sample sits in the middle of the germanium crystal and almost 4π geometry can be obtained. However, only small sample can be measured in this detector due to the limited size of the well of the detector. In connection with appropriate electronic equipment such as high voltage supply, preamplifier and main amplifier, and a pulse height analyzer with 4096 or 8192 channels, it is possible to determine simultaneously many different radionuclides if their gamma ray energies differ by 2–3 keV. Additional discrimination is achieved between radionuclides with different half-lives by counting the sample at different decay times. The measurement of gamma rays is based on the interactions of gamma rays with matter, such as the photoelectric effect, Compton scattering and pair production. In the photoelectric effect, the gamma ray ejects a shell electron from a germanium atom and produces a vacancy; the ejected electron has a kinetic energy equal to the energy of the gamma ray less the binding energy of the electron. It may interact with other electrons, thus causing secondary ionization, and produce more vacancies in the shells of the germanium atoms. The number of the produced photoelectrons and corresponding vacancies is determined by the energy of the gamma ray. In this way, the gamma ray energy is converted to an electric signal in the detector, which is then amplified by pre-amplifier and main amplifier. The output from the main amplifier is a peak of nearly Gaussian shape with an amplitude proportional to the gamma ray energy which enters the detector. This electric signal is finally registered by the multichannel analyzer (MCA) as a count, the number of the channel in the MCA corresponds to the gamma ray energy and the counts in a channel corresponds to the number of gamma rays with the same energy. The peak of the gamma spectrum registered in the MCA is normally also Gaussian distribution, the width of the peak, or the energy resolution being an important parameter of the detector. Besides the photoelectric effect, the pair production process results
Activation analysis for the determination of long-lived radionuclides
379
in a gamma-ray with energy less by 0.511 and 1.02 MeV than the original one. Compton scattering results in an energy continuum, which increases the baseline counts of the gamma peaks, and therefore worsens the detection limit of the radionuclide of interest. Most gammaemitting radionuclides also emit beta particles, and the interaction of the beta particle with the detector results also in a continuum of energy under the gamma spectrum. A beta absorber can be used to reduce the interference of the beta particles. The utilization of anti-Compton techniques will reduce Compton interference; the recently developed multiparameter coincidence spectrum technique has significantly improved the detection limit of the radionuclide of interest (Hatsukawa et al., 2003). 3.6. Post-irradiation separation NAA for very low-levels of radionuclides is limited by the presence of other elements in the sample. Some minor elements in the environmental samples such as Na, Cl, Br, and P contribute to a high radioactivity in the irradiated sample, and the signals of the radionuclides produced from the radionuclides of interest are overlapped by an increased Compton continuum under the gamma peaks. In order to measure the trace amount of a radionuclide of interest and improve the detection limit and analytical accuracy, it is required to eliminate these interfering elements before counting. Since the irradiated sample is radioactive, this procedure is called radiochemical or post-irradiation separation. Radiochemical separation is normally carried out by the following steps: addition of carrier and tracer, decomposition or dissolution of irradiated sample, chemical separation, and preparation of separated samples for counting. A detailed development of RNAA for individual long-lived radionuclides is presented in Section 5. Since the activated radionuclide is a new nuclide, this procedure does not create any blank problem, but may easily result in losses of the radionuclides to be determined. In order to minimize these losses, a suitable amount of carrier of the element is always added to the sample before radiochemical separation. Since the amount of carrier element is much higher than the same element existing in the original sample, it can then be used to monitor the chemical recovery of the elements determined during the radiochemical separation. In many cases, some carriers of interfering elements are also added to improve the removal of these interferences, which are so-called holdback carriers. A complete equilibrium between an inactive carrier and radioactive element is necessary to be sure that both suffer the same physical and chemical procedure, so the same chemical recoveries apply during the radiochemical separation, which is normally carried out by a comprehensive oxidation/reduction cycle. If the pre-separation is applied and carrier and tracer were added in this step, it can be directly used in the post-irradiation separation step as carrier and yield tracer. 3.7. Chemical recovery In comprehensive pre-separation and post-irradiation separation, the addition of carrier cannot entirely prevent losses of the elements of interest, even in a simple separation procedure. For accurate results, a correction for losses should therefore always be made, which can be carried out by the measurement of the chemical recovery of the determined radionuclide in the separation procedure. Chemical recovery can be measured by carrier and radiotracer. Before decomposition, carriers are added to the sample in macro-amounts compared to the element originally present
380
X. Hou
in the sample. When the carriers behave as the probed element in the sample, their chemical recoveries should be the same. Thus the chemical recovery of the carrier element is taken as the recovery of the determined radionuclide. The carrier element can be easily measured by classic analytical methods, such as gravimetry, calorimetry and titration methods, especially when a single carrier is added and separated, and the contribution from other elements is negligible. When more than one carrier is added and multielements are separated and determined, a re-activation method can be used for determination of the carrier content in the separated sample. After the measurement, the separated sample containing the activated elements and the carriers is re-activated by irradiation with neutrons, and the amount of carrier is determined by NAA. As the amount of the radionuclide originating from the sample is negligible compared to the added carrier, and the amount of added carrier is known, the chemical recovery can be calculated. For many long-lived radionuclides, such as 99 Tc, 231 Pa, 232 Th, 238 U and 237 Np, there is no stable isotope of these radionuclides, and a carrier of another element with similar chemical properties to the radionuclide of interest may be used, such as rhenium for 99 Tc. But the best way is to use other radioisotopes of the radionuclides of interest, such as 99m Tc for 99 Tc, 125 I for 129 I, and 239 Np for 237 Np as yield tracers. Using a radiotracer is a more direct method of monitoring chemical recovery. In this case, radioisotopes are added to the sample before the chemical separation. These radiotracers are not the same as those produced by the neutron activation, and can be measured simultaneously by the gamma detector due to the different energies of the gamma rays. Since the radiotracer is another indicator of the same element as the radionuclide produced from the radionuclide of interest, the chemical recovery of the radiotracer is the same as the indicator. If the radiochemical separation is carried out in a well-controlled manner, a constant chemical recovery can be assumed, and the recovery correction may be carried out without measurement of chemical recovery for every individual sample. But the constant chemical recovery has to be determined first by processing an un-irradiated sample to which irradiated carriers or other radiotracers are added using the same chemical separation procedure. In order to evaluate the precision of the correction at least 10 determinations should be made. 3.8. Analysis of gamma spectrum and NAA calculation The data acquired by a germanium detector are registered in the multichannel analyzer (MCA), and may be transferred directly or via an analyzer buffer to a computer for processing. At present, an MCA can even be made as an interface card, which can be directly inserted into a personal computer. Then the computer can control the gamma detector and the gamma spectra can be acquired and analyzed by computer software. Data acquisition software is chiefly concerned with the handling of the MCA system and its components; the programs will connect the hardware of the spectrometry system with the storage memory for the data. Critical physical parameters such as starting time, duration and dead-time of the acquisition are recorded together with the spectral data. The acquisition software can also take care of controlling sample changers and of an automatic sample transfer system for cyclic NAA, and the acquired spectrum can be stored separately or summarized. Many γ -spectrum analysis programs have been developed, which can search for γ peaks, calculate their energies and their peak areas. The special software for NAA can even identify
Activation analysis for the determination of long-lived radionuclides
381
nuclides by the energy of γ peaks and decay time, calculate the radioactivity of a radionuclide, and finally calculate the concentrations of the trace elements or radionuclides in the sample. The energy of a gamma peak can be identified by calibration of the counting system by measurement of some known γ -ray sources. By comparison with a database of γ -energy and half-life of radionuclides, the identification of the nuclides can be carried out. The calculation of the peak area is the most critical step in the NAA calculation, many techniques having been developed to calculate the peak area and to subtract the baseline under the peak, such as Gaussian shape and experimental peak shape fitting. When the peak area and the efficiency of the detector are measured, the activity of the nuclide can be easily calculated by correcting for decay and counting time. If the comparator method is used, the contents of the elements of interest can then be calculated using Equation (8) after the measurement of a comparator element standard. In NAA, overlapping peaks may sometimes occur, when the energy difference between peaks is not large enough, and a significant error may result from incorrect separation of peaks and baseline subtraction. However, with proper execution of NAA and correction of results for possible separation losses, NAA is capable of providing unbiased results with known precision for a multitude of trace elements and radionuclides. 3.9. Evaluation and quality assurance NAA has been demonstrated to be reliable and under statistical control due to the absence of unknown sources of variability. This means that is possible to predict the standard deviation of an analytical results so well that the observed and expected variation among replicate data are in agreement. The special qualities of NAA make the method one of the best for the certification of reference materials in the environmental field. The common sources of uncertainty in NAA come from irradiation, counting, spectra acquisition and analysis, blank and interference from nuclear reactions and gamma rays. The uncertainty of irradiation is contributed from the variation of neutron flux during irradiation and the inhomogeneous distribution of neutron flux in the irradiation container. The instability of the neutron flux may significantly influence the analytical results of elements determined by measuring short-lived radioisotopes, because the sample and standard are not irradiated simultaneously. This problem is normally overcome by on-line monitoring of the neutron flux, and most research reactors, especially small reactors, have installed such a system. In most cases, the uncertainty from this source is low (<0.5%). The neutron flux gradient is sometimes very significant in a big irradiation container, so that a flux monitor foil or wire of Fe, Co or Au alloy may be used to make a neutron flux correction. Uncertainty in counting mainly results from the variation of counting geometry. The uncertainty due to counting geometry can be controlled very reliably by good efficiency calibration of the detector for various source shapes; software is available for the appropriate calculation. Matrix effects can suppress gamma rays, specially low-energy gamma-rays by self-absorption, and this effect can be calibrated by measurement of a standard with a similar matrix along with sample. Uncertainty in the spectral acquisition may come from dead-time and pile-up losses of signals. A quickly varying and high counting rate often occurs in instrumental NAA of environmental materials due to high concentrations of Al, Cl and Na, and this may cause a
382
X. Hou
high and varying dead time during counting. Dead time is normally measured by the gamma spectra system, and live time is used for activity calculation. But when the dead time quickly changes during counting, the activity may be underestimated by this method. This problem can be solved by using a loss-free counting method, which estimates the number of counts lost during a dead-time interval and adds this number to the channel of the just-processed pulse, and so presents a loss-corrected spectrum. The pile-up losses are corrected by electronic or computational means built in to the counting system. For the NAA of radionuclides, since pre-separation and post-irradiation separation are normally used and most of the matrix and interfering elements and radionuclides are removed before measurement, the dead-time and pile-up losses of signals are very small and therefore negligible. However, for some radionuclides, which are measured without post-irradiation separation, such as 100 Tc from 99 Tc, the uncertainty from this source is important. Uncertainty in the spectra analysis results mainly from the evaluation of peak area and subtraction of baseline, especially for the analysis of multi-peaks. A major effort has been made to develop effective software to improve the analysis of gamma spectra. The blank problem can usually be ignored in NAA. But a blank correction will be necessary when pre-separation and sample preparation are used before irradiation. A given radionuclide can often be produced in more than one way. If the indicator nuclide used in NAA is produced from nuclides other than the radionuclide determined, then nuclear reaction interference occurs. Such interference is mainly produced by (n, 2n), (n, p) or (n, α) reactions with fast neutrons and elements with an atomic number one or two above the element to be determined. A typical example of such interference is the formation of 130 I from 133 Cs(n, α)130 I and 238 Np from 238 U(n, 2n)237 Np(n, γ )238 Np. Double and triple neutron activation reactions can also result in an interference, for example, the interference from 127 I(3n, γ )130 I and 128 Te(n, γ )129 Te(β − )129 I(n, γ )130 I to the determination of 129 I by 129 I(n, γ )130 I reaction. But interference from triple neutron reactions is normally negligible due to its very low probability. However if the amount of interfering nuclide is much higher than the radionuclides determined, such as 127 I (stable iodine) in the separated 129 I sample, this interference has to be corrected by experiment. A special type of nuclear reaction interference is caused by the presence of uranium, which yields a large number of radionuclides as a result of fission. A typical example is the interference of U to the determination of 129 I by reactions 235 U(n, f )129 I(n, γ )130 I and 235 U(n, f )130 I. Chemical separation before irradiation can prevent the interference from these reactions. As an effective method for analytical quality control, certified reference materials (CRMs) with a matrix similar to the samples, which have appropriate combinations of elements to be determined and with concentrations bracketing the range of interest, are normally analyzed to evaluate the analytical accuracy and precision. A large number of CRMs have been prepared by many countries and international organizations (IAEA, 2004), which are not only for analytical quality control of NAA, but actually for all other analytical methods. Uncertainty from sampling procedure and inhomogeneity of sample should also be considered as probably a main contribution to the total uncertainty of the analytical data, although it is not directly related to the NAA method itself. For the radionuclides, especially at very low concentration, their distribution in the samples is normally inhomogeneous and so selection of representative samples is very important to analytical quality.
Activation analysis for the determination of long-lived radionuclides
383
4. Advantage of activation analysis for the determination of long-lived radionuclides The benefit of activation analysis compared with direct radiometric methods is the improved analytical sensitivity and accuracy. For some long-lived radionuclides, activation can produce a radionuclide with a shorter half-life than the target nuclide. If the reaction cross-section and other nuclear characteristics are favorable, a considerable improvement can be obtained. When a γ -emitting nuclide is produced from a nuclide that only emits β-particles or low energy γ - and X-rays, an especially significant improvement of analytical accuracy can be obtained because the measurement of γ -radiation is more specific than that of β-radiation that does not have a discrete energy. Meanwhile, interference from other impurity radionuclides can be omitted without a tedious purification procedure. The advantage of activation analysis with respect to the radiometric method for the longlived radionuclides can be quantified in terms of an advantage factor (Byrne and Benedik, 1999). The original activity of the target radionuclide is given by A1 = Nλ1 . Activation produces a new radionuclide 2, whose activity A2 is given by A2 = N(Φth σth + Φepi I )(1 − e−λti ). Hence, the advantage factor (AF) can be calculated by the ratio of A2 /A1 : AF =
(Φth σth + Φepi I )(1 − e−λ2 ti ) A2 = . A1 λ1
Some long-lived radionuclides, their neutron activation-produced radionuclides, the relevant parameters and the calculated AF values are listed in Table 1. It shows that a very significant improvement can be obtained for these long-lived radionuclides by NAA, especially for the very long-lived nuclides such as 238 U, 232 Th and 129 I. There is of course a large literature on NAA of U, Th and 129 I. A 10−11 g (10−6 Bq) level of 238 U and 232 Th and 10−13 g Table 1 Advantage factors of NAA and relative nuclear parameters of NAA for some long-lived radionuclides (Φth = 1013 n cm−2 s−1 , Φepi = 5 × 1011 n cm−2 s−1 , Ti = 15 h) Target nuclide
T1/2 of target nuclide (years)
Product nuclide
T1/2 of product nuclide
Eγ (keV)
σth (b)
I (b)
AF
99 Tc
2.1 × 105 1.57 × 107 2.06 × 106 7.54 × 104 1.4 × 1010 4.46 × 109 2.14 × 106 8.27 × 104 3.75 × 105
100 Tc
15.8 s 12.3 h 13.2 d 25.5 h 27 d 23.5 min (2.36 d) 2.12 d 1.31 d 4.96 h
539.5 536 818.5 84 312 75 (228) 984 969 84
20 30.4 8.7 23 7.4 2.7 169 210 19.1
30 27.6 90 1010 72.4 277 660 1500 100
2.1 × 104 1.3 × 106 3.86 × 103 8.6 × 103 2.5 × 106 3.4 × 108 3.6 × 105 1.2 × 104 3.6 × 104
129 I 135 Cs 230 Th 232 Th 238 U 237 Np 231 Pa 242 Pu
130 I 136 Cs 231 Th 233 Pa 239 U (239 Np) 238 Np 232 Pa 243 Pu
384
X. Hou
(10−6 Bq) level of 129 I can be determined by radiochemical NAA, whereas assay by the radiometric method at this level is impossible. Since the detection limit of NAA is proportional to the counting efficiency, counting time and the radiation emission probability of the product radionuclide, the actual improvement of NAA may be lower than the level shown in Table 1. However, there is no doubt that NAA is a more sensitive method for the determination of 129 I, 237 Np, 232 Th and 238 U and it has already been widely applied in the analysis of environmental samples. A detailed discussion of NAA for some long-lived radionuclides is highlighted below.
5. Application of activation analysis in the determination of long-lived radionuclides in environmental samples 5.1. Iodine-129 Iodine-129 (T1/2 = 1.57×107 years) is a naturally occurring as well as a man-made long-lived radioisotope of iodine. The ratio of 129 I/127 I in the environment was estimated to be 10−12 in the pre-nuclear era (Fehn et al., 1986), but the present ratio has exceeded 10−10 because of the release from nuclear weapons testing and the peaceful utilization of nuclear energy, especially the reprocessing of spent fuel elements (Raisbeck and Yiou, 1999; Hou et al., 2000a). 129 I emits a beta particle with maximum energy of 154.4 keV, a γ -ray of 39.6 keV with an emission probability of 7.5% and X-rays of 29.46 keV (20.4%), 29.78 keV, (37.7%), 33.62 keV (6.69%). The activity of 129 I can therefore be measured by β- and γ -spectrometry. In this case, very pure 129 I must be separated from the sample. However, even with this the sensitivity of the radiometric method is too low to be used for the analysis of real environmental samples, except for special environmental samples, such as seaweed and thyroid collected from the surrounding area of La Hague reprocessing plant (Bouisset et al., 1999; Frechou et al., 2001). In order to determine environmental levels of 129 I, radiochemical neutron activation analysis (RNAA) and mass spectrometry in different forms were developed. Table 2 lists the methods used for 129 I and their detection limits, in which RNAA is the first and very widely used method for the determination of 129 I in environmental samples. Table 2 Detection limits for 129 I obtained by different methods (Rosenberg, 1993) Method
Liquid scintillation Ge-γ -spectrometry ICP-MS TIMS AMS NAA
Detection limits Number of atoms
129 I/127 I ratio
3.5 × 1013 1013 3 × 1011 107 2 × 106 108
>10−6 >10−6 >10−8 10−13 10−11 –10−10
Activation analysis for the determination of long-lived radionuclides
385
NAA of 129 I is based on the following nuclear reaction: 129 (n,γ ), σ =30 b, I =27.6 b 130
I −−−−−−−−−−−−−→
β − , 12.3 h
I −−−−−−→
130
Xe.
The irradiated 129 I absorbs a neutron, in the meantime emits a photon and becomes 130 I with a half-life of 12.3 h, which decays by emitting beta particles and gamma rays with energies of 418.0 keV (34.2%), 539.1 keV (99.0%), 668.5 keV (96.3%), and 739.5 keV (82.3%). There are some interfering nuclear reactions from some nuclides existing in the samples other than iodine isotopes through which the production of 130 I in the samples can be simulated, these nuclides including 235 U, 128 Te and 133 Cs. Table 3 lists these reactions and the magnitude of various interferences calculated under the experimental condition used in Risø National Laboratory, Denmark (Hou et al., 1999). Because of the extremely low concentration of 129 I in the environmental samples (10−17 –10−11 g g−1 ), these interfering nuclides have to be separated before irradiation to prevent their nuclear interference which would generate spurious results. The radioactivity produced from the activation products of matrix elements of the sample, such as 24 Na and 82 Br, is more than 10 orders of magnitude higher than that of 130 I, making the direct measurement of 130 I impossible after irradiation. In addition, because of the low concentration of 129 I in the samples, a large quantity of sample has to be analyzed, such as 20–50 g of seaweed or grass, 5–10 g of soil or sediment, 10–50 l of seawater or freshwater (Hou et al., 1999, 2000a, 2000b, 2001, 2002, 2003a, 2003b; Hou, 2004). If no separation and pre-concentration of iodine is carried out, no such large irradiation channel is available in the reactor, and even if it could be irradiated, a very high radioactivity from the matrix elements would make the post-separation of iodine very difficult. Thus, a pre-irradiation chemical separation of iodine from the samples is necessary. NAA cannot only determine 129 I, but also 127 I. 127 I is normally determined by measurement of 128 I produced by the 127 I(n, γ )128 I reaction. However, it cannot be simultaneously determined with 129 I, because the half-life of 128 I (25 min) is much shorter than 130 I, and post-irradiation separation is normally carried out some hours after irradiation to reduce the radiation dose mainly from the irradiation container (Al). Therefore, the fast neutron reaction 127 I(n, 2n)126 I is normally used for the simultaneous determination of 127 I and 129 I. In this case, the longer half-life radionuclide 126 I (13.1 days) is separated from the irradiated sample and measured with 130 I. Since bromine is chemically similar to iodine and the γ -rays of 82 Br produced by the reaction of 81 Br(n, γ )82 Br interTable 3 Nuclear interferences in the determination of 129 I (Hou et al., 1999) (Φth = 4 × 1013 n cm−2 s−1 , Φepi = 2 × 1011 n cm−2 s−1 , Φf = 1 × 1011 n cm−2 s−1 ; irradiation time: 10 h; activities at the end of the irradiation are given) Nuclear reaction
Induced activities of iodine isotopes
Amount of element required to produce 130 I equivalent to 10−12 g 129 I
235 U(n, f )129 I(n, γ )130 I and 235 U(n, f )130 I
4.6 × 105 Bq 130 I/g U 5.2 × 105 Bq 130 I/g Te 6.4 × 102 Bq 130 I/g Cs 7.2 Bq 130 I/g 127 I
2.6 µg 2.3 µg 1.9 mg 170 mg
128 Te(n, γ )129 Te(β − )129 I(n, γ )130 I 133 Cs(n, α)130 I 127 I(n, γ )128 I(n, γ )129 I(n, γ )130 I
386
X. Hou
feres with the measurement of 130 I, pre-irradiation separation cannot completely separate it and a post-irradiation chemical purification has to be directed to providing the necessary decontamination with respect to this nuclide. The analytical procedure for the determination of 129 I by RNAA therefore involves four steps to minimize the effects of the above-mentioned nuclear and chemical interferences; they are pre-irradiation separation of iodine, neutron activation in a nuclear reactor, post-irradiation purification of iodine, and measurement of 130 I by γ -spectrometry. A typical procedure used for the pre-separation of iodine from solid samples (vegetation, soil, sediment, and tissues) is shown in Figure 2 (Muramatsu et al., 1984; Hou et al., 1999, 2000a, 2000b, 2002, 2003a, 2003b; Hou, 2004). In this method, the sample is first mixed with alkali solution, such as KOH, the mixed sample is ashed or fused at 650 ◦ C. Since iodine is not easily converted to volatile elemental iodine in alkali media at high temperature, the sample is decomposed under this condition without loss of iodine. Our experimental results showed that the recovery of iodine in ashing or fusion procedures is higher than 90%. Besides ashing and fusion, a combustion method is also widely used for the separation of iodine from solid samples (Aumann, 1981; Aumann et al., 1985, 1987; Muramatsu et al., 1985, 1988; Handl et al., 1990, 1993). Figure 3 shows a diagram of the combustion apparatus. In this method, the sample is combusted at higher temperature (>800 ◦ C), the released iodine, mainly as I2 , is trapped with alkali solution (KOH) or active charcoal. The trapped iodine is extracted with CCl4 after being acidified and oxidized to I2 , and then back-extracted with H2 SO3 . After conversion of separated iodine to MgI2 , it is subjected to neutron irradiation. Iodine trapped in active charcoal is directly irradiated in reactor. For aqueous samples including milk and urine, iodine can be separated by an anion exchange method. Iodine is first converted to iodide and then absorbed by an anion exchange resin (AG1) and separated from the matrix elements, the iodine absorbed on the resin being eluted by nitrate solution, and concentrated by extraction with CCl4 (Hou et al., 2001, 2003a; Parry et al., 1995). For large water samples, the anion exchange method can also be used; the author has analyzed 50 l of seawater for 129 I by this method (Hou, 2004). In this case, the anion exchange resin was converted to the NO− 3 form to enlarge the exchange capacity of the resin for iodide. In addition, a big column (∅ 2 × 40 cm) and repeat loading and elution have to be applied (Hou et al., 2001; Hou, 2004). For small volume water samples (1–2 l), a direct extraction using CCl4 after conversion of iodine to elemental iodine using NaNO2 in acidic solution can be used to separate iodine from the water. However, for seawater samples, stable iodine carrier has to be added to improve the extraction recovery of iodine (Hou et al., 1999). In the pre-separation procedure, most of the matrix and interfering elements, such as uranium, tellurium, and cesium, can be removed; the nuclear interference from these elements can be eliminated. As a consequence, no 134 Cs, the neutron activation product of 133 Cs and 133 I, one of the fission products of 235 U, were observed in the spectra (Hou et al., 1999). The main spectral interference for NAA of 129 I is from the gamma-rays of 126 I (388.6 keV (34.1%), 666.3 keV (33.1%), 753.8 keV (4.3%)). Due to the low 129 I/127 I ratio (10−10 –10−6 ), the radioactivity of 126 I will be higher than 130 I, although the cross-section of 129 I(n, γ )130 I (30 b) is 30 times higher than that of 127 I(n, 2n)126 I (0.9 b). For suppression of the radioactivity of 126 I, the sample should be irradiated in a thermal neutron irradiation channel, because 127 I(n, 2n)126 I is a fast neutron reaction. Since iodine is a volatile element and iodide can be oxidized to elemental iodine during neutron irradiation, the separated iodine as MgI2
Activation analysis for the determination of long-lived radionuclides
387
Fig. 2. Pre-irradiation separation procedure of iodine from environmental samples (Hou et al., 1999).
has therefore to be sealed into a tight container, from which elemental iodine cannot be lost. A quartz ampoule is normally used for this purpose, because elemental iodine does not absorb onto and penetrate the wall, and the quartz ampoule can also stand high pressure and contains low-level impurities. For iodine absorbed on activated charcoal, it is also sealed in a quartz ampoule with the charcoal to prevent any loss of iodine during irradiation, although iodine absorbed in charcoal is less volatile than iodide powder. The irradiated iodine sample exists
388
X. Hou
Fig. 3. Diagram of sample combustion apparatus for separation of iodine.
as elemental iodine to some extent in the ampoule; the ampoule should thus be opened under sulfite solution to prevent the loss of I2 from the ampoule. In this case, I2 can be quickly converted to iodide by the sulfite and retained in the solution. In addition to the nuclides mentioned above, stable iodine, 127 I, can also cause interference in the determination of 129 I by three continuous neutron capture reactions, 127 I(3n, γ )130 I. This interference varies as the square of the neutron flux and increases with the length of the irradiation time. By simultaneous determination of 127 I concentration via the fast reaction 127 I(n, 2n)126 I, this interference can be corrected. But this interference limits the analysis of samples with very low 129 I/127 I ratio. However, for normal environmental samples, this interference is very small. A simple post-irradiation separation method (Figure 4) is usually used for the further purification of 130 I, which is based on the fact that I2 can be easily extracted by CCl4 , and iodide back-extracted by sulfite is precipitated as PdI2 (Hou et al., 1999). By this procedure with pre-separation, a spectrum of pure isotopes of iodine can be obtained. For the measurement of chemical yield, some radioisotopes of iodine can be used as tracers. Because 129 I can be simultaneously produced during the production of 131 I, the 131 I tracer contains some amount of 129 I and the addition of 131 I as a tracer will interfere with the determination of 129 I. It was reported that the isotope ratio 129 I/131 I ranges from 2.6×10−13 g/kBq 131 I to 1.8×10−12 g 129 I/kBq 131 I in 131 I tracer solution (Chao et al., 1998; Hou et al., 1999). Therefore, 125 I is usually used for this purpose because no 129 I is produced in the production of 125 I. A chemical yield of 70–90% in the RNAA of 129 I was obtained in the author’s laboratory, and the detection limit of 129 I is 10−13 g (5×108 atom), while 129 I/127 I ratios of 10−10 –10−6 have been found in environmental samples from different locations (Hou et al., 1999, 2000a, 2000b, 2001, 2002, 2003a, 2003b; Hou, 2004). Pre-irradiation chemical separation and post-irradiation purification of iodine can eliminate most nuclear and chemical interferences; a HpGe gamma detector is good enough for the measurement of 130 I in normal environmental samples. However, interference from some iodine isotopes, such as 126 I, in the measurement of 130 I cannot be eliminated by chemical separation. When the ratio of 129 I/127 I is low, or the fast neutron composition is high in the
Activation analysis for the determination of long-lived radionuclides
389
Fig. 4. Post-irradiation purification procedure for the measurement of 130 I (Hou et al., 1999).
irradiation site, the interference from Compton background of 126 I will be very serious; it will make the detection limit of 129 I worse. To eliminate this interference and improve the detection limit of 129 I, a coincidence measurement technique, such as 3γ -ray coincidence, sum coincidence or both and isotopic separation after irradiation can be applied (Rook et al., 1975; Aumann, 1981; Aumann et al., 1987). With chemical separation before irradiation, the chemical speciation of 129 I can also be analyzed by NAA. The author has developed a NAA method for the determination of 129 I in iodide and iodate forms in seawater. In this method, seawater is passed through an anion exchange column, iodide is absorbed on the column, while iodate and organic iodine pass through the column. Iodide on the column is then eluted by KNO3 solution. The iodate in the
390
X. Hou Table 4 Detection limits of three methods used for the determination of 237 Np (Rosenberg, 1993) Method
Detection limit (mBq)
α-Spectrometry RIMS ICP-MS NAA
0.1 0.1 0.02 0.02
effluent is then reduced to iodide by NaHSO3 after acidification, and then passes through a new anion exchange column. After eluting using KNO3 solution, iodate can be separated. The separated iodide and iodate are purified and analyzed using the same method as that for total iodine (Hou et al., 2001). A large literature on the NAA of 129 I in various environmental samples is available (Edwards, 1962; Handl et al., 1990, 1993; Aumann, 1981; Aumann et al., 1985, 1987; Hou et al., 1999, 2000a, 2000b, 2001, 2002, 2003a, 2003b; Hou, 2004; Raisbeck and Yiou, 1999; Chao et al., 1998). 5.2. Neptunium-237 237 Np
is an α-emitting long-lived radionuclide with a specific activity of 26 mBq/ng. In environmental samples, it originates from nuclear weapons testing and reprocessing of nuclear fuel. In addition to α-spectrometry, 237 Np has been determined by NAA, RIMS, AMS and recently ICP-MS. Rosenberg (1993) summarized three analytical methods for the determination of 237 Np, the detection limits of these being shown in Table 4. The following nuclear reaction is used for the NAA of 237 Np: 237
(n,γ ), σ =169 b, I =600 b 238
Np −−−−−−−−−−−−−−→
β − , 2.1 d
Np −−−−−−→
238
Pu.
Uranium may interfere with the analysis by the following reactions: β−
238
U(n, 2n)237 U −→
235
U(n, γ )236 U(n, γ )237 U −→
237
Np(n, γ )238 Np, β−
237
Np(n, γ )238 Np.
Although these interferences are said to be small, uranium has to be separated from the sample prior to the neutron irradiation. In addition, the pre-irradiation separation can also be used to concentrate Np from a large sample, and to avoid working with high activity. Many chemical procedures have been developed for the separation of Np from matrix elements and interfering elements, Figure 5 showing a procedure used in the author’s laboratory (Chen et al., 2001a). Three techniques are mainly used to separate Np from matrix and interfering elements, i.e. hydroxide precipitation, solvent extraction and ion exchange (Ruf and Friedrich, 1978; Byrne, 1986; May et al., 1987; Germain et al., 1987; Germain and Pinte, 1990; Kim et al., 1988; Hursthouse et al., 1992; Jha and Bhat, 1994). In these, adjustment of the valence state of Np 2+ plays an important role. Neptunium can exist as Np3+ , Np4+ , NpO+ 2 , and NpO2 , of which
Activation analysis for the determination of long-lived radionuclides
391
Fig. 5. Pre-separation of Np for NAA (Chen et al., 2001a).
NpO+ 2 is the most stable form in the environment. For solid samples, such as soil, sediment, plant and animal tissue, the samples have to be decomposed. Because Np is normally stable at high temperature, the sample is first ashed to decompose the organic component in the sample, and the ashed sample is then digested with acids, such as concentrated HNO3 or aqua regia to release Np from the sample. The hydroxide co-precipitate method is widely used to sepa-
392
X. Hou
rated actinides from the matrix. However, NpO+ 2 cannot be completely co-precipitated with Fe(OH)3 so it has therefore to be reduced to Np4+ before hydroxide precipitation. Fe2+ is a 2+ 4+ 2+ is not stable in the solution, very good agent to reduce NpO+ 2 and NpO2 to Np . Since Fe 3+ Fe and another reductant, such as hydrazine, hydroxylamine and sulfite, are normally used; in this case Fe3+ is first reduced to Fe2+ by the reductant, which then reduces the oxidized forms of Np to Np4+ . Solvent extraction using tri-n-octylphosphine oxide (TOPO) in toluene (Kalmykov et al., 2004) or thenoyltrifluoroacetone (TTA) in oxylene (Raghavan et al., 1976; Ruf and Friedrich, 1978; Hursthouse et al., 1992; Chen et al., 2001a) is used to separate Np from metals, uranium, thorium and other transuranics. In this step, the oxidation state of Np 4+ can also needs to be adjusted to Np4+ , because NpO+ 2 is not extracted in these solvents. Np be extracted to TTA in a low concentration of HNO3 (1–2 mol/l), as well as a high concentration of HCl (8–12 mol/l), and then back-extracted by a high concentration of HNO3 (8 mol/l) 2− and 2 mol/l HCl, respectively. In this step, NpCl− 2 , UO2 Cl4 , Th, rare earth elements and Fe will be extracted to the organic phase, while Pu as Pu3+ , Am3+ and Cm3+ remain in the aqueous phase. The extracted Np in the organic phase is then back-extracted to the aqueous phase, while U, Th, Fe and other elements remain in the organic phase. However, due to insufficient decontamination from some elements, an anion exchange method follows after solvent − extraction. Based on the formation of a complex anion of Np4+ with NO− 3 and Cl at high 4+ concentration of HNO3 and HCl media, Np is first adjusted to Np and absorbed onto an anion exchange column, while uranium, which does not form a stable complex anion in HNO3 solution, passes through the column and is washed from the column by 8–10 mol/l HNO3 , and Th, which does not form a complex anion in HCl solution, is removed from the column by washing with 12 mol/l HCl. The Np remaining on the column is finally eluted from the column by 4–6 mol/l HCl with NH2 OH·HCl, or 4 mol/l HCl with 0.1 mol/l HF. Extraction chromatography has also been used to separate Np from interfering elements. The separated Np has to be dried for neutron irradiation to reduce the volume of sample and to avoid the problem of radiolysis of water during irradiation, which may result in a broken irradiation container. The separated Np solution is therefore transferred to a quartz ampoule and dried by heating. Due to there being no loss during irradiation, the separated sample can be also dropped on to a pure Al foil and heat-dried. The foil with sample is then wrapped and put into an Al irradiation can for neutron irradiation. In this case, the Al foil has to be highly pure to prevent any contamination of sample. Due to the relatively longer half-life of 238 Np (2.1 days), a longer irradiation time (20–70 h) is normally used for improvement of analytical sensitivity (Ruf and Friedrich, 1978; Germain et al., 1987; Germain and Pinte, 1990; Jha and Bhat, 1994; May et al., 1987). Since analytical sensitivity for 237 Np is proportional to the neutron flux, a high flux neutron irradiation of samples is always beneficial to the determination of low-level 237 Np. The higher resonance integral cross-section (600 b) compared to the thermal the neutron cross-section (169 b) of the reaction makes epithermal neutron irradiation useful for suppression of the radioactivity from interfering elements such as Na, Sc, and Cr and improvement of the detection limit of Np if no post-irradiation separation is carried out. The relative comparison method is usually used to calculate the concentration of 237 Np and so a comparison standard of 237 Np has to be prepared and irradiated with the samples in the same irradiation channel. However, if the k0 method is applied for the calculation of 237 Np content, no such 237 Np standard is necessary, but in this case, some parameters, such as the
Activation analysis for the determination of long-lived radionuclides
393
effective resonance energy and k0 value, have to be measured before the analysis. A k0 method for 237 Np determination has been reported (Piccot et al., 2002). A post-irradiation chemical separation is also necessary in order to achieve adequate decontamination from other gamma and beta emitters produced by activation of residual impurity elements, the main being 82 Br, 24 Na, with traces of 60 Co, and 59 Fe. A similar procedure as that used for pre-separation can also be used for the post-irradiation separation, such as TTA extraction (Byrne, 1986) and anion exchange chromatography (Germain and Pinte, 1990; May et al., 1987). In addition, extraction chromatography using Eichrom UTEVA resin has been used for post-irradiation separation of Np (Kalmykov et al., 2004). Since the most important interference in the irradiated sample comes from 24 Na and 82 Br, hydrated antimony pentoxide (HAP) and evaporation methods have also been applied for post-irradiation separation of Np. The HAP method is a well known technique for removal of 24 Na from the irradiated sample, whereby the sample is loaded on to a HAP column, 24 Na will be absorbed on the column, while most of the other elements will pass through the column and are collected in the effluent. Bromine is a volatile element which can be removed from the sample by oxidizing Br to Br2 using NaClO4 and evaporation (Jha and Bhat, 1994). The separated and purified Np sample is then counted using a HpGe detector. 238 Np can be identified from its major gamma rays with energies of 984.4 keV (24.0%), 1028.5 keV (16.1%) and 1026.0 keV (9.5%), the concentration of 237 Np in the sample then being calculated by comparing with a 237 Np standard. 239 Np (T 1/2 = 2.35 days) is normally used as a tracer for the measurement of chemical yield. It emits gamma rays with energies of 106.1 keV (21.0%), 103.7 keV (18.0%), 277.6 keV (12.1%) and 99.5 keV (11.0%), and therefore can be simultaneously measured with 238 Np. 239 Np can be prepared by irradiation of uranium (Germain et al., 1987), as well as from 243 Am as a decay product. In the first method, the depleted uranium is irradiated with epithermal neutrons and directly added to the sample before chemical separation (Ruf and Friedrich, 1978). In the second method, an 243 Am–239 Np generator can be prepared, and 239 Np can be routinely obtained from this generator. The latter method is better than that from irradiation of uranium, especially for the analysis of low-level samples, because trace amounts of 237 Np and 238 Np can be produced in the irradiation of uranium, while addition of neutron-irradiated depleted uranium as a 239 Np tracer can introduce small amounts of 237 Np and 238 Np. 239 Np tracer can be added before sample treatment and used for the chemical yield of the whole procedure including both pre-separation and post-irradiation separation. In this case a higher level of 239 Np needs to be added to obtain good counting statistics, because a relatively long time is needed for pre-separation, irradiation, cooling and post-irradiation separation (5–10 days). 239 Np is also added individually for pre-separation and post-irradiation procedures. In this case, the contribution of 239 Np added in the pre-separation procedure to the counts of 239 Np added in the post-irradiation procedure should be considered. The reported chemical yield in the radiochemical NAA is 50–90%, and a detection limit of 237 Np as low as 0.01 mBq (0.5 fg) in environmental and biological samples has been reported (Germain et al., 1987; Germain and Pinte, 1990). NAA has been used for the determination of 237 Np in various environmental samples, such as soil, sediment, seawater, seaweed and fish by many authors (May et al., 1987; Germain et al., 1987; Germain and Pinte, 1990; Hursthouse et al., 1992; Jha and Bhat, 1994; Ruf and
394
X. Hou
Friedrich, 1978; Kim et al., 1988; Byrne, 1986; Assinder, 1999) and a good agreement was obtained by NAA with α-spectrometry and ICP-MS. 5.3. Technetium-99 Technetium-99 decays by low energy beta emission with a specific activity of 0.625 mBq/pg. The concentration of 99 Tc in uncontaminated environmental samples is of the order of a few mBq/g or less. In addition to radiometric methods, many other methods, such as atomic absorption spectrometry, NAA, ICP-MS and AMS can be used for the determination of 99 Tc. The NAA of 99 Tc is based on two reactions: β − , 15.8 s 100
99
Tc(n, γ )100 Tc −−−−−−→
99
Tc(n, n )99m Tc −−−−−−→
Ru
and IT, 6.0 h
100 Tc,
99
Tc.
When counting which has a very short half-life (15.8 s), 99 Tc has to be separated from the sample prior to irradiation; in addition, a fast sample transfer system has to be used to shorten the decay time. NAA is theoretically a more sensitive method for 99 Tc (Table 1), and a detection limit of 5 × 10−11 g (or 2.5 mBq) has been obtained using pre-separation and RNAA (Foti et al., 1972a, 1972b). Because of the quick decay of the product nuclide (100 Tc) and some interference from the activation products, such as 19 O, 28 Al, 38 Cl, 24 Na, 80 Br, 82 Br, and 41 Ar (Goerner et al., 1988), Tc has to be separated from the matrix and other elements and radionuclides before and after irradiation to obtain a good detection limit. Many procedures have been developed for the separation of technetium from matrix and other elements. They are mainly based on precipitation of Fe(OH)3 to remove metals, coprecipitation with tetraphenylarsonium perrhenate, anion exchange, cation exchange, solvent extraction with cyclohexanone and methylothyl ketone, and extraction chromatography using Eichrom TEVA (Foti et al., 1972a, 1972b; Ikeda et al., 1989; Goerner et al., 1988; Houdek et al., 1979; Mincher and Baker, 1990). Most separation procedures used in the determination of 99 Tc by radiometric and ICP-MS methods can also be used for the pre-separation of Tc from matrix; Figure 6 shows a procedure used in the author’s laboratory (Chen et al., 1990; KeithRoach et al., 2002). 99 Tc in TcO− 4 form has a high absorption capacity on anion exchange resins and the 99 Tc absorbed on the anion exchange resin can therefore be directly irradiated and determined afterwards (Mincher and Baker, 1990). In this case, the anion exchanger has to be purified to remove all other elements and transferred to the NO− 3 form to avoid interference 38 from Cl. This can be carried out by washing the column with 1 mol/l HNO3 and NH4 OH solution after loading the 99 Tc solution on the column. If the irradiated sample is measured directly, the interference of 41 Ar, 19 O, and 18 F from the activation products of Ar2 , O2 , and F, which exist in the atmosphere and the container, is the main problem causing worsening of the detection limit of 99 Tc. This problem can be overcome by sealing the sample in a polyethylene vial under He2 . The cyclic activation model is useful for the improvement of the detection limit of 99 Tc, if no post-irradiation is applied. Post-irradiation separation can improve the detection limit of 99 Tc by reducing the background. Due to the very short half-life of 100 Tc, a fast separation procedure must be applied,
Activation analysis for the determination of long-lived radionuclides
395
Fig. 6. Pre-separation of 99 Tc for activation analysis.
otherwise the quick decay of 100 Tc will counteract the benefit of reduction of the background. Foti et al. (1972a) have developed a fast radiochemical separation procedure based on the precipitation of Fe(OH)3 to remove metals, co-precipitation of Tc with tetraphenylarsonium perrhenate, and evaporation of tetraphenylarsonium perrhenate by heating. The whole separation procedure can be completed in 40–45 s. The 100 Tc formed is normally measured by HpGe detector using its gamma rays with energies of 539.5 keV (7%) and 590.8 keV (5.7%). The 539.5 keV gamma rays are normally used to calculate the content of 99 Tc due to its relatively higher emission probability and because of the possible interference at 590.8 keV from 101 Mo (14.6 min, 590.1 keV). Due to the low gamma ray emission probability of 100 Tc, the measurement of 100 Tc by low-background beta counter has been carried out to improve the counting efficiency (Foti et al., 1972a). In this case, post-irradiation separation has to be completed to get a high purity 100 Tc source; the purity of the separated 100 Tc can be appraised by the decay curve of the prepared source. Since a short irradiation time (<1 min) is normally used, the nuclear interference from the reaction of
396
X. Hou
235 U(n, f )99 Tc(n, γ )100 Tc,
and gamma ray interference from 101 Mo is normally not serious, especially with a pre-separation procedure for technetium. 95m Tc (61 days) and 99m Tc (6 h) can be used as tracers for chemical yield monitoring. 95m Tc is produced by the Mo(p, xn) reaction with cyclotron-produced proton irradiation. The Mo matrix and other impurities must be removed before addition to the sample. 99m Tc can be easily obtained from a 99 Mo–99m Tc generator. However, 99m Tc decays to 99 Tc and special attention should be given to avoiding the contamination of 99 Tc from this tracer, especially for the analysis of low-level samples. For this purpose, a fresh eluate of 99m Tc from the generator with short in-growth time should always be used. In addition, 99 Mo in the eluate should be removed by passing the eluate through an extra Al2 O3 column. Our experience shows that 99 Tc in 40 kBq of 99m Tc tracer is lower than 0.01 mBq (<10−14 g), which is well below the detection limit of 99 Tc by most analytical methods. Compared to 100 Tc, 99m Tc, a neutron activation product of 99 Tc by the (n, n ) reaction, has a longer half-life, and this makes the post-irradiation separation easier. However, the low neutron activation cross-section (0.24 b) of the 99 Tc(n, n )99m Tc reaction limits its analytical sensitivity (Ikeda et al., 1989). The reported detection limit of this method is more than 1 Bq (>1 ng), which is even higher than for the liquid scintillation counting method, and is not good enough for the analysis of environmental samples (Goerner et al., 1988; Ikeda et al., 1989). In addition to NAA, photon activation analysis can be used for the determination 99 Tc by the nuclear reaction 99 Tc(γ , γ )99m Tc. The detection limit of this method for 99 Tc was reported to be 0.6 Bq (or 1 ng) (Sekine et al., 1989; Yagi et al., 1991), which is also not good enough for the analysis of environmental samples. Thus far, activation analysis was mainly used for the analysis of samples with a high 99 Tc concentration, such as radioactive waste. The application of this method for the analysis of real environmental samples is limited (Foti et al., 1972b) because (1) the best detection limit of activation analysis (2.5 mBq) is similar to that of the radiometric method such as low background beta counting (Chen et al., 1990); (2) a reactor is required for the neutron irradiation; (3) an especially quick post-irradiation separation method and appropriate apparatus are needed to achieve a low detection limit, (4) the detection limit of the newly developed ICP-MS method (3 × 10−14 g) is much lower than NAA, and ICP-MS equipment becomes ever more readily available. 5.4. Cesium-135 (T1/2 = 2.3 × 106 years), a soft β-emitter with a maximum beta energy of 205 keV, is produced by nuclear fission as one of the main products with a total fission yield of 6.5% on mass 135, and an individual fission yield of 0.55%. 137 Cs (T1/2 = 30.07 years), another βradioisotopes of Cs with maximum energies of 514 keV (94.4%) and 1176 keV (5.6%), is also a fission product of 235 U with a similar fission yield (6.2%). The atom ratio of 137 Cs/135 Cs produced in a reactor varies with the neutron flux since the precursor of 135 Cs, 135 Xe, has a very high neutron capture cross-section (2.6 × 106 b). 135 Cs, as well as 137 Cs, in the environment were released by nuclear weapons testing in the 1960’s and by nuclear industries, such as reprocessing plants and nuclear power plants, and nuclear accidents. The reported atomic ratios of 135 Cs/137 Cs are 0.1–0.4 in discharge from reprocessing plants, 0.7–1.6 in ra-
135 Cs
Activation analysis for the determination of long-lived radionuclides
397
dioactive waste (Chao and Tseng, 1996), 1–2.8 in environmental samples (Lee et al., 1993), and 2.7 in sediment samples (Karam et al., 2002). This corresponds to a 135 Cs/137 Cs activity ratio of 1–40 × 10−4 . The determination of 135 Cs by beta counting is therefore impossible due to the serious interference from 137 Cs. 135 Cs can also be measured by gammaspectrometry by counting its 268.2 keV γ -ray (15.5%). However, the low specific activity of 135 Cs (43 mBq/ng) and the low counting efficiency of gamma-spectrometry for 135 Cs makes its determination by radiometric methods very difficult. Some other methods can be used for determination of 135 Cs, but only NAA and mass-spectrometry have been applied for the analysis of real samples (Lee et al., 1993; Chao and Tseng, 1996; Stamm, 1973; Song et al., 2001; Karam et al., 2002; Moreno et al., 1999; Pibida et al., 2004). The NAA of 135 Cs is based on the reaction 135
(n,γ )
Cs −−−−−−→
136
β, 13.16 d
Cs −−−−−−→
131
Ba.
Because of the very low concentration of 135 Cs in the environmental samples, a preirradiation separation has to be carried out to separate Cs from the large amount of sample. The most used method for the separation and concentration of Cs in water samples is based on its specific absorption on ammonium phosphomolybdate (AMP) and copper ferrocyanide (CFC). However, before irradiation, Cs has to be separated from AMP and CFC to reduce the radioactivity of 32 P, 99 Mo, 59 Fe, 64 Cu. Cs concentrated in AMP can be released by dissolution of AMP with dilute KOH, and the released Cs is then absorbed on a cation exchange resin for neutron irradiation (Chao and Tseng, 1996). Cs concentrated in CFC can be leached out by 5–7 mol/l HNO3 , the released Cs can be purified by cation exchange chromatography, and the separated Cs is then transferred to a quartz ampoule and dried for neutron irradiation. For the separation of Cs from soil and sediment, the sample must be completely decomposed, because Cs in soil and sediment is mainly combined with the clay fraction, which cannot be leached out by acid (Hou et al., 2003b). The released Cs can be separated from the matrix and other elements by cation exchange chromatography. Due to the relatively longer half-life of 136 Cs, a long irradiation time and high neutron flux should be applied for the analysis of low-level samples. For improvement of the detection limit, a post-irradiation separation of Cs can be carried out. The Cs on the irradiated cation exchange resin can be eluted by 5–8 mol/l HNO3 , and the released Cs is then separated by AMP precipitation or cation exchange chromatography (Chao and Tseng, 1996). The separated Cs sample is measured by a HpGe detector, the content of 135 Cs being calculated by counting the radioactivity of 136 Cs using its main gamma rays with energies of 818.5 keV (100%) and 1048.1 keV (80.3%). The interference in the measurement of 136 Cs is mainly from 134 Cs (2.06 yr), and 86 Rb (18.65 days). 134 Cs is produced by the neutron activation of stable 133 Cs (isotopic abundance of 100%). Due to the high atomic ratio of 133 Cs/135 Cs in environmental samples (>109 ), the activity ratio of 134 Cs/135 Cs in the irradiated sample is normally higher than 106 with 2 days irradiation. 86 Rb is the neutron-activation product of stable 85 Rb (abundance of 72.2%) by the reaction 85 Rb(n, γ )86 Rb and emits a γ -ray of 1077 keV. This interference is very important for the analysis of environmental samples, because the chemical properties of Rb and Cs are very similar. Rb can also be absorbed by AMP and CFC, and concentrated together with Cs, and the concentration of Rb is higher than Cs in most environmental samples. The Compton background produced by 86 Rb will overlap the γ -lines of 135 Cs (818.5 and 1048 keV). Cation
398
X. Hou
exchange chromatography is a promising method for the separation of these two elements (Massart, 1971). The 137 Cs that is normally present in a sample can be used as chemical yield tracer. The whole chemical recovery of Cs can be calculated by measuring the 137 Cs in the original sample and in the separated sample. The detection limit of NAA for 135 Cs depends on the concentrations of the other Cs isotopes, especially 133 Cs and 137 Cs. A detection limit as low as 10−4 Bq (10−12 g) 135 Cs has been reported for a sample with a ratio of 133 Cs:135 Cs:137 Cs = 1:1:1 (Chao and Tseng, 1996) and irradiated at 7×1012 n cm−2 s−1 for 24 h. This detection limit can be improved by increasing the neutron flux and extending the irradiation time. To date, only two papers on NAA of 135 Cs are available for the analysis of radioactive waste and sodium coolant (Chao and Tseng, 1996; Stamm, 1973), but it is a potentially useful method for the determination of 135 Cs in environmental samples by developing a suitable pre-separation and concentration procedure. 5.5. Uranium-235, 238 and thorium-232 Uranium-235, 238 and thorium-232 are important natural radionuclides of long half-lives (>108 yr). Radiometric methods based on α-spectrometry are time-consuming and need large samples. Although recently the ICP-MS method has gained a high sensitivity for the determination of these radionuclides, and can significantly reduce the amount of sample used, the sample has to be decomposed before measurement, and the reagent, laboratory and instrumental blanks limit the detection limits. NAA is an excellent method for the determination of low-level 238 U, 235 U and 232 Th due to its high sensitivity and virtual freedom from blank problems when either instrumental neutron activation analysis (INAA) or RNAA is used. The NAA of 238 U and 232 Th is based on the following nuclear reactions: 238 232 239 Np
(n,γ ), σ =2.7 b, I =277 b
U −−−−−−−−−−−−−−−→ (n,γ ), σ =7.4 b, I =72 b
Th −−−−−−−−−−−−−−−→ 233 Pa
239
β − , 23.5 min
U −−−−−−−→
233
β−,
22.3 min
239
Th −−−−−−−→
β − , 2.36 d
Np −−−−−−−→
233
β−,
27.0 d
239
Pa −−−−−−−→ 238 U
Pu,
233
U.
and are usually used for the determination of and 232 Th, respec24 38 tively, because the interferences of Na (15 h), Cl (37.2 min) and other short-lived nuclides produced from the matrix elements make the measurement of 239 U and 233 Th more difficult in instrumental NAA. The reported detection limits of INAA for 238 U and 232 Th are 10−8 g, and it can therefore be used for the determination of 238 U and 232 Th in environmental samples, such as soil, plants, and air particles (Hou and Yan, 1998; Pulhani et al., 2000; Ochsenkuhn and Ochsenkuhn-Petropoulou, 2003; Rosamilia et al., 2004; Ramlia et al., 2005). Due to the higher resonance integral cross-sections of the above two reactions, epithermal NAA can largely reduce the interferences from 24 Na, 38 Cl, and 32 P, and therefore significantly improve the detection limit of 238 U and 232 Th. U and Th can be rapidly determined by measurement of the short-lived radionuclides 239 U and 233 Th using epithermal NAA (Hou et al., 1996a, 1996b). For the analysis of samples with very low concentrations of 238 U and 232 Th, radiochemical NAA has to be used. In this case, a detection limit of 10−9 –10−11 g for 238 U and 232 Th can be obtained (Repinc and Benedik, 2005; Byrne and Benedik, 1988; Dang et al., 1992, 1994;
Activation analysis for the determination of long-lived radionuclides
399
Becker and LaFleur, 1972; Danko and Dybcynski, 1995; Picer and Strohal, 1968; Glover et al., 2001; Yonezawa and Kurosawa, 1996; Benedik and Byrne, 1995). Because the separation is carried out after irradiation, the contamination and reagent blank can be avoid. For some samples, such as blood, urine and tissue, RNAA can supply a very accurate result. Solvent extraction, ion exchange and extraction chromatography have been used for the separation of 239 Np, 233 Pa, as well as 239 U and 233 Th. When 239 Np is used for the RNAA of 238 U, the radiochemical separation procedure is the same as that for 237 Np. Dang et al. (1992) reported a radiochemical procedure for the RNAA of both U and Th. The irradiated sample is first decomposed and prepared in a 9 mol/l HCl solution. This solution is loaded on to anion exchange column. The column is then washed with 9 mol/l HCl and 8 mol/l HNO3 to remove other interfering radionuclides. 233 Pa and 239 Np, which are absorbed on the column, are then eluted with 1 mol/l HCl–1 mol/l NH2 OH·HCl solution. The eluate is evaporated to dryness, Np and Pa are oxidized with K2 Cr2 O7 and then reduced to low valence states. Reduced Np and Pa are co-precipitated with BaSO4 . The precipitate containing 239 Np and 233 Pa is filtered and measured using a HpGe detector. Besides ion exchange, extraction chromatography has also been used to separate 233 Pa and 239 Np using Eichrom TRU and TEVA (Glover et al., 2001). For the rapid determination of U and Th, radiochemical separation of 239 U and 233 Th has to be completed after irradiation. Solvent extraction using HDEHP, TBP and TOPO (Benedik and Byrne, 1995; Becker and LaFleur, 1972) are usually applied. The radioisotopes 238 Np, produced by irradiation of 237 Np, 231 Pa, 235 U, and 230 Th have been used as yield tracers of 239 Np, 233 Pa, 239 U, and 233 Th during the chemical separation (Benedik and Byrne, 1995; Byrne and Benedik, 1988; Huh and Bacon, 1985). In addition, chemical yield can also be measured by standard addition of U and Th and repeating the procedure several times to get an average chemical recovery (Dang et al., 1992). Pre-separation methods have also been widely used to separate U and Th before irradiation, the advantage of this method being no limitation to the sample amount. Thus a large sample can be used, and a very low concentration of U and Th can be determined. This is more useful for water samples, especially seawater samples, which contain high concentrations of Na and Cl, and as a consequence the radioactivity from 24 Na and 36 Cl is extremely high after irradiation, making the post-irradiation separation more difficult. Co-precipitation of U and Th with Al(OH)3 is a simple method to separate U and Th from matrix elements and some interfering elements, but the decontamination from some elements, such as P, Sc, As, and rare earths, is not good, and the detection limit is not very satisfactory (Honda et al., 1990; Ogiwara et al., 1995). Chelating resin has also been used to separate U and Th (Greenberg and Kingston, 1983; Gladney et al., 1983). The pH of the solution is adjusted to 4–6, and in this case U and Th will be absorbed on the column, while most of the salts pass through the column and are washed out by pure water. The resin can be directly irradiated in the reactor and then measured. The advantage of this method is that many other trace elements, such as Co, Cr, Cu, Eu, Fe, Mn, Ni, Sn, Ti, V, and Zn can also be determined, but as a consequence the detection limit for Th and U is not very satisfactory. Bem and Ryan (1984) have reported a simple and sensitive method for the pre-concentration of uranium using precipitation of U with 1-(2-pyridylazo)-2-naphthol (PAN), 1,2-cyclohexylenedinitrilotetraacetic acid (CyDTA) being used as masking agent to prevent the precipitation of other elements. For large volumes of water, ion exchange combined with hydroxide precipitation is a suitable method for the separation of U and Th (Huh and Bacon, 1985; Chen et al., 2001b). Figure 7 shows a procedure
400 X. Hou
Fig. 7. Chemical procedure for separation of U and Th.
Activation analysis for the determination of long-lived radionuclides
401
used in the author’s laboratory for the separation of U and Th. 100–200 l water samples have been treated using this method for determination of ultra-low levels of the isotopes of U and Th (Chen et al., 2001b). The measurement of 239 Np and 233 Pa can be carried out by counting their gamma rays of energies 106.5 (27.2%) and 277.6 (14.4%), and 228.2 (10.8%) for 239 Np and 312.17 (38.6%) for 233 Pa. NAA of 235 U can be carried out by measuring some fission products, such as 140 La, 141 Ce, 143 Ce, 133 I and 134 I (Dermelj and Byrne, 1997; Augustson et al., 1980; Gladney et al., 1983). For post-irradiation separation NAA, one of the fission product radionuclides has to be separated. For example, if 133 I or 134 I is measured, the irradiated sample can be combusted and the released iodine is trapped in an alkaline solution, which is then extracted in CCl4 solution for measurement (Dermelj and Byrne, 1997). In addition, delayed neutron counting and fission tracks are used for the determination of 235 U (Vasconcellos et al., 1987; Armelin and Vasconcellos, 1986). The delayed neutron counting method is based on the emission of neutrons from the neutron activated elements, which are normally fissionable elements, such as 235 U, 238 U, 233 U, 232 Th, 239 U. In this method, neutrons instead of gamma-rays are measured using a neutron detector.
6. Conclusion (1) Neutron activation analysis can be used for the determination of some long-lived radionuclides as an alternative to radiometrics and mass-spectrometry. (2) NAA has been proved to be a very sensitive method for the determination of 129 I, 238 U, 235 U, and 232 Th, a useful method for 237 Np, and a potentially useful method for 135 Cs in environmental samples. (3) Compared to radiometric methods and ICP-MS, the disadvantage of NAA is the requirement for a nuclear research reactor. (4) NAA has been recognized as a very important radioanalytical tool for production and evaluation of certified reference materials for radionuclides (Povinec, 2004).
References Armelin, M.J.A., Vasconcellos, M.B.A. (1986). An evaluation of the delayed neutron counting method for simultaneous analysis of uranium and thorium for 235 U/238 U isotopic ratio determination. J. Radioanal. Nucl. Chem. 100 (1), 37–47. Assinder, D.J. (1999). A review of the occurrence and behaviour of neptunium in the Irish Sea. J. Environ. Radiat. 44, 335–347. Augustson, J.H., Haynes, J.W., Sanders, T.W. (1980). The determination of uranium in biological material by neutron activation analysis using the fission product 134 I. J. Radioanal. Chem. 60, 373–383. Aumann, D.C. (1981). Determination of 129 I and 127 I at natural levels by neutron activation analysis. Radiochim. Acta 29, 209–215. Aumann, D.C., Buheitel, F., Faleschini, H., Friedmann, L., Mullen, G. (1985). Iodine-129 in mammalian thyroid in the Federal Republic of Germany. Health Phys. 49, 495–498. Aumann, D.C., Buheitel, F., Hausclid, J., Robens, E., Wershofen, H. (1987). Chemical and nuclear interferences in neutron activation of 129 I and 127 I in environmental samples. J. Radioanal. Nucl. Chem. 109, 261–273.
402
X. Hou
Becker, D.A., LaFleur, P.D. (1972). Determination of trace quantities of uranium in biological materials by NAA using a rapid radiochemical separation. Anal. Chem. 44 (8), 1508–1511. Bem, H., Ryan, D.E. (1984). Determination of uranium in seawater and biological materials by neutron activation analysis after selective preconcentration with 1-(2-pyridylazo)-2-naphthol. Anal. Chim. Acta 138, 124–129. Benedik, L., Byrne, A.R. (1995). Simultaneous determination of trace uranium and thorium by radiochemical neutron activation analysis. J. Radioanal. Nucl. Chem. 189 (2), 325–331. Bouisset, P., Lefevre, O., Cagnat, X., Kerlau, G., Ugron, A., Calmet, D. (1999). Direct gamma-X spectrometry measurement of I-129 in environmental samples using experimental self-absorption corrections. Nucl. Instrum. Methods Phys. Res. A 437 (1), 114–127. Byrne, A.R. (1986). Determination of 237 Np in Cambrian (U.K.) sediments by neutron activation analysis: preliminary results. J. Environ. Radioact. 4, 133–144. Byrne, A.R., Benedik, L. (1988). Determination of uranium at trace levels by radiochemical neutron activation analysis employing radioisotopic yield evaluation. Talanta 35 (3), 161–166. Byrne, A.R., Benedik, L. (1999). Application of neutron activation analysis in determination of natural and man-made radionuclides, including Pa-231. Czechoslovak J. Phys. 49 (S1), 265–271. Chao, J.H., Tseng, C.L. (1996). Determination of 135 Cs by neutron activation analysis. Nucl. Instrum. Methods Phys. Res. A 272, 275–279. Chao, J.H., Tseng, C.L., Chou, F.C. (1998). Determination of I-129 in I-131 pharmaceuticals produced in THOR. Appl. Radiat. Isotopes 49 (12), 1587–1590. Chen, Q.J., Dahlgaard, H., Hansen, H.J.M., Aarkrog, A. (1990). Determination of 99 Tc in environmental samples by anion exchange and liquid–liquid extraction at controlled valency. Anal. Chim. Acta 228, 163–167. Chen, Q.J., Dahlgaard, H., Nielsen, S.P., Aarkrog, A., Christensen, I., Jensen, A. (2001a). Determination of 237 Np in marine sediment and seawater. J. Radioanal. Nucl. Chem. 249, 527–533. Chen, Q.J., Aarkrog, A., Nielsen, S.P., Dahlgaard, H., Lind, B., Kolstad, A.K., Yu, Y.X. (2001b). Procedures for determination of 239,240 Pu, 241 Am, 237 Np, 234,238 U, 228,230,232 Th, 99 Tc and 210 Pb–210 Po in environmental samples. Risø-R-1263(EN), Risø National Laboratory, Denmark. Chichester, D.L., Simpson, J.D. (2004). Advanced compact accelerator neutron generator technology for FNAA/PGNAA field work. In: Abstracts of 11th International Conference on Modern Trends in Activation Analysis, Guildford, UK. Conklin, A.R. (2004). Field Sampling—Principles and Practices in Environmental Analysis. Marcel Dekker, New York. Dang, H.S., Pullat, V.R., Pillai, K.C. (1992). Simultaneous determination of 232 Th and 238 U in biological samples. Application to the estimation of their daily intake through diet. J. Radioanal. Nucl. Chem. 162 (1), 163–169. Dang, H.S., Pullat, V.R., Sharma, R.C. (1994). Distribution of uranium in human organs of an urban Indian population and its relationship with clearance half-life. Health Phys. 68 (3), 328–331. Danko, B., Dybcynski, R. (1995). Radiochemical scheme for the determination of Mo and U in biological materials by NAA. J. Radioanal. Nucl. Chem. 192 (1), 117–129. deCorte, F., Simnits, A.S. (2003). Recommendation data for use in the k(0) standardization of neutron activation analysis. At. Data Nucl. Data Tables 85 (1), 47–67. Dermelj, M., Byrne, A.R. (1997). Simultaneous radiochemical neutron activation analysis of iodine, uranium and mercury in biological and environmental samples. J. Radioanal. Nucl. Chem. 216 (1), 13–18. Edwards, R.R. (1962). Iodine-129. Its occurrence in nature and its utility as a tracer. Science 137, 851–853. Eriksson, M. (2002). On weapons plutonium in the arctic environment (Thule, Greenland). Risø R-1321(EN), Risø National Laboratory, Denmark. Fehn, U., Holdren, G.R., Elmore, D., Brunelle, T., Teng, R., Kubik, P.W. (1986). Determination of natural and anthropogenic 129 I in marine sediments. Geophys. Res. Lett. 13, 137–139. Foti, S., Delucchi, E., Akamian, V. (1972a). Determination of picogram amounts of 99 Tc by neutron activation analysis. Anal. Chim. Acta 60, 261–268. Foti, S., Delucchi, E., Akamian, V. (1972b). Determination of picogram amounts of technetium in environmental samples by neutron activation analysis. Anal. Chim. Acta 60, 269–276. Frechou, C., Calmet, D., Bouisset, P., Piccot, D., Gaudry, A., Yiou, F., Raisbeck, G. (2001). I-129 and I-129/I-127 ratio determination in environmental biological samples by RNAA, AMS and direct gamma-X spectrometry measurements. J. Radioanal. Nucl. Chem. 249 (1), 133–138.
Activation analysis for the determination of long-lived radionuclides
403
Germain, P., Pinte, G. (1990). Neptunium-237 in the marine environment, determination in animal and plant species in the English Channel: Biological indicators and trophic relationships. J. Radioanal. Nucl. Chem. 138 (1), 49–61. Germain, P., Guegueniat, P., May, S., Pinte, G. (1987). Measurement of transuranium elements, chiefly 237 Np (by neutron activation analysis), in the physical and biological compartments of the French shore of the English Channel. J. Environ. Radioact. 5, 319–331. Gladney, E.S., Peters, R.J., Perrin, D.R. (1983). Determination of 235 U/238 U ratio in natural water by Chelex-100 ion exchange and neutron activation analysis. Anal. Chem. 55, 976–977. Glover, S.E., Qu, H., LaMont, S.P., Grimn, C.A., Filby, R.H. (2001). Comparison of two ultra-sensitive methods for the determination of 232 Th by recovery corrected pre-concentration radiochemical neutron activation analysis. J. Radioanal. Nucl. Chem. 248 (1), 29–33. Goerner, W., Noll, B., Spies, H., Thieme, K. (1988). Determination of technetium in 99 Mo/99m Tc generators. J. Radioanal. Nucl. Chem. 122 (2), 291–298. Greenberg, R.R., Kingston, H.M. (1983). Trace element analysis of natural water samples by neutron activating analysis with chelating resin. Anal. Chem. 55, 1160–1165. Guinn, V.P., Garzanov, E., Cortes, E. (1978). Further-studies in advance prediction of gamma ray spectra and detection limits in instrumental neutron activation analysis. J. Radioanal. Chem. 43 (2), 599–609. Handl, J., Pfau, A., Huth, F.W. (1990). Measurement of 129 I in human and bovine thyroids in Europe—transfer of 129 I into the food chain. Health Phys. 58 (5), 609–618. Handl, J., Oliver, E., Jakob, D., Johanson, K.J., Schuller, P. (1993). Biospheric 129 I concentration in the pre-nuclear and nuclear age. Health Phys. 65 (3), 265–271. Hatsukawa, Y., Toh, Y., Oshima, M., Hayakawa, T., Shinohara, N., Kushita, K., Ueno, T., Toyoda, K. (2003). New technique for the determination of trace elements using multiparameter coincidence spectrometry. J. Radioanal. Nucl. Chem. 255 (1), 111–113. Honda, T., Oi, T., Ossaka, T., Nozaki, T., Kakihana, H. (1990). Determination of thorium and uranium in hot spring and crater lake waters by neutron activation analysis. J. Radioanal. Nucl. Chem. 139 (1), 65–77. Hou, X.L. (2000). Cyclic activation analysis. In: Meyers, R.A. (Ed.), Encyclopedia of Analytical Chemistry. Applications, Theory and Instrumentation. John Wiley and Sons, Chichester. Hou, X.L. (2004). Application of 129 I as an environmental tracer. J. Radioanal. Nucl. Chem. 262, 67–75. Hou, X.L., Yan, X.J. (1998). Study on the concentration and seasonal variation of inorganic elements in 35 species of marine algae. Sci. Total Environ. 222, 141–156. Hou, X.L., Wang, K., Chai, C.F. (1996a). Epithermal neutron activation analysis and its application in the miniature neutron source reactor. J. Radioanal. Nucl. Chem. 210 (1), 137–148. Hou, X.L., Chai, Z.F., Wang, K. (1996b). Epithermal neutron activation analysis of trace uranium using short-lived nuclide 239 U. Nucl. Techniques 20 (9), 568–571 (in Chinese). Hou, X.L., Dalhgaard, H., Rietz, B., Jacobsen, U., Nielsen, S.P., Aarkrog, A. (1999). Determination of 129 I in seawater and some environmental materials by neutron activation analysis. Analyst 124, 1109–1114. Hou, X.L., Dalhgaard, H., Nielsen, S.P. (2000a). Iodine-129 time series in Danish, Norwegian and northwest Greenland coast and the Baltic Sea by Seaweed. Estuar. Coast. Shelf Sci. 51, 571–584. Hou, X.L., Dalhgaard, H., Nielsen, S.P., Ding, W.J. (2000b). Iodine-129 in human thyroid and seaweed in China. Sci. Total Environ. 246, 285–291. Hou, X.L., Dalhgaard, H., Nielsen, S.P. (2001). Chemical speciation analysis of 129 I in seawater and a preliminary investigation to use it as a tracer for geochemical cycle study of stable iodine. Mar. Chem. 74 (2–3), 145–155. Hou, X.L., Dahlgaard, H., Nielsen, S.P., Kucera, J. (2002). Level and origin of iodine-129 in the Baltic Sea. J. Environ. Radioact. 61, 331–343. Hou, X.L., Malencheko, A.F., Kucera, J., Dahlgaard, H., Nielsen, S.P. (2003a). Iodine-129 in thyroid and urine in Ukraine and Denmark. Sci. Total Environ. 302, 63–73. Hou, X.L., Fogh, C.L., Kucera, J., Andersson, K.G., Dahlgaard, H., Nielsen, S.P. (2003b). Iodine-129 and caesium137 in Chernobyl contaminated soil and their chemical fractionation. Sci. Total Environ. 308, 97–109. Houdek, F., Obrusník, I., Svoboda, K. (1979). Determination of technetium in 99m Tc radiopharmaceutical solutions by means of activation analysis. Radiochem. Radioanal. Lett. 39 (4–5), 343–352. Huh, C.A., Bacon, B.P. (1985). Determination of thorium concentration in seawater by neutron activation analysis. Anal. Chem. 57, 2138–2142. Hursthouse, A.S., Baxter, M.S., McKay, K., Livens, F.R. (1992). Evaluation of methods for the assay of neptunium and other long-lived actinides in environmental matrices. J. Radioanal. Nucl. Chem. 157 (2), 281–294.
404
X. Hou
IAEA (2004). Analytical Quality Control Services. International Atomic Energy Agency, http://www.iaea.org/ programmes/aqcs/database/database_search_start.htm. Ikeda, N., Seki, R., Kamemoto, M., Otsuji, M. (1989). Activation analysis for technetium-99 by the use of a neutron excitation reaction. J. Radioanal. Nucl. Chem. 131 (1), 65–71. Isaksson, M. (2000). Sampling methods. A survey of methods in use in the Nordic countries. Nordisk Kernesikkerhedsforskning, Roskilde (DK), NKS-17. Jha, S.K., Bhat, I.S. (1994). Measurement of 237 Np in the marine environment of coastal nuclear sites in India. J. Radioanal. Nucl. Chem. 182 (1), 5–10. Kalmykov, St.N., Aliev, R.A., Sapozhnikov, D.Yu., Sapozhnikov, Yu.A., Afinogenov, A.M. (2004). Determination of Np-237 by radiochemical neutron activation analysis combined with extraction chromatography. Appl. Radiat. Isotopes 60, 595–599. Karam, L.R., Pibida, L., McMahon, C.A. (2002). Use of resonance ionization mass spectrometry for determination of Cs ratios in solid samples. Appl. Radiat. Isotopes 56, 369–374. Keith-Roach, M.J., Stürup, S., Oughton, D.H., Dahlgaard, H. (2002). Comparison of two ICP-MS set-ups for measuring 99 Tc in large volume water samples. Analyst 127, 70–75. Kim, C.K., Seki, R., Ikeda, N. (1988). Determination of 237 Np in soil samples by neutron activation analysis. Radioisotopes 37, 229–230. Lee, T., Ku, T.L., Lu, H.L., Chen, J.C. (1993). First determination of fallout Cs-135 and potential applications of 137 Cs/135 Cs ratios. Geochim. Cosmochim. Acta 57, 3493–3497. Markert, B. (Ed.) (1994). Environmental Sampling for Trace Analysis. VCH Verlagsgesellschaft, Weinheim. Massart, D.L. (1971). Cation-exchange techniques in radiochemistry. NAS-NS-3113, National Academy of Science, National Research Council. May, S., Engelmann, Ch., Pinte, G. (1987). Neutron activation determination of 237 Np in irradiated experimental fuels and waste solutions and distribution studies in sea-water and submarine fauna and flora of disposal areas. J. Radioanal. Nucl. Chem. 113 (2), 343–350. Mincher, B.J., Baker, J.D. (1990). The separation of 99 Tc from mixed fission products and its determination by neutron activation analysis. J. Radioanal. Nucl. Chem. 139 (2), 273–276. Moreno, J.M.B., Betti, M., Nicolaou, G. (1999). Determination of caesium and its isotopes composition in nuclear samples using isotope dilution–ion chromatography–inductively coupled plasma mass spectrometry. J. Anal. At. Spectrom. 14 (5), 875–879. Muramatsu, Y., Ohmomo, Y., Christoffers, D. (1984). Neutron activation analysis of iodine-129 and iodine-127 in environmental samples. J. Radioanal. Nucl. Chem. 83 (2), 353–361. Muramatsu, Y., Uchida, S., Sumiya, M., Ohmomo, Y. (1985). Iodine separation procedure for the determination of 129 I and 127 I in soil by neutron activation analysis. J. Radioanal. Nucl. Chem. Lett. 94 (5), 329–338. Muramatsu, Y., Ohmomo, Y., Sumiya, M. (1988). Determination of iodine-129 and iodine-127 in environmental samples collected in Japan. J. Radioanal. Nucl. Chem. 123 (1), 181–189. Nielsen, S.P. (Ed.) (2000). Seminar on sampling techniques, radiochemical methods, detectors and measurements. Risø National Laboratory, Roskilde (DK). Ochsenkuhn, K.M., Ochsenkuhn-Petropoulou, M. (2003). Determination of leachable uranium in airborne particulates. J. Radioanal. Nucl. Chem. 256 (1), 87–91. Ogiwara, K., Oi, T., Ossaka, T., Mukaida, M., Honda, T. (1995). Determination of uranium and thorium in solar salts by neutron activation analysis. J. Radioanal. Nucl. Chem. 191 (2), 273–278. Parr, R.M. (1986). Technical consideration for sampling and sample preparation of biomedical samples for trace element analysis. J. Res. Natl. Bureau Stand. 91 (2), 51–57. Parry, S.J., Bennett, B.A., Benzing, R., Lally, A.E., Birch, C.P., Fulker, M.J. (1995). The determination of I-129 in milk and vegetation using neutron activation analysis. Sci. Total Environ. 173 (1–6), 351–360. Pibida, L., McMahon, C.A., Bushaw, B.A. (2004). Laser resonance ionization mass spectrometry measurements of cesium in nuclear burn-up and sediment samples. Appl. Radiat. Isotopes 60 (2–4), 567–570. Piccot, D., Ayrault, S., Gaudry, A. (2002). Determination of 237 Np by k0 RNAA. J. Radioanal. Nucl. Chem. 254 (1), 213–215. Picer, M., Strohal, P. (1968). Determination of thorium and uranium in biological materials. Anal. Chim. Acta 40, 131–136. Povinec, P.P. (2004). Developments in analytical technologies for marine radionuclide studies. In: Livingston, H.D. (Ed.), Marine Radioactivity. Elsevier, Amsterdam, pp. 237–294.
Activation analysis for the determination of long-lived radionuclides
405
Pulhani, V., Kayasth, V., More, A.K., Mishra, U.C. (2000). Determination of trace uranium and thorium in environmental matrices by neutron activation analysis. J. Radioanal. Nucl. Chem. 243 (3), 625–629. Raghavan, R., Ramakrishna, V.V., Patil, S.K., Ramaniah, M.V. (1976). Neutron activation analysis method for the determination of neptunium in sub-microgram quantities in process solutions. J. Radioanal. Chem. 33, 31–38. Raisbeck, G.M., Yiou, F. (1999). I-129 in the ocean: origins and applications. Sci. Total Environ. 238, 31–41. Ramlia, A.T., Wahab, A., Hussein, M.A., Wood, A.K. (2005). Environmental U-238 and Th-232 concentration measurements in an area of high level natural background radiation at Palong, Johor, Malaysia. J. Environ. Radioact. 80 (3), 287–304. Reijonen, J., Leung, K.N., Firestone, R.B., English, J.A., Perry, D.L., Smith, A., Gicquel, F., Sun, M., Koivunoro, H., Lou, T.P., Bandong, B., Garabedian, G., Revay, Z., Szentmiklosi, L., Molnar, G. (2004). First PGAA and NAA experimental results from a compact high intensity D–D neutron generator. Nucl. Instrum. Methods A 522 (3), 598–602. Repinc, U., Benedik, L. (2005). Simultaneous determination of trace uranium and vanadium in biological samples by radiochemical neutron activation analysis. J. Radioanal. Nucl. Chem. 264 (1), 77–81. Rook, H.L., Suddueth, J.E., Becker, D.A. (1975). Determination of iodine-129 at natural levels using neutron activation and isotopic separation. Anal. Chem. 47 (9), 1557–1562. Rosamilia, S., Gaudino, S., Sansone, U., Belli, M., Jeran, Z., Ruisi, S., Zucconi, L. (2004). Uranium isotopes, metal and other elements in lichens and tree barks collected in Bosnia–Herzegovina. J. Atmos. Chem. 49 (1–3), 447– 460. Rosenberg, R.J. (1993). Non-conventional measurement techniques for the determination of some long-lived radionuclides produced in nuclear fuel: A literature survey. J. Radioanal. Nucl. Chem. 171 (2), 465–482. Ruf, M., Friedrich, M. (1978). Neutron activation analysis determination of small amounts of neptunium-237 in solution containing uranium, plutonium and fission products. Nucl. Technol. 37, 79–83. Sekine, T., Yoshihara, K., Nemeth, Zs., Lakosi, L., Veres, A. (1989). A new determination method of 99 Tc by nuclear excitation. J. Radioanal. Nucl. Chem. 130 (2), 269–278. Song, M., Probst, T.U., Berryman, N.G. (2001). Rapid and sensitive determination of radiocaesium (Cs-135, Cs-137) in the presence of excess barium by electrothermal vaporization–inductively coupled plasma–mass spectrometry (ETV-ICP-MS) with potassium thiocyanate as modifier. Fresenius J. Anal. Chem. 370, 744–751. Stamm, H.H. (1973). Determination of 135 Cs in sodium from an in-pile loop by activation analysis. J. Radioanal. Chem. 14, 367–373. Vasconcellos, M.B.A., Armelin, M.J.A., Figueoredo, A.M.G., Mazzilli, B.P., Saiki, M. (1987). A comparative study of some nuclear method for 235 U/238 U isotopic ratios determination. J. Radioanal. Nucl. Chem. 113 (2), 357– 370. Wyse, E.J., Lee, S.H., La Rosa, J., Povinec, P., de Mora, S.J. (2001). ICP-sector field mass spectrometry analysis of plutonium isotopes: recognizing and resolving potential interferences. J. Anal. At. Spectrom. 16, 1107–1111. Yagi, M., Sekine, T., Yoshihara, K. (1991). Determination limit of technetium in 99 Tc(γ , γ )99m Tc radioactivation analysis in the presence of molybdenum. J. Radioanal. Nucl. Chem. Lett. 155 (6), 435–443. Yonezawa, C., Kurosawa, T. (1996). Determination of ultra trace amounts of uranium and thorium in reference materials of silicon dioxide and aluminum by radiochemical NAA. Bunseki Kagaku 45 (5), 435–440.
This page intentionally left blank
407
In situ and airborne gamma-ray spectrometry Andrew N. Tyler∗ Environmental Radioactivity Laboratory, School of Biological and Environmental Sciences, University of Stirling, Stirling, FK9 4LA, United Kingdom
1. Introduction 1.1. The philosophy and scope It is generally appreciated that conventional sampling methodologies combined with laboratory based analyses of environmental media are hindered by issues associated with the representative nature of the sample or samples, access, coverage, time required for laboratory based analyses and delays in obtaining results. The ability to make rapid real time measurements of environmental radioactivity therefore brings immediate benefits to surveys for the purposes of prospecting, baseline monitoring, contamination mapping and site characterization. In addition, a detector based at 1 m (e.g., in situ gamma spectrometry, IGS) (Figure 1) and 100 m (e.g., for airborne gamma spectrometry, AGS) (Figure 2) above the ground will typically measure 4×104 and 1.3×107 times, respectively, more soil than a soil core of 10 cm diameter and about 30 cm depth. IGS and AGS not only give more spatially representative measurements, averaging out small scale heterogeneity, but a spatial context emerges when measurements are spatially juxtaposed. This spatial context is crucial for effective interpretation in many aspects of environmental radioactivity and radioecology. Further more, the substantially larger sample gained through IGS and AGS measurements will yield similar counting statistics in a fraction of the time of typical laboratory measurements, enabling maps of environmental radioactivity to be rapidly constructed. IGS and AGS have therefore found significant application in geological mapping and uranium prospecting, baseline mapping, identification and characterization of contaminated areas, dose reconstruction and the determination of dose specific radionuclide quantities, monitoring of planned releases of radioactivity and the search for point sources in the environment. For environmental purposes, there are distinct advantages and value when the results derived from different techniques are comparable and IGS and AGS are no exception. In this case these techniques are often compared with each other and against conventional soil sample ∗ E-mail address:
[email protected]
RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11013-5
© 2008 Elsevier B.V. All rights reserved.
408
A.N. Tyler
Fig. 1. Example of a portable HPGe n-type detector and soil coring for empirical validation. Detector supported from a purpose built tripod.
derived measurements of environmental radioactivity. However, this can only be effectively achieved when sampling plans are designed to match the spatial response of the IGS or AGS detector, and sufficient account is given to spatial heterogeneity (Tyler et al., 1996a). One of the most important disadvantages of IGS and AGS is that the accuracy of the analyses depends on a separate knowledge of the vertical activity distribution, and to a lesser extent, soil density, moisture content and for low energy gamma photons (≈200 keV) on chemical composition (Beck et al., 1972; Tyler et al., 1996b; Tyler, 1999). Other influences that might affect the comparison between soil sample and IGS and AGS measurements include: (i) tree canopy effects, although minor include shielding and bringing the source closer to the detector (Kogan et al., 1971); and (ii) surface roughness and topographic unevenness (Laedermann et al., 1998; Schwarz et al., 1992; Kogan et al., 1971). Laedermann et al. (1998) describe a marked reduction in the effect of surface topography as the vertical activity distribution changes from a surface distribution to one with increasing depth distribution. Gering et al. (2002) demonstrate that tree canopies only contribute a few percent to the total kerma rate and thus photon fluence for IGS detectors, but the trees themselves can provide significant shielding of photons compared with grassland areas. However, the leaf litter in forests and woods tend to keep 137 Cs higher in the soil profile of coniferous stands compared with open grassland (Miller et al., 1990; Bunzl and Kracke, 1988) counterbalancing the effect of the attenuation of the trees themselves. However, some evidence suggests that deciduous forest enable greater depth penetra-
In situ and airborne gamma-ray spectrometry
409
Fig. 2. Airborne gamma spectrometry system deployed by the Scottish Universities Environmental Research Centre. System comprising a 16 l pack NaI(Tl) detector inside the aircraft and two Lo-Axial HPGe detectors deployed externally (from Sanderson and McLeod, 2000).
tion of 137 Cs (Veen and Mejer, 1989), which combined with the attenuation characteristics of tree trunks may result in a substantially smaller photon fluence compared with open grassland areas with similar contamination levels. This chapter will provide a historical perspective to the development and implementation of mobile gamma ray spectrometry techniques. The chapter will then focus on the state of the art of IGS and AGS systems, perhaps the two end members of environmental gamma spectrometry, although similar systems have been deployed in road vehicles (e.g., Sanderson et al., 2000) and hovercrafts (e.g., Jones et al., 1984, 1988, 1999). Solutions to overcoming problems of making effective comparisons between techniques with different spatial response characteristics and overcoming problems of stratification of the vertical activity distribution will be dealt with. Examples of the application will also be presented. The application of IGS to borehole and marine applications will not be addressed. 1.2. Historical perspective IGS and AGS systems were originally fundamental geophysical tools used in geological mapping of large areas, primarily in the search for enriched uranium deposits and often in combination with other geophysical instrumentation. IGS was often used as a tool for overcoming issues of sampling error when undertaking mapping or prospecting survey work. However, the calibration of IGS methods for measuring environmental radioactivity introduces a new
410
A.N. Tyler
set of problems. Historically, IGS systems used for geological exploration relied upon calibrations or sensitivity estimates determined from fixed concrete calibration facilities, typically enriched with 40 K, equilibrated U and Th. For AGS calibration, these calibrated IGS detectors were used to characterize calibration areas typically 1 km by 3 km (IAEA, 1976, 1979, 1991; Alexander and Kosanke, 1978; Løvborg et al., 1978). The calibrations were quantified in terms of activity per unit dry weight or as concentration as determined directly from the concrete calibration pads. Comparisons of AGS and IGS measurements with associated soil or sediment samples analyzed in the laboratory were often found to be less favorable than comparisons between ground based IGS measurements and AGS measurements (Darnley, 1984; Løvborg et al., 1978; IAEA, 1976). This was attributed to changes in the environmental conditions between soil overburden and soil type, water concentrations and density, and problems associated with uranium series disequilibrium in the field and within the sample collected and analyzed in the laboratory (Alexander and Kosanke, 1978). Even in the mapping of a region of exposed rock outcrop with a collimated in situ NaI(Tl) detector, Løvborg et al. (1971) observed difficulties in relying on a calibration derived from rock samples collected with each IGS measurement. This was due to the heterogeneity of U and Th and their daughters in the rock, and as a result Løvborg, among others, relied upon calibrations derived from concrete calibration pads. Comparisons made by Løvborg (1982) between five international calibration facilities provided good traceability between similarly calibrated detection systems for K and Th with a 2% accuracy, whilst U calibrations were consistent between 4 and 6%. However, the comparison between laboratory determined soil concentrations and IGS determinations remained problematic. With the growth of the nuclear industry and the subsequent dispersal into the environment of anthropogenic radionuclides through weapons testing, licensed discharges and nuclear accidents, these former geophysical techniques were adapted to monitor gamma emitting anthropogenic radionuclides in the environment, especially 137 Cs. These techniques have been applied to the detection of point sources and more famously in the search for radioactive debris from Cosmos 954 (Grasty, 1980, 1981; Bristow, 1978) as well as to the mapping of environmental radioactivity following the Windscale fire in 1957 (Williams et al., 1957; Chamberlain et al., 1961) and other nuclear accidents including Chernobyl (Mellander, 1989; Lindahl and Habrekke, 1987). The Chernobyl accident and its impact across Europe in particular led to a rapid expansion in the development and application of IGS and AGS systems primarily for the purposes of environmental monitoring. Given this new environmental application, the problem of relating IGS and AGS measurements of environmental radioactivity with soil sample derived estimates re-emerged. Deposition of radioactivity as a result of fallout from local (including sea spray), regional or global sources of tropospheric or stratospheric fallout when coupled with effects of topography, surface roughness and the soil’s own natural variability in chemical, physical, hydrological and formation characteristics (Livens and Baxter, 1988) can lead to highly heterogenic distribution over a range of spatial scales (e.g., Tyler and Heal, 2000; Tyler et al., 1996a) and environmental contexts, including grassland and forests (Sutherland and de Jong, 1990; Bunzl and Kracke, 1988), salt marshes (Parkinson and Horrill, 1984) and nuclear weapons test sites (Gilbert and Simpson, 1985; Barnes et al., 1980; Barnes, 1978; Fowler et al., 1974; Gilbert et al., 1974).
In situ and airborne gamma-ray spectrometry
411
The typical solution adopted to determine detector calibrations for anthropogenic radionuclides, principally 137 Cs, was to analytically derive solutions through photon fluence equations. Such equations, initially used for uniform geological sources (e.g., Duval et al., 1971), were modified to account for assumed negative exponential depth distributions of radionuclides in soil (ICRU, 1994; Grasty, 1979; Beck et al., 1972; Beck, 1978). The photon fluence equation derived by Beck et al. (1972) has since been extensively applied (e.g., Jacob et al., 1994; Sowa et al., 1989; Helfer and Miller, 1988; Jacob et al., 1986; Dickson et al., 1976). However, the assumption of a constant negative exponential profile can lead to loss in accuracy in the estimation of radionuclide concentrations, and with time the exponential profile assumption tends to fail. Solving the problem of accounting for variations in the vertical activity distribution has perhaps become the “Holy Grail” in environmental gamma ray spectrometry. 1.3. Principles and definitions 1.3.1. Soil compositions The unscattered photon flux traveling through soil is not only dependent on density but, at low photon energy (<150 keV), the mass attenuation coefficient (μm ) can be influenced by variations in soil composition. Beck et al. (1972) indicated that this may result in a typical variation between 6–7%. Tyler (1994) and Allyson (1994) undertook a series of calculations to estimate the variation in μm for a series of world soil compositions chosen from the Geostandards News Letter (Govindara, 1989) which provides the major and trace elemental composition of soils as well as other reference materials. This provided the major weight percent oxide of soils, which includes SiO2 , Al2 O3 , Fe2 O3 , FeO, MnO, MgO, CaO, Na2 O, K2 O, TiO2 , P2 O5 , H2 O, and CO2 . To this C was also added as a major soil component. For completeness, trace elements were also selected and rationalized to those with a potential for making a significant contribution to changes in μm . Thus those more abundant elements with a Z greater than 50 were selected. These included parts per million (ppm) concentrations of Ar, Ba, Bi, Ce, Cu, Dy, Eu, Er, Gd, Hf, Ho, La, Lu, Nd, Os, Pb, Pt, Sn, Th, U, and W. In addition, N and S were also included. The contributions of coherent and incoherent scattering, photoelectric effect and pair production contributions and total μm were tabulated in a spread sheet for the above elements from Storm and Israel (1970) with later contributions from Hubbell (1982). The oxide fractions were then split by weight in order to provide their concentrations in ppm. Elemental compositions of individual soils were then tabulated and the individual photon interactions and total μm were calculated by weighting each elemental photon interaction and total μm appropriately for each elemental member of the soil as shown in Equation (1): μm =
N
wi μmi ,
(1)
i=1
where μmi is the mass attenuation coefficient of element i, N is the number of elements present in the soil fraction, wi represents the abundance by weight of the ith element, such that the sum of wi = 1. Zmean of the soil was also calculated simply by the weighted average of the individual Z’s.
412
A.N. Tyler
Fig. 3. (A) Photon cross-sections for Beck’s soil composition (Beck et al., 1972) from 30 to 6000 keV. (B) Comparison of μm for a range of soil compositions relative to Beck’s soil composition. After Tyler (1999), Allyson (1994) and Tyler (1994).
Figure 3A shows these contributions for Beck’s soil composition and Figure 3B compares these to the other soils compositions. Figure 1B shows that above about 150 keV, incoherent scattering dominates μm and is independent of soil composition. Below 150 keV, μm becomes dominated by the photoelectric cross-section (μpe ) of the soils, which is directly dependent upon the soil composition. Thus below 150 keV, photon fluence rates from soil profiles are likely to be controlled by soil or sediment composition in addition to density and the vertical depth distribution. Hence changes in organic content, moisture content and mineral composition will have an important implication on IGS measurements of 241 Am and 210 Pb for example. The magnitude of this effect at low energy may at best render IGS measurements of 241 Am and 210 Pb semi-quantitative when measuring across soils or sediments with heterogeneous composition characteristics. 1.3.2. Photon fluence equations and fields of view There are two principle techniques of quantifying the field of view of a detector, also referred to as the circle of investigation (COI). The first is to use point sources or areas spiked with quantified amounts of activity of known photon energy, and measure the signal with distance from the detector to determine the detector response (e.g., Cutshall and Larsen, 1986). The detector spatial response is determined by integrating these responses over the area represented
In situ and airborne gamma-ray spectrometry
413
by each radial distance. However, it is difficult to adapt this technique experimentally to allow for the effect of source burial. The alternative and more efficient technique is to solve photon transport equations analytically. Beck et al. (1972), Anspaugh et al. (1972), Helfer and Miller (1988), Sowa et al. (1989) and ICRU (1994) present and discuss the fundamental quantities used for IGS. The parameters are: N the photopeak count rate (s−1 ), A the activity per unit area (Bq m−2 ), per unit volume (Bq m−3 ) or per unit mass (Bq kg−1 ) in the soil, and ψ the fluence rate (cm−2 s−1 ). The calibration coefficient at energy E is defined as N/A, which is given by Equation (2): N No N ψ − = − , A No ψ A
(2)
where: N/No is the angular correction factor of the detector at energy E for a given source distribution in the soil. This is measured relative to θ = 0◦ or 0 radians, and is dependent upon detector shape as well as the source distribution profile within the soil. No /ψ is the full energy peak count rate per unit fluence rate for a plane parallel beam of photons at energy E, that is normal to the detector face. This is purely dependent upon detector characteristics. ψ/A is the efficiency, at energy E, from photons arriving at the detector unscattered due to a gamma transition for a particular nuclide. This is purely dependent upon the source distribution characteristics within the soil. Of importance from these photon fluence equations is the determination of the circle of investigation. The contribution of fluence rate to a point above the ground is given by plotting the fraction of the total fluence rate (ψ/A) with radius. By incorporating the detector angular correction factor, the circles of investigation can be calculated. One solution to calculating the circle of investigation was described by Duval et al. (1971) and Figure 4 describes the typical integration limits over which the circle of investigations are calculated. However, since Duval et al.’s work, the importance of the vertical depth distribution of radionuclides (Helfer and Miller, 1988; Dickson et al., 1976; Beck et al., 1972) and angular response of detectors has been incorporated into calculations (Helfer and Miller, 1988; Grasty, 1979; Beck et al., 1972). Allyson (1994) describes in addition the influence of the angular response of detectors. Instead of integrating by parts, the circle of investigation of radius R is defined as COI in Equation (3): COIR =
CR , C∞
(3)
where C is given by Equation (4) (Beck et al., 1972): π
"2π " 2 "∞ C= 0
where:
0
h cos θ
Ae[−(α/ρs )ρs z] γ100 σdet,θ,ϕ 2 Ptot sin θ e−μa ρa Pa e−μs ρs Ps dR dθ dϕ, 2 4πPtot
(4)
414
A.N. Tyler
Fig. 4. The spherical coordinate system for in situ gamma ray detector with integration limits R1 R ∞, θ1 θ π/2 and 0 ϕ 2π for field of view calculations (Tyler, 1994).
σdet,ϕ = effective detection cross-section and its angular dependence; Ptot = total gamma-ray path length = ra + rs ; α = the reciprocal of the relaxation length of the assumed exponential-distributed source activity with depth, cm−1 ; μa = mass attenuation coefficient in air; μs = mass attenuation coefficient in soil; ρa = density of air; ρs = density of soil or rock; z = vertical depth from surface; R = radius of circular area; A = activity; γ100 = number of gamma-rays emitted per 100 decays. Examples of the results from Equations (3) and (4) are given in Figure 5. The COI may be usefully defined as the horizontal radial distance from a point directly underneath detector on the ground surface to a point from which 90% of the infinite yield originates. Comparing Figures 5A and 5B demonstrates that the COI is very strongly dependent on the vertical activity distribution for measurements made 1 m above the ground and less so at AGS altitudes (∼100 m). This is primarily due to the increased photon path traveled through soil due to the wider field with respect to detector height, observed as a large solid angle θ . For an AGS based detector, the distance traveled by photons in the soil is not so great given the smaller θ, and thus the distribution of the source in the soil is not as important. In addition, the COI for 137 Cs, also increases from around 9 m radius for a detector at 1 m height increasing to about 180 m radius for a detector at 100 m altitude. Conversely, the influence of photon energy
In situ and airborne gamma-ray spectrometry
415
(A)
(B) Fig. 5. Graphs showing the influence of the 137 Cs depth distribution (mass relaxation per unit area) on the field of view of (A) an in situ detectors and (B) aerial survey detector (after Tyler et al., 1996a; Allyson, 1994).
416
A.N. Tyler
on the COI is more important at AGS altitudes (180 m at 662 keV and 270 m at 2614 keV) than at 1 m (9 m at 662 keV and 9.5 m at 2614 keV). For the purposes of this discussion, the influence of detector shape on the field of view is not considered, particularly when other environmental factors involved in empirical comparisons are quite considerable. However, the effect of detector shape is effectively characterized through Monte Carlo simulation (Allyson and Sanderson, 2000, 2001). 1.3.3. Depth distributions The fluence rate in soils is a function of the soil density in addition to source depth z. For relating exposure, dose and IGS measurements with source burial, it is more appropriate to determine depth distribution functions in terms of mass per unit area given by ρz, rather than linear depth (Tyler et al., 2001, 1996a, 1996b; Hillman et al., 1996; ICRU, 1994; Rybacek et al., 1991). Given that soil density ρ varies with depth, and depth intervals selected tend also to vary, and the combined effect will lead to further variations in z, activities should be expressed in terms of activity per unit mass (Bq kg−1 ). Thus the function mass per unit area, defined as x is given by the function in Equation (5): "z x=
ρ(z ) dz ,
(5)
0
where x is the total mass depth or effective mass per unit area (g cm−2 ) for the soil sample depth provided. Following deposition from the atmosphere, radioactivity can quickly penetrate into the soil and the vertical activity distribution can initially be approximated by a negative exponential. From ICRU (1994), Rybacek et al. (1991) and Beck et al. (1972) the activity distribution with depth A(x) (Bq g−1 ) is given by Equation (6): A(x) = Ao · e(−x/β) , x ln A(x) = ln Ao · , β 1 ∴ slope = − , β
(6)
where Ao is the activity concentration (Bq kg−1 ) at the surface and β is the relaxation mass per unit area (g cm−2 ) or mean mass depth. From Beck et al. (1972), Miller et al. (1990) and Helfer and Miller (1988), density is assumed to be constant with depth and β is equivalent to ρ/α, where α is the reciprocal of the relaxation length (cm−1 ). However, often considerable change in density is observed, especially in soils where the organic content can vary markedly. As β will be derived with changes in soil density taken into consideration, the activity per unit area Aa may be obtained by an integration of the specific activity A(x) over the mass per unit area, resulting in Equation (7) (ICRU, 1994; Rybacek et al., 1991): Aa = β · Ao , where Ao is the massic activity at the surface (Bq g−1 ).
(7)
In situ and airborne gamma-ray spectrometry
417
With time a subsurface maxima often develops from aged cesium deposits derived from atmospheric deposition and from buried source layers in salt marsh environments. Thus there is a requirement to be able to describe the depth characteristic and relate it back to previous work and β. Zombori et al. (1992) overcame this by combining a positive and negative exponential to describe such features. However, the shape of the distributions observed is unlikely to fit exponential profiles and there is a requirement to derive more than a single function to resolve the double exponential. A single parameter to describe the vertical activity distribution has the distinct advantage that it may be related to the forward scattered region to the left of the full energy peak of interest, commonly 137 Cs. This approach was first introduced by Zombori et al. (1992) and usually referred to as the peak to valley method (Tyler et al., 1996b; Hillman et al., 1996). Hillman et al. (1996) show that the Lorenz function can be used to fit subsurface maxima profiles well, and provides a numerical description of the depth (A2 ) at which the subsurface maximum of activity occurs and a measure of the FWHM (A3 ) of the subsurface maxima, which if assumed to be a normal distribution, approximates to 2.35σ . Hillman et al. (1996) show that there is a reasonable correlation between the coefficients, A2 and A3 , enabling the potential for IGS to estimate the parameter for depth distribution. Here, a conceptually simpler model is considered, relying on fewer assumptions. From the discussion thus far: x1/2 ρ = = mean mass depth. β= (8) α ln 2 By fitting a polynomial to the subsurface maximum and integrating with respect to mass depth, the parameter β, also referred to as the mean mass depth (Tyler, 1999; Tyler et al., 1996a, 1996b), can be estimated from Equation (9): &∞ x · A(x) dx β = 0& ∞ (9) , 0 A(x) dx where x is the mass depth (g cm−2 ) and A(x) is the massic activity per unit mass (e.g., Bq g−1 ). This can be solved by fitting a polynomial curve to the depth distribution A(x) and calculating the mean mass depth from the area under the curve. However, given equal mass depth intervals, this could be simplified to Equation (10): x 0 A(x) · x β= (10) , x 0 A(x) where x are equal mass depth interval in g cm−2 . It is common to assume a uniform distribution with depth for natural radionuclides. However, it is not uncommon to find some stratification as a result of natural on anthropic influence. These distributions may be difficult to characterize using, activity per unit area, mean mass depth or Lorentz function. Thummerer and Jacob (1998) describe an alternative two slab model, described by three parameters: (i) the activity per unit mass in the upper layer; (ii) the ratio of the activities of the upper and lower layer; (iii) the thickness of the upper layer. This was then related to spectral measures of forward scattering and differential attenuation of energetically very different full energy peaks (two line method) derived from the same natural
418
A.N. Tyler
Fig. 6. Comparison of a 35% relative efficiency n type HPGe detector (FWHM 1.36 keV) and a 76 × 76 mm NaI(Tl) detector (FWHM 6.6%). Measurement of a 200 g IAEA 375 reference material in a lead shield.
radionuclide, e.g., 214 Bi, 214 Pb. These spectral methods are discussed in more detail later in this chapter. 1.3.4. Spectral processing As discussed, the primary systems adopted for IGS and AGS are based on, but not exclusively, HPGe or NaI(Tl) technology and frequently in combination with AGS systems (e.g., Sanderson et al., 2004a, 2000; Gutierrez et al., 2002). A comparison of spectral capability is illustrated in Figure 6. Given the high spectral resolution of HPGe detectors, the full energy peak counts are relatively straightforward to process. For NaI(Tl) based systems, where the FWHM can range from 6 to 10%, the full energy peaks of interest typically also contain scattered secondary contributions from higher energy primary gamma photons and other primary photons of similar gamma photon energies. Figure 7 shows the full energy peaks derived from concrete calibrations pads at the Scottish Universities Environmental Research Centre for 40 K, U and Th. The windows used to measure U from 214 Bi and Th from 208 Tl are also shown. In addition, the full energy peaks from point sources of 137 Cs and 134 Cs are also shown to illustrate the influence on these windows from the natural series. To determine the net count rates within each window, the spectrum has to be stripped through a process known as deconvolution or spectral stripping. It has become the standard practice to use a set of concrete calibrations
In situ and airborne gamma-ray spectrometry
419
Fig. 7. The net spectra from concrete calibration pads and regions of interest for a 76 × 76 mm NaI(Tl) detector (Tyler, 1994).
420
A.N. Tyler
pads to simulate the scattering of natural radionuclides from rock and soils (Grasty, 1981; Løvborg, 1983; IAEA, 1979, 1990, 1991). The net count rate in each window can be determined by Gaussian elimination, matrix inversion or simple window stripping as given in Equations (11) for the naturals. This was subsequently extended to include for 137 Cs and 134 Cs after the Chernobyl accident (e.g., Sanderson et al., 1989) as shown in Equations (12):
NTh = NTh ,
, NU = NU − αNTh
− γ NU , NK = NK − βNTh
(11)
NCs137 = NCs137 − aNTh − bNU − cNK − cNCs134 ,
− eNTh − f NU − gNK − hNCs137 , NCs134 = NCs134
(12)
where: NTh , NU , NK , NCs137 = stripped or true counts,
, N , N , N NTh U K Cs137 = measured counts, α, β, γ , a, b, c, d, e, f, g, h = the stripping coefficients. The stripping approach can be applied to window counts as well as full spectral stripping (Tyler et al., 1996b). Similar full spectrum analysis (FSA) has also been successfully adopted for use with Bismuth Germanate (BGO) detectors (Hendricks et al., 2001). However, this approach, whilst widely accepted is not without its limitations. Some residual contribution or over stripping may result as the scattering contributions of concrete may not resemble the real environment. As suggested by Potts (1978), the uncertainty on the final stripped count rate can be strongly influenced by the stripping ratios and the relative intensities of the signals within that widow. Allyson and Sanderson (2001) also demonstrate the relationship between stripping ratios and changes in spectral resolution, detector gain and changes in aircraft altitude through Monte Carlo simulation. Alternative approaches to stripping have been proposed and are gradually being adopted through either commercial channels or adoption of scientific principle, especially for the processing of AGS data. This includes the alternative approaches of the Noise Adjusted Single Value Decomposition (NASVD) (Hovgaard, 2000; Aage et al., 1999) and spectral profiling (Guillot, 2001). The NASVD method analyses a complete data set for all independent spectral shapes. The spectral shapes or components are ordered in accordance with their overall importance. Spectral component zero is the mean spectral shape and spectral component 1 is the spectral shape that best explains the residual spectra after removal of the mean spectrum. In an ideal case, the spectral components should range from 0 to k, where k is the number of independent gamma ray fields in the survey area. If this should be 137 Cs, 40 K, U and Th, the k will equal 4. However, there is a Poisson noise component to these data sets and if all observed spectra are scaled inversely to the standard error (Noise Adjustment), a singular Value Decomposition of the data will result in Noise Adjusted spectral components ordered in accordance with their importance. In this case, only spectral components 0–4 will contain real spectral information, the remaining components will be dominated by spectral noise (Hovgaard, 2000; Aage et al., 1999).
In situ and airborne gamma-ray spectrometry
421
Guillot (2001) presents an alternative approach of spectral profiling. Here digital filters are used, taking into consideration the characteristics (e.g., FWHM) of the absorption peaks for 512 channel spectra, significantly reducing the counting fluctuations, making detection possible based on the first and second derivatives. Guillot shows that whilst this approach has little benefit over the conventional stripping approach for the naturals, the results demonstrate that it is particularly suited to the detection of man-made nuclides, without any prior assumption about the number and nature of radionuclides presents. This is particularly useful for the detection of an anomaly in a very large number of measurements and thus the detection of radioactive sources. Given the difficulty of simulating the scattering contributions at low energy, this approach has a distinct advantage at low energy, e.g., the detection of 241 Am (Guillot, 2001).
2. Gamma ray spectrometry systems and calibration 2.1. Instrumentation The spectral advantages of HPGe over NaI(Tl) detector technology has been demonstrated in Figure 3, and have the potential of yielding a wealth of data on the presence of gamma emitting radionuclides compared with NaI(Tl) based systems. This is also the case for post accident scenarios in AGS systems. However, NaI(Tl) have the advantage that they are relatively cheap, more robust and thus relatively more portable, and can be manufactured in a range of sizes and consequently able to provide substantially higher relative detection efficiencies at a fraction of the cost. The ultimate choice of instrumentation is highly dependent on the final application and the desirable detection limits, in addition to the budget available. The choice of detector is not exclusive to HPGe or NaI(Tl) detectors. Other inorganic scintillation detectors which present a real alternative include BGO and Cs(Tl) detectors (Knoll, 1989). The BGO detector (e.g., Hendricks et al., 2001) has poorer resolution but superior gamma absorption characteristics to NaI(Tl). Cs(Tl) detectors that also have superior gamma absorption characteristics and are more shock resistant compared to NaI(Tl) and also have the capability of differentiating between different charged particles. However, Cs(Tl) emission spectra are poorly matched to photomultiplier tubes. Here, therefore, the focus is given to the most popular systems in common usage, HPGe and NaI(Tl). For most IGS and AGS applications the normal configuration is to have the detector pointing downwards with associated electronics and cryostat above the detector. The special requirements of AGS detectors will be discussed later. For IGS applications the detector is held at about 1 m above the surface and live counting times are selected on the basis of the counting precision required. The field of view of the detector, can be adjusted by lowering and raising the detector or by collimation with lead shielding. In contrast to laboratory based gamma spectrometry, the environmental source geometry is not fixed and is a function of soil type and composition, soil moisture, the vertical activity distribution and surface roughness and evenness. In addition, environmental conditions, including temperature, humidity and exposure to sunlight are constantly changing which may influence detector performance, such as gain setting, and thus need to be regularly checked. All systems should be energy calibrated.
422
A.N. Tyler
Fig. 8. Example of a portable 76 × 76 mm NaI(Tl) detector coupled with Digital Electronics and portable computer, along side a 50 × 50 mm NaI(Tl) detector.
For portable systems, the data acquisition should also include a means for integrating positional information, usually from a global positioning system (GPS) or differential GPS (DGPS). 2.1.1. NaI(Tl) detectors NaI(Tl) detectors are configured such that one side of the crystal is in direct contact with the photomultiplier tube (PMT). The remaining sides of the crystal are surrounded by graded MgO to reflect light into the PMT and the unit is hermetically sealed in an aluminum can, typically 0.5 mm thick. The gain across the PMT a sensitive function of the applied voltage across the PMT and the interdynode voltage can be adjusted by gain or potentiometer. A pre-amplifier is primarily used for signal processing and sends output to the multi-channel analyzer (MCA) or multi-channel buffer (MCB). Today, digital units are available which connect to the fourteen pin photomultiplier. These digital units comprise the high voltage potentiometer, pre-amplifier and multi-channel analyzer and have the advantage of faster signal processing capabilities, multiple synchronizing, overlapping and sub one second sampling of the detector output. Figure 8 illustrates the typical detector configuration coupled with a portable computer through USB communications. Blue tooth technology is also simplifying remote communication with detector systems. 2.1.2. HPGe detectors The electronics of HPGe semiconductor system is essentially the same as a laboratory based system, comprising a High Voltage bias across the crystal via a high voltage filter in the preamplifier. The pre-amplifier is primarily responsible for pulse shaping and is placed close to the detector to minimize the capacitive load on the detector. The main part of the ampli-
In situ and airborne gamma-ray spectrometry
423
Fig. 9. Example of a portable in situ HPGe p-type detector with digital electronics, supported by a camera tripod.
fier is positioned further away with the rest of the nuclear electronics, which can otherwise affect the signal to noise ratio. The gain setting is adjusted through the amplifier and the output is captured in a MCA. The amplifier, high voltage bias supply and MCA are captured in portable electronics (Figure 1) which can usually be controlled through software and a portable computer. With the advance of digital technology the external electronics are becoming increasingly miniaturized as shown in Figure 9. The final handicap to the portability of HPGe detectors is the necessity to keep the detectors cooled at 77 K with liquid nitrogen. The size of the cryostat largely dictates the portability as well as the longevity of the system between re-fills. Power supply issues required by electrical cooled systems remain a barrier for its implementation in IGS.
424
A.N. Tyler
Fig. 10. The expanding hexagonal sampling and typical sampling sets A and B (Tyler et al., 1996a).
2.2. Calibration As described above, calibrations can be derived theoretically through analytical solutions to photon transport equations. However, usually validation of IGS and AGS is made with reference to soil or sediment sample derived estimates. Where variations in lateral activity distribution is suspected within the field of view of IGS and AGS systems, comparisons and empirical calibrations may be affected by random and systematic uncertainties introduced by the environmental heterogeneity. Under these circumstances, which tend to persist in most environmental contexts, structured sampling plans (Tyler et al., 1996a) are to be preferred to: (i) the use of single or limited numbers of samples which cannot account for the common situation of significant lateral variation in activity distribution, (ii) other sampling systems, e.g., random, which are not suitably weighted for an effective comparison and are at best inefficient and are liable to bias, and (iii) alternatives such as rotor-tilling sites to improve homogeneity (e.g., Cutshall and Larsen, 1986). The sampling strategy proposed by Tyler et al. (1996a) and widely adopted through European AGS Intercomparison exercises (Sanderson et al., 2004a, 2004b) is shown in Figure 10. The typical hexagonal plan includes a single central core and five hexagonal shells spaced at 2, 8, 32, 128 and 256 m. The mean activity from each shell should be weighted such that the it matches the circle of investigation as illustrated in Figure 11. Ideally, the circle of investigation should pass through the mid point m between each shell r, calculated from Equation (13): r2 − r1 + r1 , (13) 2 where r1 is the radius of the inner shell and r2 is the radius of the outer of the two shells. The intercept of m with the circle of investigation provides the weighting attached to the shell r1 . m=
In situ and airborne gamma-ray spectrometry
425
Fig. 11. The coincidence between the circles of investigation for 137 Cs and the hexagonal sampling plan (Tyler et al., 1996a).
The final weighting for the 256 m shell simply represents the remainder of the data such that 100% weighting is achieved. Tyler et al. (1996a) demonstrate how the weighting of the outer shells marginally increases with photon energy. Figure 11 is an example of this matching for 137 Cs. The mean activity or mean mass depth (β) can be calculated for each shell or for each set of sample spacing. The six points on each shell can be used to calculate the standard deviation and the standard error of the mean. By using the appropriate shell weighting the error contributions from each shell can be summed appropriately to provide an estimate of the overall uncertainty on the weighted mean activity estimate for each detector altitude comparison. Such an approach enables the effect of other environmental factors, such as the mean mass depth, to be confidently observed on calibration coefficients. An example of calibration coefficients for the 16 l AGS detector is given in Table 1. 2.3. Conversion to gamma dose quantities Gamma spectrometers have distinct advantages over single parameter systems for the determination of dose, usually air kerma (nGy h−1 ), as they can identify the contributing source or sources and derive a number of useful radiological quantities (Tompson et al., 1999). There are several procedures for converting the gamma spectra height spectrum into measures of air kerma and the key approaches are detailed in ICRU (1994) and Tompson et al. (1999). Here, commonly used approaches are summarized.
426
A.N. Tyler
Table 1 Comparison of derived calibration coefficients from hexagonal calibration sites. All errors quoted as 1σ standard error except b which is 1σ standard deviation (from Tyler et al., 1996b) Calibration site
Mean mass depth β (g cm−2 )
Ax at 100 m (kBq m−2 )
N/Ax (cps/kBq m−2 )
Vesivehmaa Airport, Finland, 1995 Longbridgemuir, Dumfries, SW Scotland. Terrestrial—Basin and Valley Peat. Aug. 1993 Caerlaverock, Dumfries, SW Scotland. Salt Marsh. February 1992 Wigtown Bay, Wigtown, SW Scotland. Salt Marsh. Aug. 1993
1.31 ± 0.15
47.2 ± 8.5
10.0 ± 2.0
3.6 ± 1.1
9.5 ± 2.2
6.75 ± 0.67
13.2 ± 2.1a
77.2 ± 8.2
2.71 ± 0.40
12.8 ± 5.1b
160 ± 13
2.88 ± 0.42
a Estimated from data recorded in Tyler et al. (1996a, 1996b). b Estimated from 5 cores sampled along transect perpendicular to the coastline.
2.3.1. Full energy peak method Having derived an energy dependent efficiency calibration for the IGS detector, spectral analyses should provide estimates of the radionuclide concentrations in soil. This approach is especially appropriate for HPGe detectors. For the natural radionuclides, the assumption of a uniform distribution of depth provides a relatively straightforward conversion from Bq g−1 to kerma rate, given in ICRU (1994). However, knowledge of the depth distribution for the anthropogenic radionuclides is required and ICRU (1994) publish a range of conversion factors for sources distributed exponentially (β) in soil. The main shortcoming here is that knowledge of the depth distribution is required to account for the secondary gamma photon fluence on the air kerma contribution, which in more shielded circumstances can give a significant contribution to the dose rate (Jacob et al., 1994). 2.3.2. Energy band method The energy band approach is mainly used with poor resolution systems such as NaI(Tl) detectors. Here, a procedure such as spectral deconvolution or stripping is required to calculate the net contributions to the spectrum from K, U, Th and 137 Cs. These can then be converted to activities and their contributions to kerma rate estimated. Again, some knowledge of the depth distribution is desirable (Tompson et al., 1999). 2.3.3. Total or partial energy method This method returns a single parameter estimate kerma and uses either a large proportion or the entire spectrum. EML (1992) suggest a range from about 150 to 3000 keV, assuming that the energy response is relatively similar across the defined energy range. Once calibrated
In situ and airborne gamma-ray spectrometry
427
against a standard reference source, air kerma rates in the environment can be estimated. The main disadvantage of this approach is that it does not use the full spectral capabilities. 2.3.4. Energy dependent air kerma conversion Grasty et al. (2001) presented a procedure of converting full gamma ray spectra on a 76 × 76 mm NaI(Tl) detector into air kerma with an energy sensitive calibration conversion. Ten radioactive sources covering the energy range from 60 to 1836 keV were used to measure the energy deposited in the detector, which combined with Monte Carlo simulation was used to estimate the energy deposited in the detector from a semi-infinite plane. The known air kerma rates for each source were used to provide calibrations for the 10 regions in the spectrum. Any energy calibrated spectra could then be converted to an air kerma rate without a prior knowledge of the source distribution. 2.3.5. Spectral unfolding ICRU (1994) describe the spectral unfolding procedure in detail and Tompson et al. (1999) provide a shorter summary. The measured spectrum is a distribution of pulse heights that is a spectrum of absorption in the detector and is thus affected by partial interaction through Compton scattering of gamma photons out of the detector. In this context, differentiating between photons that are partially deposited in the detector from photons that are already scattered in the environment is very difficult. Furthermore, the full energy peak response of the detector is energy dependent. It would therefore be desirable to know what the incident spectrum on the detector is. These affects have to be taken into account through the unfolding procedure to measure the pulse height distribution in which partial energy deposition events are subtracted and the conversion to fluence is made using the peak response (ICRU, 1994). The unfolding procedure requires an accurate energy dependent response function of the detector from mono-energetic photons within the range of the spectrum. This can be derived experimentally or through Monte Carlo simulation.
3. In situ gamma-ray spectrometry 3.1. Calibration corrections As already outlined, the primary limitation of IGS has been the need for independent knowledge of the vertical activity distribution. A means of overcoming this limitation, without the need for further soil sampling, has been one of the primary focuses of IGS detector research over the last decade or more. Three main approaches to this problem have been pursued: (i) the differential attenuation of gamma photon emission lines, or two line method (Rybacek et al., 1991; Miller et al., 1994); (ii) the forward scattering or peak to valley method (Zombori et al., 1992; Karlberg, 1990); and (iii) the use of lead collimators (Benke and Kearfoot, 2002; Fülöp and Ragan, 1997) or lead shielding at various distances in front of the detector (Korun et al., 1994) coupled with repeat IGS measurements to reconstruct depth distributions. The third approach has the potential for characterizing complex depth distributions, but the collimation necessitates longer counting times and repeat measurements with differing collimation configurations. The first and second approach requires a general assumption to
428
A.N. Tyler
be made of the nature of the depth distribution shape, i.e. “exponential” like or subsurface maximum. The first two approaches have proven popular and are discussed here in more detail. 3.1.1. The two line method The two line method is applicable to those radionuclides emitting photons of more than one energy, which are energetically very different. Rybacek et al., (1991) used the 137m Ba 32 keV X-ray and gamma-photon at 662 keV to estimate the 137 Cs vertical activity distribution. However, the 32 keV X-ray is strongly attenuated and highly susceptible to photo-electric absorption and thus soil composition. Miller et al. (1994) used the 234m Pa line at 1001 keV and 234 Th line at 92.6 keV to provide accurate estimates of the distribution of 238 U contamination in surface soils of nuclear weapons production sites. Thummerer and Jacob (1998) take this approach further with the 583 and 2615 keV lines for 208 Tl for 232 Th determination. They identify four possibilities of 238 U decay series from the 295 keV 214 Pb line and the 609, 2204 and 2448 keV 214 Bi lines. They compared the two line method with the peak to valley approach for the same radionuclides and demonstrated that the two line method tended to yield better counting statistics and thus better precision. However, the two line method cannot be applied to mono-energetic radionuclides such as 40 K. 3.1.2. The full energy peak to forward scattered ratio (peak to valley method) The peak to valley method provides, in many cases, a more flexible approach to account for systematic variations in the vertical activity distribution and was demonstrated by Karlberg (1990) and Zombori et al. (1992). Zombori demonstrated that the build-up of secondary scattered gamma photons in the valley region (between the full energy peak and Compton edge) relative to the full energy peak for a point 137 Cs source buried beneath a tank of water over a range of water depths. In principle, the extent of forward scattering is systematically related to the interaction probability associated with the photon trajectory between source and detector and hence to the amount and type of material intervening. Thus the amount of forward scattering relative to the size of the full energy peak will provide a measure of the extent of source burial (Tyler et al., 1996b). Figure 12 identifies the valley region between the 137 Cs full-energy peak and the Compton edge of a gamma ray spectrum, and indicates the origins of its principle contributions. Examples of interactions corresponding to these spectral contributions are also shown schematically in Figure 13. Given constant detector characteristics, source configuration and ground clearance, the spectral contributions from multiple Compton scattering (C) within the detector crystal prior to photon escape is illustrated by b1 in Figure 13. Forward scattering from the material surrounding the detector can also be denoted by b1 . The contribution from b1 is directly proportional to the full energy peak A. The contribution from forward scattering interactions in the source and air path is given by b2 and is also proportional to the source activity and thus also to the full-energy peak intensity A. However, as source burial increases, the contributions to the valley region from forward scattering of primary gamma photons within the soil b2 will increase relative to the size of the full energy peak. Thus the spectral step BT will increase relative to the area under the full energy peak A.
In situ and airborne gamma-ray spectrometry
429
Fig. 12. Example of a gamma ray emission spectra demonstrating the quantification of forward scattering relative to the area under the full energy peak. See Fig. 13 for definition of b1 and b2 (Tyler, 1999).
Fig. 13. Diagram showing examples of possible gamma photon interactions from a 137 Cs source buried in a soil and with a detector (Tyler et al., 1996b).
430
A.N. Tyler
This approach to estimating source burial provides only a single parameter Q with which to describe the source depth, where Q is given by Equation (14): Q=
A . BT
(14)
In general, it may be expected that environmental source profiles are better described by two or more parameters as profile shapes may vary considerably. However, at present Q potentially provides a means of monitoring variations in mass depth for an activity distribution which varies systematically from place to place. This approach has been used successfully to provide more accurate IGS determinations of both natural (e.g., Tyler et al., 2001; Thummerer and Jacob, 1998) and anthropogenic radionuclides, primarily 137 Cs (e.g., Kastlander and Bargholtz, 2005; Tyler et al., 2001, 1996b; Tyler, 1999; Zombori et al., 1992; Karlberg, 1990). The ability to characterize the vertical activity remotely, not only lends itself to improved accuracy IGS determinations of environmental radioactivity, but greatly extends the potential applications and flexibility of the IGS technique. The peak to valley method is therefore becoming a widely adopted approach. Here we summarize some of the general applications of IGS and then focus on a few examples. 3.2. Examples of implementation 3.2.1. Brief review of general applications Unsurprisingly the majority of published applications of IGS have a radiological emphasize, including the determination of air kerma in a variety of landscapes (e.g., Anastasis et al., 2005; Gering et al., 2002; Grasty et al., 2001; Jacob et al., 1994) and background assessments (e.g., Malczewski et al., 2004). Intercomparison exercises have highlighted the usefulness of IGS in interpreting results from single parameter dosimetry systems (Wissman and Sáez Vergara, 2007) and provided comparable results (Shebell et al., 2003). Shebel et al. demonstrated that 84% of the teams reproduced 226 Ra, 232 Th and 40 K within 20% of the soil sample mean. In addition to basic geological mapping applications, Hadley et al. (2000) used IGS to refine the differentiate between geological units in Northern Ireland, Lindsay et al. (2004) used IGS to predict radon exhalation in the Witwatersrand region of South Africa and Chen and Chan (2002) used IGS to evaluate the weathering grade of volcanic rocks in Hong Kong. Ruffel et al. (2006) used IGS to help resolve the location of subsurface structures, such as buried pipelines and foundations, using the K/Th and Th/U ratios. 3.2.2. Mapping the accumulation of 137 Cs in salt marsh environments Radionuclide bearing effluents discharged into the Irish Sea have resulted in the accumulation of radionuclides in salt marsh environments, which can contribute to critical group exposures. A field investigation was undertaken at Caerlaverock salt marsh, Dumfries, Scotland to test the accuracy with which the peak to valley method could resolve the accuracy of 137 Cs inventories, surface 241 Am concentrations, the mass depth (β) of the 137 Cs subsurface maximum and the sediment accumulation rates (Tyler, 2004, 1999). An IGS n-type HPGe detector was used to estimate a spectrally derived coefficient QCs , derived from the 137 Cs full energy peak. Figure 2 shows the relationship between QCs and
In situ and airborne gamma-ray spectrometry
431
Fig. 14. The relationship between QCs and β of the vertical 137 Cs activity distribution, r 2 = 0.94 (Tyler, 1999).
the mean mass per unit area of the vertical 137 Cs activity distribution derived from 11 sediment cores distributed across the survey area. Whilst small variations in the 137 Cs depth profile shape were observed in addition to overall changes in the depth distribution, Figure 14 demonstrates that the approach is robust enough to compensate for these variations for a given β. A similar relationship is also derived for QCs and the IGS calibration coefficient N/ACs (r 2 = 0.95). Subsequent comparisons between 16 single sediment core derived 137 Cs inventories and IGS derived 137 Cs inventories revealed an improvement in the comparison from an r 2 = 0.0009 (uncorrected) to an r 2 = 0.87 (QCs corrected). Given a typical sampling error of between 8 and 10% (1 standard deviation) at Caerlaverock (Tyler et al., 1996a), this demonstrated a good agreement between the new IGS approach and conventional soil coring methods. However, the QCs calibration correction failed to fully correct the IGS estimates at sites close to the salt marsh edge, where rapid sedimentation rates resulted in effectively uniform 137 Cs depth distribution profiles. This resulted in the introduction of a systematic error in the measurements at these extreme locations. Figure 15A shows an example of the results from the IGS survey and shows an accurate representation of the distribution of 137 Cs across the salt marsh. 241 Am specific activity concentrations were also mapped across the salt marsh. The IGS results compared favorably with surface (0–5 cm) sediment core derived estimates, typically within 25% of core sample estimates, accuracies comparable with sampling errors (Tyler et al., 1996a). As demonstrated in Figure 15, the vertical activity distribution β (g cm−2 ) can also be estimated directly from QCs as shown in Figure 15B, demonstrating the increase in the effective depth of the subsurface 137 Cs maximum towards the coastal edge of the salt marsh. As anticipated, this demonstrates a systematic increase in sediment accretion rates (g cm−2 a−1 ) towards the coastal edge of the salt marsh (Figure 15C). These Irish Sea salt marshes have sufficient 137 Cs to enable this approach to work effectively with NaI(Tl) detectors combined with full spectral stripping (Tyler et al., 1996b).
432
A.N. Tyler
Fig. 15. In situ derived maps using QCs of 137 Cs to estimate inventory (A), mean mass depth β (B), and sediment accretion rates (C) (Tyler, 1999).
In situ and airborne gamma-ray spectrometry
433
Fig. 16. Schematic diagram illustrating the application of in situ gamma ray spectrometry to the monitoring of soil redistribution on arable land (not drawn to scale). (A) Stable reference site exhibiting a –ve exponential 137 Cs profile. (B) Stable cultivated site, same total activity as A. (C) Site of soil loss, 137 Cs depletion, same ploughed layer depth as B. (D) Transitional site, balance between soil loss and gain. (E) soil accumulation, increase in soil depth and total 137 Cs input (Tyler et al., 2001).
3.2.3. Mapping soil erosion and field moist bulk density across cultivated fields through in situ gamma-ray spectrometry The combination of recent developments in IGS with the 137 Cs method for quantifying soil erosion has provided the potential for rapid, spatially representative and non-invasive measurement of soil erosion suitable for monitoring sensitive sites such as archaeological crop marks (Tyler et al., 2001). The methodology presented here also uses a 35% n-type HPGe detector. A spectrally derived coefficient (QCs ), derived from the forward gamma ray scattering around the 137 Cs full energy peak (662 keV), was used to correct for changes in the vertical 137 Cs activity distribution observed in ploughed fields and thus provide accurate estimates of 137 Cs activity (Bq m−2 ). Figure 16 illustrates the vertical source characteristics one might expect in cultivated soils. The successful application of IGS relies on the ability to compensate for increases in the “effective plough depth”, the depth over which 137 Cs activity has been homogenized by ploughing. This is most apparent in areas of soil accumulation, i.e. at the bottom of a slope sequence (Figure 17). Adopting a similar philosophy to that described above in salt marshes, the correction for variations in the “effective ploughed layer” thickness can be made with the spectral coefficient QCs . Following detailed soil profiling of the vertical 137 Cs activity distribution at each site, a correction procedure was validated. IGS and soil sample estimates of activity were converted to erosion rates through the power function model (Zhang et al., 1990; Kachanoski, 1993), where the erosion rate E is given by ' 1/n ( M Ci E= (15) 1− R Cr
434
A.N. Tyler
Fig. 17. Estimates of soil erosion—a comparison between uncorrected and corrected in situ with soil sample estimates (Tyler et al., 2001).
and R is the ratio of the concentration of 137 Cs in the eroding sediment to that in the ploughed layer. In this case, R should be estimated from field data and reflects the particle selectivity of the erosion process. M is the thickness of the ploughed layer, Ci is the 137 Cs activity measured, Cr is the reference 137 Cs measured at a stable (non-eroding) control site. To derive an erosion rate, the number of years since peak weapons testing deposition is given by n. The results demonstrate the ability of IGS to map 137 Cs activities to levels of precision equivalent to the errors associated with soil sampling. Estimates of field-moist bulk density were also made from a separate coefficient (QK ), derived from the forward scattering around the 40 K full energy peak (1460 keV). IGS erosion rate estimates, using the power function model, were in excellent agreement with those derived from soil cores (Figure 17). The addition of IGS derived field moist bulk density estimates provides the potential of converting estimates of the vertical activity distribution from units of mass per unit area (g cm−2 ) to linear depth units (cm). However, the fundamental limitation of this approach is the low activity levels associated with weapons testing fallout. This necessitates long counting times or the requirement of very high efficiency HPGe detectors (e.g., 100% relative efficiencies). Working on cultivated soils has the advantage that the cultivation layer has been reasonably well homogenized. In these circumstances, the peak to valley method for 40 K should relate well to the field moist bulk density of the soil. Eight sites were carefully sampled to estimate the mean bulk density to 30 cm depth. Substantial heterogeneity was observed within sample
In situ and airborne gamma-ray spectrometry
435
Fig. 18. Regression between QK and field moist bulk density (Tyler et al., 2001).
sites as indicated in Figure 18, indicating that in situ measurements could yield more precise estimates. This approach may have applicability in precision farming techniques. 3.2.4. The UK’s national soil and herbage pollution survey A national soil and herbage pollution survey of the United Kingdom (UKSHS) was undertaken in 2002. The aim of the UKSHS was to provide a 50 km grid baseline of contaminants across the UK, including metals, PAHs, PCBs, dioxins and natural and anthropogenic radionuclides. This was achieved by sampling relatively undisturbed sites within urban and rural settings. IGS was introduced for the first time as part of this national survey to evaluate its potential for replacing the conventional soil sampling approach in environmental monitoring. The UKSHS provided an opportunity to assess the accuracy and precision of IGS measurements of both natural and anthropogenic radioactivity across a range of soils types and landscapes in the UK (Tyler and Copplestone, 2007). A total of 128 points were measured across the whole of the UK, encompassing a complete spectrum of soil types, geology (providing a range of natural radioactive backgrounds) and anthropogenic radioactivity (derived from atmospheric weapons testing and Chernobyl fallout). The preliminary assessment of measurement accuracy was undertaken on the eleven calibration sites. An excellent comparison between soil sample and in situ estimates were observed for the natural radionuclides. The good comparison (r 2 = 0.932) between 40 K specific activities for soil sample and in situ measurements were observed. The estimates of measure-
436
A.N. Tyler
Fig. 19. Preliminary map of the variation in air kerma and the relative contributions from 238 U series, 232 Th series, 40 K and 137 Cs (Tyler and Copplestone, 2007).
ment precision were substantially better for in situ measurements, the soil samples measurements being affected by spatial variability for the soil core measurements. The natural series 238 U and 232 Th also demonstrate good equilibrium characteristics, although the precision and accuracy on laboratory determination of some low energy 238 U series daughters, e.g., 234 Th, 226 Ra and 210 Pb, make empirical in situ calibration difficult. The specific activity concentrations from the natural series and 137 Cs inventories were converted to air kerma (nGy h−1 ) for each site using ICRU (1994) conversion units. Mini-monitor Series 680 results derived simultaneously were estimated from the mean of three measurements and a standard deviation estimated. The results are in good agreement (r 2 = 0.711). Figure 19 demonstrates the additional advantage of IGS systems over single parameter de-
In situ and airborne gamma-ray spectrometry
437
vices, such that the relative contributions from U, Th, K and 137 Cs (and any other radionuclide) can be estimated in real time. The size of each pie chart in Figure 19 is directly proportional to the air kerma measurement and the relative contribution from each source is directly proportional to the size of each pie segment. The contribution to total dose rate from Chernobyl and weapons testing fallout can be instantaneously seen and its relative importance presented in an easily digestible format.
4. Aerial gamma-ray spectrometry 4.1. Instrumentation There are a range of AGS systems operated by a number of groups throughout the world. Many of these systems are commercially available and some are purpose built. Examples of AGS systems and applications including emergency response are summarized in Sanderson et al. (2000) and specific instrumental set-ups are reported in Pavlik and Engelsmann (2004); Sanderson et al. (2004a); Winklemann et al. (2004); Sanderson et al. (1995) and ICRU (1994). Under nuclear accident emergency conditions it may well be sufficient to fly solely with high relative efficiency HPGe detectors (Gutierrez et al., 2002). More recently, the availability of digital electronics has enabling more flexible and intelligent sampling of detector output. Figure 20 provides a sketch of the typical components of an AGS system. The back bone of any AGS system for non-emergency applications must be a detector with sufficient detection efficient to enable short (1–5 s) integration times, thereby permitting a reasonable ground resolution (or footprint) during the survey. NaI(Tl) detectors are the standard equipment used, usually supplied as multiples of 10 ×10×40 cm crystals. Typically 16 l of NaI(Tl) are used in an AGS system and for geological mapping purposes two extra crystals are added (Figure 21) which are upward and sideways looking and used to substract the airborne 222 Rn daughter component from the measurement, improving the accuracy of Uranium maps. The detectors should normally be housed in boxes with sufficient insulation and protection against physical and thermal shock. Spectral drift is usually monitored through the position of the 40 K full energy peak position at 1460 keV and an automatic gain adjustment implemented according to the measured peak location. Some AGS systems now routinely incorporate high resolution gamma spectrometry capability through high efficiency (>50% relative efficiency) HPGe technology, greatly facilitating the recognition of more exotic radionuclides (Creswell et al., 2000), low energy gammaemitters (e.g., 241 Am) in addition to confirming and helping to interpret the results derived from the NaI(Tl) detectors. Most environmental and baseline mapping purposes require deployment from helicopter based platforms and shock mounting to avoid high frequency interference in HPGe systems is fundamental. A range of commercially available portable electronics is now available for power and signal processing purposes, in many cases miniaturizing the electronics and thus space and weight requirements on the aircraft. Central to all AGS systems must be the data acquisition and storage system which is usually PC based. The system must incorporate information about the aircrafts position, usually through GPS including differential systems, and height above
438
A.N. Tyler
Fig. 20. Schematic diagram of the typical components of airborne gamma spectrometry systems.
Fig. 21. Gamma ray dose rates recorded in the IRE–Fleurus area (Sanderson et al., 2004a).
In situ and airborne gamma-ray spectrometry
439
the ground through radar (Figure 21). AGS systems may also be linked with digital inputs from other airborne geophysical instrumentation (e.g., Hewson and Taylor, 2000; Airo and Loukola-Ruskeeniemi, 2004). 4.2. Examples of implementation 4.2.1. Brief review of general application Many examples of AGS systems and application to environmental problems are summarized in Sanderson et al. (2002) especially from a European and North American perspective. With roots very firmly in geophysics, the application of AGS in the earth sciences continues especially in arid and semi-arid landscapes. Application includes the use of AGS in helping to map geological and geomorphological features (Hewson and Taylor, 2000; Fernandez-Alonso and Theunissen, 1998), mapping soils and regolith (Wilford et al., 1997) and helping to develop regional scale models of sedimentation processes (Pickup and Marks, 2001). From an environmental perspective, AGS has been used also to map uranium ore mining, milling and processing areas of Germany (Winkelmann et al., 2001). However, as previously described, the major advances in AGS have been in application to the nuclear industry in the wake of the Chernobyl accident. Across Europe in particular the rapid advances were also accompanied by a divergence of methodologies and within a European context where cross-boundary comparability is likely to be important in the event of future nuclear accidents, there as a need to harmonize the different approaches to AGS. This was addressed through a Concerted Action funded through the Fourth Framework program coordinated by David Sanderson (Sanderson and Ferguson, 1997). As part of this Concerted Action a bibliography of AGS was published (Sanderson et al., 2001). AGS systems are now routinely used for baseline mapping purposes in relation to the environment and nuclear installations (e.g., Rybach et al., 2001) and they are proving themselves to be excellent tools for mapping contamination in a timely manner (Toivonen, 2004). Here, two examples will be explored in a little more detail relating to characterizing the radiation field around a nuclear site (Sanderson et al., 2004a) and the integration and comparability of the European capability through the EU Fifth Framework program funded ECCOMAGS project (Sanderson et al., 2003, 2004b). 4.2.2. Airborne gamma ray surveys of nuclear sites: Mol Dessel, Belgium The operation of AGS systems from twin engine helicopters lends itself to baseline mapping of often complex radiation fields around nuclear installations, from a range of operational altitudes (30–100 m). Sanderson et al. (2004a) operated a 16 l NaI(Tl) spectrometer system, a single 50% relative efficiency HPGe (GMX) detector, a differentially corrected GPS system and associated electronics and power supply. Operating a UK chartered twin engine helicopter necessitated the need for Belgium recognition of UK aviation approval of the equipment installation. Spectra were recorded with 2 s integration times for the NaI(Tl) detector and 4 s for the GMX detector. GPS position and radar altimetry were recorded simultaneously with the spectra. Background measurements were made over open water and the NaI(Tl) spectra were stripped using stripping coefficients determined from a series of concrete calibration pads and sources. The calibration factors were determined by a combination of response
440
A.N. Tyler
Fig. 22. GMX and NaI(Tl) gamma-ray spectra.
modeling and empirical comparisons on calibration sites (Allyson and Sanderson, 1998; Tyler et al., 1996a). The survey was undertaken in the vicinity of the Mol, some 90 km northeast of Brussels, and at the Institut National des Radioelements (IRE site near Fleurus) to characterize the distribution of anthropogenic and natural radionuclides and ground level air kerma rate. The Mol-Dessel area includes the Belgian Nuclear Research Centre (SCK-CEN), the Belgoprocess sites, Belgonucléaire MOX production plant, the FBFC fuel fabrication plant, and the IRMM institute for reference materials. Sanderson et al. (2004a) provide a detailed report of the survey results, which are summarized here. A total of 7500 NaI(Tl) and 3750 GMX spectra were acquired in 5½ survey hours. 137 Cs fallout was typical of weapons testing fallout, with some contributions from Chernobyl. The distribution of fly ash from a nearby coal power station is clearly detected, otherwise the 40 K, 214 B and 208 Tl reflect the geology and soils. Very localized signals were detected with the various nuclear sites, including 134 Cs and 41 Ar at the SCK-CEN site, 137 Cs, 60 Co and U-series at the Belgoprocess site, and 234m Pa at the FBFC site. The survey at Fleurus area resulted in 5400 NaI(Tl) and 2700 GMX spectra collected in 4 survey hours. Figure 21 shows the gamma dose rate (air kerma) mapped, broadly reflecting the variation in geology and soils in the area. The 137 Cs inventories varied from levels associated with nuclear weapons testing fallout (1–2 kBq m−2 ) to Chernobyl associated levels of around 10 kBq m−2 . Isolated signals can also be located over the IRE site and Figure 22 demonstrates the advantage of the GMX detector in identifying 133 Xe, 131 I and 99 Mo. In one or two locations during the survey the very sensitive NaI(Tl) detectors suffered from spectral distortion as a result of coincidence summing in very high radiation fields, whilst the less sensitive GMX detector continued to operate correctly. The results demonstrated the value of a combined NaI(Tl) and HPGe system in often complex radiation fields whilst also providing a rapid baseline against which change can be compared for future surveys.
In situ and airborne gamma-ray spectrometry
441
Fig. 23. Map of the exercise area, showing the composite mapping areas A to J, common mapping areas X, Y and Z, and the three calibration sites (Sanderson et al., 2003, 2004b).
4.2.3. ECCOMAGS European Intercomparison Exercise Having defined the European capability in AGS through EU 4th Framework Concerted Action (Sanderson and McLeod, 1999), the EU Fifth Framework funded ECCOMAGS exercise was designed to test the metrological capability and comparability between the European AGS teams against protocols of dose rate and deposition mapping and establish the traceability of the AGS methods against ground based techniques as a means of validating the AGS method for potential standardization (Sanderson et al., 2004b, 2003). The exercise was performed in SW Scotland encompassing a range of natural and anthropogenic features including granite bodies and uranium mineralization, weapons testing (1–2 kBq m−2 ) and Chernobyl fallout (up to 30 kBq m−2 ), and strong 137 Cs and 241 Am radiometric features on salt marshes (up to 120 kBq m−2 ) derived from historical discharges from the Sellafield reprocessing plant in Cumbria. Figure 23 shows the survey areas and the division of mapping tasks into 10 composite areas, three common areas and three calibration sites. The calibrations sites comprised expanding hexagons enabling the cores to be spatially matched to IGS and AGS measurements, as described previously (Tyler et al., 1996a). 500 cores were split equally between ten labs from three nations, along with IAEA reference materials and a sub sample of homogenized bulk sample from each site to enable the results from each lab to be standardized. During the exercise in (May 2002), an additional 40 ground locations were sampled with 4 cores on a triangular pattern spaced 8 m from the central core. The cores were similarly distributed for high resolution gamma spectrometry analysis. IGS measurements and dose rate measurements were also taken by ten organizations from five European countries. More than 150 people from 18 institutions and 10 European countries participated in the exercise. The AGS teams collected more than 120,000 NaI(Tl) and 20,000 HPGe spectra. The AGS teams
442
A.N. Tyler
Fig. 24. Composite map of air kerma rate for south-west Scotland with topography from the ECCOMAGS project. Areas labeled X, Y and Z show the common mapping areas (Sanderson et al., 2003, 2004b).
reported their data promptly and the first maps were produced within one day of task completion (Sanderson et al., 2003, 2004a, 2004b). Descriptive statistics of the initial results showed broad agreements between measurements in the common areas, with a factor of variation of less than 2 in the range of 137 Cs mean values and a factor of 2 for dose rate. The most common explanation for the differences in results could be explained by variations in the 137 Cs depth distribution assumed. Thereafter the data was leveled with general differences reduced to <25% for Chernobyl contaminated soils decreasing to 40 and 50% for lower and higher contaminated sites, respectively. These draft comparisons were distributed to the AGS teams who then submitted and revised data set and answers to specific questions. The final data sets showed a good agreement for 137 Cs in the terrestrial areas, also with IGS and soil core derived estimates (Sanderson et al., 2004b). Attempts to account for the variation in mean mass depth (β) on the comparison between AGS and soil sample estimates showed that sample heterogeneity was the more important controlling variable (Sanderson et al., 2003, 2004b). The simple spot leveling methodologies employed demonstrated that seamless data sets could be produced through collaboration, as shown in Figure 24 for air kerma rate. The results show that AGS clearly has an important role to play in nuclear emergency response.
5. Summary and conclusions This chapter has demonstrated that good comparisons generally exist between conventional sampling methodologies and appropriately calibrated IGS and AGS systems when consideration is given to the spatial response of IGS and AGS systems and the influences of spatial heterogeneity on soil sampling methodologies. Increasing wider recognition of the capabilities and advantages of IGS and AGS technologies over conventional methodologies is leading to their wider adoption for routine monitoring in countries of Europe and North America. Nevertheless, examples still exist where conventional sampling methodologies remain the favored approach in circumstances that may be better served or complemented by IGS or AGS methodologies. Influences of the vertical activity distribution generally have a greater impact on IGS measurements than AGS measurements due to the larger solid angle between the IGS detector and source, unless the detector is collimated. Nevertheless, solutions to providing independent
In situ and airborne gamma-ray spectrometry
443
means of measuring changes in the vertical activity distribution through IGS are becoming more commonly used. Techniques such as the two peak and peak to valley ratio need to make some assumptions about the nature of the systematic changes in the vertical activity distribution, but provide a rapid and robust solution. Other approaches involving multiple counting with differing fields of view through shielding are more time consuming but have the potential of being almost completely independent of the knowledge of the vertical activity distribution. Interestingly, these approaches greatly extend the potential application of IGS, including investigations of soil erosion and sediment accumulation. The substantially improved accuracy of IGS, comparable and often better than soil sampling uncertainties, confirm that IGS is a useful platform for detailed small scale or generalized large scale investigations of environmental radioactivity. Significant developments in AGS technology have occurred making it ideally suited to base line mapping, providing a spatial context for more detail surveys and sampling. The realistic integration of HPGe with NaI(Tl) technology lends itself well to emergency response applications and the identification of more exotic gamma emitting sources. In addition, the complete spatial coverage, effectively provided by AGS, provides a suitable data set against which to monitor change in the environmental radiation field. Realistically, environmental monitoring and surveys of radioactivity are best achieved when two or more approaches are used concomitantly. The ultimate choice of technique will depend on the objectives set, but soil samples are more effectively interpreted when placed in a spatial context, overcoming issues of spatial heterogeneity, more easily achieved through IGS and or AGS techniques. Equally, the validation and interpretation of IGS and AGS surveys is substantially improved when measurements are comparable with soil or sediment samples. Survey methodologies would therefore benefit from a combined approach adding substantially to the value of the data retrieved.
Acknowledgements Dr. David Sanderson and the airborne gamma spectrometry team at SUERC, University of Glasgow and the EU FW5 ECCOMAGS project for the provision of the aerial survey data presented in this chapter.
References Aage, H.K., Korsbech, U., Bargholtz, K., Hovgaard, J. (1999). A new technique for processing airborne gamma ray spectrometry data for mapping low level contaminations. Appl. Radiat. Isotopes 51, 651–662. Airo, M.L., Loukola-Ruskeeniemi, K. (2004). Characterization of sulfide deposits by airborne magnetic and gammaray responses in eastern Finland. Ore Geol. Rev. 24 (1–2), 67–84. Alexander, P., Kosanke, K. (1978). The quality of airborne radiometric data. In: Aerial Techniques for Environ. Monitoring, Topical Symp. Proc. Am. Nucl. Soc., Nevada Section. Allyson, J.D. (1994). Environmental gamma ray spectrometry: Simulations of absolute calibration of in-situ and airborne spectrometry for natural and anthropogenic sources. Ph.D. Thesis, Glasgow University. Allyson, J.D., Sanderson, D.C.W. (1998). Monte Carlo simulation of environmental airborne gamma spectrometry. J. Environ. Radioact. 38 (3), 259–282.
444
A.N. Tyler
Allyson, J.D., Sanderson, D.C.W. (2000). Response modeling of airborne survey detectors. In: Sanderson, D.C.W., McLeod, J.J. (Eds.), Recent Applications and Developments in Mobile and Airborne Gamma Spectrometry: Proceedings of the RADMAGS Symposium, University of Stirling, 15–18 June. SURRC, University of Glasgow. Allyson, J.D., Sanderson, D.C.W. (2001). Spectral deconvolution and operational use of stripping ratios in airborne radiometrics. J. Environ. Radioact. 53 (3), 351–363. Anastasis, B.P., Vogiannis, E., Nikolopoulos, D., Halvadakis, C.P. (2005). Preliminary survey of outdoor gamma dose rates in Lesvos Island (Greece). Radiat. Protect. Dosim. 113 (3), 336–341. Anspaugh, L.R., Phelps, P.L., Huckabay, G.W. (1972). IV Methods for the in situ Measurement of Radionuclides in Soil. Pico-Medical Division, Lawrence Livermore Lab. University of California, Livermore, CA. Barnes, M.G. (1978). Statistical Design and Analysis in the Cleanup of Environmental Radionuclide Contamination. Water Resources Centre, Desert Research Institute, University of Nevada System, Reno, NV. DRI Publication No. 45012, NVO 1253-12, 53 pp. Barnes, M.G., Giacomini, J.J., Belman, R.T., Elliot, B. (1980). NTS Radiological Assessment Project: Results from the Frenchman Lake Region of Area 5. Water Resources Center, Desert Research Institute, University of Nevada System, Reno, NV. DOE/DP/01253-17, 105 pp. Beck, H.L. (1978). The physical properties of environmental properties of environmental radiation fields and their utility for interpretation aerial measurements. In: Aerial Techniques for Environ. Monitoring, Topical Symp. Proc. Am. Nucl. Soc., Nevada Section. Beck, H.L., Decampo, J., Gogolak, C. (1972). In situ Ge(Li) and NaI(Tl) gamma-ray spectrometry. HASL-258, United States Atomic Energy Commission—Health and Safety (TID-4500), 75 pp. Benke, R.R., Kearfoot, K.J. (2002). Demonstration of a collimated in situ method for determining depth distributions using gamma ray spectrometry. Nucl. Instrum. Methods Phys. Res. A 483, 814–831. Bristow, G. (1978). The application of airborne gamma-ray spectrometry in the search for radioactive debris from the Russian satellite Cosmos 954. Curr. Res. Geol. Survey Can. Part B 1, 151–162. Bunzl, K., Kracke, W. (1988). Cumulative deposition of 137 Cs, 238 Pu and 239+240 Pu and 241 Am from global fallout in soils from forest, grassland and arable land in Bavaria (FRG). J. Environ. Radioact. 8, 1–14. Chamberlain, A.C., Garner, R.J., Williams, D. (1961). Environmental monitoring after accidental deposition of radioactivity. Reactor Sci. Technol. 14, 155–167. Chen, M.Q.F., Chan, L.S. (2002). In situ gamma ray spectrometry study of weathered volcanic rocks in Hong Kong. Earth Surf. Process. Landforms 27, 613–625. Cutshall, N.H., Larsen, I.I. (1986). Calibration of a portable instrinsic Ge gamma-ray detector using point sources and testing for field applications. Health Phys. 51 (1), 53–59. Creswell, A.J., Sanderson, D.C.W., Allyson, J.D. (2000). Experimental measurements and computer simulation of fission product gamma-ray spectra. In: Sanderson, D.C.W., McLeod, J.J. (Eds.), Recent Applications and Development in Mobile and Airborne Gamma Spectrometry. Proceedings of the RADMAGS Symposium, University of Stirling, 1–18 June. SURRC, University of Glasgow. Darnley, A.G. (1984). Natural Background Radiation in Canada. Geol. Survey Can. Bull. 360, 39 pp. Dickson, H.W., Kerr, G.D., Perdue, P.T., Abdullah, S.A. (1976). Environmental γ -ray measurements using in-situ and core sampling techniques. Health Phys. (Feb.), 1221–1227. Duval, J.S., Cook, B., Adams, J.A.S. (1971). Circles of investigation of an airborne gamma ray spectrometer. J. Geophys. Res. 76, 35. EML (1992). EML Procedures Manual, vol. 1, 27th ed. Environmental Measurements Laboratory, U.S. Department of Energy, New York. Reference quoted from Thompson et al. (1999). Fernandez-Alonso, M., Theunissen, K. (1998). Airborne geophysics and geochemistry provide new insights in the intercontinental evolution of the Mesoproterozoic Kibaran belt (Central Africa). Geol. Mag. 135 (2), 203–216. Fowler, E.B., Gilbert, R.O., Essington, E.H. (1974). Sampling of soils for radioactivity: Philosophy, experience and results. In: Atmosphere—Surface Exchange of Particles and Gaseous Pollutants. In: ERDA Symposium Series, vol. 38, pp. 706–727. Fülöp, M., Ragan, P. (1997). In situ measurements of 137 Cs in soil by unfolding method. Health Phys. 72 (6), 923– 930. Gering, F., Kiefer, P., Fesenko, S., Voigt, G. (2002). In situ gamma ray spectrometry in forests: determination of kerma rate in air from 137 Cs. J. Environ. Radioact. 61 (1), 75–89. Gilbert, R.O., Simpson, J.C. (1985). Kriging for estimating spatial pattern of contaminants: Potential and problems. Environ. Monit. Assess. 5, 113–135.
In situ and airborne gamma-ray spectrometry
445
Gilbert, R.O., Eberhardt, L.L., Fowler, E.B., Essington, E.H. (1974). Statistical design aspects of sampling soil for plutonium. In: Atmosphere—Surface Exchange of Particles and Gaseous Pollutants. In: ERDA Symposium Series, vol. 38, pp. 689–705. Govindara, K. (1989). In: Analytical Standards of Minerals, Ores and Rocks. Geostandards News Letter (Special Issue). ISSN: 0150-5505. Grasty, R.L. (1979). One flight snow-water equivalent measurement by airborne gamma-ray spectrometry. In: Proc. Int. Symp. on Remote Sensing of Snow and Soil Moisture by Nuclear Means. World Meteorological Organisation, Geneva. INIS-Mf-5776, pp. 105–112. Grasty, R.L. (1980). The search for COSMOS-954. In: Proceedings of the NATO Conference on Search Theory and Applications, Portugal, 1979. Plenum Press, New York/London, pp. 211–220. Grasty, R.L. (1981). Skyshine and the calibration of ground gamma-ray spectrometry. Geol. Soc. Can. Curr. Res. Part A, 80-1-A. Grasty, R.L., Walters, B.R.B., Hovgaard, J., LaMarre, J.R. (2001). Calibration of a 7.6 cm × 7.6 cm (3 inch × 3 inch) sodium iodide gamma ray spectrometer for air kerma rate. Radiat. Protect. Dosim. 94 (4), 309–316. Guillot, L. (2001). Extraction of full absorption peaks in airborne gamma-spectrometry by filtering techniques couped with a study of the derivatives. Comparison with the window method. J. Environ. Radioact. 53, 381–398. Gutierrez, S., Guillot, L., Bourgeois, C. (2002). Contributions of a germanium detector in mobile gamma-ray spectrometry. Spectral analysis and performance. Nucl. Instrum. Methods Phys. Res. A 482, 425–433. Hadley, M.J., Ruffell, A., Leslie, A.G. (2000). Gamma-ray spectrometry in structural correlations: an example from the Neoproterozoic Dalradian succession of Donegal (NW Ireland). Geol. Mag. 137 (3), 319–333. Helfer, I.K., Miller, K.M. (1988). Calibration factors for the Ge detectors used for field spectrometry. Health Phys. 55 (1), 15–29. Hendricks, P.H.G.M., Limburg, R.J., de Meijer, R.J. (2001). Full spectrum analysis of natural gamma ray spectra. J. Environ. Radioact. 53, 365–380. Hewson, R.D., Taylor, G.R. (2000). An investigation of the geological and geomorphological features of fowlers gap using thermal infrared, radar and airborne geophysical remote sensing techniques. Rangeland J. 22 (1), 105–123. Hillman, U., Schimmak, W., Jacob, P., Bunzl, K. (1996). In situ gamma spectrometry several years after deposition of radiocesium. 1. Approximation of depth distributions by the Lorentz function. Radiat. Environ. Biophys. 35 (4), 297–303. Hovgaard, J. (2000). NASCD—A new method for processing airborne spectra gamma ray. In: Sanderson, D.C.W., McLeod, J.J. (Eds.), Recent Applications and Developments in Mobile and Airborne Gamma Spectrometry: Proceedings of the RADMAGS Symposium. SURRC, University of Glasgow. Hubbell, J.H. (1982). Photon mass attenuation and energy absorption coefficients from 1 keV and 20 keV. Int. J. Appl. Radiat. Isotopes 33, 1269–1290. IAEA (1976). Radiometric Reporting Methods and Calibration in Uranium Exploration. Tech. Report Series, vol. 174. IAEA, Vienna. IAEA (1979). Gamma-Ray Surveys in Uranium Exploration. Tech. Report Series, vol. 186. IAEA, Vienna. IAEA (1990). The Use of Gamma Ray Data to Define the Natural Radiation Environment. Tech. Report Series, vol. 566. IAEA, Vienna. IAEA (1991). Airborne Gamma Ray Spectrometer Surveying. Tech. Report Series, vol. 323. IAEA, Vienna, ISBN 920-125 291-9. ICRU (1994). In-Situ Gamma-Ray Spectrometry in the Environment. International Committee for Radiological Units, Report No. 53. Jacob, P., Herwig, P., Herwig, G. (1986). Gamma-ray exposure from contaminated soil. Nucl. Sci. Eng. 93, 248–261. Jacob, P., Meckbach, Paretzke, H.G., Likhtarev, I., Los, I., Kovgan, L., Komarikov, I. (1994). Attenuation effects on the kerma rates in air after cesium depositions on grasslands. Radiat. Environ. Biophys. 33 (3), 251–267. Jones, D.G., Miller, J.M., Roberts, P.D. (1984). The distribution of Cs-137 in surface intertidal sediments from the Solway Firth. Mar. Pollut. Bull. 15, 187–194. Jones, D.G., Roberts, P.D., Miller, J.M. (1988). The distribution of gamma-emitting radionuclides in surface subtidal sediments near the Sellafield plant. Estuar. Coast. Shelf Sci. 27, 143–161. Jones, D.G., Roberts, P.D., Strutt, M.H., Higgo, J.J.W., Davis, J.R. (1999). Distribution of Cs-137 and inventories of Pu-238, Pu-239/240, Am-241 and Cs-137 in Irish Sea intertidal sediments. J. Environ. Radioact. 44, 159–189. Kachanoski, R.G. (1993). Estimating soil loss from the changes in soil caesium-137. Can. J. Soil Sci. 73, 629–632.
446
A.N. Tyler
Karlberg, O. (1990). In-situ gamma spectrometry of the Chernobyl fallout on urban and rural surfaces. STUDSVIK Nuclear Report 89/108, Sweden. Kastlander, J., Bargholtz, C. (2005). Efficient in situ method to determine radionuclide concentration in soil. Nucl. Instrum. Methods Phys. Res. A 547, 400–410. Knoll, G.F. (1989). Radiation Detection and Measurement, second ed., John Wiley and Sons, Canada, 754 pp. Kogan, R.M., Nazarov, I.M., Fridman, Sh.D. (1971). Gamma Spectrometry of the Natural Environment and Formations. Theory and Method Applications to Geology and Geophysics. Keter Press, Jerusalem. Available from the U.S. Department of Commerce—translated from Russian. National Technical Information Service, Springfield, VA. Korun, M., Likar, A., Lipoglavsek, M., Martincic, R., Pucelj, B. (1994). In situ measurement of Cs distribution in soil. Nucl. Instrum. Methods Phys. Res. B 93 (4), 485–491. Laedermann, J.P., Byrde, F., Murith, C. (1998). In-situ gamma-ray spectrometry: the influence of topography on the accuracy of activity determination. J. Environ. Radioact. 38 (1), 1–16. Lindahl, I., Habrekke, H. (1987). Mapping the radioactive fallout after the Chernobyl accident. Geoexploration 24 (3), 267. Lindsay, R., de Meijer, R.J., Joseph, A.D., Motlhabane, T.G.K., Newman, R.T., Tsela, S.A., Speelma, W.J. (2004). Measurement of radon exhalation from a gold-mine tailings dam by gamma-ray mapping. Radiat. Phys. Chem. 71 (3–4), 797–798. Livens, F.R., Baxter, M.S. (1988). Particle size and radionuclide levels in some west Cumbrian soils. Sci. Total Environ. 70, 1–17. Løvborg, L. (1982). Error analysis of calibration and field trials with spectrometers and counters. In: Symp. on Uranium Exploration Methods. OECD Nuclear Energy Agency in collaboration with IAEA, Vienna. Løvborg, L. (1983). Total Count Calibration Blocks for the Use in Uranium Exploration. Risø R-490, Risø Natl. Lab., Roskilde, Denmark. Løvborg, L., Wollenberg, H., Søernsen, P., Hansen, J. (1971). Field determination of uranium and thorium by gamma ray spectrometer, exemplified by measurement in the Ilimaussag, Alkali Intrusion, S Greenland. Econ. Geol. 88, 3. Løvborg, L., Grasty, R.L., Kirkegaard, P. (1978). A guide to the calibration constants for aerial gamma ray surveys in geoexploration. In: Aerial Techniques for Environmental Monitoring. Am. Nucl. Soc., New York. Malczewski, D., Teper, L., Dorda, J. (2004). Assessment of natural and anthropogenic radioactivity levels in rocks and soils in the environs of Swieradow Zdroj in Sudetes, Poland, by in situ gamma-ray spectrometry. J. Environ. Radioact. 73, 233–245. Mellander, H. (1989). Airborne gamma spectrometric measurements of the fallout over Sweden after the nuclear reactor accident at Chernobyl, USSR. IAEA/NENF/NM-89-1. Miller, K.M., Kuiper, J.L., Helfer, I.K. (1990). 137 Cs fallout depth distributions in forest soils versus field sites. Implication for external gamma dose rates. J. Environ. Radioact. 12, 23–47. Miller, K.M., Shebell, P., Klemic, G.A. (1994). In situ gamma ray spectrometry for the measurement of uranium in surface soils. Health Phys. 67 (2), 140–150. Parkinson, J.A., Horrill, D.A. (1984). Assessment of variation due to latitude and field conditions in the measurement of radionuclides. Nucl. Instrum. Methods Phys. Res. 223, 598–601. Pavlik, B., Engelsmann, J. (2004). Experience with airborne detection of radioactive pollution (ENMOS, IRIS). J. Environ. Radioact. 72, 203–211. Pickup, G., Marks, A. (2001). Regional scale sedimentation process models from airborne gamma ray remote sensing and digital elevation data. Earth Surf. Process. Landform 26, 273–293. Potts, M.J. (1978). Accuracy and precision in airborne radiometric measurement. In: Aerial Techniques for Environmental Monitoring. Am. Nucl. Soc., New York. 79/4327. Ruffel, A., McKinley, J.M., Lloyd, C.D., Graham, C. (2006). Tgh/K and Th/U ratios from spectral gamma-ray surveys improve the mapped definition of subsurface structures. JEEG 11 (1), 53–61. Rybacek, K., Jacob, P., Macklach, R. (1991). In situ determination of deposition radionuclide activities: Improvement of the method by deriving depth distributions from the measured photon spectra. Health Phys. 62, 519–528. Rybach, L., Bucher, B., Schwarz, G. (2001). Airborne surveys of Swiss nuclear facility sites. J. Environ. Radioact. 53 (3), 291–300. Sanderson, D.C.W., Ferguson, J.M. (1997). The European capability for environmental airborne gamma ray spectrometry. Radiat. Protect. Dosim. 73 (1–4), 213–218.
In situ and airborne gamma-ray spectrometry
447
Sanderson, D.C.W., McLeod, J.J. (1999). European Coordination of Environmental Airborne Gamma Ray Spectrometry. Final Report of EC Concerted Action, Contact No. F14P-CT95-0017. Sanderson, D.C.W., McLeod, J.J. (Eds.) (2000). Recent Applications and Developments in Mobile and Airborne Gamma Spectrometry: Proceedings of the RADMAGS Symposium, University of Stirling, 15–18 June. SURRC, University of Glasgow. Sanderson, D.C.W., Baxter, M.S., Scott, E.M. (1989). The use and potential of radiometrics for monitoring environmental radioactivity. In: COGER Conferences at the University of Lancaster, UK. Sanderson, D.C.W., Allyson, J.D., Tyler, A.N., Scott, E.M. (1995). Environmental applications of airborne gamma spectrometry. In: Application of Uranium Exploration Data and Techniques in Environmental Studies. Proceedings of a Technical Committee Meeting, Vienna, 9–12 November 1993. IAEA, Vienna, pp. 71–91. IAEATECDOC-827. Sanderson, D.C.W., Allyson, J.D., Cresswell, A.J. (2000). The use of combined Ge and NaI systems for airborne and vehicular surveys. In: Sanderson, D.C.W., McLeod, J.J. (Eds.), Recent Applications and Developments in Mobile and Airborne Gamma Spectrometry: Proceedings of the RADMAGS Symposium, University of Stirling, 15–18 June. SURRC, University of Glasgow. Sanderson, D.C.W., McLeod, J.J., Ferguson, J.M. (2001). A European bibliography on airborne gamma-ray spectrometry. J. Environ. Radioact. 53, 411–422. Sanderson, D.C.W., Cresswell, A.J., McLeod, J.J., Giannitrapani, M., Scott, E.M. (2002). Resumé 2002—International Intercomparison Exercise. Report of Phase 1 Pre-characterization Conducted November 2001. Scottish Universities Environmental Research Centre Report, Glasgow, 49 pp. Sanderson, D.C.W., Cresswell, A.J., McLeod, J.J., Giannitrapani, M., Scott, E.M. (2003). Report on phase 1 precharacterization conducted November 2001. In: Proceedings of an International Comparison of Airborne and Ground Based Gamma Ray Spectrometry. Results of the ECCOMAGS 2002 Exercise held 24 May to June 2002. University of Glasgow, Dumfries and Galloway, Scotland, ISBN: 0-85261-783-6, pp. 319–370. Sanderson, D.C.W., Cresswell, A.J., Hardeman, F., Debauche, A. (2004a). An airborne gamma-ray spectrometry survey of nuclear sites in Belgium. J. Environ. Radioact. 72, 213–224. Sanderson, D.C.W., Cresswell, A.J., Scott, E.M., Lang, J.J. (2004b). Demonstrating the European capability for airborne gamma spectrometry: results from the ECCOMAGS Exercise. Radiat. Protect. Dosim. 109, 119–125. Schwarz, G.F., Klingele, E.E., Rybach, L. (1992). How to handle rugged topography in airborne gamma-ray spectrometry surveys. First Break 10 (1), 11–17. Shebell, P., Faller, S., Monetti, M., Bronson, F., Hagenauer, R., Jarrell, C.L., Keefer, D., Moos, J.R., Panzarino, N., Reimanm, R.T., Sparks, B.J., Thisell, M. (2003). An in situ gamma-ray spectrometry intercomparison. Health Phys. 85 (6), 662–677. Sowa, W., Martini, E., Gehrike, K., Marsher, P., Naziry, M.J. (1989). Uncertainty of in-situ gamma spectrometry for environmental monitoring. Radiat. Protect. Dosim. 27 (2), 93–101. Storm, E., Israel, H.I. (1970). Photon cross sections from 1 keV to 100 MeV for elements Z = 1 to Z = 100. Nucl. Data Tables A 7, 565–681. Sutherland, R.A., de Jong, E. (1990). Statistical analysis of γ -emitting radionuclide concentrations for three fields in southern Saskatchewan, Canada. Health Phys. 59 (4), 417–428. Thummerer, S., Jacob, P. (1998). Determination of depth distributions of natural radionuclides with in situ gamma ray spectrometry. Nucl. Instrum. Methods Phys. Res. A 416, 161–178. Toivonen, H. (2004). Airborne gamma spectrometry—towards integration of European operational capability. Radiat. Protect. Dosim. 109, 137–140. Tompson, I.M.G., Bøtter-Jensen, L., Deme, S., Pernicka, F., Sáez-Vergara, J.C. (1999). Technical recommendations on measurements of external environmental gamma radiation doses. A report of EURADOS Working Group 12 “Environmental radiation monitoring”. Radiat. Protect. 12, 191 pp. Tyler, A.N. (1994). Environmental influences on gamma ray spectrometry. Ph.D. Thesis, Glasgow University. Tyler, A.N. (1999). Monitoring anthropogenic radioactivity in salt marsh environments through in situ gamma ray spectrometry. J. Environ. Radioact. 45 (3), 235–252. Tyler, A.N. (2004). High accuracy in situ radiometric mapping. J. Environ. Radioact. 72, 195–202. Tyler, A.N., Heal, K.V. (2000). Predicting areas of Cs-137 loss and accumulation in upland catchments. Water Air Soil Pollut. 121, 273–291. Tyler, A.N., Copplestone, D. (2007). Preliminary results from the first national in-situ gamma spectrometry survey of the United Kingdom. J. Environ. Radioact., submitted for publication.
448
A.N. Tyler
Tyler, A.N., Sanderson, D.C.W., Scott, E.M., Allyson, J.D. (1996a). Investigations of spatial variability and fields of view in environmental gamma ray spectrometry. J. Environ. Radioact. 33 (3), 213–235. Tyler, A.N., Sanderson, D.C.W., Scott, E.M. (1996b). Estimating and accounting for 137 Cs source burial through in situ gamma spectrometry in salt marsh environments. J. Environ. Radioact. 33 (3), 195–212. Tyler, A.N., Davidson, D.A., Grieve, I.C. (2001). In situ radiometric mapping of soil erosion and field-moist bulk density on cultivated fields. Soil Use Manage. 17, 88–96. Veen, A.W.L., Mejer, R.J. (1989). Radionuclide levels at two sites in a water extraction area in the Netherlands after Chernobyl. Water Air Soil Pollut. 44, 83–92. Wilford, J.R., Bierwirth, P.N., Craig, M.A. (1997). Application of airborne gamma-ray spectrometry in soil-regolith mapping and applied geomorphology. AGISO J. Austral. Geol. Geophys. 17 (2), 201–216. Williams, D., Cambray, R.S., Maskell, S.C. (1957). An airborne radiometric survey of the Windscale area, October 19–22nd. AERE Report R-2890. Winkelmann, I., Thomas, M., Vogl, K. (2001). Aerial measurements on uranium ore mining, milling and processing areas in Germany. J. Environ. Radioact. 53, 301–311. Winkelmann, I., Stoble, C., Thomas, M. (2004). Aerial measurements of artificial radionuclides in Germany in case of a nuclear accident. J. Environ. Radioact. 72 (1–2), 225–231. Wissman, F., Sáez Vergara, J.C. (2007). Dosimetry of environmental radiation—A report on the achievements of Eurados WG 3. Radiat. Protect. Dosim., in press. Zhang, X.B., Higgitt, D.L., Walling, D.E. (1990). A preliminary assessment of the potential for using cesium-137 to estimate rates of soil-erosion in the Loess Plateau of China. Hydrol. Sci. J. 35, 243–252. Zombori, P., Andrrasi, A., Nemeth, I. (1992). In situ gamma spectrometric measurements of the contamination in some selected settlements of Byelorussia (BSSR), Ukraine (UkrSSR) and the Russian Federation (RSFSR). J. Environ. Radioact. 17, 97–106.
449
Underwater gamma-ray spectrometry Pavel P. Povineca,b,∗ , Iolanda Osvatha , Jean-François Comanduccia a International Atomic Energy Agency, Marine Environment Laboratory, Monaco b Comenius University, Faculty of Mathematics, Physics and Informatics, Bratislava, Slovakia
1. Introduction It was recognized at a very early stage in the development of detectors for environmental radioactivity monitoring that a complementary detection method to traditional sampling and laboratory analysis can be via direct, in situ analysis of radionuclides in the aquatic environment (Wedekind, 1973). With the development of NaI(Tl) detectors and later also cryogenic HPGe detectors (Povinec et al., 1995) it was possible to carry out underwater analysis of gamma-emitting radionuclides at sensitivities good enough for fresh water as well as marine radioactivity studies (Povinec, 2004). As underwater gamma-ray spectrometry was recently reviewed by Jones (2001), we present here only a short literature survey, and concentrate more on results obtained by the International Atomic Energy Agency’s Marine Environmental Laboratory (IAEA-MEL). The systems most commonly used for underwater γ -ray spectrometry are based on NaI(Tl) (or BGO—bismuth germanate) and HPGe detectors. The advantage of NaI(Tl)-based systems is firstly related to the high detection efficiency of NaI(Tl) crystals at much lower cost than equivalent Ge crystals. Moreover, NaI(Tl)-based systems can be built sufficiently robust and are adequate for long-term underwater deployment. In addition polycrystalline NaI(Tl) materials can be obtained with identical optical characteristics but far better thermal and mechanical strength than single crystals, thus further improving the shock resistance and thermal stability of the system. The drawbacks of NaI(Tl)-based systems are the high power consumption for the operation of the photomultiplier tube and a relatively poor energy resolution of the detector (with peaks tens of times wider than those in spectra recorded with Ge detectors). HPGe-based systems, on the other hand, have the advantage of good energy resolution and hence excellent radionuclide identification capability. However, in terms of ruggedness, autonomy and power consumption, they generally perform less well than NaI(Tl)-based systems. Recently electrically cooled HPGe γ -ray detectors using Stirling cycle refrigerators have been ∗ Corresponding author. Present address: Comenius University, Faculty of Mathematics, Physics and Informatics,
Mlynska dolina F1, SK-84248 Bratislava, Slovakia. E-mail address:
[email protected] RADIOACTIVITY IN THE ENVIRONMENT VOLUME 11 ISSN 1569-4860/DOI: 10.1016/S1569-4860(07)11014-7
© 2008 Elsevier B.V. All rights reserved.
450
P.P. Povinec et al.
developed (Katagiri and Itoh, 1994; Katagiri et al., 1995). The size of such a detector assembly including the crystal and the refrigerator is very small (38 × 12 × 12 cm) and its power consumption extremely low (of the order of 1 W). A submersible electrically cooled HPGe γ -ray detector system of this type, operational to 500 m depth, was successfully tested by Kobayashi et al. (1998). Due to various performance and/or cost reasons, other available types of detectors, such as CsI crystals (pure or doped with Tl or Na), fluorides (e.g., CsF, BaF2 ) or semiconductor detectors which do not require external cooling (e.g., CdTe, CdZnTe, HgI2 , GaAs), have in general not become competitive with NaI(Tl) and HPGe detectors mainly because of their lower efficiency for the detection of gamma-rays in the energy interval 30–3000 keV. The development and use of a BGO-based γ -ray spectrometer was reported by Jones (2001). Underwater gamma-ray spectrometry has clear advantages over traditional sampling and laboratory analysis for a number of applications, including: (i) Gamma-emitter mapping of large river, lake and seabed areas to estimate the levels and distributions of natural and/or anthropogenic radionuclides, with applications in geological mapping, mineral exploration and contamination assessment studies (e.g., Miller et al., 1982; Noakes and Harding, 1982; Thomas et al., 1984; Jones et al., 1988; Povinec et al., 1996; Noakes et al., 1999; Jones, 2001). In the case of mineral exploration both passive and neutron-induced gamma-spectrometry are of interest (Senftle et al., 1976; Thomas et al., 1983). (ii) Optimization and focusing of conventional sample collection. An underwater γ -ray survey performed in conjunction with a sampling campaign provides a cost-effective method for obtaining representative samples from the investigated area. It offers real-time information on the presence and levels of radionuclides of interest, thus replacing timeconsuming sampling, sample preparation and laboratory analysis operations (IAEA, 1998b; Osvath et al., 1999a, 1999b; Osvath and Povinec, 2001; Osvath et al., 2001). (iii) Investigation of the radiation field in the vicinity of sunken objects. There is no alternative to in situ measurement for providing information on the presence and nature of radioactive material contained in sunken or dumped radioactive objects (Povinec et al., 1995, 1997a; Sjöblom et al., 1999). (iv) Continuous analysis of gamma-ray emitters in seawater and fresh water in the vicinity of sources or potential sources of release of radionuclides (Wedekind, 1973; Povinec et al., 1995, 1996; Harms and Povinec, 1998, 1999; Wedekind et al., 1999, 2000; Povinec et al., 2001a; Nies and Schilling, 2003; van Put et al., 2004; Osvath et al., 2005). Such monitoring can provide early warning at nuclear facilities with authorized discharges to the aquatic environment (e.g., nuclear reprocessing plants, nuclear power plants), at underwater nuclear waste dumpsites or at sites where nuclear substances were accidentally introduced to the marine environment (e.g., accidentally sunken nuclear vessels). (v) Screening of contaminated areas in emergency situations. This particular application of underwater γ -ray spectrometry is related to accidental releases from nuclear facilities, directly or indirectly affecting the marine environment. In such emergency situations, underwater γ -ray spectrometry can be used to estimate the increases in radionuclide levels and the extent of the contaminated area of seabed sediment (detector operating on the bottom) and volume of water (detector operating in water), but also to monitor the progression of the plume of pollutant.
Underwater gamma-ray spectrometry
451
(vi) Continuous monitoring of natural radionuclides in the aquatic environment—for example, investigations of radon daughter products in groundwater, pre-earthquake studies and submarine groundwater discharge studies (Povinec et al., 2006a, 2006b). In situ gamma-ray spectrometry has been recognized by IAEA-MEL as a powerful tool for continuous analysis of gamma-ray emitters in seawater. Following the 1991–1998 period of intensive international investigations of the Arctic environment in relation to radioactive wastes dumped in the Kara and Barents Sea, carried out in the framework of the IAEA’s International Arctic Seas Assessment Project (IASAP) (Sjöblom et al., 1999), it was recommended that in situ monitoring of underwater dumpsites for potential releases to the marine environment would be the best option (Harms and Povinec, 1998, 1999). The installation of remotely operated monitors which would transmit data through a satellite in real-time was identified as a cost-effective way of providing timely information on eventual changes in the radionuclide levels in seawater. In fact, this is the sole approach allowing continuous observation of the environmental radionuclide levels and immediate alert. It also considerably reduces the requirements for sampling expeditions and laboratory-based analytical work (Hamilton et al., 1994; Povinec et al., 1997a, 1997b). IAEA-MEL has contributed to the development of prototype NaI(Tl) and HPGe γ -ray spectrometers and has applied movable and stationary in situ γ -ray spectrometers for surveys at radioactive waste dumping sites in the Kara Sea (Povinec et al., 1997a, 1997b), seabed mapping in the Irish Sea (Osvath et al., 2001; Osvath and Povinec, 2001) and at the atolls of Mururoa and Fangataufa (Osvath et al., 1999a, 1999b), stationary seawater monitoring in the eastern Irish Sea (Osvath et al., 2005) and submarine groundwater discharge studies (Povinec et al., 2006a, 2006b, 2006c). In this chapter we shall review the characteristics of underwater gamma-ray spectrometers and discuss the main results obtained during their deployment in different environments.
2. Mobile detection systems 2.1. Dual HPGe–NaI(Tl) detector system The system consisted of a 20% relative efficiency (i.e. relative to a 76 × 76 mm NaI(Tl) crystal) HPGe crystal cooled with propane. The HPGe probe was prepared before deployment by cooling with a He pump to 20 K (and creating a vacuum of 3 × 10−4 bar between the propane reservoir and the system’s outer casing). The detector after cooling operated up to temperatures of 120 K, giving an operational time up to 24 h. The 100 × 150 mm NaI(Tl) crystal was housed in a stainless steel–titanium tube together with part of the spectrum acquisition and processing electronics, which were connected both to the preamplifier of the HPGe detector and to a shipboard computer and power supply through a 1200 m long coaxial cable, used for the transmission of both data (digitized, via MODEM) and power. The shipboard computer emulated the submerged one, allowing control of data acquisition and real-time spectrum processing. The submerged electronics included two Target/Nucleus® multichannel analyzer boards of 2 and 8 channels for the NaI(Tl) and HPGe detector, respectively. The resolution was 8% for the NaI(Tl) detector (for 662 keV of 137 Cs) and 4 keV for the HPGe detector (for 1332 keV of 60 Co). The two detectors and the submerged electronics were enclosed in
452
P.P. Povinec et al.
a polyethylene sledge, protecting the assembly from mechanical shock while being towed on the seabed (Povinec et al., 1995, 1996). The detection system was calibrated with point sources using Monte-Carlo simulation. For the studies described below, post-survey calibration using radionuclide profiles measured in sediment cores was required. During fieldwork, energy calibration was performed as a rule prior and subsequent to each deployment on deck by using a set of point sources. Background spectra were recorded both in water and on deck (to check for eventual contamination of the detector housing). Gamma-spectra were treated with commercial analysis software (EURISYS Interwinner® , EG&G ORTEC GammaVision® ). A package of custom-developed codes in GW and Visual Basic was used for treating NaI(Tl) spectra and survey data. Contour maps were produced using the commercial software Surfer® for Windows and the GEBCO 97 Digital Atlas. 2.2. Small NaI(Tl) detector IAEA-MEL has also been operating a second NaI(Tl)-based system, consisting of a submersible probe, a shipboard computer and a DC power supply (Osvath and Povinec, 2001). The stainless steel tubular probe casing contains a 152 × 51 mm NaI(Tl) ruggedized crystal, a high gain photomultiplier tube, pressure, temperature and seabed roughness sensors and electronic circuit boards which provide power to the sensors and process the signals (Figure 1). The probe is operational down to depths of 2000 m. Power from the shipboard stabilized DC power supply is provided to the probe through a co-axial cable. The digitized signal from the sensors is transferred to the shipboard computer through the same cable, using frequency shift key (FSK) techniques. This computer contains a plug-in board, with an autonomous Intel 486 DX-66 microcomputer, which performs the role of a multichannel analyzer. Data are transferred from the probe to the logging computer every second and can be sequentially stored
Fig. 1. Small NaI(Tl) underwater gamma-ray spectrometer with electronics.
Underwater gamma-ray spectrometry
453
with this frequency or at other user-selected time intervals. A 300-channel spectrum is generated, covering the energy range from 100 to 3000 keV. The logging program on the PC allows the monitoring of the spectrum recording and of the total count rate, bottom ruggedness, pressure and temperature. The resulting records are used to control system functioning and the data quality both in real time and during later processing of results. The energy calibration of the gamma-ray spectrometer was carried out using 137 Cs and 60 Co radioactive sources. The energy resolution for 662 keV (137 Cs) gamma-rays was 6.5%. The efficiency calibration was done using a 137 Cs source dispersed in a polyethylene tank 1 m in diameter and 1 m high, filled with fresh water. The tank was also filled with seawater with natural concentrations of 40 K (1462 keV), 226 Ra, 222 Rn and its daughter products. 222 Rn being a pure alpha-emitter (half-life 3.83 d), decays to 218 Po (half-life 3.05 m), which then decays to 214 Pb (half-life 26.8 m) and then to 214 Bi (19.7 m), which has been used as the most suitable gamma-ray emitter for analysis of 222 Rn in environmental samples, as it emits highintensity gamma-rays of different energies. The corresponding 214 Bi peaks used in spectra evaluations were either the 609, 1120 or 1765 keV peaks, depending on background conditions during real measurements, assuring the best factor of merit (the ratio εγ /(b)−1/2 , where ε is the photopeak efficiency, γ is the photon emission intensity, and b is the corresponding background). Background measurements were carried out with the detector immersed in the tank filled with fresh water (Povinec et al., 2006b).
3. Stationary detection system The stationary underwater monitoring system, also called “NEMO” (Nautic Environment Marine Observatory) is shown in Figure 2. It records gamma-ray spectra in water together with a suite of environmental parameters including seawater temperature, conductivity, current speed and direction (Aakenes, 1995; Povinec et al., 2001b; Osvath et al., 2005). These data may be transmitted via a satellite link to a base station at the IAEA-MEL at Monaco. The monitor’s capabilities can be further expanded by adding sensors for meteorological (wind speed and direction, air temperature and pressure) and oceanographic measurements (wave height, direction and period, water temperature and salinity profiles, light attenuation, Chlorophyll-A and hydrocarbons). Such a complex data set can be useful as input to transport models. The monitoring station’s general technical characteristics are given below: (i) Radionuclides in seawater are analyzed using a RADAM (radioactivity detection and measuring system) sensor which comprises a NaI(Tl) detector with a ∅ 76 × 76 mm ruggedized crystal with 7% resolution for 662 keV gamma-rays (137 Cs), a 512-channel multichannel analyzer and a power supply. The energy range used is 50–1800 keV. The limit of detection for 137 Cs in water is 19 Bq m−3 for a 24 h spectrum integration time, 7 Bq m−3 for 7 days and 4 Bq m−3 for 30 days integration time. (ii) The current speed and direction are measured with a 3-axis ultrasonic transducer, with ranges of 0–300 cm s−1 and 0–360◦ and uncertainties of 3% and 2◦ , respectively. The temperature sensor has an operational range of −5 to +45 ◦ C with an uncertainty of 0.1 ◦ C. Salinity is determined from conductivity measurements performed with an electrode-less induction type cell, with an operational range of 2–77 m cm−1 and an uncertainty of 0.06 m cm−1 .
454
P.P. Povinec et al.
Fig. 2. The NEMO buoy in the open sea.
(iii) The sensors are suspended 4 m underwater on a structure attached to a floating buoy (Figure 3). The buoy is equipped with signal-processing electronics including a microcomputer (GENI) and a marine transceiver, and with a set of lead–acid batteries. An integrated GPS receiver continuously records the buoy’s position. These instruments and batteries are maintained in a nitrogen atmosphere in an air-tight/water-tight instrument container. The Observatory’s autonomy is supported by a set of solar panels installed on the instrument container. The satellite antenna and visual signaling devices are mounted on the top deck, about 4 m above water level. (iv) Duplex data transmission is facilitated by the Inmarsat-C Atlantic Ocean Region East (AOR-E) satellite, via Eik (Norway) Land Earth Station. A PC, based at the IAEA-MEL in Monaco, controls the electronics on board the buoy and receives and processes all measurement data. Besides being transmitted via satellite the data are also stored in the GENI. The RADAM spectrometer is designed for minimizing power consumption and maximizing its autonomy (Figure 4). It consumes 10 mW in idle mode and a maximum of 1.5 W while operating. The input voltage is 11–20 V. It contains three electronic boards: • the sensor board, which generates the necessary voltages for the system (except the high voltage, which is generated separately inside the RADAM), processes the pre-amplifier signal and provides electronic first-order compensation for the temperature dependence of the NaI(Tl) crystal response; • the AD conversion board, based on a 12 bit sampling ADC, with a total conversion time of 25 µs; • the CPU board, which reads and stores AD converter data and transmits data over the serial port to the main onboard computer (GENI). During operation the RADAM is placed 4 m under the water level and is connected to the power supply and the GENI through marine connectors and cables. Its housing, made
Underwater gamma-ray spectrometry
455
Fig. 3. Schematic construction of the NEMO system.
Fig. 4. RADAM spectrometer.
of POM C/Hostaform, is watertight down to 10 m. Besides the hardware temperature compensation mentioned above, software compensation is performed using the 137 Cs peak. The calibration of the RADAM was carried out in a tank filled with seawater spiked with 137 Cs. The efficiency was estimated as 2.1 × 10−4 s−1 /(Bq m−3 ) net count rate under the 137 Cs photopeak.
456
P.P. Povinec et al.
Fig. 5. In situ gamma-ray spectrum of seabed sediment as measured with an underwater HPGe detector.
The NEMO monitoring station is suited to operation in ice-free waters in latitudinal regions with sufficient light to recharge the batteries. Deployment sites should be chosen at locations where the buoy is easily serviceable or the budget should provide for service from large ships and/or replacement buoys.
4. Deployment of underwater gamma-ray spectrometers 4.1. Dual HPGe–NaI(Tl) spectrometer in the Irish Sea The spectrometer was tested in the Menai Strait of the Irish Sea. Figure 5 shows a typical gamma-ray spectrum of sediment as measured in the Irish Sea with underwater HPGe detector situated about 40 m below the sea surface. The obtained spectrum represents a first gamma-ray spectrum reported with an underwater HPGe detector. The dominant radionuclides observed in the spectrum are the anthropogenic 137 Cs (due to its releases from the Sellafield reprocessing facility) and natural 40 K (Povinec et al., 1995, 1996). 4.2. Investigation of radioactive waste dumping sites in the Kara Sea In the framework of the IAEA’s IASAP (International Arctic Seas Assessment Project) project, IAEA-MEL carried out a mapping of anthropogenic radionuclides in Novaya Zemlya bays where dumping of radioactive wastes was carried out by the former Soviet Union (Baxter et al., 1997; Sjöblom et al., 1999). Figure 6 shows a HPGe gamma-ray spectrum of sediment measured in Stepovovo Bay, confirming a leakage of 137 Cs from dumped containers filled with radioactive wastes (Povinec, 1997; Povinec et al., 1997a, 1997b). The observed distribu-
Underwater gamma-ray spectrometry
457
Fig. 6. In situ gamma-ray spectrum of seabed sediment at Stepovovo Bay of Novaya Zemlya.
Fig. 7. Distribution of 137 Cs in seabed sediment at Abrosimov Bay of Novaya Zemlya.
tion of 137 Cs in the sediment of Abrosimov Bay (Figure 7) shows that massic activities up to 7000 Bq kg−1 dry weight (d.w.) occur in the Bay (Osvath et al., 1999a). 4.3. Mapping of radionuclides in Mururoa and Fangataufa lagoons sediments IAEA-MEL participated in the IAEA’s international study of the contamination with anthropogenic radionuclides of the Mururoa and Fangataufa lagoons and surrounding ocean area
458
P.P. Povinec et al.
Fig. 8. Distribution of 60 Co in Fangataufa lagoon sediments.
with the aim of providing an independent assessment of the environmental contamination (Povinec et al., 1999). The coralligenous sediment inside the lagoons was mainly contaminated following 41 atmospheric nuclear tests and 5 atmospheric safety trials with a total yield of 10 Mt of TNT. In particular three tests in Mururoa and one in Fangataufa, carried out in 1966–1967 from barges on the lagoon, 0–10 m above water level, resulted in higher levels of contamination of sediment by transuranics (e.g., in the order of 106 Bq kg−1 dry weight (d.w.) for 239,240 Pu) and gamma-emitters (e.g., in the order of 103 Bq kg−1 d.w. for 60 Co and 137 Cs) in two areas inside the Mururoa lagoon (Dindon and Denise zones) and one in the Fangataufa lagoon (Fregate zone) (Povinec et al., 1999). The size of these hot spots was below 1 km2 (Musa et al., 1996). The distribution (Masse and Musa, 1996) and thickness of sediment varies considerably inside the lagoons, and so does the in-depth distribution of radionuclides, which at certain locations penetrate down to over 4 m in the sediments. Based on extensive sampling and analytical campaigns carried out previously (Musa et al., 1996), inventories of radionuclides in sediment had been estimated for the lagoons at large and for the identified hot spots. Underwater γ -ray spectrometry was used to guide sampling in the areas of the hot spots. The small NaI(Tl) spectrometer was deployed at 16 sites in Mururoa and 12 sites in Fangataufa at water depths between 30 and 45 m. Static deployment rather than towing was preferred to minimize the risk of snagging the spectrometer on corals. 60 Co and 137 Cs were readily identifiable in the spectra recorded. The presence of 60 Co in the environment of the lagoons is due to neutron activation of structural materials during the atmospheric nuclear tests, while 137 Cs is a fission product. Maximum 60 Co concentrations of 1.4 kBq kg−1 d.w. had been recorded in Mururoa surface sediment (Dindon zone) and 2.4 kBq kg−1 d.w. in Fangataufa (Fregate zone) (Musa et al., 1996). During the 1996 gamma-ray survey the count-rates under the 60 Co photopeaks (1173.2 and 1332.5 keV) were used, together with the full energy range countrate, to identify the most contaminated sites. Based on a simple calibration obtained with 60 Co profiles in sediment cores collected from three of the in situ measurement locations, estimates of 60 Co concentrations in surface sediment were given. Figure 8 shows that activities of 60 Co up to 60 kBq m−2 were measured in the Fregate zone of the Fangataufa lagoon (Osvath et
Underwater gamma-ray spectrometry
459
al., 1999b). The Dindon zone at Mururoa showed even higher activities, up to 400 kBq m−2 . A fairly good correlation (r 2 = 0.8) was found between 60 Co and 239,240 Pu concentrations measured in sediment cores collected at the site (Povinec et al., 1999). Based on the 60 Co data, a 239,240 Pu inventory of 8 × 1012 Bq was estimated for the Dindon hot spot, which is in reasonable agreement with the value of 4.9 × 1012 Bq previously given by Musa et al. (1996). 4.4. Mapping of 137 Cs in Irish Sea sediments A sediment gamma-ray survey was carried out in the Irish Sea off the British Nuclear Fuels plc (BNFL) Sellafield reprocessing plant (formerly Windscale). This establishment is authorized to discharge low-level liquid radioactive waste into the sea. Discharges are released at about 20 m depth through a pipeline extending 2.1 km into the sea from the low water mark (Kershaw et al., 1992). Between 1952, when it started discharging, and 1998, Sellafield released over 40 PBq of 137 Cs, (Aarkrog, 2003). The discharges of 137 Cs peaked in 1977 with 4.5 PBq; the recent annual releases were below 1 TBq. The environmental behavior of the released radionuclides and their radiological impact have been extensively studied over several decades (Kershaw et al., 1992). One issue of particular interest is to estimate the export of radionuclides from the Irish Sea, and, in this context, to assess the remobilization of radionuclides from seabed sediment. Extensive sampling campaigns and analytical work (Woodhead, 1988; McCartney et al., 1994; Poole et al., 1997; Kershaw et al., 1999) were carried out to assess the distribution and inventories of gammaemitters and actinides in Irish Sea sediments. Gamma-ray surveys (Miller et al., 1982; Jones et al., 1984, 1988, 1999) were also undertaken to obtain better detail on the distribution patterns of gamma-emitting radionuclides. According to Jones et al. (1988, 1999) and McCartney et al. (1994), the general features of the distribution patterns are similar for all radionuclides. Based on this observation, results of hovercraft-based gamma-ray surveys were used to estimate inventories of alpha-emitting actinides in intertidal sediments (Jones et al., 1999). 137 Cs has been remobolized from Irish Sea sediments following the decrease of its concentration in water associated with the fall in discharges from BNFL Sellafield (McCartney et al., 1994). Hunt and Kershaw (1990) estimated that 1300 ± 700 TBq 137 Cs were remobilized from Irish Sea sediments in 1983–1988, while for the period 1988–1995 Poole et al. (1997) estimated a corresponding value between 350 and 573 TBq. Inventories of 137 Cs in Irish Sea sediment were previously estimated at about 1530 TBq in 1988 and 960 TBq in 1995 (Poole et al., 1997). In 1995 IAEA-MEL participated in a cruise in the eastern Irish Sea with the purpose of carrying out a gamma-ray survey of the bottom sediment (Osvath and Povinec, 2001). Based on the high resolution information resulting from towing a spectrometer on the bottom sediment, more accurate estimates of radionuclide distributions are obtainable than when using point samples. The NaI(Tl) underwater gamma spectrometer was towed at speeds between 2.5 and 4 knots on the seabed at water depths up to 45 m along an array of 14 transects (Figure 9) totaling almost 300 km. During the survey, over 1800 individual 1-min spectra were recorded. The ship’s GPS position, her speed and the water depth were recorded every minute, along with other parameters, by the ship’s logging system. The overall position accuracy was estimated to be within 30 m. The distance corresponding to the spectrum integration time varied between 80 and 120 m. A typical 1-min spectrum allowed immediate identification of
460
P.P. Povinec et al.
Fig. 9. Transects covered by the Irish Sea seabed γ -ray survey (locations of BNFL Sellafield and the discharge pipeline are represented on the map).
137 Cs. Besides 137 Cs and 40 K, no other radionuclides were detectable for this integration time.
A summed spectrum corresponding to 120 min integration time showed in addition the presence of natural radionuclides in the U–Ra and Th decay series. Spectrum analysis was carried out to estimate count rates corresponding to the full energy range, and gross and net 137 Cs and 40 K peak areas for individual 1-min spectra. No correction was applied for contributions from the water radioactivity, as these were estimated to be below 3% based on measurements taken with the probe towed in water. Calibration was carried out using a set of eight cores collected along the survey transects which were analyzed later at IAEA-MEL. In order to allow comparison with previously reported data, results were given for 137 Cs concentrations in surface sediments (r 2 = 0.9 for linear fit procedure using 137 Cs photopeak net counts in survey spectra and average concentrations in the top 5 cm sediment layer).
Underwater gamma-ray spectrometry
461
Fig. 10. 137 Cs, 40 K and integral counting rates along transects depicted in Figure 9.
The profiles of full energy range and gross 137 Cs and 40 K count rates along transects (a few examples of which are given in Figure 10) show that 137 Cs dominates the gamma-radiation field, with its profile closely followed by that of the total count rate. Isoline maps of 137 Cs count rate and dry weight concentration in surface sediment decay-corrected to June 1995 (Figure 11) were obtained from 1600 individual data points using the Kriging interpolation method. A combined uncertainty of 20% was estimated, including maximum 8% in situ counting statistical uncertainty and contributions from the positioning, calibration and interpolation procedures. An overlying patchy smaller-scale sediment contamination pattern is attributed to a multitude of factors related to discharge practices in conjunction with local variability of water flows, scavenging and sediment transport processes (Woodhead, 1988). On the regional scale, maximum levels were attained in the vicinity of the discharge point and to the north of it along the coast. This northerly pattern appears to be related to persistent northward trans-
462
P.P. Povinec et al.
Fig. 10. (continued)
port driven by topography and wind during strong northerly and westerly gales (Kershaw et al., 1992). The sediment type reinforces this effect in the area covered by the present survey, with sands dominating the area nearer to shore and to the south of the outfall and sandy mud offshore and northward of it. The steepest gradients were observed north of the outfall, perpendicular to the shoreline, with count rates decreasing to half of the maximum recorded value over a distance of less than 2000 m. Due to difficulty in towing nearshore, the survey did not cover the respective area, where, based on extensive sampling and analyses, radionuclide levels were reported to decrease in relation with the large grain-size of sediments (McCartney et al., 1994). A decrease in 137 Cs concentrations in the top 5 cm of sediment along the Cumbrian coast of approximately 75% between 1983 and 1988 was reported by McCartney et al. (1994). Jones et al. (1988) observed a decrease of 137 Cs levels of up to one order of magnitude in the time
Underwater gamma-ray spectrometry
463
Fig. 11. 137 Cs concentrations in surface sediment (Bq kg−1 d.w.). The contour plot extends over the area covered by the survey transects in Figure 9. Maximum concentrations are observed to the north of the outfall along the coast (the BNFL Sellafield discharge pipeline is represented to scale on the figure).
interval between their 1978 and 1985 sea-bed gamma-ray surveys. For the 1988–1995 time interval a 40% reduction in the total 137 Cs inventory was estimated by Poole et al. (1997) for an area off the Cumbrian coast which includes the area covered by this survey. For the same time period (1988–1995) and for a large part of the area of this survey, Kershaw et al. (1999) found decreases of approximately 25% in surface sediment concentrations of 239,240 Pu and 241 Am. As compared to the 1988 isoline map of 137 Cs concentrations in surface sediment given in McCartney et al. (1994), Figure 11 indicates that the 1995 levels decreased by 40– 70% in the area extending up to 10 km (in the upper part of map in Figure 11) to 20 km (lower part) offshore. In general, the gradient in reduction follows the gradient in concentration, that is, more important reductions were found in the higher 137 Cs level areas along the shore while lower levels (200 Bq kg−1 d.w. 137 Cs and below) further away have remained practically unchanged. This observation is consistent with the reported change of 137 Cs levels in sediment on a wider scale in the Irish Sea as related to its remobilization from sediment. The general distribution patterns are quite similar for 1995 and 1988, with steeper gradients nearshore north of Sellafield and off Saint Bees Head, but an important reduction in the size of the areas with relatively high concentrations can be observed. Maximum 137 Cs levels recorded during the 1995 sea-bed gamma-ray survey in surface sediment in the investigated area are about 900 Bq kg−1 d.w. and are restricted to a small patch some 2 km northwest of the outfall (on transect 8 of this survey), whereas for 1988 an extended area with levels above 1000 Bq kg−1 d.w. 137 Cs was reported by McCartney et al. (1994).
464
P.P. Povinec et al.
Fig. 12. Location of NEMO deployment site in the Irish Sea. The histograms superimposed on the map represent the evolution of 137 Cs concentrations in seawater as measured during 1985–2000 at various locations in the Irish Sea in the framework of RPII’s routine monitoring program (RPII, 2000).
4.5. In situ stationary monitoring of radionuclides in seawater in the NW Irish Sea A continuous remote monitoring of radionuclides in aquatic systems, where authorized discharges from nuclear facilities occur, or more generally, surveillance of aquatic environmental radionuclides in areas of interest and recording of time series of measurements, has been one of the driving forces for the development of this new technology. With the purpose of testing and developing this technology for further deployment at sites of interest, IAEA-MEL acquired a NEMO observatory equipped for gamma-ray spectrometry, oceanographic measurements and satellite data transmission. The NEMO was tested offshore of Monaco in 1999 (Povinec et al., 2001b) and was then deployed in the Irish Sea in the year 2000 as part of a joint program between IAEA-MEL and the Radiological Protection Institute of Ireland (RPII) (Osvath et al., 2005). The Irish Sea has important fishing grounds shared by several European countries and therefore the quality of its marine environment is being closely monitored. Radioactivity has been of interest particularly in relation to the authorized discharges from the Sellafield nuclear reprocessing plant on the Cumbrian coast. Measurement campaigns have been carried out regularly by British and Irish institutes (e.g., McCartney et al., 1994; Kershaw et al., 1999; RPII, 2000) to assess radionuclide levels and their dynamics. The NEMO was deployed in the open Irish Sea (53◦ 46.404 N; 5◦ 38.362 W) from August 2 to December 14, 2000 at 87 (+3) m water depth (Figure 12). The joint IAEA-RPII project aimed to provide continuous data on seawater radioactivity in an area usually monitored by RPII through annual sampling and analysis campaigns. In particular, the project would aim to measure the variation of 137 Cs levels, as 137 Cs being the main anthropogenic gammaemitter monitored by RPII in the marine environment in relation to the authorized releases of radionuclides from the Sellafield nuclear reprocessing facility.
Underwater gamma-ray spectrometry
465
Fig. 13. Gamma-ray spectra recorded in seawater, integrated over one month. The spectra have been slightly shifted along both axes with respect to each other to allow easier viewing.
The western Irish Sea gyre is a stably located seasonal circulation feature persistent during the summer months (Hill et al., 1994, 1996) discovered in 1990 and studied also in relation to valuable fisheries resources in the area. It develops in relation with a dome-like density structure of the water mass, present in the warm season (spring–summer) due to thermal stratification. In this area tidal currents are weak. The residual surface cyclonic flow entrains water southward along the Irish coast as described above. The gyre is thought to generate a retention mechanism for marine plankton and any conservative contaminants introduced into its central area, this potentially exposing the respective biota populations, including commercially valuable species and their consumers, to risk. It was therefore found of interest to locate the buoy’s mooring station in relation to this gyre as described further in this chapter. The deployment and simultaneous sampling was carried out using R/V Celtic Voyager, with assistance provided by the Marine Technical and Development Services (MTDS) based in Galway. The targeted area is intensively trawled and the risk of losing moored equipment is therefore considerable. The site of deployment was chosen in an area with wrecks on the bottom and therefore avoided by trawlers. For safety reasons, the mooring was constructed from steel chain and therefore required two large volume subsurface buoys to achieve buoyancy. Anti-fouling and anti-corrosion treatment and measures were applied to the monitoring system. A racon active radar target enhancer was added to the passive radar reflector to improve the buoy’s visibility to passing ships. Data were recorded 4 times a day following a synoptic time scheme (00:00, 06:00, 12:00 and 18:00 UTM). Over 350 6-h gamma-ray spectra were analyzed. In a typical 6-h spectrum the 1460.73 keV 40 K and the 609.3 keV 214 Bi lines are well visible. The monthly averaged output spectra are presented in Figure 13. 137 Cs is the only anthropogenic gamma-emitting radionuclide that could be identified in the spectra. The monthly mean activity 137 Cs concentrations in seawater at the deployment position during September, October and November were 17 ± 7, 19 ± 8 and 23 ± 10 Bq m−3 , respectively. This indicates that on the timescale of months no significant changes were detectable in 137 Cs concentration in seawater. Fluc-
466
P.P. Povinec et al.
Fig. 14. Gamma-ray spectrum recorded in water (6.11.2000, 18:00–24:00 UTM) displaying higher activity of U–Ra and Th series radionuclides as compared to typical gamma-ray spectra (see Figure 12).
tuations of 137 Cs levels over shorter time-scales would have been within the range of uncertainty of the determinations. The counting rate integrated over the whole energy range (50–1900 keV) is a sensitive parameter used for immediate information on any changes in the activity and prompting a detailed analysis of the gamma-ray spectrum. This counting rate has been noted to vary by as much as 15% due to variations of concentrations of natural radionuclides of the U–Ra and Th series in seawater (Figure 14). Since no significant simultaneous variation of 40 K activity is observed, it can be deduced that these variations are due to atmospheric washout by rainfall. Even higher such variations (up to 40% of the average counting rate) were observed during the buoy’s operation in a coastal environment, where atmospheric concentrations of radon and thoron daughters as well that of scavenging aerosols are higher than in the open sea (Povinec et al., 2001b). They were attributed to subsequent increases in seawater of the concentrations of cosmogenic 7 Be and of radionuclides in the decay chains of the primordial U–Ra and Th series observed in the spectra. The data for 137 Cs were in agreement with the monitoring data collected by the RPII during its routine marine monitoring program in the region (RPII, 2000), which are for recent years summarized in Figure 12. The NEMO observatory provided continuous information to complement the year 2000 sampling campaigns. The data recorded and received in real-time indicated that no significant changes occurred in the concentrations of 137 Cs at the mooring site in the period from September to November, 2000. Other anthropogenic gamma-emitters were not detected. The real-time seawater radioactivity, temperature, salinity and the current speed and direction data derived from the NEMO deployment also provided useful additional information for the hydrography of the area. 4.6. Submarine groundwater discharge studies Following successful operations of IAEA-MEL’s in situ fully automated gamma-ray spectrometers for continuous monitoring of anthropogenic radionuclides in seawater, it has been
Underwater gamma-ray spectrometry
467
possible to apply this technique also for monitoring of natural radionuclides such as 40 K, 228 Ra and radon daughter products, which are commonly used as tracers for studying marine and coastal processes (Povinec et al., 2001a; Levy-Palomo et al., 2004). One of the frequently studied coastal process is submarine groundwater discharge (SGD), because of its potential importance for management of fresh water resources in coastal areas (Taniguchi et al., 2002; Kontar et al., 2002; Burnett et al., 2006). Several isotope techniques for SGD quantification have been developed using stable (2 H, 18 O, 87/86 Sr, etc.) and radioactive (3 H, 14 C, Ra isotopes, 222 Rn, etc.) isotopes (e.g., Moore, 2000; Burnett et al., 2001, 2002, 2006; Moore, 2006; Schiavo et al., 2006; Weinstein et al., 2005; Povinec et al., 2006a). New technologies developed in recent years have been used to carry out temporal and spatial monitoring of SGD via analysis of radon or its daughter products emitting alpharays (Burnett et al., 2001; Burnett and Dulaiova, 2003, 2006; De Oliveira et al., 2003), or by analyzing radon decay products emitting gamma-rays (Povinec et al., 2001a; Levy-Palomo et al., 2004; Povinec, 2004, 2005; Povinec et al., 2006a, 2006b, 2006c, 2007). A Coordinated Research Project (CRP) on “Nuclear and Isotopic Techniques for the Characterization of SGD in Coastal Zones” has been jointly organized by the IAEA’s Marine Environment Laboratory (Monaco) and the Isotope Hydrology Section (Vienna), with the aim of developing new isotope techniques for studying SGD. The CRP has been carried out in cooperation with UNESCO’s Intergovernmental Oceanographic Commission (IOC) and the International Hydrological Program (IHP), and collaboration with several laboratories in Brazil, India, Italy, Japan, Russia, Slovenia, Turkey and USA was established (Povinec et al., 2006a). In the framework of the CRP two expeditions were carried out (in June 2001 and March 2002) to the Ionian Sea (offshore of Sicily), and one in November 2003 to Ubatuba (Sao Paulo region, Brazil). The choice of such geologically different regions was based on the strategy, developed in the framework of the IAEA–UNESCO SGD cooperation, to visit and study SGD sites with different geological and hydrological conditions, which could primarily affect SGD in a region. It has been a great challenge to investigate SGD using underwater gamma-ray spectrometry in such different geological and hydrological environments, as presented by the Sicilian and Brazilian coasts. 4.6.1. Monitoring of SGD using an underwater gamma-ray spectrometer is one of the radionuclides which has been frequently used as a tracer for studying marine and coastal processes (Burnett and Dulaiova, 2003, 2006; De Oliveira et al., 2003; Levy-Palomo et al., 2004; Povinec et al., 2006a, 2006b, 2006c). Radon is a conservative tracer and because its concentration in groundwater is much higher than in seawater, it is an ideal tracer for studying groundwater–seawater interactions. 222 Rn is a decay product of 226 Ra (half-life 1600 y) in the 238 U natural decay chain, and thanks to its short half-life (3.82 d), it is a suitable tracer for studying dynamic systems that are usually found in coastal regions. 222 Rn daughters are short-lived radionuclides such as 214 Pb, 214 Bi, etc., which further decay by alpha and beta decays to medium-lived 210 Pb (22.2 y) and 210 Po (138 d), and finally to the stable 206 Pb. In the 232 Th natural decay chain there is another radon isotope, 220 Rn (also called thoron), with a shorter half-life (55.6 s). While 228 Ra (228 Ac) has been (as a part of the radium quartet, together with 226 Ra, 223 Ra and 224 Ra) very often used as a tracer of coastal processes (Moore, 222 Rn
468
P.P. Povinec et al.
2000, 2006), 220 Rn is still waiting for its applications in oceanography. Especially in coastal areas rich in thorium rocks observed, e.g., along the south-eastern Brazilian coast, 220 Rn may be a useful tracer of fast coastal processes. 220 Rn is a decay product of 224 Ra (half-life 3.66 d) which further decays to several short-lived daughter products including 208 Tl (half-life 183 s), and finally to the stable 208 Pb. Analysis of 222 Rn in the marine environment has been a time-consuming process. As shortlived daughter products of both radon isotopes, 222 Rn and 220 Rn, emit gamma-rays, in situ gamma-ray spectrometry would be a suitable non-destructive technique for their analysis in seawater. In situ gamma-ray spectrometry has the advantage that it can also simultaneously analyze other radionuclides, e.g. 40 K, which is the dominant radionuclide in seawater (half-life 1.96 × 109 y, gamma-line 1461 keV), but also 228 Ra (half-life 5.75 y, via its daughter 228 Ac (half-life 6.13 h, gamma-line 911 keV), 234 Th–234 Pa (half-lives 24.1 d and 71 s, gammalines 63, 93 and 1001 keV), and others, especially when highly sensitive Ge detectors with electrical cooling operating in seawater would be used (Kobayashi et al., 1998). The best visible gamma-lines in the spectrum of seawater are due to 214 Bi (609, 1120 and 1765 keV) for 222 Rn daughters, and 212 Pb (239 keV) and 208 Tl (583 and 2615 keV) for 220 Rn decay products. The small NaI(Tl) underwater gamma-ray spectrometer was used in SGD studies. Additional sensors for monitoring of temperature, water depth and wave impacts were located in front of the NaI(Tl) detector. The detector unit was connected via a 70 m long, doublearmored steel coaxial cable to a PC with processing electronics and multichannel analyzer. The PC and an auxiliary low voltage power supply were located on a ship or in a car when operating close to the coast. The efficiency calibration was done using 137 Cs, 40 K and 226 Ra sources dispersed in a polyethylene tank. Background measurements were carried out with the detector immersed in the tank filled with fresh water. The detection limit for 222 Rn measurements in seawater is ∼0.05 kBq m−3 and the reported uncertainties are ∼20% (Povinec et al., 2006b). The corresponding 214 Bi peaks used in spectral evaluations were either at 609, 1120 or 1765 keV, depending on background conditions during real measurements. The data acquisition system evaluates gamma-ray spectra every minute. Later the obtained spectra are integrated to 1 h intervals (depending on the type of measurement, e.g., continuous long-term monitoring at one site, or spatial mapping), and the activity concentrations of selected radionuclides in fresh water or seawater are calculated. The system is fully automated and can operate without any surveillance. 4.6.2. Study area offshore of SE Sicily The studied area belongs to the Hyblean Plateau that forms the south-eastern (SE) part of Sicily (Aureli, 1992). It is primarily of carbonate origin (Triassic–Jurassic and partly Cretaceous). The eastern part was influenced by volcanic activity, while the western part was formed from carbonate sediments. The groundwater circulation appears mostly along cracks, fractures and karst hollows, producing numerous springs on beaches, as well as submarine springs in the sea (Aureli, 1992). In situ underwater gamma-ray spectrometry measurements were carried out from 16 to 25 March 2002 during the 2002 expedition. The continuous SGD monitoring was done mainly in the Donnalucata boat basin (Figure 15), where several sites, situated close to manual or automatic seepage meter posts, were visited. Continuous salinity measurements during the
Underwater gamma-ray spectrometry
469
Fig. 15. Sampling sites in the Donnalucata boat basin (Sicily).
Fig. 16. Radon concentration in seawater versus salinity in the Donnalucata boat basin (Sicily) (for sampling sites see Figure 15).
Sicily expedition were carried out using a small conductivity/temperature/depth DST-CTD sensor (Star-Oddi, Iceland). The precision of salinity measurements was ±0.01. Seepage rates up to ∼30 cm day−1 were observed (Taniguchi et al., 2006). Highest 222 Rn activity concentrations (up to 3.7 kBq m−3 ) were recorded at the sites close to the coast (Figure 16), where salinities were the lowest (35.8); on the contrary, in open sea conditions, where salinity was the highest (38.7), the 222 Rn activity concentrations were the lowest (∼0.1 kBq m−3 ). A recirculation of groundwater + seawater mixture, having higher salinities
470
P.P. Povinec et al.
Fig. 17. Time series of radon concentration in seawater in the Donnalucata boat basin (Sicily) versus salinity and tide.
but lower 222 Rn activity concentrations (due to its radioactive decay), may be responsible for the observed lower 222 Rn concentrations at sites where high seepage rates were observed. Time-series of 222 Rn in seawater, salinity and tide were recorded at a site close to the coast. Results presented in Figure 17 document that after the maximum tide the 222 Rn activity concentration of seawater was minimum (down to 2.3 kBq m−3 ), and after the minimum tide the 222 Rn activity concentration was at maximum (up to 4.8 kBq m−3 ) with a delay of about 1 h. A shift of approximately 2 h was observed between the maximum tide and the salinity maximum. It is interesting to see that even small changes in the water level, observed, e.g., during March 21 and 22, made corresponding changes in 222 Rn activity concentrations. 4.6.3. Study area offshore of SE Brazil The tropical coastal area in south-eastern Brazil, covering latitudes between 23◦ 26 S and 23◦ 46 S, and longitudes between 45◦ 02 W and 45◦ 11 W in the northernmost part of the São Paulo Bight (about 270 km north of São Paulo), and comprising a series of small embayments near Ubatuba town, was visited during the expeditions in August 2002 and November 2003 (Figure 18). The humid tropical and subtropical climate, as well as the absence of great river basins in the area, gives the rainfall regime great importance in the contribution of freshwater from the continental areas to the ocean. In the study area, the coastal aquifer system can be classified as a fractured rock aquifer, covered by Pleistocene and Holocene sediments. The discharge pattern of this kind of aquifer is spatially heterogeneous, with preferential flow paths along rock fractures. Wave action is the most effective hydrodynamic phenomenon responsible for the bottom sedimentary processes in the coastal area as well as in the adjacent inner continental shelf. Three water masses occur in the area: (i) Coastal Water (CW), characterized by high temperature (>25 ◦ C) and low salinity (32–33); (ii) Tropical Water (TW) with intermediate temperature (20–23 ◦ C) and high salinity (∼36); and (iii) South Atlantic Central Water (SACW) with low temperature (16–18 ◦ C) and high salinity (35–36). During the summer (November
Underwater gamma-ray spectrometry
471
Fig. 18. Ubatuba coast (SE Brazil).
through February), nutrient-rich SACW moves onshore and is often found in the central and outer portions of the continental shelf (20–100 m), while CW is found along a narrow band inshore. These water movements result in a vertical stratification over the inner shelf, with a strong thermocline at middle depths. In the winter (May through August), when SACW is restricted to the outer shelf, horizontal and vertical thermal gradients are reduced and almost no stratification is observed on the inner shelf. In summer, the advance of the SACW over the coast leads to the displacement of the CW, rich in continental suspended materials. The mean annual rainfall is around 1 800 mm, the maximum rainfall rates being observed in February. Sea level varies from 0.5 to 1.5 m, the highest values occurring in months August–September due to the greater volume of warm waters of the Brazil Current (Mesquita, 1997). Continuous underwater gamma-ray spectrometry measurements were carried out in several bays along the Ubatuba coast, especially in Flamengo and Picinguaba Bays, where spatial mapping was carried out, and at the Oceanographic Base of the University of Sao Paolo, about 20 m from the coast, where time-series of radon daughters in seawater were investigated. Salinity and temperature measurements were done using a CTD (MCTD-DBP, Falmouth Scientific Inc., precision ±0.001). Water level data measured by Burnett and Dulaiova (2006) during the Sicily mission and data recorded by the Oceanographic Institute of the University of Sao Paulo for the Brazilian mission have been used in the present study. In Flamengo Bay the 222 Rn activity concentration observed at five stations varied between 50 and 200 Bq m−3 , the highest values being found close to the Oceanographic Base, and at Perequê-Mirim Beach. Four stations visited in Picinguaba Bay showed 222 Rn activity concentrations between 50 and 140 Bq m−3 . An inverse relationship between the observed 222 Rn activity concentration of seawater and salinity is demonstrated in Figure 19, with correlation
472
P.P. Povinec et al.
Fig. 19. Radon concentration versus salinity in Picinguaba and Flamengo Bays (Brazil).
Fig. 20. Time series of radon concentration in seawater in Flamengo Bay (Brazil) versus salinity and tide.
coefficients of r 2 = 0.93 and 0.92 for Flamengo and Picinguaba Bay, respectively (Povinec et al., 2007). Time-series of 222 Rn activity concentration in seawater, salinity and tide recorded from November 22 to 26, 2003 in Flamengo Bay are shown in Figure 20. The 222 Rn activity concentration in seawater varied between 1.0 and 5.2 kBq m−3 , while the tide varied between 4.4 and 5.6 m. Generally, the salinity record follows the tide record; however, on November 26 a delay in the salinity record was observed. The usual inverse relationship between the 222 Rn activity concentration in seawater and tide/salinity was not observed during November 22, despite large variations in water level. The observed changes in salinity during this time were, however, also smaller than during 25th and 26th November. The inverse relationship between the 222 Rn activity concentration in seawater and tide/salinity was, however, again established from 23rd to 25th November, when a few hours shift between the tide minimum/maximum and the 222 Rn maximum/minimum activity concentration was observed (Povinec et al., 2007). Thanks to elevated concentrations of 232 Th in the region it was possible to measure for the first time non-destructively 228 Ra in seawater (via its daughter gamma-ray emitter 228 Ac).
Underwater gamma-ray spectrometry
473
228 Ra
activity concentrations measured in Flamengo and Picinguaba bays varied between 1.6 and 3.3 Bq m−3 . They followed the 222 Rn concentrations measured at the same stations. 4.6.4. Submarine groundwater discharge estimates offshore of Brazil We are developing a general model of groundwater–seawater interactions in coastal areas, which is based on oceanographic parameters (tide, waves, salinity, temperature, seepage rates) as well as on chemical tracers (stable isotopes, tritium, radium quartet, radon, trace elements and nutrients). As this is a very complex approach, in the present paper we shall focus on radon data only. We shall follow the Burnett model developed for estimation of SGD fluxes in coastal waters (Burnett and Dulaiova, 2003, 2006; De Oliveira et al., 2003). We assume that radon may be advected via groundwater discharge into coastal water. On the basis of measured radon concentrations in coastal seawater we calculate radon inventories averaged over several hours. The radon flux FRn may then be expressed as FRn =
IRn 1−
e−λt λ
,
where IRn is the 222 Rn inventory and λ is the 222 Rn decay constant (7.54 × 10−3 h−1 ). After several 222 Rn half-lives this equation reduces to λIRn , and for the SGD flux FSGD we can write λIRn − λIRa + E ± M , FSGD = CRn where λIRa is the production rate of 222 Rn from 226 Ra, E is the atmospheric evasion of radon, M represents the horizontal mixing of radon, and CRn is the 222 Rn concentration in the pore water. Several approximations have to be made for estimation of SGD fluxes from measured 222 Rn concentrations: (i) The correction for supported radon was done by subtracting an equivalent activity of 226 Ra measured in seawater close to the Base (8 Bq m−3 ). (ii) The average atmospheric evasion in Flamengo Bay was estimated to be 0.8 ± 0.5 Bq m−2 h−1 following the previous work of De Oliveira et al. (2003), and the more recent study of Burnett et al. (2007). (iii) The losses due to mixing of SGD with offshore seawater were estimated to be 12 ± 9 Bq m−2 h−1 using data published by De Oliveira et al. (2003) and Burnett et al. (2007). (iv) We expect that benthic fluxes of radon are driven mainly by groundwater (pore water) advection. The mean 222 Rn concentration in pore water was estimated to be 5.5 kBq m−3 on the basis of measurements of Cable and Martin (2007) carried out at two sites (MS-3 and MS-5), close to our monitoring site (Figure 8). The estimated SGD fluxes from underwater gamma-spectrometry monitoring show similar patterns with time as the 222 Rn concentrations presented in Figure 20. The SGD fluxes varied during November 22–26 between 8 and 40 cm d−1 (the unit is cm3 /cm2 day), with an average value of 21 cm d−1 . The obtained SGD flux is similar to that obtained between November 16 and 20, 2003 at Flamengo Bay by Burnett et al. (2007), who found values between 1 and 29 cm d−1 (average value 13 cm d−1 ). The SGD fluxes estimated at a nearby site MS-3 by Cable and Martin (2007) from pore water inventories (9–24 cm d−1 ) and from artificial
474
P.P. Povinec et al.
tracers (53 cm d−1 ) are also in agreement with our values. Much larger SGD fluxes (up to 300 cm d−1 ) were measured at the low tide shoreline by manual and continuous seepage meters (Bokuniewicz et al., 2007; Taniguchi et al., 2007); however, comparable results to our data were obtained for a site situated about 10 m from the shoreline. Generally, these results showed large temporal and spatial variations of SGD fluxes observed in the small bay area. Using a two end-member mixing model we estimated that at the monitoring site the contribution of fresh water in seawater varied between 4 and 20%, with an average value of 10%. However, with offshore distance, as documented by the isotopic composition of collected samples, there was only a small contribution of fresh water in the submarine groundwater discharge. 4.6.5. Comparison of Sicilian and Brazilian results In contrast to the Sicily sites, which represent a typical karstic region with low massic activities of rocks (238 U ∼10 Bq kg−1 and 232 Th ∼5 Bq kg−1 ), the Brazilian site is a tropical coastal area characterized by granite rocks where the concentrations of 238 U and 232 Th in collected rock samples were higher by a factor of 5 and 10, respectively. The results obtained for Sicilian and Brazilian waters confirm the inverse relationship between tide and 222 Rn activity concentration. During decreasing sea level 222 Rn concentration increases and conversely during high tides 222 Rn concentration decreases. While 222 Rn activity concentration in Sicilian waters followed the tide with a delay of only 1 h, in Brazilian waters a delay of several hours was observed. This may be caused by the different geological/hydrological conditions at both sites. The Sicilian coast is characterized by carbonate rocks with cracks which facilitate groundwater transport to the sea, while for the Brazilian coast, granite rocks have lower transport capabilities (IAEA, 1998). In spite of different geological/hydrological settings, the 222 Rn activity concentrations in seawater at the Sicilian and Brazilian sites were very similar, between 2.3 and 4.8 kBq m−3 , and 1.0 and 5.2 kBq m−3 , respectively. Due to the factor of 5 higher 238 U concentration in granite than in carbonate rocks, higher 222 Rn activity concentrations along the Brazilian coast would be expected. However, in Flamengo Bay, similarly as observed in the Donnalucata boat basin, the SGD may be represented by a mixture of recirculated groundwater and seawater, having a lower 222 Rn concentration. In the Donnalucata region we have an end-member represented by the groundwater beach spring, having an average 222 Rn activity concentration of ∼16 kBq m−3 (varying between 12 and 18 kBq m−3 , depending on tide; Povinec et al., 2006b). In the case of Flamengo Bay, the groundwater end-member may be represented by a spring found at the Oceanographic Base which showed a similar isotopic composition (Povinec et al., 2007). The variations in 222 Rn activity concentrations in the Donnalucata boat basin, measured at several sites using alpha-ray spectrometry, were also reported by Burnett and Dulaiova (2006). They found spatial changes in 222 Rn activity concentrations from 0.05 to 2.5 kBq m−3 (the maximum concentration was measured on March 21st at a site about 10 m from our site). They showed that the observed variations in 222 Rn activity concentrations can be related to SGD fluxes, and thus can be used for characterization of SGD. Such large changes in SGD, observed in a relatively small area, document again why the isotopic characterization of SGD is important for estimation of real groundwater fluxes to the sea.
Underwater gamma-ray spectrometry
475
5. Conclusions Underwater gamma-ray instruments are most suitable for environmental monitoring in the areas where discharges from nuclear facilities occur. It would be of great public interest if such monitoring devices would accompany in real-time all releases of radionuclides from nuclear installations to the aquatic environment, so information on radionuclide levels would be accessible, e.g. by Internet, to the public. Such monitoring systems can also be adapted for environmental monitoring for safeguards to check for undeclared nuclear activities in regions of interests. Potential releases from radioactive waste dumpsites or accidents at sea can also be under surveillance and control using underwater monitoring systems. Underwater gamma-ray spectrometry is a cost-effective technique for applications requiring surveys of large areas, an immediate assessment or long-term observations of marine radioactivity. IAEA-MEL has applied underwater gamma-ray spectrometry in surveys of seabed sediment contamination in a variety of marine environments: nuclear waste dumpsites in the Arctic; areas influenced by authorized liquid waste discharges in the Irish Sea offshore of Sellafield nuclear reprocessing plant; nuclear test sites in the South Pacific atolls of Mururoa and Fangataufa; and Mediterranean background levels monitoring offshore of Monaco. The survey of seabed contamination in Mururoa and Fangataufa lagoons identified areas of high surface sediment activities, which were subsequently sampled for detailed laboratory-based analyses. In the eastern Irish Sea patterns of 137 Cs in seabed sediment obtained in 1995 yielded a similar distribution to that previously reported, but the levels of 137 Cs concentration in surface sediment showed significant decreases in the nearshore area as compared to those reported for 1988 and earlier surveys. An autonomous NaI(Tl)-based system with a relatively low power consumption (1 W), operational on an oceanographic buoy with satellite data transmission (the NEMO observatory), was deployed in the NW Irish Sea from August to December 2000, searching for 137 Cs variations in the Irish Sea gyre. In situ underwater gamma-ray spectrometry measurements carried out during the expeditions in coastal waters offshore of SE Sicily and SE Brazil showed the ability of the method to monitor SGD and to study its temporal and spatial variations. This new method of SGD investigations represents a robust technique that can be applied for long-term, continuous monitoring of radon in seawater and/or groundwater. The time-series measurements of 222 Rn activity concentrations generally confirmed an anticorrelation between the 222 Rn activity concentration and tide/salinity. It has been found that variations in 222 Rn activity concentrations are caused by sea-level variations as tide effects induce variations of hydraulic gradients, which increase 222 Rn concentrations during decreasing sea level and vice versa.
Acknowledgements The authors are indebted to Prof. M.S. Baxter who, as director of IAEA-MEL, initiated the work on underwater gamma-ray spectrometry. Further, the authors acknowledge collaboration with Dr. C. Musa of SMSRB-CEA, Drs. D. Woodhead and P. Kershaw of CEFAS, Dr. T.P. Ryan of RPII during various stages of expeditions to Mururoa and to the Irish Sea. The authors
476
P.P. Povinec et al.
would like to thank members of the IAEA–UNESCO team who participated in the expeditions to Sicily and to Brazil for fruitful collaboration. The authors also thank Prof. W.C. Burnett and the Oceanographic Institute of the University of Sao Paulo for providing water level data. Prof. A. Aureli and Dr. A. Privitera and the University of Palermo (Italy), as well as Dr. J. de Oliveira, the Instituto de Pesquisas Energeticas e Nucleares at Sao Paulo (Brazil), and the University of Sao Paulo are acknowledged for logistical support during the field-work on SGD. The assistance of Captains and crews of the R/Vs Rari, Cirolana, Celtic Voyager, Veliger II and Albacora during expeditions to Mururoa, the Irish Sea and offshore of Brazil are acknowledged as well. The authors are also grateful to Dr. I. Levy-Palomo for assistance during data evaluation for the Sicilian and Brazilian expeditions. The IAEA is grateful for the support provided to its Marine Environment Laboratory by the Government of the Principality of Monaco.
References Aakenes, U.R. (1995). Radioactivity monitored from moored oceanographic buoys. Chem. Ecol. 10, 61–69. Aarkrog, A. (2003). Input of anthropogenic radionuclides into the World Ocean. Deep-Sea Res. II 50, 2597–2606. Aureli, A. (1992). The submarine springs in Sicily. Reporto di Instituto di Geologia e Geofisica, Universita di Catania, 21 pp. Baxter, M.S., Harms, I., Osvath, I., Povinec, P.P., Scott, E.M. (1997). Modelling the potential radiological consequences of radioactive waste dumping in the Kara Sea. J. Environ. Radioact. 39, 161–181. Bokuniewicz, H., Taniguchi, M., Ishitoibi, T., Charette, M., Allen, M., Kontar, E.A. (2007). Direct measures of submarine groundwater discharge over a fractured rock aquifer in Ubatuba, Brazil. Estuar. Coast. Shelf Sci., in press. Burnett, W.C., Dulaiova, H. (2003). Estimating the dynamics of groundwater input into the coastal zone via continuous radon-222 measurements. J. Environ. Radioact. 69, 21–35. Burnett, W.C., Dulaiova, H. (2006). Radon as a tracer of submarine groundwater discharge into a boat basin in Donnalucata, Sicily. Cont. Shelf Res. 26, 862–873. Burnett, W.C., Kim, G., Lane-Smith, D. (2001). A continuous radon monitor for assessment of radon in coastal ocean waters. J. Radioanal. Nucl. Chem. 249, 167–172. Burnett, W.C., Chanton, J., Christoff, J., Kontar, E., Krupa, S., Lambert, M., Moore, W., O’Rourke, D., Paulsen, R., Smith, C., Smith, L., Taniguchi, M. (2002). Assessing methodologies for measuring groundwater discharge to the ocean. EOS 83, 117–123. Burnett, W.C., Aggarwal, P.K., Aureli, A., Bokuniewicz, H., Cable, J.E., Charette, M.A., Kontar, E., Krupa, S., Kulkarni, K.M., Loveless, A., Moore, W.S., Oberdorfer, J.A., Oliveira, J., Ozyurt, N., Povinec, P.P., Privitera, A.M.G., Rajar, R., Ramessur, R.T., Scholten, J., Stieglitz, T., Taniguchi, M., Turner, J.V. (2006). Quantifying submarine groundwater discharge in the coastal zone via multiple methods. Sci. Total Environ. 367, 498–543. Burnett, W.C., Peterson, R., Moore, W.S., de Oliveira, J. (2007). Radon and radium isotopes as tracers of submarine groundwater discharge—results from the Ubatuba, Brazil SGD assessment intercomparison. Estuar. Coast. Shelf Sci., in press. Cable, J.E., Martin, J.B. (2007). In situ evaluation of nearshore marine and fresh porewater transport into Flamengo Bay, Brazil. Estuar. Coast. Shelf Sci., in press. De Oliveira, J., Burnett, W.C., Mazzilli, B.P., Braga, E.S., Farias, L.A., Christoff, J., Furtado, V.V. (2003). Reconnaissance of submarine groundwater discharge at Ubatuba coast, Brazil, using 222 Rn as a natural tracer. J. Environ. Radioact. 69, 37–52. Hamilton, T.H., Ballestra, S., Baxter, M.S., Gastaud, J., Osvath, I., Parsi, P., Povinec, P.P., Scott, E.M. (1994). Radiometric investigations of Kara Sea sediments and preliminary radiological assessment related to dumping of radioactive wastes in the Arctic Seas. J. Environ. Radioact. 25, 113–134. Harms, I., Povinec, P. (1998). The outflow of radionuclides from Novaya Zemlya bays—modelling and monitoring strategies. In: Int. Symp. on Marine Pollution. IAEA, Vienna, pp. 730–731.
Underwater gamma-ray spectrometry
477
Harms, I.H., Povinec, P.P. (1999). The outflow of radionuclides from Novaya Zemlya Bays—modelling and monitoring strategies. Sci. Total Environ. 237/238, 193–201. Hill, A.E., Durazo, R., Smeed, D.A. (1994). Observations of a cyclonic gyre in the western Irish Sea. Cont. Shelf Res. 14, 479–490. Hill, A.E., Brown, J., Fernand, L. (1996). The western Irish Sea gyre: a retention system for Norway lobster (Nephrops norvegicus). Oceanol. Acta 19, 357–368. Hunt, G.J., Kershaw, P.J. (1990). Remobilisation of artificial radionuclides from the sediment of the Irish Sea. J. Radiol. Protect. 10, 147–151. IAEA (1998). The Radiological Situation at the Atolls of Mururoa and Fangataaufa, The Main Report. IAEA, Vienna, 282 pp. Jones, D.G. (2001). Development and application of offshore gamma-ray spectrometry measurements: a review. J. Environ. Radioact. 53, 313–333. Jones, D.G., Miller, J.M., Roberts, P.D. (1984). The distribution of 137 Cs in surface intertidal sediments from the Solvay Firth. Mar. Pollut. Bull. 15, 187–194. Jones, D.G., Roberts, P.D., Miller, J.M. (1988). The distribution of gamma-emitting radionuclides in surface subtidal sediments near the Sellafield Plant. Estuar. Coast. Shelf Sci. 27, 143–161. Jones, D.G., Roberts, P.D., Strutt, M.H., Higgo, J.J., Davis, J.R. (1999). Distribution of 137 Cs and inventories of 238 Pu, 239/240 Pu, 241 Am and 137 Cs in Irish Sea intertidal sediments. J. Environ. Radioact. 44, 159–189. Katagiri, M., Itoh, H. (1994). Small electric-cooled germanium gamma-ray detector. In: Proc. 8th Workshop on Radiation Detectors and Their Uses. KEK, Tsukuba, KEK-PROC-94-7. Katagiri, M., Kobayashi, Y., Taguchi, Y., Uchida, T. (1995). General-purpose germanium gamma-ray detector cooled with two Stirling refrigerators. In: Proc. 9th Workshop on Radiation Detectors and Their Uses. KEK, Tsukuba, KEK-PROC-95-1. Kershaw, P., Pentreath, R., Woodhead, D., Hunt, G. (1992). A review of radioactivity in the Irish Sea. MAFF Aquatic Environment Monitoring Report No. 32. MAFF, Lowestoft. Kershaw, P., Denoon, D., Woodhead, D. (1999). Observations on the redistribution of plutonium and americium in the Irish Sea sediments, 1978 to 1996: concentrations and inventories. J. Environ. Radioact. 44, 191–221. Kobayashi, Y., Takahashi, R., Shima, S., Katagiri, M., Takahashi, K. (1998). Development of a submersible Ge gamma-ray detector system. In: Int. Symp. on Marine Pollution. IAEA, Vienna, pp. 530–531. Kontar, E.A., Burnett, W.C., Povinec, P.P. (2002). Submarine groundwater discharge and its influence on hydrological trends in the Mediterranean Sea. In: Tracing Long-Term Hydrological Change in the Mediterranean Sea. In: CIESM Series, vol. 16. CIESM, Monaco, pp. 109–113. Levy-Palomo, I., Comanducci, J.-F., Povinec, P.P. (2004). Investigation of submarine groundwater discharge in Sicilian and Brazilian coastal waters using underwater gamma-spectrometer. In: Aquatic Forum ’2004. IAEA, Vienna, IAEA-Cn-118, pp. 228–229. Masse, L., Musa, C. (1996). Etude Granulometrique des Sediments de Surface du Lagon de Mururoa. SMSRB Report, Paris. Mesquita, A.R. (1997). Marés, circulação e nível do mar na Costa Sudeste do Brasil. Relatório Fundespa, São Paulo. Miller, J.M., Thomas, B.W., Roberts, P.D., Creamer, S.C. (1982). Measurement of marine radionuclide distributions using a towed seabed spectrometer. Mar. Pollut. Bull. 13, 315–319. Moore, W.S. (2000). Ages of continental shelf waters determined from 223 Ra and 224 Ra. J. Geophys. Res. 105, 22117–22122. Moore, W.S. (2006). Radium isotopes as tracers of submarine groundwater discharge in Sicily. Cont. Shelf Res. 26, 852–861. Musa, C., Bourlat, Y., Millies-Lacroix, J.-C. (1996). Lagoon sediment radioactivity in Mururoa and Fangataufa. SMSRB, CEA, Monthlery. Nies, H., Schilling, J.G. (2003). Aktuelle Belastung von Nord- und Ostsee durch Radopnuklide und Informationen aus dem automatischen Radioaktivitätsmessnetz des BSH. In: Proc. 12 Fachgespräch Überwachung der Umweltradioaktivität. BMU, Bonn, pp. 445–454. Noakes, J.E., Harding, J.L. (1982). Nuclear techniques for seafloor mineral exploration. In: Proc. Oceanology International Exhibition and Conference, Brighton. Noakes, J.E., Noakes, S.E., Dvoracek, D.K., Culp, R.A. (1999). Rapid coastal survey of anthropogenic radionuclides, metals, and organic compounds in surface marine sediments. Sci. Total Environ. 237/238, 449–454.
478
P.P. Povinec et al.
McCartney, M., Kershaw, P., Woodhead, D., Denoon, D. (1994). Artificial radionuclides in the surface sediments of the Irish Sea, 1968–1988. Sci. Total Environ. 141, 103–138. Osvath, I., Povinec, P.P. (2001). Seabed γ -ray spectrometry: applications at IAEA-MEL. J. Environ. Radioact. 53, 335–349. Osvath, I., Povinec, P.P., Baxter, M.S. (1999a). Kara Sea radioactivity assessment. Sci. Total Environ. 237/238, 167– 179. Osvath, I., Povinec, P.P., Huynh-Ngoc, L., Comanducci, J.-F. (1999b). Underwater gamma surveys of Mururoa and Fangataufa lagoons. Sci. Total Environ. 237/238, 277–286. Osvath, I., Povinec, P.P., Baxter, M.S., Huynh-Ngoc, L. (2001). Mapping of the distribution of 137 Cs in Irish Sea sediments. J. Radioanal. Nucl. Chem. 248, 735–739. Osvath, I., Povinec, P.P., Livingston, H.D., Ryan, T.P., Mulsow, S., Comanducci, J.F. (2005). Monitoring of radioactivity in NW Irish Sea water using a stationary underwater gamma-ray spectrometer with satellite data transmission. J. Radioanal. Nucl. Chem. 263, 437–440. Poole, A.J., Denoon, D.C., Woodhead, D.S. (1997). The distribution and inventory of 137 Cs in sub-tidal sediments of the Irish Sea. Radioprotection—Colloquies 32, C2-263–C2-270. Povinec, P.P. (1997). Radioactivity of the Barents and Kara Seas. In: World Directory of Environmental Testing, Monitoring and Treatment 1997/8. James & James (Science Publishers), London, pp. 142–146. Povinec, P.P. (2004). Developments in analytical technologies for marine radionuclide studies. In: Livingston, H.D. (Ed.), Marine Radioactivity. Elsevier, Amsterdam, pp. 237–294. Povinec, P.P. (2005). Ultra-sensitive radionuclide spectrometry: Radiometrics and mass spectrometry synergy. J. Radioanal. Nucl. Chem. 263, 413–417. Povinec, P.P., Osvath, I., Baxter, M.S., Harms, I. (1995). In situ monitoring of gamma-radioactivity in the Kara Sea. In: Proc. of the Workshop on Monitoring of Nuclear Contamination in Arctic Seas. Naval Research Laboratory, Washington, DC, NRL/MR/6610-95-7674, pp. IV2–IV17. Povinec, P.P., Osvath, I., Baxter, M.S. (1996). Underwater gamma-spectrometry with HPGe and NaI (Tl) detectors. Appl. Radiat. Isotopes 47, 1127–1133. Povinec, P.P., Osvath, I., Baxter, M.S., Harms, I., Huynh-Ngoc, L., Liong Wee Kwong, L., Pettersson, H.B.L. (1997a). IAEA-MEL’s contribution to the investigation of Kara Sea radioactivity and radiological assessment. Mar. Pollut. Bull. 35, 235–241. Povinec, P.P., Baxter, M.S., Gayol, J., Huynh-Ngoc, L., Liong, W.K.L., Osvath, I. (1997b). Inventories and distributions of radionuclides in the Kara Sea. In: Germain, P., Guarry, J.C., Guegueniat, P., Metivier, H. (Eds.), Radionuclides in the Ocean. IPSN, Les Ulis, pp. C2.29–C2.34. Povinec, P.P., Woodhead, D., Blowers, P., Bonfield, R., Cooper, M., Chen, Q.J., Dahlgaard, H., Dovlete, C., Fox, V., Froehlich, K., Gastaud, J., Gröning, M., Hamilton, T., Ikeuchi, Y., Kanisch, G., Krüger, A., Liong Wee Kwong, L., Matthews, M., Morgenstern, U., Mulsow, S., Pettersson, H.B.L., Smedley, P., Taylor, B., Taylor, C., Tinker, R. (1999). Marine radioactivity assessment of Mururoa and Fangataufa atolls. Sci. Total Environ. 237/238, 249–267. Povinec, P.P., La Rosa, J.J., Lee, S.H., Mulsow, S., Osvath, I., Wyse, E. (2001a). Recent developments in radiometric and mass spectrometry methods for marine radioactivity measurements. J. Radioanal. Nucl. Chem. 248, 713–718. Povinec, P.P., Osvath, I., Mulsow, S., Comanducci, J.-F. (2001b). La surveillance in situ de la radioactivite marine au large de Monaco. In: Rapport du 36e Congres de la CIESM. In: CIESM Series, vol. 36. CIESM, Monaco, 155 pp. Povinec, P.P., Aggarwal, P.K., Aureli, A., Burnett, W.C., Kontar, E.A., Kulkarni, K.M., Moore, W.S., Rajar, R., Taniguchi, M., Comanducci, J.-F., Cusimano, G., Dulaiova, H., Gatto, L., Groening, M., Hauser, S., LevyPalomo, I., Oregioni, B., Ozorovich, Y.R., Privitera, A.M.G., Schiavo, M.A. (2006a). Characterisation of submarine groundwater discharge offshore south-eastern Sicily. J. Environ. Radioact. 89, 81–101. Povinec, P.P., Comanducci, J.F., Levy-Palomo, I., Oregioni, B. (2006b). Monitoring of submarine groundwater discharge along the Donnalucata coast in the south-eastern Sicily using underwater gamma-ray spectrometry. Cont. Shelf Res. 26, 874–884. Povinec, P.P., Levy-Palomo, I., Comanducci, J.-F., de Oliveira, J., Oregioni, B., Privitera, A.M.G. (2006c). Submarine groundwater discharge investigations in Sicilian and Brazilian coastal waters using an underwater gamma-ray spectrometer. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 373–381. Povinec, P.P., de Oliveira, J., Braga, E.S., Comanducci, J.F., Gastaud, J., Groening, M., Levy-Palomo, I., Morgenstern, U., Top, Z. (2007). Isotopic characterization of groundwater and seawater in the Ubatuba coastal area (Brazil). Estuar. Coast. Shelf Sci., in press.
Underwater gamma-ray spectrometry
479
Radiological Protection Institute of Ireland, RPII (2000). Radioactivity Monitoring of the Irish Marine Environment, 1998 and 1999. RPII, Dublin, Ireland. Schiavo, M.A., Hauser, S., Cusimano, G., Gatto, L. (2006). Geochemical characterization of groundwater and submarine discharge in the south-eastern Sicily (Italy). Cont. Shelf Res. 26, 826–834. Senftle, F.E., Tanner, A.B., Philbin, P.W., Noakes, J.E., Spaulding, J.D., Harding, J.L. (1976). In-situ capture gammaray analyses for seabed exploration. In: Proc. Symp. on Nuclear Techniques in Geochemistry and Geophysics. IAEA, Vienna, pp. 75–89. Sjöblom, K.-L., Salo, A., Bewers, J.M., Cooper, J., Dyer, R.S., Lynn, N.M., Mount, M.E., Povinec, P.P., Sazykina, T.G., Schwarz, J., Scott, E.M., Sivintsev, Y.V., Tanner, J.E., Warden, J.M., Woodhead, D. (1999). International Arctic Seas Assessment Project. Sci. Total Environ. 237/238, 153–166. Taniguchi, M., Burnett, W.C., Cable, J.E., Turner, J.V. (2002). Investigation of submarine groundwater discharge. Hydrol. Processes 16, 2115–2129. Taniguchi, M., Burnett, W.C., Dulaiova, H., Kontar, E.A., Povinec, P.P., Moore, W.S. (2006). Submarine groundwater discharge measured by seepage meters in Sicilian coastal waters. Cont. Shelf Res. 26, 835–842. Taniguchi, M., Stieglitz, T., Ishitobi, T. (2007). Temporal variability of water quality of submarine groundwater discharge in Ubatuba, Brazil. Estuar. Coast. Shelf Sci., in press. Thomas, B.W., Clayton, C.G., Ranasinghe, V., Blair, I.M. (1983). Mineral exploration of the sea bed by towed sea bed spectrometers. Int. J. Appl. Radiat. Isotopes 34, 437–449. Thomas, B.W., Miller, J.M., Malcolm, A. (1984). Radiometric surveys of the seabed. In: Oceanology International Exhibition and Conference, Brighton. van Put, P., Debauche, A., De Lellis, C., Adam, V. (2004). Performance level of an autonomous system of continuous monitoring of radioactivity in seawater. J. Environ. Radioact. 72, 177–186. Wedekind, Ch. (1973). Gamma-ray spectrometer probe for the measurements of radioactive pollution in the sea. Health Phys. 25, 51–55. Wedekind, Ch., Schilling, G., Grüttmüller, M., Becker, K. (1999). Gamma-radiation monitoring network at sea. Appl. Radiat. Isotopes 50, 733–738. Wedekind, Ch., Schilling, G., Grüttmüller, M., Becker, K. (2000). Marine environmental radioactivity monitoring by “in-situ” radiation detection. Kerntechnik 65, 190–194. Weinstein, Y., Less, G., Kafri, U., Herut, B. (2005). Submarine groundwater discharge in the Southeastern Mediterranean. In: Povinec, P.P., Sanchez-Cabeza, J.A. (Eds.), Radionuclides in the Environment. Elsevier, Amsterdam, pp. 360–372. Woodhead, D.S. (1988). Mixing processes in nearshore marine sediments as inferred from the distributions of radionuclides discharged into the north-east Irish Sea from BNFL, Sellafield. In: Guary, J.C., Guegueniat, P., Pentreath, R.J. (Eds.), Radionuclides: A Tool for Oceanography. Elsevier, London, pp. 331–340.
This page intentionally left blank
481
Author Index
Aage, H.K. 420 Aakenes, U.R. 453 Aarkrog, A. 362, 459 Aarkrog, A., see Chen, Q.J. 152, 153, 312, 390–392, 394, 396, 399, 401 Aarkrog, A., see Holm, E. 311 Aarkrog, A., see Hou, X.L. 318, 385– 390 Aarkrog, A., see Povinec, P.P. 142, 148, 154 Aarnio, P.A. 182 Abdallah, A.M.A., see Sturchio, N.C. 91 Abdullah, S.A., see Dickson, H.W. 411, 413 Abe, T., see Komura, K. 234 Abelson, P.H., see McMillan, E.M. 310 Abney, K.D., see Conradson, S.D. 366 Abrahamsen, L., see Marsden, O.J. 287 Absy, M.L., see Turcq, B. 252 Adam, V., see van Put, P. 450 Adams, F., see Salbu, B. 357, 367 Adams, J.A.S., see Duval, J.S. 411, 413 Adrianov, V., see Kraft, S. 272 Afinogenov, A.M., see Kalmykov, St.N. 392, 393 Agarande, M. 322 Aggarwal, P.K., see Burnett, W.C. 467 Aggarwal, P.K., see Povinec, P.P. 451, 467 Ahmad, I., see Paul, M. 280 Ahmad, I., see Vockenhuber, C. 279 Ahmad, I., see Winkler, S. 279
Airo, M.L. 439 Aitken, C.G.G. 25 Akamian, V., see Foti, S. 394–396 Al Mahamid, I., see Conradson, S.D. 366 Alberts, R. 19 Aldave de las Heras, L., see Betti, M. 359 Aldave de las Heras, L., see Perna, L. 361 Alexander, P. 410 Aliev, R.A., see Kalmykov, St.N. 392, 393 Alkhazov, G.D. 342 Allen, M., see Bokuniewicz, H. 474 Alley, R., see Hughen, K.A. 249 Alley, R.A. 251 Alley, R.B. 251, 253 Allkofer, O.C. 172–174, 184 Allyson, J.D. 411–413, 415, 416, 420, 440 Allyson, J.D., see Creswell, A.J. 437 Allyson, J.D., see Sanderson, D.C.W. 409, 418, 437 Allyson, J.D., see Tyler, A.N. 408, 410, 415–417, 424–426, 431, 440, 441 Almasi, P., see Bond, G. 251 Alonso, A., see Richter, S. 268 Alsecz, A., see Eriksson, M. 357, 362 Alsecz, A., see Jernström, J. 357, 361, 367 Amano, H., see Hirose, K. 138, 154, 155 Amano, H., see Ikeuchi, Y. 155, 157 Ambartzumian, R.V. 333
482
Author Index
American Chemical Society, ACS, 21 American National Standards Institute (ANSI) and American Society for Quality Control, ANSI/ASQC, 18 Amiel, A.J., see Margaritz, M. 40 Anandakrishnan, S., see Alley, R.B. 251, 253 Anastasis, B.P. 430 Ancich, V.G. 310 Anderson, R.F., see Livingston, H.D. 138, 155 Anderson, T.J. 313 Andersson, K.G., see Hou, X.L. 385, 386, 388, 390, 397 Andrews, J.E., see Dai, M.H. 332 Andrews, J.T., see Dyke, A.S. 254 Andrrasi, A., see Zombori, P. 417, 427, 428, 430 Anspaugh, L.R. 413 Aono, T., see Yamada, M. 155 Aoyama, M. 138, 140, 142, 143, 145, 146 Aoyama, M., see Hirose, K. 138, 139, 141, 143, 144, 146, 150, 151, 154, 155, 235 Aoyama, M., see Ikeuchi, Y. 155, 157 Aoyama, M., see Miyao, T. 138 Aoyama, M., see Povinec, P.P. 138, 148, 149, 155 Aoyama, M., see Tsumune, D. 138, 139, 158 Appleby, L.J. 355 Ardila, G., see Meltzer, D.J. 253 Arianer, J., see Sauvage, J. 337 Armelin, M.J.A. 401 Armelin, M.J.A., see Vasconcellos, M.B.A. 401 Armstrong, M. 29 Arnold, D. 234 Arnold, D., see Hult, M. 235 Arnold, D., see Laubenstein, M. 108, 163, 202, 203 Arnold, D., see Neumaier, S. 194, 202, 231, 236 Arnold, E., see Schulz, C. 344
Arnold, J. 231 Arnold, J.R. 241 Arnold, M., see Bard, E. 249 Arnold, M., see Goslar, T. 254 Arpesella, C. 232 Arthur, R.J. 222, 224, 225 Asmerom, Y., see Cheng, H. 302 Assinder, D.J. 312, 394 ASTM, 18, 87, 116, 117 Aubert, M., see Chartier, F. 322 Augustson, J.H. 401 Aumann, D.C. 386, 389, 390 Aureli, A. 468 Aureli, A., see Burnett, W.C. 467 Aureli, A., see Povinec, P.P. 451, 467 Avignone, F.T., see Brodzinski, R.L. 168 Avignone III, T., see Brodzinski, R.L. 218, 231 Ayliffe, L.K., see Bird, M.I. 245 Ayranov, M. 311 Ayrault, S., see Piccot, D. 393 Äystö, J., see Moore, I.D. 337 Baba, N., see Kilius, L.R. 267 Backe, H. 347 Backe, H., see Sewtz, M. 347 Bacon, B.P., see Huh, C.A. 399 Baer, M., see Liao, C.-L. 297 Baglan, N. 304, 311 Baglan, N., see Pointurier, F. 305 Bailey, K. 91, 92 Bailey, K., see Chen, C.Y. 91, 342 Bailey, K., see Du, X. 91, 92 Bailey, K., see Sturchio, N.C. 91 Bailey, M.R., see Stradling, N. 287 Baillie, M.G.L., see Hughen, K.A. 249, 250 Baillie, M.G.L., see Reimer, P.J. 247, 249, 251 Baillie, M.G.L., see van der Plicht, J. 247, 250 Baker, J.D., see Mincher, B.J. 394 Ballestra, S., see Fowler, S.W. 138 Ballestra, S., see Fukai, R. 116 Ballestra, S., see Hamilton, T.H. 451
Author Index
Ballestra, S., see Holm, E. 311 Bamford, A., see Eriksson, M. 357, 362 Bandong, B., see Reijonen, J. 377 Bandura, D.R. 300 Bandura, D.R., see Tanner, S.D. 298, 310 Baranov, V.I., see Bandura, D.R. 300 Baranov, V.I., see Tanner, S.D. 298, 310 Bard, E. 249, 250, 253 Bard, E., see Hughen, K.A. 249, 250 Bard, E., see Reimer, P.J. 247, 249, 251 Bard, E., see van der Plicht, J. 247, 250 Bargholtz, C., see Kastlander, J. 430 Bargholtz, K., see Aage, H.K. 420 Bari, A., see Semkow, T.M. 164, 166, 194 Barinaga, C.J., see Eiden, G.C. 297, 320, 321 Barinaga, C.J., see Farmer, O.T. 319 Barker, A.W., see Meltzer, D.J. 253 Barnes, M.G. 410 Barnes, R.K., see Hill, D.M. 315 Barr, D.W., see Hicks, H.G. 150 Barre, N., see Sauvage, J. 337 Barrows, T.T., see Bird, M.I. 245 Bartomolei, P., see Plastino, W. 233, 234 Bartoš, P. 18 Barzakh, A.E. 342 Bastiaansen, J. 334 Bastiaansen, J., see Vandeweert, E. 334 Batist, L.K., see Alkhazov, G.D. 342 Baule, B. 19 Baumgartner, S. 250 Baxter, D.C., see Nygren, U. 306, 322 Baxter, M.S. 456 Baxter, M.S., see Hamilton, T.H. 451 Baxter, M.S., see Hirose, K. 138, 154, 155 Baxter, M.S., see Hursthouse, A.S. 390, 392, 393 Baxter, M.S., see Livens, F.R. 410 Baxter, M.S., see Osvath, I. 450, 451, 457 Baxter, M.S., see Povinec, P.P. 449–452, 456 Baxter, M.S., see Sanderson, D.C.W. 420
483
Bayliss, A., see Hughen, K.A. 249, 250 Bayliss, A., see Reimer, P.J. 247, 249, 251 Beals, D.M. 314 Beasley, T.M. 311 Beasley, T.M., see Kelley, J.M. 311 Beaumont, P.B., see Bird, M.I. 245 Beck, H.L. 408, 411–413, 416 Beck, J.W. 250, 252 Beck, J.W., see Biddulph, D.L. 257 Beck, J.W., see Burr, G.S. 249 Beck, J.W., see Cutler, K.B. 249 Beck, J.W., see Edwards, R.L. 249, 253 Beck, J.W., see Hughen, K.A. 249, 250 Beck, J.W., see Jull, A.J.T. 242, 243, 245 Beck, J.W., see Reimer, P.J. 247, 249, 251 Beck, J.W., see Richards, D.A. 250 Beck, J.W., see van der Plicht, J. 247, 250 Becker, D.A. 399 Becker, D.A., see Rook, H.L. 389 Becker, J.S. 305, 332 Becker, J.S., see Boulyga, S.B. 300, 320 Becker, J.S., see Boulyga, S.F. 305, 324, 332 Becker, J.S., see Desideri, D. 269, 324 Becker, J.S., see Izmer, A.V. 320 Becker, J.S., see Kerls, W. 319 Becker, J.S., see Vonderheide, A.P. 297, 321 Becker, J.S., see Zoriy, M.V. 297, 305, 321, 324, 332 Becker, K., see Wedekind, Ch. 450 Becker, St., see Krönert, U. 337 Beer, J., see Baumgartner, S. 250 Beer, J., see Oeschger, H. 210, 229, 234 Beer, J., see Vojtyla, P. 164, 166, 167, 171, 182, 194 Begg, B.D., see Conradson, S.D. 366 Beláˇn, T. 171 Bella, F., see Plastino, W. 233, 234 Belli, M., see Rosamilia, S. 398 Bellotti, E. 233 Belman, R.T., see Barnes, M.G. 410
484
Author Index
Belt, V.F., see Vance, D.E. 311 Belti, M., see Perna, L. 361 Bem, H. 399 Benedetti-Pichler, A., see Baule, B. 19 Benedict, M. 338 Benedik, L. 399 Benedik, L., see Byrne, A.R. 372, 383, 398, 399 Benedik, L., see Repinc, U. 398 Benke, R.R. 427 Benkhedda, K. 323 Bennett, B.A., see Parry, S.J. 386 Bennett, S.J., see McGeehin, J. 245 Benninghoven, A., see Erdmann, N. 334, 348 Benson, W.K., see Krey, P.W. 150 Benzing, R., see Parry, S.J. 386 Benzoubir, S., see Agarande, M. 322 Bercovitch, M. 223 Berezhnov, V.I., see Ikeuchi, Y. 155, 157 Berg, T. 349 Berg, T., see Maul, J. 334, 348 Berger, M.J., see Kim, L. 181 Berger, M.J., see Seltzer, S.M. 181 Berglund, M., see Denk, E. 346 Bergquist, B.A. 268 Berkovits, D. 280 Berkovits, D., see Paul, M. 280, 344 Bernhard, P., see Berg, T. 349 Berryman, N. 321 Berryman, N.G., see Song, M. 321, 397 Bertine, K.K., see Koide, M. 150 Bertl, W., see Kozlowski, T. 179 Bertrand, C.J.H., see Reimer, P.J. 247, 249, 251 Betti, M. 348, 355–359 Betti, M., see Erdmann, N. 334, 348, 349, 358 Betti, M., see Eriksson, M. 357, 362 Betti, M., see Jernström, J. 357, 361, 367 Betti, M., see Moreno, J.M.B. 397 Betti, M., see Tamborini, G. 348, 358 Betti, M., see Török, S. 357, 358, 360, 367
Bewers, J.M., see Sjöblom, K.-L. 450, 451, 456 Bexon, A., see Betti, M. 359 Bhat, I.S., see Jha, S.K. 390, 392, 393 Bhaumik, D.K., see Gibbons, R.D. 87 Bhowmick, G.K., see Wendt, K. 344 Biddulph, D., see Beck, J.W. 250, 252 Biddulph, D., see Currie, L.A. 89 Biddulph, D., see Jull, A.J.T. 245 Biddulph, D., see McHargue, L.R. 256 Biddulph, D.L. 257 Biddulph, D.L., see Jull, A.J.T. 243 Bienvenu, P. 319 Biersack, J.P., see Ziegler, J.F. 275 Bierwirth, P.N., see Wilford, J.R. 439 Bildstein, O., see Tamborini, G. 348, 358 Bilinski, H., see Fukai, R. 150 Billowes, J., see Moore, I.D. 337 Birch, C.P., see Parry, S.J. 386 Birchall, A., see Stradling, N. 287 Bird, J.R., see Tuniz, C. 241, 242, 332 Bird, M.I. 245 Birnstein, D., see Helbig, W. 210, 229 Blackwell, P.G., see Hughen, K.A. 249, 250 Blackwell, P.G., see Reimer, P.J. 247, 249, 251 Blackwell, P.G., see van der Plicht, J. 247, 250 Blair, I.M., see Thomas, B.W. 450 Blaise, J., see Worden, E.F. 339 Blaum, K. 299, 345 Blaum, K., see Müller, P. 337, 346 Blaum, K., see Wies, K. 342 Bleile, A., see Kraft, S. 272 Bleise, A., see Danesi, P.R. 288 Bloom, A.L., see Edwards, R.L. 249, 253 Bloom, A.L., see Fairbanks, R.G. 249– 251 Blowers, P., see Povinec, P.P. 458, 459 Bogaerts, A. 334 Böhm, J., see Neumaier, S. 194, 202, 231, 236 Bohn, L.L. 30, 35 Bokuniewicz, H. 474
Author Index
Bokuniewicz, H., see Burnett, W.C. 467 Bolivar, J.P., see Mas, J.L. 314–317 Bolle, W.D., see Richter, S. 268 Bollen, G., see Krönert, U. 337 Bonani, G., see Bond, G. 251 Bond, G. 251 Bond, L.A., see Beasley, T.M. 311 Bond, L.A., see Cooper, L.W. 301, 311 Bond, L.A., see Kelley, J.M. 311 Bonfield, R., see Leonard, K.S. 267 Bonfield, R., see Povinec, P.P. 458, 459 Bonn, J., see Peuser, P. 332 Bonne, A., see Wordel, R. 230 Bonté, P., see Reyss, J.-L. 194, 235 Boos, N., see Sauvage, J. 337 Boqué, R. 81 Borchers, W., see Schulz, C. 344 Bordeanu, C., see Paul, M. 280 Borgman, L.E. 29 Borretzen, P. 285 Bøtter-Jensen, L., see Tompson, I.M.G. 425–427 Bouisset, P. 384 Bouisset, P., see Agarande, M. 322 Bouisset, P., see Frechou, C. 384 Boulyga, S.B. 300, 320 Boulyga, S.F. 305, 324, 332 Boulyga, S.F., see Desideri, D. 269, 324 Boulyga, S.F., see Izmer, A.V. 320 Bourgeois, C., see Gutierrez, S. 418, 437 Bourlat, Y., see Musa, C. 458, 459 Bourles, D.L., see Braucher, R. 255 Bourlès, J., see Raisbeck, G.M. 255 Bouvier-Capely, C., see Baglan, N. 311 Bouwen, W., see Vandeweert, E. 334 Bowen, V.T. 137–139, 146, 151, 155 Bowen, V.T., see Nelson, D.M. 150 Bowen, V.T., see Sugihara, T.T. 146 Bowers, D.L., see Wolf, S.F. 322 Braga, E.S., see De Oliveira, J. 467, 473 Braga, E.S., see Povinec, P.P. 467, 472, 474 Brand, T.L., see Salbu, B. 19 Braucher, R. 255 Brauer, F.P., see Kaye, J.H. 210, 230
485
Braun, T., see Tölgyessy, J. 20, 36 Brinkmann, H.J., see Müller, G. 221 Bristow, G. 410 Brochard, E., see Bienvenu, P. 319 Brodzinski, R.L. 168, 171, 218, 231, 234 Broecker, W.S., see Quay, P.D. 154 Broecker, W.S., see Rocco, G.G. 138, 139, 146 Bronk Ramsey, C., see Reimer, P.J. 247, 249, 251 Bronk Ramsey, C., see van der Plicht, J. 247, 250 Bronson, F., see Shebell, P. 430 Brown, C.F. 320 Brown, E., see Lind, O.C. 285 Brown, E.T., see Braucher, R. 255 Brown, J., see Hill, A.E. 465 Brown, J., see Skipperud, L. 285, 286 Brown, T.A. 265, 269, 277 Brown, T.A., see McAninch, J.E. 277 Brubaker, L., see Lertzman, K. 252 Brück, K., see Wies, K. 342 Bruhn, F. 245 Brunelle, T., see Fehn, U. 384 Bruyneel, B., see Kudryavtsev, Y. 337 Bryan, N.D., see Keith-Roach, M.J. 287 Bryan, N.D., see Marsden, O.J. 287 Bucher, B., see Rybach, L. 439 Buck, C.E., see Hughen, K.A. 249, 250 Buck, C.E., see Reimer, P.J. 247, 249, 251 Buck, C.E., see van der Plicht, J. 247, 250 Budagov, Yu.A., see Povinec, P. 233 Buesseler, K.O. 144, 150, 151, 155, 157, 301 Buesseler, K.O., see Dai, M.H. 332 Buesseler, K.O., see Povinec, P.P. 142, 148, 154 Buheitel, F., see Aumann, D.C. 386, 389, 390 Bullister, J.L., see Warner, M.J. 138 Bunzl, K. 31, 43, 348, 408, 410 Bunzl, K., see Hillman, U. 416, 417 Bunzl, K., see Tschiersch, J. 39
486
Author Index
Buonicore, A.J., see Theodore, L. 38 Burchuladze, A., see Pagava, S. 202 Burger, M., see Ayranov, M. 311 Burkarrt, W., see Danesi, P.R. 358 Burkart, W., see Danesi, P.R. 288 Burke, J.C. 42 Burnett, W., see La Rosa, J.J. 139, 152, 153 Burnett, W.C. 467, 471, 473, 474 Burnett, W.C., see De Oliveira, J. 467, 473 Burnett, W.C., see Kontar, E.A. 467 Burnett, W.C., see Povinec, P.P. 451, 467 Burnett, W.C., see Taniguchi, M. 467, 469 Burney, D.A. 253 Burney, L.P., see Burney, D.A. 253 Burr, G., see McHargue, L.R. 256 Burr, G.S. 245, 249 Burr, G.S., see Beck, J.W. 250, 252 Burr, G.S., see Biddulph, D.L. 257 Burr, G.S., see Currie, L.A. 89 Burr, G.S., see Cutler, K.B. 249 Burr, G.S., see Edwards, R.L. 249, 253 Burr, G.S., see Hughen, K.A. 249, 250 Burr, G.S., see Jull, A.J.T. 242, 243, 245, 257 Burr, G.S., see McGeehin, J. 245 Burr, G.S., see McNichol, A.P. 242, 245 Burr, G.S., see Reimer, P.J. 247, 249, 251 Burrough, P.A. 11 Busby, R.G., see MacCartney, M. 314 Busby, R.G., see McCartney, M. 267 Bushaw, B.A. 336, 344 Bushaw, B.A., see Blaum, K. 299, 345 Bushaw, B.A., see Müller, P. 337, 346 Bushaw, B.A., see Pibida, L. 397 Bushaw, B.A., see Schumann, P.G. 347 Bushaw, B.A., see Wendt, K. 342, 344 Bykov, A.A., see Alkhazov, G.D. 342 Byrde, F., see Laedermann, J.P. 408 Byrne, A.R. 372, 383, 390, 393, 394, 398, 399 Byrne, A.R., see Benedik, L. 399 Byrne, A.R., see Dermelj, M. 401
Byrnes, M.E.
18
Cabaret, L., see Sauvage, J. 337 Cabianca, T., see Danesi, P.R. 288 Cabioch, G., see Bard, E. 249 Cabioch, G., see Burr, G.S. 249 Cabioch, G., see Cutler, K.B. 249 Cable, J.E. 473 Cable, J.E., see Burnett, W.C. 467 Cable, J.E., see Taniguchi, M. 467 Cagnat, X., see Bouisset, P. 384 Calmet, D., see Agarande, M. 322 Calmet, D., see Bouisset, P. 384 Calmet, D., see Frechou, C. 384 Calsoyas, L., see Beck, J.W. 250, 252 Cambray, R.S., see Williams, D. 410 Campbell, M.J., see Danesi, P.R. 288 Campbell, P., see Moore, I.D. 337 Cao, L., see Fairbanks, R.G. 249–251 Carchon, R., see Müller, G. 221 Carey, A.E., see Nelson, D.M. 150 Carlin, R.P. 25 Carling, R.S., see Fifield, L.K. 268 Carmichael, H., see Bercovitch, M. 223 Caruso, J.A., see Vonderheide, A.P. 297, 321 Casamiquela, R., see Nuñez, L.A. 253 Casas, M., see Miró, M. 147 Casso, S.A., see Livingston, H.D. 138 Cechova, A., see Cimbák, Š. 39 Cecil, L.D., see Paul, M. 344 CELLAR, 108, 112 Cerdà, V., see Miró, M. 147 CERN, 175–177, 180–182 Chai, C.F., see Hou, X.L. 377, 398 Chai, Z.F., see Hou, X.L. 377, 398 Chamberlain, A.C. 410 Chamizo, E., see Wacker, L. 272–274, 279, 280, 332 Chan, L.S., see Chen, M.Q.F. 430 Chang, B.U., see Kim, C.K. 151 Chanton, J., see Burnett, W.C. 467 Chao, J.H. 388, 390, 397, 398 Chappell, J.M.A., see Edwards, R.L. 249, 253
Author Index
Charette, M., see Bokuniewicz, H. 474 Charette, M.A., see Burnett, W.C. 467 Chartier, F. 322 Chaykovskaya, E., see Hirose, K. 138, 154, 155 Chaykovskaya, E., see Ikeuchi, Y. 155, 157 Chebotayev, V.P., see Letokhov, V.S. 335 Chen, C.Y. 91, 342 Chen, C.Y., see Bailey, K. 91, 92 Chen, J.C., see Lee, T. 397 Chen, J.H. 302, 304 Chen, J.J., see Edwards, R.L. 41 Chen, L.-T., see Kelly, W.R. 86, 95 Chen, M.Q.F. 430 Chen, Q.J. 152, 153, 312, 390–392, 394, 396, 399, 401 Chen, Q.J., see Povinec, P.P. 458, 459 Chen, Z., see Bogaerts, A. 334 Cheng, H. 302 Cheng, H., see Cutler, K.B. 249 Cheng, Y.-S. 38 Cherkezyan, V.O., see Khitrov, L.M. 43 Cheseby, M., see Bond, G. 251 Chevallier, A., see Hubert, Ph. 217 Chiappini, R. 155 Chiappini, R., see Baglan, N. 304 Chichester, D.L. 377 Child, D., see Hotchkis, M.A.C. 270, 278, 287 Child, D.P. 281, 282 Chine-Cano, E., see Danesi, P.R. 358 Chinea-Cano, E., see Eriksson, M. 357, 362 Chiu, T.-C., see Fairbanks, R.G. 249–251 Cho, Y.H., see Lee, C.W. 147 Choi, G.S., see Lee, C.W. 147 Choi, S.W., see Kim, C.K. 151 Choi, Y.W., see Lee, C.W. 147 Choppin, G.R. 150 Chou, F.C., see Chao, J.H. 388, 390 Chow, T.J., see Koide, M. 150 Christensen, G.C., see Yiou, F. 266 Christensen, I., see Chen, Q.J. 390–392 Christoff, J., see Burnett, W.C. 467
487
Christoff, J., see De Oliveira, J. 467, 473 Christoffers, D., see Muramatsu, Y. 386 Chudý, M., see Beláˇn, T. 171 Chudý, M., see Hlinka, V. 198 Chudý, M., see Pagava, S. 202 Chumichev, V.B., see Hirose, K. 138, 154, 155 Chumichev, V.B., see Ikeuchi, Y. 155, 157 Chung, C.S., see Ikeuchi, Y. 155, 157 Chung, C.S., see Kim, C.K. 151 Cimbák, Š. 39 Ciurapinski, A., see Danesi, P.R. 358 Clacher, A.P., see Fifield, L.K. 265, 269, 271, 280, 282, 332 Clark, D.L., see Conradson, S.D. 366 Clark, P.U., see Alley, R.A. 251 Clark, P.U., see Dyke, A.S. 254 Clayton, C.G., see Thomas, B.W. 450 Clough, A., see Alley, R.B. 251, 253 Cocconi, G. 214 Cochran, W.G. 3, 7, 9, 14 Cockburn, H.A.P. 254 Cohen, A.S. 323 Cohen, C.A. 117 Colaresi, J.F., see Semkow, T.M. 164, 166, 194 Coleman, G.H. 150 Colin, F., see Braucher, R. 255 Collon, P. 91 Coluzza, J., see Krey, P.W. 150 Comanducci, J.-F., see Levy-Palomo, I. 467 Comanducci, J.-F., see Osvath, I. 450, 451, 458 Comanducci, J.-F., see Povinec, P.P. 228, 451, 453, 464, 466, 467 Comanducci, J.F., see Osvath, I. 450, 451, 453, 464 Comanducci, J.F., see Povinec, P.P. 108, 164, 166, 169, 194, 196, 202, 228, 234, 451, 453, 467, 468, 472, 474 Conally, R.E., see Kaye, J.H. 210, 230 Coniglio, W.A., see Hess, C.T. 40 Conklin, A.R. 376
488
Author Index
Conney, S., see Leonard, K.S. 267 Connolly, M.V., see Currie, L.A. 83 Conradson, S.D. 366 Cook, B., see Duval, J.S. 411, 413 Cooper, J., see Sjöblom, K.-L. 450, 451, 456 Cooper, J.A. 164 Cooper, L.W. 301, 311 Cooper, M., see Povinec, P.P. 458, 459 Copplestone, D., see Tyler, A.N. 435, 436 Cornett, J., see Vais, V. 305, 310 Cornett, R.J., see Epov, V.N. 322, 323 Cornett, R.J., see Larivière, D. 322, 323 Cornett, R.J., see Lariviere, D. 332 Corrège, T., see Burr, G.S. 249 Cortes, E., see Guinn, V.P. 374 Cossonnet, C., see Baglan, N. 311 Courtney, C., see Jull, A.J.T. 242 Cox, C.C., see Brown, T.A. 265, 269, 277 Cox, D.R. 62, 104, 108 Craig, M.A., see Wilford, J.R. 439 Crawford, J.E., see Sauvage, J. 337 Creamer, S.C., see Miller, J.M. 450, 459 Cremers, A. 18 Cremers, A., see Wauters, J. 18 Cremonesi, O., see Bellotti, E. 233 Cressie, N. 10, 11 Cressie, N.A. 18 Cresswell, A.J., see Sanderson, D.C.W. 409, 418, 424, 437–442 Cresswell, R.G., see Bird, M.I. 245 Cresswell, R.G., see Fifield, L.K. 265, 268, 269, 271, 280, 282, 332 Cresswell, R.G., see Oughton, D.H. 285, 286, 303, 332 Creswell, A.J. 437 Croudace, I.W., see Taylor, R.N. 305 Croudace, I.W., see Warneke, T. 150 Crummett, W.B., see Kurtz, D.A. 64 Csongor, E. 96 Cullen, H., see Bond, G. 251 Culp, R.A., see Noakes, J.E. 450 Cunnane, J.C., see Wolf, S.F. 322
Currie, L.A. 49, 50, 52, 53, 57–60, 65, 66, 72, 74, 75, 83–85, 88, 89, 91, 93, 94, 98, 100, 108, 109, 114, 115, 117, 120, 121, 123, 128, 130, 131 Cusimano, G., see Povinec, P.P. 451, 467 Cusimano, G., see Schiavo, M.A. 467 Cutler, K., see Burr, G.S. 249 Cutler, K.B. 249 Cutler, K.B., see Hughen, K.A. 249, 250 Cutler, K.B., see Reimer, P.J. 247, 249, 251 Cutshall, N.H. 412, 424 CX/MAS, 53 Da Silva, A. 221 Dahlgaard, H., see Aarkrog, A. 362 Dahlgaard, H., see Chen, Q.J. 152, 153, 312, 390–392, 394, 396, 399, 401 Dahlgaard, H., see Eriksson, M. 357, 362 Dahlgaard, H., see Hou, X.L. 318, 385, 386, 388, 390, 397 Dahlgaard, H., see Keith-Roach, M. 317 Dahlgaard, H., see Keith-Roach, M.J. 394 Dahlgaard, H., see Lind, O.C. 367 Dahlgaard, H., see McMahon, C.A. 362 Dahlgaard, H., see Povinec, P.P. 458, 459 Dahlgaard, H., see Sturup, S. 305 Dai, M.H. 332 Dalhgaard, H., see Hou, X.L. 384–390 Damon, P.E. 251 Damon, P.E., see Donahue, D.J. 65 Damon, P.E., see Hughen, K.A. 249, 250 Damon, P.E., see Klein, J. 247 Damon, P.E., see McHargue, L.R. 256 Damon, P.E., see Reimer, P.J. 247, 249, 251 Danesi, P.R. 288, 358 Danesi, P.R., see Salbu, B. 357, 358, 367 Dang, H.S. 398, 399 Danko, B. 399 Dannecker, W., see Kerls, W. 319 Danzer, K., see Eckschlager, K. 21 Darnley, A.G. 410 Das Neves, J., see Hult, M. 202
Author Index
Dassie, D., see Hubert, Ph. 217 Dau, W.D., see Allkofer, O.C. 173 David, B., see Bird, M.I. 245 David, G., see Semkow, T.M. 164, 166, 194 Davidson, D.A., see Tyler, A.N. 416, 430, 433–435 Davis, C.B. 75 Davis, J.R., see Jones, D.G. 409, 459 Davis, P.T., see McGeehin, J. 245 Dawood, Y., see Sturchio, N.C. 91 Day, J.P., see Fifield, L.K. 265, 268, 269, 271, 280, 282, 332 Day, J.P., see Keith-Roach, M.J. 265, 287 Day, J.P., see Marsden, O.J. 270, 287 Day, J.P., see Salbu, B. 19 Day, J.P., see Oughton, D.H. 285, 286, 303, 332 De Geer, L.-E. 107, 108 de Hoog, J., see Ro, C.-U. 363 de Jong, E., see Sutherland, R.A. 410 De Lellis, C., see van Put, P. 450 de Meijer, R.J., see Hendricks, P.H.G.M. 420, 421 de Meijer, R.J., see Lindsay, R. 430 de Mora, S.J., see Wyse, E.J. 264, 269, 306, 371 De Oliveira, J. 467, 473 de Oliveira, J., see Burnett, W.C. 473 de Oliveira, J., see Povinec, P.P. 451, 467, 472, 474 de Pablo, J., see Perna, L. 361 De Preter, P., see Cremers, A. 18 Debauche, A., see Sanderson, D.C.W. 418, 424, 437–440, 442 Debauche, A., see van Put, P. 450 Debertin, K. 164, 169 Deboffle, D., see Raisbeck, G.M. 255 Decampo, J., see Beck, H.L. 408, 411– 413, 416 deCorte, F. 374 Delfanti, R., see Povinec, P.P. 142, 148, 154 Delucchi, E., see Foti, S. 394–396 Deme, S., see Tompson, I.M.G. 425–427
489
deMenocal, P., see Bond, G. 251 Deming, S.N., see Massart, D.L. 75 den Auwer, Ch., see Conradson, S.D. 366 Deng, L., see Payne, M.G. 343 Denisov, V.P., see Barzakh, A.E. 342 Denk, E. 346 Denoon, D., see Kershaw, P. 459, 463, 464 Denoon, D., see McCartney, M. 459, 462–464 Denoon, D.C., see Poole, A.J. 459, 463 Denschlag, H.O., see Peuser, P. 332 Dep, L. 221 Department of Energy, DOE, 35, 37, 39 Dermelj, M. 401 Deschamps, P. 304 Desideri, D. 269, 324 Desideri, D., see Boulyga, S.F. 332 di Tada, M.I., see Bird, M.I. 245 di Tada, M.L., see Fifield, L.K. 265, 268, 269, 271, 280, 282, 332 Di Tada, M.L., see Oughton, D.H. 285 Diamond, H., see Horwitz, E.P. 147, 153 Dibb, J.E., see Currie, L.A. 83, 89 Dickin, A.P., see Li, W.-X. 41 Dickson, H.W. 411, 413 Diel, S., see Müller, P. 337, 346 Dietl, F., see Tschiersch, J. 39 Dietz, M.L., see Horwitz, E.P. 147, 153 Dietze, H.-J., see Becker, J.S. 305, 332 Dietze, H.J., see Kerls, W. 319 Dincauze, D.F., see Meltzer, D.J. 253 Ding, M., see Conradson, S.D. 366 Ding, W.J., see Hou, X.L. 385, 386, 388, 390 Ditze, H.-J., see Boulyga, S.B. 300 Dixon, P.M. 6 Döbeli, M., see Grajcar, M. 243 Döbeli, M., see Suter, M. 272 Donahue, D.J. 65, 242, 245 Donahue, D.J., see Beck, J.W. 250, 252 Donahue, D.J., see Biddulph, D.L. 257 Donahue, D.J., see Burr, G.S. 245, 249 Donahue, D.J., see Currie, L.A. 83, 89
490
Author Index
Donahue, D.J., see Edwards, R.L. 249, 253 Donahue, D.J., see Jull, A.J.T. 245 Donahue, D.J., see McHargue, L.R. 256 Dong, C.Z., see Sewtz, M. 347 Donoghue, J.C., see Currie, L.A. 89 Donohue, D., see Danesi, P.R. 358 Donohue, D.L. 348 Donohue, D.L., see Tamborini, G. 358 Dorda, J., see Malczewski, D. 430 Dorhout, P.K., see Conradson, S.D. 366 Doucelance, R., see Deschamps, P. 304 Douglas, D.J. 297, 300 Dovlete, C. 131 Dovlete, C., see Povinec, P.P. 458, 459 Dowdall, M., see Borretzen, P. 285 Downey, S.W. 313 Drady, E.D., see Conradson, S.D. 366 Drakopoulos, M., see Salbu, B. 357, 367 Dretzke, A., see Sewtz, M. 347 Drozcho, E., see Oughton, D.H. 285 Druffel, E.R.M., see Burr, G.S. 245 Druffel, E.R.M., see Edwards, R.L. 249, 253 Du, X. 91, 92 Du, X., see Bailey, K. 91, 92 Du, X., see Sturchio, N.C. 91 Ducourtieux, M., see Sauvage, J. 337 Duhr, A., see Bruhn, F. 245 Dulaiova, H., see Burnett, W.C. 467, 471, 473, 474 Dulaiova, H., see Povinec, P.P. 451, 467 Dulaiova, H., see Taniguchi, M. 469 Duncan, A.J. 18 Duong, H.T., see Sauvage, J. 337 ˇ Durana, L., see Beláˇn, T. 171 Durazo, R., see Hill, A.E. 465 Durˇcík, M., see Sýkora, I. 169, 171, 196 Dutton, M., see Betti, M. 359 Duval, J.S. 411, 413 Dvoracek, D.K., see Noakes, J.E. 450 Dvorak, Z., see Csongor, E. 96 Dybcynski, R., see Danko, B. 399 Dyck, W., see Steinkopff, T. 92
Dyer, R.S., see Sjöblom, K.-L. 450, 451, 456 Dyke, A.S. 254 Eastgate, A., see Haldiman, M. 318 Eberhardt, K., see Backe, H. 347 Eberhardt, K., see Eichler, B. 333 Eberhardt, K., see Erdmann, N. 339 Eberhardt, K., see Maul, J. 334, 348 Eberhardt, K., see Sewtz, M. 347 Eberhardt, L.L., see Gilbert, R.O. 410 Eckerman, K.F., see Leggett, R.W. 287, 288 Eckschlager, K. 21 Edwards, R.L. 41, 249, 253 Edwards, R.L., see Beck, J.W. 250, 252 Edwards, R.L., see Burr, G.S. 245, 249 Edwards, R.L., see Chen, J.H. 302, 304 Edwards, R.L., see Cheng, H. 302 Edwards, R.L., see Cutler, K.B. 249 Edwards, R.L., see Hughen, K.A. 249, 250 Edwards, R.L., see Reimer, P.J. 247, 249, 251 Edwards, R.R. 390 Efron, B. 32 Egeberg, P.K., see Fehn, U. 267 Egelhof, P., see Kraft, S. 272 Egorov, O.B. 311 Egorov, O.B., see Grate, J.W. 153 Ehrlich, G. 51 Eichler, B. 333 Eiden, G.C. 297, 320, 321 Eijgenhuijsen, E.M., see Currie, L.A. 108, 109, 120, 121 Eisenbud, M. 169 Eisenhart, C. 72, 81 Eisenhut, S., see Heumann, K.G. 336 El Alfy, Z., see Sturchio, N.C. 91 El Kaliouby, B., see Sturchio, N.C. 91 Elliot, B., see Barnes, M.G. 410 Ellison, A.M., see Dixon, P.M. 6 Elmore, D. 241, 243 Elmore, D., see Dep, L. 221 Elmore, D., see Fehn, U. 384
Author Index
Elsen, A., see Cremers, A. 18 Elster, J. 210 EML, 426 Engelmann, Ch., see May, S. 390, 392, 393 Engelsmann, J., see Pavlik, B. 437 Engfer, R., see Kozlowski, T. 179 Engfer, R., see Schröder, W.U. 179, 180 England, J., see Dyke, A.S. 254 English, J.A., see Reijonen, J. 377 Environmental Protection Agency, EPA, 18, 23, 24, 26, 30, 40, 41 Epov, V.N. 322, 323 Epov, V.N., see Larivière, D. 322, 323 Erdmann, N. 334, 339, 348, 349, 358 Erdmann, N., see Berg, T. 349 Erdmann, N., see Eichler, B. 333 Erdmann, N., see Nunnemann, M. 338 Erdmann, N., see Passler, G. 337, 338, 341 Erdmann, N., see Waldek, A. 339 Eriksson, M. 357, 362, 363, 376 Eriksson, M., see Jernström, J. 357, 361, 367 Eriksson, M., see Moring, M. 362 Espinosa-Faller, F.J., see Conradson, S.D. 366 Essington, E.H., see Fowler, E.B. 410 Essington, E.H., see Gilbert, R.O. 410 Estela, J.M., see Miró, M. 147 Etherington, G. 287 Etherington, G., see Stradling, N. 287 Evans, D., see Benkhedda, K. 323 Evans, R.D., see Epov, V.N. 322, 323 Evans, R.D., see Larivière, D. 322, 323 Evans, R.D., see Lariviere, D. 332 Everett, S.J. 290 Excoffier, E., see Bienvenu, P. 319 Fabryka-Martin, J., see Dep, L. 221 Faestermann, T., see Wallner, C. 279 Fairbank, W.M., see LaBelle, R.D. 344 Fairbanks, R.G. 249–251 Fairbanks, R.G., see Bard, E. 249, 253
491
Fairbanks, R.G., see Hughen, K.A. 249, 250 Fairbanks, R.G., see Reimer, P.J. 247, 249, 251 Fairbanks, T.W., see Fairbanks, R.G. 249–251 Fairman, W.D., see Horwitz, E.P. 147, 153 Faleschini, H., see Aumann, D.C. 386, 390 Falkenberg, G., see Eriksson, M. 357, 362 Falkenberg, G., see Jernström, J. 357, 361, 367 Faller, S., see Shebell, P. 430 Farias, L.A., see De Oliveira, J. 467, 473 Farmer, O.T. 319 Farmer III, O.T., see Egorov, O.B. 311 Fattinger, K., see Denk, E. 346 Faure, G. 41 Fedorov, D.V., see Barzakh, A.E. 342 Fedoseyev, V.N., see Alkhazov, G.D. 342 Fehn, U. 267, 384 Fehn, U., see Moran, J.E. 267 Fehn, U., see Snyder, G. 267 Fehn, U., see Snyder, G.T. 267 Feigl, F. 51 Feldmann, I. 298, 299 Feldmann, R., see Backe, H. 347 Feldstein, H., see Berkovits, D. 280 Feldstein, H., see Paul, M. 344 Ferguson, J.M., see Sanderson, D.C.W. 439 Fernand, L., see Hill, A.E. 465 Fernandez-Alonso, M. 439 Fesefeldt, H.C. 180, 182 Fesenko, S., see Gering, F. 408, 430 Fettweis, B., see Müller, G. 221 Fifield, K., see Stradling, N. 287 Fifield, L.K. 242, 243, 265, 268–271, 279, 280, 282, 332 Fifield, L.K., see Bird, M.I. 245 Fifield, L.K., see Borretzen, P. 285 Fifield, L.K., see Etherington, G. 287 Fifield, L.K., see Everett, S.J. 290
492
Author Index
Fifield, L.K., see Keith-Roach, M.J. 265, 287 Fifield, L.K., see Lind, O.C. 285 Fifield, L.K., see Marsden, O.J. 270, 287 Fifield, L.K., see Newton, D. 287, 288 Fifield, L.K., see Oughton, D.H. 285, 286, 303, 332 Fifield, L.K., see Skipperud, L. 285, 286 Fifield, L.K., see Tims, S.G. 266, 284 Fifield, L.K., see Wacker, L. 272–274, 279, 280, 313, 332 Figueoredo, A.M.G., see Vasconcellos, M.B.A. 401 Filby, R.H., see Glover, S.E. 399 Fink, D., see Hotchkis, M. 278, 279, 287 Fink, D., see Hotchkis, M.A.C. 270, 278, 287 Fink, D., see Tuniz, C. 241, 242, 332 Fiorini, E., see Bellotti, E. 233 Firestone, R.B., see Reijonen, J. 377 Flanagan, S., see Lane, S.L. 39 Fletcher, R.A., see Currie, L.A. 83, 89 Fogh, C.L., see Hou, X.L. 385, 386, 388, 390, 397 Folsom, T.R. 138, 139 Folsom, T.R., see Hodge, V.F. 151, 152 Fonger, W., see Simpson, J.A. 223 Forcina, V., see Tamborini, G. 358 Ford, D.C., see Li, W.-X. 41 Foreman, R.L., see Hodge, V.F. 151, 152 Forkel-Wirth, D., see Sauvage, J. 337 Foster, D.A. 323 Foti, S. 394–396 Fourie, P.J., see Wasserman, H.J. 314 Fowler, E.B. 410 Fowler, E.B., see Gilbert, R.O. 410 Fowler, S.W. 138 Fox, A., see O’Donnell, R.G. 263, 269 Fox, V., see Povinec, P.P. 458, 459 Franchoo, S., see Kudryavtsev, Y. 337 Frank, G., see Steinkopff, T. 92 Franks, S., see Snyder, G.T. 267 Frechou, C. 384 Fred, M., see Worden, E.F. 339 Freimann, S.B., see Simpson, J.A. 223
French, J.B., see Douglas, D.J. 300 Frenzel, S., see Steinkopff, T. 92 Fridman, Sh.D., see Kogan, R.M. 408 Friedmann, L., see Aumann, D.C. 386, 390 Friedrich, M., see Hughen, K.A. 249, 250 Friedrich, M., see Liechtenstein, V.K. 275 Friedrich, M., see Reimer, P.J. 247, 249, 251 Friedrich, M., see Ruf, M. 390, 392, 393 Friedrich, M., see van der Plicht, J. 247, 250 Fritzsche, S., see Sewtz, M. 347 Froehlich, K., see Povinec, P.P. 458, 459 Fuchs, N.A. 38 Fuge, R., see Rucklidge, J. 267 Fujita, A., see Mamuro, T. 31 Fukai, R. 116, 150 Fukai, R., see Fowler, S.W. 138 Fulker, M.J., see Parry, S.J. 386 Fülöp, M. 427 Funck, E., see Neumaier, S. 194, 202, 231, 236 Funk, H., see Eichler, B. 333 Furtado, V.V., see De Oliveira, J. 467, 473 Gabelmann, H., see Peuser, P. 332 Gaiser, T. 165 Galang, C., see Burr, G.S. 249 Gallup, C., see Burr, G.S. 249 Gallup, C.D., see Cheng, H. 302 Gallus, S., see Heumann, K.G. 336 Gammon, R.H., see Warner, M.J. 138 Gann, J., see Jull, A.J.T. 243 Gann, J., see Steadman, D.W. 253, 254 Garabedian, G., see Reijonen, J. 377 Garcia-Leon, M., see Mas, J.L. 314–317 Garcia-León, M., see Santos, F.J. 318 Garner, R.J., see Chamberlain, A.C. 410 Garwan, M.A., see Kilius, L.R. 267 Garzanov, E., see Guinn, V.P. 374 Gascard, J.C. 266
Author Index
Gasparro, J., see Hult, M. 235 Gasparro, J., see Laubenstein, M. 108, 163, 202, 203 Gastaud, J., see Hamilton, T.H. 451 Gastaud, J., see Ikeuchi, Y. 155, 157 Gastaud, J., see La Rosa, J.J. 139, 152, 153, 311 Gastaud, J., see Povinec, P.P. 138, 148, 149, 155, 458, 459, 467, 472, 474 Gatto, L., see Povinec, P.P. 451, 467 Gatto, L., see Schiavo, M.A. 467 Gaudino, S., see Rosamilia, S. 398 Gaudry, A., see Frechou, C. 384 Gaudry, A., see Piccot, D. 393 Gavin, D., see Lertzman, K. 252 Gayol, J., see Povinec, P.P. 451, 456 Geckeis, H., see Kuczewski, B. 311 Geertsema, M., see Jull, A.J.T. 252–254 Geertsema, M., see Sanborn, P. 252, 253 Gehrike, K., see Sowa, W. 411, 413 Geiszler, K.N., see Brown, C.F. 320 Geitel, F., see Elster, J. 210 Genevey, J., see Sauvage, J. 337 Gent, C., see Marsden, O.J. 287 Gentens, J., see Kudryavtsev, Y. 337 Geppert, C., see Blaum, K. 299, 345 Geppert, C., see Denk, E. 346 Geppert, C., see Müller, P. 337, 346 Geppert, C., see Wies, K. 342 Geppert, Ch., see Moore, I.D. 337 Gerber, M., see Krönert, U. 337 Gerchikov, M., see Betti, M. 359 Gering, F. 408, 430 Germain, P. 390, 392, 393 Gerstmann, U., see Wallner, C. 279 Gervasio, G., see Bellotti, E. 233 Ghaleb, B., see Deschamps, P. 304 Ghelberg, S., see Berkovits, D. 280 Ghelberg, S., see Paul, M. 280 Giacomini, J.J., see Barnes, M.G. 410 Giannitrapani, M., see Sanderson, D.C.W. 439, 441, 442 Gibbons, R.D. 75, 87, 113
493
Gicquel, F., see Reijonen, J. 377 Gijbels, R., see Bogaerts, A. 334 Gijsels, L., see Salbu, B. 357, 358 Gilbert, R.O. 3, 8, 9, 14, 18, 410 Gilbert, R.O., see Fowler, E.B. 410 Gilbert, R.O., see Hassig, N.L. 18, 31 Gilbert, R.O., see Pulsipher, B.A. 13 Girod, M., see Sauvage, J. 337 Gladney, E.S. 399, 401 Gleisberg, B., see Niese, S. 194, 202, 234, 235 Glover, S.E. 399 Glower, R.N. 210 Godfrey, L.R., see Burney, D.A. 253 Goerner, W. 394, 396 Gogolak, C., see Beck, H.L. 408, 411– 413, 416 Goldberg, E.D., see Koide, M. 150 Goldbrunner, T. 234 Goldstein, S.I., see Schramm, A. 250 Golser, R., see Kraft, S. 272 Golser, R., see Liechtenstein, V.K. 275 Golser, R., see Steier, P. 275, 278, 279 Golser, R., see Vockenhuber, C. 279 Golser, R., see Winkler, S. 279 Gómez, E., see Miró, M. 147 Goodal, P.S., see Marsden, O.J. 270, 287 Goodman, S.M., see Burney, D.A. 253 Gordon, P.L., see Conradson, S.D. 366 Gore, B., see Donahue, D.J. 65 Görner, W., see Helbig, W. 229 Gorody, A.W., see Snyder, G.T. 267 Goroncy, I., see Povinec, P.P. 138, 148, 149, 155 Goslar, T. 254 Gosse, J. 254 Gosse, J., see McGeehin, J. 245 Gotelli, N.J., see Dixon, P.M. 6 Govindara, K. 411 Grab, C.H., see Kozlowski, T. 179 Graham, C., see Ruffel, A. 430 Grajcar, M. 243 Grajcar, M., see Suter, M. 272 Grasty, R.L. 410, 411, 413, 420, 427, 430 Grasty, R.L., see Løvborg, L. 410
494
Author Index
Grate, J.W. 153 Grate, J.W., see Egorov, O.B. 311 Gray, A.L. 296 Gray, S.C., see Cutler, K.B. 249 Grayson, D.K., see Meltzer, D.J. 253 Grebmeier, J.M., see Cooper, L.W. 301, 311 Green, R.H. 19 Greenberg, R.R. 399 Greenwell, R.D., see Navratil, J.D. 18 Grgula, M., see Beláˇn, T. 171 Grgula, M., see Cimbák, Š. 39 Grgula, M., see Csongor, E. 96 Grieder, P.K.F., see Allkofer, O.C. 172, 174, 184 Grieve, I.C., see Tyler, A.N. 416, 430, 433–435 Grimn, C.A., see Glover, S.E. 399 Groening, M., see Povinec, P.P. 451, 467, 472, 474 Gröning, M., see Povinec, P.P. 458, 459 Grootes, P.M., see Bruhn, F. 245 Grootes, P.M., see Fairbanks, R.G. 249– 251 Grootes, P.M., see Voelker, A.H.L. 249 Grüning, C. 337–339 Grüning, C., see Erdmann, N. 334, 348 Grüning, C., see Sewtz, M. 347 Grüning, C., see Waldek, A. 339 Grüttmüller, M., see Wedekind, Ch. 450 Guegueniat, P., see Germain, P. 390, 392, 393 Guilderson, T.P., see Fairbanks, R.G. 249–251 Guilderson, T.P., see Hughen, K.A. 249, 250 Guilderson, T.P., see Reimer, P.J. 247, 249, 251 Guilderson, T.P., see van der Plicht, J. 247, 250 Guillot, L. 420, 421 Guillot, L., see Gutierrez, S. 418, 437 Guinn, V.P. 374 Guryn, W., see Semkow, T.M. 164, 166, 194
Gutierrez, S. 418, 437 Gutknecht, D., see Hubert, Ph. 217 Gy, P.M. 19 Habrekke, H., see Lindahl, I. 410 Hadley, M.J. 430 Hafer, K.M., see Ketterer, M.E. 268, 302, 332 Hagenauer, R., see Shebell, P. 430 Haire, R.G., see Conradson, S.D. 366 Haire, R.G., see Sewtz, M. 347 Hajdas, I., see Bond, G. 251 Haldiman, M. 318 Halicz, L., see Becker, J.S. 305 Halicz, L., see Boulyga, S.F. 305, 324 Halicz, L., see Zoriy, M.V. 297, 305, 321, 324 Hallett, D., see Lertzman, K. 252 Hallett, D.J. 252 Halliday, A.N., see Luo, X. 324 Halvadakis, C.P., see Anastasis, B.P. 430 Halverson, J.E., see Buesseler, K.O. 150, 157, 301 Hamada, T., see Kimura, T. 147 Hamajima, Y., see Komura, K. 140, 202 Hamelin, B., see Bard, E. 249, 253 Hamilton, T., see Muramatsu, Y. 150 Hamilton, T., see Povinec, P.P. 458, 459 Hamilton, T.F., see Brown, T.A. 265, 269, 277 Hamilton, T.F., see McAninch, J.E. 277 Hamilton, T.H. 451 Hancock, G.J., see Everett, S.J. 290 Hancock, G.J., see Tims, S.G. 266, 284 Handl, J. 386, 390 Handl, J., see Schmidt, A. 318 Hanna, G.C., see Bercovitch, M. 223 Hanneman, W.W., see Kurtz, D.A. 64 Hanor, J.S., see Moran, J.E. 267 Hansen, C.S., see LaBelle, R.D. 344 Hansen, H.J.M., see Chen, Q.J. 394, 396 Hansen, J., see Løvborg, L. 410 Hanson, W.C. 338 Hardeman, F., see Sanderson, D.C.W. 418, 424, 437–440, 442
Author Index
Harding, J.L., see Noakes, J.E. 450 Harding, J.L., see Senftle, F.E. 450 Hardy, E.P., see Krey, P.W. 150 Harley, J.H. 138 Harms, I. 450, 451 Harms, I., see Baxter, M.S. 456 Harms, I., see Povinec, P.P. 449–452, 456 Harms, I.H. 450, 451 Hartley, H.O., see Pearson, E.S. 80 Hartsock, W.J., see Ketterer, M.E. 268 Harvey, B.R. 315 Hashemi-Nezhad, S.R. 221–223, 229 Hashimoto, Y., see Paul, M. 280 HASL, 150 Hasse, H.-U., see Nunnemann, M. 338 Hasse, H.-U., see Passler, G. 337, 338, 341 Hassig, N.L. 18, 31 Hatheway, A., see Donahue, D.J. 65 Hatheway, A.L., see Jull, A.J.T. 243, 245 Hatsukawa, Y. 379 Hatté, C. 245 Hatté, C., see Goslar, T. 254 Hausclid, J., see Aumann, D.C. 386, 389, 390 Hauser, S., see Povinec, P.P. 451, 467 Hauser, S., see Schiavo, M.A. 467 Hausladen, P.A., see Bird, M.I. 245 Hausladen, P.A., see Fifield, L.K. 268 Hawkes, B., see Sanborn, P. 252, 253 Hawkesworth, C.J., see Beck, J.W. 250 Hayakawa, T., see Hatsukawa, Y. 379 Haynes, C.V. 253 Haynes, D.V., see Meltzer, D.J. 253 Haynes, J.W., see Augustson, J.H. 401 Heal, K.V., see Tyler, A.N. 410 Hebeda, E.H., see Heumann, K.G. 336 Hedberg, M., see Danesi, P.R. 358 Heisinger, B. 255 Helbig, W. 210, 229 Helbig, W., see Niese, S. 234, 235 Helfer, I.K. 411, 413, 416 Helfer, I.K., see Miller, K.M. 408, 416 Helmer, R.G., see Debertin, K. 164, 169 Hemet, P., see Baglan, N. 304
495
Hemet, P., see Chiappini, R. 155 Hemet, P., see Pointurier, F. 305 Henck, R., see Hubert, Ph. 217 Henderson, G.M., see Foster, D.A. 323 Hendricks, P.H.G.M. 420, 421 Hennessy, C., see Denk, E. 346 Henrich, E., see Betti, M. 359 Henshaw, D.L., see O’Donnell, R.G. 263, 269 Hermes, E.A., see Kozlowski, T. 179 Herrera-Osterheld, S., see Beck, J.W. 250, 252 Herring, C., see Hughen, K.A. 249, 250 Herrmann, G., see Eichler, B. 333 Herrmann, G., see Erdmann, N. 339 Herrmann, G., see Passler, G. 337, 338, 341 Herrmann, G., see Peuser, P. 332 Herrmann, G., see Wendt, K. 344 Hershkowitz, A., see Berkovits, D. 280 Hershkowitz, A., see Paul, M. 280, 344 Hershkowitz, Y.H., see Paul, M. 280 Herut, B., see Weinstein, Y. 467 Herwig, G., see Jacob, P. 411 Herwig, P., see Jacob, P. 411 Herzog, G.F., see Tuniz, C. 241, 242, 332 Hess, C.T. 40 Hess, N.J., see Conradson, S.D. 366 Hess, R.F., see Conradson, S.D. 366 Hess, V.M. 210 Hesselink, W.H.A., see Kozlowski, T. 179 Heuberger, H., see Tschiersch, J. 39 Heumann, K.G. 336 Heusser, G. 163–167, 169, 171, 185, 189, 194, 202, 211, 228, 229, 233 Heusser, G., see Laubenstein, M. 108, 163, 202, 203 Heusser, G., see Neder, H. 194, 232 Heusser, G., see Simgen, H. 94 Hewson, R.D. 439 Hicks, H.G. 150 Hiernaut, T., see Tamborini, G. 358 Hiersche, L., see Kammerer, L. 42 Hies, M., see Backe, H. 347
496
Author Index
Higgitt, D.L., see Zhang, X.B. 433 Higgo, J.J.W., see Jones, D.G. 409, 459 Hilberath, Th., see Krönert, U. 337 Hill, A.E. 465 Hill, D.M. 315 Hille, R., see Vonderheide, A.P. 297, 321 Hille, R., see Zoriy, M.V. 297, 321, 324, 332 Hillebrandt, W., see Wallner, C. 279 Hillegonds, D., see Denk, E. 346 Hillman, U. 416, 417 Hincks, E.P., see Bercovitch, M. 223 Hindorf, C., see Eriksson, M. 363 Hinds, W.C. 38 Hirano, S., see Yamada, M. 155 Hirose, K. 138, 139, 141, 143, 144, 146, 150–155, 235 Hirose, K., see Aoyama, M. 138, 140, 142, 143, 145, 146 Hirose, K., see Ikeuchi, Y. 155, 157 Hirose, K., see Ito, T. 138, 155 Hirose, K., see Kim, C.K. 151 Hirose, K., see Miyake, Y. 138 Hirose, K., see Miyao, T. 138 Hirose, K., see Povinec, P.P. 138, 142, 148, 149, 154, 155 Hirose, K., see Tsumune, D. 138, 139, 158 Hisamatsu, S. 150 Hlavay, J. 18 Hlinka, V. 198 Hoaglin, D.C. 117 Hoaglin, D.C., see Velleman, P.F. 98, 99 Hodge, V.F. 151, 152 Hodgins, G., see McGeehin, J. 245 Hodgins, G.W., see Steadman, D.W. 253, 254 Hodgins, G.W.L., see Jones, K.B. 245 Hodgins, G.W.L., see Jull, A.J.T. 243 Hodgson, A., see Etherington, G. 287 Hodgson, A., see Stradling, N. 287 Hodgson, S., see Stradling, N. 287 Hoff, J., see Cheng, H. 302 Hoffman, F.L., see Hodge, V.F. 151, 152 Hoffmann, D.L., see Beck, J.W. 250
Hofmann, D.L., see Richards, D.A. 250 Hogg, A.G., see Reimer, P.J. 247, 249, 251 Holdren, G.R., see Fehn, U. 384 Holm, E. 311 Holm, E., see Johansson, L. 301 Holm, E., see Yiou, F. 266 Holý, K., see Beláˇn, T. 171 Honda, T. 399 Honda, T., see Ogiwara, K. 399 Honda, T., see Povinec, P.P. 138, 142, 154 Hong, G.H., see Hirose, K. 138, 154, 155 Hong, G.H., see Ikeuchi, Y. 155, 157 Hong, G.H., see Kim, C.K. 151 Hong, G.H., see Povinec, P.P. 142, 148, 154 Hong, K.H., see Lee, C.W. 147 Hoog, I., see Maul, J. 334, 348 Horn, J. 32 Horrill, D.A., see Parkinson, J.A. 410 Horton, T.R., see Hess, C.T. 40 Horwitz, E.P. 147, 153 Horwitz, W. 113 Horwitz, W., see Alberts, R. 19 Horwitz, W., see Kurtz, D.A. 64 Hoshi, M., see Hult, M. 235 Hotchkis, M. 278, 279, 287 Hotchkis, M., see Danesi, P.R. 288 Hotchkis, M.A.C. 270, 278, 287 Hotchkis, M.A.C., see Child, D.P. 281, 282 Hou, X. 318 Hou, X.L. 318, 377, 384–390, 397, 398 Houdek, F. 394 Houk, R.S., see Jiang, S.-J. 296 Houk, R.S., see Rowan, J.T. 297, 298 Houtermans, H., see Parr, R. 75 Hovgaard, J. 420 Hovgaard, J., see Aage, H.K. 420 Hovgaard, J., see Grasty, R.L. 427, 430 Hrnecek, E. 279 Hrnecek, E., see Perna, L. 361 Hubaux, A. 70 Hubbell, J.H. 411
Author Index
Hübener, S., see Eichler, B. 333 Huber, G. 332 Huber, G., see Erdmann, N. 339 Huber, G., see Grüning, C. 337–339 Huber, G., see Maul, J. 334, 348 Huber, G., see Nunnemann, M. 338 Huber, G., see Passler, G. 337, 338, 341 Huber, G., see Sauvage, J. 337 Huber, G., see Waldek, A. 339 Hubert, Ph. 217 Hubert, Ph., see Torres, R. 231 Huckabay, G.W., see Anspaugh, L.R. 413 Hudson, E.A., see Conradson, S.D. 366 Hughen, K.A. 249, 250 Hughen, K.A., see Reimer, P.J. 247, 249, 251 Hughen, K.A., see van der Plicht, J. 247, 250 Hughey, B.J. 243 Huh, C.A. 399 Huijbregts, C.J., see Journel, A.G. 18 Hult, M. 202, 235 Hult, M., see Laubenstein, M. 108, 163, 202, 203 Hunt, G., see Kershaw, P. 459, 462 Hunt, G.J. 459 Hunter, G., see Betti, M. 359 Hurrell, R.F., see Denk, E. 346 Hurst, G.S. 332 Hursthouse, A.S. 390, 392, 393 Hurtgen, C. 54 Hussein, M.A., see Ramlia, A.T. 398 Hutchinson, J.M.R., see Lee, S.C. 313 Huth, F.W., see Handl, J. 386, 390 Huynh-Ngoc, L., see Osvath, I. 450, 451, 458 Huynh-Ngoc, L., see Povinec, P.P. 450, 451, 456 Huyse, M., see Kudryavtsev, Y. 337 Hyman, M., see Russ, J. 245 Hynh-Ngoc, L., see Povinec, P.P. 138, 148, 149, 155
497
IAEA, 18, 19, 124, 125, 131, 142, 149, 382, 410, 420, 450, 474 Ibbett, R.D., see Harvey, B.R. 315 Ibrahim, F., see Sauvage, J. 337 ICRM, 108 ICRP, 18, 287, 288 ICRU, 31, 411, 413, 416, 425–427, 436, 437 Igarashi, S., see Kim, C.K. 313 Igarashi, Y., see Aoyama, M. 138, 140, 146 Igarashi, Y., see Hirose, K. 138, 139, 141, 143, 151, 155, 235 Igarashi, Y., see Kim, C.K. 151, 313 Igarashi, Y., see Miyao, T. 138 Ikäheimonen, T.K., see Moring, M. 362 Ikäheimonen, T.K., see Pöllänen, R. 43, 348 Ikeda, N. 394, 396 Ikeda, N., see Kim, C.K. 313, 390, 394 Ikeuchi, Y. 155, 157 Ikeuchi, Y., see Povinec, P.P. 138, 148, 149, 155, 458, 459 Ilus, E., see Moring, M. 362 Imai, T. 151 Imbaud, H., see Yiou, F. 266 Inn, K.G.W., see Lee, S.C. 313 Inoue, M., see Komura, K. 234 International Commission on Radiation Units and Measurements, ICRU, 3, 4, 6, 9, 10, 14 Isaak, H.P., see Kozlowski, T. 179 Isaaks, E.H. 18 Isaksson, M. 376 Ishimori, T., see Kimura, T. 147 Ishitobi, T., see Taniguchi, M. 474 Ishitoibi, T., see Bokuniewicz, H. 474 ISO, 18, 49, 50, 53, 54, 61, 88, 114, 116, 117, 121 ISO-IUPAC, 53 Isogai, K., see Hirose, K. 138, 154, 155 Israel, H.I., see Storm, E. 411 Ito, T. 138, 155 Ito, T., see Livingston, H.D. 138, 139, 144, 155, 157
498
Author Index
Ito, T., see Povinec, P.P. 138, 142, 148, 149, 154, 155 Itoh, H., see Katagiri, M. 450 IUPAC, 49, 50, 52–54, 56, 66, 68, 83, 86, 87, 116, 117, 121 Ivkova, T.M., see Liechtenstein, V.K. 275 Ivy-Ochs, S., see Heisinger, B. 255 Iwashima, K., see Kimura, T. 147 Izmer, A.V. 320 Izmer, A.V., see Vonderheide, A.P. 297, 321 Izrael, Yu.A. 356 Jackson, J.E. 82 Jacob, P. 411, 426, 430 Jacob, P., see Hillman, U. 416, 417 Jacob, P., see Rybacek, K. 416, 427, 428 Jacob, P., see Thummerer, S. 417, 428, 430 Jacob, S.W.A., see Suter, M. 243 Jacobs, J.J., see Van R. Smit, J. 139 Jacobsen, G.E., see Hotchkis, M.A.C. 270, 278, 287 Jacobsen, U., see Hou, X.L. 318, 385– 390 Jahnke, U., see Schröder, W.U. 179, 180 Jakob, D., see Handl, J. 386, 390 Jakob, D., see Schmidt, A. 318 Jakubowski, N., see Feldmann, I. 298, 299 James, H.I., see Sugihara, T.T. 146 Janssens, A., see Betti, M. 359 Janssens, K., see Lind, O.C. 367 Janssens, K., see Salbu, B. 357, 358, 367 Jarrell, C.L., see Shebell, P. 430 Jeffrey, D.A., see Jull, A.J.T. 242 Jensen, A., see Chen, Q.J. 390–392 Jeran, Z., see Rosamilia, S. 398 JERN, 357 Jernström, J. 357, 361, 367 Jernström, J., see Eriksson, M. 357, 362 Jernström, J., see Perna, L. 361 Jerome, S., see Hurtgen, C. 54 Jha, S.K. 390, 392, 393
Ji, Y.-Q. 311 Jiang, S., see Paul, M. 280 Jiang, S.-J. 296 Joannon, S. 323 Johanson, K.J., see Handl, J. 386, 390 Johansson, L. 301 Johnson, T.C., see Schwartz, M. 256 Johnston, P.N., see Hult, M. 202 Jokela, T.A., see McAninch, J.E. 277 Jokinen, A., see Moore, I.D. 337 Jones, D.G. 409, 449, 450, 459, 462 Jones, K.B. 245 Jordan, C.F., see Stewart, M.L. 42 Joseph, A.D., see Lindsay, R. 430 Journel, A.G. 18 Jouzel, J., see Raisbeck, G.M. 250, 255 Juhanoja, J., see Moring, M. 362 Juhanoja, J., see Pöllänen, R. 43, 348 Jull, A.J.T. 241–243, 245, 252–254, 257 Jull, A.J.T., see Beck, J.W. 250, 252 Jull, A.J.T., see Burney, D.A. 253 Jull, A.J.T., see Currie, L.A. 83 Jull, A.J.T., see Donahue, D.J. 242, 245 Jull, A.J.T., see Heisinger, B. 255 Jull, A.J.T., see Lifton, N.A. 254, 255 Jull, A.J.T., see McGeehin, J. 245 Jull, A.J.T., see McNichol, A.P. 242, 245 Jull, A.J.T., see Meyer, G.A. 252 Jull, A.J.T., see Pierce, J.L. 252, 253 Jull, A.J.T., see Povinec, P.P. 318 Jull, A.J.T., see Sanborn, P. 252, 253 Jull, A.J.T., see Steadman, D.W. 253, 254 Jull, A.T.J., see Donahue, D.J. 65 Jung, P., see Alley, R.B. 251, 253 Jungers, W.L., see Burney, D.A. 253 Kachanoski, R.G. 433 Kafri, U., see Weinstein, Y. 467 Kaihola, L., see Plastino, W. 233, 234 Kaiser, H. 51, 75 Kakihana, H., see Honda, T. 399 Kalin, R.M. 228, 234 Kalinina, G.V., see Kashkarov, L.L. 357 Kalmykov, St.N. 392, 393 Kamemoto, M., see Ikeda, N. 394, 396
Author Index
Kammerer, L. 42 Kanapilly, G.M., see Cheng, Y.-S. 38 Kanazawa, T., see Miyake, Y. 138 Kanazawa, T., see Saruhashi, K. 138, 146 Kanisch, G., see Povinec, P.P. 458, 459 Kaplan, A., see Fairbanks, R.G. 249–251 Karam, L.R. 397 Karlberg, O. 427, 428, 430 Karpuk, S., see Maul, J. 334, 348 Karstensen, K., see Allkofer, O.C. 173 Kashgarian, M., see Hughen, K.A. 249 Kashiv, Y., see Paul, M. 344 Kashkarov, L.L. 357 Kashparov, V.A., see Salbu, B. 19, 357, 367 Kastlander, J. 430 Katagiri, M. 450 Katagiri, M., see Kobayashi, Y. 450, 468 Katayama, Y., see Yamamoto, M. 150 Katsuragi, K., see Hirose, K. 143, 155 Katsuragi, Y., see Saruhashi, K. 138, 146 Kaufman, L., see Massart, D.L. 75 Kawabata, K., see Sakata, K. 296 Kawamura, H., see Kim, C.K. 313 Kayasth, V., see Pulhani, V. 398 Kaye, J.H. 210, 230 Kearfoot, K.J., see Benke, R.R. 427 Keating, J.A., see Cheng, Y.-S. 38 Keefer, D., see Shebell, P. 430 Keigwin, L.D., see Alley, R.A. 251 Keith-Roach, M.J. 265, 287, 317, 394 Kelley, J.M. 311 Kelley, J.M., see Beasley, T.M. 311 Kelley, J.M., see Cooper, L.W. 301, 311 Kelley, J.M., see Dai, M.H. 332 Kelly, W.R. 86, 95 Kenna, T.C. 311 Kennedy, B.M., see Sturchio, N.C. 91 Keogh, D.W., see Conradson, S.D. 366 Kerlau, G., see Bouisset, P. 384 Kerls, W. 319 Kerr, G.D., see Dickson, H.W. 411, 413 Kershaw, P. 459, 462–464 Kershaw, P., see McCartney, M. 459, 462–464
499
Kershaw, P.J., see Hunt, G.J. 459 Kessler, J.D., see Currie, L.A. 83, 89 Kessler, T., see Moore, I.D. 337 Ketterer, M.E. 268, 302, 332 Ketterer, M.E., see Zoriy, M.V. 305, 324 Keusch, M., see Tschiersch, J. 39 Khan, A.J., see Semkow, T.M. 164, 166, 194 Khitrov, L., see Tcherkezian, V. 19 Khitrov, L.M. 43 Khokhryakov, V.F., see Leggett, R.W. 287, 288 Kiefer, P., see Gering, F. 408, 430 Kieser, W.E. 318 Kieser, W.E., see Hughey, B.J. 243 Kieser, W.E., see Povinec, P.P. 318 Kilius, L., see Rucklidge, J. 267 Kilius, L.R. 267 Kilius, L.R., see Purser, K.H. 289 Kilius, L.R., see Raisbeck, G.M. 266 Kilius, L.R., see Yiou, F. 266 Kim, C.-K., see Kim, C.-S. 305 Kim, C.-S. 305 Kim, C.K. 151, 313, 315, 390, 394 Kim, C.K., see Hirose, K. 138, 141, 154, 155 Kim, C.K., see Ikeuchi, Y. 155, 157 Kim, C.K., see Kim, C.S. 151, 153, 157, 312 Kim, C.S. 151, 153, 157, 312 Kim, C.S., see Hirose, K. 141 Kim, C.S., see Kim, C.K. 151, 315 Kim, G., see Burnett, W.C. 467 Kim, L. 181 Kim, S.H., see Hirose, K. 138, 154, 155 Kim, S.H., see Ikeuchi, Y. 155, 157 Kimura, T. 147 Kindl, P., see Schroettner, T. 164, 166, 194 King, S.J., see Fifield, L.K. 265, 269, 271, 280, 282, 332 Kingston, H.M., see Greenberg, R.R. 399 Kirchhoff, K. 19 Kirkegaard, P., see Løvborg, L. 410
500
Author Index
Kiselev, O., see Kraft, S. 272 Kiseleva, A., see Kraft, S. 272 Kitagawa, H. 250 Kladiva, E., see Povinec, P. 233 Kleeberg, H., see Niese, S. 234 Klein, J. 247 Klein, R.G., see Martin, P.S. 253, 254 Kleine, J.R., see Stewart, M.L. 42 Klemic, G.A., see Miller, K.M. 427, 428 Klemola, S., see Moring, M. 362 Klemola, S., see Pöllänen, R. 43, 348 Klinedinst, D.B., see Currie, L.A. 83 Klingele, E.E., see Schwarz, G.F. 408 Klinkowstein, R.E., see Hughey, B. 243 Klinkowstein, R.E., see Hughey, B.J. 243 Klopp, P., see Grüning, C. 337–339 Klouda, G.A., see Currie, L.A. 83, 93, 94, 98, 108, 109, 120, 121 Kluge, H.-J., see Krönert, U. 337 Kluge, H.-J., see Wies, K. 342 Knezovich, J.P., see Brown, T.A. 265, 269, 277 Knezovich, J.P., see McAninch, J.E. 277 Knie, K., see Heisinger, B. 255 Knie, K., see Wallner, C. 279 Knoll, G.F. 421 Kobayashi, Y. 450, 468 Kobayashi, Y., see Katagiri, M. 450 Koch, L., see Betti, M. 348, 357, 358 Koch, L., see Tamborini, G. 358 Koch, L., see Wallenius, M. 332 Koehler, M., see Hult, M. 202 Kofuji, H., see Yamamoto, M. 311 Kogan, R.M. 408 Köhler, M., see Laubenstein, M. 108, 163, 202, 203 Köhler, M., see Niese, S. 194, 202 Köhler, S., see Eichler, B. 333 Köhler, S., see Erdmann, N. 339 Köhler, S., see Passler, G. 337, 338, 341 Koide, M. 150 Koivunoro, H., see Reijonen, J. 377 Kolb, W. 225 Kolb, W.A. 169 Kolesov, G., see Tcherkezian, V. 19
Kollmer, F., see Erdmann, N. 334, 348 Kolstad, A.K., see Chen, Q.J. 152, 153, 399, 401 Kolwaite, D., see Ketterer, M.E. 302, 332 Komamura, M., see Yamamoto, M. 311 Komarikov, I., see Jacob, P. 411, 426, 430 Komura, K. 140, 150, 202, 230, 234, 235 Komura, K., see Aoyama, M. 145 Komura, K., see Hirose, K. 138, 141, 151, 235 Komura, K., see Yamamoto, M. 311 Konishi, M., see Sekine, T. 315 Kontar, E., see Burnett, W.C. 467 Kontar, E.A. 467 Kontar, E.A., see Bokuniewicz, H. 474 Kontar, E.A., see Povinec, P.P. 451, 467 Kontar, E.A., see Taniguchi, M. 469 Koppenaal, D.W., see Eiden, G.C. 297, 320, 321 Koppenaal, D.W., see Farmer, O.T. 319 Koprda, V. 18 Korsbech, U., see Aage, H.K. 420 Korschinek, G., see Wallner, C. 279 Korun, M. 427 Kosanke, K., see Alexander, P. 410 Köster, U. 342 Kovach, J.L. 39 Kovalchuk, E.L. 230 Kovgan, L., see Jacob, P. 411, 426, 430 Kozlowski, T. 179 Kra, R., see Stuiver, M. 247 Kracke, W., see Bunzl, K. 408, 410 Kraft, S. 272 Krahenbuhl, M.P., see Leggett, R.W. 287, 288 Krähenbühl, U., see Ayranov, M. 311 Kratochvil, B. 35 Kratz, J.V., see Erdmann, N. 339, 349, 358 Kratz, J.V., see Grüning, C. 337–339 Kratz, J.V., see Huber, G. 332 Kratz, J.V., see Kuczewski, B. 311 Kratz, J.V., see Nunnemann, M. 338 Kratz, J.V., see Passler, G. 337, 338, 341
Author Index
Kratz, J.V., see Waldek, A. 339 Kratz, J.V., see Wendt, K. 344 Kreissig, U., see Liechtenstein, V.K. 275 Krekling, T., see Salbu, B. 19, 348, 357, 367 Kretchmark, T., see Beck, J.W. 250 Krey, P.W. 150 Krieg, M., see Sauvage, J. 337 Krogner, K., see Niese, S. 235 Kromer, B., see Hughen, K.A. 249, 250 Kromer, B., see Reimer, P.J. 247, 249, 251 Kromer, B., see van der Plicht, J. 247, 250 Krönert, U. 337 Krüger, A., see Povinec, P.P. 458, 459 Kruglov, K., see Kudryavtsev, Y. 337 Krupa, S., see Burnett, W.C. 467 Ku, T.L., see Lee, T. 397 Kube, G., see Sewtz, M. 347 Kubik, P.W., see Fehn, U. 384 Kubik, P.W., see Grajcar, M. 243 Kubik, P.W., see Heisinger, B. 255 Kucera, J., see Hou, X.L. 385, 386, 388, 390, 397 Kuczewski, B. 311 Kudo, H., see Sekine, T. 315 Kudrjashov, V., see Mironov, V. 267 Kudrjasov, V.P., see Boulyga, S.F. 305, 324 Kudryavtsev, Y.A. 337, 343 Kudryavtsev, Y.A., see Schulz, C. 344 Kuiper, J.L., see Miller, K.M. 408, 416 Kulkarni, K.M., see Burnett, W.C. 467 Kulkarni, K.M., see Povinec, P.P. 451, 467 Kulkarni, S., see Smith, F. 19 Kunz, H., see Backe, H. 347 Kunz, P., see Grüning, C. 337–339 Kunz, P., see Nunnemann, M. 338 Kunz, P., see Sewtz, M. 347 Kunz, P., see Waldek, A. 339 Kurosawa, T., see Yonezawa, C. 399 Kurtz, D. 64, 68, 119
501
Kurunczi, S., see Török, S. 357, 358, 360, 367 Kushita, K., see Hatsukawa, Y. 379 Kutschera, W. 342 Kutschera, W., see Collon, P. 91 Kutschera, W., see Kraft, S. 272 Kutschera, W., see Liechtenstein, V.K. 275 Kutschera, W., see Steier, P. 275, 278, 279 Kutschera, W., see Vockenhuber, C. 279 Kutschera, W., see Winkler, S. 279 Kuzminov, V.V., see Kovalchuk, E.L. 230 Kwong, L.L.W., see Povinec, P.P. 138, 148, 149, 155 Kyrš, M., see Tölgyessy, J. 20, 36 La Rosa, J. 311 La Rosa, J., see Fowler, S.W. 138 La Rosa, J., see Wyse, E.J. 306, 371 La Rosa, J.J. 139, 152, 153 La Rosa, J.J., see Povinec, P.P. 157, 235, 450, 467 LaBelle, R.D. 344 Laedermann, J.P. 408 LaFleur, P.D., see Becker, D.A. 399 Lagan, L., see La Rosa, J. 311 Laird, N.M., see Maher, E.F. 38 Lakosi, L., see Sekine, T. 396 Lal, D., see Heisinger, B. 255 Lally, A.E., see Parry, S.J. 386 LaMarre, J.R., see Grasty, R.L. 427, 430 Lambert, D. 116 Lambert, M., see Burnett, W.C. 467 LaMont, S.P., see Glover, S.E. 399 Lander, G.H., see Conradson, S.D. 366 Lane, S.L. 39 Lane-Smith, D., see Burnett, W.C. 467 Lang, J.J., see Sanderson, D.C.W. 424, 439, 441, 442 Lange, T.E., see Jones, K.B. 245 Lange, T.E., see Jull, A.J.T. 243, 245 Langland, J.K., see Lindström, R.M. 223, 228, 235
502
Author Index
Lantzsch, J., see Wendt, K. 344 Larivière, D. 322, 323, 332 Larivière, D., see Benkhedda, K. 323 Larivière, D., see Epov, V.N. 322, 323 LaRosa, J., see Povinec, P.P. 138, 148, 149, 155 LaRosa, J.J., see Horwitz, E.P. 147, 153 Larrieu, P., see Hubert, Ph. 217 Larsen, I.I., see Cutshall, N.H. 412, 424 Lassen, J., see Sewtz, M. 347 Lassen, J., see Waldek, A. 339 Laubenstein, M. 108, 163, 202, 203, 232, 233 Laubenstein, M., see Heusser, G. 202, 233 Laubenstein, M., see Neder, H. 194, 232 Lauth, W., see Backe, H. 347 Lauth, W., see Sewtz, M. 347 Lazarev, V., see Heisinger, B. 255 Le Blanc, F., see Sauvage, J. 337 Le Scornet, G., see Sauvage, J. 337 Leccia, F., see Hubert, Ph. 217 Lee, C.W. 147 Lee, C.W., see Lee, M.W. 150 Lee, D.-C., see Luo, X. 324 Lee, J.-I., see Kim, C.-S. 305, 315 Lee, J.K.P., see Sauvage, J. 337 Lee, K.J., see Kim, C.S. 151, 153, 157, 305, 312 Lee, M.H., see Lee, C.W. 147 Lee, M.W. 150 Lee, P.J., see Hotchkis, M.A.C. 270, 278, 287 Lee, S.C. 313 Lee, S.H., see La Rosa, J.J. 139, 152, 153, 311 Lee, S.H., see Povinec, P.P. 138, 148, 149, 155, 157, 235, 450, 467 Lee, S.H., see Wyse, E.J. 264, 269, 306, 332, 371 Lee, T. 397 Lefevre, O., see Bouisset, P. 384 Legeleux, F., see Reyss, J.-L. 194, 235 Leggett, R.W. 287, 288 Lehman, S.J., see Hughen, K.A. 249
Lehmann, B.E., see Du, X. 92 Lehmann, B.E., see Sturchio, N.C. 91 Leonard, K.S. 267 Leontiev, G.G., see Moskvin, L.N. 39 Lepetit, G., see Chiappini, R. 155 Lepofksy, D.S., see Hallett, D.J. 252 Lepofsky, D., see Lertzman, K. 252 Lerch, M., see Peuser, P. 332 Lerman, J.C., see Klein, J. 247 Lertzman, K. 252 Lertzman, K.P., see Hallett, D.J. 252 Leslie, A.G., see Hadley, M.J. 430 Less, G., see Weinstein, Y. 467 Lestringuez, J., see Raisbeck, G.M. 255 Letokhov, V.S. 332, 335 Letokhov, V.S., see Alkhazov, G.D. 342 Letokhov, V.S., see Ambartzumian, R.V. 333 Letokhov, V.S., see Kudryavtsev, Y.A. 343 Letokhov, V.S., see Schulz, C. 344 Lettry, J., see Sauvage, J. 337 Leung, K.N., see Reijonen, J. 377 Levaiova, D., see Beláˇn, T. 171 Levin, D.J., see Smith, P.F. 169 Levy, I., see La Rosa, J.J. 139, 152, 153 Levy-Palomo, I. 467 Levy-Palomo, I., see La Rosa, J. 311 Levy-Palomo, I., see Povinec, P.P. 108, 164, 166, 169, 194, 196, 202, 228, 234, 451, 453, 467, 468, 472, 474 Lewis, P.A.W., see Cox, D.R. 62, 104, 108 Li, C., see Tanner, S.D. 310 Li, C., see Vais, V. 305, 310 Li, J.-Y., see Ji, Y.-Q. 311 Li, W.-X. 41 Li, Y.M., see Bailey, K. 91, 92 Li, Y.M., see Chen, C.Y. 91, 342 Liao, C.-L. 297 Libby, W.F. 241, 247 Libby, W.F., see Arnold, J.R. 241 Libert, J., see Sauvage, J. 337 Liechtenstein, V., see Vockenhuber, C. 279
Author Index
Liechtenstein, V.K. 275 Liechtenstein, V., see Steier, P. 275, 278, 279 Lierse, C., see Wallner, C. 279 Lievens, P. 334 Lievens, P., see Bastiaansen, J. 334 Lievens, P., see Erdmann, N. 334, 348, 349 Lievens, P., see Schulz, C. 344 Lievens, P., see Vandeweert, E. 334 Lifton, N.A. 254, 255 Lifton, N.A., see Jull, A.J.T. 243, 245 Lifton, N.A., see Pigati, J.S. 255 Likar, A., see Korun, M. 427 Likhtarev, I., see Jacob, P. 411, 426, 430 Limburg, R.J., see Hendricks, P.H.G.M. 420, 421 Lind, B., see Chen, Q.J. 152, 153, 399, 401 Lind, O.C. 285, 367 Lind, O.C., see Salbu, B. 348, 357, 358, 367 Lind, O.C., see Skipperud, L. 285, 286 Lindahl, I. 410 Lindenberger, K.H., see Schröder, W.U. 179, 180 Lindsay, R. 430 Lindström, D.J., see Lindström, R.M. 223, 228, 235 Lindström, R.M. 223, 228, 235 Lindstrom, R.M., see Zeissler, C.J. 43 Linick, T.W., see Donahue, D.J. 65, 242, 245 Link, C.L., see Ketterer, M.E. 268, 302, 332 Liong, W.K.L., see Povinec, P.P. 450, 451, 456, 458, 459 Lipoglavsek, M., see Korun, M. 427 Liteanu, C. 72 Litherland, A.E. 283 Litherland, A.E., see Hughey, B.J. 243 Litherland, A.E., see Kilius, L.R. 267 Litherland, A.E., see Purser, K.H. 289 Littlejohn, D., see Ure, M. 18 Liu, J.-L., see Ji, Y.-Q. 311
503
Livens, F.R. 410 Livens, F.R., see Fifield, L.K. 265, 269, 280 Livens, F.R., see Hursthouse, A.S. 390, 392, 393 Livens, F.R., see Keith-Roach, M.J. 265, 287 Livens, F.R., see Marsden, O.J. 270, 287 Livingston, H.D. 137–139, 142, 144, 148, 154, 155, 157 Livingston, H.D., see Bowen, V.T. 137– 139, 146, 155 Livingston, H.D., see Osvath, I. 450, 451, 453, 464 Livingston, H.D., see Povinec, P.P. 138, 142, 148, 149, 154, 155 Ljunggren, K., see Eriksson, M. 363 Lloyd, C.D., see Ruffel, A. 430 Lohrmann, E., see Schopper, E. 211 Long, A., see Kalin, R.M. 228, 234 Long, A., see Stuiver, M. 247 Long, G.L. 52, 53 Long, S.C., see O’Donnell, R.G. 263, 269 Loos, G., see Niese, S. 234 Loosli, H.H. 91, 232 Loosli, H.H., see Oeschger, H. 210, 229, 234 Lopez, J.M., see Schmidt, A. 318 López-Gutiérrez, J.M., see Santos, F.J. 318 Lorenzo, R., see Du, X. 92 Lorenzo, R., see Sturchio, N.C. 91 Los, I., see Jacob, P. 411, 426, 430 Los Alamos National Laboratory, LANL, 182 Lotti, R., see Bond, G. 251 Lou, T.P., see Reijonen, J. 377 Louis, A.T., see Carlin, R.P. 25 Loukola-Ruskeeniemi, K., see Airo, M.L. 439 Løvborg, L. 410, 420 Loveless, A., see Burnett, W.C. 467 Lovett, M.B., see Harvey, B.R. 315 Lovett, M.B., see Nelson, D.W. 150
504
Author Index
Lu, H.L., see Lee, T. 397 Lu, Z.-T. 342 Lu, Z.-T., see Bailey, K. 91, 92 Lu, Z.-T., see Chen, C.Y. 91, 342 Lu, Z.-T., see Collon, P. 91 Lu, Z.-T., see Du, X. 91, 92 Lu, Z.-T., see Sturchio, N.C. 91 Lund, S.P., see Schwartz, M. 256 Lundberg, J., see Li, W.-X. 41 Lunney, D., see Sauvage, J. 337 Luo, S.-G., see Ji, Y.-Q. 311 Luo, X. 324 Lupinetti, A.J., see Conradson, S.D. 366 Luttrell, S.P., see Appleby, L.J. 355 Lynn, N.M., see Sjöblom, K.-L. 450, 451, 456 Macášek, F. 18, 20–22, 32, 35–37 Macášek, F., see Bartoš, P. 18 Macášek, F., see Navratil, J.D. 18 MacCartney, M. 314 Macklach, R., see Rybacek, K. 416, 427, 428 Maden, C., see Grajcar, M. 243 Madruga, M.J., see Wauters, J. 18 Maes, A., see Cremers, A. 18 Maesday, D.F., see Suzuki, T. 179 Maher, E.F. 38 Mai, H., see Ehrlich, G. 51 Maiti, T.C., see Beasley, T.M. 311 Makarewicz, M., see Danesi, P.R. 288 Malcolm, A., see Thomas, B.W. 450 Malczewski, D. 430 Malencheko, A.F., see Hou, X.L. 385, 386, 388, 390 Mamuro, T. 31 Mandel, J. 114 Mankowski, M.M., see LaBelle, R.D. 344 Mann, D.K., see Vance, D.E. 311 Mann, D.R., see Livingston, H.D. 138 Manning, S., see Reimer, P.J. 247, 249, 251 Mansel, A., see Nunnemann, M. 338 Mansel, A., see Passler, G. 337, 338, 341
Marchetti, A.A., see Bergquist, B.A. 268 Marchetti, A.A., see Brown, T.A. 265, 269, 277 Margaritz, M. 40 Markert, B. 376 Markewich, H., see McGeehin, J. 245 Markgraf, V. 251 Markowicz, A., see Danesi, P.R. 358 Markowicz, S., see Eriksson, M. 357, 362 Marks, A., see Pickup, G. 439 MARLAP, 66, 108 Marolf, J.V., see Currie, L.A. 89 Marquardt, C.M., see Kuczewski, B. 311 Marsden, O.J. 270, 287 Marsh, B., see Moore, I.D. 337 Marsher, P., see Sowa, W. 411, 413 Martin, J.B., see Cable, J.E. 473 Martin, L., see Turcq, B. 252 Martin, P.J. 154 Martin, P.S. 253, 254 Martin, P.S., see Steadman, D.W. 253, 254 Martin, R., see Backe, H. 347 Martincic, R., see Korun, M. 427 Martineau, O. 231 Martinelli, R.E., see Bergquist, B.A. 268 Martinelli, R.E., see Brown, T.A. 265, 269, 277 Martinez Canet, M.J., see Hult, M. 202 Martini, E., see Sowa, W. 411, 413 Maruyama, K., see Tsumune, D. 138 Mas, J.L. 314–317 Masarik, J., see Baumgartner, S. 250 Masarik, J., see Dep, L. 221 Maskell, S.C., see Williams, D. 410 Massart, D.L. 75, 398 Masse, L. 458 Mathewes, R., see Lertzman, K. 252 Mathewes, R.W., see Hallett, D.J. 252 Matsumoto, R., see Fehn, U. 267 Matsunami, T., see Mamuro, T. 31 Mattey, D.P., see Beck, J.W. 250 Mattey, D.P., see Richards, D.A. 250 Matthews, M., see Povinec, P.P. 458, 459
Author Index
Mattsson, S. 314 Matusevich, J.L., see Boulyga, S.F. 305, 324 Mauck, G., see Schopper, E. 211 Maul, J. 334, 348 Maul, J., see Berg, T. 349 May, S. 390, 392, 393 May, S., see Germain, P. 390, 392, 393 Mayer, K., see Wallenius, M. 332 Mazzilli, B.P., see De Oliveira, J. 467, 473 Mazzilli, B.P., see Vasconcellos, M.B.A. 401 McAninch, J.E. 277 McAninch, J.E., see Bergquist, B.A. 268 McCartney, M. 267, 459, 462–464 McCormac, F.G., see Hughen, K.A. 249, 250 McCormac, F.G., see van der Plicht, J. 247, 250 McCormac, G., see Reimer, P.J. 247, 249, 251 McCormack, M. 64 McCubbin, D., see Leonard, K.S. 267 McDonald, H.G., see Steadman, D.W. 253, 254 McDonald, P., see Leonard, K.S. 267 McDonald, P., see MacCartney, M. 314 McDonald, P., see McCartney, M. 267 McDonnell, R.A., see Burrough, P.A. 11 McGeehin, J. 245 McHargue, L.R. 256 McHargue, L.R., see Jull, A.J.T. 245 McKay, K., see Hursthouse, A.S. 390, 392, 393 McKinley, J.M., see Ruffel, A. 430 McLean, J.A., see Boulyga, S.F. 305, 324 McLeod, J.J., see Sanderson, D.C.W. 409, 439, 441, 442 McMahon, C.A. 362 McMahon, C.A., see Karam, L.R. 397 McMahon, C.A., see Pibida, L. 397 McMillan, E.M. 310 McNichol, A.P. 242, 245 McNichol, A.P., see Currie, L.A. 89
505
McNutt, R.H., see Li, W.-X. 41 McPhee, R.D.E., see Steadman, D.W. 253, 254 Meckbach, see Jacob, P. 411, 426, 430 Meier, J.P., see Kraft, S. 272 Mejer, R.J., see Veen, A.W.L. 409 Meli, M.A., see Boulyga, S.F. 332 Meli, M.A., see Desideri, D. 269, 324 Mellander, H. 410 Mel’nikov, V.A., see Moskvin, L.N. 39 Meloun, M. 33, 34 Meltzer, D.J. 253 Mena, F., see Meltzer, D.J. 253 Mende, O., see Kirchhoff, K. 19 Mennrath, P., see Hubert, Ph. 217 Ménot-Combes, G., see Bard, E. 250 Men’shova, N.P., see Nalimov, V.V. 51 Méray, L. 108 Mesquita, A.R. 471 Messner, M.J., see Smith, F. 19 Meyer, G.A. 252 Meyer, G.A., see Pierce, J.L. 252, 253 Meynadier, L., see Baumgartner, S. 250 Meynendonckx, P., see Wordel, R. 230 Michel, J., see Hess, C.T. 40 Michel, R. 53, 54 Michel, R., see Kirchhoff, K. 19 Michel, R., see Schmidt, A. 318 Michelot, J.-L., see Deschamps, P. 304 Michotte, Y., see Massart, D.L. 75 Midkiff Jr., C.R., see Kurtz, D.A. 64 Mietelski, J.W., see Ketterer, M.E. 302, 332 Mika, J. 19, 29 Miley, H.S., see Arthur, R.J. 222 Miley, M.S., see Brodzinski, R.L. 168, 218, 231 Militký, J., see Meloun, M. 33, 34 Miller, C.M., see Downey, S.W. 313 Miller, G.H., see Dyke, A.S. 254 Miller, J.M. 450, 459 Miller, J.M., see Jones, D.G. 409, 450, 459, 462 Miller, J.M., see Thomas, B.W. 450 Miller, K.M. 408, 416, 427, 428
506
Author Index
Miller, K.M., see Helfer, I.K. 411, 413, 416 Miller, R.G. 75 Millerc, S.C., see Leggett, R.W. 287, 288 Millies-Lacroix, J.-C., see Musa, C. 458, 459 Millies-Lacroix, J.C., see Chiappini, R. 155 Milton, J.A., see Taylor, R.N. 305 Mincher, B.J. 394 Mino, N., see Hotchkis, M.A.C. 270, 278, 287 Mintrop, A., see Bruhn, F. 245 Miró, M. 147 Mironov, V. 267 Mironov, V.P., see Boulyga, S.F. 305, 324 Miroshnikov, V.S., see Moskvin, L.N. 39 Mishin, V.I., see Alkhazov, G.D. 342 Mishin, V.I., see Schulz, C. 344 Mishra, U.C., see Pulhani, V. 398 Mitchell, P.I., see McMahon, C.A. 362 Mitchell, P.I., see O’Donnell, R.G. 263, 269 Miyake, Y. 138, 146 Miyake, Y., see Saruhashi, K. 138, 146 Miyake, Y., see Sugiura, Y. 138 Miyao, T. 138 Miyao, T., see Aoyama, M. 138, 140, 146 Miyao, T., see Hirose, K. 138, 139, 143, 154, 155 Miyao, T., see Ikeuchi, Y. 155, 157 Moens, L., see Vanhaecke, F. 295 Mokrov, Y., see Oughton, D.H. 285 Molnar, G., see Reijonen, J. 377 Mondry, G., see Müller, G. 221 Monetti, M., see Shebell, P. 430 Montaser, A., see Boulyga, S.F. 305, 324 Moon, S.H., see Lee, C.W. 147 Moore, I.D. 337 Moore, J., see Cutler, K.B. 249 Moore, W.S. 467, 468 Moore, W.S., see Burnett, W.C. 467, 473 Moore, W.S., see Povinec, P.P. 451, 467
Moore, W.S., see Taniguchi, M. 469 Moos, J.R., see Shebell, P. 430 Mora, S.J.d., see Wyse, E.J. 332 Morales, J., see Hubert, Ph. 217 Morales, L.A., see Conradson, S.D. 366 Morales, R., see Hubert, Ph. 217 Moran, J.E. 267 Moran, J.E., see Snyder, G.T. 267 More, A.K., see Pulhani, V. 398 Morehead, N., see McGeehin, J. 245 Moreno, J., see Danesi, P.R. 288 Moreno, J.M.B. 397 Morgenstern, A., see Choppin, G.R. 150 Morgenstern, U., see Povinec, P.P. 458, 459, 467, 472, 474 Morimoto, T., see Hirose, K. 138, 154, 155 Morimoto, T., see Ikeuchi, Y. 155, 157 Moring, M. 362 Morita, M. 178, 179 Moriya, H., see Povinec, P.P. 138, 148, 149, 155 Mork, K.A., see Gascard, J.C. 266 Morris, K., see Fifield, L.K. 265, 269, 280 Morris, K., see Marsden, O.J. 287 Mortlock, R.A., see Fairbanks, R.G. 249–251 Morvan, J., see Hatté, C. 245 Moskvin, L.N. 39 Mosteller, F., see Hoaglin, D.C. 117 Motlhabane, T.G.K., see Lindsay, R. 430 Mott, N.S., see Noshkin, V.E. 147 Mouchel, D., see Wordel, R. 230 Mount, M.E., see Sjöblom, K.-L. 450, 451, 456 Mueller, P., see Du, X. 91, 92 Mueller, P., see Sturchio, N.C. 91 Mueller, W.F., see Kudryavtsev, Y. 337 Muguntha Manikandan, N., see Komura, K. 234 Muhs, D., see McGeehin, J. 245 Mukaida, M., see Ogiwara, K. 399 Mullen, G., see Aumann, D.C. 386, 390 Müller, A., see Suter, M. 272
Author Index
Muller, C., see Becker, J.S. 305 Müller, G. 221 Müller, P. 337, 346 Müller, P., see Blaum, K. 299, 345 Müller, P., see Wendt, K. 344 Mulsow, S., see Osvath, I. 450, 451, 453, 464 Mulsow, S., see Povinec, P.P. 138, 148, 149, 155, 157, 450, 453, 458, 459, 464, 466, 467 Muramatsu, Y. 150, 318, 386 Muramatsu, Y., see Fehn, U. 267 Murata, Y., see Komura, K. 234 Murith, C., see Laedermann, J.P. 408 Murphy, K.E., see Kelly, W.R. 86, 95 Murray, C.N., see Fukai, R. 116 Musa, C. 458, 459 Musa, C., see Masse, L. 458 Myers, J.C. 18 Myers, L.E., see Smith, F. 19 Myroslav, V.Z., see Vonderheide, A.P. 297, 321 Nadeau, M.-J., see Fairbanks, R.G. 249– 251 Nadeau, M.-J., see Kilius, L.R. 267 Nadeau, M.J., see Bruhn, F. 245 Nadeau, M.J., see Voelker, A.H.L. 249 Nagaya, Y. 138, 155 Nagel, P., see Berg, T. 349 Nähler, A., see Müller, P. 337, 346 Nakamura, K., see Nagaya, Y. 138, 155 Nakamura, N., see Komura, K. 234 Nakanishi, T., see Paul, M. 280 Nakano, M. 158 Nakashiki, N., see Tsumune, D. 138 Nalimov, V.V. 51 Natrella, M. 70, 113 Navon, E., see Berkovits, D. 280 Navratil, J.D. 18 Nazarov, I.M., see Kogan, R.M. 408 Naziry, M.J., see Sowa, W. 411, 413 NCRP, 211, 220 Neder, H. 194, 232 Neder, H., see Heusser, G. 202, 233
507
Nedler, V.V., see Nalimov, V.V. 51 Nelson, D.M. 150 Nelson, D.M., see Orlandini, K.A. 150 Nelson, D.W. 150 Nemeth, I., see Zombori, P. 417, 427, 428, 430 Nemeth, Zs., see Sekine, T. 396 Nemoto, K., see Aoyama, M. 145 Nesbitt, R.W., see Taylor, R.N. 305 Neu, M.P., see Conradson, S.D. 366 Neu, W., see Schulz, C. 344 Neugart, R., see Schulz, C. 344 Neumaier, S. 194, 202, 231, 236 Neumaier, S., see Arnold, D. 234 Neumaier, S., see Heisinger, B. 255 Neumaier, S., see Hult, M. 235 Neumaier, S., see Laubenstein, M. 108, 163, 202, 203 Neuroth, M., see Schulz, C. 344 Newman, R.T., see Lindsay, R. 430 Newton, D. 287, 288 Newton, D., see Stradling, N. 287 Ng, C.Y., see Liao, C.-L. 297 Nicolaou, G., see Moreno, J.M.B. 397 Nielsen, S., see Betti, M. 359 Nielsen, S.C., see Sturup, S. 305 Nielsen, S.P. 376 Nielsen, S.P., see Chen, Q.J. 152, 153, 312, 390–392, 399, 401 Nielsen, S.P., see Hou, X.L. 318, 384– 390, 397 Nieminen, A., see Moore, I.D. 337 Nies, H. 450 Nies, H., see Povinec, P.P. 142, 148, 154 Niese, S. 194, 202, 218, 223, 229, 234, 235 Niese, S., see Helbig, W. 210, 229 Nikitin, A., see Hirose, K. 138, 154, 155 Nikitin, A., see Ikeuchi, Y. 155, 157 Nikolopoulos, D., see Anastasis, B.P. 430 Nilsson, C., see Nygren, U. 306, 322 Nilsson, K., see Aarkrog, A. 362 Nimz, G.J., see Bergquist, B.A. 268 Noakes, J.E. 450
508
Author Index
Noakes, J.E., see Senftle, F.E. 450 Noakes, S.E., see Noakes, J.E. 450 Noertershaeuser, W., see Blaum, K. 345 Nogami, K., see Yamakoshi, K. 210 Nogar, N.S., see Downey, S.W. 313 Noll, B., see Goerner, W. 394, 396 Nolte, E., see Heisinger, B. 255 Nörtershäuser, W., see Blaum, K. 299 Nörtershäuser, W., see Bushaw, B.A. 336 Nörtershauser, W., see Müller, P. 337, 346 Nörtershäuser, W., see Wendt, K. 344 Noshkin, V.E. 147 Noshkin, V.E., see Bowen, V.T. 137–139, 146, 151, 155 Noshkin, V.E., see Povinec, P.P. 142, 148, 154 Noshkin, V.E., see Robison, W.L. 148 Nourbakhsh, G.D., see Liao, C.-L. 297 Noury, C., see Hatté, C. 245 Nozaki, T., see Honda, T. 399 Nuñez, L.A. 253 Nuñez, L.A., see Meltzer, D.J. 253 Nunez-Lagoz, R., see Hubert, Ph. 217 Nunnemann, M. 338 Nunnemann, M., see Erdmann, N. 339 Nunnemann, M., see Passler, G. 337, 338, 341 Nusko, R., see Heumann, K.G. 336 Nygren, U. 306, 322 Oatts, T.J., see Vance, D.E. 311 Oberdorfer, J.A., see Burnett, W.C. 467 Obert, J., see Sauvage, J. 337 Obrusník, I., see Houdek, F. 394 Ochsenkuhn, K.M. 398 Ochsenkuhn-Petropoulou, M., see Ochsenkuhn, K.M. 398 O’Conner, T.P., see Sturchio, N.C. 91 O’Connor, T.P., see Bailey, K. 91, 92 O’Connor, T.P., see Chen, C.Y. 91, 342 O’Connor, T.P., see Du, X. 91, 92 Oda, K., see Hirose, K. 138, 154, 155 Oda, K., see Ikeuchi, Y. 155, 157 O’Donnell, R.G. 263, 269
Oeschger, H. 210, 229, 234 Oeschger, H., see Loosli, H.H. 232 Oeschger, H.O., see Loosli, H.H. 91 Oestby, G., see Salbu, B. 19 Ogiwara, K. 399 Ognibene, T.J., see McAninch, J.E. 277 O’Hara, M.J., see Egorov, O.B. 311 Ohmomo, Y., see Muramatsu, Y. 386 Oi, T., see Honda, T. 399 Oi, T., see Ogiwara, K. 399 Olesik, J.W., see Stewart, I.I. 309 Olive, V., see McCartney, M. 267, 314 Oliveira, J., see Burnett, W.C. 467 Oliver, E., see Handl, J. 386, 390 Oliver, M.A., see Webster, R. 10, 11 Olshanski, E.D., see Liechtenstein, V.K. 275 O’Malley, J.M., see Burr, G.S. 249 Oms, J., see Sauvage, J. 337 O’Nions, R.K., see Cohen, A.S. 323 Ophel, T.R., see Fifield, L.K. 265, 269, 271, 280, 282, 332 Oregioni, B., see Povinec, P.P. 138, 148, 149, 155, 318, 451, 453, 467, 468, 474 Orlandini, K.A. 150 Orlandini, K.A., see Cooper, L.W. 301, 311 Orlandini, K.A., see Winkler, S. 279 Orlov, S.Y., see Barzakh, A.E. 342 O’Rourke, D., see Burnett, W.C. 467 Osán, J., see Eriksson, M. 357, 362 Osán, J., see Jernström, J. 357, 361, 367 Osán, J., see Ro, C.-U. 363 Osán, J., see Török, S. 357, 358, 360, 367 Oshima, M., see Hatsukawa, Y. 379 Ossaka, T., see Honda, T. 399 Ossaka, T., see Ogiwara, K. 399 Ostapczuk, P., see Rossbach, M. 35 Ostapczuk, P., see Vonderheide, A.P. 297, 321 Ostapczuk, P., see Zoriy, M.V. 297, 305, 321, 324, 332 Osvath, I. 450–453, 457–459, 464 Osvath, I., see Baxter, M.S. 456
Author Index
Osvath, I., see Hamilton, T.H. 451 Osvath, I., see Povinec, P.P. 157, 235, 449–453, 456, 464, 466, 467 Otsuji, M., see Ikeda, N. 394, 396 Otsuji, M., see Kim, C.K. 313 Ott, U., see Berg, T. 349 Otten, E.W., see Blaum, K. 299, 345 Otten, E.W., see Schulz, C. 344 Otten, E.W., see Wendt, K. 344 Oughton, D.H. 285, 286, 303, 332 Oughton, D.H., see Borretzen, P. 285 Oughton, D.H., see Keith-Roach, M. 317 Oughton, D.H., see Keith-Roach, M.J. 394 Oughton, D.H., see Lind, O.C. 285 Oughton, D.H., see Salbu, B. 19, 348, 357, 367 Oughton, D.H., see Skipperud, L. 285, 286 Overpeck, J.T., see Hughen, K.A. 249 Ozorovich, Y.R., see Povinec, P.P. 451, 467 Ozyurt, N., see Burnett, W.C. 467 Pachucki, C., see Krey, P.W. 150 Pagava, S. 202 Pajo, L., see Tamborini, G. 348 Palmer, Ph.D., see Conradson, S.D. 366 Panteleyev, V.N., see Alkhazov, G.D. 342 Panzarino, N., see Shebell, P. 430 Pappas, R.S., see Ting, B.G. 306 Parekh, P.P., see Semkow, T.M. 164, 166, 194 Paretzke, H.G., see Jacob, P. 411, 426, 430 Parkinson, J.A. 410 Parr, R.M. 75, 375, 376 Parr, R.M., see Currie, L.A. 52 Parry, S.J. 386 Parsi, P., see Hamilton, T.H. 451 Paschal, D.C., see Ting, B.G. 306 Passler, G. 337, 338, 341 Passler, G., see Eichler, B. 333 Passler, G., see Erdmann, N. 339, 349
509
Passler, G., see Grüning, C. 337–339 Passler, G., see Huber, G. 332 Passler, G., see Krönert, U. 337 Passler, G., see Maul, J. 334, 348 Passler, G., see Nunnemann, M. 338 Passler, G., see Sewtz, M. 347 Passler, G., see Trautmann, N. 332, 338 Passler, G., see Waldek, A. 339 Paterne, M., see Goslar, T. 254 Paterne, M., see Hatté, C. 245 Paterson, B.A., see Beck, J.W. 250 Paterson, B.A., see Richards, D.A. 250 Patil, G.P. 30, 33, 34 Patil, S.K., see Raghavan, R. 392 Patnaik, P.B. 80 Patterson, H.W., see Yamashita, M. 214 Patterson, L.J., see Sturchio, N.C. 91 Paul, M. 280, 344 Paul, M., see Berkovits, D. 280 Paul, M., see Winkler, S. 279 Paule, R.C., see Mandel, J. 114 Paulsen, P.J., see Kelly, W.R. 86, 95 Paulsen, R., see Burnett, W.C. 467 Paviet-Hartmann, P., see Conradson, S.D. 366 Pavlik, B. 437 Payne, M.G. 343 Payne, M.G., see Hurst, G.S. 332 Peak, L.S., see Hashemi-Nezhad, S.R. 221–223, 229 Pearson, E.S. 80 Peerani, P., see Wallenius, M. 332 Pellow, P., see Stradling, N. 287 Pelzmann, W.L., see Snyder, G.T. 267 Penrose, W.R., see Orlandini, K.A. 150 Pentreath, R., see Kershaw, P. 459, 462 Pentreath, R.J. 138, 154 Penttilä, H., see Moore, I.D. 337 Perdue, P.T., see Dickson, H.W. 411, 413 Perelygin, V.P., see Kashkarov, L.L. 357 Perkins, R.W. 138 Perkins, R.W., see Cooper, J.A. 164 Perna, L. 361 Pernicka, F., see Tompson, I.M.G. 425– 427
510
Author Index
Perrin, D.R., see Gladney, E.S. 399, 401 Perry, D.L., see Reijonen, J. 377 Peru, S., see Sauvage, J. 337 Peters, R.J., see Gladney, E.S. 399, 401 Peterson, B., see Lambert, D. 116 Peterson, J.R., see Erdmann, N. 339 Peterson, L.C., see Hughen, K.A. 249 Peterson, R., see Burnett, W.C. 473 Petit, J.R., see Raisbeck, G.M. 250, 255 Petrunin, V.V., see Kudryavtsev, Y.A. 343 Petrunin, V.V., see Schulz, C. 344 Pettersson, H.B.L., see Hirose, K. 138, 154, 155 Pettersson, H.B.L., see Ikeuchi, Y. 155, 157 Pettersson, H.B.L., see Kim, C.K. 151 Pettersson, H.B.L., see Povinec, P.P. 138, 148, 149, 155, 450, 451, 456, 458, 459 Peuser, P. 332 Pfau, A., see Handl, J. 386, 390 Phelps, P.L., see Anspaugh, L.R. 413 Philbin, P.W., see Senftle, F.E. 450 Philipsen, V., see Bastiaansen, J. 334 Philipsen, V., see Erdmann, N. 334, 348 Philipsen, V., see Lievens, P. 334 Philipsen, V., see Vandeweert, E. 334 Phillips, F.M., see Elmore, D. 241, 243 Phillips, F.M., see Gosse, J. 254 Phinney, D., see Tamborini, G. 358 Pibida, L. 397 Pibida, L., see Karam, L.R. 397 Piccione, M., see Bienvenu, P. 319 Piccot, D. 393 Piccot, D., see Frechou, C. 384 Picer, M. 399 Pich, G.M., see Newton, D. 287, 288 Pickering, W.F. 21, 32 Pickhardt, C., see Becker, J.S. 305 Pickhardt, C., see Vonderheide, A.P. 297, 321 Pickhardt, C., see Zoriy, M.V. 305, 324, 332 Pickup, G. 439 Pierce, J.L. 252, 253
Pierce, J.L., see Meyer, G.A. 252 Pigati, J., see Lifton, N.A. 255 Pigati, J.S. 255 Pike, S., see Dai, M.H. 332 Pilier, M., see Chartier, F. 322 Pillai, K.C., see Dang, H.S. 398, 399 Pin, C., see Joannon, S. 323 Pinard, J., see Sauvage, J. 337 Pinte, G., see Germain, P. 390, 392, 393 Pinte, G., see May, S. 390, 392, 393 Pitard, F.F. 29 Plastino, W. 233, 234 Platzner, I.T., see Becker, J.S. 305 Pliml, A., see Tschiersch, J. 39 Pointurier, F. 305 Pointurier, F., see Baglan, N. 304 Pointurier, F., see Chiappini, R. 155 Pöllänen, R. 43, 348 Pöllänen, R., see Moring, M. 362 Pomansky, A.A. 230 Pomansky, A.A., see Kovalchuk, E.L. 230 Pomar, C., see Wallner, C. 279 Poole, A.J. 459, 463 Potts, M.J. 420 Povel, H.P., see Kozlowski, T. 179 Povinec, P.P. 21, 43, 108, 138, 142, 148, 149, 154, 155, 157, 163, 164, 166, 169, 194, 196, 198, 201, 202, 206, 210, 226, 228, 233, 234, 235, 318, 371, 401, 449– 453, 456, 458, 459, 464, 466–468, 472, 474 Povinec, P.P., see Aoyama, M. 138, 146 Povinec, P.P., see Baxter, M.S. 456 Povinec, P.P., see Beláˇn, T. 171 Povinec, P.P., see Bellotti, E. 233 Povinec, P.P., see Burnett, W.C. 467 Povinec, P.P., see Cimbák, Š. 39 Povinec, P.P., see Csongor, E. 96 Povinec, P.P., see Dovlete, C. 131 Povinec, P.P., see Hamilton, T.H. 451 Povinec, P.P., see Harms, I.H. 450, 451 Povinec, P.P., see Hirose, K. 138, 141, 154, 155 Povinec, P.P., see Hlinka, V. 198
Author Index
Povinec, P.P., see Ikeuchi, Y. 155, 157 Povinec, P.P., see Ito, T. 138, 155 Povinec, P.P., see Kontar, E.A. 467 Povinec, P.P., see La Rosa, J.J. 139, 152, 153, 311 Povinec, P.P., see Laubenstein, M. 108, 163, 202, 203 Povinec, P.P., see Levy-Palomo, I. 467 Povinec, P.P., see Livingston, H.D. 137– 139, 142, 144, 148, 154, 155, 157 Povinec, P.P., see Nakano, M. 158 Povinec, P.P., see Osvath, I. 450–453, 457–459, 464 Povinec, P.P., see Pagava, S. 202 Povinec, P.P., see Sjöblom, K.-L. 450, 451, 456 Povinec, P.P., see Staníˇcek, J. 196 Povinec, P.P., see Sýkora, I. 169, 171, 196, 198 Povinec, P.P., see Taniguchi, M. 469 Povinec, P.P., see Vojtyla, P. 163, 164, 182, 191, 194, 228 Povinec, P.P., see Wyse, E.J. 264, 269, 306, 332, 371 Povinec, P.P., see Zvara, I. 163, 198 Prasad, N.V.S.V., see Kudryavtsev, Y. 337 Pratt, R.H., see Kim, L. 181 Prichard, H.M., see Hess, C.T. 40 Priest, N.D., see Fifield, L.K. 265, 269, 271, 280, 282, 332 Priest, N.D., see Newton, D. 287, 288 Priest, N.D., see O’Donnell, R.G. 263, 269 Priller, A., see Kraft, S. 272 Priller, A., see Steier, P. 275, 278, 279 Priller, A., see Vockenhuber, C. 279 Priller, A., see Winkler, S. 279 Priore, P., see Bond, G. 251 Privitera, A.M.G., see Burnett, W.C. 467 Privitera, A.M.G., see Povinec, P.P. 451, 467 Probst, T., see Berryman, N. 321 Probst, T.U., see Song, M. 316, 321, 397 Proctor, I.D., see Bergquist, B.A. 268
511
Proctor, I.D., see McAninch, J.E. 277 Prohaska, T., see Hlavay, J. 18 Proost, K., see Lind, O.C. 367 Proost, K., see Salbu, B. 357, 358, 367 Przyborowski, J. 103–105, 107 Pucelj, B., see Korun, M. 427 Pulhani, V. 398 Pullat, V.R., see Dang, H.S. 398, 399 Pulsipher, B.A. 13 Pulsipher, B.A., see Hassig, N.L. 18, 31 Purser, K.H. 289 Purtschert, R., see Du, X. 92 Purtschert, R., see Sturchio, N.C. 91 Putaux, J.C., see Sauvage, J. 337 Qu, H., see Glover, S.E. 399 Quade, J., see Lifton, N.A. 254, 255 Quay, P.D. 154 Quimby, W.F., see Borgman, L.E. 29 Quirk, W., see Burr, G.S. 249 Raabe, R., see Kudryavtsev, Y. 337 Radiological Protection Institute of Ireland, RPII, 464, 466 Ragan, P., see Fülöp, M. 427 Ragazzi, S., see Bellotti, E. 233 Raghavan, R. 392 Raisbeck, G., see Frechou, C. 384 Raisbeck, G., see Gascard, J.C. 266 Raisbeck, G., see Yiou, F. 266 Raisbeck, G.M. 250, 255, 266, 267, 384, 390 Raisbeck, G.M., see Mironov, V. 267 Rajar, R., see Burnett, W.C. 467 Rajar, R., see Povinec, P.P. 451, 467 Rajendran, K., see McCartney, M. 267, 314 Ralph, E.K., see Klein, J. 247 Ralska-Jasiewiczowa, M., see Goslar, T. 254 Ramakrishna, V.V., see Raghavan, R. 392 Ramaniah, M.V., see Raghavan, R. 392 Ramebäck, H., see Nygren, U. 322 Ramessur, R.T., see Burnett, W.C. 467
512
Author Index
Ramlia, A.T. 398 Ramsden, D., see Watt, D.E. 189 Ramsey, N.F. 333 Ramsey, S.W., see Hughen, K.A. 249, 250 Ranasinghe, V., see Thomas, B.W. 450 Rance, E., see Stradling, N. 287 Récy, J., see Burr, G.S. 249 Reedy, R.C., see Dep, L. 221 Reeve, C.P. 80 Reeves, J.H., see Arthur, R.J. 222, 224, 225 Reeves, J.H., see Brodzinski, R.L. 168, 218, 231 Rehkämper, M., see Luo, X. 324 Reiber, K.M., see Larivière, D. 322, 323 Reijonen, J. 377 Reilly, S.D., see Conradson, S.D. 366 Reimanm, R.T., see Shebell, P. 430 Reimer, P.J. 247, 249, 251 Reimer, P.J., see Hughen, K.A. 249, 250 Reimer, P.J., see van der Plicht, J. 247, 250 Reimer, R.W., see Hughen, K.A. 249, 250 Reimer, R.W., see Reimer, P.J. 247, 249, 251 Reimer, R.W., see van der Plicht, J. 247, 250 Reines, D., see McGeehin, J. 245 Reiter, E.R. 137 Remedy, W.R. 19, 22 Remmele, S., see Hughen, K.A. 249, 250 Remmele, S., see Reimer, P.J. 247, 249, 251 Remmele, S., see van der Plicht, J. 247, 250 Repinc, U. 398 Repnow, R., see Liechtenstein, V.K. 275 Reusen, I., see Kudryavtsev, Y. 337 Revay, Z., see Reijonen, J. 377 Review of Particle Properties, RPP, 180 Reyss, J.-L. 194, 235 Reyss, J.-L., see Laubenstein, M. 108, 163, 202, 203
Rho, B.H., see Kim, C.K. 315 Riˇca, I., see Liteanu, C. 72 Richard-Serre, C., see Sauvage, J. 337 Richards, D.A. 250 Richards, D.A., see Beck, J.W. 250, 252 Richards, D.A., see Cheng, H. 302 Richards, D.A., see van der Plicht, J. 247, 250 Richtáriková, M., see Beláˇn, T. 171 Richter, S. 268 Rieck, H.G., see Kaye, J.H. 210, 230 Riese, W.C., see Snyder, G.T. 267 Rietz, B., see Hou, X.L. 318, 385–390 Rindi, A. 221 Rink, W.J. 41 Rinta-Antila, S., see Moore, I.D. 337 Rius, F.X., see Boqué, R. 81 Ro, C.-U. 363 Roalsvik, J.P., see Suzuki, T. 179 Robakidze, T., see Pagava, S. 202 Robb, W., see Van R. Smit, J. 139 Robbins, J.A., see McGeehin, J. 245 Robens, E., see Aumann, D.C. 386, 389, 390 Roberts, M.L., see McAninch, J.E. 277 Roberts, P.D., see Jones, D.G. 409, 450, 459, 462 Roberts, P.D., see Miller, J.M. 450, 459 Robinson, W., see Muramatsu, Y. 150 Robison, W.L. 148 Roca-Saiz, M.C., see Maul, J. 334, 348 Rocco, G.G. 138, 139, 146 Rodushkin, I., see Nygren, U. 306 Röllin, S., see Ayranov, M. 311 Ronen, D., see Margaritz, M. 40 Rook, H.L. 389 Roos, J.B. 51 Rosa, J.L., see Wyse, E.J. 264, 269, 332 Rosamilia, S. 398 Röschert, G., see Schröder, W.U. 179, 180 Roselli, C., see Desideri, D. 269, 324 Rosenberg, R.J. 372, 384, 390 Rossbach, M. 35 Rossi, L., see Bellotti, E. 233
Author Index
Rostek, F., see Bard, E. 250 Rourke, F., see Krey, P.W. 150 Roussière, B., see Sauvage, J. 337 Rowan, J.T. 297, 298 Rowe, M., see Russ, J. 245 Royden, C.S., see Ketterer, M.E. 268 Rubin, K.H. 302 Rucklidge, J. 267 Rucklidge, J.C., see Kilius, L.R. 267 Rudenauer, F.G., see Tamborini, G. 358 Ruedenauer, F., see Danesi, P.R. 358 Ruf, M. 390, 392, 393 Ruffel, A. 430 Ruffell, A., see Hadley, M.J. 430 Rugel, G., see Wallner, C. 279 Ruhm, W., see Uschida, S. 314 Ruisi, S., see Rosamilia, S. 398 Rumyantsev, O.V., see Khitrov, L.M. 43 Runde, W.H., see Conradson, S.D. 366 Rusetski, L., see Pagava, S. 202 Russ, J. 245 Ruster, W., see Peuser, P. 332 Ryan, D.E., see Bem, H. 399 Ryan, T.P., see Osvath, I. 450, 451, 453, 464 Rybacek, K. 416, 427, 428 Rybach, L. 439 Rybach, L., see Schwarz, G.F. 408 Sabol, J., see Tykva, R. 331 Sáez Vergara, J.C., see Wissman, F. 430 Sáez-Vergara, J.C., see Tompson, I.M.G. 425–427 Sahli, H., see Ayranov, M. 311 Saiki, M., see Vasconcellos, M.B.A. 401 Sakamoto, K., see Paul, M. 280 Sakamoto, K., see Tanaka, S. 210, 234 Sakanoue, M., see Hisamatsu, S. 150 Sakanoue, M., see Imai, T. 151 Sakanoue, M., see Komura, K. 150 Sakata, K. 296 Salbu, B. 19, 348, 356–358, 367 Salbu, B., see Danesi, P.R. 358 Salbu, B., see Lind, O.C. 285, 367
513
Salbu, B., see Oughton, D.H. 285, 286, 303, 332 Salo, A., see Sjöblom, K.-L. 450, 451, 456 Salvamoser, J., see Steinkopff, T. 92 Sanborn, P. 252, 253 Sanchez-Angulo, C., see Mas, J.L. 315 Sandalls, F.J. 31 Sanders, T.W., see Augustson, J.H. 401 Sanderson, D.C.W. 10, 409, 418, 420, 424, 437–442 Sanderson, D.C.W., see Allyson, J.D. 416, 420, 440 Sanderson, D.C.W., see Creswell, A.J. 437 Sanderson, D.C.W., see Tyler, A.N. 408, 410, 415–417, 420, 424–426, 428–431, 440, 441 Sansone, U., see Rosamilia, S. 398 Santos, F.J. 318 Santos, G.M., see Bird, M.I. 245 Sapozhnikov, D.Yu., see Kalmykov, St.N. 392, 393 Sapozhnikov, Yu.A., see Kalmykov, St.N. 392, 393 Sarntheim, M., see Voelker, A.H.L. 249 Saruhashi, K. 138, 146 Saruhashi, K., see Miyake, Y. 138 Saruhashi, K., see Sugiura, Y. 138 Sauvage, J. 337 Sazykina, T., see Betti, M. 359 Sazykina, T.G., see Sjöblom, K.-L. 450, 451, 456 Schaerf, K., see Parr, R. 75 Scherf, M., see Schulz, C. 344 Schiavo, M.A. 467 Schiavo, M.A., see Povinec, P.P. 451, 467 Schilling, E.G. 18 Schilling, G., see Wedekind, Ch. 450 Schilling, J.G., see Nies, H. 450 Schimmak, W., see Hillman, U. 416, 417 Schladot, J.D., see Rossbach, M. 35 Schmidt, A. 318 Schmidt, S., see Reyss, J.-L. 194, 235
514
Author Index
Schmitt, A., see Blaum, K. 299, 345 Schnabel, Ch., see Schmidt, A. 318 Scholkovitz, E.R., see Buesseler, K.O. 150 Scholten, J., see Burnett, W.C. 467 Schönhense, G., see Berg, T. 349 Schöpe, H., see Backe, H. 347 Schopper, E. 211 Schotterer, U., see Oeschger, H. 210, 229, 234 Schramm, A. 250 Schröder, F., see Müller, G. 221 Schröder, W.U. 179, 180 Schroettner, T. 164, 166, 194 Schuert, E.A., see Shirasawa, T.H. 138, 146 Schuettelkopf, H., see Wilhelm, J.G. 39 Schuller, P., see Handl, J. 386, 390 Schulz, C. 344 Schumacher, M., see Müller, G. 221 Schumann, P.G. 347 Schuppler, S., see Berg, T. 349 Schwaiger, M., see Laubenstein, M. 108, 163, 202, 203 Schwaigerm, M., see Schroettner, T. 164, 166, 194 Schwalbach, R., see Wendt, K. 344 Schwamb, P., see Backe, H. 347 Schwamb, P., see Sewtz, M. 347 Schwarcz, H.P., see Li, W.-X. 41 Schwartz, M. 256 Schwarz, G., see Rybach, L. 439 Schwarz, G.F. 408 Schwarz, J., see Sjöblom, K.-L. 450, 451, 456 Schwarz, S., see Wies, K. 342 Schwenker, C.D., see Semkow, T.M. 164, 166, 194 Schwieters, J., see Weyer, S. 296 Science and Technology Agency, STA, 139 Scott, E.M., see Baxter, M.S. 456 Scott, E.M., see Hamilton, T.H. 451 Scott, E.M., see Povinec, P.P. 138, 142, 154
Scott, E.M., see Sanderson, D.C.W. 10, 420, 424, 437, 439, 441, 442 Scott, E.M., see Sjöblom, K.-L. 450, 451, 456 Scott, E.M., see Tyler, A.N. 408, 410, 415–417, 420, 424–426, 428–431, 440, 441 Scott, S., see Benkhedda, K. 323 Scottish Environment Protection Agency, SEPA, 42 Sebastian, V., see Sauvage, J. 337 Seber, G.A.F. 30, 33 Seber, G.A.F., see Thompson, S.K. 13 Segal, I., see Becker, J.S. 305 Segal, I., see Boulyga, S.F. 305, 324 Segal, M.G., see Sandalls, F.J. 31 Seibert, A., see Kuczewski, B. 311 Seibert, U.-A., see Wendt, K. 344 Sekatsky, S.K., see Alkhazov, G.D. 342 Seki, R., see Ikeda, N. 394, 396 Seki, R., see Kim, C.K. 390, 394 Sekine, T. 315, 396 Sekine, T., see Yagi, M. 396 Seliverstov, M.D., see Barzakh, A.E. 342 Seltzer, S.M. 181 Seltzer, S.M., see Kim, L. 181 Seman, M., see Povinec, P. 233 Semenov, A.A., see Povinec, P. 233 Semkow, T.M. 164, 166, 194 Senftle, F.E. 450 Sennhauser, U., see Kozlowski, T. 179 Sequeira, S., see Gascard, J.C. 266 Service, M., see Leonard, K.S. 267 Seto, Y., see Hirose, K. 138, 154, 155 Sewtz, M. 347 Sewtz, M., see Backe, H. 347 Shaban, I.S., see Macášek, F. 18 Sharma, R.C., see Dang, H.S. 398 Shaw, J., see Dyke, A.S. 254 Shebell, P. 430 Shebell, P., see Miller, K.M. 427, 428 Shefer, R.E., see Hughey, B.J. 243 Sheffield, A.E., see Currie, L.A. 83 Shima, S., see Kobayashi, Y. 450, 468
Author Index
Shima, S., see Povinec, P.P. 138, 142, 148, 149, 154, 155 Shinohara, N., see Hatsukawa, Y. 379 Shinonaga, T., see Tschiersch, J. 39 Shipman, W.H., see Weiss, H.V. 147 Shiraishi, K., see Kim, C.K. 313 Shirasawa, T.H. 138, 146 Shizuma, K., see Hult, M. 235 Shkinev, V., see Tcherkezian, V. 19 Showers, W., see Bond, G. 251 Shrivistava, A., see Kraft, S. 272 Shutt, A.L., see Stradling, N. 287 Sickel, M., see Skipperud, L. 285, 286 Sickel, M.A., see Lind, O.C. 285 Sideras-Haddad, E., see McAninch, J.E. 277 Sidhu, R.S. 152, 153 Siffedine, A., see Turcq, B. 252 Sigman, D.M., see Hughen, K.A. 249 Silverans, R.E., see Bastiaansen, J. 334 Silverans, R.E., see Erdmann, N. 334, 348, 349 Silverans, R.E., see Lievens, P. 334 Silverans, R.E., see Vandeweert, E. 334 Silverman, B.W., see Beck, J.W. 250, 252 Sima, O., see Arnold, D. 234 Simgen, H. 94 Simionovici, A., see Salbu, B. 357, 367 Simmonds, J., see Betti, M. 359 Simnits, A.S., see deCorte, F. 374 Simon, R., see Eriksson, M. 357, 362 Simon, R., see Jernström, J. 357, 361, 367 Simpson, J.A. 223 Simpson, J.C., see Gilbert, R.O. 410 Simpson, J.D., see Chichester, D.L. 377 Sinha, A.K., see Patil, G.P. 30, 33, 34 Sitár, B., see Povinec, P. 233 Sitte, K. 213 Sivintsev, Y.V., see Sjöblom, K.-L. 450, 451, 456 Šivo, A., see Beláˇn, T. 171 Sivo, A., see Cimbák, Š. 39 Sjöblom, K.-L. 450, 451, 456
515
Skipper, P.L., see Hughey, B. 243 Skipperud, L. 285, 286 Skipperud, L., see Lind, O.C. 285 Skipperud, L., see Oughton, D.H. 285, 286, 303, 332 Skipperud, L., see Salbu, B. 348, 357 Slaback, L.A., see Lindström, R.M. 223, 228, 235 Slutskii, G.K., see Moskvin, L.N. 39 Smart, P.L., see Beck, J.W. 250, 252 Smart, P.L., see Richards, D.A. 250 Smedley, P., see Povinec, P.P. 458, 459 Smeed, D.A., see Hill, A.E. 465 Smekal, F.G. 210 Smend, F., see Müller, G. 221 Smith, A., see Reijonen, J. 377 Smith, A.M., see Hotchkis, M.A.C. 270, 278, 287 Smith, C., see Burnett, W.C. 467 Smith, F. 19 Smith, L., see Burnett, W.C. 467 Smith, P.F. 169 Smith-Briggs, J.L. 18 Snigirev, A., see Salbu, B. 357, 367 Snigireva, I., see Salbu, B. 357, 367 Snyder, G.T. 267 Snyder, G.T., see Fehn, U. 267 Søernsen, P., see Løvborg, L. 410 Sohnius, B., see Peuser, P. 332 Sonett, C.P., see McHargue, L.R. 256 Song, M. 316, 321, 397 Sonnett, C.P., see Damon, P.E. 251 Soto, C., see Kilius, L.R. 267 Soto, C.Y., see Kieser, W.E. 318 Soubies, F., see Turcq, B. 252 Southon, J., see McGeehin, J. 245 Southon, J.R., see Bergquist, B.A. 268 Southon, J.R., see Hughen, K.A. 249, 250 Southon, J.R., see McAninch, J.E. 277 Southon, J.R., see Reimer, P.J. 247, 249, 251 Southon, J.R., see van der Plicht, J. 247, 250 Sowa, W. 411, 413
516
Author Index
Sparks, B.J., see Shebell, P. 430 Spaulding, J.D., see Senftle, F.E. 450 Speelma, W.J., see Lindsay, R. 430 Spies, H., see Goerner, W. 394, 396 Sreekumaran, C., see Folsom, T.R. 139 Srivastava, R.M., see Isaaks, E.H. 18 Stamm, H.H. 397, 398 Standring, W.J.F., see Borretzen, P. 285 Stanford, D.J., see Meltzer, D.J. 253 Staníˇcek, J. 196 Staníˇcek, J., see Pagava, S. 202 Staníˇcek, J., see Sýkora, I. 169, 171, 196 Staubwasser, M., see Foster, D.A. 323 Stavina, P., see Vojtyla, P. 164, 166, 167, 171, 182, 194 Steadman, D.W. 253, 254 Steier, P. 275, 278, 279 Steier, P., see Hrnecek, E. 279 Steier, P., see Kraft, S. 272 Steier, P., see Liechtenstein, V.K. 275 Steier, P., see Vockenhuber, C. 279 Steier, P., see Winkler, S. 279 Stein, M., see Schramm, A. 250 Stein, M.L. 30 Steiner, M., see Uschida, S. 314 Steinkopff, T. 92 Stephens, L.D., see Yamashita, M. 214 Sternheimer, R.M. 180 Stetzer, O., see Erdmann, N. 358 Stetzer, O., see Nunnemann, M. 338 Stevens, M.A., see Jiang, S.-J. 296 Stewart, I.I. 309 Stewart, M.L. 42 Stieglitz, T., see Burnett, W.C. 467 Stieglitz, T., see Taniguchi, M. 474 Stingeder, G.J., see Hlavay, J. 18 Stoble, C., see Winkelmann, I. 437 Stocker, M., see Suter, M. 272 Stocker, M., see Synal, H.A. 243 Stocker, M., see Wacker, L. 272–274, 279, 280, 332 Storm, E. 411 Strachnov, I., see Maul, J. 334, 348 Stradling, G.N., see Etherington, G. 287 Stradling, N. 287
Strand, P., see Oughton, D.H. 285 Strange, L., see O’Donnell, R.G. 263, 269 Streli, C., see Eriksson, M. 357, 362 Strohal, P., see Picer, M. 399 Strutt, M.H., see Jones, D.G. 409, 459 Stuart, D.R., see Currie, L.A. 89 Stuewer, D., see Feldmann, I. 298, 299 Stuiver, M. 247–249 Stuiver, M., see Hughen, K.A. 249, 250 Stuiver, M., see Quay, P.D. 154 Stuiver, M., see Reimer, P.J. 247, 249, 251 Stuiver, M., see van der Plicht, J. 247, 250 Stukheil, K., see Wilhelmova, L. 96, 97 Sturchio, N.C. 91 Sturchio, N.C., see Du, X. 92 Sturdivan, L., see Kurtz, D.A. 64 Sturup, S. 305 Stürup, S., see Keith-Roach, M.J. 317, 394 Suddueth, J.E., see Rook, H.L. 389 Sudek, C., see Berg, T. 349 Sugihara, T.T. 146 Sugimura, Y., see Hirose, K. 138, 143, 150, 152–155 Sugimura, Y., see Miyake, Y. 138 Sugimura, Y., see Saruhashi, K. 138, 146 Sugiura, Y. 138 Sugiura, Y., see Miyake, Y. 138, 146 Suguio, K., see Turcq, B. 252 Sultan, M., see Sturchio, N.C. 91 Sumiya, M., see Muramatsu, Y. 386 Summerfield, M.A., see Cockburn, H.A.P. 254 Sun, G., see Suter, M. 272 Sun, M., see Reijonen, J. 377 Suslova, K.G., see Leggett, R.W. 287, 288 Suter, M. 243, 272 Suter, M., see Fifield, L.K. 243, 279, 332 Suter, M., see Grajcar, M. 243 Suter, M., see Santos, F.J. 318 Suter, M., see Schmidt, A. 318
Author Index
Suter, M., see Synal, H.A. 243 Suter, M., see Wacker, L. 272–274, 279, 280, 332 Sutherland, R.A. 410 Suzuki, T. 179 Sverzelatti, P.P., see Bellotti, E. 233 Svoboda, K., see Houdek, F. 394 Sýkora, I. 169, 171, 196, 198 Sýkora, I., see Zvara, I. 163, 198 Synal, A., see Denk, E. 346 Synal, H.-A. 243 Synal, H.-A., see Baumgartner, S. 250 Synal, H.-A., see Grajcar, M. 243 Synal, H.-A., see Fifield, L.K. 243, 279, 332 Synal, H.-A., see Santos, F.J. 318 Synal, H.-A., see Schmidt, A. 318 Synal, H.-A., see Suter, M. 243, 272 Synal, H.-A., see Wacker, L. 272–274, 279, 280, 332 Szalóki, I., see Ro, C.-U. 363 Szarka, J., see Bellotti, E. 233 Szarka, J., see Povinec, P. 201, 233 Szentmiklosi, L., see Reijonen, J. 377 Tabarelli de Fatis, T., see Bellotti, E. 233 Tagami, K. 315 Tagami, K., see Mas, J.L. 314, 315 Tagami, K., see Muramatsu, Y. 150 Tagami, K., see Sekine, T. 315 Tagami, K., see Uschida, S. 314 Taguchi, Y., see Katagiri, M. 450 Taillie, C., see Patil, G.P. 30, 33, 34 Tait, C.D., see Conradson, S.D. 366 Takagi, J., see Tanaka, S. 210, 234 Takahashi, K., see Kobayashi, Y. 450, 468 Takahashi, R., see Kobayashi, Y. 450, 468 Takaku, Y. 318 Takaku, Y., see Kim, C.K. 313 Talamo, S., see Hughen, K.A. 249, 250 Talamo, S., see Reimer, P.J. 247, 249, 251 Talbot, R.J., see Newton, D. 287, 288
517
Talvitie, N.A. 153 Tamborini, G. 348, 358 Tamborini, G., see Betti, M. 348, 357, 358 Tamborini, G., see Erdmann, N. 358 Tamborini, G., see Eriksson, M. 357, 362 Tamborini, G., see Jernström, J. 357, 361, 367 Tamborini, G., see Török, S. 357, 358, 360, 367 Tanaka, S. 210, 234 Taniguchi, M. 467, 469, 474 Taniguchi, M., see Bokuniewicz, H. 474 Taniguchi, M., see Burnett, W.C. 467 Taniguchi, M., see Povinec, P.P. 451, 467 Tannenbaum, S.R., see Hughey, B. 243 Tanner, A.B., see Senftle, F.E. 450 Tanner, J.E., see Sjöblom, K.-L. 450, 451, 456 Tanner, S. 296 Tanner, S.D. 298, 310 Tanner, S.D., see Bandura, D.R. 300 Taylor, B., see Povinec, P.P. 458, 459 Taylor, C., see Povinec, P.P. 458, 459 Taylor, F.W., see Burr, G.S. 245, 249 Taylor, F.W., see Edwards, R.L. 249, 253 Taylor, F.W., see Hughen, K.A. 249, 250 Taylor, F.W., see Reimer, P.J. 247, 249, 251 Taylor, G.R., see Hewson, R.D. 439 Taylor, J.K., see Kratochvil, B. 35 Taylor, J.K., see Kurtz, D.A. 64 Taylor, P.D.P., see Richter, S. 268 Taylor, R.E. 254 Taylor, R.N. 305 Taylor, R.N., see Warneke, T. 150 Taylor, V., see Epov, V.N. 322, 323 Taylor, V.F., see Lariviere, D. 332 Tayor, F.W., see Cutler, K.B. 249 Tcherkezian, V. 19 Tench, O.K., see Semkow, T.M. 164, 166, 194 Teng, R., see Fehn, U. 384 Teper, L., see Malczewski, D. 430 Teplyakov, N., see Becker, J.S. 305
518
Author Index
Terpenning, I., see Lambert, D. 116 Testa, C., see Boulyga, S.F. 332 Testa, C., see Desideri, D. 269, 324 the ISOLDE Collaboration, see Krönert, U. 337 the ISOLDE Collaboration, see Schulz, C. 344 Thein, M., see Fukai, R. 150 Thein, M., see Lee, S.C. 313 Theodore, L. 38 Theodórsson, P. 112, 163, 211, 220, 221 Theodórsson, P., see Laubenstein, M. 108, 163, 202, 203 Theunissen, K., see Fernandez-Alonso, M. 439 Thieme, K., see Goerner, W. 394, 396 Thirlwall, M.F. 304 Thisell, M., see Shebell, P. 430 Thoen, P., see Vandeweert, E. 334 Thomas, B.W. 450 Thomas, B.W., see Miller, J.M. 450, 459 Thomas, C., see Feldmann, I. 299 Thomas, C.W., see Perkins, R.W. 138 Thomas, M., see Winkelmann, I. 437, 439 Thomas, R., see Ure, M. 18 Thompson, S.K. 3, 8, 9, 13, 14, 30, 33, 35 Thompson, S.K., see Seber, G.A.F. 30, 33 Thonnard, N., see Payne, M.G. 343 Thörle, P., see Backe, H. 347 Thörle, P., see Sewtz, M. 347 Thummerer, S. 417, 428, 430 Tiku, M.L. 81 Tims, S., see Skipperud, L. 285, 286 Tims, S.G. 266, 284 Tims, S.G., see Everett, S.J. 290 Tims, S.G., see Lind, O.C. 285 Tims, S.G., see Wacker, L. 313 Ting, B.G. 306 Tinker, R., see Povinec, P.P. 458, 459 Tisnerat-Laborde, N., see Bard, E. 249 Tisnérnat-Laborde, N., see Goslar, T. 254 Tkalin, A., see Hirose, K. 138, 154, 155
Tkalin, A., see Ikeuchi, Y. 155, 157 Togawa, O., see Hirose, K. 138, 154, 155 Togawa, O., see Ikeuchi, Y. 155, 157 Togawa, O., see Ito, T. 138, 155 Togawa, O., see Livingston, H.D. 138, 139, 144, 155, 157 Togawa, O., see Povinec, P.P. 138, 142, 148, 149, 154, 155 Toh, Y., see Hatsukawa, Y. 379 Toivonen, H. 439 Tölgyessy, J. 20, 36 Tomaru, H., see Fehn, U. 267 Tomasek, M., see Wilhelmova, L. 96, 97 Tompson, I.M.G. 425–427 Tongiorgi, V., see Cocconi, G. 214 Toolin, L.J., see Donahue, D.J. 65, 242, 245 Top, Z., see Povinec, P.P. 467, 472, 474 Tordoff, B., see Moore, I.D. 337 Török, S. 357, 358, 360, 367 Török, S., see Eriksson, M. 357, 362 Török, S., see Jernström, J. 357, 361, 367 Torres, R. 231 Toyoda, K., see Hatsukawa, Y. 379 Tracy, B., see Kieser, W.E. 318 Trautmann, N. 332, 338 Trautmann, N., see Backe, H. 347 Trautmann, N., see Blaum, K. 299, 345 Trautmann, N., see Eichler, B. 333 Trautmann, N., see Erdmann, N. 339, 349, 358 Trautmann, N., see Grüning, C. 337–339 Trautmann, N., see Huber, G. 332 Trautmann, N., see Kuczewski, B. 311 Trautmann, N., see Maul, J. 334, 348 Trautmann, N., see Müller, P. 337, 346 Trautmann, N., see Nunnemann, M. 338 Trautmann, N., see Passler, G. 337, 338, 341 Trautmann, N., see Peuser, P. 332 Trautmann, N., see Sewtz, M. 347 Trautmann, N., see Waldek, A. 339 Trautmann, N., see Wendt, K. 332, 342, 344 Trautmann, N., see Worden, E.F. 339
Author Index
Troianello, E.J., see Sugihara, T.T. 146 Tschiersch, J. 39 Tschiersch, J., see Bunzl, K. 43 Tsela, S.A., see Lindsay, R. 430 Tseng, C.L., see Chao, J.H. 388, 390, 397, 398 Tsintsadze, D., see Pagava, S. 202 Tsukatani, T., see Yamamoto, M. 150 Tsumune, D. 138, 139, 158 Tsumura, A., see Yamamoto, M. 150, 311 Tukey, J.W. 26, 33, 98, 99 Tukey, J.W., see Hoaglin, D.C. 117 Tuniz, C. 241, 242, 332 Tuniz, C., see Danesi, P.R. 288 Tuniz, C., see Hotchkis, M.A.C. 270, 278, 279, 287 Turcq, B. 252 Turner, J.V., see Burnett, W.C. 467 Turner, J.V., see Taniguchi, M. 467 Turney, C.S.M., see Bird, M.I. 245 Tykva, R. 331 Tyler, A.N. 408, 410–412, 414–417, 419, 420, 424–426, 428–436, 440, 441 Tyler, A.N., see Sanderson, D.C.W. 437 U.S. Nuclear Regulatory Commission, 77 Uchida, S., see Muramatsu, Y. 150, 386 Uchida, S., see Sekine, T. 315 Uchida, S., see Tagami, K. 315 Uchida, T., see Katagiri, M. 450 Ueno, K., see Yamamoto, M. 311 Ueno, T., see Hatsukawa, Y. 379 Ugron, A., see Bouisset, P. 384 United States Environmental Protection Agency, US EPA, 3 UNSCEAR, 137, 143, 144, 155 Urban, F.-J., see Eichler, B. 333 Ure, M. 18 USACE, 18 Usacev, S., see Hlinka, V. 198 Usacev, S., see Povinec, P. 201, 233 USAF Nuclear Safety, 362
Uschida, S. 314 Uschida, S., see Mas, J.L.
519
314, 315
Vais, V. 305, 310 Vais, V., see Tanner, S.D. 310 Vajda, N. 43 Valenta, A., see Paul, M. 280 Valenta, A., see Steier, P. 275, 278, 279 Valenta, A., see Winkler, S. 279 Van den Bergh, P., see Kudryavtsev, Y. 337 Van den Haute, P., see Wagner, G.A. 42 van der Plicht, J. 247, 250 van der Plicht, J., see Hughen, K.A. 249, 250 van der Plicht, J., see Kitagawa, H. 250 van der Plicht, J., see Reimer, P.J. 247, 249, 251 van der Plicht, J., see Stuiver, M. 247– 249 Van Der Pluym, J., see Kozlowski, T. 179 Van Der Schaaf, A., see Kozlowski, T. 179 Van Duppen, P., see Kudryavtsev, Y. 337 van Geel, J., see Erdmann, N. 358 Van Grieken, R., see Ro, C.-U. 363 van Put, P. 450 Van R. Smit, J. 139 Van Roosbroeck, J., see Kudryavtsev, Y. 337 van Soest, M., see Sturchio, N.C. 91 van Weers, A.W., see Betti, M. 359 Vance, D.E. 311 Vandeginste, B.G.M., see Massart, D.L. 75 Vandeweert, E. 334 Vandeweert, E., see Bastiaansen, J. 334 Vandeweert, E., see Erdmann, N. 334, 348, 349 Vandeweert, E., see Lievens, P. 334 Vanhaecke, F. 295 Vanmarcke, H., see Wordel, R. 230 Varela, J., see Nuñez, L.A. 253 Vasconcellos, M.B.A. 401
520
Author Index
Vasconcellos, M.B.A., see Armelin, M.J.A. 401 Vasselli, R., see Hult, M. 235 Veen, A.W.L. 409 Veillette, J.J., see Dyke, A.S. 254 Veirs, D.K., see Conradson, S.D. 366 Veletova, N.K., see Hirose, K. 138, 154, 155 Veletova, N.K., see Ikeuchi, Y. 155, 157 Velleman, P.F. 98, 99 Vengosh, A., see Heumann, K.G. 336 Veres, A., see Sekine, T. 396 Vermeeren, L., see Kudryavtsev, Y. 337 Verney, D., see Sauvage, J. 337 Veron, C., see Sauvage, J. 337 Verplancke, J. 220, 224 Vertes, A., see Bogaerts, A. 334 Vickerman, T.S., see Brown, C.F. 320 Victorova, N., see Sandalls, F.J. 31 Vidal, M., see Wauters, J. 18 Villagrán, C., see Nuñez, L.A. 253 Villar, J.A., see Hubert, Ph. 217 Vincze, L., see Török, S. 357, 358, 360, 367 Vintró, L.L., see McMahon, C.A. 362 Vitman, V.D., see Alkhazov, G.D. 342 Vocke Jr., R.D., see Kelly, W.R. 86, 95 Vockenhuber, C. 279 Vockenhuber, C., see Kraft, S. 272 Vockenhuber, C., see Liechtenstein, V.K. 275 Vockenhuber, C., see Steier, P. 275, 278, 279 Vockenhuber, C., see Winkler, S. 279 Voelker, A.H.L. 249 Vogel, J., see Denk, E. 346 Vogel, J.S., see Bergquist, B.A. 268 Vogel, J.S., see McAninch, J.E. 277 Vogiannis, E., see Anastasis, B.P. 430 Vogl, K., see Winkelmann, I. 439 Vogt, S., see Danesi, P.R. 358 Vogt, S., see Hotchkis, M. 278, 279, 287 Vogt, S., see Paul, M. 344 Voigt, G., see Gering, F. 408, 430
Vojtyla, P. 163, 164, 166, 167, 171, 182, 191, 194, 228 Volchok, H.L., see Bowen, V.T. 137– 139, 146, 155 Volkmer-Ribeiro, C., see Turcq, B. 252 von Hahn, R., see Liechtenstein, V.K. 275 Vonderheide, A.P. 297, 321 Vos, G., see Hubaux, A. 70 Wacker, J.F., see Dai, M.H. 332 Wacker, L. 272–274, 279, 280, 313, 332 Wacker, L., see Suter, M. 272 Wacker, L., see Tims, S.G. 266, 284 Wackernagel, H. 10, 11 Wagner, G.A. 42 Wagner, G.A., see Baumgartner, S. 250 Wahab, A., see Ramlia, A.T. 398 Walczyk, T., see Denk, E. 346 Walczyk, T., see Heumann, K.G. 336 Waldek, A. 339 Waldek, A., see Erdmann, N. 339 Waldek, A., see Grüning, C. 337–339 Waldek, A., see Nunnemann, M. 338 Waldek, A., see Passler, G. 337, 338, 341 Waldek, A., see Wendt, K. 344 Walker, R.C., see Hallett, D.J. 252 Walker, R.L., see Anderson, T.J. 313 Wallace, D., see Kratochvil, B. 35 Wallenius, M. 332 Wallenius, M., see Tamborini, G. 348 Walling, D.E., see Zhang, X.B. 433 Wallner, A., see Hrnecek, E. 279 Wallner, C. 279 Walter, H.K., see Kozlowski, T. 179 Walter, H.K., see Schröder, W.U. 179, 180 Walter, W.W., see Hlavay, J. 18 Walters, B.R.B., see Grasty, R.L. 427, 430 Wang, K., see Hou, X.L. 377, 398 Warden, J.M., see Sjöblom, K.-L. 450, 451, 456 Warneke, T. 150 Warneke, T., see Taylor, R.N. 305
Author Index
Warner, M.J. 138 Warwick, P.E., see Taylor, R.N. 305 Warwick, P.E., see Warneke, T. 150 Wasserburg, G.J., see Chen, J.H. 302, 304 Wasserburg, G.J., see Edwards, R.L. 41 Wasserman, H.J. 314 Wastin, F., see Conradson, S.D. 366 Waterscience, 119 Watt, D.E. 189 Watt, D.E., see Glower, R.N. 210 Watters Jr., R.L., see Kurtz, D.A. 64 Wauters, J. 18 Webb, R.S., see Alley, R.A. 251 Webber, W.R. 213 Weber, M., see Peuser, P. 332 Webster, R. 10, 11 Wedekind, Ch. 449, 450 Wegrzynek, D., see Eriksson, M. 357, 362 Weidele, H., see Vandeweert, E. 334 Weinstein, Y. 467 Weiss, H.V. 147 Weiss, R.F., see Warner, M.J. 138 Weissman, L., see Kudryavtsev, Y. 337 Weisz, M., see Hlavay, J. 18 Weitkamp, T., see Salbu, B. 357, 367 Wells, M., see Margaritz, M. 40 Wells, S.G., see Meyer, G.A. 252 Wellum, R., see Richter, S. 268 Wendt, K. 332, 342, 344 Wendt, K., see Blaum, K. 299, 345 Wendt, K., see Bushaw, B.A. 336 Wendt, K., see Denk, E. 346 Wendt, K., see Grüning, C. 337–339 Wendt, K., see Huber, G. 332 Wendt, K., see Maul, J. 334, 348 Wendt, K., see Moore, I.D. 337 Wendt, K., see Müller, P. 337, 346 Wendt, K., see Schulz, C. 344 Wendt, K., see Wies, K. 342 Wendt, K.D.A., see Lu, Z.-T. 342 Wendt, K.D.A., see Schumann, P.G. 347 Wendt, K.D.A., see Trautmann, N. 332, 338
521
Werner, C.D., see Niese, S. 235 Wershofen, H., see Aumann, D.C. 386, 389, 390 Weyer, S. 296 Weyhenmeyer, C.E., see Hughen, K.A. 249, 250 Weyhenmeyer, C.E., see Reimer, P.J. 247, 249, 251 Weyhenmeyer, C.E., see van der Plicht, J. 247, 250 Wies, K. 342 Wight, S.A., see Zeissler, C.J. 43 Wilde, F.D., see Lane, S.L. 39 Wilenski, H., see Przyborowski, J. 103– 105, 107 Wilford, J.R. 439 Wilhelm, J.G. 39 Wilhelmova, L. 96, 97 Wilhelmova, L., see Csongor, E. 96 Williams, D. 410 Williams, D., see Chamberlain, A.C. 410 Williams, D., see Li, W.-X. 41 Williams, K.J., see Harvey, B.R. 315 Williams, M.L., see Child, D.P. 281, 282 Wilson, G.C., see Kilius, L.R. 267 Wilson, J., see Ketterer, M.E. 302, 332 Wilson, J.E., see Hassig, N.L. 18, 31 Wilson, J.E., see Pulsipher, B.A. 13 Wilson, N. 39 Winefordner, J.D., see Long, G.L. 52, 53 Winkelmann, I. 437, 439 Winkler, S. 279 Winkler, S., see Vockenhuber, C. 279 Wirth, E., see Kammerer, L. 42 Wirth, E., see Uschida, S. 314 WISE Uranium Project, 357 Wisegarver, D.P., see Warner, M.J. 138 Wishnook, J.S., see Hughey, B. 243 Wissman, F. 430 Wissmann, F., see Müller, G. 221 Wobrauschek, P., see Eriksson, M. 357, 362 Wolf, S.F. 322 Wolfe, D.A., see Bohn, L.L. 30, 35 Wollenberg, H., see Løvborg, L. 410
522
Author Index
Wong, H.K.Y., see Hill, D.M. 315 Wong, K.M., see Bowen, V.T. 151 Wood, A.K., see Ramlia, A.T. 398 Wood, L.J., see Kurtz, D.A. 64 Wood, S.H., see Meyer, G.A. 252 Woodhead, D., see Kershaw, P. 459, 462–464 Woodhead, D., see McCartney, M. 459, 462–464 Woodhead, D., see Povinec, P.P. 458, 459 Woodhead, D., see Sjöblom, K.-L. 450, 451, 456 Woodhead, D.S. 459, 461 Woodhead, D.S., see Poole, A.J. 459, 463 Woodruff, J.F., see Remedy, W.R. 19, 22 Woods, C.A., see Steadman, D.W. 253, 254 Woods, M., see Hurtgen, C. 54 Wordel, R. 230 Worden, E.F. 339 Worobiec, A., see Ro, C.-U. 363 Wright, H.T., see Burney, D.A. 253 Wu, T., see Ji, Y.-Q. 311 Wyart, J.-F., see Worden, E.F. 339 Wyse, E., see La Rosa, J. 311 Wyse, E., see Povinec, P.P. 157, 235, 450, 467 Wyse, E.J. 264, 269, 306, 332, 371 Xu, R., see Liao, C.-L.
297
Yagi, M. 396 Yamada, M. 155 Yamagata, N. 139 Yamagata, T., see Yamagata, N. 139 Yamaguchi, Y., see Komura, K. 234 Yamakoshi, K. 210 Yamamoto, M. 150, 311 Yamamoto, M., see Komura, K. 150 Yamasaki, S., see Yamamoto, M. 311 Yamashita, M. 214 Yamato, A., see Fukai, R. 150 Yan, X.J., see Hou, X.L. 398
Yiou, F. 266 Yiou, F., see Frechou, C. 384 Yiou, F., see Gascard, J.C. 266 Yiou, F., see Mironov, V. 267 Yiou, F., see Raisbeck, G.M. 250, 255, 266, 267, 384, 390 Yonezawa, C. 399 Yoshida, S., see Muramatsu, Y. 150, 318 Yoshihara, K., see Sekine, T. 396 Yoshihara, K., see Yagi, M. 396 Youden, W.J. 115 Young, L., see Bailey, K. 91, 92 Young, L., see Chen, C.Y. 91, 342 Young, L., see Du, X. 91, 92 Young, L., see Sturchio, N.C. 91 Yu, Y.X., see Chen, Q.J. 152, 153, 399, 401 Yuita, K., see Yamamoto, M. 311 Zahradnik, P., see Danesi, P.R. 358 Zanotti, L., see Bellotti, E. 233 Zatsepin, G.T., see Kovalchuk, E.L. 230 Zauner, S., see Backe, H. 347 Zawadzki, A.W., see Hill, D.M. 315 Zdesenko, Yu.G. 230 Zeisler, R., see Donohue, D.L. 348 Zeissler, C.J. 43 Zelen, M., see Eisenhart, C. 72, 81 Zemlyanoi, S., see Sauvage, J. 337 Zglinski, A., see Kozlowski, T. 179 Zhang, X.B. 433 Zhao, X.-L., see Kilius, L.R. 267 Zhao, X.-L., see Povinec, P.P. 318 Zhao, X.-L., see Purser, K.H. 289 Zhao, X.L., see Hughey, B.J. 243 Zhao, X.L., see Kieser, W.E. 318 Zhou, Y., see Bird, M.I. 245 Zhou, Z.Q., see Raisbeck, G.M. 266 Zhou, Z.Q., see Yiou, F. 266 Ziegler, J. 211, 212 Ziegler, J.F. 275 Zimmerli, B., see Haldiman, M. 318 Zindler, A., see Bard, E. 249, 253 Zoeger, N., see Eriksson, M. 357, 362 Zombori, P. 417, 427, 428, 430
Author Index
Zoriy, M., see Becker, J.S. 305 Zoriy, M.V. 297, 305, 321, 324, 332 Zoriy, M.V., see Izmer, A.V. 320
Zucconi, L., see Rosamilia, S. 398 Zuzel, G., see Simgen, H. 94 Zvara, I. 163, 198
523
This page intentionally left blank
525
Subject Index
7 Be
94, 164, 466 234, 241, 243, 244, 250, 254–257 14 C 5, 6, 17, 41, 43, 83–85, 88, 89, 130, 164, 214, 228, 229, 231, 234, 241–244, 246–255, 257, 263, 467 26 Al 210, 234, 241, 243, 244, 254, 257, 303 39 Ar 40, 41, 120, 121, 229, 343 41 Ca 241, 331, 344–346, 350 60 Co 6, 8, 75–77, 124, 167, 196, 218, 235, 393, 440, 451, 453, 458, 459 85 Kr 84, 91–99, 343 90 Sr 2, 17, 36, 42, 138, 139, 141, 142, 146–149, 157, 297, 320, 321, 331, 344, 350 99 Tc 263, 267, 268, 313–315, 317, 318, 371, 377, 380, 382, 383, 394–396 129 I 42, 84, 241, 243, 244, 257, 263, 266, 267, 296, 300, 303, 318–320, 371, 376, 377, 380, 382– 386, 388–390, 401 135 Cs 320, 321, 371, 377, 383, 396–398, 401 137 Cs 2, 9, 10, 18, 137–146, 148, 149, 154, 157, 168, 170, 204, 205, 235, 288–290, 320, 321, 396–398, 409–411, 414, 415, 417, 418, 420, 425, 426, 428–434, 436, 437, 440–442, 451, 453, 455–466, 468, 475 210 Pb 169, 170, 194, 219, 228, 322, 412, 436, 467 210 Po 169, 322, 467 220 Rn 171, 219, 467, 468 222 Rn 171, 219, 229, 323, 437, 453, 467, 468, 470– 475 226 Ra 170, 171, 216, 218, 219, 266, 323, 324, 372, 430, 436, 453, 467, 468, 473 228 Ra 263, 266, 277, 323, 467, 468, 472, 473 234 U 311, 324, 358, 360 235 U 268, 272, 275, 276, 281, 283, 284, 289, 311, 313, 314, 318, 320, 324, 325, 358–360, 371, 372, 382, 385, 386, 390, 396, 398, 399, 401 10 Be
236 U
263–265, 268–270, 272, 275–277, 279, 280, 283, 285, 287–290, 300, 311, 324, 347, 358– 360, 368 237 Np 263, 265, 277, 281, 287, 310–313, 371, 372, 377, 380, 383, 384, 390, 392, 393, 399, 401 238 U 36, 168, 170, 171, 216, 217, 219, 265, 266, 268, 269, 272, 275–277, 279–285, 287, 289, 290, 300, 301, 303, 304, 307–310, 312, 313, 318, 324, 325, 348, 358–360, 371, 372, 377, 380, 383, 384, 390, 398, 399, 401, 428, 436, 474 239 Pu 26, 149–151, 264, 265, 269, 270, 273, 277, 281–283, 285, 286, 289, 290, 296, 300–303, 305, 306, 320, 331, 332, 338, 339 239,240 Pu 36, 138, 142, 143, 148, 150, 151, 154– 157, 265, 458, 459, 463 240 Pu 36, 138, 142, 143, 148–151, 154–157, 264, 265, 269, 270, 273, 274, 277, 281–283, 285, 286, 301, 302, 306, 312, 331, 338, 339, 458, 459, 463 241 Pu 149, 150, 301, 306, 338, 348, 363 242 Pu 152, 153, 265, 270, 273, 277, 283–285, 301, 311, 312, 338, 372, 383 244 Pu 263, 265, 287, 288, 331, 339, 350 241 Am 36, 149, 224, 287, 301, 322, 348, 363, 412, 421, 430, 431, 437, 441, 463 Abrosimov bay 457 accelerator mass spectrometry (AMS) 41, 88, 241, 263, 303, 332 activation analysis 2, 67, 125, 235, 371, 372, 375, 377, 383, 384, 395, 396, 398, 401 activation method 380 active veto shielding 194 activity 3, 5–8, 11, 17, 19–21, 30, 31, 35, 36, 38, 39, 43, 130, 147–151, 168, 169, 194, 216–219, 228–231, 266, 267, 287, 312–315, 363, 373, 374, 382–385, 397, 408–414, 416, 417, 424–
526
Subject Index
428, 430, 431, 433–436, 442, 443, 465, 466, 470–475 activity concentration 17, 93, 116, 130, 219, 416, 470–472, 474, 475 ages 41, 248, 250–252, 255 airborne gamma-ray spectrometry (AGS) 407–410, 416, 418, 420, 424, 425, 437–439, 441–443 alpha counting 322 alpha-ray spectrometry 474 aluminum-26 210, 234, 241, 243, 244, 254, 257, 303 americium-241 36, 149, 224, 287, 301, 322, 348, 363, 412, 421, 430, 431, 437, 441, 463 ammonium molybdophosphate (AMP) 139–141, 162, 317, 397 analyte 20–22, 29, 32, 33, 36, 39, 54, 55, 74, 81, 83, 85, 296–300, 307, 309, 311, 314, 317, 333, 334, 336 analytical measurement 242, 49–53, 64, 88, 109, 110, 114, 241, 314, 345, 371, 380, 386 method 23, 66, 139, 141, 146, 149, 157, 296, 314, 371, 380–383, 396 anion exchange chromatography 393 anthropogenic radioactivity 1, 137, 138, 140, 157, 235, 263, 318, 331, 355, 410, 430, 435, 464, 466 radionuclides 1, 137, 143, 144, 157, 235, 331, 347, 355, 410, 411, 426, 430, 435, 440, 450, 456, 457, 464, 466, 467 anti-Compton spectrometer 164, 166, 191, 192, 198, 204 anticosmic shielding 164, 166, 167, 175, 194, 199– 202, 204, 205 aquifer 470 argon-39 40, 41, 120, 121, 229, 343 ashing 376 assessment of the blank 85 Atlantic Ocean 454 atom source 333 atmosphere 18, 31, 38, 93, 97, 110, 164, 165, 168, 173, 210–215, 234, 247–249, 255, 256, 318, 319, 356, 394, 416, 454 autoradiography 43, 363 background simulation 172, 174, 184 spectrum 127, 203, 204, 220, 224 reduction 94, 167, 192, 193, 201, 210 barometric effect 175 beryllium -7 94, 164, 466 -10 234, 241, 243, 244, 250, 254–257
beta counting 120, 121, 296, 317, 320, 322, 396, 397 Bikini atoll 150 bismuth-209 314, 322 bismuth germanate (BGO) detector 168, 420 blanks 23, 37, 74, 81, 83–85, 89, 90, 95, 96, 282, 302, 313 bottom sediment 459 Brazil 467, 470–473, 475, 476 building material 216 cesium -135 320, 321, 371, 377, 383, 396–398, 401 -137 2, 9, 10, 18, 137–146, 148, 149, 154, 157, 168, 170, 204, 205, 235, 288–290, 320, 321, 396–398, 409–411, 414, 415, 417, 418, 420, 425, 426, 428–434, 436, 437, 440–442, 451, 453, 455–466, 468, 475 analysis 139, 157 in seawater 138, 139, 157, 205 calibration 54, 55, 67, 68, 70–72, 74–76, 84, 87, 114, 121, 124, 130, 236, 247–251, 255, 301, 304, 305, 381, 409, 410, 419, 421, 424–427, 431, 435, 436, 439–441, 452, 453, 455, 458, 460, 461, 468 carbon-14 age 2482, 250, 252 calibration 84, 247 dating 84, 241, 247, 253 cation exchange chromatography 397 certified reference materials (CRM) 382, 401 chemical recovery 2672, 379, 380, 398, 399 separation 54, 150, 153, 157, 300, 312–315, 317, 323, 333, 345, 376, 379, 380, 382, 385, 386, 388–390, 393, 398–401 Chernobyl accident 19, 168, 267, 301, 339, 341, 410, 420, 439 fallout 9, 10, 12, 31, 142, 148, 170, 267, 301, 338, 339, 410, 435, 437, 440, 441 coastal area 470, 475 sediment 431 water 467, 470, 473 zone 467 cobalt-60 6, 8, 75–77, 124, 167, 196, 218, 235, 393, 440, 451, 453, 458, 459 coincidence–anticoincidence spectrometry 196 collaboration of European low-level underground laboratories (CELLAR) 202, 203, 232, 233 Compton suppression 193, 203–205 confidence interval 25, 34, 61–63, 70, 91, 114, 129 corals 241, 247, 249–251, 257
Subject Index cosmic particles electrons 164, 165, 175, 176, 178–182, 194, 201, 211, 214, 215, 222, 225, 226, 265, 270, 275, 297, 334, 378 muons 109, 164–168, 173–180, 182–186, 189, 194, 201, 209–211, 213–215, 220–223, 225, 226, 229, 236 neutrons 140, 163–169, 171, 174, 179, 182, 209, 211–215, 220–226, 229, 230, 235, 236, 289, 310, 318, 356, 371–373, 377, 378, 380, 383, 385, 393, 401 pions 164, 165, 180, 211–214, 226 protons 164, 165, 211–216, 220–223, 226, 372 cosmic ray background 167 cosmic rays 121, 163–165, 167, 169, 172, 174, 194, 201, 209–213, 215, 216, 220–223, 226, 228, 229, 233, 236 cosmogenic radionuclides 164, 168, 170, 228, 234, 254, 255 counting efficiency 67, 88, 89, 93, 94, 130, 148, 163, 190, 196, 201, 224, 269, 301, 332, 374, 381, 384, 395, 397, 434 statistics 61, 75, 92, 110, 120, 301, 358, 393, 407, 428 data evaluation 476 data reporting 121 database 117, 119, 142, 144, 157, 248, 249, 310 dendrochronology 247 depleted uranium 268, 287, 288, 290, 356–359, 368, 393 penetrator 358, 367 detection decision 232, 51–56, 58, 69, 71, 75, 79, 80, 108, 130 limit 19, 21, 23, 49, 51–57, 60, 62, 65, 67, 69– 71, 73, 75, 77–80, 83, 86, 91, 93, 94, 100, 101, 105, 106, 130, 317–321, 331, 332, 347, 349, 350, 374, 376, 377, 379, 388–390, 392–394, 396–399 detector background 2, 75, 76, 163–172, 176, 182–185, 187–205, 209, 210, 218, 220–226, 228, 229, 231–235, 265, 271, 281–283, 395, 439, 452, 468 resolution 167, 195, 196, 271–273, 275, 279, 281, 283, 302, 304, 338, 339, 378, 420, 421, 437, 451 shielding 164–171, 182, 185, 187, 192, 194, 196, 197, 199–205, 209, 210, 218, 220, 222, 224, 225, 228, 229, 232, 377, 421, 427 disintegration 373 diurnal variation 174
527
Doppler shift 333–345 Dowex 152, 153 efficiency 6, 38, 67, 88, 89, 93, 94, 130, 140, 148, 163, 165, 166, 168, 170, 186–188, 190–192, 194–196, 198, 200, 201, 204, 231, 269, 270, 280, 285, 297–299, 307, 309, 332–334, 336, 337, 364, 437, 450, 451 electrodeposition 153, 154 electron 2, 165, 175–182, 214, 215, 225, 226, 268, 334, 338, 358, 361–363, 368, 378 electron multiplier 268 electronics 166, 167, 195, 197–199, 273, 421–423, 437, 439, 451, 452, 454, 468 electrophoretic technique 376 energy resolution 171, 271, 378, 449, 453 energy-dispersive X-ray spectrometry (EDX) 362 Enewetak atoll 139, 143, 148, 157 environment 1–3, 7, 10, 14, 17, 18, 21, 22, 28, 43, 49, 137, 138, 149, 150, 157, 158, 163, 229, 235, 236, 266, 287, 310, 311, 313, 331, 347, 348, 355–357, 366–368, 407, 410, 427, 449– 451, 464, 466–468, 475, 476 environmental analysis 2, 5, 7, 13, 17, 19, 21, 22, 33, 35–37, 77, 95, 108, 150, 157, 194, 216, 257, 273, 301, 306, 307, 313, 318, 319, 321–323, 362, 371, 372, 374–376, 384, 385, 389, 390, 396–398, 449 radioactivity 1, 2, 9, 17, 19, 22, 37, 77, 96, 97, 108, 117, 137, 142, 157, 164, 168, 171, 216, 230, 263, 318, 323, 331, 339, 371, 372, 385, 386, 397, 407–410, 435, 443, 449 radionuclides 1–3, 9, 14, 17, 18, 21, 36, 77, 79, 137, 151, 157, 168, 171, 194, 263, 273, 291, 321, 331, 355–357, 371, 372, 374–376, 379, 384, 385, 397, 398, 401, 410, 449, 459, 464 samples 2, 3, 5, 9, 17, 19–22, 29, 31–33, 35–37, 43, 93, 142, 150, 151, 157, 257, 290, 301, 306, 313, 316, 321, 322, 338, 339, 362, 367, 371, 374–376, 384, 385, 387, 388, 393, 394, 396– 398, 410 ETV technique 316, 319, 321 European Organization for Nuclear Research (CERN) 163, 182, 204, 232 extinct megafauna 253 extraction chromatography 147, 312, 392–394, 399 factor of merit 163, 175, 453 fallout 9, 10, 12, 31, 42, 137, 138, 142–144, 146, 148–151, 155, 170, 267, 273, 301, 316, 338, 339, 342, 410, 434, 435, 437, 440, 441 Fangataufa atoll 458, 475 fg-range 333
528
Subject Index
filter 38, 77, 83, 90, 126, 142, 147, 151, 153, 265, 271, 272, 277, 281, 282, 297, 303, 319, 324, 337, 343, 422 filtration 37, 39, 142, 153 forest fires 252 fractionation 18, 39, 356, 376 fresh water 449, 450, 453, 467, 468, 474 gamma-ray spectrometry 108, 125, 194, 205, 407, 427, 433, 437, 449–451, 464, 468, 471, 475 gas counters 1, 109, 228, 232 Gaussian distribution 175, 378 Ge detector 163, 166, 168, 182, 184, 189, 224, 225 GEANT 175–182, 185, 186, 204 geostatistical sampling 29 global fallout 137, 138, 143, 144, 146, 148–150, 155, 273, 342 graphite sample 243, 245, 283, 284 gravimetry 380 groundwater 39, 40, 91, 290, 343, 451, 466–468, 473–475 guard detector 196 hard component of cosmic rays 165, 194, 220 heteroscedasticity 58, 59, 63, 69, 120 homogenization 19–22, 40 homoscedasticity 69–71 hot particles 19, 31, 43, 312, 334, 348, 356, 361, 363, 375 HPGe detector 164, 165, 171, 172, 186, 187, 189, 190, 192, 194, 196, 198, 200, 203–205, 228, 393, 395, 397, 399, 418, 430, 433, 451, 456 IAEA -414 205 -MEL 142, 148, 154, 194, 200, 202, 203, 228, 233, 449, 451–454, 456, 457, 459, 460, 464, 466, 475 -reference materials 414 inductively coupled plasma mass spectrometry (ICPMS) 1, 2, 157, 322 interactions of electrons 176, 180, 201, 378 muons 176, 187, 201, 221, 223, 228 neutrons 169, 221, 223, 378 photons 180 positrons 180, 201 in situ gamma-ray spectrometry (IGS) 407–410, 412, 413, 416, 418, 421, 423, 424, 426, 427, 430, 431, 433–436, 441–443 intercomparison exercise 116, 441 interferences 54, 64, 121, 127, 196, 264, 267, 269, 283, 295–300, 302, 303, 305, 306, 310, 312– 318, 320, 322–324, 332, 333, 336, 348, 349,
377, 379, 381, 382, 385, 386, 388–390, 394, 395, 397, 398 Intergovernmental Oceanographic Commission (IOC) 467 International Atomic Energy Agency (IAEA) 50, 75, 100, 348, 449 Committee for Radiological Units (ICRU) 14 Committee for Radionuclide Metrology (ICRM) 108, 133 Standardization Organization (ISO) 49 Union of Pure and Applied Chemistry (IUPAC) 49 international harmonization 52, 53 intrinsic radioactivity 169, 235 iodine -125 17, 380, 388 -129 42, 84, 241, 243, 244, 257, 263, 266, 267, 296, 300, 303, 318–320, 371, 376, 377, 380, 382–386, 388–390, 401 ion cyclotron resonance (ICR) 338 ion detection 333, 337, 338 ion source 243, 264, 267, 269, 270, 279–281, 283– 285, 295, 302, 338, 340–342, 347, 348 ionization chamber 268, 271–273, 275, 277, 279, 282 Irish Sea 142, 148, 154, 205, 267, 312, 359, 361, 368, 430, 431, 456, 459, 460, 463–465, 475, 476 irradiation 67, 182, 373, 376–382, 385–390, 392– 399, 401 isobaric interference 321, 324 isotopic selectivity 91, 331–333, 336, 337, 342, 344 Kara Sea 286, 456 Kerma 425–427, 430, 436, 437, 440, 442 Khyshtym accident 285 krypton-85 84, 91–99, 343 lake sediment 250 lakes 39, 42, 230, 450 laser cw 333, 334, 336, 337, 342, 344, 345 Nd:YAG 337, 339, 340, 349 pulsed 333, 336–340, 343, 348, 349 pumped 337, 340, 349 Ti–Sa 341 laser ablation microprobe mass analysis (LAMMA) 334 lead -210 169, 170, 194, 219, 228, 322, 412, 436, 467 dating 228
Subject Index shield 164, 165, 169, 171, 172, 179, 182, 186– 189, 194–196, 201, 204, 205, 217, 222–224, 226, 228, 231, 236, 357, 418 liquid chromatography 376 liquid scintillation counting (LSC) 67, 150, 160, 233, 301, 396 liquid scintillation spectrometry (LSS) 39, 150, 228, 234 low-level counting 2, 54, 61, 65, 84, 88, 91, 93, 100, 103, 104, 106–112, 119–121, 125, 130, 150, 163– 165, 170, 171, 189, 209, 224, 226, 227, 230, 232, 233, 236, 301, 393, 397 gamma-spectrometry 234 Ge spectrometry 163 spectrometry 88, 95, 108, 125, 130, 194, 231, 234, 301, 323, 359 lyophilization 376 mapping of radionuclides 457 marine radioactivity 1, 2, 138, 142, 154, 158, 235, 449, 475 Marinelli beaker 40, 223 Marsh 287, 417, 426, 430, 431 mass analyzer 295, 307, 337 mass spectrometry 1, 2, 41, 65, 67, 74, 88, 95, 150, 205, 241–243, 249, 251, 253, 257, 263, 264, 267, 268, 290, 297, 301, 322–324, 331–337, 343–345, 347–349 massic activity 17, 20, 21, 169, 216, 218, 416, 417 Mediterranean Sea 154 meteorites 210, 228, 234 Monte Carlo simulation 163, 165, 167, 169, 171, 173, 175, 177, 179, 181–183, 185, 187, 189, 191, 193–195, 197, 199, 201, 203–205, 416, 420, 427 multichannel analyzer (MCA) 363, 380 multicomponent detection 74, 75 multi-dimensional spectrometry 164 multiparameter spectrum 379 multivariate detection 74, 79–82 Mururoa atoll 339, 341, 364 NaI(Tl) detector 194, 196, 198, 409, 410, 418, 419, 421, 422, 427, 439, 451–453, 468 natrium -22 196 -24 385, 393, 394, 398, 399 natural background 37, 91, 150, 199, 201, 202, 204, 209, 216, 229, 266, 314 radioactivity 1, 17, 22, 37, 216, 221, 223, 225, 229, 231, 235, 331, 355, 410, 430, 435, 460
529
radionuclides 1, 3, 5, 14, 17, 199, 202, 204, 223, 235, 283, 287, 321, 331, 355, 398, 417, 420, 426, 430, 435, 440, 450, 451, 456, 460, 466, 467 nebulizer 305, 307–309, 316, 323 neutron activation analysis (NAA) 2, 371, 372 neutrons 140, 163–169, 171, 174, 179, 182, 209, 211–215, 220–226, 229, 230, 235, 236, 289, 310, 318, 356, 371–373, 377, 378, 380, 383, 385, 393, 401 normal approximation 60–64, 66, 101, 104, 107, 120, 125 nuclear energy 50, 75, 109, 124, 163, 165, 179, 180, 182, 209, 211–214, 225–227, 232, 265, 278, 313, 348, 362, 372, 377, 384, 396, 428 explosion 212, 235, 285, 356, 366 reaction 231, 310, 372, 377, 382, 383, 385, 390, 395, 396 reactor 31, 287, 339, 346, 355, 371, 372, 376, 377, 386, 396, 401 test 58, 75, 82, 91, 107, 126, 144, 150, 227, 286, 339, 366, 410, 475 waste 117, 285, 286, 355, 356, 359, 371, 372, 396, 397, 450, 459, 475 weapon 31, 137, 155, 286 optical excitation 332–334, 337, 343 Pacific Ocean 139, 143, 148, 149, 156 paleoclimate studies 251 partitioning 114, 115 passive shield 204, 205 peak/Compton ratio 195, 196 Pelletron accelerator 242, 257, 271 photon fluence 409, 411–413, 426 plasma 1, 150, 212, 241, 245, 246, 268, 295–300, 302, 303, 306–310, 312, 314, 316, 318, 319, 321–323, 332 plastic scintillation detector 196 plastic scintillator 224, 234 plutonium 137–139, 144, 149–151, 153, 155, 263, 265, 266, 268–272, 275, 277, 279–281, 283– 290, 301, 305, 307, 310–312, 316, 331, 332, 334, 338, 339, 341, 350, 357, 361, 363–367 analysis 139, 150, 268–270, 301, 307, 310–312, 331, 332, 334, 339, 341 in water 138, 285, 312 plutonium isotopes 238 Pu 149–151, 156, 301, 338, 348, 350, 356 239 Pu 26, 149–151, 264, 265, 269, 270, 273, 277, 281–283, 285, 286, 289, 290, 296, 300– 303, 305, 306, 320, 331, 332, 338, 339
530
Subject Index
239,240 Pu
36, 138, 142, 143, 148, 150, 151, 154– 157, 265, 458, 459, 463 240 Pu 36, 138, 142, 143, 148–151, 154–157, 264, 265, 269, 270, 273, 274, 277, 281–283, 285, 286, 301, 302, 306, 312, 331, 338, 339, 458, 459, 463 241 Pu 149, 150, 301, 306, 338, 348, 363 242 Pu 152, 153, 265, 270, 273, 277, 283–285, 301, 311, 312, 338, 372, 383 Poisson distribution 54, 60, 63, 65, 100, 101, 108, 109, 111, 125 polonium-210 169, 322, 467 polyatomic interferences 299, 305, 306, 312, 318, 323 post-irradiation separation 379, 382, 385, 388, 392–397, 399, 401 potassium-40 75–77, 125–131, 166, 168–171, 201, 202, 204, 216–219, 229, 410, 418, 420, 428, 430, 434–437, 440, 453, 456, 460, 461, 465– 468 precipitation 39, 142, 147, 148, 151–153, 157, 255, 376, 390, 392, 394, 395, 397, 399 pre-concentration of 2, 147, 151, 376, 385, 399 strontium 147 plutonium 151 preparation of samples 35, 89, 266, 268, 284, 376, 382 primary cosmic rays 172, 174 processes with muons 175, 177, 178, 182 proportional counter 222, 232, 364 pulse shape analysis 150 quadrupole 295, 297–300, 302, 305, 312, 322, 324, 337, 338, 343–345 quality assurance 18, 21, 33, 36, 75, 381 quantification limit 55, 70, 89, 93, 94 radioactive particles 355, 357, 359, 361–363, 365, 367, 368 radioactivity of building materials 2162, 226 Ge detectors 168 lead shield 236 radioanalytical methods 37 radiocesium, see cesium-137 radiocarbon, see carbon-14 radiochemical analysis 1, 17, 19, 37, 157, 313, 317, 320, 384, 395, 398 separation 1, 146, 147, 152, 153, 157, 313, 317, 320, 379, 380, 393, 395, 398, 399 radioecological studies 355, 357, 368 radiofrequency quadrupole analyzer (RFQ) 337
radiogenic radionuclides 234 radiometrics 10, 17, 267, 311, 317, 394, 401 radionuclide residence time 138 tracer 241, 379 radium -226 170, 171, 216, 218, 219, 266, 323, 324, 372, 430, 436, 453, 467, 468, 473 -228 263, 266, 277, 323, 467, 468, 472, 473 radon -220 171, 219, 467, 468 -222 171, 219, 229, 323, 437, 453, 467, 468, 470–475 daughters 323, 466, 467, 471 in the air 2, 219, 229, 231 random sampling 4, 6–8, 11, 12, 18, 21, 22, 24, 27– 29, 32 ratio 228 Ra/226 Ra 226 236 U/238 U 265, 272, 275, 276, 285, 287, 290, 324, 360 240 Pu/239 Pu 150, 151, 264, 269, 285, 286, 301, 339 reference material 52, 75, 124, 205, 315, 322, 339, 418 method 29, 37, 255, 381, 382, 401 relative efficiency 186–188, 190, 194–196, 198, 200, 221, 224, 231, 418, 437, 439, 451 remobilization of radionuclides 459 reprocessing plant 287, 318, 357, 359, 441, 459, 464, 475 residence time 138, 155 resonance ionization mass spectrometry (RIMS) 331, 332, 348 safeguards 263, 270, 287, 290, 348, 475 salinity 453, 466, 468–473, 475 sampling concepts 3 homogenization 19, 21, 22, 40 objectives 3, 5, 11, 22, 23, 32, 43 optimization 1, 23 plan 13, 24, 25, 28–32, 37, 424, 425 strategy 6, 7, 14, 22, 40, 424 techniques 2, 10, 12, 17, 19, 21, 23, 25, 27, 29– 33, 35, 37, 39–41, 43, 147, 296, 376, 435 sampling of aerosols 37 air 39 biota 43 sediment 43 water 39, 140 scanning electron microscopy (SEM) 2, 362
Subject Index scavenging 155, 284, 461, 466 scintillation detector 196, 204 scintillator 94, 224, 234 sea 139, 140, 142, 143, 145, 148, 154, 155, 157, 164–166, 168, 169, 172–174, 182, 185, 187, 191–193, 205, 210, 211, 213–215, 229, 230, 253, 255, 267, 286, 289, 359, 361, 362, 366, 430, 431, 456, 459, 460, 463–468, 474–476 seabed gamma-ray spectrometry 463 secondary cosmic rays 201 secondary ionization mass spectrometry (SIMS) sediment 1, 6–8, 18, 35, 42, 43, 131, 148, 155, 205, 248–250, 254, 255, 285, 288–290, 301, 302, 312, 339, 341, 361–364, 366, 368, 375, 376, 385, 386, 391, 393, 397, 410, 430–432, 443, 456–463, 475 sediment transport 254, 288, 290, 461 Sellafield 143, 148, 154, 266, 267, 287, 318, 359, 441, 456, 459, 463, 464, 475 shield 163–169, 171, 172, 176, 179, 180, 182–184, 186–189, 191, 194–196, 201, 204, 205, 209, 217, 219, 222–224, 226, 228, 229, 231, 236, 296, 302, 307, 346, 357, 418 Sicily 467–471, 474–476 simulation code 163, 180, 184 soft component of cosmic rays 165, 226, 236 soil 4, 9–11, 18, 28, 29, 35, 37, 40, 41, 138, 148, 150, 168, 220, 228, 246, 254, 288–290, 301, 312, 358, 359, 367, 375, 376, 385, 386, 397, 398, 407, 408, 410–413, 416, 421, 426–431, 433–436, 442, 443 soil erosion 288, 433, 434, 443 speciation 19–22, 32, 150, 357, 376, 389 spectral processing 418 speleothems 241, 250 stalagmites 250, 257 statistics of measurement 92 sampling 33 Stepovovo bay 456, 457 stratified sampling 6, 7, 12, 21, 32, 39, 40, 43 strontium-90 analysis 157 in water 138, 147 submarine groundwater discharge (SGD) 467 systematic sampling 4, 7, 8, 11, 12, 28, 37 tandem accelerator 270 temporal variations of cosmic rays 174, 212 TEVA resin 153, 267, 284, 312, 315, 394, 399 thenoyltrifluoroacetone (TTA) extraction 392, 393 thermal ionization mass spectrometry (TIMS) 41, 301, 332
531
Th-series 171, 436, 466 time-of-flight mass spectrometry (TOF-MS) 338, 339, 349 time-of-flight method (TOF) 272, 275–277, 280, 295, 338, 340 titration 380 trace analysis 91, 95, 331, 333, 335, 337, 339, 341– 345, 347, 349 traceability 36, 116, 410, 441 tracer 1, 2, 98, 137, 138, 140, 152, 153, 157, 234, 241, 245, 257, 267, 287, 288, 290, 301, 311– 315, 318, 339, 343, 379, 380, 388, 393, 396, 398, 399, 467, 468, 474 transuranics 392, 458 TRU resin 153 uncertainty of analysis 21 of results 61, 62 underground laboratory 94, 193, 194, 199, 200, 202, 205, 206, 210, 211, 216, 219, 225, 228– 233, 235, 236 underwater gamma-ray spectrometry 449, 450, 467, 468, 471, 475 United Nations Educational, Scientific and Cultural Organization (UNESCO) 467 uranium -234 311, 324, 358, 360 -235 268, 272, 275, 276, 281, 283, 284, 289, 311, 313, 314, 318, 320, 324, 325, 358–360, 371, 372, 382, 385, 386, 390, 396, 398, 399, 401 -236 263–265, 268–270, 272, 275–277, 279, 280, 283, 285, 287–290, 300, 311, 324, 347, 358–360, 368 -238 36, 168, 170, 171, 216, 217, 219, 265, 266, 268, 269, 272, 275–277, 279–285, 287, 289, 290, 300, 301, 303, 304, 307–310, 312, 313, 318, 324, 325, 348, 358–360, 371, 372, 377, 380, 383, 384, 390, 398, 399, 401, 428, 436, 474 uranium particles 348, 357–359 UTEVA resin 284, 393 variance function 88, 121, 123 variations of cosmic rays 212 varved sediments 247–249, 251 veto shielding 194 water column 40, 139, 146, 148, 149, 151, 155, 392, 399 equivalent 140, 210, 226, 473 weapons fallout 286, 338, 410, 434, 437, 440, 441
532
Subject Index
Wien filter 271, 272, 277, 281, 282 Windscale 356, 410, 459 X-ray fluorescence microscopy (XRF) 361, 362, 364, 365, 368 X-ray spectrometry 319, 368
yield
2, 355, 358,
36, 153, 279, 380, 458 yttrium-90
95, 137, 139, 140, 142, 147, 150, 221, 223, 224, 230, 233, 265, 266, 280, 297, 313, 315, 322, 339, 377, 388, 393, 396, 398, 399, 407, 428, 146–148, 320
152, 271, 379, 435,