|
1. |
Geoanalysis: Past, Present and Future† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1179-1186
Philip J. Potts,
Preview
|
|
摘要:
Geoanalysis: Past, Present and Future† Philip J. Potts‡ Department of Geology, The Australian National University, Canberra, ACT 0200, Australia A review of trends in the use of techniques for the chemical analysis of geological samples has been undertaken by abstracting details from reference material characterisation programmes published over the period 1980–1997 and from papers published in Geochimica et Cosmochimica Acta over a similar period. An analysis of these details shows that whereas during the 1980s, XRF and INAA were the two most popular techniques chosen for the major and trace element analysis of silicate rocks, this role continues to be filled by XRF but now complemented by ICP-MS. An equivalent survey of the geochemical use of isotope measurements shows that improvements in the sensitivity and precision of mass spectrometry instrumentation has led to new analytical capabilities: for example, in the application to the Sm–Nd and U–Pb isotope systems in the 1980s and to Re–Os in the 1990s. If these instrument developments continue, the measurement of new isotope systems may become possible including, for example, those related to nucleosynthetic processes.However, it is argued that the greatest influence in future years will arise from the further development of microprobe techniques. Already some of these techniques have demonstrated the importance of providing data that takes into account zoning effects in individual minerals, as in the U–Pb dating of zircon by ion probe and the dating of selected minerals using the laser Ar–Ar technique.Indeed, in these applications, equivalent determinations on bulk samples are now largely discredited, because they mask variations at the mineralogical level. It is likely, therefore, that with further developments in microprobe instrumentation and refinements in geochemical models for interpreting microanalysis data, this area of endeavour will play an increasingly important role in future geochemical studies.Keywords: Trends; future developments; analytical techniques; geochemistry; isotope techniques; microprobe techniques Of all the sciences supported by analytical measurements, the modern day demands of geochemistry are more stringent than those of almost any other application. Not only is there a requirement for the routine determination of all the major elements, and a range of trace elements, that are routinely used in geochemical modelling, but there are also geochemical applications for almost every other element in the periodic table.Furthermore, important branches of geochemistry are based on the determination of isotope ratios derived from the decay of naturally occurring radioactive isotopes (e.g., Rb–Sr, Sm–Nd, Re–Os, U–Pb, Pb–Pb), stable isotopes (particularly of the elements H, N, C, O, S), the noble gases and naturally occurring artificial isotopes (such as 10Be, 14C, 26Al, 36Cl, 129I), which are formed by cosmic ray interactions in the upper atmosphere and can be used to model the rates of more recent geochemical processes. In addition, recent advances in microbeam techniques now permit many major, trace and isotope ratio measurements to be made on individual minerals to spatial resolutions in the range 1–50 mm.What distinguishes geochemical research is not only the wide range of analytical measurements required, but also the high degree of performance demanded (in terms of precision, accuracy and low detection limits).Furthermore, there are few other areas of research where important applications exist for all these measurements using microbeam instrumentation. This present-day degree of sophistication has not always been the case. Fifty years ago, routine analytical techniques were based mainly on chemical separation procedures followed by a gravimetric determination, the so-called ‘classical’ method of analysis.From any cursory examination of the subsequent development of geochemical research it is apparent that advances in analytical instrumentation have had a substantial impact on its development but that equally the requirements of geochemical research have had a significant effect on the development of geoanalytical techniques. One example of this synergy is the development of rare earth element geochemistry. In 1924, Goldschmidt and Thomassen1 published rare earth element concentration data plotted as a function of atomic number.However, it was not until the 1970s that REE geochemistry achieved prominence. This development arose because of: (i) the development of high-purity germanium detectors, which facilitated the routine determination of selected REEs to low detection limits by neutron activation analysis; (ii) the development of a geochemical framework within which to interpret REE data, especially the convention of normalising REE concentrations to chondrite abundances (Masuda2 and others) to overcome the ‘saw-tooth’ effect when un-normalised concentration data are plotted as a function of atomic number; (iii) the realisation that REE patterns could shed light on geochemical processes, such as those associated with the then new ideas of plate tectonics.Arguably, the substantial impact of REE geochemistry required the coincidence of all three of these developments. Accepting the importance of analytical technology on progress in geochemistry, the purpose of this paper is to review trends in the development of analytical techniques over the last 50 years and to use these data as a basis for predicting which techniques are likely to make the most significant contribution to future geochemical research.This evaluation has, in part, been based on an earlier review (Potts et al.3). However, one of the aims of the present paper is to show that more quantitative predictions can be made of future trends, based on an evaluation of the use of techniques over the last two decades.In developing this approach, it soon became apparent more conclusions depend very much on the perspective adopted by the reviewer. Several papers covering developments in atomic spectrometry techniques have appeared in recent years. For example, Boumans4 evaluated trends in a range of modern analytical spectrochemical techniques and included the significant contemporary use of glow-discharge mass spectrometry and dc-arc atomic emission spectrometry. Whereas it is true that the former has had an influential role in the characterisation of † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.‡ On study leave from: Department of Earth Sciences, The Open University, Walton Hall, Milton Keynes, UK MK7 6AA. Analyst, November 1997, Vol. 122 (1179–1186) 1179semiconductor materials, and the latter in the metallurgical industry, neither currently plays a significant role in geochemical research.Hall5 recently reviewed trends in techniques that have influenced, to a significant extent, developments in geochemical exploration research, placing particular emphasis on the contribution of atomic absorption spectrometry and ICP techniques. The present paper will show that this emphasis is not supported to the same extent by the contribution these techniques have made to pure geochemical research.Hiefje,6 in a well known review of atomic absorption spectrometry, undertook an assessment of trends in research publications in that topic. When extrapolated to predict publication rates in future years, the trend in these data could be interpreted as demonstrating that the AAS technique would die out by the end of the century, indicating the misunderstandings and pitfalls that can arise when attempting to predict the future.Since the present review is based on an analysis of trends in the use of analytical techniques that are employed in pure geochemical research applications, it must be emphasised that any conclusions are not necessarily directly applicable to other fields of geochemical research. Trends in Geochemical Techniques Bulk Analytical Techniques When considering the contribution made by bulk geochemical techniques, it is relatively easy to identify a series of publications that include details of contemporary analytical technology.The most useful of these publications are reports of reference material characterisation programmes of which the following have been used to abstract details for the present assessment. Fairbairn et al.7 This USGS Bulletin presented results from 34 laboratories that were asked to analyse two samples, a granite (G-1) and a diabase (W-1). Subsequently, these samples achieved the status of being the first geochemical reference materials, although the original intention was to provide two well-characterised samples that could be used for the calibration of dc arc atomic emission spectrography (Ahrens8).The majority of results presented in this Bulletin were for the major elements, and almost all determinations were made using the so-called ‘classical’ wet chemistry methods of analysis. The details included in this Bulletin give an insight into the ultimate development of classical methods of analysis as practised in geochemistry laboratories up to 1950. Shapiro and Brannock9 During the 1950s, the capabilities of geoanalytical laboratories were still generally restricted to the determination of the major elements, although a number of refinements were introduced to improve element specificity and the rate of analytical production.Several comprehensive schemes of analysis detailing socalled ‘rapid’ methods were published, of which that of Shapiro and Brannock can be regarded as the most authoritative, representing the analytical capabilities of laboratories at the end of the 1950s, before procedures were modified to include determinations made by atomic absorption spectrometry. Reference material characterisation studies Subsequent to the 1950s, a series of instrumental techniques were progressively introduced into geoanalytical laboratories, including atomic absorption spectrometry, X-ray fluorescence, instrumental neutron activation analysis (INAA), ICP-atomic emission spectrometry and ICP-mass spectrometry.Several of these techniques have overlapping capabilities in the determination of major and trace elements. The way in which these techniques were integrated into geoanalytical laboratories can be judged from a series of reference material characterisation reports published between 1980 and 1994. Of these, the series by Govindaraju, which appeared in Geostandards Newsletter, may be regarded as the most comprehensive.Laboratories were asked to analyse candidate reference materials and contribute results together with brief details of the analytical techniques used. A period of 6–9 months was typically allowed between the distribution of samples and the deadline for the submission of data. These characterisation reports represent, therefore, a snapshot of the techniques employed by contributing laboratories in the period immediately preceding publication of the report. Comparison of data presented in a series of reports reveals, therefore, trends in the popular use of techniques.For the present review, data from the following reference material characterisation studies have been abstracted: Govindaraju10: data for anorthosite, AN-G; Govindaraju11: data for Ailsa Craig microgranite, AC-E; Govindaraju et al.12: data for Whin Sill dolerite, WS-E. These data have been supplemented by details abstracted from the most recent proficiency testing round for geochemical laboratories, GeoPT2 (Thompson et al., 199713), based on the analysis of Bardon volcanic tuff, OU-1.The extent to which data abstracted from these reports are representative of contemporary laboratory facilities may be judged from the relatively large number of laboratories participating in each of these characterisation studies as listed in Table 1. The number of times individual techniques were used to contribute data for selected elements to these characterisation studies is listed in Table 2.Isotopic Studies Although the reports listed in the previous section provide a record of the use of bulk geoanalytical techniques, some of the most influential advances in geochemistry over the last 20 years have been based on the isotopic analysis of samples using techniques such as thermal ionisation mass spectrometry, gas source mass spectrometry (GSMS), accelerator mass spectrometry (AMS) and (particularly in earlier studies) alpha spectrometry.The first two of these techniques, in particular, received a significant boost from the geochemical studies associated with the Lunar Landing Programme. It is more difficult to evaluate trends in the use of these techniques, since isotope laboratories have not normally contributed to reference material characterisation studies. However, some evidence can be obtained by examining papers published in annual volumes of an influential geochemical research journal, Geochimica et Cosmochimica Acta, noting that the scope of this journal covers a wide range of pure geochemical (and cosmochemical) applications.To obtain this evidence, the title of each research paper published in Table 1 Number of laboratories participating in the cited reference material characterisation and proficiency testing programmes Number of participating Year Sample laboratories 1980 Anorthosite: AN-G 121 1987 Ailsa Craig microgranite: AC-E 128 1994 Whin Sill dolerite: WS-E 104 1997 Bardon volcanic tuff: OU-1 60 1180 Analyst, November 1997, Vol. 122selected annual volumes of this journal has been examined and a note made of all isotopic analysis measurements cited or inferred in the title. The volumes abstracted were for the years 1970, 1980, 1987, 1994 and 1996 (selected largely to match the dates of the reference material characterisation studies listed above, but with additional information extended back to 1970). Each volume contained a relatively large number of research contributions (typically 300–400).However, the extent to which the abstracted information can be considered to represent the contemporary state of isotopic analysis will be affected by the scope of the journal. Furthermore, the abstracted information may under-represent the true usage since no account is taken of techniques that contributed to a published work where the use of that technique could not reasonably be inferred from the title of the paper.Results and Principal Trends Trends in Bulk Analytical Techniques When comparing analytical data presented in the series of reference material characterisation studies, one clear trend is the substantial increase in the number of trace elements reported routinely in studies published subsequent to that of Govindaraju in 1980.10 This trend is illustrated by data presented in Table 3 which lists the trace elements for which 20 or more laboratories reported results in each characterisation study.In the 198010 study, only 13 trace elements satisfied this criterion. In subsequent reports, although there was some variation, the number approximately doubled, with a significant extension in the representation of the rare earth elements. Other trends in analytical techniques may be evaluated by examining data listed in Table 2 and plotted in Fig. 1 [(a) to (g)]. Data in Fig. 1 show the proportion of results for selected elements determined by individual techniques plotted as a function of the year of publication of the reference material characterisation programme. As can be seen, the main contributions were made by classical/rapid chemical methods, AAS, INAA, XRF, ICP-AES and ICP-MS.Data are plotted for the elements SiO2 and Fe2O3 (representing the major elements), Ba, Sr, Zr and Hf (representing the important geochemical discriminatory trace elements) and Eu (representing the REE). Trends for the two major elements are extended back to 1950 to take account of the G-1/W-1 characterisation study.However, since the routine determination of the trace elements was not undertaken until the 1960s and 1970s, trends in these data have only been plotted back to 1975. When interpreting these data, account should be taken of the years in which the various categories of commercial instrumentation became widely available in geochemical laboratories. These dates are approximately as follows: AAS, 1960; XRF, 1963; INAA, 1968; ICPTable 2 Number of techniques* that have contributed to selected characterisation programmes over the period 1950–1997 Sample/ Concentration element Year (% m/m) Chem.Phot. AAS Arc-AES ICP-AES DCP-AES SSMS TIMS ICP-MS XRF XRF-g XRF-p INAA Probe Total G-1 SiO2 1950 72.51 34 34 AN-G SiO2 1980 46.3 47 15 1 8 41 40 1 112 AC-E SiO2 1987 70.35 19 18 1 14 71 66 5 1 124 WS-E SiO2 1994 50.7 13 3 16 60 54 6 92 OU-1 SiO2 1997 58.25 2 4 8 37 51 G-1 Fe2O3T 1950 1.95 30 30 AN-G Fe2O3T 1980 3.36 38 32 1 11 40 38 2 5 127 AC-E Fe2O3T 1987 2.53 14 26 2 17 75 68 7 11 1 146 WS-E Fe2O3T 1994 13.15 5 9 22 57 51 6 5 98 OU-1 Fe2O3T 1997 8.99 2 4 11 35 2 54 mg g21 G-1 Ba 1950 1080 3 3 AN-G Ba 1980 34 6 6 4 1 2 13 2 11 4 36 AC-E Ba 1987 55 10 4 16 3 34 3 31 11 78 WS-E Ba 1994 338 6 21 12 46 18 28 5 90 OU-1 Ba 1997 131.4 1 15 8 25 3 52 G-1 Eu 1950 1.22 0 AN-G Eu 1980 0.37 1 8 9 AC-E Eu 1987 2 1 13 2 1 3 21 41 WS-E Eu 1994 2.25 13 18 8 39 OU-1 Eu 1997 0.52 5 19 6 30 G-1 Hf 1950 5.4 0 AN-G Hf 1980 0.38 1 8 9 AC-E Hf 1987 27.9 1 13 2 1 3 21 41 WS-E Hf 1994 5.3 1 16 9 4 5 9 35 OU-1 Hf 1997 1.65 1 16 7 24 G-1 Sr 1950 248 3 3 AN-G Sr 1980 76 12 6 4 1 1 28 2 26 1 53 AC-E Sr 1987 3 1 9 1 12 2 1 35 3 32 61 WS-E Sr 1994 410 1 7 20 18 60 20 40 2 108 OU-1 Sr 1997 104.8 2 12 8 29 1 52 G-1 Zr 1950 201 3 3 AN-G Zr 1980 11 4 2 1 23 2 21 30 AC-E Zr 1987 780 3 14 2 1 53 3 50 6 79 WS-E Zr 1994 195 1 16 11 54 16 38 2 84 OU-1 Zr 1997 55 1 8 9 29 47 * Chem.= classical/rapid techniques; Phot. = flame photometry; AAS = atomic absorption spectrometry; Arc-AES = dc arc atomic emission spectrometry spectrography; ICP-AES = inductively coupled plasma atomic emission spectrometry; DCP-AES = dc coupled plasma atomic emission spectrometry; SSMS = spark source mass spectrometry/ spectrography; TIMS = thermal ionisation mass spectrometry; ICP-MS = inductively coupled plasma-mass spectrometry; XRF = X-ray fluorescence analysis; XRF-g = laboratories undertaking XRF analyses on a glass disc; XRF-p = laboratories undertaking XRF analyses on a powder pellet; INAA = neutron activation analysis; Probe = electron microprobe analysis.Analyst, November 1997, Vol. 122 1181AES, 1975; ICP-MS, 1985. To simplify the interpretation of trends, tie-lines have been drawn between successive data points, although these tie lines do not necessarily imply a continuous trend in the popular use of the respective techniques.The principal features of the trends in these data are as follows. Major elements (represented by SiO2 and Fe2O3) [Figs. 1(a) and (b)] Although wet chemical procedures were the only reliable methods available in 1950, these procedures declined rapidly in popular use as the majority of laboratories turned to XRF methods once appropriate procedures had been fully established. The earlier interest in the use of AAS techniques (especially indicated in the determination of total iron), has declined in recent years.Apart from XRF, the other technique that now makes a significant contribution to the determination of the major elements is ICP-AES, although the proportion of laboratories reporting results by this technique is significantly lower than that by XRF. Trends in these data indicate that XRF will continue to be the technique chosen by the majority of laboratories for the routine determination of the major elements. Sr, Ba, Zr [Figs. 1(c), (d) and (e)] This group of elements is also dominated by contributions of laboratories using the XRF technique with ICP-AES also making a significant contribution.In recent years there has been Table 3 Elements for which 20 or more laboratories reported results in the cited reference material characterisation and proficiency testing programmes Year Sample Elements with 20 or more reported results 1980 AN-G Ba, Co, Cr, Li, Ni, Pb, Rb, Sr, V, Y, Zn, Zr 1987 AC-E As above for 1980 plus: Be, Ce, Cs, Dy, Eu, Ga, Gd, Hf, La, Lu, Nb, Nd, Sc, Sm, Ta, Tb, Th, U, Yb 1994 WS-E As above for 1987 plus: Er, Ho, Mo, Pr, Sn, Tm (but not including: Be, Cs, Li) 1997 OU-1 As, Ba, Ce, Co, Cr, Cu, Er, Eu, Ga, Gd, Hf, Ho, La, Lu, Nb, Nd, Ni, Pb, Pr, Rb, Sc, Sm, Sr, Tb Th, U, V, Y, Yb, Zn, Zr Fig. 1 Trends in the use of selected analytical techniques. The proportion of laboratories reporting results by specified techniques are plotted as a function of the year of publication of the appropriate reference material characterisation study of proficiency test round.Data are plotted for the elements (a) SiO2 (b) Fe2O3 (c) Sr (d) Ba, (e) Zr (f) Hf and (g) Eu. 1182 Analyst, November 1997, Vol. 122a decline in the popularity of AAS and INAA (where applicable), but an increase in the use of the ICP-MS technique. Trends in the data would indicate a continuing dominant role for XRF analysis. However, these trends also indicate that the role of ICP-MS in the determination of these elements will increase and it will be interesting to see if part of this increase will be at the expense of ICP-AES.Hf and Eu [Figs 1(f) and (g)] The most popular technique for the determination of these elements in the early 1980s was INAA, with a significant contribution from ICP-AES during the mid-1980s. However, there has been a dramatic and progressive decrease in the popularity of INAA in recent years and an equally dramatic increase in the reporting of data by ICP-MS.This change may, in part, be associated with a decline in the number of research reactors with facilities for the irradiation of INAA samples. However, a significant contribution is likely to be a recognition of the extended capabilities of ICP-MS for the determination of a wide range of trace elements as well as the attractive option of extending the capability of the technique using laser ablation. Trends in data indicate that this change is likely to continue, with ICP-MS increasingly becoming the dominant technique for this category of measurement.Overall Although data are plotted in Figs. 1 (a)–(g) only for selected elements, these results indicate that whereas in the 1980s XRF and INAA represented the combination of techniques that were most frequently selected by geochemical laboratories to contribute most routine geochemical data, XRF and ICP-MS now represent the two techniques with largely complementary capabilities that are increasingly chosen to satisfy this role.Trends in Isotopic Techniques The treatment of data representing the use of isotopic techniques is presented in a different way in Figs. 2 (a)–(d). Data abstracted from Geochimica et Cosmochimica Acta for individual categories of measurement are plotted as a series of histograms, where each bar of the histogram represents the number of contributions made in annual volumes for the years 1970, 1980, 1987, 1994 and 1996, respectively.The categories of measurement are as follows. Fig. 2(a): stable isotopes (H, C, N, O and S). Fig. 2(b): noble gases (He, Ne, Ar, Kr and Xe). Fig. 2(c): radiogenic isotopes and isotopes used in dating applications (14C, K–Ar, Ar–Ar, Rb–Sr, Pb–Pb, U–Pb, Th–Pb, U disequilibrium studies, Sm– Nd, Re–Os). Rare earth element geochemistry studies have been included as an indicator of continuing use of conventional geochemical modelling as well as citations for artificial and cosmogenic isotopes and the use of Pb isotopes in tracer studies.Fig. 2(d): more specialised studies using other isotopic systems, mainly related to ‘difficult’ stable isotope systems and isotope systems where minute differences in isotopic abundance can be related to nucleosynthetic processes relevant to the early formation of the Solar System. These data indicate a sustained contribution of isotopic techniques to geochemical research over the years in question.Particular features worthy of note are as follows. (i) Continuing interest in the contribution of stable isotope studies (based on gas source mass spectrometry) with a significant increase in the popularity of oxygen isotope techniques [Fig. 2(a)]. (ii) A relatively consistent contribution from noble gas studies, no doubt reflecting the source journal’s interest in cosmochemistry [Fig. 2(b)]. (iii) A comparative decline in the popularity of K– Ar dating studies, replaced by more the precise Ar–Ar measurements, which in term have been augmented in recent years by the laser extraction Ar-Ar technique [Fig. 2(c)]. (iv) The comparative recent exploitation of Nd–Sm and U–Pb procedures following the development of higher precision thermal ionisation mass spectrometers in the 1980s [Fig. 2(c)]. (v) The even more recent exploitation of the Re–Os geochronometer, following the development of negative ion TIMS instrumentation in the 1990s [Fig. 2(c)]. (vi) A comparative increase in popularity of uranium disequilibrium procedures, noting that the data plotted in Fig. 2(c) represents the combination of determinations made by alpha counting in earlier years and the more recent higher precision measurements made by TIMS instrumentation. (v) It is also interesting to note that despite the introduction of higher sensitivity mass spectrometer instrumentation capable of exploiting new isotope systems, interest in the more traditional Rb–Sr and REE measurements has continued to expand, although the former would now normally be undertaken on individual mineral separates, rather than on bulk samples as in former years [Fig. 2(c)]. (vi) These data do not differentiate and therefore adequately credit the contribution of ion probe techniques, including SHRIMP, to the U–Pb dating studies, particularly of zircons. Overall Trends in Isotopic Analysis One trend that is apparent from data presented in Fig. 2 is that technical advances in mass spectrometer instrumentation have had a significant impact by permitting the exploitation of new isotope systems.Two examples, mentioned above, relate to advances in thermal ionisation mass spectrometry instrumentation. Higher precision instrumentation developed in the 1980s permitted routine characterisation of the Sm–Nd (and U–Pb) isotope systems and negative ionisation technology, developed in the 1990s, allowed routine measurement of the Re–Os system. In terms of future trends, therefore, it seems likely that technical developments, particularly improvements in the precision of measurements, could be influential in allowing measurements to be made of isotope systems that are beyond the capabilities of the present generation of instrumentation.In so doing, new categories of geochemical data would become available. The impact of these developments is most likely to see the exploitation of isotope systems such as those listed in Fig. 2(d). As an example of this category of measurement, Sharma et al.14 investigated the 146Sm isotope system. This isotope has a half-life of 149 Ma and is now extinct. However, if TIMS measurements can be made to a precision that is three times better than at present, the potential exists to investigate variations in progeny 142Nd isotopic composition that occurred during the first few hundred million years after the formation of the Universe, thus providing new geochemical data on nucleosynthetic processes.The second trend illustrated by data in Fig. 2 is the continuing role for isotopic (and other geochemical) measurements. Although the pioneering days in developing and applying both Rb–Sr isotope system and rare earth element geochemistry are long since passed, the evidence is that both of these categories of measurement have a continuing and enduring role to play in geochemical research. One can conclude, therefore, that almost all the isotope measurements that have had a significant impact in the past will continue to contribute to future geochemical studies.However, the influence of technical advances that offer improvements in the quality of geochemical information should not be overlooked. Of particular relevance are the improvements that would arise if all isotope measurements could be obtained on a microprobe scale. One example that has already arisen through both advances in the instrumentation capabilities and the development of appropriate techniques is the progression from K–Ar to the Ar–Ar dating technique and the Analyst, November 1997, Vol. 122 1183subsequent improvements associated with the further development of the laser extraction Ar–Ar microprobe technique.Evolution of Microanalytical Techniques Although not directly apparent from the data presented above for bulk and isotopic techniques, the advance in analytical technology that is likely to have the most influential and enduring effect on geochemical research is that of microprobe instrumentation.To a large extent, the capability now exists to undertake many of the analytical measurements normally performed on bulk samples using advanced microprobe techniques (see contributions in Potts et al.15). These techniques provide analytical information on individual minerals, normally Fig. 2 Histograms showing the number of citations of isotopic abundance or ratio measurement for the years 1970, 1980, 1987, 1994, 1996.Data were collated by examining the titles of research contributions published in the corresponding annual volumes of the journal Geochimica et Cosmochimica Acta. Citations of the following isotopic systems are plotted: (a) stable isotopes; (b) noble gases; (c) radiogenic isotopes (plus 14C, REE, artificial isotopes, cosmogenic isotopes and Pb isotope tracer studies); and (d) other isotopic measurements. 1184 Analyst, November 1997, Vol. 122in situ in thin section, and are capable, therefore, of providing much more detail about the geochemical history of a sample in comparison with bulk measurements.A summary of the analytical capabilities of some of these techniques is as follows. Electron microprobe Although restricted to the determination of major and minor elements at spatial resolutions of about 2–3 mm, the electron microprobe is the only microprobe method capable of quantitative analysis without having to resort to internal standardisation of the analytical signal or comparison with matrix matched calibration samples.In fact, a reliable physical model exists for the excitation process, and forms the basis for the most common type of matrix correction procedure. The electron microprobe, therefore, occupies a pivotal role in microanalysis. Modern instrumentation offers the capability of high resolution backscattered electron imaging and X-ray mapping, and is capable of revealing zoning effects that were overlooked on earlier generations of instrumentation. The electron microprobe occupies an analogous position in microprobe studies to XRF in the routine bulk analysis of silicate samples.Ion probe By ablating and exciting a sample with a beam of ions (often O2), focused to typically 10–25 mm and measuring the secondary ion mass spectrum using a double focusing (high resolution) mass spectrometer, high precision measurements can be made of the isotopic and trace element composition of samples to sub-mg g21 detection limits.An important influence on these studies has been the application of the SHRIMP ion probe to geochronology studies, mainly based on the analysis of zircon. Trace element calibration generally has to be achieved by comparing normalised intensities to those of a matrix matched mineral standard. Laser ablation ICP-MS This technique is already widely used for the trace element characterisation of minerals by the ablation of material using a laser focused to, typically, 50 mm diameter and analysing directly the ablated material using an ICP-MS.Calibration is normally achieved by internal standardisation. Conventional instrumentation is based on a general purpose quadrupole mass spectrometer with a single electron multiplier detector. However, laser ablation ICP-MS instrumentation using a high resolution magnetic sector or double focusing mass spectrometer with multiple collection has as yet an only partially fulfilled potential for measuring isotopic compositions to precisions approaching that of TIMS and the ion probe, so providing a general capability for the determination of isotope ratios on a microprobe scale.Laser Ar–Ar mass spectrometry The quality of data obtained when a laser is used to extract argon from a small region of an irradiated sample is far superior to that from the bulk analysis of samples using the 39Ar : 40Ar technique. The confidence of these measurements can be further increased if the laser power is controlled to give stepwise heating and extraction of the argon gas.In this way, the isotopic history of a sample can be characterised in much greater detail by the detection (if present) of variations within individual mineral grains, as well as the identification of atmospheric contamination. This microanalysis development demonstrates the power of laser extraction procedures and has discredited, to some extent, conventional bulk measurements of isotopic composition in the same way that the detection of zoning effects in zircon using the ion probe has discredited isotopic measurements based on the bulk isotopic composition of this mineral. Laser fluorination An active area of development is in the stable isotope characterisation of individual minerals using laser induced fluorination and laser extraction techniques and high-precision gas-source mass spectrometric detection. Although, like LA– ICP-MS, the spatial resolution of such measurements is currently not as good as with other microbeam techniques (see, for example, Wright16), these developments are likely to lead to similar advances in the geochemical interpretation of stable isotope ratios, as has occurred with other microprobe techniques.Synchrotron X-ray microprobe A synchrotron is a facility offering X-ray beams of very high intensity that have very low divergence (i.e., are near parallel) and are polarised.These beams can be focused onto the surface of a sample, facilitating both the determination of the trace elements by X-ray fluorescence and nearest-neighbour coordination information by X-ray absorption measurements. Submicrometre beams can be achieved on the latest, third generation, synchrotrons, although taking into account the penetrating properties of X-rays, the depth analysed is usually controlled by the thickness of the sample. Future Development of Microprobe Techniques Much of the earlier developments in geochemistry have been based on analytical data obtained from bulk methods of analysis.However, with the introduction of successive microprobe techniques, it has been clearly demonstrated that a much greater wealth of information can often be obtained from the microanalysis of individual minerals. This information can show the presence of mineral zoning effects that were often previously unsuspected and indeed could not readily be identified from bulk methods of analysis.In fact, results from these microprobe techniques have, in many cases, discredited equivalent determinations on bulk samples (as in the ion probe analysis of zircon or laser Ar–Ar studies on individual minerals). It seems likely, therefore, that as these techniques are developed further, geochemistry will increasingly be based on results obtained by microprobe analysis because of the greater detail that can then be obtained about the geochemical history of samples.Conclusions An analysis of trends in the use of geoanalytical techniques over the last 25 years shows that XRF and ICP-MS are likely to represent the combination of instrumentation dominating the production of routine major and trace element data in future years. In an equivalent evaluation of trends in isotopic methods of analysis, improvements in mass spectrometry instrumentation have had a large influence on the geochemical applications. This observation is illustrated by the introduction of both higher sensitivity thermal ionisation instrumentation in the 1980s, which facilitated the development of the Sm–Nd and U–Pb radiogenic isotope systems, and negative ionisation instrumentation in the 1990s and its application to the Re–Os system. If these advances in sensitivity continue, it is likely that analytical capabilities will extend to new isotope systems that are beyond the capabilities of current instrumentation.However, the greatest impact in future is likely to result from the further development and geochemical application of data from Analyst, November 1997, Vol. 122 1185a range of microprobe techniques. An overview of the current capabilities of techniques like the ion probe and laser Ar–Ar dating already shows that microprobe techniques have discredited equivalent measurements made on bulk samples. This trend is likely to continue with further refinements in microprobe instrumentation and geochemical models for interpreting microanalysis data.The author would like to thank the organising committee of Geoanalysis 97 for the opportunity to present this paper and to B. Chappell (ANU) for valued comments on an earlier version of the manuscript. References 1 Goldschmidt, V. M., and Thomassen, L., ‘Geochemische Verteilungsgeesetze der Elemente III’, Videnskabsselsk Skr. 1. Mat.- Naturvidensk Kl., 1924 (5), 24; as quoted in Mason, B., Victor Moritz Goldschmidt: Father of Modern Geochemistry, The Geochemical Society, San Antonio , TX, USA, 1992, pp. 1–184. 2 Masuda, A., Earth Sci. Nagoya Univ., 1962, 10, 173. 3 Potts, P. J., Hawkesworth, C. J., van Calsteren, P., and Wright, I. P., in Magmatic Processes and Plate Tectonics, ed. Prichard, H. M., Alabaster, T., Harris, N. B. W., and Neary, C. R., Geological Society, London, UK, 1993, p. 501. 4 Boumans, P., J. Anal. At. Spectrom., 1993, 8, 767. 5 Hall, G.E. M., J. Geochem. Explor., 1996, 57, 1. 6 Hiefje, G. M., J. Anal. At. Spectrom., 1989, 4, 117. 7 Fairbairn, H. W., Schlect, W. G., Stevens, R. E., Dennen, W. H., Ahrens, L. H. and Chayes, F. US Geol. Surv. Bull., 1951, 980. 8 Ahrens, L. H., Geostand. Newsl., 1977, 1, 157. 9 Shapiro, L., and Brannock, W. W., U.S. Geol. Surv. Bull., 1962, 1144-A. 10 Govindaraju, K., Geostand. Newsl., 1980, 4, 49. 11 Govindaraju, K., Geostand. Newsl., 1987, 11, 203. 12 Govindaraju, K., Potts, P.J., Webb, P. C., and Watson, J. S., Geostand. Newsl., 1994, 18, 211. 13 Thompson, M., Potts, P. J., Kane, J. S., Webb, P. C., and Watson, J. S., Geostand. Newsl. J. Geostand. Geoanal., 1997, in the press. 14 Sharma, M., Papanastassiou, D. A., Wasserburg, G. J., and Dymek, R. F., Geochim. Cosmochim. Acta, 1996, 60, 2037. 15 Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., The Mineralogical Society and Chapman and Hall, London, UK 1995, pp. 1–419. 16 Wright, I. P., in Micropole Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., The Mineralogical Society and Chapman and Hall, London, UK, 1995, ch. 9. Paper 7/04856D Received July 8, 1997 Accepted August 26, 1997 1186 Analyst, November 1997, Vol. 122 Geoanalysis: Past, Present and Future† Philip J. Potts‡ Department of Geology, The Australian National University, Canberra, ACT 0200, Australia A review of trends in the use of techniques for the chemical analysis of geological samples has been undertaken by abstracting details from reference material characterisation programmes published over the period 1980–1997 and from papers published in Geochimica et Cosmochimica Acta over a similar period.An analysis of these details shows that whereas during the 1980s, XRF and INAA were the two most popular techniques chosen for the major and trace element analysis of silicate rocks, this role continues to be filled by XRF but now complemented by ICP-MS.An equivalent survey of the geochemical use of isotope measurements shows that improvements in the sensitivity and precision of mass spectrometry instrumentation has led to new analytical capabilities: for example, in the application to the Sm–Nd and U–Pb isotope systems in the 1980s and to Re–Os in the 1990s. If these instrument developments continue, the measurement of new isotope systems may become possible including, for example, those related to nucleosynthetic processes.However, it is argued that the greatest influence in future years will arise from the further development of microprobe techniques. Already some of these techniques have demonstrated the importance of providing data that takes into account zoning effects in individual minerals, as in the U–Pb dating of zircon by ion probe and the dating of selected minerals using the laser Ar–Ar technique.Indeed, in these applications, equivalent determinations on bulk samples are now largely discredited, because they mask variations at the mineralogical level. It is likely, therefore, that with further developments in microprobe instrumentation and refinements in geochemical models for interpreting microanalysis data, this area of endeavour will play an increasingly important role in future geochemical studies. Keywords: Trends; future developments; analytical techniques; geochemistry; isotope techniques; microprobe techniques Of all the sciences supported by analytical measurements, the modern day demands of geochemistry are more stringent than those of almost any other application.Not only is there a requirement for the routine determination of all the major elements, and a range of trace elements, that are routinely used in geochemical modelling, but there are also geochemical applications for almost every other element in the periodic table. Furthermore, important branches of geochemistry are based on the determination of isotope ratios derived from the decay of naturally occurring radioactive isotopes (e.g., Rb–Sr, Sm–Nd, Re–Os, U–Pb, Pb–Pb), stable isotopes (particularly of the elements H, N, C, O, S), the noble gases and naturally occurring artificial isotopes (such as 10Be, 14C, 26Al, 36Cl, 129I), which are formed by cosmic ray interactions in the upper atmosphere and can be used to model the rates of more recent geochemical processes.In addition, recent advances in microbeam techniques now permit many major, trace and isotope ratio measurements to be made on individual minerals to spatial resolutions in the range 1–50 mm. What distinguishes geochemical research is not only the wide range of analytical measurements required, but also the high degree of performance demanded (in terms of precision, accuracy and low detection limits). Furthermore, there are few other areas of research where important applications exist for all these measurements using microbeam instrumentation.This present-day degree of sophistication has not always been the case. Fifty years ago, routine analytical techniques were based mainly on chemical separation procedures followed by a gravimetric determination, the so-called ‘classical’ method of analysis. From any cursory examination of the subsequent development of geochemical research it is apparent that advances in analytical instrumentation have had a substantial impact on its development but that equally the requirements of geochemical research have had a significant effect on the development of geoanalytical techniques. One example of this synergy is the development of rare earth element geochemistry.In 1924, Goldschmidt and Thomassen1 published rare earth element concentration data plotted as a function of atomic number. However, it was not until the 1970s that REE geochemistry achieved prominence.This development arose because of: (i) the development of high-purity germanium detectors, which facilitated the routine determination of selected REEs to low detection limits by neutron activation analysis; (ii) the development of a geochemical framework within which to interpret REE data, especially the convention of normalising REE concentrations to chondrite abundances (Masuda2 and others) to overcome the ‘saw-tooth’ effect when un-normalised concentration data are plotted as a function of atomic number; (iii) the realisation that REE patterns could shed light on geochemical processes, such as those associated with the then new ideas of plate tectonics.Arguably, the substantial impact of REE geochemistry required the coincidence of all three of these developments. Accepting the importance of analytical technology on progress in geochemistry, the purpose of this paper is to review trends in the development of analytical techniques over the last 50 years and to use these data as a basis for predicting which techniques are likely to make the most significant contribution to future geochemical research.This evaluation has, in part, been based on an earlier review (Potts et al.3). However, one of the aims of the present paper is to show that more quantitative predictions can be made of future trends, based on an evaluation of the use of techniques over the last two decades. In developing this approach, it soon became apparent more conclusions depend very much on the perspective adopted by the reviewer.Several papers covering developments in atomic spectrometry techniques have appeared in recent years. For example, Boumans4 evaluated trends in a range of modern analytical spectrochemical techniques and included the significant contemporary use of glow-discharge mass spectrometry and dc-arc atomic emission spectrometry. Whereas it is true that the former has had an influential role in the characterisation of † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.‡ On study leave from: Department of Earth Sciences, The Open University, Walton Hall, Milton Keynes, UK MK7 6AA. Analyst, November 1997, Vol. 122 (1179–1186) 1179semiconductor materials, and the latter in the metallurgical industry, neither currently plays a significant role in geochemical research.Hall5 recently reviewed trends in techniques that have influenced, to a significant extent, developments in geochemical exploration research, placing particular emphasis on the contribution of atomic absorption spectrometry and ICP techniques. The present paper will show that this emphasis is not supported to the same extent by the contribution these techniques have made to pure geochemical research. Hiefje,6 in a well known review of atomic absorption spectrometry, undertook an assessment of trends in research publications in that topic.When extrapolated to predict publication rates in future years, the trend in these data could be interpreted as demonstrating that the AAS technique would die out by the end of the century, indicating the misunderstandings and pitfalls that can arise when attempting to predict the future. Since the present review is based on an analysis of trends in the use of analytical techniques that are employed in pure geochemical research applications, it must be emphasised that any conclusions are not necessarily directly applicable to other fields of geochemical research.Trends in Geochemical Techniques Bulk Analytical Techniques When considering the contribution made by bulk geochemical techniques, it is relatively easy to identify a series of publications that include details of contemporary analytical technology. The most useful of these publications are reports of reference material characterisation programmes of which the following have been used to abstract details for the present assessment.Fairbairn et al.7 This USGS Bulletin presented results from 34 laboratories that were asked to analyse two samples, a granite (G-1) and a diabase (W-1). Subsequently, these samples achieved the status of being the first geochemical reference materials, although the original intention was to provide two well-characterised samples that could be used for the calibration of dc arc atomic emission spectrography (Ahrens8).The majority of results presented in this Bulletin were for the major elements, and almost all determinations were made using the so-called ‘classical’ wet chemistry methods of analysis. The details included in this Bulletin give an insight into the ultimate development of classical methods of analysis as practised in geochemistry laboratories up to 1950. Shapiro and Brannock9 During the 1950s, the capabilities of geoanalytical laboratories were still generally restricted to the determination of the major elements, although a number of refinements were introduced to improve element specificity and the rate of analytical production. Several comprehensive schemes of analysis detailing socalled ‘rapid’ methods were published, of which that of Shapiro and Brannock can be regarded as the most authoritative, representing the analytical capabilities of laboratories at the end of the 1950s, before procedures were modified to include determinations made by atomic absorption spectrometry.Reference material characterisation studies Subsequent to the 1950s, a series of instrumental techniques were progressively introduced into geoanalytical laboratories, including atomic absorption spectrometry, X-ray fluorescence, instrumental neutron activation analysis (INAA), ICP-atomic emission spectrometry and ICP-mass spectrometry. Several of these techniques have overlapping capabilities in the determination of major and trace elements.The way in which these techniques were integrated into geoanalytical laboratories can be judged from a series of reference material characterisation reports published between 1980 and 1994. Of these, the series by Govindaraju, which appeared in Geostandards Newsletter, may be regarded as the most comprehensive. Laboratories were asked to analyse candidate reference materials and contribute results together with brief details of the analytical techniques used.A period of 6–9 months was typically allowed between the distribution of samples and the deadline for the submission of data. These characterisation reports represent, therefore, a snapshot of the techniques employed by contributing laboratories in the period immediately preceding publication of the report. Comparison of data presented in a series of reports reveals, therefore, trends in the popular use of techniques. For the present review, data from the following reference material characterisation studies have been abstracted: Govindaraju10: data for anorthosite, AN-G; Govindaraju11: data for Ailsa Craig microgranite, AC-E; Govindaraju et al.12: data for Whin Sill dolerite, WS-E.These data have been supplemented by details abstracted from the most recent proficiency testing round for geochemical laboratories, GeoPT2 (Thompson et al., 199713), based on the analysis of Bardon volcanic tuff, OU-1. The extent to which data abstracted from these reports are representative of contemporary laboratory facilities may be judged from the relatively large number of laboratories participating in each of these characterisation studies as listed in Table 1.The number of times individual techniques were used to contribute data for selected elements to these characterisation studies is listed in Table 2. Isotopic Studies Although the reports listed in the previous section provide a record of the use of bulk geoanalytical techniques, some of the most influential advances in geochemistry over the last 20 years have been based on the isotopic analysis of samples using techniques such as thermal ionisation mass spectrometry, gas source mass spectrometry (GSMS), accelerator mass spectrometry (AMS) and (particularly in earlier studies) alpha spectrometry.The first two of these techniques, in particular, received a significant boost from the geochemical studies associated with the Lunar Landing Programme.It is more difficult to evaluate trends in the use of these techniques, since isotope laboratories have not normally contributed to reference material characterisation studies. However, some evidence can be obtained by examining papers published in annual volumes of an influential geochemical research journal, Geochimica et Cosmochimica Acta, noting that the scope of this journal covers a wide range of pure geochemical (and cosmochemical) applications.To obtain this evidence, the title of each research paper published in Table 1 Number of laboratories participating in the cited reference material characterisation and proficiency testing programmes Number of participating Year Sample laboratories 1980 Anorthosite: AN-G 121 1987 Ailsa Craig microgranite: AC-E 128 1994 Whin Sill dolerite: WS-E 104 1997 Bardon volcanic tuff: OU-1 60 1180 Analyst, November 1997, Vol. 122selected annual volumes of this journal has been examined and a note made of all isotopic analysis measurements cited or inferred in the title. The volumes abstracted were for the years 1970, 1980, 1987, 1994 and 1996 (selected largely to match the dates of the reference material characterisation studies listed above, but with additional information extended back to 1970). Each volume contained a relatively large number of research contributions (typically 300–400).However, the extent to which the abstracted information can be considered to represent the contemporary state of isotopic analysis will be affected by the scope of the journal.Furthermore, the abstracted information may under-represent the true usage since no account is taken of techniques that contributed to a published work where the use of that technique could not reasonably be inferred from the title of the paper. Results and Principal Trends Trends in Bulk Analytical Techniques When comparing analytical data presented in the series of reference material characterisation studies, one clear trend is the substantial increase in the number of trace elements reported routinely in studies published subsequent to that of Govindaraju in 1980.10 This trend is illustrated by data presented in Table 3 which lists the trace elements for which 20 or more laboratories reported results in each characterisation study.In the 198010 study, only 13 trace elements satisfied this criterion. In subsequent reports, although there was some variation, the number approximately doubled, with a significant extension in the representation of the rare earth elements.Other trends in analytical techniques may be evaluated by examining data listed in Table 2 and plotted in Fig. 1 [(a) to (g)]. Data in Fig. 1 show the proportion of results for selected elements determined by individual techniques plotted as a function of the year of publication of the reference material characterisation programme.As can be seen, the main contributions were made by classical/rapid chemical methods, AAS, INAA, XRF, ICP-AES and ICP-MS. Data are plotted for the elements SiO2 and Fe2O3 (representing the major elements), Ba, Sr, Zr and Hf (representing the important geochemical discriminatory trace elements) and Eu (representing the REE). Trends for the two major elements are extended back to 1950 to take account of the G-1/W-1 characterisation study. However, since the routine determination of the trace elements was not undertaken until the 1960s and 1970s, trends in these data have only been plotted back to 1975.When interpreting these data, account should be taken of the years in which the various categories of commercial instrumentation became widely available in geochemical laboratories. These dates are approximately as follows: AAS, 1960; XRF, 1963; INAA, 1968; ICPTable 2 Number of techniques* that have contributed to selected characterisation programmes over the period 1950–1997 Sample/ Concentration element Year (% m/m) Chem.Phot. AAS Arc-AES ICP-AES DCP-AES SSMS TIMS ICP-MS XRF XRF-g XRF-p INAA Probe Total G-1 SiO2 1950 72.51 34 34 AN-G SiO2 1980 46.3 47 15 1 8 41 40 1 112 AC-E SiO2 1987 70.35 19 18 1 14 71 66 5 1 124 WS-E SiO2 1994 50.7 13 3 16 60 54 6 92 OU-1 SiO2 1997 58.25 2 4 8 37 51 G-1 Fe2O3T 1950 1.95 30 30 AN-G Fe2O3T 1980 3.36 38 32 1 11 40 38 2 5 127 AC-E Fe2O3T 1987 2.53 14 26 2 17 75 68 7 11 1 146 WS-E Fe2O3T 1994 13.15 5 9 22 57 51 6 5 98 OU-1 Fe2O3T 1997 8.99 2 4 11 35 2 54 mg g21 G-1 Ba 1950 1080 3 3 AN-G Ba 1980 34 6 6 4 1 2 13 2 11 4 36 AC-E Ba 1987 55 10 4 16 3 34 3 31 11 78 WS-E Ba 1994 338 6 21 12 46 18 28 5 90 OU-1 Ba 1997 131.4 1 15 8 25 3 52 G-1 Eu 1950 1.22 0 AN-G Eu 1980 0.37 1 8 9 AC-E Eu 1987 2 1 13 2 1 3 21 41 WS-E Eu 1994 2.25 13 18 8 39 OU-1 Eu 1997 0.52 5 19 6 30 G-1 Hf 1950 5.4 0 AN-G Hf 1980 0.38 1 8 9 AC-E Hf 1987 27.9 1 13 2 1 3 21 41 WS-E Hf 1994 5.3 1 16 9 4 5 9 35 OU-1 Hf 1997 1.65 1 16 7 24 G-1 Sr 1950 248 3 3 AN-G Sr 1980 76 12 6 4 1 1 28 2 26 1 53 AC-E Sr 1987 3 1 9 1 12 2 1 35 3 32 61 WS-E Sr 1994 410 1 7 20 18 60 20 40 2 108 OU-1 Sr 1997 104.8 2 12 8 29 1 52 G-1 Zr 1950 201 3 3 AN-G Zr 1980 11 4 2 1 23 2 21 30 AC-E Zr 1987 780 3 14 2 1 53 3 50 6 79 WS-E Zr 1994 195 1 16 11 54 16 38 2 84 OU-1 Zr 1997 55 1 8 9 29 47 * Chem.= classical/rapid techniques; Phot. = flame photometry; AAS = atomic absorption spectrometry; Arc-AES = dc arc atomic emission spectrometry spectrography; ICP-AES = inductively coupled plasma atomic emission spectrometry; DCP-AES = dc coupled plasma atomic emission spectrometry; SSMS = spark source mass spectrometry/ spectrography; TIMS = thermal ionisation mass spectrometry; ICP-MS = inductively coupled plasma-mass spectrometry; XRF = X-ray fluorescence analysis; XRF-g = laboratories undertaking XRF analyses on a glass disc; XRF-p = laboratories undertaking XRF analyses on a powder pellet; INAA = neutron activation analysis; Probe = electron microprobe analysis.Analyst, November 1997, Vol. 122 1181AES, 1975; ICP-MS, 1985. To simplify the interpretation of trends, tie-lines have been drawn between successive data points, although these tie lines do not necessarily imply a continuous trend in the popular use of the respective techniques. The principal features of the trends in these data are as follows. Major elements (represented by SiO2 and Fe2O3) [Figs. 1(a) and (b)] Although wet chemical procedures were the only reliable methods available in 1950, these procedures declined rapidly in popular use as the majority of laboratories turned to XRF methods once appropriate procedures had been fully established. The earlier interest in the use of AAS techniques (especially indicated in the determination of total iron), has declined in recent years. Apart from XRF, the other technique that now makes a significant contribution to the determination of the major elements is ICP-AES, although the proportion of laboratories reporting results by this technique is significantly lower than that by XRF.Trends in these data indicate that XRF will continue to be the technique chosen by the majority of laboratories for the routine determination of the major elements. Sr, Ba, Zr [Figs. 1(c), (d) and (e)] This group of elements is also dominated by contributions of laboratories using the XRF technique with ICP-AES also making a significant contribution.In recent years there has been Table 3 Elements for which 20 or more laboratories reported results in the cited reference material characterisation and proficiency testing programmes Year Sample Elements with 20 or more reported results 1980 AN-G Ba, Co, Cr, Li, Ni, Pb, Rb, Sr, V, Y, Zn, Zr 1987 AC-E As above for 1980 plus: Be, Ce, Cs, Dy, Eu, Ga, Gd, Hf, La, Lu, Nb, Nd, Sc, Sm, Ta, Tb, Th, U, Yb 1994 WS-E As above for 1987 plus: Er, Ho, Mo, Pr, Sn, Tm (but not including: Be, Cs, Li) 1997 OU-1 As, Ba, Ce, Co, Cr, Cu, Er, Eu, Ga, Gd, Hf, Ho, La, Lu, Nb, Nd, Ni, Pb, Pr, Rb, Sc, Sm, Sr, Tb Th, U, V, Y, Yb, Zn, Zr Fig. 1 Trends in the use of selected analytical techniques. The proportion of laboratories reporting results by specified techniques are plotted as a function of the year of publication of the appropriate reference material characterisation study of proficiency test round. Data are plotted for the elements (a) SiO2 (b) Fe2O3 (c) Sr (d) Ba, (e) Zr (f) Hf and (g) Eu. 1182 Analyst, November 1997, Vol. 122a decline in the popularity of AAS and INAA (where applicable), but an increase in the use of the ICP-MS technique. Trends in the data would indicate a continuing dominant role for XRF analysis. However, these trends also indicate that the role of ICP-MS in the determination of these elements will increase and it will be interesting to see if part of this increase will be at the expense of ICP-AES.Hf and Eu [Figs 1(f) and (g)] The most popular technique for the determination of these elements in the early 1980s was INAA, with a significant contribution from ICP-AES during the mid-1980s. However, there has been a dramatic and progressive decrease in the popularity of INAA in recent years and an equally dramatic increase in the reporting of data by ICP-MS. This change may, in part, be associated with a decline in the number of research reactors with facilities for the irradiation of INAA samples.However, a significant contribution is likely to be a recognition of the extended capabilities of ICP-MS for the determination of a wide range of trace elements as well as the attractive option of extending the capability of the technique using laser ablation. Trends in data indicate that this change is likely to continue, with ICP-MS increasingly becoming the dominant technique for this category of measurement. Overall Although data are plotted in Figs. 1 (a)–(g) only for selected elements, these results indicate that whereas in the 1980s XRF and INAA represented the combination of techniques that were most frequently selected by geochemical laboratories to contribute most routine geochemical data, XRF and ICP-MS now represent the two techniques with largely complementary capabilities that are increasingly chosen to satisfy this role. Trends in Isotopic Techniques The treatment of data representing the use of isotopic techniques is presented in a different way in Figs. 2 (a)–(d). Data abstracted from Geochimica et Cosmochimica Acta for individual categories of measurement are plotted as a series of histograms, where each bar of the histogram represents the number of contributions made in annual volumes for the years 1970, 1980, 1987, 1994 and 1996, respectively. The categories of measurement are as follows. Fig. 2(a): stable isotopes (H, C, N, O and S). Fig. 2(b): noble gases (He, Ne, Ar, Kr and Xe).Fig. 2(c): radiogenic isotopes and isotopes used in dating applications (14C, K–Ar, Ar–Ar, Rb–Sr, Pb–Pb, U–Pb, Th–Pb, U disequilibrium studies, Sm– Nd, Re–Os). Rare earth element geochemistry studies have been included as an indicator of continuing use of conventional geochemical modelling as well as citations for artificial and cosmogenic isotopes and the use of Pb isotopes in tracer studies. Fig. 2(d): more specialised studies using other isotopic systems, mainly related to ‘difficult’ stable isotope systems and isotope systems where minute differences in isotopic abundance can be related to nucleosynthetic processes relevant to the early formation of the Solar System.These data indicate a sustained contribution of isotopic techniques to geochemical research over the years in question. Particular features worthy of note are as follows. (i) Continuing interest in the contribution of stable isotope studies (based on gas source mass spectrometry) with a significant increase in the popularity of oxygen isotope techniques [Fig. 2(a)]. (ii) A relatively consistent contribution from noble gas studies, no doubt reflecting the source journal’s interest in cosmochemistry [Fig. 2(b)]. (iii) A comparative decline in the popularity of K– Ar dating studies, replaced by more the precise Ar–Ar measurements, which in term have been augmented in recent years by the laser extraction Ar-Ar technique [Fig. 2(c)]. (iv) The comparative recent exploitation of Nd–Sm and U–Pb procedures following the development of higher precision thermal ionisation mass spectrometers in the 1980s [Fig. 2(c)]. (v) The even more recent exploitation of the Re–Os geochronometer, following the development of negative ion TIMS instrumentation in the 1990s [Fig. 2(c)]. (vi) A comparative increase in popularity of uranium disequilibrium procedures, noting that the data plotted in Fig. 2(c) represents the combination of determinations made by alpha counting in earlier years and the more recent higher precision measurements made by TIMS instrumentation.(v) It is also interesting to note that despite the introduction of higher sensitivity mass spectrometer instrumentation capable of exploiting new isotope systems, interest in the more traditional Rb–Sr and REE measurements has continued to expand, although the former would now normally be undertaken on individual mineral separates, rather than on bulk samples as in former years [Fig. 2(c)]. (vi) These data do not differentiate and therefore adequately credit the contribution of ion probe techniques, including SHRIMP, to the U–Pb dating studies, particularly of zircons. Overall Trends in Isotopic Analysis One trend that is apparent from data presented in Fig. 2 is that technical advances in mass spectrometer instrumentation have had a significant impact by permitting the exploitation of new isotope systems. Two examples, mentioned above, relate to advances in thermal ionisation mass spectrometry instrumentation.Higher precision instrumentation developed in the 1980s permitted routine characterisation of the Sm–Nd (and U–Pb) isotope systems and negative ionisation technology, developed in the 1990s, allowed routine measurement of the Re–Os system. In terms of future trends, therefore, it seems likely that technical developments, particularly improvements in the precision of measurements, could be influential in allowing measurements to be made of isotope systems that are beyond the capabilities of the present generation of instrumentation.In so doing, new categories of geochemical data would become available. The impact of these developments is most likely to see the exploitation of isotope systems such as those listed in Fig. 2(d). As an example of this category of measurement, Sharma et al.14 investigated the 146Sm isotope system. This isotope has a half-life of 149 Ma and is now extinct.However, if TIMS measurements can be made to a precision that is three times better than at present, the potential exists to investigate variations in progeny 142Nd isotopic composition that occurred during the first few hundred million years after the formation of the Universe, thus providing new geochemical data on nucleosynthetic processes. The second trend illustrated by data in Fig. 2 is the continuing role for isotopic (and other geochemical) measurements.Although the pioneering days in developing and applying both Rb–Sr isotope system and rare earth element geochemistry are long since passed, the evidence is that both of these categories of measurement have a continuing and enduring role to play in geochemical research. One can conclude, therefore, that almost all the isotope measurements that have had a significant impact in the past will continue to contribute to future geochemical studies. However, the influence of technical advances that offer improvements in the quality of geochemical information should not be overlooked.Of particular relevance are the improvements that would arise if all isotope measurements could be obtained on a microprobe scale. One example that has already arisen through both advances in the instrumentation capabilities and the development of appropriate techniques is the progression from K–Ar to the Ar–Ar dating technique and the Analyst, November 1997, Vol. 122 1183subsequent improvements associated with the further development of the laser extraction Ar–Ar microprobe technique. Evolution of Microanalytical Techniques Although not directly apparent from the data presented above for bulk and isotopic techniques, the advance in analytical technology that is likely to have the most influential and enduring effect on geochemical research is that of microprobe instrumentation. To a large extent, the capability now exists to undertake many of the analytical measurements normally performed on bulk samples using advanced microprobe techniques (see contributions in Potts et al.15).These techniques provide analytical information on individual minerals, normally Fig. 2 Histograms showing the number of citations of isotopic abundance or ratio measurement for the years 1970, 1980, 1987, 1994, 1996. Data were collated by examining the titles of research contributions published in the corresponding annual volumes of the journal Geochimica et Cosmochimica Acta.Citations of the following isotopic systems are plotted: (a) stable isotopes; (b) noble gases; (c) radiogenic isotopes (plus 14C, REE, artificial isotopes, cosmogenic isotopes and Pb isotope tracer studies); and (d) other isotopic measurements. 1184 Analyst, November 1997, Vol. 122in situ in thin section, and are capable, therefore, of providing much more detail about the geochemical history of a sample in comparison with bulk measurements.A summary of the analytical capabilities of some of these techniques is as follows. Electron microprobe Although restricted to the determination of major and minor elements at spatial resolutions of about 2–3 mm, the electron microprobe is the only microprobe method capable of quantitative analysis without having to resort to internal standardisation of the analytical signal or comparison with matrix matched calibration samples. In fact, a reliable physical model exists for the excitation process, and forms the basis for the most common type of matrix correction procedure.The electron microprobe, therefore, occupies a pivotal role in microanalysis. Modern instrumentation offers the capability of high resolution backscattered electron imaging and X-ray mapping, and is capable of revealing zoning effects that were overlooked on earlier generations of instrumentation. The electron microprobe occupies an analogous position in microprobe studies to XRF in the routine bulk analysis of silicate samples.Ion probe By ablating and exciting a sample with a beam of ions (often O2), focused to typically 10–25 mm and measuring the secondary ion mass spectrum using a double focusing (high resolution) mass spectrometer, high precision measurements can be made of the isotopic and trace element composition of samples to sub-mg g21 detection limits. An important influence on these studies has been the application of the SHRIMP ion probe to geochronology studies, mainly based on the analysis of zircon.Trace element calibration generally has to be achieved by comparing normalised intensities to those of a matrix matched mineral standard. Laser ablation ICP-MS This technique is already widely used for the trace element characterisation of minerals by the ablation of material using a laser focused to, typically, 50 mm diameter and analysing directly the ablated material using an ICP-MS. Calibration is normally achieved by internal standardisation.Conventional instrumentation is based on a general purpose quadrupole mass spectrometer with a single electron multiplier detector. However, laser ablation ICP-MS instrumentation using a high resolution magnetic sector or double focusing mass spectrometer with multiple collection has as yet an only partially fulfilled potential for measuring isotopic compositions to precisions approaching that of TIMS and the ion probe, so providing a general capability for the determination of isotope ratios on a microprobe scale. Laser Ar–Ar mass spectrometry The quality of data obtained when a laser is used to extract argon from a small region of an irradiated sample is far superior to that from the bulk analysis of samples using the 39Ar : 40Ar technique. The confidence of these measurements can be further increased if the laser power is controlled to give stepwise heating and extraction of the argon gas. In this way, the isotopic history of a sample can be characterised in much greater detail by the detection (if present) of variations within individual mineral grains, as well as the identification of atmospheric contamination.This microanalysis development demonstrates the power of laser extraction procedures and has discredited, to some extent, conventional bulk measurements of isotopic composition in the same way that the detection of zoning effects in zircon using the ion probe has discredited isotopic measurements based on the bulk isotopic composition of this mineral.Laser fluorination An active area of development is in the stable isotope characterisation of individual minerals using laser induced fluorination and laser extraction techniques and high-precision gas-source mass spectrometric detection. Although, like LA– ICP-MS, the spatial resolution of such measurements is currently not as good as with other microbeam techniques (see, for example, Wright16), these developments are likely to lead to similar advances in the geochemical interpretation of stable isotope ratios, as has occurred with other microprobe techniques.Synchrotron X-ray microprobe A synchrotron is a facility offering X-ray beams of very high intensity that have very low divergence (i.e., are near parallel) and are polarised. These beams can be focused onto the surface of a sample, facilitating both the determination of the trace elements by X-ray fluorescence and nearest-neighbour coordination information by X-ray absorption measurements.Submicrometre beams can be achieved on the latest, third generation, synchrotrons, although taking into account the penetrating properties of X-rays, the depth analysed is usually controlled by the thickness of the sample. Future Development of Microprobe Techniques Much of the earlier developments in geochemistry have been based on analytical data obtained from bulk methods of analysis.However, with the introduction of successive microprobe techniques, it has been clearly demonstrated that a much greater wealth of information can often be obtained from the microanalysis of individual minerals. This information can show the presence of mineral zoning effects that were often previously unsuspected and indeed could not readily be identified from bulk methods of analysis. In fact, results from these microprobe techniques have, in many cases, discredited equivalent determinations on bulk samples (as in the ion probe analysis of zircon or laser Ar–Ar studies on individual minerals).It seems likely, therefore, that as these techniques are developed further, geochemistry will increasingly be based on results obtained by microprobe analysis because of the greater detail that can then be obtained about the geochemical history of samples. Conclusions An analysis of trends in the use of geoanalytical techniques over the last 25 years shows that XRF and ICP-MS are likely to represent the combination of instrumentation dominating the production of routine major and trace element data in future years. In an equivalent evaluation of trends in isotopic methods of analysis, improvements in mass spectrometry instrumentation have had a large influence on the geochemical applications. This observation is illustrated by the introduction of both higher sensitivity thermal ionisation instrumentation in the 1980s, which facilitated the development of the Sm–Nd and U–Pb radiogenic isotope systems, and negative ionisation instrumentation in the 1990s and its application to the Re–Os system. If these advances in sensitivity continue, it is likely that analytical capabilities will extend to new isotope systems that are beyond the capabilities of current instrumentation. However, the greatest impact in future is likely to result from the further development and geochemical application of data from Analyst, November 1997, Vol. 122 1185a range of microprobe techniques. An overview of the current capabilities of techniques like the ion probe and laser Ar–Ar dating already shows that microprobe techniques have discredited equivalent measurements made on bulk samples. This trend is likely to continue with further refinements in microprobe instrumentation and geochemical models for interpreting microanalysis data. The author would like to thank the organising committee of Geoanalysis 97 for the opportunity to present this paper and to B. Chappell (ANU) for valued comments on an earlier version of the manuscript. References 1 Goldschmidt, V. M., and Thomassen, L., ‘Geochemische Verteilungsgeesetze der Elemente III’, Videnskabsselsk Skr. 1. Mat.- Naturvidensk Kl., 1924 (5), 24; as quoted in Mason, B., Victor Moritz Goldschmidt: Father of Modern Geochemistry, The Geochemical Society, San Antonio , TX, USA, 1992, pp. 1–184. 2 Masuda, A., Earth Sci. Nagoya Univ., 1962, 10, 173. 3 Potts, P. J., Hawkesworth, C. J., van Calsteren, P., and Wright, I. P., in Magmatic Processes and Plate Tectonics, ed. Prichard, H. M., Alabaster, T., Harris, N. B. W., and Neary, C. R., Geological Society, London, UK, 1993, p. 501. 4 Boumans, P., J. Anal. At. Spectrom., 1993, 8, 767. 5 Hall, G. E. M., J. Geochem. Explor., 1996, 57, 1. 6 Hiefje, G. M., J. Anal. At. Spectrom., 1989, 4, 117. 7 Fairbairn, H. W., Schlect, W. G., Stevens, R. E., Dennen, W. H., Ahrens, L. H. and Chayes, F. US Geol. Surv. Bull., 1951, 980. 8 Ahrens, L. H., Geostand. Newsl., 1977, 1, 157. 9 Shapiro, L., and Brannock, W. W., U.S. Geol. Surv. Bull., 1962, 1144-A. 10 Govindaraju, K., Geostand. Newsl., 1980, 4, 49. 11 Govindaraju, K., Geostand. Newsl., 1987, 11, 203. 12 Govindaraju, K., Potts, P. J., Webb, P. C., and Watson, J. S., Geostand. Newsl., 1994, 18, 211. 13 Thompson, M., Potts, P. J., Kane, J. S., Webb, P. C., and Watson, J. S., Geostand. Newsl. J. Geostand. Geoanal., 1997, in the press. 14 Sharma, M., Papanastassiou, D. A., Wasserburg, G. J., and Dymek, R. F., Geochim. Cosmochim. Acta, 1996, 60, 2037. 15 Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., The Mineralogical Society and Chapman and Hall, London, UK 1995, pp. 1–419. 16 Wright, I. P., in Micropole Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., The Mineralogical Society and Chapman and Hall, London, UK, 1995, ch. 9. Paper 7/04856D Received July 8, 1997 Accepted August 26, 1997 1186 Analyst, November 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a704856d
出版商:RSC
年代:1997
数据来源: RSC
|
2. |
A Brilliant Future for Microanalysis?† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1187-1192
Richard W. Hinton,
Preview
|
|
摘要:
A Brilliant Future for Microanalysis?† Richard W. Hinton Department of Geology and Geophysics, University of Edinburgh, West Mains Road, Edinburgh, UK EH9 3JW The ability to measure trace element distributions on the microscale is often critical to our understanding of large scale geological processes. As the number of different instruments and techniques capable of trace element analysis increases the choice of analytical method is becoming less clear. Although most techniques have 1 micrometre spatial resolution as their ultimate goal it is clear that the analytical depth is determined by sample thickness (proton probe and synchrotron XRF) or analytical volume (ion probe and laser ablation-ICP-MS) considerations.Thus ability to produce brighter and smaller beams also requires significant improvement in the detection of secondary particles. Provided analytical considerations, such as precision or detectibility are met, the eventual choice between competing techniques may well be more mundane aspects such as ease of access and cost.Keywords: Microanalytical techniques; ion probe; laser ablation inductively coupled plasma mass spectrometry; proton probe; synchrotron X-ray fluorescence; geological analyses The ability to analyse a wide range of trace elements on the microscale has become an important part of our understanding of geological processes. Both instrument development and microanalytical methods have progressed steadily over the past two decades to give greater element coverage, lower detection levels and smaller beam sizes. Earth scientists can now choose between a variety of microanalytical methods, however, information is often more readily available on ultimate performance (under ideal conditions) than routine capability.The general rule of thumb, that the instrument size must increase to permit greater microanalytical sensitivity, has resulted in very large (and expensive) instruments and facilities. There is a danger that the smaller number of very costly facilities may result in significant unsatisfied demand.However, it is clear from the continued use of older, smaller instruments that many applications do not require the most sensitive instruments and that access, flexibility, and cost are often the dominant factors affecting choice of technique. Future management of access may require directing research between techniques to make the most efficient use of limited resources.Microanalytical Techniques for Multi-element Analysis In this paper only techniques which permit the simultaneous microanalysis of a number of trace elements will be considered, namely the ion probe, the laser ablation (LA)-ICP-MS, the proton probe and synchrotron XRF. An extended discussion of all these techniques can be found in Potts et al.1 General descriptions are given below: (1) The ion probe2 uses secondary ion mass spectrometry (SIMS) to detect ions physically removed from the surface by a high energy ion beam. Nearly all geological laboratories use a focused beam of 16O2 to sputter atoms from a (polished) flat surface.Ions which are formed in the sputtering process can be accelerated into a double focusing mass spectrometer and separated on the basis of both mass and energy. All masses from H to U are ejected, with the degree of ionisation principally being related to the elements’ first ionisation potential.The mass spectrum created by ion bombardment contains not only the single ions of individual elements but also many molecular species. These molecular species must be separated either by measuring only high energy ions (and thereby losing 95–97% of the original signal) or high mass resolution (where losses are related to the size of the instrument). Although ion probes with very large mass spectrometers are often capable of separating individual elements from molecular species, without significantly reducing the number of secondary ions transmitted, these expensive instruments are principally used for isotopic analyses (e.g., U–Pb dating of zircons).Samples are coated (generally with Au) as the surface of the sample is held at a high voltage. (2) LA-ICP-MS3 combines a laser source with a conventional inductively coupled plasma mass spectrometer. A finely focused laser beam is used to ablate the surface of a sample held in an argon filled cell.Solid particles ablated from the surface are transported in Ar gas into an inductively coupled plasma where they are ionised. As the ionisation process produces a lower proportion of molecular species than SIMS most trace element measurements can be made using low resolution quadrupole mass spectrometers. (3) In a manner similar to the electron probe,4 the proton probe5 uses a finely focused beam of very high energy protons (2–4 MeV) to excite atoms and generate X-rays. This technique is often called particle (or proton) induced X-ray emission (PIXE).Characteristic X-ray lines of individual elements are detected by either solid state detectors (known as energy dispersive spectrometer or EDS) or a combination of curved crystal and proportional counter (known as wavelength dispersive spectrometer or WDS). The EDS system detects the entire range of X-rays simultaneously whilst the WDS system relies on the positioning of a curved crystal to measure one Xray line at a time.The electron microprobe has a high X-ray background (Bremsstrahlung); the X-rays are generated as the electrons decelerate. Heavier protons decelerate much more slowly consequently the proton probe has a much lower intrinsic X-ray background. However, the background is not completely absent as some X-rays are generated by electrons ejected by the proton beam interaction. The commonly used EDS method of X-ray detection has relatively poor X-ray energy resolution therefore the unresolved overlap of X-rays lines of different elements is not uncommon, especially in complex substrates.(4) The synchrotron X-ray fluorescence technique6 is analogous to the laboratory XRF except that a synchrotron source provides many orders of magnitude more X-ray photons to excite X-ray fluorescence in the target material. The synchrotron X-ray source provides polarised light either as a ‘white’ Xray continuum or at well defined energies (monochromatic excitation).The photon beam is highly concentrated and † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1187–1192) 1187Instrument Time 1993 - 1997 Edinburgh NERC Ion Microprobe Facility micrometre sized X-ray beams can be produced either by apertures or by focusing using mirrors, curved lenses or capillaries. X-rays are detected using EDS and WDS systems similar to those used for both the proton and electron probes; the detectors used on the new generation of instruments must be capable of detecting very high energy X-rays.Most of the X-ray background is generated by scattering, however, as the synchrotron photons are polarised the scattering is at a minimum at 90° to the orbital plane of the electrons or positrons which generate the X-ray photons (i.e., horizontal plane). Thus by having the detector at right angles to the incident X-ray beam the scattered X-ray background is reduced (effectively by a factor of 10, Haller and Kn�ochel7).This geometry requires that the sample be at 45° to the incident beam; this has obvious implications for the micro-analysis of thick samples. Although not capable of element analysis at very low concentrations, the electron probe is invariably used to determine the major and minor chemistry of all samples analysed by these techniques and is an essential part of all microanalytical studies.Access An earth or environmental scientist who has a specific analytical problem is principally concerned with two factors: (1) can a particular method make the analysis to the precision and accuracy required on their samples? and (2) can they obtain access at either no, or reasonable, cost? There are numerous examples in the literature where a particular technique has been used, not becausit is the best method, but because of ready access. Unlike the electron microprobe and SEM which are generally accessible for most earth scientists, either within-house or at local centres, access to the modern microanalytical instruments usually requires formal application to a national centre or an informal approach to the principal scientist in charge of a laboratory.It is inevitable that such instruments will be oversubscribed and access may well require a formal proposal which is subject to peer review. For example the Edinburgh Ion Microprobe Facility (IMF) has been operating as a national centre for ten years using a Cameca ims-4f controlled by Charles Evans and Associates operating system.UK based earth scientists can obtain access to the IMF by making a formal proposal to the Scientific Services division of the Natural Environment Research Council (NERC). Large projects requiring substantial use of the instrument are funded directly by NERC Research Grants. Thus virtually all access is by peer reviewed proposals and is in competition with other earth scientists.In contrast, techniques such as the synchrotron XRF are dependent on access to very large facilities which may be funded by a number of different Government agencies (or indeed a number of Governments). Unless part of the facility (e.g., a beam line) is specifically funded for earth science applications, access may not only be in competition with scientists from many disciplines but may appear very expensive on a per project basis (see Dove and Redfern8).Many laboratories expect running costs to be met by external users, a proportion of staff costs may also be included if analyses are carried out by ‘in-house’ staff. The approach to providing access for a wide number of scientists to expensive microanalytical techniques varies from laboratory to laboratory. Many laboratories provide only limited access to external users while others provide the majority of the time to ‘external’ users.The IMF at Edinburgh averages over 35 different projects a year ranging through oxygen isotopes, trace element and light element analysis, to diffusion studies. Training is as an important part of the laboratory operation and a ‘hands-on’ approach is adopted for nearly all users. However, while improving the users knowledge of the technique, the ‘hands-on’ approach does have the drawback that any advances in technique must become relatively routine before being offered to inexperienced users.Thus techniques are often developed by ‘research grant’ users before being offered as part of the IMF service. Other laboratories have ‘in-house’ operators with the user choosing the elements required and the points to be analysed. While less time is lost in setting up the instrument, unless the user is on site when analyses are made, there is potential for some analyses to be of incorrect locations. Many laboratories will have a dedicated research programme with a restricted number of researchers, including post-graduate students, having ‘handson’ use of an instrument.Access to outside users will in this case be limited, however, it is often still possible to gain access to such laboratories if the project is scientifically challenging. Over the 10 years that the IMF has been in operation at Edinburgh there have been in excess of 300 requests for information about use of the ion microprobe, with over 100 resulting in formal applications for access.Initial approaches for access vary through very specific problems from those knowledgeable about the technique, to vague inquiries about its general capabilities. However, it is clear that even when applicants have reasonable knowledge of a technique, each study may have unique features which can create analytical difficulties. A breakdown of the ion microprobe usage over the last four years (Fig. 1) demonstrates not only the strengths of the technique but the awareness within the geological community.About 50% of the time was spent on rare earth element (REE) and high field strength element (HFSE) characterisation of igneous and metamorphic systems including both natural rocks and experimental run products. In principal many of these analyses could have been done on the first generation instruments over the last 15 years. Demand stays strong due to the continued demonstration of the importance of the microscale variability in light and trace elements both within and between individual minerals.The ion probe continues to be one of the most sensitive instruments for the measurement of volatile elements e.g., hydrogen and carbon (11% of instrument time). The relatively low instrument time (17%) devoted to isotopic analysis in part reflects the difficulty of these analyses (especially for inexperienced operators) but also the higher operational costs of this type of analysis.Although sedimentalogical applications have been relatively low, these are now showing signs of increase. While this may be largely due to improved awareness of the technique, it may also reflect increased funding for projects related to environmental change. Somewhat surprisingly, even though the ims-4f was designed as an ion microscope very little instrument time has been dedicated to imaging. The reason for this is relatively simple: the nonconducting nature of silicates leads to sample charging problems when large areas of the sample are bombarded with Fig. 1 Instrument usage of the ion microprobe over the last four years. 1188 Analyst, November 1997, Vol. 122charged particles (the conductive coating must be removed by the incident beam prior to analysis). Sample Preparation One of the first analytical concerns is whether samples of the rock or mineral can be prepared for each particular method and whether the analysis must be non-destructive. This aspect may be far from trivial as sometimes the sample has already been prepared and is so ‘unique’ or highly valued by its owner that little or no modifications can be made to it.Ideally specimens are prepared for the ion microprobe as 1 inch round, polished, thin sections. All samples must be compatible with high vacuum operation. Samples prepared for the electron microprobe are usually compatible with the ion microprobe (subject to high vacuum compatibility). The sample surface must be optically flat, not just over the area to be analysed but over an extended region (mm) since the ims-4f operates with high voltage extraction field and the (Au coated) sample surface acts as part of the secondary ion extraction optics.As it is quite possible that a sample will be subject to a number of different analytical techniques, compatibility between techniques should be considered at an early stage. Clearly the most destructive technique must be kept until last.The ion probe sputters the sample away relatively slowly and rarely sputters to greater than 10 mm depth. An ion beam of 15–25 KeV energy will mix a layer of approximately 0.01 mm deep beneath the surface. Atoms beneath this depth will be unaffected by the ion beam therefore the volume sampled is essentially that sputtered. Samples for LA-ICP-MS analysis are usually made relatively thick (100s mm) as the laser ablation process often reaches over 50 mm depth in the few minutes required for an analysis. As the efficiency of this method improves the thick sample requirement is relaxing and samples of normal thickness (30 mm) have been analysed.Samples for synchrotron XRF must be mounted with the assumption that the beam goes right through the sample (i.e., penetrates greater than 100 mm). Mounting thin slices on organic mounting medium such as Kapton ensures that X-rays are generated from the sample alone.7 Samples are usually held in air, however, compatibility with other techniques may require that the sample be compatible with high vacuum operation.The proton probe samples depths of between 25 and 100 mm depending on the sample’s composition and the element measured. Samples for the proton probe are preferably thin (@10 mm thick) and mounted on a substrate which will not give additional X-ray background. Mounting on pure silica slides is one possible method. In some cases samples which have already been mounted on normal glass slides have been analysed and then the contribution from the glass slide subtracted from the concentrations measured.9 Clearly, although this must result in reduced precision, and should be avoided if at all possible, it may be the only practical solution for a unique sample.The quantitative analysis of thick samples using both the synchrotron XRF or the proton probe creates greater difficulties for correction procedures but is not a major barrier to analysis.Although it may appear obligatory that the phase to be analysed is at the surface of the sample there are a few situations where this is not possible. For example, fluid inclusions can only be analysed in their original location within their host mineral. The ion probe primary beam can sputter away the sample until the fluid inclusion is reached releasing the gas and liquid in a short burst.10 However, the signal obtained as the inclusion bursts is very short lived and measurement of the individual ion intensities is very difficult. Further, quantification is extremely difficult as the release cannot be simulated under standard conditions.In contrast the deep penetration of synchrotron radiation or the proton beam is a positive advantage since the inclusion can be analysed in situ.5,11 In these studies analytical problems are associated with the attenuation of Xrays within the host mineral and quantification is far from routine.Beam damage to geological materials is generally not a problem for either the synchrotron XRF or the proton probe. While the LA-ICP-MS and ion probe are destructive techniques the amounts of material removed is less than 10 ng for the ion probe and 100–500 ng for the LA-ICP-MS. Analytical Spot Sizes and Volumes Analytical spot sizes and volumes can be a relatively contentious issue, in part because comparisons can rarely be made where all the parameters are the same for each technique.Clearly, while micro-XRF using the synchrotron is predicted to achieve sub-micrometre spatial resolution, the analytical volume is limited by the thickness of the section presented to the beam or the absorption of the secondary X-rays by the sample itself. Similarly, while the proton probe can achieve 1 mm spatial resolution, X-rays are generated from at least 25 to 100 mm into the sample depending on the element (the sampled volume being element and matrix specific).In contrast, the ‘destructive’ techniques excavate a crater which can be measured optically or with an SEM. Reasonable estimates can also be made on crater depths by either drilling through the entire sample of known thickness (LA-ICP-MS) or physically measuring the depth of large pits (using a rastered ion beam) to determine sputter rates (ion probe). In both cases the number of atoms removed from the sample can be calculated and estimates made of the efficiency of both ion formation and collection. Definition of spot widths can vary from those preferred by manufacturers (most optimistic, the width at half peak maximum, Fig. 2) to the more realistic (the width at 10% of maximum beam intensity). Overlap of the low intensity fringe of the beam (Fig. 2) may be highly significant when large differences exist over short distances (e.g., grain boundaries). All techniques should perhaps include the total width or volume with which the beam interacts either through direct bombardment (i.e., ion probe), heating (LA-ICP-MS) or secondary Xrays and electrons (synchrotron XRF and proton probe).Sources of Energetic Particles For all techniques there will be an anti-correlation between spatial resolution and the intensity of the excitation source as measured at the sample. Thus although the proton probe is capable of a spatial resolution of 1 mm, the beam diameter is Fig. 2 Beam current distribution of primary ions. True beam diameter is total area from which ions are removed rather than the optimistic width at 50% peak height (not infrequently used by manufacturers) and the width at 10% peak height used by the laboratories.Analyst, November 1997, Vol. 122 1189normally !20 mm (approximately 0.1 compared to 6–8 nA beam currents, respectively) to give a sufficient number of Xrays to detect very low concentrations. Similarly, an ion microprobe beam of less than 5 mm is feasible with an O2 source but normally 10–25 mm is used for trace element analysis.A proton beam of 3 MeV excites the K X-ray lines only up to about z = 42. Heavier elements must therefore be analysed by their L lines (which may be overlapped by the K lines of lighter elements); the generation of X-rays from elements between Mo and Lu is particularly poor. The new generation synchrotron X-ray facilities, e.g., the Advanced Photon Source (APS) and European Synchrotron Radiation Facility (ESRF) produce not only a higher total numbers of photons but they are also concentrated into a smaller area (i.e., not only have high brightness but also high brilliance6).Focusing of X-ray beams has become more advanced with focusing using curved crystals or capillaries giving orders of magnitude higher intensities than apertures. Small diameter beams therefore do not necessarily entail prohibitive decreases in beam intensity. While the new generation sources only produce about factors of 2 to 3 more photons at low energy they produce many orders of magnitude more high energy photons.Both the APS and the ESRF facilities will therefore be able to excite the K lines of all elements, including U. The ability to measure the heavier elements (z > 40) on their K lines should improve their minimum detection limits (MDL) by at least an order of magnitude. In contrast, the duoplasmatron gas source which provides the O2 primary beam for the ion probe has not changed significantly over the last 20 years.However, since SIMS is a destructive technique, unless the measurement efficiency is improved the volume of material sputtered must be kept constant. Thus if it were possible to focus a 10 nA beam into 1 mm area, the beam would be required to penetrate 100 mm depth to maintain the same total secondary ion counts. Despite these reservations a more intense source would considerably improve instrument operation, including better beam shapes and greater flexibility, including faster analyses.The proton source also uses a duoplasmatron source and is not expected to undergo any major changes in the near future. Originally LA-ICP-MS relied on infra-red lasers to ablate the surface, however, they did not always create small pits, especially in transparent or highly cleaved minerals. Fortunately UV lasers have been found to give much more controlled ablation and smaller craters even in transparent materials. Laser technology for the LA-ICP-MS is still improving and it is likely that new pico- and femtosecond lasers will permit smaller beams, better coupling with transparent materials and lower matrix effects.If these improvements can be coupled with improvements in sensitivity of the ICP-MS system as a whole, smaller beams will not necessarily have to be compensated for by increasing depth of sampling. Secondary Losses The ultimate limitation for both the ICP-MS and the ion probe is the efficiency of the ionisation process.In SIMS only a relatively small proportion of atoms sputtered are ionised. The secondary ionisation is strongly dependent on the first ionisation potential of each element and variations of over 3 orders of magnitude exist in the secondary ionisation efficiency. This has major consequences on the detection limits for certain elements, particularly the heavier transition elements. The laser ablation creates solids and gases which are transported in an Ar gas flow into the ICP-MS plasma.It is thought that 25–100% of the material ablated by the laser beam is transported to the plasma of the ICP-MS. However, losses within the plasma interface and within the mass spectrometer must be high as less than 0.1% of atoms ablated are measured. The ionisation in the inductively coupled plasma is far less dependent on the first ionisation potential than for SIMS and the variations in ionisation efficiency between elements is generally less than a factor of 10. However, while the secondary ion current produced by ion bombardment is relatively stable, the laser ablation process produces an ion signal which changes rapidly with time even for homogeneous substrates.These time dependent effects in the laser ablation and ionisation process are not the same for all elements. Time dependent variations of between factors of 2 to 3 can occur between elements (particularly between siderophile and chalcophile elements12).Instrument tuning for maximum ionisation efficiency in the ICP-MS is also mass specific and if a large element range is required there must be some compromise over the efficiency of ionisation at the extremes of mass range. Overall, the strong relationship between first ionisation potential and detection efficiencies for the ion microprobe makes this technique’s capabilities much more element specific than those of the LA-ICP-MS. In particular the heavier transition elements Cu, Zn, Rh–Ag, Ir–Au all have poor sensitivity on the ion microprobe.Conversely the alkalis, the alkaline earths and the rare earth elements have higher sensitivity on the ion probe than the LA-ICP-MS. Detector Systems/Mass Spectrometer Limitations For trace element analyses by destructive techniques coupled with mass spectrometric measurement on a single collector the ultimate detection level is in part dependent on the number of elements analysed (i.e., the duty cycle).The ultimate mass spectrometer would therefore be one that operates with high transmission at very high mass resolving power, measuring all elements of interest simultaneously. This is far from being realised. High sensitivity at high mass resolving power has been possible on the ion microprobe for a number of years using large radius instruments, e.g., SHRIMP I and II, the VG Isolab 120 and recently the Cameca 1270 ion probes. The next generation instrument the reverse-geometry SHRIMP III? (being built at ANU, Canberra, with Matsuda designed mass spectrometer) will greatly increase the maximum resolving power (by a least a factor of 4) and permit higher overall transmission for the same resolving power.However, the reverse-geometry design of this instrument will not permit multi-collector operation. Multi-collector operation for single ion counting has proved to be very difficult. In particular, the large radius ion microprobes, despite their size, do not create a sufficiently large physical separation of masses to permit standard ion counting electron multipliers to be used at high mass (i.e., Pb).Similarly, the LAICP- MS systems are all single collector instruments or at best have multiple Faraday cups. Future designs which incorporate multichannel plate detectors with fast counting electronics for low resolution operation (as in LA-ICP-MS) must be a distinct possibility. The electron multiplier detector used for single ion counting detects all ions with high efficiency.While there are differences in efficiency between elements these should be much less than a factor of 2 provided the detector is set correctly. The background generated by the electron multiplier, and associated electronics, is usually less than 5 31023 to 1 31024 counts s21 and (for an ion microprobe analysis of 26 elements using energy filtering) is equivalent to about 1 ppb for Si. The overall efficiency of the X-ray detectors for both the proton probe and the synchrotron XRF is low ( < 10% of the total X-rays generated are measured by the EDS and 0.1–1.0% by WDS6) therefore increased detector performance could have almost as much impact as improving the brightness of the source.The EDS detector’s ability to simultaneously record Xrays of all energies creates additional problems when using 1190 Analyst, November 1997, Vol. 122intense beams for trace element determination.The EDS amplifier time constant is only 4–8 ms, therefore, dead-time losses will limit the total number of X-rays detected (for all elements) to less than 10 000 counts s21. This problem is often circumvented by putting filters in front of the detectors to virtually eliminate the X-rays from the (relatively light) major elements, e.g., Kapton for Ca and K. Simply having an air gap between sample and detector reduces the X-rays generated by the major elements Si, Al and Mg.This has the additional advantage that the sample does not have to be in a vacuum chamber. Unfortunately use of filters reduces the elements which can be analysed to only those which generate X-ray energies greater than those absorbed by the filter. One method for increasing sensitivity is to increase the number of individual EDS detectors. For example the X26A station on the Brookhaven National Synchrotron Light Source has a 13 EDS detector array6 permitting total count rates to be substantially higher without running into dead-time limitations.Although the low energy resolution of the EDS detector creates X-ray line overlap problems, measurement of at least 9 REE in mid ocean ridge basalt (MORB) glasses (2–25 ppm, approximately 2 ppm detection limit) has been shown to be possible using an EDS detector on the Hasylab synchrotron.13 Overall Measurement Efficiency The measuring efficiency of the ion microprobe in terms of the total number of ions measured per atom sputtered can be estimated from the volume sputtered per unit time and the observed count rate for an alkali element (the alkali elements have the highest ionisation efficiencies).Measurements made on glass2 demonstrated that at least 20% of Na atoms are transmitted through the mass spectrometer at low mass resolution (this assumes 100% ionisation of Na). Losses due to high mass resolution operation on large radius mass spectrometers are between factors of 2 and 10 (below 10 000 resolution), therefore, if single elements are measured it is possible that as many as 1 Na atom in 10 is counted.If the ionisation efficiencies, relative to those of Na, are considered then it should be possible to measure 1 atom in 200 for Si or La, or 1 in 300 for Pb. Although it is possible to measure isotope ratios, or single trace elements, down to sub-ppb levels, high mass resolution mass spectrometers are rarely used for multi-element analysis.Most analyses which require the measurement of a large number of elements (over a large mass range) invariably use energy, rather than mass, filtering. While this method reduces the ion transmission by factors of 20 to 30, many molecular species can be virtually eliminated and analyses are less affected by both instrument tuning and matrix compositions. A direct comparison has been made of Th sensitivity for the ion probe (measuring in energy filtering mode) and the LA-ICPMS. 14 It was shown that the ion microprobe is approximately 10 times more sensitive than the LA-ICP-MS. However, since the laser system removes material some 200 times faster than the ion probe the overall time required to achieve the same precision is 10–20 times faster by LA-ICP-MS (1 to 2 min per analysis compared to 30–45 min by ion probe, albeit on 10 times larger volume). Thus when operating at high mass resolution, and low ion energy, the transmission efficiency of the large radius ion probes is approximately 2 orders of magnitude higher than presently available LA-ICP-MS instruments.Analysis Time Analysis time can significantly affect the type of study attempted. Therefore projects which require the analysis of a large number of samples or require large data-sets for statistical analysis may be rejected not because they cannot be done but because they are not an effective use of instrument time. As primary beam intensities improve, and analysis times are reduced, quantitative trace element imaging at 1 mm spatial resolution becomes a possibility.Where analytical times are short it is possible to build up a substantial database of analyses, for example > 20 000 analyses of mantle garnet have been made on the CSIRO proton probe.15 In general, analysis times for the X-ray techniques are measured in minutes. Limits of Detection for Multi-element Analysis Detection of single elements can be significantly enhanced under specialised analytical conditions, sometimes to the exclusion of all other elements.Single element MDLs may thus be orders of magnitude better than those applicable to multielement analysis. Although duty cycle losses are inevitable where elements are measured sequentially, e.g., in mass spectrometer or WDS X-ray detectors, such losses can be minimised if the measurement time can be adjusted in favour of the least sensitive elements. The deficiencies of the EDS detector, including dead-time restrictions, poor energy resolution and higher peak-to-background ratios, create higher MDLs than might be anticipated for a system which analyses all elements simultaneously. For example in some situations the sensitivity of WDS and EDS analysis can have comparable MDLs (0.1–10 ppm on the Brookhaven Synchrotron with 10 mm white beam).16 The MDLs of the various techniques (in ppb) are given in Fig. 3; those for the destructive techniques are for analyses of more than 20 elements. Both proton and synchrotron XRF probes are presently capable of ppm sensitivity for the period 4 and 5 transition elements.As MDLs improve with the square root of primary excitation intensity17 the new high brilliance sources for the synchrotron-XRF should be capable of detecting elements at < 0.1 ppm in 2 mm spots. At present the most efficient LA-ICP-MS instruments are capable of detecting 1–10 ppb with 50 mm spots (depth of 240 mm) and > 10 ppb for 10 mm.14 The ion probe MDLs are strongly dependent on the 1st ionisation potential of the element and, where energy filtering is employed, their secondary ion energy distribution.For elements having similar ionisation efficiency to Si, the MDLs (15–25 mm holes, 5 mm deep, energy filtering conditions) can be as low as 0.3 ppb at Li, 1 at Y, 2 at La through to 8 at Th (Fig. 3). In contrast, the MDLs for the halogens and the heavier transitional elements (not shown on Fig. 3) are in the 1–100 ppm range.Detection limit calculations made for each technique do not usually include the effects of spectral overlaps, even for the simplest of substrates. Where groups of elements are always present together in natural materials, e.g., the REE, spectral Fig. 3 Approximate minimum detection levels, in ppb, for a 15 mm beam diameter and assuming multi-element analysis. Spectral overlaps are assumed to be either mimimal or absent. Analyst, November 1997, Vol. 122 1191overlaps are not uncommon. In such cases there may be some elements which are invariably overlapped by neighbouring, chemically similar elements, and which either have much poorer detection limits or simply cannot be detected. The Future As new generations of instruments become available it may be expected that existing instruments will become obsolete. However, some ‘new instruments’ are simply refinements on existing instruments which, although offering greater flexibility and (usually) greater ease of measurement, give no major advances in either detection limits or precision of measurement.For example the Cameca ims-3f to 6f series of ion probes have been used for earth science applications for some 20 years and despite little intrinsic change to the overall secondary ion transmission can be expected to continue to be used for the foreseeable future. In contrast the LA-ICP-MS is undergoing such rapid change that obsolescence of earlier instruments (if not other competing techniques) might be expected.Large increases in sensitivity through changes in instrument design (possibly by ‘factors of 10’) are being claimed. If these claims can be sustained on commercially available instruments, and are coupled with improved laser technology, ppb detection limits at sub-10 micrometre resolutions are a distinct possibility. Changes in synchrotron XRF are principally dependent on the development of large facilities at several $100 millions and it is likely that changes in this technique will be by much larger steps rather than the continued development observed on the smaller instruments. Advances in X-ray focusing are giving significant increases in brightness in small spots and may further enhance the capabilities of these new facilities. However, where destructive analytical methods samples are permitted it is unlikely that the synchrotron or proton probe can compete with either the LA-ICP-MS or ion probe, especially if ppb detection of a large number of elements is required.Conclusions Although all instruments would benefit from improvement in source intensity and size, the X-ray techniques should benefit most since they are non-destructive and not limited by the volume of material removed. In particular ability of the new synchrotron X-ray sources to excite the K lines of all elements will strongly enhance detection of elements above z = 40. However, even for the higher atomic number elements, the complexity of the X-ray spectrum may limit the number of elements which can be analysed if detector resolution is not improved.The ion probe is effectively limited by the secondary ionisation process and the complexity of the molecular ion spectrum therefore improved detection relies on the manufacture of ever larger instruments. In contrast the recent changes in ICP-MS technology suggest that increases at least an order of magnitude in efficiency may shortly be possible on commercially available instruments.As such, new generation instruments would be capable of a similar performance to the small ion probes (although still factors of 10 to 30 lower than the SHRIMP-type probes) and will outperform them in analysis of the heavy transition elements. The development of multicollector mass spectrometers are likely to lead to major improvements to the precision of isotope ratio measurements, however, such technology may be difficult to apply to trace element analysis.The development of mass spectrometers compatible with multi-channel detection (essentially electronic versions of the old photoplate instruments) may hold the key to cheap and fast trace element analysis. References 1 Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, p. 419. 2 Hinton, R. W., in Microprobe Techniques in the Earth Sciences, ed.Potts, P. J. Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 235–289. 3 Perkins, W. T., and Pearce N. J., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 290–325. 4 Reed, S. J. B., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J. Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 49–90. 5 Fraser, D.G., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 141–162. 6 Smith, J. V., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 163–234. 7 Haller, M., and Kn�ochel, A., J. Trace Microprobe Tech., 1996, 143, 461. 8 Dove, M., and Redfern, S., Min. Soc. Bull., 1997, 113, 21. 9 Griffin, W.L., Slack, J. F., Ramsden, A. R., Win, T. T., and Ryan, C. G., Econ. Geol., 91, 657. 10 Diamond, L. W., Marshall, D. D. Jackman, J. A., and Skippen, G. B., Geochim. Cosmochim. Acta, 1990, 54, 545. 11 Vanko, D. A., Sutton, S. R., Rivers, M. L., and Bodnar, R. J., Chem. Geol., 1993, 109, 125. 12 Fryer, B. J., Jackson, S. E., and Longerich, H. P., Can. Min., 1995, 33, 303. 13 Lechtenberg, F., Garbe, S., Bauch, J., Dingwell, D. B., Freitag, J., Haller, M., Hansteen, T. H., Ippach, P., Knochel, A., Radtke, M., Romano, C., Sachs, P.M., Schmincke, H. U., and Ullrich, H. J., J. Trace Microprobe Tech., 1996, 14, 561. 14 Horn, I., Hinton, R. W., Jackson, S. E., and Longerich, H. E., Geostand. Newsl., 1997, in the press. 15 Griffin, W. L., and Ryan, C. G., Contrib. Min. Petrol.,996, 124, 216. 16 Rakovan, J., and Reeder, R. J., Geochim. Cosmochim. Acta, 1996, 60, 4435. 17 Chevalier, P., Dhez, P., Legrand, F., Erko, A., Agafonov, Y., Panchenko, L. A., and Yakshin, A.Y., J. Trace Microbeam Tech., 1996, 14, 517. Paper 7/06063G Received August 18, 1997 Accepted October 9, 1997 1192 Analyst, November 1997, Vol. 122 A Brilliant Future for Microanalysis?† Richard W. Hinton Department of Geology and Geophysics, University of Edinburgh, West Mains Road, Edinburgh, UK EH9 3JW The ability to measure trace element distributions on the microscale is often critical to our understanding of large scale geological processes. As the number of different instruments and techniques capable of trace element analysis increases the choice of analytical method is becoming less clear.Although most techniques have 1 micrometre spatial resolution as their ultimate goal it is clear that the analytical depth is determined by sample thickness (proton probe and synchrotron XRF) or analytical volume (ion probe and laser ablation-ICP-MS) considerations. Thus ability to produce brighter and smaller beams also requires significant improvement in the detection of secondary particles.Provided analytical considerations, such as precision or detectibility are met, the eventual choice between competing techniques may well be more mundane aspects such as ease of access and cost. Keywords: Microanalytical techniques; ion probe; laser ablation inductively coupled plasma mass spectrometry; proton probe; synchrotron X-ray fluorescence; geological analyses The ability to analyse a wide range of trace elements on the microscale has become an important part of our understanding of geological processes.Both instrument development and microanalytical methods have progressed steadily over the past two decades to give greater element coverage, lower detection levels and smaller beam sizes. Earth scientists can now choose between a variety of microanalytical methods, however, information is often more readily available on ultimate performance (under ideal conditions) than routine capability.The general rule of thumb, that the instrument size must increase to permit greater microanalytical sensitivity, has resulted in very large (and expensive) instruments and facilities. There is a danger that the smaller number of very costly facilities may result in significant unsatisfied demand. However, it is clear from the continued use of older, smaller instruments that many applications do not require the most sensitive instruments and that access, flexibility, and cost are often the dominant factors affecting choice of technique. Future management of access may require directing research between techniques to make the most efficient use of limited resources.Microanalytical Techniques for Multi-element Analysis In this paper only techniques which permit the simultaneous microanalysis of a number of trace elements will be considered, namely the ion probe, the laser ablation (LA)-ICP-MS, the proton probe and synchrotron XRF. An extended discussion of all these techniques can be found in Potts et al.1 General descriptions are given below: (1) The ion probe2 uses secondary ion mass spectrometry (SIMS) to detect ions physically removed from the surface by a high energy ion beam.Nearly all geological laboratories use a focused beam of 16O2 to sputter atoms from a (polished) flat surface. Ions which are formed in the sputtering process can be accelerated into a double focusing mass spectrometer and separated on the basis of both mass and energy. All masses from H to U are ejected, with the degree of ionisation principally being related to the elements’ first ionisation potential.The mass spectrum created by ion bombardment contains not only the single ions of individual elements but also many molecular species. These molecular species must be separated either by measuring only high energy ions (and thereby losing 95–97% of the original signal) or high mass resolution (where losses are related to the size of the instrument).Although ion probes with very large mass spectrometers are often capable of separating individual elements from molecular species, without significantly reducing the number of secondary ions transmitted, these expensive instruments are principally used for isotopic analyses (e.g., U–Pb dating of zircons). Samples are coated (generally with Au) as the surface of the sample is held at a high voltage. (2) LA-ICP-MS3 combines a laser source with a conventional inductively coupled plasma mass spectrometer.A finely focused laser beam is used to ablate the surface of a sample held in an argon filled cell. Solid particles ablated from the surface are transported in Ar gas into an inductively coupled plasma where they are ionised. As the ionisation process produces a lower proportion of molecular species than SIMS most trace element measurements can be made using low resolution quadrupole mass spectrometers. (3) In a manner similar to the electron probe,4 the proton probe5 uses a finely focused beam of very high energy protons (2–4 MeV) to excite atoms and generate X-rays.This technique is often called particle (or proton) induced X-ray emission (PIXE). Characteristic X-ray lines of individual elements are detected by either solid state detectors (known as energy dispersive spectrometer or EDS) or a combination of curved crystal and proportional counter (known as wavelength dispersive spectrometer or WDS).The EDS system detects the entire range of X-rays simultaneously whilst the WDS system relies on the positioning of a curved crystal to measure one Xray line at a time. The electron microprobe has a high X-ray background (Bremsstrahlung); the X-rays are generated as the electrons decelerate. Heavier protons decelerate much more slowly consequently the proton probe has a much lower intrinsic X-ray background. However, the background is not completely absent as some X-rays are generated by electrons ejected by the proton beam interaction.The commonly used EDS method of X-ray detection has relatively poor X-ray energy resolution therefore the unresolved overlap of X-rays lines of different elements is not uncommon, especially in complex substrates. (4) The synchrotron X-ray fluorescence technique6 is analogous to the laboratory XRF except that a synchrotron source provides many orders of magnitude more X-ray photons to excite X-ray fluorescence in the target material. The synchrotron X-ray source provides polarised light either as a ‘white’ Xray continuum or at well defined energies (monochromatic excitation).The photon beam is highly concentrated and † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1187–1192) 1187Instrument Time 1993 - 1997 Edinburgh NERC Ion Microprobe Facility micrometre sized X-ray beams can be produced either by apertures or by focusing using mirrors, curved lenses or capillaries.X-rays are detected using EDS and WDS systems similar to those used for both the proton and electron probes; the detectors used on the new generation of instruments must be capable of detecting very high energy X-rays. Most of the X-ray background is generated by scattering, however, as the synchrotron photons are polarised the scattering is at a minimum at 90° to the orbital plane of the electrons or positrons which generate the X-ray photons (i.e., horizontal plane).Thus by having the detector at right angles to the incident X-ray beam the scattered X-ray background is reduced (effectively by a factor of 10, Haller and Kn�ochel7). This geometry requires that the sample be at 45° to the incident beam; this has obvious implications for the micro-analysis of thick samples. Although not capable of element analysis at very low concentrations, the electron probe is invariably used to determine the major and minor chemistry of all samples analysed by these techniques and is an essential part of all microanalytical studies.Access An earth or environmental scientist who has a spific analytical problem is principally concerned with two factors: (1) can a particular method make the analysis to the precision and accuracy required on their samples? and (2) can they obtain access at either no, or reasonable, cost? There are numerous examples in the literature where a particular technique has been used, not because it is the best method, but because of ready access.Unlike the electron microprobe and SEM which are generally accessible for most earth scientists, either within-house or at local centres, access to the modern microanalytical instruments usually requires formal application to a national centre or an informal approach to the principal scientist in charge of a laboratory.It is inevitable that such instruments will be oversubscribed and access may well require a formal proposal which is subject to peer review. For example the Edinburgh Ion Microprobe Facility (IMF) has been operating as a national centre for ten years using a Cameca ims-4f controlled by Charles Evans and Associates operating system. UK based earth scientists can obtain access to the IMF by making a formal proposal to the Scientific Services division of the Natural Environment Research Council (NERC).Large projects requiring substantial use of the instrument are funded directly by NERC Research Grants. Thus virtually all access is by peer reviewed proposals and is in competition with other earth scientists. In contrast, techniques such as the synchrotron XRF are dependent on access to very large facilities which may be funded by a number of different Government agencies (or indeed a number of Governments). Unless part of the facility (e.g., a beam line) is specifically funded for earth science applications, access may not only be in competition with scientists from many disciplines but may appear very expensive on a per project basis (see Dove and Redfern8).Many laboratories expect running costs to be met by external users, a proportion of staff costs may also be included if analyses are carried out by ‘in-house’ staff. The approach to providing access for a wide number of scientists to expensive microanalytical techniques varies from laboratory to laboratory.Many laboratories provide only limited access to external users while others provide the majority of the time to ‘external’ users. The IMF at Edinburgh averages over 35 different projects a year ranging through oxygen isotopes, trace element and light element analysis, to diffusion studies. Training is as an important part of the laboratory operation and a ‘hands-on’ approach is adopted for nearly all users. However, while improving the users knowledge of the technique, the ‘hands-on’ approach does have the drawback that any advances in technique must become relatively routine before being offered to inexperienced users.Thus techniques are often developed by ‘research grant’ users before being offered as part of the IMF service. Other laboratories have ‘in-house’ operators with the user choosing the elements required and the points to be analysed. While less time is lost in setting up the instrument, unless the user is on site when analyses are made, there is potential for some analyses to be of incorrect locations. Many laboratories will have a dedicated research programme with a restricted number of researchers, including post-graduate students, having ‘handson’ use of an instrument.Access to outside users will in this case be limited, however, it is often still possible to gain access to such laboratories if the project is scientifically challenging.Over the 10 years that the IMF has been in operation at Edinburgh there have been in excess of 300 requests for information about use of the ion microprobe, with over 100 resulting in formal applications for access. Initial approaches for access vary through very specific problems from those knowledgeable about the technique, to vague inquiries about its general capabilities. However, it is clear that even when applicants have reasonable knowledge of a technique, each study may have unique features which can create analytical difficulties.A breakdown of the ion microprobe usage over the last four years (Fig. 1) demonstrates not only the strengths of the technique but the awareness within the geological community. About 50% of the time was spent on rare earth element (REE) and high field strength element (HFSE) characterisation of igneous and metamorphic systems including both natural rocks and experimental run products. In principal many of these analyses could have been done on the first generation instruments over the last 15 years.Demand stays strong due to the continued demonstration of the importance of the microscale variability in light and trace elements both within and between individual minerals. The ion probe continues to be one of the most sensitive instruments for the measurement of volatile elements e.g., hydrogen and carbon (11% of instrument time). The relatively low instrument time (17%) devoted to isotopic analysis in part reflects the difficulty of these analyses (especially for inexperienced operators) but also the higher operational costs of this type of analysis.Although sedimentalogical applications have been relatively low, these are now showing signs of increase. While this may be largely due to improved awareness of the technique, it may also reflect increased funding for projects related to environmental change. Somewhat surprisingly, even though the ims-4f was designed as an ion microscope very little instrument time has been dedicated to imaging.The reason for this is relatively simple: the nonconducting nature of silicates leads to sample charging problems when large areas of the sample are bombarded with Fig. 1 Instrument usage of the ion microprobe over the last four years. 1188 Analyst, November 1997, Vol. 122charged particles (the conductive coating must be removed by the incident beam prior to analysis). Sample Preparation One of the first analytical concerns is whether samples of the rock or mineral can be prepared for each particular method and whether the analysis must be non-destructive.This aspect may be far from trivial as sometimes the sample has already been prepared and is so ‘unique’ or highly valued by its owner that little or no modifications can be made to it. Ideally specimens are prepared for the ion microprobe as 1 inch round, polished, thin sections. All samples must be compatible with high vacuum operation.Samples prepared for the electron microprobe are usually compatible with the ion microprobe (subject to high vacuum compatibility). The sample surface must be optically flat, not just over the area to be analysed but over an extended region (mm) since the ims-4f operates with high voltage extraction field and the (Au coated) sample surface acts as part of the secondary ion extraction optics. As it is quite possible that a sample will be subject to a number of different analytical techniques, compatibility between techniques should be considered at an early stage.Clearly the most destructive technique must be kept until last. The ion probe sputters the sample away relatively slowly and rarely sputters to greater than 10 mm depth. An ion beam of 15–25 KeV energy will mix a layer of approximately 0.01 mm deep beneath the surface. Atoms beneath this depth will be unaffected by the ion beam therefore the volume sampled is essentially that sputtered.Samples for LA-ICP-MS analysis are usually made relatively thick (100s mm) as the laser ablation process often reaches over 50 mm depth in the few minutes required for an analysis. As the efficiency of this method improves the thick sample requirement is relaxing and samples of normal thickness (30 mm) have been analysed. Samples for synchrotron XRF must be mounted with the assumption that the beam goes right through the sample (i.e., penetrates greater than 100 mm).Mounting thin slices on organic mounting medium such as Kapton ensures that X-rays are generated from the sample alone.7 Samples are usually held in air, however, compatibility with other techniques may require that the sample be compatible with high vacuum operation. The proton probe samples depths of between 25 and 100 mm depending on the sample’s composition and the element measured. Samples for the proton probe are preferably thin (@10 mm thick) and mounted on a substrate which will not give additional X-ray background. Mounting on pure silica slides is one possible method.In some cases samples which have already been mounted on normal glass slides have been analysed and then the contribution from the glass slide subtracted from the concentrations measured.9 Clearly, although this must result in reduced precision, and should be avoided if at all possible, it may be the only practical solution for a unique sample.The quantitative analysis of thick samples using both the synchrotron XRF or the proton probe creates greater difficulties for correction procedures but is not a major barrier to analysis. Although it may appear obligatory that the phase to be analysed is at the surface of the sample there are a few situations where this is not possible. For example, fluid inclusions can only be analysed in their original location within their host mineral. The ion probe primary beam can sputter away the sample until the fluid inclusion is reached releasing the gas and liquid in a short burst.10 However, the signal obtained as the inclusion bursts is very short lived and measurement of the individual ion intensities is very difficult.Further, quantification is extremely difficult as the release cannot be simulated under standard conditions. In contrast the deep penetration of synchrotron radiation or the proton beam is a positive advantage since the inclusion can be analysed in situ.5,11 In these studies analytical problems are associated with the attenuation of Xrays within the host mineral and quantification is far from routine.Beam damage to geological materials is generally not a problem for either the synchrotron XRF or the proton probe. While the LA-ICP-MS and ion probe are destructive techniques the amounts of material removed is less than 10 ng for the ion probe and 100–500 ng for the LA-ICP-MS. Analytical Spot Sizes and Volumes Analytical spot sizes and volumes can be a relatively contentious issue, in part because comparisons can rarely be made where all the parameters are the same for each technique. Clearly, while micro-XRF using the synchrotron is predicted to achieve sub-micrometre spatial resolution, the analytical volume is limited by the thickness of the section presented to the beam or the absorption of the secondary X-rays by the sample itself.Similarly, while the proton probe can achieve 1 mm spatial resolution, X-rays are generated from at least 25 to 100 mm into the sample depending on the element (the sampled volume being element and matrix specific). In contrast, the ‘destructive’ techniques excavate a crater which can be measured optically or with an SEM.Reasonable estimates can also be made on crater depths by either drilling through the entire sample of known thickness (LA-ICP-MS) or physically measuring the depth of large pits (using a rastered ion beam) to determine sputter rates (ion probe).In both cases the number of atoms removed from the sample can be calculated and estimates made of the efficiency of both ion formation and collection. Definition of spot widths can vary from those preferred by manufacturers (most optimistic, the width at half peak maximum, Fig. 2) to the more realistic (the width at 10% of maximum beam intensity). Overlap of the low intensity fringe of the beam (Fig. 2) may be highly significant when large differences exist over short distances (e.g., grain boundaries).All techniques should perhaps include the total width or volume with which the beam interacts either through direct bombardment (i.e., ion probe), heating (LA-ICP-MS) or secondary Xrays and electrons (synchrotron XRF and proton probe). Sources of Energetic Particles For all techniques there will be an anti-correlation between spatial resolution and the intensity of the excitation source as measured at the sample.Thus although the proton probe is capable of a spatial resolution of 1 mm, the beam diameter is Fig. 2 Beam current distribution of primary ions. True beam diameter is total area from which ions are removed rather than the optimistic width at 50% peak height (not infrequently used by manufacturers) and the width at 10% peak height used by the laboratories. Analyst, November 1997, Vol. 122 1189normally !20 mm (approximately 0.1 compared to 6–8 nA beam currents, respectively) to give a sufficient number of Xrays to detect very low concentrations.Similarly, an ion microprobe beam of less than 5 mm is feasible with an O2 source but normally 10–25 mm is used for trace element analysis. A proton beam of 3 MeV excites the K X-ray lines only up to about z = 42. Heavier elements must therefore be analysed by their L lines (which may be overlapped by the K lines of lighter elements); the generation of X-rays from elements between Mo and Lu is particularly poor. The new generation synchrotron X-ray facilities, e.g., the Advanced Photon Source (APS) and European Synchrotron Radiation Facility (ESRF) produce not only a higher total numbers of photons but they are also concentrated into a smaller area (i.e., not only have high brightness but also high brilliance6).Focusing of X-ray beams has become more advanced with focusing using curved crystals or capillaries giving orders of magnitude higher intensities than apertures. Small diameter beams therefore do not necessarily entail prohibitive decreases in beam intensity.While the new generation sources only produce about factors of 2 to 3 more photons at low energy they produce many orders of magnitude more high energy photons. Both the APS and the ESRF facilities will therefore be able to excite the K lines of all elements, including U. The ability to measure the heavier elements (z > 40) on their K lines should improve their minimum detection limits (MDL) by at least an order of magnitude.In contrast, the duoplasmatron gas source which provides the O2 primary beam for the ion probe has not changed significantly over the last 20 years. However, since SIMS is a destructive technique, unless the measurement efficiency is improved the volume of material sputtered must be kept constant. Thus if it were possible to focus a 10 nA beam into 1 mm area, the beam would be required to penetrate 100 mm depth to maintain the same total secondary ion counts.Despite these reservations a more intense source would considerably improve instrument operation, including better beam shapes and greater flexibility, including faster analyses. The proton source also uses a duoplasmatron source and is not expected to undergo any major changes in the near future. Originally LA-ICP-MS relied on infra-red lasers to ablate the surface, however, they did not always create small pits, especially in transparent or highly cleaved minerals.Fortunately UV lasers have been found to give much more controlled ablation and smaller craters even in transparent materials. Laser technology for the LA-ICP-MS is still improving and it is likely that new pico- and femtosecond lasers will permit smaller beams, better coupling with transparent materials and lower matrix effects. If these improvements can be coupled with improvements in sensitivity of the ICP-MS system as a whole, smaller beams will not necessarily have to be compensated for by increasing depth of sampling.Secondary Losses The ultimate limitation for both the ICP-MS and the ion probe is the efficiency of the ionisation process. In SIMS only a relatively small proportion of atoms sputtered are ionised. The secondary ionisation is strongly dependent on the first ionisation potential of each element and variations of over 3 orders of magnitude exist in the secondary ionisation efficiency.This has major consequences on the detection limits for certain elements, particularly the heavier transition elements. The laser ablation creates solids and gases which are transported in an Ar gas flow into the ICP-MS plasma. It is thought that 25–100% of the material ablated by the laser beam is transported to the plasma of the ICP-MS. However, losses within the plasma interface and within the mass spectrometer must be high as less than 0.1% of atoms ablated are measured.The ionisation in the inductively coupled plasma is far less dependent on the first ionisation potential than for SIMS and the variations in ionisation efficiency between elements is generally less than a factor of 10. However, while the secondary ion current produced by ion bombardment is relatively stable, the laser ablation process produces an ion signal which changes rapidly with time even for homogeneous substrates. These time dependent effects in the laser ablation and ionisation process are not the same for all elements.Time dependent variations of between factors of 2 to 3 can occur between elements (particularly between siderophile and chalcophile elements12). Instrument tuning for maximum ionisation efficiency in the ICP-MS is also mass specific and if a large element range is required there must be some compromise over the efficiency of ionisation at the extremes of mass range. Overall, the strong relationship between first ionisation potential and detection efficiencies for the ion microprobe makes this technique’s capabilities much more element specific than those of the LA-ICP-MS. In particular the heavier transition elements Cu, Zn, Rh–Ag, Ir–Au all have poor sensitivity on the ion microprobe.Conversely the alkalis, the alkaline earths and the rare earth elements have higher sensitivity on the ion probe than the LA-ICP-MS. Detector Systems/Mass Spectrometer Limitations For trace element analyses by destructive techniques coupled with mass spectrometric measurement on a single collector the ultimate detection level is in part dependent on the number of elements analysed (i.e., the duty cycle).The ultimate mass spectrometer would therefore be one that operates with high transmission at very high mass resolving power, measuring all elements of interest simultaneously. This is far from being realised. High sensitivity at high mass resolving power has been possible on the ion microprobe for a number of years using large radius instruments, e.g., SHRIMP I and II, the VG Isolab 120 and recently the Cameca 1270 ion probes.The next generation instrument the reverse-geometry SHRIMP III? (being built at ANU, Canberra, with Matsuda designed mass spectrometer) will greatly increase the maximum resolving power (by a least a factor of 4) and permit higher overall transmission for the same resolving power. However, the reverse-geometry design of this instrument will not permit multi-collector operation.Multi-collector operation for single ion counting has proved to be very difficult. In particular, the large radius ion microprobes, despite their size, do not create a sufficiently large physical separation of masses to permit standard ion counting electron multipliers to be used at high mass (i.e., Pb). Similarly, the LAICP- MS systems are all single collector instruments or at best have multiple Faraday cups. Future designs which incorporate multichannel plate detectors with fast counting electronics for low resolution operation (as in LA-ICP-MS) must be a distinct possibility.The electron multiplier detector used for single ion counting detects all ions with high efficiency. While there are differences in efficiency between elements these should be much less than a factor of 2 provided the detector is set correctly. The background generated by the electron multiplier, and associated electronics, is usually less than 5 31023 to 1 31024 counts s21 and (for an ion microprobe analysis of 26 elements using energy filtering) is equivalent to about 1 ppb for Si.The overall efficiency of the X-ray detectors for both the proton probe and the synchrotron XRF is low ( < 10% of the total X-rays generated are measured by the EDS and 0.1–1.0% by WDS6) therefore increased detector performance could have almost as much impact as improving the brightness of the source. The EDS detector’s ability to simultaneously record Xrays of all energies creates additional problems when using 1190 Analyst, November 1997, Vol. 122intense beams for trace element determination. The EDS amplifier time constant is only 4–8 ms, therefore, dead-time losses will limit the total number of X-rays detected (for all elements) to less than 10 000 counts s21. This problem is often circumvented by putting filters in front of the detectors to virtually eliminate the X-rays from the (relatively light) major elements, e.g., Kapton for Ca and K.Simply having an air gap between sample and detector reduces the X-rays generated by the major elements Si, Al and Mg. This has the additional advantage that the sample does not have to be in a vacuum chamber. Unfortunately use of filters reduces the elements which can be analysed to only those which generate X-ray energies greater than those absorbed by the filter. One method for increasing sensitivity is to increase the number of individual EDS detectors. For example the X26A station on the Brookhaven National Synchrotron Light Source has a 13 EDS detector array6 permitting total count rates to be substantially higher without running into dead-time limitations.Although the low energy resolution of the EDS detector creates X-ray line overlap problems, measurement of at least 9 REE in mid ocean ridge basalt (MORB) glasses (2–25 ppm, approximately 2 ppm detection limit) has been shown to be possible using an EDS detector on the Hasylab synchrotron.13 Overall Measurement Efficiency The measuring efficiency of the ion microprobe in terms of the total number of ions measured per atom sputtered can be estimated from the volume sputtered per unit time and the observed count rate for an alkali element (the alkali elements have the highest ionisation efficiencies).Measurements made on glass2 demonstrated that at least 20% of Na atoms are transmitted through the mass spectrometer at low mass resolution (this assumes 100% ionisation of Na).Losses due to high mass resolution operation on large radius mass spectrometers are between factors of 2 and 10 (below 10 000 resolution), therefore, if single elements are measured it is possible that as many as 1 Na atom in 10 is counted. If the ionisation efficiencies, relative to those of Na, are considered then it should be possible to measure 1 atom in 200 for Si or La, or 1 in 300 for Pb.Although it is possible to measure isotope ratios, or single trace elements, down to sub-ppb levels, high mass resolution mass spectrometers are rarely used for multi-element analysis. Most analyses which require the measurement of a large number of elements (over a large mass range) invariably use energy, rather than mass, filtering. While this method reduces the ion transmission by factors of 20 to 30, many molecular species can be virtually eliminated and analyses are less affected by both instrument tuning and matrix compositions.A direct comparison has been made of Th sensitivity for the ion probe (measuring in energy filtering mode) and the LA-ICPMS. 14 It was shown that the ion microprobe is approximately 10 times more sensitive than the LA-ICP-MS. However, since the laser system removes material some 200 times faster than the ion probe the overall time required to achieve the same precision is 10–20 times faster by LA-ICP-MS (1 to 2 min per analysis compared to 30–45 min by ion probe, albeit on 10 times larger volume).Thus when operating at high mass resolution, and low ion energy, the transmission efficiency of the large radius ion probes is approximately 2 orders of magnitude higher than presently available LA-ICP-MS instruments. Analysis Time Analysis time can significantly affect the type of study attempted. Therefore projects which require the analysis of a large number of samples or require large data-sets for statistical analysis may be rejected not because they cannot be done but because they are not an effective use of instrument time.As primary beam intensities improve, and analysis times are reduced, quantitative trace element imaging at 1 mm spatial resolution becomes a possibility. Where analytical times are short it is possible to build up a substantial database of analyses, for example > 20 000 analyses of mantle garnet have been made on the CSIRO proton probe.15 In general, analysis times for the X-ray techniques are measured in minutes.Limits of Detection for Multi-element Analysis Detection of single elements can be significantly enhanced under specialised analytical conditions, sometimes to the exclusion of all other elements. Single element MDLs may thus be orders of magnitude better than those applicable to multielement analysis. Although duty cycle losses are inevitable where elements are measured sequentially, e.g., in mass spectrometer or WDS X-ray detectors, such losses can be minimised if the measurement time can be adjusted in favour of the least sensitive elements.The deficiencies of the EDS detector, including dead-time restrictions, poor energy resolution and higher peak-to-background ratios, create higher MDLs than might be anticipated for a system which analyses all elements simultaneously. For example in some situations the sensitivity of WDS and EDS analysis can have comparable MDLs (0.1–10 ppm on the Brookhaven Synchrotron with 10 mm white beam).16 The MDLs of the various techniques (in ppb) are given in Fig. 3; those for the destructive techniques are for analyses of more than 20 elements. Both proton and synchrotron XRF probes are presently capable of ppm sensitivity for the period 4 and 5 transition elements. As MDLs improve with the square root of primary excitation intensity17 the new high brilliance sources for the synchrotron-XRF should be capable of detecting elements at < 0.1 ppm in 2 mm spots.At present the most efficient LA-ICP-MS instruments are capable of detecting 1–10 ppb with 50 mm spots (depth of 240 mm) and > 10 ppb for 10 mm.14 The ion probe MDLs are strongly dependent on the 1st ionisation potential of the element and, where energy filtering is employed, their secondary ion energy distribution. For elements having similar ionisation efficiency to Si, the MDLs (15–25 mm holes, 5 mm deep, energy filtering conditions) can be as low as 0.3 ppb at Li, 1 at Y, 2 at La through to 8 at Th (Fig. 3). In contrast, the MDLs for the halogens and the heavier transitional elements (not shown on Fig. 3) are in the 1–100 ppm range. Detection limit calculations made for each technique do not usually include the effects of spectral overlaps, even for the simplest of substrates. Where groups of elements are always present together in natural materials, e.g., the REE, spectral Fig. 3 Approximate minimum detection levels, in ppb, for a 15 mm beam diameter and assuming multi-element analysis.Spectral overlaps are assumed to be either mimimal or absent. Analyst, November 1997, Vol. 122 1191overlaps are not uncommon. In such cases there may be some elements which are invariably overlapped by neighbouring, chemically similar elements, and which either have much poorer detection limits or simply cannot be detected. The Future As new generations of instruments become available it may be expected that existing instruments will become obsolete.However, some ‘new instruments’ are simply refinements on existing instruments which, although offering greater flexibility and (usually) greater ease of measurement, give no major advances in either detection limits or precision of measurement. For example the Cameca ims-3f to 6f series of ion probes have been used for earth science applications for some 20 years and despite little intrinsic change to the overall secondary ion transmission can be expected to continue to be used for the foreseeable future.In contrast the LA-ICP-MS is undergoing such rapid change that obsolescence of earlier instruments (if not other competing techniques) might be expected. Large increases in sensitivity through changes in instrument design (possibly by ‘factors of 10’) are being claimed. If these claims can be sustained on commercially available instruments, and are coupled with improved laser technology, ppb detection limits at sub-10 micrometre resolutions are a distinct possibility.Changes in synchrotron XRF are principally dependent on the development of large facilities at several $100 millions and it is likely that changes in this technique will be by much larger steps rather than the continued development observed on the smaller instruments. Advances in X-ray focusing are giving significant increases in brightness in small spots and may further enhance the capabilities of these new facilities.However, where destructive analytical methods samples are permitted it is unlikely that the synchrotron or proton probe can compete with either the LA-ICP-MS or ion probe, especially if ppb detection of a large number of elements is required. Conclusions Although all instruments would benefit from improvement in source intensity and size, the X-ray techniques should benefit most since they are non-destructive and not limited by the volume of material removed. In particular ability of the new synchrotron X-ray sources to excite the K lines of all elements will strongly enhance detection of elements above z = 40. However, even for the higher atomic number elements, the complexity of the X-ray spectrum may limit the number of elements which can be analysed if detector resolution is not improved. The ion probe is effectively limited by the secondary ionisation process and the complexity of the molecular ion spectrum therefore improved detection relies on the manufacture of ever larger instruments. In contrast the recent changes in ICP-MS technology suggest that increases at least an order of magnitude in efficiency may shortly be possible on commercially available instruments. As such, new generation instruments would be capable of a similar performance to the small ion probes (although still factors of 10 to 30 lower than the SHRIMP-type probes) and will outperform them in analysis of the heavy transition elements. The development of multicollector mass spectrometers are likely to lead to major improvements to the precision of isotope ratio measurements, however, such technology may be difficult to apply to trace element analysis. The development of mass spectrometers compatible with multi-channel detection (essentially electronic versions of the old photoplate instruments) may hold the key to cheap and fast trace element analysis. References 1 Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, p. 419. 2 Hinton, R. W., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J. Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 235–289. 3 Perkins, W. T., and Pearce N. J., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 290–325. 4 Reed, S. J. B., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J. Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 49–90. 5 Fraser, D. G., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 141–162. 6 Smith, J. V., in Microprobe Techniques in the Earth Sciences, ed. Potts, P. J., Bowles, J. F. W., Reed, S. J. B., and Cave, M. R., Chapman and Hall, 1995, pp. 163–234. 7 Haller, M., and Kn�ochel, A., J. Trace Microprobe Tech., 1996, 143, 461. 8 Dove, M., and Redfern, S., Min. Soc. Bull., 1997, 113, 21. 9 Griffin, W. L., Slack, J. F., Ramsden, A. R., Win, T. T., and Ryan, C. G., Econ. Geol., 91, 657. 10 Diamond, L. W., Marshall, D. D. Jackman, J. A., and Skippen, G. B., Geochim. Cosmochim. Acta, 1990, 54, 545. 11 Vanko, D. A., Sutton, S. R., Rivers, M. L., and Bodnar, R. J., Chem. Geol., 1993, 109, 125. 12 Fryer, B. J., Jackson, S. E., and Longerich, H. P., Can. Min., 1995, 33, 303. 13 Lechtenberg, F., Garbe, S., Bauch, J., Dingwell, D. B., Freit, J., Haller, M., Hansteen, T. H., Ippach, P., Knochel, A., Radtke, M., Romano, C., Sachs, P. M., Schmincke, H. U., and Ullrich, H. J., J. Trace Microprobe Tech., 1996, 14, 561. 14 Horn, I., Hinton, R. W., Jackson, S. E., and Longerich, H. E., Geostand. Newsl., 1997, in the press. 15 Griffin, W. L., and Ryan, C. G., Contrib. Min. Petrol., 1996, 124, 216. 16 Rakovan, J., and Reeder, R. J., Geochim. Cosmochim. Acta, 1996, 60, 4435. 17 Chevalier, P., Dhez, P., Legrand, F., Erko, A., Agafonov, Y., Panchenko, L. A., and Yakshin, A. Y., J. Trace Microbeam Tech., 1996, 14, 517. Paper 7/06063G Received August 18, 1997 Accepted October 9, 1997 1192 Analyst, November 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a706063g
出版商:RSC
年代:1997
数据来源: RSC
|
3. |
Some Thoughts on Problems Associated With Various Sampling Media Used for Environmental Monitoring† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1193-1200
Arthur J. Horowitz,
Preview
|
|
摘要:
Some Thoughts on Problems Associated With Various Sampling Media Used for Environmental Monitoring† Arthur J. Horowitz US Geological Survey, Peachtree Business Center, Suite 130, 3039 Amwiler Road, Atlanta, GA 30360, USA. E-mail: horowitz@usgs.gov Modern analytical instrumentation is capable of measuring a variety of trace elements at concentrations down into the single or double digit parts-per-trillion (ng l21) range. This holds for the three most common sample media currently used in environmental monitoring programs: filtered water, whole-water and separated suspended sediment.Unfortunately, current analytical capabilities have exceeded the current capacity to collect both uncontaminated and representative environmental samples. The success of any trace element monitoring program requires that this issue be both understood and addressed. The environmental monitoring of trace elements requires the collection of calendar- and event-based dissolved and suspended sediment samples.There are unique problems associated with the collection and chemical analyses of both types of sample media. Over the past 10 years, reported ambient dissolved trace element concentrations have declined. Generally, these decreases do not reflect better water quality, but rather improvements in the procedures used to collect, process, preserve and analyze these samples without contaminating them during these steps. Further, recent studies have shown that the currently accepted operational definition of dissolved constituents (material passing a 0.45 mm membrane filter) is inadequate owing to sampling and processing artifacts.The existence of these artifacts raises questions about the generation of accurate, precise and comparable ‘dissolved’ trace element data. Suspended sediment and associated trace elements can display marked short- and long-term spatial and temporal variability. This implies that spatially representative samples only can be obtained by generating composites using depth- and width-integrated sampling techniques.Additionally, temporal variations have led to the view that the determination of annual trace element fluxes may require nearly constant (e.g., high-frequency) sampling and subsequent chemical analyses. Ultimately, sampling frequency for flux estimates becomes dependent on the time period of concern (daily, weekly, monthly, yearly) and the amount of acceptable error associated with these estimates.Keywords: Environmental monitoring; water quality; sampling; trace elements Since the early to mid-1980s, trace element analyses of a variety of environmental samples have been increasingly performed using inductively coupled plasma (ICP)-based instrumentation as opposed to atomic absorption-based instrumentation.1,2 ICP techniques provide several distinct advantages; the two most obvious are (1) multi-element analyses from a single sample aspiration, as opposed to single element analyses, and (2) calibration curves that tend to be linear over much larger ranges, thus reducing the need to dilute and re-analyze numerous samples.Initial ICP instrumentation was based on atomic emission (ICP-AES); this limited the number of elements that could be determined and their associated reporting/detection limits. The subsequent introduction of ICP instrumentation using mass spectrometry (ICP-MS) increased the number of trace elements that could be determined, and further improved some reporting/detection limits.2 In the field of fluvial environmental trace element monitoring, at least some of the advantages accrued by switching to ICP-based instrumentation have been mitigated by samplingrelated issues.Of particular concern are (1) sample contamination during collection, processing, preservation and subsequent chemical analyses and (2) insufficient numbers of samples to encompass the range of spatial and temporal trace element variability. In other words, current laboratory analytical capabilities have exceeded the capacity of field personnel to collect uncontaminated and representative environmental samples.As such, environmental trace element monitoring requirements, and also regulatory limits, should not be based solely on the capabilities of analytical instrumentation, but on the limitations associated with the collection, processing and preservation of field samples.The various problems associated with typical sample media used to monitor environmental trace element levels, and the approaches employed by the US Geological Survey (USGS) to deal with them, have been the subject of a substantial amount of research over the past 5–7 years. Portions of that work are summarized below. Specific details of the various field and laboratory studies used in this summary can be found in the publications cited in the various sections of this paper.Sampling Media Concentrations of dissolved and suspended sediment-associated trace elements, during baseflow and major events (e.g., spring runoff, floods) are necessary to encompass the range of trace element variability in many natural or anthropogenically impacted systems. However, there are unique problems associated with the collection and subsequent chemical analyses of both types of sample media. The success of any monitoring program, regardless of spatial (local, regional, national, global) or temporal (hourly, daily, weekly, monthly, yearly) scale, requires that these various problems be both understood and addressed.Dissolved Trace Elements Associated With Water Samples Contamination Over the past 10 years, reported ambient (background) dissolved trace element concentrations have declined from tens of parts-per-billion (mg l21) through single digit parts-perbillion, to the parts-per-trillion (ng l21) range.3–6 These decreases generally do not reflect improved water quality, but † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Analyst, November 1997, Vol. 122 (1193–1200) 1193rather, reductions in contamination introduced during sampling, processing and analysis. The procedures required to reduce contamination have increased the costs for all types of waterquality studies and have made it difficult for many developed countries, let alone undeveloped countries, to generate accurate dissolved trace element data.However, monitoring programs typically require data on dissolved constituents; hence a relatively simple, user-friendly protocol for the collection and processing of uncontaminated samples must be available to permit trace element quantitation at ambient levels.7 Further, laboratory facilities must be sufficiently clean to prevent contamination during sample preparation and analyses (e.g., portions of the work must be performed in ‘clean’ rooms).Even with such protocols and facilities, processed field blanks indicate that sampling and/or laboratory personnel are currently unable to reach the lowest concentration levels achievable with ICP-based instrumentation, for all trace elements of concern. 18 Processing artifacts Reductions in contamination and concomitant decreases in reported ambient dissolved trace element levels have led to the discovery that the currently accepted operational definition of dissolved constituents (filtration of unspecified volumes of natural water through unspecified 0.45 mm membrane filters) is inadequate owing to sampling and processing (filtration) artifacts.9,10 In fact, once significant sample contamination has been eliminated, ‘dissolved’ trace element concentrations appear to result predominantly from colloidal material passing through 0.45 mm membrane filters, rather than to truly dissolved constituents.11 As such, the inclusion/exclusion of colloids has a major impact on reported ‘dissolved’ trace element concentrations.Studies have indicated that the inclusion/exclusion of colloidal material can be affected by such diverse factors as (1) filter type, (2) filter diameter, (3) filtration method, (4) suspended sediment concentration, (5) suspended sediment grain-size distribution, (6) concentration of colloids, (7) concentration of organic matter, (8) volume of sample processed because of points (4)–(6) and (9) method of sample collection, again because of points (4)–(6).9 The issue of filtration artifacts raises questions about the utility/useability of dissolved trace element data for trend identification, for the generation of accurate chemical averages for comparison/ regulatory purposes and for the determination of national and/or worldwide chemical averages, since these all rely on information from multiple sources which may or may not be using equivalent sample-processing procedures. Hence any successful monitoring program that calls for data on ‘dissolved’ concentrations requires consistent sampling and processing protocols that address the problem(s) of filtration (processing) artifacts.As an example, examine the data from a field study designed to evaluate the effect(s) of using different (e.g., manufacturer, diameter) 0.45/0.40 mm membrane filters to field process whole-water samples for subsequent ‘dissolved’ trace element quantification (Table 1).Three filters were used: (1) a 47 mm, 0.40 mm polycarbonate Nuclepore‡ plate filter with a surface area of 17.3 cm2, (2) a 142 mm, 0.45 mm cellulose nitrate MicroFiltration Systems (MFS) plate filter with a surface area of 158 cm2 and (3) a 47 mm, 0.45 mm polyethersulfone Gelman capsule filter with a surface area of 600 cm2. The Nuclepore filter was selected because it is used by numerous aquatic chemists;3,5 the MFS was selected because it represented a filter type used extensively by the USGS and other US monitoring agencies until the development of a new sampling and processing protocol;7 and the Gelman capsule was selected because it was used by Windom et al.4 and currently is used by the USGS7 and recommended by the US Environmental Protection Agency (EPA).Whole-water samples were processed through each filter type, and sequential aliquots (250 ml if possible) collected.The Nuclepore filters were the slowest, and processed the smallest filtrate volume (approximately 100 ml). MFS filters were faster, but the filtration rates declined after 750 to 800 ml had been processed. Gelman capsules did not appear to slow even after 3500 ml had been processed. Although the Tangipahoa River sample did not have as much suspended sediment as the Mississippi River sample, it contained more organic matter. In fact, the processing rates for all three filters were slower for the Tangipahoa than for the Mississippi samples.In almost every case, field blank concentrations were either below the detection limits or were sufficiently low relative to the measured concentrations as to be insignificant. The analytical precision, based on split field samples, typically was better than 5%. The detectable Mississippi dissolved trace element data fell into one of three categories: (1) affected by filtration artifacts; (2) possibly affected by filtration artifacts and/or dilution; and (3) not affected by either filtration artifacts or dilution (Table 1).The first group includes Al, Fe, Ni, Cu and Zn. The general pattern always was the same: the lowest concentrations occurred in the Nuclepore filtrates, the next highest occurred in the MFS filtrates and the highest occurred in the Gelman filtrates. This pattern is proportional to the respective surface areas of the three filters, and may imply a correlation with filtration rate and/or filter clogging.The effects on Al and Fe are pronounced (Table 1). Note that the 8.2 mg l21 Al level in the 100 ml Nuclepore filtrate was not reached in the MFS filtrate until 500 ml had been processed, and was not reached in the Gelman filtrate until over 2000 ml had been processed. In fact, the Gelman filtrates eventually contained lower Al levels than the Nuclepore filtrates, but only after 2600 ml had been processed. The patterns for Ni, Cu and Zn are not as strong as those for Al and Fe.The Cu concentration (50 mg l21) in the first 250 ml Gelman aliquot was ‘verified’ (it was determined in both halves of a split field sample) but almost certainly represented some form of contamination introduced during sample handling. Even so, the second Gelman aliquot (250–500 ml) contained more Cu than either the first Nuclepore or MFS aliquots. There appears to be a correlation between filter surface area and the Fe and Al concentrations in the filtrates.However, although filtration artifacts appear to have affected Ni, Cu and Zn, and despite the sorptive capacity of Fe and Al oxides and hydroxides for these elements, there does not appear to be a direct correlation between Ni, Cu and Zn concentrations and filter surface area. These conclusions are similar to others reported elsewhere. 9,11 Co, Mo, Pb and possibly Cr concentrations appear to have been affected by a different type of artifact associated only with Gelman filters (Table 1), and may be a function of their high surface area and/or composition (polyethersulfone). Note that the filtrate concentrations of these elements increased after processing 1500 ml.This may indicate that some proportion of these elements had been sorbed to the filter during initial processing and then, once all the potential sorption sites had been filled, higher concentrations began to appear in the filtrate. The concentrations of several elements (e.g., Sr, Ba and Ca) appear to have been affected by dilution due to some entrained deionized water that was used to condition the filters.Note the concentration increases in the second and third filtrate aliquots, relative to the first aliquot, for these elements for the Gelman and MFS filters (Table 1). This effect has been noted previously.7 If dilution affects trace element concentrations, it ‡ The use of brand names is for identification purposes only, and does not represent an endorsement by the US Geological Survey. 1194 Analyst, November 1997, Vol. 122should affect them all equally; however, as a result of low concentrations and/or limited analytical sensitivity, it is not significant below 50 mg l21. On the other hand, this may indicate that some proportion of these elements initially had been sorbed to the filter and then, once all the potential sorption sites had been filled, higher concentrations began to appear in the filtrate.Although data were not included, the concentrations of a variety of dissolved nutrients (nitrate, nitrite, ammonium, orthophosphate and total phosphorus) did not appear to be affected by filtration artifacts or dilution. In fact, constituents occurring at !mg l21 levels apparently are not affected by these artifacts (Table 1). The Tangipahoa River data are similar to those for the Mississippi River (Table 1). In most cases, the elements affected by filtration artifacts in the Mississippi sample also were affected in the Tangipahoa sample, with the exception of Pb: in the Tangipahoa sample Pb appears to be affected by filtration artifacts in the same way as Al, Fe, Cu, and Zn, whereas in the Mississippi sample Pb did not appear to be affected by the exclusion of colloids due to filter clogging, but may have been affected by initial sorption on the Gelman filter.The sorption artifacts noted in the Mississippi sample for Cr, Co and Mo, could not be evaluated in the Tangipahoa sample owing to the lower concentrations present.All the unaffected elements, or those possibly affected by dilution in the Mississippi sample, also displayed similar patterns in the Tangipahoa sample (Table 1). Three viable options are available for dealing with filtration artifacts; the last two entail substantive changes in the current operational definition of dissolved constituents. (1) The arti- Table 1 Comparison of selected dissolved trace element concentrations in sample filtrates from the Mississippi River at St.Francisville and the Tangipahoa River at Robert, LA. All concentrations in mg l21 except for Ca, Mg, Na and Si (mg l21). All filters were washed/preconditioned with deionized water.7 Blanks were run through each filter prior to processing any samples and were either below the method detection limit and/or insignificant relative to the measured concentrations Analytical method* and constituent Aliquot MS OES MS MS MS MS MS MS MS MS MS MS OES OES OES OES OES OES River Filter† volume/ml Al Fe Cr Mn Co Ni Cu Zn Mo Ba Pb U B Sr Ca Mg Na Si Mississippi‡ Nuclepore 0–100 8.2 7 0.5 14 0.4 < 0.8 1.8 2.1 1.3 57 < 0.2 0.9 33 160 39 11.1 19.5 7.0 MicroFiltration systems 0–250 26 27 1.3 14 0.4 < 0.8 2.2 4.6 1.3 54 < 0.2 0.9 31 153 37 10.7 18.2 6.7 250–500 8.6 14 0.8 14 0.5 < 0.8 2.1 3.8 1.5 58 < 0.2 1.0 34 164 39 11.3 19.8 7.1 500–750 4.9 4 0.9 14 0.6 < 0.8 2.4 4.1 1.5 58 < 0.2 0.9 31 161 39 11.3 19.4 7.1 Gelman capsule 0–250 43 58 < 0.3 13 0.2 2.2 50 6.1 1.2 52 < 0.2 0.9 31 152 37 10.6 18.6 6.9 250–500 41 56 0.3 14 0.4 2.0 3.9 2.7 1.2 57 < 0.2 1.0 32 160 39 11.2 19.6 7.2 500–750 32 50 < 0.3 14 0.3 1.7 2.2 2.5 1.1 59 < 0.2 1.1 34 163 39 11.3 19.7 7.2 750–1000 27 43 0.3 14 0.3 2.0 1.9 1.5 1.3 55 0.6 1.1 37 155 39 11.3 19.7 7.2 1125–137 22 34 0.4 14 0.4 2.1 2.2 2.4 1.5 56 0.6 1.1 31 161 39 11.2 19.3 7.1 1375–1625 18 36 0.4 13 0.4 2.5 2.2 1.9 1.7 56 0.8 1.1 34 163 40 11.3 19.2 7.2 1625–1875 15 19 0.5 14 0.4 3.0 2.5 2.2 1.8 57 0.7 1.1 32 161 39 11.2 19.5 7.0 1875–2125 11 24 0.6 13 0.5 2.3 2.7 2.8 2.0 58 0.7 1.1 32 161 39 11.2 19.5 7.1 2125–2375 8.6 20 0.6 14 0.5 2.6 2.5 2.1 1.8 57 0.9 1.0 32 162 39 11.2 19.4 7.1 2375–2625 8.2 20 0.7 13 0.5 2.8 3.0 2.7 1.9 58 1.0 1.0 34 164 40 11.4 19.6 7.1 2625–2875 7.4 17 0.6 14 0.5 2.6 2.7 2.5 1.9 59 0.9 1.1 32 161 39 11.2 19.9 7.0 2875–3125 6.8 15 0.6 14 0.5 2.9 2.7 2.6 2.0 59 1.0 1.1 31 161 39 11.1 19 7.0 3125–3375 3.4 13 < 0.3 14 0.3 1.9 2.1 2.4 1.2 58 < 0.2 1.1 33 162 39 11.2 19.6 7.0 Analytical method and constituent MS OES MS MS MS MS MS MS MS OES OES OES OES OES OES Al Fe Mn Co Ni Cu Zn Ba Pb B Sr Ca Mg Na Si Tangipahoa Nuclepore 0–100 36 55 75 < 0.2 < 0.8 1.0 2.2 24 < 0.2 10 13 1.3 0.7 2.3 4.2 MicroFiltration systems 0–250 185 252 81 0.3 < 0.8 1.4 4.1 28 0.5 13 13 1.4 0.7 2.3 4.7 250–500 66 109 81 < 0.2 < 0.8 0.7 2.5 27 < 0.2 11 14 1.4 0.7 2.4 4.4 625–875 42 74 77 < 0.2 < 0.8 0.8 1.8 28 < 0.2 11 14 1.4 0.7 2.4 4.4 Gelman capsule 0–250 320 302 72 0.3 1.4 4.8 5.6 27 0.8 11 12 1.5 0.6 2.2 4.4 250–500 308 289 89 0.3 < 0.8 1.6 3.1 33 0.5 10 14 1.5 0.7 2.4 4.7 500–750 153 248 81 0.3 < 0.8 2.0 3.0 30 0.4 11 14 1.5 0.7 2.4 4.7 750–1000 145 228 85 0.4 < 0.8 1.5 2.6 31 0.3 12 14 1.4 0.7 2.4 4.6 1125–1375 64 169 82 0.2 < 0.8 1.6 2.2 30 < 0.2 13 14 1.5 0.7 2.4 4.6 1375–1625 75 143 83 0.3 < 0.8 1.4 2.2 30 < 0.2 12 14 1.5 0.7 2.4 4.5 1625–1875 50 142 81 < 0.2 < 0.8 0.9 2.3 28 < 0.2 11 13 1.5 0.7 2.4 4.3 1875–2125 54 105 80 < 0.2 < 0.8 0.8 0.9 27 < 0.2 12 14 1.2 0.7 2.4 4.4 2250–2500 54 98 79 < 0.2 < 0.8 0.8 0.8 28 < 0.2 12 14 1.5 0.7 2.4 4.4 2500–2750 45 95 79 < 0.2 < 0.8 0.7 0.8 28 < 0.2 12 14 1.4 0.7 2.4 4.4 2750–3000 44 93 79 < 0.2 < 0.8 0.3 0.6 27 < 0.2 10 14 1.4 0.7 2.3 4.4 3000–3250 44 87 80 0.2 < 0.8 0.6 1.0 28 < 0.2 11 14 1.4 0.7 2.4 4.3 * Analytical methods were inductively coupled plasma mass spectrometry (MS) and inductively coupled plasma optical emission spectrometry (OES).† Filters were a 47 mm, 0.40 mm Nucleopore, a 142 mm, 0.45 mm MicroFiltration Systems and a 47 mm, 0.45 mm Gelman capsule. ‡ The concentrations for Be (0.6 mg l21), Ag (0.2 mg l21), Cd (0.2 mg l21), Sb (0.2 mg l21), Li (1 mg l21) and V (3 mg l21) were excluded because they were all below the method detection limit (given in parentheses).The concentration of suspended sediment was 157 mg l21. § The concentrations for Be (0.6 mg l21), Cr (0.3 mg l21), Mo (0.6 mg l21), Ag (0.2 mg l21), Cd (0.2 mg l21), Sb (0.2 mg l21), U (0.1 mg l21), Li (1 mg l21) and V (3 mg l21) were excluded because they were all below the method detection limit (given in parentheses). The concentration of suspended sediment was 39 mg l21. Analyst, November 1997, Vol. 122 1195facts can be reduced/eliminated by using 0.45 mm filters with very high surface areas (e.g., capsule filters) and collecting initial aliquots for the quantitation of artifact-affected constituents. (2) Artifact-induced differences in trace element concentrations can be limited by pre-treating the sample (e.g., filtration or centrifugation) prior to filtration with a 0.45 mm membrane; however, pre-treatment enhances the chances of random contamination due to increased sample handling, and markedly lowers a number of filtrate trace element concentrations. 10 (3) Colloids should be viewed as contaminants which should be excluded from ‘dissolved’ samples. Although there is controversy over what constitutes a colloid, current data indicate that material coarser than 0.015–0.005 mm would have to be removed.11,12 This approach represents the most substantive departure from the current ‘dissolved’ definition and would require the use of multiple filters, more expensive equipment (e.g., tangential-flow filtration systems) or the use of such subjective procedures as ‘exhaustive filtration’.3,8,10–12 The magnitude of the observed chemical differences associated with variations in sample processing raises questions regarding acceptable levels of analytical imprecision and bias.In other words, considering the degree of chemical variability which can occur as a result of processing artifacts, are the low reporting/ detection limits and highly precise and unbiased analytical results achievable with ICP-based instrumentation justified when analyzing filtered water? Trace Elements Associated With Suspended Sediment Samples Recent evidence indicates that the majority of fluvial trace element transport occurs in association with suspended sediment. 13 Even in waters with suspended sediment concentrations < 10 mg l21, these solids can represent the major carrier for many trace elements; at 100 mg l21, the solid phase dominates the transport of most trace elements (Fig. 1). In water quality studies, as in other studies, the collection of representative samples is of paramount importance as it is impossible to sample and analyze an entire water body.14 Spatial variability in suspended sediment and associated trace elements When both sand-sized ( > 63 mm) and silt/clay-sized ( < 63 mm) particles are present in a stream, the concentration of suspended sediment tends to increase with increasing distance from river banks (Fig. 2). This is a common pattern15 and results from an increase in stream velocity (discharge) due to decreasing frictional resistance from the river banks and the river bed (in shallow water). Note that the cause of the increased suspended sediment concentrations is an increase in the amount of sandsized ( > 63 mm) material (Fig. 2). These concentration changes can occur over relatively short distances; in the case of the Arkansas River sample, the distances were as short as 3 m (Fig. 2). For suspended sediment, there is a concomitant decrease in the concentrations of most trace elements with increasing distance from the river banks (Fig. 3). This decrease occurs as a result of the increase in sand-sized ( > 63 mm) particle concentration because the coarser material typically contains lower trace element concentrations than the finer siltand clay-sized material.16–18 Vertical concentrations of fluvial suspended sediment tend to increase with increasing depth; this also is due to an increase in sand-sized ( > 63 mm) material (Fig. 4). This occurs because the velocity (discharge) in most rivers, under normal flow conditions, is insufficient to distribute coarse material homogeneously. Hence the majority of sand-sized particles tend to be transported near the river bed. The increase in sand-sized particle concentration, from top to bottom, leads to a concomitant decrease in sediment-associated trace element concentrations (Fig. 4). As with the horizontal variations noted above, these changes can occur over very short distances. In the case of the Arkansas River sample, the changes in suspended sediment and associated trace element concentrations occurred over distances as small as 0.2–0.5 m (Fig. 4). Fig. 1 Solid-phase contributions to the concentration of trace elements for a typical whole-water (suspended plus dissolved phases) sample for (upper) 10 and (lower) 100 mg l21 suspended sediment concentrations (from ref. 18). Fig. 2 Horizontal cross-sectional changes in suspended sediment concentration for the Arkansas (total width = 32 m) and Cowlitz (128.1 m) Rivers based on isokinetic depth-integrated vertical samples. The numbers following the D are distances from the left bank of the river, in meters. 1196 Analyst, November 1997, Vol. 122It should be apparent that the collection of a ‘grab’ sample at a single depth, from the centroid of flow or from one bank is unlikely to produce representative samples of suspended sediment and associated trace elements. Experience has indicated that representative suspended sediment sampling requires a composite of a series of depth- and width-integrated, isokinetic samples obtained either at equal discharge or at equal width increments across a river.19–22 Without committing the Fig. 3 Horizontal cross-sectional changes in suspended sediment-associated trace element concentrations in depth-integrated isokinetic vertical samples for the Arkansas River (total width = 32 m).The numbers following the D are distances from the left bank of the river, in meters. Fig. 4 Vertical cross-sectional changes in suspended sediment and associated trace element concentrations for the Arkansas River based on isokinetic point samples at 20, 40, 60 and 80% of depth. Vertical VB was 6.45 m from the left bank and vertical VA was 17.4 m from the left bank. The entire cross-section was 32 m wide.Hence VA was much nearer the centroid of flow than VB. The numbers following either VA or VB refer to the percentage of depth. Analyst, November 1997, Vol. 122 1197requisite resources to collect representative whole-water samples, are the precise and unbiased analytical results achievable with ICP-based instrumentation justified when analyzing fluvial suspended sediment samples? Temporal variability in suspended sediment and associated trace elements Although a number of factors other than just discharge (velocity) are involved (e.g., grain-size distribution, shear stress, turbulence, stream-bed gradient), there is a widely held belief that in fluvial systems, as discharge increases, suspended sediment concentration also increases.15–17 The presumption is that as discharge increases, the percentage of sand-sized particles will increase and there will be a concomitant decrease in associated trace element concentrations.16 However, these decreases in sample chemical concentrations do not necessarily imply decreases in trace element fluxes becuse the lower concentrations are accompanied by increases in both discharge and suspended sediment concentration.On the other hand, several studies have indicated that as discharge increases, at least initially, suspended sediment concentrations can become finer grained and the concentrations of both the associated trace elements and their fluxes can increase.22–24 Data from the Arkansas River, collected over a 2 h period when discharge remained constant at 56.5 m3 s21, indicate that the posited interrelations between discharge, suspended sediment and associated trace element concentrations probably do not hold in all cases (Fig. 5).During the initial 85 min of the sampling program, suspended sediment concentrations were nearly constant at 556 ± 19 mg l21; the grain size distribution also remained nearly constant (proportion < 63 mm = 37 ± 1.2% and > 63 mm = 63 ± 1.2%).However, during the last 25 min of sampling, suspended sediment concentrations increased by nearly 60% to 886 mg l21. This concentration change was caused by increases in both the > 63 mm and the < 63 mm fractions. The relative proportions of the two size ranges shifted slightly; there was a marginal increase in the > 63 mm percentage (Fig. 5). Interestingly, the change in suspended sediment concentration and the relative proportions of the > 63 mm and the < 63 mm fractions did not produce a significant change in suspended sediment chemistry.22 Similar results have been observed in other fluvial systems.25–27 While the Arkansas River was being sampled on May 29, 1987, a thunderstorm occurred upstream from the sampling site.Storm runoff increased turbidity in the river during and after the fourth set of samples (Fig. 6). The effects of the storm were reflected in the analyses for the 80 and particularly for the 100 and 105 min samples.In the 80 min sample, the most apparent effects were increased concentrations of Pb and Zn and, to a lesser extent, Cu. In the 100 min sample, the effects displayed for these elements were more substantial (Fig. 6). Comparison of the data from the 80 and 100 min samples with those from samples collected earlier indicates that suspended sediment concentration increased by 26%, the concentration of < 63 mm material increased by 60% and the concentrations of Cu, Zn, and Pb increased 2–9-fold. Although the percentage changed, the actual concentration of the > 63 mm fraction either remained constant or decreased slightly.Hence the observed changes in suspended sediment and associated trace element concentrations occurred as a result of a significant increase in the amount of < 63 mm material in suspension. The increases noted in the 100 min sample continued into the 105 min sample.It should be noted that the marked changes in chemistry became noticeable over a very short time period, of the order of only 20–40 min. Under normal sampling conditions in a river, a depth- and width-integrated sample can be obtained within 0.5–2 h. Based on an examination of the discharge records for the site, the effects of the storm lasted for a total of about 5 h. If a sample had been collected at the time of the first composite (the 20 min sample), or 2–3 h after the 105 min sample, the effects of the thunderstorm would not have been detected.If sampling had begun around the time of the 40–60 min sample, then only a portion of the storm’s contribution would have been collected and quantified. Although the suspended sediment-associated trace element chemical changes that occurred in the Arkansas River during the storm were marked, their duration was limited. Hence, in the context of annual transport at this site, where this storm event represented only 0.06% of the year (5 h out of 8760 h), and the discharge represented an even smaller percentage, the impact of the storm probably was insignificant.On the other hand, if a sample had been collected during the storm as part of a scheduled sampling program (e.g., once per month) for Fig. 5 Temporal variations in suspended sediment concentration in the entire cross-section of the Arkansas River for a nearly 2 h period. During the entire sampling operation, discharge remained constant at 56.5 m3 s21 based on a constant stage of 1.30 m.Fig. 6 Temporal changes in suspended sediment and associated trace element concentrations during a storm event on the Arkansas River. 1198 Analyst, November 1997, Vol. 122estimating annual flux, it would have almost certainly led to a major overestimate of annual transport. Here again, as with the discussion concerning spatial variability, it should be apparent that suspended sediment and associated trace elements can display marked temporal variability.Most monitoring programs lack the requisite resources to sample with sufficient frequency to encompass the degree of temporal variability typical in most fluvial systems. As such, are the low reporting/detection limits and precise and unbiased analytical results achievable with ICP-based instrumentation justified when analyzing only a limited number of fluvial suspended sediment samples that do not encompass potential temporal variability? Choice of sample media (whole-water versus separated suspended sediment) for the determination of associated trace element concentrations Historically, most water-quality investigations have attempted to assess suspended sediment-associated trace elements in aquatic systems by determining the concentrations of total recoverable (whole-water) and dissolved trace elements through the collection and analysis, respectively, of unfiltered and filtered water.However, at typical suspended sediment (@70 mg l21) and associated trace element concentrations, whole-water total recoverable analyses generally do not provide an accurate measure of trace element concentrations owing to dilution effects and limitations in analytical techniques and equipment.Hence the use of whole-water total recoverable trace element data to estimate suspended sediment-associated trace element concentrations should be discouraged. As an example, examine the data for a sample from the Susquehanna River in which the sediment concentration was 4 mg l21 (Table 2).Suspended sediment chemical analyses were conducted after the solids had been physically separated from their water matrix by flow-through centrifugation. The chemical concentrations for several trace elements are elevated (Ag, Zn, Ni, Co, Cd, Cr and As; ‘Concentration’, Table 2). On the other hand, when the chemical data for the suspended sediment are converted back to whole-water sample values, the concentrations appear low (‘Recalculated whole-water concentration’, Table 2).Comparison of the recalculated whole-water concentrations with those for currently accepted dissolved concentrations from unimpacted areas indicates that the suspended sediment accounts for a significant proportion of the total concentration of many of the trace elements (‘Calculated % of solid-phase contributions’, Table 2). Despite this, if the wholewater concentrations are compared with typical reporting limits for many water quality laboratories, no sediment-associated trace element concentrations would have been detected because the whole-water values were less than their respective reporting limits [‘USGS National Water Quality Laboratory (NWQL) reporting limit, Table 2’].These data show a major problem with the determination of suspended sediment-associated trace element concentrations using whole-water samples and the ‘method of difference’ [subtracting the concentrations from a filtered (‘dissolved’) sample from the concentrations from a whole-water (suspended sediment plus water) sample].To place this problem in an appropriate context, it helps to see what chemical concentrations would occur if a whole water sample were ‘created’ using a sediment containing average trace element concentrations (Average sediment-associated trace element concentration’, Table 2). The values represent typical chemical levels associated with fine-grained sediment samples collected in unimpacted areas.21 Using currently available reporting limits, it is possible to calculate the minimum suspended sediment concentration required before each trace element could be detected in a whole-water sample (‘Mass required to reach NWQL reporting limits’, Table 2).Considering the typical median suspended sediment concentration (e.g., about 70 mg l21 (ref. 28), relative to the requisite masses listed, many of the trace elements would be at or below current reporting limits.The problem actually is far worse because the calculated mass requirements assume that all the trace elements are quantified (a total analysis). Typically, this is not the case for whole-water analyses because the presence of water in the samples, and also the digestion procedures used, preclude complete solubilization/quantification of all the entrained trace elements. Table 2 Sediment-associated trace element data for a suspended sediment sample from the Susquehanna River containing 4 mg l21 suspended sediment Parameter Ag Cu Pb Zn Ni Co Cd Cr As Sb Se Fe Mn Al Ti Concentration/mg g21* 1.8 50 58 450 120 77 1.2 123 17.2 2.0 1.3 57 000 6400 93 000 4800 Recalculated whole-water concentration/mg l21† 0.01 0.2 0.2 1.8 0.5 0.3 0.01 0.5 0.07 0.01 0.01 228 25.6 372 19.2 Average dissolved trace element concentration/ mg l21‡ 0.2 0.05 0.2 0.3 0.05 0.01 0.1 0.5 0.05 0.08 Calculated % of solidphase contributions to whole-water concentration 50 82 90 62 51 50 83 12 17 11 NWQL reporting limit/ mg l21¶ 1 1 1 3 1 1 1 1 1 1 1 3 1 3 Average sedimentassociated trace element concentration/mg g21· 0.5 25 50 100 25 18 0.6 20 7.0 0.6 0.4 Mass required to reach NWQL reporting limits/ mg** 2000 30 20 33 40 55 1650 250 140 1650 2500 * Trace element concentrations in a separated (centrifuged), freeze-dried and totally digested suspended sediment sample from the Susquehanna River.† Calculated whole-water cocentrations of the suspended sediment sample (as above) based on a suspended sediment concentration of 4 mg l21.‡ Average dissolved concentrations from unimpacted areas. § Calculated percentage contributions of the suspended sediment-associated trace element concentrations using dissolved concentrations reported for relatively clean areas. ¶ Current reporting limits for the USGS NWQL for dissolved and/or digested whole-water samples. · Average total trace element concentrations for fine-grained, unimpacted bed sediments.** Calculated suspended sediment concentration required to produce a whole-water concentration equal to current USGS NWQL reporting limits using the average total trace element concentrations according to footnote ¶, in conjunction with the reporting limits given according to footnote §. Analyst, November 1997, Vol. 122 1199Based on the foregoing, it is apparent that the low reporting/ detection limits and highly precise and unbiased analytical results achievable with ICP-based instrumentation are a requisite for dealing with whole-water samples containing low suspended sediment concentrations; otherwise, suspended sediment- associated trace element contributions would not be quantifiable.However, there are two caveats to this conclusion: (1) contamination levels associated with sampling and sample processing must be sufficiently low as to be insignificant at achieveable reporting/detection limits; and (2) spatially and temporally representative samples must be obtained.Conclusions Modern analytical instruments, particularly ICP-AES and ICPMS, are capable of measuring numerous trace elements down to the single or double digit parts-per-trillion range. This holds for the three most common sample media currently used in environmental monitoring programs: filtered water, wholewater and separated suspended sediment. However, current analytical capabilities have exceeded the capacity to collect both uncontaminated and representative environmental samples from fluvial systems. For ‘dissolved’ (filtered water) trace element concentrations, problems of sample contamination during collection, processing, preservation and subsequent chemical analyses, and also variations introduced by processing artifacts, are of primary concern.For suspended sedimentassociated trace element concentrations, problems related to collecting representative samples (ones that encompass the range of spatial and temporal variability at a site) are of primary concern. In either case (‘dissolved’ or suspended sedimentassociated trace elements), laboratory chemists, in conjunction with field personnel and data end-users, need to determine if the low reporting/detection limits and highly precise and unbiased analytical results achievable with modern ICP-based instrumentation are justified in the light of the variety of potential problems associated with sampling and sample processing.References 1 van Loon, J. C., Selected Methods of Trace Metal Analysis: Biological and Environmental Samples, Wiley, New York, 1985, pp. 36–39. 2 Thompson, M., and Walsh, J. N., Handbook of Inductively Coupled Plasma Spectrometry, Chapman and Hall, New York, 2nd edn., pp. 16–43 and 238–269. 3 Shiller, A. M., and Boyle, E., Geochim. Cosmoschim. Acta, 1987, 51, 3273. 4 Windom, H. L., Byrd, J. T., Smith, R.G., Jr., and Feng, H., Environ. Sci. Technol., 1991, 25, 1137. 5 Benoit, G., Environ. Sci. Technol., 1994, 28, 1987. 6 Nriagu, J. O., Lawson, G., Wong, H. K., and Cheam, V., Environ. Sci. Technol., 1996, 30, 178. 7 Horowitz, A. J., Demas, C. R., Fitzgerald, K. K., Miller, T. L., and Rickert, D. A., US Geological Survey Protocol for the Collection and Processing of Surface-Water Samples for the Subsequent Determination of Inorganic Constituents, US Geological Survey Open-File Report 94-539, US Government Printing Office, Washington, DC 1994. 8 Horowitz, A. J., Lum, K. R., Garbarino, J. R., Hall, G. E. M., Lemieux, C., and Demas, C. R., Environ. Sci. Technol., 1996, 30, 3398. 9 Horowitz, A. J., Elrick, K. A., and Colberg, M. R., Water Res., 1992, 26, 753. 10 Horowitz, A. J., Lum, K. R., Garbarino, J. R., Hall, G. E. M., Lemieux, C., and Demas, C. R., Environ. Sci. Technol., 1996, 30, 954. 11 Karlsson, S., Peterson, A., Hakansson, K., and Ledin, A., Sci.Total Environ., 1994, 194, 215. 12 Taylor, H. R., and Shiller, A. M., Environ. Sci. Technol., 1995, 29, 1313. 13 Horowitz, A. J., The Use of Suspended Sediments and Associated Trace Elements in Water Quality Studies, IAHS Special Publication No. 4, IAHS Press, Wallingford, 1995. 14 Childress, C. J., Chaney, T. H., Myers, D., Norris, J. M., and Hren, J., Water Data Collection Activities in Colorado and Ohio: Phase II— Evaluation of 1984 Field and Laboratory Quality Assurance Practices, US Geological Survey Open-File Report 87–33, US Geological Survey, Columbus, OH, 1987, pp. 22–25. 15 Vanoni, V. A., Sedimentation Engineering, American Society of Civil Engineers Manuals and Reports on Engineering Practice No. 54, American Society of Civil Engineers, New York, 1977, pp. 154–190 and 317–349. 16 Forstner, U., and Wittmann, G. T. W., Metal Pollution in the Aquatic Environment, Springer, New York, 1981, pp. 71–196. 17 Salomons, W., and Forstner, Metals in the Hydrocycle, Springer, New York, 1984, pp. 63–92. 18 Horowitz, A. J., A Primer on Sediment-Trace Element Chemistry, Lewis, Chelsea, MI, 2nd edn., 1991. 19 Feltz, H. R., and Culbertson, J. K., Pestic. Monit. J., 1972, 6, 171. 20 Office of Water Data Coordination, National Handbook of Recommended Methods for Water Data Acquisition, US Geological Survey, Reston, VA, 1982, ch. 5, pp. 3-17–3-27. 21 Horowitz, A. J., Elrick, K. A., and Hooper, R. P., Hydrological Processes, 1989, 3, 347. 22 Horowitz, A.J., Rinella, F. A., Lamothe, P., Miller, T. L., Edwards, T. K., Roche, R. L., and Rickert, D. A., Environ. Sci. Technol., 1990, 24, 1313. 23 Walling, D. E., and Moorhead, P. W., Geograf. Ann., 1987, 69A, 47. 24 Mossa, J., in Sediment and the Environment, ed. Hadley, R. F., and Ongley, E. D., IAHS Publication No. 184, IAHS Press, Wallingford, 1989, pp. 105–112. 25 de Groot, A. J., and Allersma, E., in Heavy Metals in the Aquatic Environment, ed. Krenkel, P. A., Pergamon Press, Oxford, 1975, pp. 85–95. 26 Chapman, P. M., Romberg, G. P., and Vigers, G. A., J. Water Pollut. Control Fed., 1982, 54, 292. 27 Walling, D. E., Webb, B. W., and Woodward, J. C., in Erosion and Sediment Transport Monitoring Programmes in River Basins, ed. Bogen, J., Walling, D. E., and Day, T. J., IAHS Publication No. 210, IAHS Press, Wallingford, 1992, pp. 279–288. 28 Alexander, R., Ludtke, A., Fitzgerlad, K., and Schertz, T., Data from Selected US Geological Survey National Stream Water-Quality Monitoring Networks (WQN) on CD-ROM, US Geological Survey Open-File Report 96-337 US Geological Survey, Reston, VA, 1996.Paper 7/04604I Received July 1, 1997 Accepted September 17, 1997 1200 Analyst, November 1997, Vol. 122 Some Thoughts on Problems Associated With Various Sampling Media Used for Environmental Monitoring† Arthur J. Horowitz US Geological Survey, Peachtree Business Center, Suite 130, 3039 Amwiler Road, Atlanta, GA 30360, USA.E-mail: horowitz@usgs.gov Modern analytical instrumentation is capable of measuring a variety of trace elements at concentrations down into the single or double digit parts-per-trillion (ng l21) range. This holds for the three most common sample media currently used in environmental monitoring programs: filtered water, whole-water and separated suspended sediment. Unfortunately, current analytical capabilities have exceeded the current capacity to collect both uncontaminated and representative environmental samples. The success of any trace element monitoring program requires that this issue be both understood and addressed.The environmental monitoring of trace elements requires the collection of calendar- and event-based dissolved and suspended sediment samples. There are unique problems associated with the collection and chemical analyses of both types of sample media. Over the past 10 years, reported ambient dissolved trace element concentrations have declined.Generally, these decreases do not reflect better water quality, but rather improvements in the procedures used to collect, process, preserve and analyze these samples without contaminating them during these steps. Further, recent studies have shown that the currently accepted operational definition of dissolved constituents (material passing a 0.45 mm membrane filter) is inadequate owing to sampling and processing artifacts. The existence of these artifacts raises questions about the generation of accurate, precise and comparable ‘dissolved’ trace element data.Suspended sediment and associated trace elements can display marked short- and long-term spatial and temporal variability. This implies that spatially representative samples only can be obtained by generating composites using depth- and width-integrated sampling techniques. Additionally, temporal variations have led to the view that the determination of annual trace element fluxes may require nearly constant (e.g., high-frequency) sampling and subsequent chemical analyses.Ultimately, sampling frequency for flux estimates becomes dependent on the time period of concern (daily, weekly, monthly, yearly) and the amount of acceptable error associated with these estimates. Keywords: Environmental monitoring; water quality; sampling; trace elements Since the early to mid-1980s, trace element analyses of a variety of environmental samples have been increasingly performed using inductively coupled plasma (ICP)-based instrumentation as opposed to atomic absorption-based instrumentation.1,2 ICP techniques provide several distinct advantages; the two most obvious are (1) multi-element analyses from a single sample aspiration, as opposed to single element analyses, and (2) calibration curves that tend to be linear over much larger ranges, thus reducing the need to dilute and re-analyze numerous samples. Initial ICP instrumentation was based on atomic emission (ICP-AES); this limited the number of elements that could be determined and their associated reporting/detection limits.The subsequent introduction of ICP instrumentation using mass spectrometry (ICP-MS) increased the number of trace elements that could be determined, and further improved some reporting/detection limits.2 In the field of fluvial environmental trace element monitoring, at least some of the advantages accrued by switching to ICP-based instrumentation have been mitigated by samplingrelated issues.Of particular concern are (1) sample contamination during collection, processing, preservation and subsequent chemical analyses and (2) insufficient numbers of samples to encompass the range of spatial and temporal trace element variability. In other words, current laboratory analytical capabilities have exceeded the capacity of field personnel to collect uncontaminated and representative environmental samples.As such, environmental trace element monitoring requirements, and also regulatory limits, should not be based solely on the capabilities of analytical instrumentation, but on the limitations associated with the collection, processing and preservation of field samples. The various problems associated with typical sample media used to monitor environmental trace element levels, and the approaches employed by the US Geological Survey (USGS) to deal with them, have been the subject of a substantial amount of research over the past 5–7 years.Portions of that work are summarized below. Specific details of the various field and laboratory studies used in this summary can be found in the publications cited in the various sections of this paper. Sampling Media Concentrations of dissolved and suspended sediment-associated trace elements, during baseflow and major events (e.g., spring runoff, floods) are necessary to encompass the range of trace element variability in many natural or anthropogenically impacted systems.However, there are unique problems associated with the collection and subsequent chemical analyses of both types of sample media. The success of any monitoring program, regardless of spatial (local, regional, national, global) or temporal (hourly, daily, weekly, monthly, yearly) scale, requires that these various problems be both understood and addressed. Dissolved Trace Elements Associated With Water Samples Contamination Over the past 10 years, reported ambient (background) dissolved trace element concentrations have declined from tens of parts-per-billion (mg l21) through single digit parts-perbillion, to the parts-per-trillion (ng l21) range.3–6 These decreases generally do not reflect improved water quality, but † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Analyst, November 1997, Vol. 122 (1193–1200) 1193rather, reductions in contamination introduced during sampling, processing and analysis. The procedures required to reduce contamination have increased the costs for all types of waterquality studies and have made it difficult for many developed countries, let alone undeveloped countries, to generate accurate dissolved trace element data. However, monitoring programs typically require data on dissolved constituents; hence a relatively simple, user-friendly protocol for the collection and processing of uncontaminated samples must be available to permit trace element quantitation at ambient levels.7 Further, laboratory facilities must be sufficiently clean to prevent contamination during sample preparation and analyses (e.g., portions of the work must be performed in ‘clean’ rooms).Even with such protocols and facilities, processed field blanks indicate that sampling and/or laboratory personnel are currently unable to reach the lowest concentration levels achievable with ICP-based instrumentation, for all trace elements of concern. 18 Processing artifacts Reductions in contamination and concomitant decreases in reported ambient dissolved trace element levels have led to the discovery that the currently accepted operational definition of dissolved constituents (filtration of unspecified volumes of natural water through unspecified 0.45 mm membrane filters) is inadequate owing to sampling and processing (filtration) artifacts.9,10 In fact, once significant sample contamination has been eliminated, ‘dissolved’ trace element concentrations appear to result predominantly from colloidal material passing through 0.45 mm membrane filters, rather than to truly dissolved constituents.11 As such, the inclusion/exclusion of colloids has a major impact on reported ‘dissolved’ trace element concentrations. Studies have indicated that the inclusion/exclusion of colloidal material can be affected by such diverse factors as (1) filter type, (2) filter diameter, (3) filtration method, (4) suspended sediment concentration, (5) suspended sediment grain-size distribution, (6) concentration of colloids, (7) concentration of organic matter, (8) volume of sample processed because of points (4)–(6) and (9) method of sample collection, again because of points (4)–(6).9 The issue of filtration artifacts raises questions about the utility/useability of dissolved trace element data for trend identification, for the generation of accurate chemical averages for comparison/ regulatory purposes and for the determination of national and/or worldwide chemical averages, since these all rely on information from multiple sources which may or may not be using equivalent sample-processing procedures.Hence any successful monitoring program that calls for data on ‘dissolved’ concentrations requires consistent sampling and processing protocols that address the problem(s) of filtration (processing) artifacts.As an example, examine the data from a field study designed to evaluate the effect(s) of using different (e.g., manufacturer, diameter) 0.45/0.40 mm membrane filters to field process whole-water samples for subsequent ‘dissolved’ trace element quantification (Table 1). Three filters were used: (1) a 47 mm, 0.40 mm polycarbonate Nuclepore‡ plate filter with a surface area of 17.3 cm2, (2) a 142 mm, 0.45 mm cellulose nitrate MicroFiltration Systems (MFS) plate filter with a surface area of 158 cm2 and (3) a 47 mm, 0.45 mm polyethersulfone Gelman capsule filter with a surface area of 600 cm2.The Nuclepore filter was selected because it is used by numerous aquatic chemists;3,5 the MFS was selected because it represented a filter type used extensively by the USGS and other US monitoring agencies until the development of a new sampling and processing protocol;7 and the Gelman capsule was selected because it was used by Windom et al.4 and currently is used by the USGS7 and recommended by the US Environmental Protection Agency (EPA).Whole-water samples were processed through each filter type, and sequential aliquots (250 ml if possible) collected. The Nuclepore filters were the slowest, and processed the smallest filtrate volume (approximately 100 ml). MFS filters were faster, but the filtration rates declined after 750 to 800 ml had been processed. Gelman capsules did not appear to slow even after 3500 ml had been processed.Although the Tangipahoa River sample did not have as much suspended sediment as the Mississippi River sample, it contained more organic matter. In fact, the processing rates for all three filters were slower for the Tangipahoa than for the Mississippi samples. In almost every case, field blank concentrations were either below the detection limits or were sufficiently low relative to the measured concentrations as to be insignificant.The analytical precision, based on split field samples, typically was better than 5%. The detectable Mississippi dissolved trace element data fell into one of three categories: (1) affected by filtration artifacts; (2) possibly affected by filtration artifacts and/or dilution; and (3) not affected by either filtration artifacts or dilution (Table 1). The first group includes Al, Fe, Ni, Cu and Zn. The general pattern always was the same: the lowest concentrations occurred in the Nuclepore filtrates, the next highest occurred in the MFS filtrates and the highest occurred in the Gelman filtrates. This pattern is proportional to the respective surface areas of the three filters, and may imply a correlation with filtration rate and/or filter clogging. The effects on Al and Fe are pronounced (Table 1).Note that the 8.2 mg l21 Al level in the 100 ml Nuclepore filtrate was not reached in the MFS filtrate until 500 ml had been processed, and was not reached in the Gelman filtrate until over 2000 ml had been processed.In fact, the Gelman filtrates eventually contained lower Al levels than the Nuclepore filtrates, but only after 2600 ml had been processed. The patterns for Ni, Cu and Zn are not as strong as those for Al and Fe. The Cu concentration (50 mg l21) in the first 250 ml Gelman aliquot was ‘verified’ (it was determined in both halves of a split field sample) but almost certainly represented some form of contamination introduced during sample handling.Even so, the second Gelman aliquot (250–500 ml) contained more Cu than either the first Nuclepore or MFS aliquots. There appears to be a correlation between filter surface area and the Fe and Al concentrations in the filtrates. However, although filtration artifacts appear to have affected Ni, Cu and Zn, and despite the sorptive capacity of Fe and Al oxides and hydroxides for these elements, there does not appear to be a direct correlation between Ni, Cu and Zn concentrations and filter surface area.These conclusions are similar to others reported elsewhere. 9,11 Co, Mo, Pb and possibly Cr concentrations appear to have been affected by a different type of artifact associated only with Gelman filters (Table 1), and may be a function of their high surface area and/or composition (polyethersulfone). Note that the filtrate concentrations of these elements increased after processing 1500 ml. This may indicate that some proportion of these elements had been sorbed to the filter during initial processing and then, once all the potential sorption sites had been filled, higher concentrations began to appear in the filtrate.The concentrations of several elements (e.g., Sr, Ba and Ca) appear to have been affected by dilution due to some entrained deionized water that was used to condition the filters. Note the concentration increases in the second and third filtrate aliquots, relative to the first aliquot, for these elements for the Gelman and MFS filters (Table 1).This effect has been noted previously.7 If dilution affects trace element concentrations, it ‡ The use of brand names is for identification purposes only, and does not represent an endorsement by the US Geological Survey. 1194 Analyst, November 1997, Vol. 122should affect them all equally; however, as a result of low concentrations and/or limited analytical sensitivity, it is not significant below 50 mg l21.On the other hand, this may indicate that some proportion of these elements initially had been sorbed to the filter and then, once all the potential sorption sites had been filled, higher concentrations began to appear in the filtrate. Although data were not included, the concentrations of a variety of dissolved nutrients (nitrate, nitrite, ammonium, orthophosphate and total phosphorus) did not appear to be affected by filtration artifacts or dilution.In fact, constituents occurring at !mg l21 levels apparently are not affected by these artifacts (Table 1). The Tangipahoa River data are similar to those for the Mississippi River (Table 1). In most cases, the elements affected by filtration artifacts in the Mississippi sample also were affected in the Tangipahoa sample, with the exception of Pb: in the Tangipahoa sample Pb appears to be affected by filtration artifacts in the same way as Al, Fe, Cu, and Zn, whereas in the Mississippi sample Pb did not appear to be affected by the exclusion of colloids due to filter clogging, but may have been affected by initial sorption on the Gelman filter.The sorption artifacts noted in the Mississippi sample for Cr, Co and Mo, could not be evaluated in the Tangipahoa sample owing to the lower concentrations present. All the unaffected elements, or those possibly affected by dilution in the Mississippi sample, also displayed similar patterns in the Tangipahoa sample (Table 1).Three viable options are available for dealing with filtration artifacts; the last two entail substantive changes in the current operational definition of dissolved constituents. (1) The arti- Table 1 Comparison of selected dissolved trace element concentrations in sample filtrates from the Mississippi River at St. Francisville and the Tangipahoa River at Robert, LA. All concentrations in mg l21 except for Ca, Mg, Na and Si (mg l21). All filters were washed/preconditioned with deionized water.7 Blanks were run through each filter prior to processing any samples and were either below the method detection limit and/or insignificant relative to the measured concentrations Analytical method* and constituent Aliquot MS OES MS MS MS MS MS MS MS MS MS MS OES OES OES OES OES OES River Filter† volume/ml Al Fe Cr Mn Co Ni Cu Zn Mo Ba Pb U B Sr Ca Mg Na Si Mississippi‡ Nuclepore 0–100 8.2 7 0.5 14 0.4 < 0.8 1.8 2.1 1.3 57 < 0.2 0.9 33 160 39 11.1 19.5 7.0 MicroFiltration systems 0–250 26 27 1.3 14 0.4 < 0.8 2.2 4.6 1.3 54 < 0.2 0.9 31 153 37 10.7 18.2 6.7 250–500 8.6 14 0.8 14 0.5 < 0.8 2.1 3.8 1.5 58 < 0.2 1.0 34 164 39 11.3 19.8 7.1 500–750 4.9 4 0.9 14 0.6 < 0.8 2.4 4.1 1.5 58 < 0.2 0.9 31 161 39 11.3 19.4 7.1 Gelman capsule 0–250 43 58 < 0.3 13 0.2 2.2 50 6.1 1.2 52 < 0.2 0.9 31 152 37 10.6 18.6 6.9 250–500 41 56 0.3 14 0.4 2.0 3.9 2.7 1.2 57 < 0.2 1.0 32 160 39 11.2 19.6 7.2 500–750 32 50 < 0.3 14 0.3 1.7 2.2 2.5 1.1 59 < 0.2 1.1 34 163 39 11.3 19.7 7.2 750–1000 27 43 0.3 14 0.3 2.0 1.9 1.5 1.3 55 0.6 1.1 37 155 39 11.3 19.7 7.2 1125–137 22 34 0.4 14 0.4 2.1 2.2 2.4 1.5 56 0.6 1.1 31 161 39 11.2 19.3 7.1 1375–1625 18 36 0.4 13 0.4 2.5 2.2 1.9 1.7 56 0.8 1.1 34 163 40 11.3 19.2 7.2 1625–1875 15 19 0.5 14 0.4 3.0 2.5 2.2 1.8 57 0.7 1.1 32 161 39 11.2 19.5 7.0 1875–2125 11 24 0.6 13 0.5 2.3 2.7 2.8 2.0 58 0.7 1.1 32 161 39 11.2 19.5 7.1 2125–2375 8.6 20 0.6 14 0.5 2.6 2.5 2.1 1.8 57 0.9 1.0 32 162 39 11.2 19.4 7.1 2375–2625 8.2 20 0.7 13 0.5 2.8 3.0 2.7 1.9 58 1.0 1.0 34 164 40 11.4 19.6 7.1 2625–2875 7.4 17 0.6 14 0.5 2.6 2.7 2.5 1.9 59 0.9 1.1 32 161 39 11.2 19.9 7.0 2875–3125 6.8 15 0.6 14 0.5 2.9 2.7 2.6 2.0 59 1.0 1.1 31 161 39 11.1 19 7.0 3125–3375 3.4 13 < 0.3 14 0.3 1.9 2.1 2.4 1.2 58 < 0.2 1.1 33 162 39 11.2 19.6 7.0 Analytical method and constituent MS OES MS MS MS MS MS MS MS OES OES OES OES OES OES Al Fe Mn Co Ni Cu Zn Ba Pb B Sr Ca Mg Na Si Tangipahoa Nuclepore 0–100 36 55 75 < 0.2 < 0.8 1.0 2.2 24 < 0.2 10 13 1.3 0.7 2.3 4.2 MicroFiltration systems 0–250 185 252 81 0.3 < 0.8 1.4 4.1 28 0.5 13 13 1.4 0.7 2.3 4.7 250–500 66 109 81 < 0.2 < 0.8 0.7 2.5 27 < 0.2 11 14 1.4 0.7 2.4 4.4 625–875 42 74 77 < 0.2 < 0.8 0.8 1.8 28 < 0.2 11 14 1.4 0.7 2.4 4.4 Gelman capsule 0–250 320 302 72 0.3 1.4 4.8 5.6 27 0.8 11 12 1.5 0.6 2.2 4.4 250–500 308 289 89 0.3 < 0.8 1.6 3.1 33 0.5 10 14 1.5 0.7 2.4 4.7 500–750 153 248 81 0.3 < 0.8 2.0 3.0 30 0.4 11 14 1.5 0.7 2.4 4.7 750–1000 145 228 85 0.4 < 0.8 1.5 2.6 31 0.3 12 14 1.4 0.7 2.4 4.6 1125–1375 64 169 82 0.2 < 0.8 1.6 2.2 30 < 0.2 13 14 1.5 0.7 2.4 4.6 1375–1625 75 143 83 0.3 < 0.8 1.4 2.2 30 < 0.2 12 14 1.5 0.7 2.4 4.5 1625–1875 50 142 81 < 0.2 < 0.8 0.9 2.3 28 < 0.2 11 13 1.5 0.7 2.4 4.3 1875–2125 54 105 80 < 0.2 < 0.8 0.8 0.9 27 < 0.2 12 14 1.2 0.7 2.4 4.4 2250–2500 54 98 79 < 0.2 < 0.8 0.8 0.8 28 < 0.2 12 14 1.5 0.7 2.4 4.4 2500–2750 45 95 79 < 0.2 < 0.8 0.7 0.8 28 < 0.2 12 14 1.4 0.7 2.4 4.4 2750–3000 44 93 79 < 0.2 < 0.8 0.3 0.6 27 < 0.2 10 14 1.4 0.7 2.3 4.4 3000–3250 44 87 80 0.2 < 0.8 0.6 1.0 28 < 0.2 11 14 1.4 0.7 2.4 4.3 * Analytical methods were inductively coupled plasma mass spectrometry (MS) and inductively coupled plasma optical emission spectrometry (OES).† Filters were a 47 mm, 0.40 mm Nucleopore, a 142 mm, 0.45 mm MicroFiltration Systems and a 47 mm, 0.45 mm Gelman capsule.‡ The concentrations for Be (0.6 mg l21), Ag (0.2 mg l21), Cd (0.2 mg l21), Sb (0.2 mg l21), Li (1 mg l21) and V (3 mg l21) were excluded because they were all below the method detection limit (given in parentheses). The concentration of suspended sediment was 157 mg l21. § The concentrations for Be (0.6 mg l21), Cr (0.3 mg l21), Mo (0.6 mg l21), Ag (0.2 mg l21), Cd (0.2 mg l21), Sb (0.2 mg l21), U (0.1 mg l21), Li (1 mg l21) and V (3 mg l21) were excluded because they were all below the method detection limit (given in parentheses).The concentration of suspended sediment was 39 mg l21. Analyst, November 1997, Vol. 122 1195facts can be reduced/eliminated by using 0.45 mm filters with very high surface areas (e.g., capsule filters) and collecting initial aliquots for the quantitation of artifact-affected constituents.(2) Artifact-induced differences in trace element concentrations can be limited by pre-treating the sample (e.g., filtration or centrifugation) prior to filtration with a 0.45 mm membrane; however, pre-treatment enhances the chances of random contamination due to increased sample handling, and markedly lowers a number of filtrate trace element concentrations. 10 (3) Colloids should be viewed as contaminants which should be excluded from ‘dissolved’ samples. Although there is controversy over what constitutes a colloid, current data indicate that material coarser than 0.015–0.005 mm would have to be removed.11,12 This approach represents the most substantive departure from the current ‘dissolved’ definition and would require the use of multiple filters, more expensive equipment (e.g., tangential-flow filtration systems) or the use of such subjective procedures as ‘exhaustive filtration’.3,8,10–12 The magnitude of the observed chemical differences associated with variations in sample processing raises questions regarding acceptable levels of analytical imprecision and bias.In other words, considering the degree of chemical variability which can occur as a result of processing artifacts, are the low reporting/ detection limits and highly precise and unbiased analytical results achievable with ICP-based instrumentation justified when analyzing filtered water? Trace Elements Associated With Suspended Sediment Samples Recent evidence indicates that the majority of fluvial trace element transport occurs in association with suspended sediment. 13 Even in waters with suspended sediment concentrations < 10 mg l21, these solids can represent the major carrier for many trace elements; at 100 mg l21, the solid phase dominates the transport of most trace elements (Fig. 1). In water quality studies, as in other studies, the collection of representative samples is of paramount importance as it is impossible to sample and analyze an entire water body.14 Spatial variability in suspended sediment and associated trace elements When both sand-sized ( > 63 mm) and silt/clay-sized ( < 63 mm) particles are present in a stream, the concentration of suspended sediment tends to increase with increasing distance from river banks (Fig. 2). This is a common pattern15 and results from an increase in stream velocity (discharge) due to decreasing frictional resistance from the river banks and the river bed (in shallow water).Note that the cause of the increased suspended sediment concentrations is an increase in the amount of sandsized ( > 63 mm) material (Fig. 2). These concentration changes can occur over relatively short distances; in the case of the Arkansas River sample, the distances were as short as 3 m (Fig. 2). For suspended sediment, there is a concomitant decrease in the concentrations of most trace elements with increasing distance from the river banks (Fig. 3). This decrease occurs as a result of the increase in sand-sized ( > 63 mm) particle concentration because the coarser material typically contains lower trace element concentrations than the finer siltand clay-sized material.16–18 Vertical concentrations of fluvial suspended sediment tend to increase with increasing depth; this also is due to an increase in sand-sized ( > 63 mm) material (Fig. 4). This occurs because the velocity (discharge) in most rivers, under normal flow conditions, is insufficient to distribute coarse material homogeneously.Hence the majority of sand-sized particles tend to be transported near the river bed. The increase in sand-sized particle concentration, from top to bottom, leads to a concomitant decrease in sediment-associated trace element concentrations (Fig. 4). As with the horizontal variations noted above, these changes can occur over very short distances. In the case of the Arkansas River sample, the changes in suspended sediment and associated trace element concentrations occurred over distances as small as 0.2–0.5 m (Fig. 4). Fig. 1 Solid-phase contributions to the concentration of trace elements for a typical whole-water (suspended plus dissolved phases) sample for (upper) 10 and (lower) 100 mg l21 suspended sediment concentrations (from ref. 18). Fig. 2 Horizontal cross-sectional changes in suspended sediment concentration for the Arkansas (total width = 32 m) and Cowlitz (128.1 m) Rivers based on isokinetic depth-integrated vertical samples. The numbers following the D are distances from the left bank of the river, in meters. 1196 Analyst, November 1997, Vol. 122It should be apparent that the collection of a ‘grab’ sample at a single depth, from the centroid of flow or from one bank is unlikely to produce representative samples of suspended sediment and associated trace elements. Experience has indicated that representative suspended sediment sampling requires a composite of a series of depth- and width-integrated, isokinetic samples obtained either at equal discharge or at equal width increments across a river.19–22 Without committing the Fig. 3 Horizontal cross-sectional changes in suspended sediment-associated trace element concentrations in depth-integrated isokinetic vertical samples for the Arkansas River (total width = 32 m).The numbers following the D are distances from the left bank of the river, in meters.Fig. 4 Vertical cross-sectional changes in suspended sediment and associated trace element concentrations for the Arkansas River based on isokinetic point samples at 20, 40, 60 and 80% of depth. Vertical VB was 6.45 m from the left bank and vertical VA was 17.4 m from the left bank. The entire cross-section was 32 m wide. Hence VA was much nearer the centroid of flow than VB. The numbers following either VA or VB refer to the percentage of depth. Analyst, November 1997, Vol. 122 1197requisite resources to collect representative whole-water samples, are the precise and unbiased analytical results achievable with ICP-based instrumentation justified when analyzing fluvial suspended sediment samples? Temporal variability in suspended sediment and associated trace elements Although a number of factors other than just discharge (velocity) are involved (e.g., grain-size distribution, shear stress, turbulence, stream-bed gradient), there is a widely held belief that in fluvial systems, as discharge increases, suspended sediment concentration also increases.15–17 The presumption is that as discharge increases, the percentage of sand-sized particles will increase and there will be a concomitant decrease in associated trace element concentrations.16 However, these decreases in sample chemical concentrations do not necessarily imply decreases in trace element fluxes becuse the lower concentrations are accompanied by increases in both discharge and suspended sediment concentration.On the other hand, several studies have indicated that as discharge increases, at least initially, suspended sediment concentrations can become finer grained and the concentrations of both the associated trace elements and their fluxes can increase.22–24 Data from the Arkansas River, collected over a 2 h period when discharge remained constant at 56.5 m3 s21, indicate that the posited interrelations between discharge, suspended sediment and associated trace element concentrations probably do not hold in all cases (Fig. 5). During the initial 85 min of the sampling program, suspended sediment concentrations were nearly constant at 556 ± 19 mg l21; the grain size distribution also remained nearly constant (proportion < 63 mm = 37 ± 1.2% and > 63 mm = 63 ± 1.2%). However, during the last 25 min of sampling, suspended sediment concentrations increased by nearly 60% to 886 mg l21. This concentration change was caused by increases in both the > 63 mm and the < 63 mm fractions.The relative proportions of the two size ranges shifted slightly; there was a marginal increase in the > 63 mm percentage (Fig. 5). Interestingly, the change in suspended sediment concentration and the relative proportions of the > 63 mm and the < 63 mm fractions did not produce a significant change in suspended sediment chemistry.22 Similar results have been observed in other fluvial systems.25–27 While the Arkansas River was being sampled on May 29, 1987, a thunderstorm occurred upstream from the sampling site.Storm runoff increased turbidity in the river during and after the fourth set of samples (Fig. 6). The effects of the storm were reflected in the analyses for the 80 and particularly for the 100 and 105 min samples. In the 80 min sample, the most apparent effects were increased concentrations of Pb and Zn and, to a lesser extent, Cu. In the 100 min sample, the effects displayed for these elements were more substantial (Fig. 6). Comparison of the data from the 80 and 100 min samples with those from samples collected earlier indicates that suspended sediment concentration increased by 26%, the concentration of < 63 mm material increased by 60% and the concentrations of Cu, Zn, and Pb increased 2–9-fold. Although the percentage changed, the actual concentration of the > 63 mm fraction either remained constant or decreased slightly.Hence the observed changes in suspended sediment and associated trace element concentrations occurred as a result of a significant increase in the amount of < 63 mm material in suspension. The increases noted in the 100 min sample continued into the 105 min sample. It should be noted that the marked changes in chemistry became noticeable over a very short time period, of the order of only 20–40 min. Under normal sampling conditions in a river, a depth- and width-integrated sample can be obtained within 0.5–2 h.Based on an examination of the discharge records for the site, the effects of the storm lasted for a total of about 5 h. If a sample had been collected at the time of the first composite (the 20 min sample), or 2–3 h after the 105 min sample, the effects of the thunderstorm would not have been detected. If sampling had begun around the time of the 40–60 min sample, then only a portion of the storm’s contribution would have been collected and quantified.Although the suspended sediment-associated trace element chemical changes that occurred in the Arkansas River during the storm were marked, their duration was limited. Hence, in the context of annual transport at this site, where this storm event represented only 0.06% of the year (5 h out of 8760 h), and the discharge represented an even smaller percentage, the impact of the storm probably was insignificant. On the other hand, if a sample had been collected during the storm as part of a scheduled sampling program (e.g., once per month) for Fig. 5 Temporal variations in suspended sediment concentration in the entire cross-section of the Arkansas River for a nearly 2 h period. During the entire sampling operation, discharge remained constant at 56.5 m3 s21 based on a constant stage of 1.30 m. Fig. 6 Temporal changes in suspended sediment and associated trace element concentrations during a storm event on the Arkansas River. 1198 Analyst, November 1997, Vol. 122estimating annual flux, it would have almost certainly led to a major overestimate of annual transport. Here again, as with the discussion concerning spatial variability, it should be apparent that suspended sediment and associated trace elements can display marked temporal variability. Most monitoring programs lack the requisite resources to sample with sufficient frequency to encompass the degree of temporal variability typical in most fluvial systems.As such, are the low reporting/detection limits and precise and unbiased analytical results achievable with ICP-based instrumentation justified when analyzing only a limited number of fluvial suspended sediment samples that do not encompass potential temporal variability? Choice of sample media (whole-water versus separated suspended sediment) for the determination of associated trace element concentrations Historically, most water-quality investigations have attempted to assess suspended sediment-associated trace elements in aquatic systems by determining the concentrations of total recoverable (whole-water) and dissolved trace elements through the collection and analysis, respectively, of unfiltered and filtered water.However, at typical suspended sediment (@70 mg l21) and associated trace element concentrations, whole-water total recoverable analyses generally do not provide an accurate measure of trace element concentrations owing to dilution effects and limitations in analytical techniques and equipment.Hence the use of whole-water total recoverable trace element data to estimate suspended sediment-associated trace element concentrations should be discouraged. As an example, examine the data for a sample from the Susquehanna River in which the sediment concentration was 4 mg l21 (Table 2). Suspended sediment chemical analyses were conducted after the solids had been physically separated from their water matrix by flow-through centrifugation. The chemical concentrations for several trace elements are elevated (Ag, Zn, Ni, Co, Cd, Cr and As; ‘Concentration’, Table 2).On the other hand, when the chemical data for the suspended sediment are converted back to whole-water sample values, the concentrations appear low (‘Recalculated whole-water concentration’, Table 2). Comparison of the recalculated whole-water concentrations with those for currently accepted dissolved concentrations from unimpacted areas indicates that the suspended sediment accounts for a significant proportion of the total concentration of many of the trace elements (‘Calculated % of solid-phase contributions’, Table 2).Despite this, if the wholewater concentrations are compared with typical reporting limits for many water quality laboratories, no sediment-associated trace element concentrations would have been detected because the whole-water values were less than their respective reporting limits [‘USGS National Water Quality Laboratory (NWQL) reporting limit, Table 2’].These data show a major problem with the determination of suspended sediment-associated trace element concentrations using whole-water samples and the ‘method of difference’ [subtracting the concentrations from a filtered (‘dissolved’) sample from the concentrations from a whole-water (suspended sediment plus water) sample]. To place this problem in an appropriate context, it helps to see what chemical concentrations would occur if a whole water sample were ‘created’ using a sediment containing average trace element concentrations (Average sediment-associated trace element concentration’, Table 2).The values represent typical chemical levels associated with fine-grained sediment samples collected in unimpacted areas.21 Using currently available reporting limits, it is possible to calculate the minimum suspended sediment concentration required before each trace element could be detected in a whole-water sample (‘Mass required to reach NWQL reporting limits’, Table 2).Considering the typical median suspended sediment concentration (e.g., about 70 mg l21 (ref. 28), relative to the requisite masses listed, many of the trace elements would be at or below current reporting limits. The problem actually is far worse because the calculated mass requirements assume that all the trace elements are quantified (a total analysis). Typically, this is not the case for whole-water analyses because the presence of water in the samples, and also the digestion procedures used, preclude complete solubilization/quantification of all the entrained trace elements.Table 2 Sediment-associated trace element data for a suspended sediment sample from the Susquehanna River containing 4 mg l21 suspended sediment Parameter Ag Cu Pb Zn Ni Co Cd Cr As Sb Se Fe Mn Al Ti Concentration/mg g21* 1.8 50 58 450 120 77 1.2 123 17.2 2.0 1.3 57 000 6400 93 000 4800 Recalculated whole-water concentration/mg l21† 0.01 0.2 0.2 1.8 0.5 0.3 0.01 0.5 0.07 0.01 0.01 228 25.6 372 19.2 Average dissolved trace element concentration/ mg l21‡ 0.2 0.05 0.2 0.3 0.05 0.01 0.1 0.5 0.05 0.08 Calculated % of solidphase contributions to whole-water concentration 50 82 90 62 51 50 83 12 17 11 NWQL reporting limit/ mg l21¶ 1 1 1 3 1 1 1 1 1 1 1 3 1 3 Average sedimentassociated trace element concentration/mg g21· 0.5 25 50 100 25 18 0.6 20 7.0 0.6 0.4 Mass required to reach NWQL reporting limits/ mg** 2000 30 20 33 40 55 1650 250 140 1650 2500 * Trace element concentrations in a separated (centrifuged), freeze-dried and totally digested suspended sediment sample from the Susquehanna River.† Calculated whole-water cocentrations of the suspended sediment sample (as above) based on a suspended sediment concentration of 4 mg l21. ‡ Average dissolved concentrations from unimpacted areas. § Calculated percentage contributions of the suspended sediment-associated trace element concentrations using dissolved concentrations reported for relatively clean areas.¶ Current reporting limits for the USGS NWQL for dissolved and/or digested whole-water samples. · Average total trace element concentrations for fine-grained, unimpacted bed sediments. ** Calculated suspended sediment concentration required to produce a whole-water concentration equal to current USGS NWQL reporting limits using the average total trace element concentrations according to footnote ¶, in conjunction with the reporting limits given according to footnote §.Analyst, November 1997, Vol. 122 1199Based on the foregoing, it is apparent that the low reporting/ detection limits and highly precise and unbiased analytical results achievable with ICP-based instrumentation are a requisite for dealing with whole-water samples containing low suspended sediment concentrations; otherwise, suspended sediment- associated trace element contributions would not be quantifiable.However, there are two caveats to this conclusion: (1) contamination levels associated with sampling and sample processing must be sufficiently low as to be insignificant at achieveable reporting/detection limits; and (2) spatially and temporally representative samples must be obtained. Conclusions Modern analytical instruments, particularly ICP-AES and ICPMS, are capable of measuring numerous trace elements down to the single or double digit parts-per-trillion range.This holds for the three most common sample media currently used in environmental monitoring programs: filtered water, wholewater and separated suspended sediment. However, current analytical capabilities have exceeded the capacity to collect both uncontaminated and representative environmental samples from fluvial systems. For ‘dissolved’ (filtered water) trace element concentrations, problems of sample contamination during collection, processing, preservation and subsequent chemical analyses, and also variations introduced by processing artifacts, are of primary concern.For suspended sedimentassociated trace element concentrations, problems related to collecting representative samples (ones that encompass the range of spatial and temporal variability at a site) are of primary concern. In either case (‘dissolved’ or suspended sedimentassociated trace elements), laboratory chemists, in conjunction with field personnel and data end-users, need to determine if the low reporting/detection limits and highly precise and unbiased analytical results achievable with modern ICP-based instrumentation are justified in the light of the variety of potential problems associated with sampling and sample processing. References 1 van Loon, J.C., Selected Methods of Trace Metal Analysis: Biological and Environmental Samples, Wiley, New York, 1985, pp. 36–39. 2 Thompson, M., and Walsh, J. N., Handbook of Inductively Coupled Plasma Spectrometry, Chapman and Hall, New York, 2nd edn., pp. 16–43 and 238–269. 3 Shiller, A. M., and Boyle, E., Geochim. Cosmoschim. Acta, 1987, 51, 3273. 4 Windom, H. L., Byrd, J. T., Smith, R. G., Jr., and Feng, H., Environ. Sci. Technol., 1991, 25, 1137. 5 Benoit, G., Environ. Sci. Technol., 1994, 28, 1987. 6 Nriagu, J. O., Lawson, G., Wong, H. K., and Cheam, V., Environ. Sci. Technol., 1996, 30, 178. 7 Horowitz, A. J., Demas, C. R., Fitzgerald, K. K., Miller, T. L., and Rickert, D. A., US Geological Survey Protocol for the Collection and Processing of Surface-Water Samples for the Subsequent Determination of Inorganic Constituents, US Geological Survey Open-File Report 94-539, US Government Printing Office, Washington, DC 1994. 8 Horowitz, A. J., Lum, K. R., Garbarino, J. R., Hall, G. E. M., Lemieux, C., and Demas, C. R., Environ. Sci. Technol., 1996, 30, 3398. 9 Horowitz, A. J., Elrick, K. A., and Colberg, M. R., Water Res., 1992, 26, 753. 10 Horowitz, A. J., Lum, K. R., Garbarino, J. R., Hall, G. E. M., Lemieux, C., and Demas, C. R., Environ. Sci. Technol., 1996, 30, 954. 11 Karlsson, S., Peterson, A., Hakansson, K., and Ledin, A., Sci. Total Environ., 1994, 194, 215. 12 Taylor, H. R., and Shiller, A. M., Environ. Sci. Technol., 1995, 29, 1313. 13 Horowitz, A. J., The Use of Suspended Sediments and Associated Trace Elements in Water Quality Studies, IAHS Special Publication No. 4, IAHS Press, Wallingford, 1995. 14 Childress, C. J., Chaney, T. H., Myers, D., Norris, J. M., and Hren, J., Water Data Collection Activities in Colorado and Ohio: Phase II— Evaluation of 1984 Field and Laboratory Quality Assurance Practices, US Geological Survey Open-File Report 87–33, US Geological Survey, Columbus, OH, 1987, pp. 22–25. 15 Vanoni, V. A., Sedimentation Engineering, American Society of Civil Engineers Manuals and Reports on Engineering Practice No. 54, American Society of Civil Engineers, New York, 1977, pp. 154–190 and 317–349. 16 Forstner, U., and Wittmann, G. T. W., Metal Pollution in the Aquatic Environment, Springer, New York, 1981, pp. 71–196. 17 Salomons, W., and Forstner, Metals in the Hydrocycle, Springer, New York, 1984, pp. 63–92. 18 Horowitz, A. J., A Primer on Sediment-Trace Element Chemistry, Lewis, Chelsea, MI, 2nd edn., 1991. 19 Feltz, H. R., and Culbertson, J. K., Pestic. Monit. J., 1972, 6, 171. 20 Office of Water Data Coordination, National Handbook of Recommended Methods for Water Data Acquisition, US Geological Survey, Reston, VA, 1982, ch. 5, pp. 3-17–3-27. 21 Horowitz, A. J., Elrick, K. A., and Hooper, R. P., Hydrological Processes, 1989, 3, 347. 22 Horowitz, A. J., Rinella, F. A., Lamothe, P., Miller, T. L., Edwards, T. K., Roche, R. L., and Rickert, D. A., Environ. Sci. Technol., 1990, 24, 1313. 23 Walling, D. E., and Moorhead, P. W., Geograf. Ann., 1987, 69A, 47. 24 Mossa, J., in Sediment and the Environment, ed. Hadley, R. F., and Ongley, E. D., IAHS Publication No. 184, IAHS Press, Wallingford, 1989, pp. 105–112. 25 de Groot, A. J., and Allersma, E., in Heavy Metals in the Aquatic Environment, ed. Krenkel, P. A., Pergamon Press, Oxford, 1975, pp. 85–95. 26 Chapman, P. M., Romberg, G. P., and Vigers, G. A., J. Water Pollut. Control Fed., 1982, 54, 292. 27 Walling, D. E., Webb, B. W., and Woodward, J. C., in Erosion and Sediment Transport Monitoring Programmes in River Basins, ed. Bogen, J., Walling, D. E., and Day, T. J., IAHS Publication No. 210, IAHS Press, Wallingford, 1992, pp. 279–288. 28 Alexander, R., Ludtke, A., Fitzgerlad, K., and Schertz, T., Data from Selected US Geological Survey National Stream Water-Quality Monitoring Networks (WQN) on CD-ROM, US Geological Survey Open-File Report 96-337 US Geological Survey, Reston, VA, 1996. Paper 7/04604I Received July 1, 1997 Accepted September 17, 1997 1200 Analyst, November 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a704604i
出版商:RSC
年代:1997
数据来源: RSC
|
4. |
Comparability and Traceability in Analytical Measurements and Reference Materials† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1201-1206
Michael Thompson,
Preview
|
|
摘要:
Comparability and Traceability in Analytical Measurements and Reference Materials† Michael Thompson Department of Chemistry, Birkbeck College, University of London, Gordon House, 29 Gordon Square, London, UK, WC1H 0PP The present paper argues that, in many sectors of analytical chemistry, measurements are not and need not be traceable to the kilogramme or the mole. Instead comparability among rational methods is assured by a traceability to the pure analyte. This leads to a system that gives the appropriate metrological level to reference materials.The argument is addressed to geochemists but is of general applicability. Keywords: Traceability; metrological level; rational method; empirical method Definitions Traceability: property of the result of a measurement or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons all having stated uncertainties. 1 Uncertainty (of measurement): parameter, associated with the result of a measurement, that characterises the dispersion of the values that could reasonably be attributed to the measurand. 1 Measurand: particular quantity subject to measurement.1 Reference material (RM): material or substance one or more of whose property values are sufficiently homogeneous and well established to be used for the calibration of an apparatus, the assessment of a measurement method, or for assigning values to materials.1 Certified reference material (CRM): reference material, accompanied by a certificate, one or more of whose property values are certified by a procedure which established traceability to an accurate realisation of the unit in which the property values are expressed, and for which each certified value is accompanied by an uncertainty at a stated level of confidence. 1 Introduction Comparability is now recognised as a prerequisite of analytical measurements made on geological materials.In the past some geochemists have argued2 that a body of data for an isolated used need only be internally self-consistent. For example, data with a consistent bias might be acceptable in a mineral exploration programme, because the eventual outcome would be simply dichotomised onto a nominal scale—either ‘dig here’ or ‘do not dig here’. Such an attitude is no longer acceptable. We now see that data generated for one purpose may be eventually turned to a use originally undreamed of.An example: the Wolfson Geochemical Atlas3 was conceived as a prospecting tool but in the event, was used almost exclusively for environmental purposes. Even in primary reconnaissance we need comparability to make use fully of our previous experience in interpreting geochemical data sets. The attempt to produce a comprehensive world-wide compilation of geochemical exploration data has highlighted problems of incomparability.4 Despite the use in geoanalysis of measurement methods based on diverse physical principles the results obtained often in practice seem to be comparable.More exactly, there is no clear evidence of incomparability. For example, results for sodium obtained by participants in Round 2 of the proficiency test GeoPT (Fig. 1) provide an almost symmetric unimodal distribution. Indeed, one outlier aside, the results are consistent with being a random sample from a normal distribution, as judged by the Kolmogorov-Smirnov one-sample statistic at the 95% confidence level.These sodium results are clearly ‘comparable’ in common parlance because the variations among them are consistent with the accumulation of numerous small random errors as required by the Central Limit Theorem. But there is another sense of ‘comparability’ that is more difficult to address. Are the results comparable in the sense that they are all made with respect to a common reference? This question is non-trivial because (a) it has financial implications for geoanalysts and, in consequence, their customers, and (b) there is controversy about the answer.The financial implication stems from the investment in quality control measures undertaken in analytical laboratories. The controversy surrounds the traceability of an analytical result. To be comparable in this more philosophical sense, two measurements must be cognate in respect of traceability. To understand this controversy, we must first examine the basic notions of metrology, namely the SI system of the measurement of physical quantities and their uncertainties.† Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Fig. 1 Cumulative distribution of results for Na2O obtained by participants in Round 2 of the proficiency test GeoPT. Analyst, November 1997, Vol. 122 (1201–1205) 1201Traceability in the SI In the SI, measurements can be traced back to the base unit, which is selected arbitrarily and may or may not be realised in a physical object.The base unit of mass is a physical object, a piece of platinum with a defined mass of 1 kg. How that relates to an everyday measurement of mass is shown in Fig. 2. The international standard is used to calibrate the national standards kept in the national laboratory. This comparison has an estimated uncertainty of u1. Standard weights issued with a certificate for calibration purposes are compared at the national laboratory with the national standard, with the introduction of a further uncertainty u2 which brings the total uncertainty to Au21 + u22 .Such a certificated weight might be used to calibrate a balance in a laboratory, with a further uncertainty u3. Finally the balance is used to weigh a portion of a material, say a metal ingot, with the introduction of uncertainty u4. The mass of the ingot is thus traceable to the International Kilogramme with a combined uncertainty of Au21 + u22 + u23 + u24 .The purpose of the traceability, the unbroken chain of comparisons is to ensure that (a) all measurements of mass are comparable (they all have the same reference point) and (b) the uncertainty of a measurement can be correctly estimated. The position of each mass in the chain of comparisons can be called its ‘metrological level’, with the international standard at the highest, and the individual measurement at the lowest, point of the scale.As each successively lower metrological level has a successively higher uncertainty associated with it, metrological level can also be based on a ratio scale (a continuous scale with a true zero point) by using the uncertainty value. The SI unit of ‘amount of substance’ is the mole, which is an Avogadro number of molecules, atoms, ions etc. The ‘mole’ is therefore the trivial name of a number, like the ‘dozen’, but of course much larger and with a nonzero uncertainty.The validity of the mole as an SI base unit is disputed by some metrologists6 while others, at the opposite extreme, contend that all chemical measurements should be made traceable to the mole because it is the natural unit for them.729 An even stronger inference sometimes drawn is that chemical measurements are not comparable unless they are traceable to the mole. One can see that measurements of concentration in units such as mol dm23 must be traceable to the mole (and the metre) but what of chemical measurements in general? The great majority of measurements made in geochemical analysis (and in many other sectors of analysis) are produced as mass fractions and expressed for example as % m/m, ppm etc.The mole is not involved in the measurement process: if we want to re-express the result in mol kg21 we do it by manipulation with relative atomic masses. The primary result is not traceable to the mole.By a homologous argument we see that chemical measurements determined as mass fractions are not traceable to the kilogramme either, despite the fact that they are sometime specified in quasi-units such mg kg21. If we determine the concentration of gold in a sample of rock in ppm the result would be numerically invariant if the kilogramme were differently defined or if an alternative base unit such as the pound or the troy ounce were used for weighing operations during the assay.Contrast that situation with the measurement previously discussed, the mass of a metal ingot. That mass is traceable to the (arbitrarily defined) kilogramme and the numerical value of the mass would be different if it were recorded in pounds. Therefore a consideration of whether the numerical value of a measurement is changed by using an alternative base unit (or other reference point) is a touchstone of traceability. As a high proportion of geochemical results are not traceable to the mole or the kilogramme, does that imply that they are not comparable? For example, if atomic absorption and XRF measurements of the same analyte in the same material are not both traceable to the mole, do we have no logical basis for expecting them to be identical within the limits of uncertainty? This is apparently the belief of some metrologists. The present paper argues that, in contrast, there is a common and valid scheme of traceability for concentration measurements produced as a mass fraction, but not traceability to the SI.Before that argument is developed, we must consider the essential structure of the typical process of making an analytical measurement. Types of Analytical Measurement System For present purposes I want to use traceability to distinguish between three types of analytical method, which I will call, the absolute, the rational and the empirical. Briefly for the moment, absolute methods are traceable to the mole, rational methods to the pure analyte, and empirical methods (where the analyte is specified by the procedure) are traceable to the method.Most methods in use seem to be of the rational kind, which is therefore most familiar. The rational method is illustrated schematically in Fig. 3. The concentration of the analyte is estimated by comparing (usually by means of a calibration curve) the responses produced in a selective physical system (e.g., an instrument) by the test material and effectively matrix-matched calibrators containing known concentrations of the analyte.The response of the Fig. 2 Chain of comparisons from the SI unit of mass (the kilogramme) to a weighing made in a laboratory, showing uncertainties (ui) and metrological levels. Fig. 3 Schematic diagram of a ‘rational method’. 1202 Analyst, November 1997, Vol. 122physical system could be, for example, changes in the intensity of a light beam of a particular wavelength, changes in an electrical potential, or in any other physical process that can be made into a surrogate for mass.In principle, all rational methods provide an unbiased estimate of the true value and, on that basis, must be comparable. The reality is that many methods fall short of complete rationality if, for example, the chemical pretreatment of the test material is incomplete or the matrix matching is imperfect. A system of traceability for rational and near-rational methods can readily be devised.5 Fig. 4 shows such a traceability to the pure analyte that exactly fulfils the VIM1 definition, namely, an unbroken chain of comparisons, each with an estimable uncertainty, to a stated reference. Incomplete dissolution and matrix effects simply contribute to the uncertainty added at the stage of instrumental comparison of the test material with the calibrators. Hence the results of all rational methods applied to the same material and analyte are comparable in that they have a common origin of traceability, the pure analyte.This is an important conclusion, because it provides the justification for using a consensus of the results of many methods and laboratories as the certified value of a reference material, or a material for internal quality control or proficiency testing. In addition it shows that traceability to the kilogramme or the mole is not the sine qua non for comparability. Direct traceability to the mole could, in many circumstances, be a useful attribute of an analytical result, but in others may be irrelevant.Mass ratios are more important to end users of analytical data than mole ratios (or mixed units such as mol kg21) in the majority of situations involving decisions based on analytical data. Moreover, converting a primary dimensionless result in (say) ppm to mol kg21 by dividing by a molecular weight adds uncertainty (albeit relatively very small) to that already present in the result.That would give the converted result a slightly lower metrological level. Traceability of Certified Reference Materials By definition a reference material may be used to calibrate or validate an analytical method, but it is also clear that reference materials, especially matrix reference materials, are certified by a procedure that is based on analysis. At first sight that seems to be a ‘chicken-and-egg’ situation (Fig. 5). If reference values depend on analysis, how can they stand at a higher metrological level than analysis and thus qualify for use in calibrating or validating methods? This question can be answered by considering the relevant metrological levels in terms of the ratio scale of the uncertainty associated with an analytical result or certified value.In the hypothetical example to be discussed a particular situation is considered: the certification of an RM from the replicated results assembled from many laboratories and obtained by the use of many distinct rational analytical methods.First we consider the extremes of the scale of metrological level. At the highest level is the conceptual pure analyte (i.e., with a zero uncertainty associated with its assay of exactly 100% m/m), and at the lowest level an result produced in a particular laboratory by a specified method. This individual result is the sum of: (a) the true value; (b) the bias associated with the method; (c) the bias associated with the laboratory; (d) the bias introduced into the results produced in the particular run of the method; and (e) the random (within-run) measurement error.We are concerned with producing a ‘consensus’ from a collection of such results, as the certified value of the concentration of the analyte in a candidate reference material. Next we have to consider all of the results produced in one particular laboratory by a specified method (Figs. 6 and 7). If we formed a mean of n such results, the common method bias, laboratory bias, or run bias in the individual results would be transferred unchanged to the mean, but the standard error of the mean would be smaller than the standard deviation associated with any particular result by a factor of 1/An.That would ensure that the mean would have a smaller uncertainty associated with it and therefore stand at a higher meteorological level than an individual result. (In practice we would want to use a robust mean to downweight the influence on the statistics of any outliers resulting from gross errors.The standard error of a Fig. 4 A scheme of traceability in a rational method from the pure analyte to the result of a measurement. Fig. 5 Is analytical measurement self-referential or traceable? Fig. 6 Individual analytical results (dots) comprise the sum of the true value plus all of the systematic and random errors. Analyst, November 1997, Vol. 122 1203robust mean could be estimated, for instance, by the bootstrap).Now consider the means from a subset of laboratories that use a particular method. All of the means have a common method bias, but each has an additional distinct laboratory bias associated with it that stems from the slightly different conditions in each laboratory (e.g., different instruments and reagents, and slightly different interpretations of the method protocol). If we calculate the mean of these laboratory means, the method bias will be propagated unchanged to this new method mean.However, the standard error of the method mean will be smaller than the standard deviation of the laboratory means and hence the uncertainty of the method mean will be smaller than the uncertainty of a laboratory mean. The method mean will stand at a higher metrological level. (As before, we would want to use robust statistics to downweight the influence of any outlying laboratories. We might also want an inversevariance weighted mean if the uncertainties on the individual means varied substantially among the laboratories.Again, the standard error of such a weighted robust mean could be estimated by the bootstrap.) We now have a series of mean values, one for each analytical method represented. Each value has associated with it a unique method bias (unknown) and a standard error. How are these values combined to form a consensus? Again, a robust weighted mean would be appropriate, given a sufficiently large number of distinct methods.In forming the consensus value we have to assume that the systemic errors of the individual methods (i.e., the method biases) will tend to a weighted average of zero. We are forced to this assumption because we have no independent way of checking it. (If there were an independent result, we would want to include it in the weighted mean along with the other values.) The alternative assumption is that all of the methods have a common nonzero bias.That seems to be a philosophical deadend, even if it were true. The uncertainty associated with the consensus would be smaller than that of the method means, so the consensus would stand at an even higher metrological level. We now have all that is required of a CRM, namely, a traceable assigned value with an uncertainty that gives it a metrological level lower than the stated reference (the pure analyte), but higher than methods or laboratories (Fig. 7). This status endows the CRM with the power to be used for testing the performance of methods or laboratories, or for calibrating methods, as required in its definition. The hypothetical scheme described above might need adaptation for real applications, but is not compromised thereby. For example, in practice there is likely to be only a few independent analytical methods available, so the robust mean of the method means would have to be a simple statistic such as the median.Alternatively, it might be appropriate to obtain a ‘smart mean’ by adjusting the weight associated with each mean to represent the known strengths and weaknesses of the analytical methods. For example, the mean from a method known to interference-prone might be downweighted. Some certifying authorities prefer to use results only from their own laboratory where special facilities are available, or to use only one method because it is considered to be definitive. In such instances we simply leave out the corresponding level of the argument, and assume that either the laboratory bias or method bias is absent from the analytical system.Traceability of Reference Materials in Empirical Methods Analytical chemists sometimes say that because empirical results are traceable only to the method they are necessarily true (using ‘true’ in the technical sense of free from bias). That is incorrect. It is quite possible for different analysts or laboratories to vary in their execution of an analytical protocol and therefore produce biased results. We know that happens in fact (and not just in theory) by considering the results of collaborative trials (interlaboratory performance studies) of empirical methods.It is usually found that a statistically significant laboratory effect occurs. Perhaps after all the method is not at the highest metrological level. We can explore that possibility by means of a simplified version (Fig. 8) of the argument in the previous section.As before, we start with many laboratories, each producing replicate results on a single test material. However, as there is only one method, the last stage of the previous argument is superfluous. There is no method bias by definition, although there can still be bias in the execution in individual laboratories. Therefore the consensus is simply the weighted robust mean of the robust means of the results from each laboratory. This provides an empirical CRM with an assigned value and an uncertainty giving it a higher metrological level than the method.The CRM can be used for calibration (e.g., of a faster Fig. 7 Metrological levels in a scheme to use the ‘consensus’ as a certified value for a reference material, when the results originate in a number of laboratories using a variety of rational analytical methods. The consensus has a higher metrological level than laboratories or methods, so the reference material can be used to test the performance of either.Fig. 8 Metrological levels associated with reference materials for empirical methods. 1204 Analyst, November 1997, Vol. 122new instrumental method) or in individual laboratories for checking the correct execution of a method. Conclusions In summary, we have characterised most analytical methods currently in use as rational (traceable to the pure analyte) or empirical (traceable to the method). Fully rational methods provide the measurand (i.e., the true value) within the bounds of precision uncertainty, regardless of the nature of the physical process acting as a surrogate for mass.As a consequence, the results of all rational measurement methods are comparable. It is therefore valid to estimate an assigned value for a material as a ‘consensus’, for example a ‘smart’ or robust mean of rational results from a variety of sources and methods. The uncertainty of the consensus can also be estimated from the data. These factors in combination satisfy the definition of a CRM.Many metrologists will disagree with these conclusions on the grounds that traceability to the pure analyte is inferior to traceability to the SI. A reconciliation of the two views would be effected if the pure analyte (as a substance) could in some sense be regarded as a realisation of the mole. References 1 International Vocabulary of Basic and General Terms in Metrology (VIM), International Organisation for Standardisation, Geneva, 2nd ed., 1993. 2 Rose, A. W., Hawkes, H. E., and Webb, J. S., Geochemistry in Mineral Exploration, Academic Press, London, 2nd ed., 1979, p. 65. 3 Webb, J. S., Thornton, I., Thompson, M., Howarth, R. J., and Lowenstein, P., The Wolfson Geochemical Atlas of England and Wales, Oxford University Press, 1978. 4 International Geological Correlation Programme, A Global Geochemical Database for Environmental and Resource Management, UNESCO, Paris, 1995. 5 Thompson, M., Analyst, 1996, 121, 258. 6 Alexandrov, J. I., Analyst, 1996, 121, 1137. 7 King, B., Analyst, 1997, 122, 197. 8 De Bi`evre, P., Kaarls, R., Peiser, H. S., Rasberry, S. D., and Reed, W. P., Accred. Qual. Assur., 1966, 1, 3. 9 De Bi`evre, P., and Taylor, P. D. P., Metrologia, 1997, 34, 67. Paper 7/04906D Received July 9, 1997 Accepted October 13, 1997 Analyst, November 1997, Vol. 122 1205 Comparability and Traceability in Analytical Measurements and Reference Materials† Michael Thompson Department of Chemistry, Birkbeck College, University of London, Gordon House, 29 Gordon Square, London, UK, WC1H 0PP The present paper argues that, in many sectors of analytical chemistry, measurements are not and need not be traceable to the kilogramme or the mole.Instead comparability among rational methods is assured by a traceability to the pure analyte. This leads to a system that gives the appropriate metrological level to reference materials.The argument is addressed to geochemists but is of general applicability. Keywords: Traceability; metrological level; rational method; empirical method Definitions Traceability: property of the result of a measurement or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons all having stated uncertainties. 1 Uncertainty (of measurement): parameter, associated with the result of a measurement, that characterises the dispersion of the values that could reasonably be attributed to the measurand. 1 Measurand: particular quantity subject to measurement.1 Reference material (RM): material or substance one or more of whose property values are sufficiently homogeneous and well established to be used for the calibration of an apparatus, the assessment of a measurement method, or for assigning values to materials.1 Certified reference material (CRM): reference material, accompanied by a certificate, one or more of whose property values are certified by a procedure which established traceability to an accurate realisation of the unit in which the property values are expressed, and for which each certified value is accompanied by an uncertainty at a stated level of confidence. 1 Introduction Comparability is now recognised as a prerequisite of analytical measurements made on geological materials. In the past some geochemists have argued2 that a body of data for an isolated used need only be internally self-consistent.For example, data with a consistent bias might be acceptable in a mineral exploration programme, because the eventual outcome would be simply dichotomised onto a nominal scale—either ‘dig here’ or ‘do not dig here’. Such an attitude is no longer acceptable. We now see that data generated for one purpose may be eventually turned to a use originally undreamed of. An example: the Wolfson Geochemical Atlas3 was conceived as a prospecting tool but in the event, was used almost exclusively for environmental purposes.Even in primary reconnaissance we need comparability to make use fully of our previous experience in interpreting geochemical data sets. The attempt to produce a comprehensive world-wide compilation of geochemical exploration data has highlighted problems of incomparability.4 Despite the use in geoanalysis of measurement methods based on diverse physical principles the results obtained often in practice seem to be comparable.More exactly, there is no clear evidence of incomparability. For example, results for sodium obtained by participants in Round 2 of the proficiency test GeoPT (Fig. 1) provide an almost symmetric unimodal distribution. Indeed, one outlier aside, the results are consistent with being a random sample from a normal distribution, as judged by the Kolmogorov-Smirnov one-sample statistic at the 95% confidence level.These sodium results are clearly ‘comparable’ in common parlance because the variations among them are consistent with the accumulation of numerous small random errors as required by the Central Limit Theorem. But there is another sense of ‘comparability’ that is more difficult to address. Are the results comparable in the sense that they are all made with respect to a common reference? This question is non-trivial because (a) it has financial implications for geoanalysts and, in consequence, their customers, and (b) there is controversy about the answer.The financial implication stems from the investment in quality control measures undertaken in analytical laboratories. The controversy surrounds the traceability of an analytical result. To be comparable in this more philosophical sense, two measurements must be cognate in respect of traceability. To understand this controversy, we must first examine the basic notions of metrology, namely the SI system of the measurement of physical quantities and their uncertainties. † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Fig. 1 Cumulative distribution of results for Na2O obtained by participants in Round 2 of the proficiency test GeoPT. Analyst, November 1997, Vol. 122 (1201–1205) 1201Traceability in the SI In the SI, measurements can be traced back to the base unit, which is selected arbitrarily and may or may not be realised in a physical object.The base unit of mass is a physical object, a piece of platinum with a defined mass of 1 kg. How that relates to an everyday measurement of mass is shown in Fig. 2. The international standard is used to calibrate the national standards kept in the national laboratory. This comparison has an estimated uncertainty of u1. Standard weights issued with a certificate for calibration purposes are compared at the national laboratory with the national standard, with the introduction of a further uncertainty u2 which brings the total uncertainty to Au21 + u22 .Such a certificated weight might be used to calibrate a balance in a laboratory, with a further uncertainty u3. Finally the balance is used to weigh a portion of a material, say a metal ingot, with the introduction of uncertainty u4. The mass of the ingot is thus traceable to the International Kilogramme with a combined uncertainty of Au21 + u22 + u23 + u24 .The purpose of the traceability, the unbroken chain of comparisons is to ensure that (a) all measurements of mass are comparable (they all have the same reference point) and (b) the uncertainty of a measurement can be correctly estimated. The position of each mass in the chain of comparisons can be called its ‘metrological level’, with the international standard at the highest, and the individual measurement at the lowest, point of the scale.As each successively lower metrological level has a successively higher uncertainty associated with it, metrological level can also be based on a ratio scale (a continuous scale with a true zero point) by using the uncertainty value. The SI unit of ‘amount of substance’ is the mole, which is an Avogadro number of molecules, atoms, ions etc. The ‘mole’ is therefore the trivial name of a number, like the ‘dozen’, but of course much larger and with a nonzero uncertainty.The validity of the mole as an SI base unit is disputed by some metrologists6 while others, at the opposite extreme, contend that all chemical measurements should be made traceable to the mole because it is the natural unit for them.729 An even stronger inference sometimes drawn is that chemical measurements are not comparable unless they are traceable to the mole. One can see that measurements of concentration in units such as mol dm23 must be traceable to the mole (and the metre) but what of chemical measurements in general? The great majority of measurements made in geochemical analysis (and in many other sectors of analysis) are produced as mass fractions and expressed for example as % m/m, ppm etc. The mole is not involved in the measurement process: if we want to re-express the result in mol kg21 we do it by manipulation with relative atomic masses. The primary result is not traceable to the mole.By a homologous argument we see that chemical measurements determined as mass fractions are not traceable to the kilogramme either, despite the fact that they are sometime specified in quasi-units such mg kg21.If we determine the concentration of gold in a sample of rock in ppm the result would be numerically invariant if the kilogramme were differently defined or if an alternative base unit such as the pound or the troy ounce were used for weighing operations during the assay. Contrast that situation with the measurement previously discussed, the mass of a metal ingot. That mass is traceable to the (arbitrarily defined) kilogramme and the numerical value of the mass would be different if it were recorded in pounds.Therefore a consideration of whether the numerical value of a measurement is changed by using an alternative base unit (or other reference point) is a touchstone of traceability. As a high proportion of geochemical results are not traceable to the mole or the kilogramme, does that imply that they are not comparable? For example, if atomic absorption and XRF measurements of the same analyte in the same material are not both traceable to the mole, do we have no logical basis for expecting them to be identical within the limits of uncertainty? This is apparently the belief of some metrologists.The present paper argues that, in contrast, there is a common and valid scheme of traceability for concentration measurements produced as a mass fraction, but not traceability to the SI.Before that argument is developed, we must consider the essential structure of the typical process of making an analytical measurement. Types of Analytical Measurement System For present purposes I want to use traceability to distinguish between three types of analytical method, which I will call, the absolute, the rational and the empirical. Briefly for the moment, absolute methods are traceable to the mole, rational methods to the pure analyte, and empirical methods (where the analyte is specified by the procedure) are traceable to the method.Most methods in use seem to be of the rational kind, which is therefore most familiar. The rational method is illustrated schematically in Fig. 3. The concentration of the analyte is estimated by comparing (usually by means of a calibration curve) the responses produced in a selective physical system (e.g., an instrument) by the test material and effectively matrix-matched calibrators containing known concentrations of the analyte.The response of the Fig. 2 Chain of comparisons from the SI unit of mass (the kilogramme) to a weighing made in a laboratory, showing uncertainties (ui) and metrological levels. Fig. 3 Schematic diagram of a ‘rational method’. 1202 Analyst, November 1997, Vol. 122physical system could be, for example, changes in the intensity of a light beam of a particular wavelength, changes in an electrical potential, or in any other physical process that can be made into a surrogate for mass. In principle, all rational methods provide an unbiased estimate of the true value and, on that basis, must be comparable.The reality is that many methods fall short of complete rationality if, for example, the chemical pretreatment of the test material is incomplete or the matrix matching is imperfect. A system of traceability for rational and near-rational methods can readily be devised.5 Fig. 4 shows such a traceability to the pure analyte that exactly fulfils the VIM1 definition, namely, an unbroken chain of comparisons, each with an estimable uncertainty, to a stated reference.Incomplete dissolution and matrix effects simply contribute to the uncertainty added at the stage of instrumental comparison of the test material with the calibrators. Hence the results of all rational methods applied to the same material and analyte are comparable in that they have a common origin of traceability, the pure analyte.This is an important conclusion, because it provides the justification for using a consensus of the results of many methods and laboratories as the certified value of a reference material, or a material for internal quality control or proficiency testing. In addition it shows that traceability to the kilogramme or the mole is not the sine qua non for comparability. Direct traceability to the mole could, in many circumstances, be a useful attribute of an analytical result, but in others may be irrelevant.Mass ratios are more important to end users of analytical data than mole ratios (or mixed units such as mol kg21) in the majority of situations involving decisions based on analytical data. Moreover, converting a primary dimensionless result in (say) ppm to mol kg21 by dividing by a molecular weight adds uncertainty (albeit relatively very small) to that already present in the result. That would give the converted result a slightly lower metrological level.Traceability of Certified Reference Materials By definition a reference material may be used to calibrate or validate an analytical method, but it is also clear that reference materials, especially matrix reference materials, are certified by a procedure that is based on analysis. At first sight that seems to be a ‘chicken-and-egg’ situation (Fig. 5). If reference values depend on analysis, how can they stand at a higher metrological level than analysis and thus qualify for use in calibrating or validating methods? This question can be answered by considering the relevant metrological levels in terms of the ratio scale of the uncertainty associated with an analytical result or certified value.In the hypothetical example to be discussed a particular situation is considered: the certification of an RM from the replicated results assembled from many laboratories and obtained by the use of many distinct rational analytical methods.First we consider the extremes of the scale of metrological level. At the highest level is the conceptual pure analyte (i.e., with a zero uncertainty associated with its assay of exactly 100% m/m), and at the lowest level an result produced in a particular laboratory by a specified method. This individual result is the sum of: (a) the true value; (b) the bias associated with the method; (c) the bias associated with the laboratory; (d) the bias introduced into the results produced in the particular run of the method; and (e) the random (within-run) measurement error.We are concerned with producing a ‘consensus’ from a collection of such results, as the certified value of the concentration of the analyte in a candidate reference material. Next we have to consider all of the results produced in one particular laboratory by a specified method (Figs. 6 and 7). If we formed a mean of n such results, the common method bias, laboratory bias, or run bias in the individual results would be transferred unchanged to the mean, but the standard error of the mean would be smaller than the standard deviation associated with any particular result by a factor of 1/An.That would ensure that the mean would have a smaller uncertainty associated with it and therefore stand at a higher meteorological level than an individual result. (In practice we would want to use a robust mean to downweight the influence on the statistics of any outliers resulting from gross errors.The standard error of a Fig. 4 A scheme of traceability in a rational method from the pure analyte to the result of a measurement. Fig. 5 Is analytical measurement self-referential or traceable? Fig. 6 Individual analytical results (dots) comprise the sum of the true value plus all of the systematic and random errors. Analyst, November 1997, Vol. 122 1203robust mean could be estimated, for instance, by the bootstrap).Now consider the means from a subset of laboratories that use a particular method. All of the means have a common method bias, but each has an additional distinct laboratory bias associated with it that stems from the slightly different conditions in each laboratory (e.g., different instruments and reagents, and slightly different interpretations of the method protocol). If we calculate the mean of these laboratory means, the method bias will be propagated unchanged to this new method mean.However, the standard error of the method mean will be smaller than the standard deviation of the laboratory means and hence the uncertainty of the method mean will be smaller than the uncertainty of a laboratory mean. The method mean will stand at a higher metrological level. (As before, we would want to use robust statistics to downweight the influence of any outlying laboratories. We might also want an inversevariance weighted mean if the uncertainties on the individual means varied substantially among the laboratories.Again, the standard error of such a weighted robust mean could be estimated by the bootstrap.) We now have a series of mean values, one for each analytical method represented. Each value has associated with it a unique method bias (unknown) and a standard error. How are these values combined to form a consensus? Again, a robust weighted mean would be appropriate, given a sufficiently large number of distinct methods.In forming the consensus value we have to assume that the systemic errors of the individual methods (i.e., the method biases) will tend to a weighted average of zero. We are forced to this assumption because we have no independent way of checking it. (If there were an independent result, we would want to include it in the weighted mean along with the other values.) The alternative assumption is that all of the methods have a common nonzero bias. That seems to be a philosophical deadend, even if it were true.The uncertainty associated with the consensus would be smaller than that of the method means, so the consensus would stand at an even higher metrological level. We now have all that is required of a CRM, namely, a traceable assigned value with an uncertainty that gives it a metrological level lower than the stated reference (the pure analyte), but higher than methods or laboratories (Fig. 7). This status endows the CRM with the power to be used for testing the performance of methods or laboratories, or for calibrating methods, as required in its definition.The hypothetical scheme described above might need adaptation for real applications, but is not compromised thereby. For example, in practice there is likely to be only a few independent analytical methods available, so the robust mean of the method means would have to be a simple statistic such as the median. Alternatively, it might be appropriate to obtain a ‘smart mean’ by adjusting the weight associated with each mean to represent the known strengths and weaknesses of the analytical methods.For example, the mean from a method known to interference-prone might be downweighted. Some certifying authorities prefer to use results only from their own laboratory where special facilities are available, or to use only one method because it is considered to be definitive. In such instances we simply leave out the corresponding level of the argument, and assume that either the laboratory bias or method bias is absent from the analytical system.Traceability of Reference Materials in Empirical Methods Analytical chemists sometimes say that because empirical results are traceable only to the method they are necessarily true (using ‘true’ in the technical sense of free from bias). That is incorrect. It is quite possible for different analysts or laboratories to vary in their execution of an analytical protocol and therefore produce biased results.We know that happens in fact (and not just in theory) by considering the results of collaborative trials (interlaboratory performance studies) of empirical methods. It is usually found that a statistically significant laboratory effect occurs. Perhaps after all the method is not at the highest metrological level. We can explore that possibility by means of a simplified version (Fig. 8) of the argument in the previous section. As before, we start with many laboratories, each producing replicate results on a single test material.However, as there is only one method, the last stage of the previous argument is superfluous. There is no method bias by definition, although there can still be bias in the execution in individual laboratories. Therefore the consensus is simply the weighted robust mean of the robust means of the results from each laboratory. This provides an empirical CRM with an assigned value and an uncertainty giving it a higher metrological level than the method.The CRM can be used for calibration (e.g., of a faster Fig. 7 Metrological levels in a scheme to use the ‘consensus’ as a certified value for a reference material, when the results originate in a number of laboratories using a variety of rational analytical methods. The consensus has a higher metrological level than laboratories or methods, so the reference material can be used to test the performance of either. Fig. 8 Metrological levels associated with reference materials for empirical methods. 1204 Analyst, November 1997, Vol. 122new instrumental method) or in individual laboratories for checking the correct execution of a method. Conclusions In summary, we have characterised most analytical methods currently in use as rational (traceable to the pure analyte) or empirical (traceable to the method). Fully rational methods provide the measurand (i.e., the true value) within the bounds of precision uncertainty, regardless of the nature of the physical process acting as a surrogate for mass. As a consequence, the results of all rational measurement methods are comparable. It is therefore valid to estimate an assigned value for a material as a ‘consensus’, for example a ‘smart’ or robust mean of rational results from a variety of sources and methods. The uncertainty of the consensus can also be estimated from the data. These factors in combination satisfy the definition of a CRM. Many metrologists will disagree with these conclusions on the grounds that traceability to the pure analyte is inferior to traceability to the SI. A reconciliation of the two views would be effected if the pure analyte (as a substance) could in some sense be regarded as a realisation of the mole. References 1 International Vocabulary of Basic and General Terms in Metrology (VIM), International Organisation for Standardisation, Geneva, 2nd ed., 1993. 2 Rose, A. W., Hawkes, H. E., and Webb, J. S., Geochemistry in Mineral Exploration, Academic Press, London, 2nd ed., 1979, p. 65. 3 Webb, J. S., Thornton, I., Thompson, M., Howarth, R. J., and Lowenstein, P., The Wolfson Geochemical Atlas of England and Wales, Oxford University Press, 1978. 4 International Geological Correlation Programme, A Global Geochemical Database for Environmental and Resource Management, UNESCO, Paris, 1995. 5 Thompson, M., Analyst, 1996, 121, 258. 6 Alexandrov, J. I., Analyst, 1996, 121, 1137. 7 King, B., Analyst, 1997, 122, 197. 8 De Bi`evre, P., Kaarls, R., Peiser, H. S., Rasberry, S. D., and Reed, W. P., Accred. Qual. Assur., 1966, 1, 3. 9 De Bi`evre, P., and Taylor, P. D. P., Metrologia, 1997, 34, 67. Paper 7/04906D Received July 9, 1997 Accepted October 13, 1997 Analyst, November 1997, Vol. 122 1205
ISSN:0003-2654
DOI:10.1039/a704906d
出版商:RSC
年代:1997
数据来源: RSC
|
5. |
Determining Cadmium in Marine Sediments by Inductively Coupled Plasma Mass Spectrometry: Attacking the Problems or the Problems With the Attack?† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1207-1210
Jennifer M. Cook,
Preview
|
|
摘要:
Determining Cadmium in Marine Sediments by Inductively Coupled Plasma Mass Spectrometry: Attacking the Problems or the Problems With the Attack?† Jennifer M. Cook*, Jon J. Robinson, Simon R. N. Chenery and Douglas L. Miles Analytical and Regional Geochemistry Group, British Geological Survey, Keyworth, Nottingham, UK NG12 5GG The importance of interelement interference of one trace element on another is highlighted in the determination of low levels of Cd in marine sediments by ICP-MS, both during an interlaboratory proficiency test and in the analysis of certified reference materials.Initial values obtained for the CRMs MESS-1 and BCSS-1 after an HF-based digestion were up to 50% higher than the certificate values, although the ratio of the Cd isotopes at m/z 111 and 114 agreed with that for the natural abundance of Cd. Investigations revealed that a polyatomic interference on 111Cd, arising from the relatively large concentrations of Zr in the digests, was very similar in size to the isobaric interference of 114Sn on 114Cd in these materials.Appropriate corrections derived from the analysis of single element interferent solutions gave concentrations indistinguishable from the certificate value, with detection limits of 0.043 and 0.060 mg kg21 for 111Cd and 114Cd, respectively. The study highlights the problem of obtaining consistent datasets between laboratories using different sample preparation schemes for environmentally important trace elements close to the detection limit.Keywords: Inductively coupled plasma mass spectrometry; interferences; marine sediments; cadmium; hydrofluoric acid digestion; sample preparation Inductively coupled plasma mass spectrometry (ICP-MS)1 provides a powerful means of simultaneously determining trace amounts of many elements in complex matrices. Like all such sensitive analytical techniques it is subject to a number of interferences that need to be understood and controlled if reliable data are to be obtained.Over the last ten years, many of the fundamental principles of interference mechanisms in ICPMS have been unravelled2 and many specific interferences are now well known and can be relatively easily quantified in particular matrices. Apart from mutual mass spectral interference caused by direct isobaric overlap between elements present at similar concentrations, e.g., the interference of 92Zr on 92Mo, it is generally true that most analytically significant interferences are derived from the effect of high concentrations of matrix elements on analytes present at concentrations many orders of magnitude smaller.Examples include: the general suppression of sensitivity for light analyte elements caused by a heavy matrix; and the interference by polyatomic ions of matrix elements on trace analytes at the same nominal m/z, e.g., 40Ar 35Cl+ on 75As, 40Ar16O+ on 56Fe and ArOH+ on 57Fe. Nowadays, in any one applications area, an experienced analyst will tend to know what interferences are likely to be encountered in ICP-MS and be alert to their potential influences on data quality.Effective monitoring of the marine environment provides some of the greatest challenges to many modern analytical methods. Partly because of the scale involved, it often relies on a number of dispersed laboratories producing comparable data. Moreover, because of the variety and complexity of sample preparation schemes that marine materials such as biota and sediments are typically subjected to prior to analysis, the potential for real environmental trends being masked by analytical discrepancies between different datasets is considerable.This is particularly true for analytes determined close to their detection limits. In the UK, the National Marine Analytical Quality Control (NMAQC) Scheme was established in 1991 with the aim of demonstrating, via a series of proficiency tests, that laboratories submitting data to the UK National Marine Monitoring Plan were achieving the necessary standards of analytical accuracy.In addition, special exercises are conducted from time to time to help laboratories improve their accuracy. One such recent special exercise on the determination of metals in marine sediments was designed to examine the comparability of the different dissolution schemes in current use.3 Participating laboratories were supplied with five subsamples of each of two marine sediments, which they digested according to their normal procedures.Most of the participating laboratories employed a HNO3–HCl digestion of some description, and five laboratories also contributed solutions from an attack using HF. All the resultant digests were then returned, in bottles which had been supplied for the purpose, and analysed at a single laboratory to minimise the measurement variance. Details of the various digestion procedures used were returned with the digested sediments.At the central laboratory the sediment digests were diluted 1 + 4 to reduce matrix effects and Cd, Cr, Cu, Pb, Ni and Zn were determined by ICP-MS. Comparison of the data indicated a high degree of agreement between the HNO3–HCl digests for all these metals, in spite of the methodological variations. The conclusion was that the digestion method was unlikely to be an important source of error in any poor comparability between data produced by different laboratories using a HNO3–HCl attack.With the stronger HF attack, there was a positive bias in the data for Cd and Cr for both sediments, compared to the mean of the HNO3–HCl digests. While this was understandable for Cr which can be locked up in minerals resistant to HF, it was surprising for Cd, because this element is assumed to be associated with easily dissolved minerals such as clays. As well as participating in the special exercise, at about the same time we had been analysing a large number of marine sediment samples for trace elements, using the CRMs MESS-1 and BCSS-1 (National Research Council of Canada) as part of our quality control procedures.The data for the CRMs were acceptable for all elements determined except Cd, where values up to 50% higher than those certified (Table 1) were regularly † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Analyst, November 1997, Vol. 122 (1207–1210) 1207observed. These apparently high Cd values were consistent over thirty independent digestions of the CRMs and showed good agreement and precision between both of the isotopes monitored, 111Cd and 114Cd. This concordance was initially taken to indicate a lack of mass spectral interference. The study reported here was therefore undertaken to shed light on the cause of the apparently high Cd values obtained following HF digestion of marine sediments in both our own work and the NMAQC special exercise.Experimental Reagents Nitric acid and perchloric acid (Primar grade, Fisons, Loughborough, UK). Hydrofluoric acid (Analytical Reagent grade, Fisons). Standard tin and molybdenum solutions, 1000 mg l21. (AAS standard solutions, Johnson Matthey, Royston, UK.) Standard zirconium solution, 10 000 mg l21. (ICP standard solution, Johnson Matthey.) Calibration and monitor solutions for ICP-MS.(Claritas PPT, Spex Chemical, Metuchen, New Jersey, USA.) Marine sediment certified reference materials BCSS-1 and MESS-1. (National Research Council Canada.) All dilutions were prepared using deionised water of 18 MW quality (Millipore, Watford, UK). Sample Preparation The method employed for the dissolution of marine sediments at the BGS was the one routinely used for rendering solid geological materials into solution. The grain size of samples was reduced to < 50 mm by grinding in agate before 0.1 g was accurately weighed into a PTFE test-tube. A 1 ml volume of HF was added and the tube swirled gently to ensure wetting before being capped and left overnight.The following day, 0.8 ml of nitric acid and 0.4 ml of perchloric acid were added. The tubes were placed in a programmable heating block and heated for 5 h at 100 °C, followed by 1 h at 140 °C and 6 h at 190 °C. When the heating program had finished, the test tubes were removed from the block and allowed to cool.A 1 ml volume of 1 + 1 HNO3 was added to each tube and the tubes warmed gently at 50 °C for approximately 30 min. Then 9 ml of deionised water were added and the tubes capped and shaken vigorously. The tubes were left overnight before being centrifuged for 10 min at 2500 rpm. The solutions were decanted into clean, disposable, stoppered plastic tubes for storage prior to analysis. Certified reference materials MESS-1 and BCSS-1 and methodological blanks were prepared alongside the samples in each batch, so that 10% of the tubes contained blanks and 10% CRMs.Immediately prior to analysis the samples were diluted a further ten-fold (1 + 9) by the addition of an internal standard solution containing In and Bi in 1% v/v HNO3, to give 10 mg l21 of both elements in the final solution. Instrumentation The sediment digests were analysed by ICP-MS using a VG Plasmaquad 2+ spectrometer (VG Elemental, Winsford, Cheshire, UK) operated under the conditions given in Table 2.Spex multielement solutions were used to calibrate the instrument. In addition, Spex solutions containing the elements of interest at 10 mg l21 were inserted at regular intervals during the analytical run to monitor drift in sensitivity with time. Results and Discussion Following the discovery of apparently high values for Cd in the HF digests of both the CRMs and the NMAQC sediments, the first action taken was to send three aliquots of the digested CRMs to another laboratory which routinely determined trace metals in marine sediments. This laboratory had participated in the NMAQC special exercise and had contributed data on HF digested samples.They also analysed the BGS-digested CRMs by ICP-MS and their data are summarised in Table 3. Again there was good agreement between the Cd isotopes monitored but a large positive bias from the certified values for both CRMs. To ensure that our batches of CRMs had not been contaminated in any way, this laboratory was also asked to digest two sub-samples of our CRM powders using their normal HF-based method and determine Cd by ICP-MS. Their results were 0.30 and 0.72 mg kg21 for BCSS-1 and MESS-1 respectively, i.e., again a positive bias from the reference values.4 Meanwhile, another large batch of marine sediments and CRMs had been prepared at the BGS and the digests dispatched to yet another laboratory, who also routinely analysed marine sediments for trace metals, but who used ETAAS for the determination of Cd. Their data on the CRM solutions are summarised in Table 4 and generally show reasonable agreement with the certified values.From this evidence it was clear that the positive bias in the ICP-MS data was the result of one or more interferences. Cadmium has several small isotopes. This limits its sensitivity by ICP-MS compared to other elements. Although 114Cd is the Table 1 Initial BGS values for the Cd content of marine sediment reference materials BCSS-1 and MESS-1 measured at m/z 111 and 114.(Values in mg kg21; n, number of replicate digestions analysed with one standard deviation in brackets.) The uncertainties in the certified values represent the 95% confidence limits for an individual sample BCSS-1 MESS-1 Cd-111 (n = 30) 0.40 (0.07) 0.86 (0.05) Cd-114 (n = 30) 0.37 (0.07) 0.86 (0.06) Certified value 0.25 (0.04) 0.59 (0.10) Table 2 ICP-MS operating conditions ICP mass spectrometer VG Plasmaquad 2+ Rf power 1350 W Plasma gas flow rate 13 l min21 Auxiliary flow rate 0.80 l min21 Injector gas flow rate 0.95 l min21 Nebuliser De Galan V groove Sample uptake rate 0.8 ml min21 Autosampler Gilson 222 Measurement mode Peak jumping Number of replicate measurements 3 Acquisition time per measurement 60 s Sample uptake delay 120 s Rinsing time 180 s Table 3 Cadmium content of BGS-digested CRMs determined by an external laboratory using ICP-MS.(Values in mg kg21; n, number of replicate digestions analysed; and the uncertainties in the certified values represent the 95% confidence limits for an individual sample) BCSS-1 MESS-1 Cd-111 (n = 3) 0.42 0.95 Cd-114 (n = 3) 0.37 0.94 Certified value 0.25 (0.04) 0.59 (0.10) 1208 Analyst, November 1997, Vol. 122most abundant isotope at 28.8%, it suffers from isobaric overlap with 114Sn (0.65%), as does 112Cd with 112Sn (1%) and 116Cd with 116Sn (14.4%).Consequently 111Cd is often used when the presence of Sn in samples may be significant. Oxides and hydroxides are a common source of polyatomic interferences in ICP-MS and there are various combinations of Zr and Mo oxides and hydroxides which are isobaric with one or more of the Cd isotopes (Table 5).5 BCSS-1 and MESS-1 are not certified for Zr and Mo. The digests were analysed for these elements and Sn; Zr was also determined by XRF. The values obtained by ICP-MS for Sn of 2 and 4 mg kg21 agree well with the certified concentrations of 1.85 ± 0.2 and 3.98 ± 0.44 mg kg21 for BCSS-1 and MESS-1, respectively.In contrast, the much lower concentrations of Zr measured by ICP-MS of 127 and 178 mg kg21 for BCSS-1 and MESS-1, respectively, compared with those obtained by XRF (224 and 539 mg kg21), indicate that our normal HF-based attack was insufficiently vigorous to liberate all the Zr from resistant mineral phases. This incomplete dissolution means that it is likely that the concentration of Zr obtained from an HFbased digest will vary with the experimental conditions used in different laboratories.Separate solutions of Zr, Mo and Sn were analysed to assess the contribution from each element or its polyatomic species at the m/z values of the Cd isotopes. This revealed that the contribution from 94ZrOH at m/z 111 was of the order of 0.1% of the Zr concentration. However, because the measured Zr : Cd ratio in the CRM digests is greater than 300 : 1, the signal from the Zr polyatomic ion was significant in relation to the low concentrations of Cd, in spite of the fact that it is a three body ion and 94Zr is only 17.5% abundant.This is in accord with the observations of Wu et al.6 in their analysis of a fly ash CRM. In contrast, the Mo polyatomic ions contribute a signal at m/z 111 equivalent to approximately 0.2% of the Mo content. However, because of the Mo : Cd ratio in the CRMs is less than 10 : 1, the interference on Cd from Mo is almost negligible for these materials. The interference on 111Cd arising from the larger concentrations of Zr was not immediately obvious because the enhancement was similar in magnitude to that of 114Sn on 114Cd and thus the apparent ratio of the two Cd isotopes remained relatively constant.This is demonstrated in Table 6 which shows the successive reduction in signal at masses 111 and 114, following sequential correction for Zr, Mo and Sn polyatomic and isobaric interferences.When appropriate corrections were applied for Zr, Mo and Sn to the signals from the CRM digests at m/z 111 and 114, the Cd concentrations derived from both isotopes were indistinguishable from the certificate values and showed good precision for trace element data fairly close to the detection limits of 0.043 and 0.060 mg kg21 for 111Cd and 114Cd, respectively (Table 7). Conclusions Normally in ICP-MS, analytically significant polyatomic interferences on trace elements result from the presence of major elements.However, the present study is an example of the influence of one trace element on another. Most laboratories analysing marine sediments do not routinely determine Zr or Mo and would therefore be unaware of their relative concentra- Table 4 Cadmium content of BGS-digested CRMs determined by an external laboratory using ETAAS. (Values in mg kg21; n, number of replicate digestions analysed with one standard deviation in brackets.) The uncertainties in the certified values represent the 95% confidence limits for an individual sample BCSS-1 MESS-1 Cd (n = 34) 0.21 (0.06) 0.47 (0.10) Certified value 0.25 (0.04) 0.59 (0.10) Table 5 Spectral interferences on selected Cd isotopes Abundance Molecular Abundance Abundance Mass of Cd ion of interfering Isobaric of Sn isotope (m/z) isotope (%) interference element (%) interference (%) 110 12.5 94ZrO 17.5 94MoO 9.1 111 12.8 94ZrOH 17.5 94MoOH 9.1 95MoO 15.9 112 24.0 96MoO 16.7 Sn 0.97 96ZrO 2.8 95MoOH 15.9 113 12.3 96MoOH 16.7 97MoO 9.5 96ZrOH 2.8 114 28.8 98MoO 24.1 Sn 0.65 97MoOH 9.4 Table 6 Magnitude of interferences from Zr, Mo and Sn at m/z 111 and 114 in digests of BCSS-1 and MESS-1 determined by ICP-MS.All signals are expressed as a percentage of the uncorrected signal Zr, Mo + Mass No Zr Zr + Mo Sn CRM m/z correction corrected corrected corrected BCSS-1 111 100 87 82 82 114 100 100 100 84 MESS-1 111 100 73 72 72 114 100 100 100 80 Table 7 BGS values for the Cd content of marine sediment reference materials BCSS-1 and MESS-1 measured at m/z 111 and 114 after interference correction.(Values in mg kg21; n, number of replicate digestions analysed, with one standard deviation in brackets). Uncertainties in the certified values represent the 95% confidence limits for an individual sample. BCSS-1 MESS-1 Cd-111 (n = 12) 0.27 (0.02) 0.66 (0.03) Cd-114 (n = 12) 0.25 (0.02) 0.64 (0.02) Certified value 0.25 (0.04) 0.59 (0.10) Analyst, November 1997, Vol. 122 1209tions. In cases where interferences are suspected, the signal from more than one isotope is usually monitored but, as demonstrated here, this procedure is not foolproof. Because of the number of Cd isotopes and possible interfering polyatomic species, together with the low concentrations of Cd in the sample, examination of the isotopic fingerprint of Cd did not by itself suggest the cause of the bias in the Cd determinations.It is probable that many laboratories using ICP-MS to determine Cd in marine sediments have observed a slight positive bias in their data for these CRMs, although the values may well have been considered to be within acceptable limits for such trace levels. Certainly, a similar slight positive bias has been reported in the determination of Cd in coal fly ash (Wu et al.6). However, the magnitude of this bias will vary depending on the severity of the sample attack employed.Moreover, at higher concentrations of Cd, such as those encountered in sediments from polluted waters, the interference from Zr would be much less significant and likely to go unnoticed. Data obtained from aqua regia digests are unlikely to reveal any statistically significant deviations from the certified value because this reagent will not digest resistant Zr-containing minerals such as zircon. This study demonstrates that proficiency testing schemes may not, by themselves, highlight specific problems with a method of analysis.This is especially true when the participating laboratories use a variety of methods of sample preparation, e.g., aqua regia and HF. The authors acknowledge the valuable discussions and data provided by B. Jones (CEFAS, Burnham-on-Crouch). This paper is published with the permission of the Director, British Geological Survey (Natural Environment Research Council). References 1 Houk, R.S., Fassel, V. A., Flesch, G. D., Svec, H. J., Gray, A. L., and Taylor, C. E., Anal. Chem., 1980, 52, 2283. 2 Evans, E. H., and Giglio, J. J., J. Anal. At. Spectrom., 1993, 8, 1. 3 Cook, J. M., Gardner, M. J., Griffiths, A. H., Jessep, M. A., Ravenscroft, J. E., and Yates, R., Mar. Pollut. Bull., 1997, in the press. 4 Jones, B., Personal communication. 5 Venth, K., Danzer, K., Kundermann, G., and Blaufuß, K.-H., Fresenius J. Anal. Chem., 1996, 354, 811. 6 Wu, S., Zhao, Y.-H., Feng, X., and Wittmeier,A., J.Anal. At. Spectrom., 1996, 11, 287. Paper 7/05973F Received August 14, 1997 Accepted October 10, 1997 1210 Analyst, November 1997, Vol. 122 Determining Cadmium in Marine Sediments by Inductively Coupled Plasma Mass Spectrometry: Attacking the Problems or the Problems With the Attack?† Jennifer M. Cook*, Jon J. Robinson, Simon R. N. Chenery and Douglas L. Miles Analytical and Regional Geochemistry Group, British Geological Survey, Keyworth, Nottingham, UK NG12 5GG The importance of interelement interference of one trace element on another is highlighted in the determination of low levels of Cd in marine sediments by ICP-MS, both during an interlaboratory proficiency test and in the analysis of certified reference materials.Initial values obtained for the CRMs MESS-1 and BCSS-1 after an HF-based digestion were up to 50% higher than the certificate values, although the ratio of the Cd isotopes at m/z 111 and 114 agreed with that for the natural abundance of Cd.Investigations revealed that a polyatomic interference on 111Cd, arising from the relatively large concentrations of Zr in the digests, was very similar in size to the isobaric interference of 114Sn on 114Cd in these materials. Appropriate corrections derived from the analysis of single element interferent solutions gave concentrations indistinguishable from the certificate value, with detection limits of 0.043 and 0.060 mg kg21 for 111Cd and 114Cd, respectively.The study highlights the problem of obtaining consistent datasets between laboratories using different sample preparation schemes for environmentally important trace elements close to the detection limit. Keywords: Inductively coupled plasma mass spectrometry; interferences; marine sediments; cadmium; hydrofluoric acid digestion; sample preparation Inductively coupled plasma mass spectrometry (ICP-MS)1 provides a powerful means of simultaneously determining trace amounts of many elements in complex matrices.Like all such sensitive analytical techniques it is subject to a number of interferences that need to be understood and controlled if reliable data are to be obtained. Over the last ten years, many of the fundamental principles of interference mechanisms in ICPMS have been unravelled2 and many specific interferences are now well known and can be relatively easily quantified in particular matrices.Apart from mutual mass spectral interference caused by direct isobaric overlap between elements present at similar concentrations, e.g., the interference of 92Zr on 92Mo, it is generally true that most analytically significant interferences are derived from the effect of high concentrations of matrix elements on analytes present at concentrations many orders of magnitude smaller. Examples include: the general suppression of sensitivity for light analyte elements caused by a heavy matrix; and the interference by polyatomic ions of matrix elements on trace analytes at the same nominal m/z, e.g., 40Ar 35Cl+ on 75As, 40Ar16O+ on 56Fe and ArOH+ on 57Fe.Nowadays, in any one applications area, an experienced analyst will tend to know what interferences are likely to be encountered in ICP-MS and be alert to their potential influences on data quality. Effective monitoring of the marine environment provides some of the greatest challenges to many modern analytical methods.Partly because of the scale involved, it often relies on a number of dispersed laboratories producing comparable data. Moreover, because of the variety and complexity of sample preparation schemes that marine materials such as biota and sediments are typically subjected to prior to analysis, the potential for real environmental trends being masked by analytical discrepancies between different datasets is considerable. This is particularly true for analytes determined close to their detection limits.In the UK, the National Marine Analytical Quality Control (NMAQC) Scheme was established in 1991 with the aim of demonstrating, via a series of proficiency tests, that laboratories submitting data to the UK National Marine Monitoring Plan were achieving the necessary standards of analytical accuracy. In addition, special exercises are conducted from time to time to help laboratories improve their accuracy. One such recent special exercise on the determination of metals in marine sediments was designed to examine the comparability of the different dissolution schemes in current use.3 Participating laboratories were supplied with five subsamples of each of two marine sediments, which they digested according to their normal procedures.Most of the participating laboratories employed a HNO3–HCl digestion of some description, and five laboratories also contributed solutions from an attack using HF. All the resultant digests were then returned, in bottles which had been supplied for the purpose, and analysed at a single laboratory to minimise the measurement variance.Details of the various digestion procedures used were returned with the digested sediments. At the central laboratory the sediment digests were diluted 1 + 4 to reduce matrix effects and Cd, Cr, Cu, Pb, Ni and Zn were determined by ICP-MS. Comparison of the data indicated a high degree of agreement between the HNO3–HCl digests for all these metals, in spite of the methodological variations.The conclusion was that the digestion method was unlikely to be an important source of error in any poor comparability between data produced by different laboratories using a HNO3–HCl attack. With the stronger HF attack, there was a positive bias in the data for Cd and Cr for both sediments, compared to the mean of the HNO3–HCl digests. While this was understandable for Cr which can be locked up in minerals resistant to HF, it was surprising for Cd, because this element is assumed to be associated with easily dissolved minerals such as clays.As well as participating in the special exercise, at about the same time we had been analysing a large number of marine sediment samples for trace elements, using the CRMs MESS-1 and BCSS-1 (National Research Council of Canada) as part of our quality control procedures. The data for the CRMs were acceptable for all elements determined except Cd, where values up to 50% higher than those certified (Table 1) were regularly † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Analyst, November 1997, Vol. 122 (1207–1210) 1207observed. These apparently high Cd values were consistent over thirty independent digestions of the CRMs and showed good agreement and precision between both of the isotopes monitored, 111Cd and 114Cd.This concordance was initially taken to indicate a lack of mass spectral interference. The study reported here was therefore undertaken to shed light on the cause of the apparently high Cd values obtained following HF digestion of marine sediments in both our own work and the NMAQC special exercise. Experimental Reagents Nitric acid and perchloric acid (Primar grade, Fisons, Loughborough, UK). Hydrofluoric acid (Analytical Reagent grade, Fisons). Standard tin and molybdenum solutions, 1000 mg l21.(AAS standard solutions, Johnson Matthey, Royston, UK.) Standard zirconium solution, 10 000 mg l21. (ICP standard solution, Johnson Matthey.) Calibration and monitor solutions for ICP-MS. (Claritas PPT, Spex Chemical, Metuchen, New Jersey, USA.) Marine sediment certified reference materials BCSS-1 and MESS-1. (National Research Council Canada.) All dilutions were prepared using deionised water of 18 MW quality (Millipore, Watford, UK). Sample Preparation The method employed for the dissolution of marine sediments at the BGS was the one routinely used for rendering solid geological materials into solution.The grain size of samples was reduced to < 50 mm by grinding in agate before 0.1 g was accurately weighed into a PTFE test-tube. A 1 ml volume of HF was added and the tube swirled gently to ensure wetting before being capped and left overnight. The following day, 0.8 ml of nitric acid and 0.4 ml of perchloric acid were added.The tubes were placed in a programmable heating block and heated for 5 h at 100 °C, followed by 1 h at 140 °C and 6 h at 190 °C. When the heating program had finished, the test tubes were removed from the block and allowed to cool. A 1 ml volume of 1 + 1 HNO3 was added to each tube and the tubes warmed gently at 50 °C for approximately 30 min. Then 9 ml of deionised water were added and the tubes capped and shaken vigorously. The tubes were left overnight before being centrifuged for 10 min at 2500 rpm.The solutions were decanted into clean, disposable, stoppered plastic tubes for storage prior to analysis. Certified reference materials MESS-1 and BCSS-1 and methodological blanks were prepared alongside the samples in each batch, so that 10% of the tubes contained blanks and 10% CRMs. Immediately prior to analysis the samples were diluted a further ten-fold (1 + 9) by the addition of an internal standard solution containing In and Bi in 1% v/v HNO3, to give 10 mg l21 of both elements in the final solution.Instrumentation The sediment digests were analysed by ICP-MS using a VG Plasmaquad 2+ spectrometer (VG Elemental, Winsford, Cheshire, UK) operated under the conditions given in Table 2. Spex multielement solutions were used to calibrate the instrument. In addition, Spex solutions containing the elements of interest at 10 mg l21 were inserted at regular intervals during the analytical run to monitor drift in sensitivity with time.Results and Discussion Following the discovery of apparently high values for Cd in the HF digests of both the CRMs and the NMAQC sediments, the first action taken was to send three aliquots of the digested CRMs to another laboratory which routinely determined trace metals in marine sediments. This laboratory had participated in the NMAQC special exercise and had contributed data on HF digested samples. They also analysed the BGS-digested CRMs by ICP-MS and their data are summarised in Table 3.Again there was good agreement between the Cd isotopes monitored but a large positive bias from the certified values for both CRMs. To ensure that our batches of CRMs had not been contaminated in any way, this laboratory was also asked to digest two sub-samples of our CRM powders using their normal HF-based method and determine Cd by ICP-MS. Their results were 0.30 and 0.72 mg kg21 for BCSS-1 and MESS-1 respectively, i.e., again a positive bias from the reference values.4 Meanwhile, another large batch of marine sediments and CRMs had been prepared at the BGS and the digests dispatched to yet another laboratory, who also routinely analysed marine sediments for trace metals, but who used ETAAS for the determination of Cd.Their data on the CRM solutions are summarised in Table 4 and generally show reasonable agreement with the certified values. From this evidence it was clear that the positive bias in the ICP-MS data was the result of one or more interferences.Cadmium has several small isotopes. This limits its sensitivity by ICP-MS compared to other elements. Although 114Cd is the Table 1 Initial BGS values for the Cd content of marine sediment reference materials BCSS-1 and MESS-1 measured at m/z 111 and 114. (Values in mg kg21; n, number of replicate digestions analysed with one standard deviation in brackets.) The uncertainties in the certified values represent the 95% confidence limits for an individual sample BCSS-1 MESS-1 Cd-111 (n = 30) 0.40 (0.07) 0.86 (0.05) Cd-114 (n = 30) 0.37 (0.07) 0.86 (0.06) Certified value 0.25 (0.04) 0.59 (0.10) Table 2 ICP-MS operating conditions ICP mass spectrometer VG Plasmaquad 2+ Rf power 1350 W Plasma gas flow rate 13 l min21 Auxiliary flow rate 0.80 l min21 Injector gas flow rate 0.95 l min21 Nebuliser De Galan V groove Sample uptake rate 0.8 ml min21 Autosampler Gilson 222 Measurement mode Peak jumping Number of replicate measurements 3 Acquisition time per measurement 60 s Sample uptake delay 120 s Rinsing time 180 s Table 3 Cadmium content of BGS-digested CRMs determined by an external laboratory using ICP-MS.(Values in mg kg21; n, number of replicate digestions analysed; and the uncertainties in the certified values represent the 95% confidence limits for an individual sample) BCSS-1 MESS-1 Cd-111 (n = 3) 0.42 0.95 Cd-114 (n = 3) 0.37 0.94 Certified value 0.25 (0.04) 0.59 (0.10) 1208 Analyst, November 1997, Vol. 122most abundant isotope at 28.8%, it suffers from isobaric overlap with 114Sn (0.65%), as does 112Cd with 112Sn (1%) and 116Cd with 116Sn (14.4%). Consequently 111Cd is often used when the presence of Sn in samples may be significant. Oxides and hydroxides are a common source of polyatomic interferences in ICP-MS and there are various combinations of Zr and Mo oxides and hydroxides which are isobaric with one or more of the Cd isotopes (Table 5).5 BCSS-1 and MESS-1 are not certified for Zr and Mo.The digests were analysed for these elements and Sn; Zr was also determined by XRF. The values obtained by ICP-MS for Sn of 2 and 4 mg kg21 agree well with the certified concentrations of 1.85 ± 0.2 and 3.98 ± 0.44 mg kg21 for BCSS-1 and MESS-1, respectively. In contrast, the much lower concentrations of Zr measured by ICP-MS of 127 and 178 mg kg21 for BCSS-1 and MESS-1, respectively, compared with those obtained by XRF (224 and 539 mg kg21), indicate that our normal HF-based attack was insufficiently vigorous to liberate all the Zr from resistant mineral phases.This incomplete dissolution means that it is likely that the concentration of Zr obtained from an HFbased digest will vary with the experimental conditions used in different laboratories. Separate solutions of Zr, Mo and Sn were analysed to assess the contribution from each element or its polyatomic species at the m/z values of the Cd isotopes. This revealed that the contribution from 94ZrOH at m/z 111 was of the order of 0.1% of the Zr concentration.However, because the measured Zr : Cd ratio in the CRM digests is greater than 300 : 1, the signal from the Zr polyatomic ion was significant in relation to the low concentrations of Cd, in spite of the fact that it is a three body ion and 94Zr is only 17.5% abundant. This is in accord with the observations of Wu et al.6 in their analysis of a fly ash CRM. In contrast, the Mo polyatomic ions contribute a signal at m/z 111 equivalent to approximately 0.2% of the Mo content.However, because of the Mo : Cd ratio in the CRMs is less than 10 : 1, the interference on Cd from Mo is almost negligible for these materials. The interference on 111Cd arising from the larger concentrations of Zr was not immediately obvious because the enhancement was similar in magnitude to that of 114Sn on 114Cd and thus the apparent ratio of the two Cd isotopes remained relatively constant.This is demonstrated in Table 6 which shows the successive reduction in signal at masses 111 and 114, following sequential correction for Zr, Mo and Sn polyatomic and isobaric interferences. When appropriate corrections were applied for Zr, Mo and Sn to the signals from the CRM digests at m/z 111 and 114, the Cd concentrations derived from both isotopes were indistinguishable from the certificate values and showed good precision for trace element data fairly close to the detection limits of 0.043 and 0.060 mg kg21 for 111Cd and 114Cd, respectively (Table 7).Conclusions Normally in ICP-MS, analytically significant polyatomic interferences on trace elements result from the presence of major elements. However, the present study is an example of the influence of one trace element on another. Most laboratories analysing marine sediments do not routinely determine Zr or Mo and would therefore be unaware of their relative concentra- Table 4 Cadmium content of BGS-digested CRMs determined by an external laboratory using ETAAS.(Values in mg kg21; n, number of replicate digestions analysed with one standard deviation in brackets.) The uncertainties in the certified values represent the 95% confidence limits for an individual sample BCSS-1 MESS-1 Cd (n = 34) 0.21 (0.06) 0.47 (0.10) Certified value 0.25 (0.04) 0.59 (0.10) Table 5 Spectral interferences on selected Cd isotopes Abundance Molecular Abundance Abundance Mass of Cd ion of interfering Isobaric of Sn isotope (m/z) isotope (%) interference element (%) interference (%) 110 12.5 94ZrO 17.5 94MoO 9.1 111 12.8 94ZrOH 17.5 94MoOH 9.1 95MoO 15.9 112 24.0 96MoO 16.7 Sn 0.97 96ZrO 2.8 95MoOH 15.9 113 12.3 96MoOH 16.7 97MoO 9.5 96ZrOH 2.8 114 28.8 98MoO 24.1 Sn 0.65 97MoOH 9.4 Table 6 Magnitude of interferences from Zr, Mo and Sn at m/z 111 and 114 in digests of BCSS-1 and MESS-1 determined by ICP-MS.All signals are expressed as a percentage of the uncorrected signal Zr, Mo + Mass No Zr Zr + Mo Sn CRM m/z correction corrected corrected corrected BCSS-1 111 100 87 82 82 114 100 100 100 84 MESS-1 111 100 73 72 72 114 100 100 100 80 Table 7 BGS values for the Cd content of marine sediment reference materials BCSS-1 and MESS-1 measured at m/z 111 and 114 after interference correction.(Values in mg kg21; n, number of replicate digestions analysed, with one standard deviation in brackets).Uncertainties in the certified values represent the 95% confidence limits for an individual sample. BCSS-1 MESS-1 Cd-111 (n = 12) 0.27 (0.02) 0.66 (0.03) Cd-114 (n = 12) 0.25 (0.02) 0.64 (0.02) Certified value 0.25 (0.04) 0.59 (0.10) Analyst, November 1997, Vol. 122 1209tions. In cases where interferences are suspected, the signal from more than one isotope is usually monitored but, as demonstrated here, this procedure is not foolproof. Because of the number of Cd isotopes and possible interfering polyatomic species, together with the low concentrations of Cd in the sample, examination of the isotopic fingerprint of Cd did not by itself suggest the cause of the bias in the Cd determinations. It is probable that many laboratories using ICP-MS to determine Cd in marine sediments have observed a slight positive bias in their data for these CRMs, although the values may well have been considered to be within acceptable limits for such trace levels.Certainly, a similar slight positive bias has been reported in the determination of Cd in coal fly ash (Wu et al.6). However, the magnitude of this bias will vary depending on the severity of the sample attack employed. Moreover, at higher concentrations of Cd, such as those encountered in sediments from polluted waters, the interference from Zr would be much less significant and likely to go unnoticed. Data obtained from aqua regia digests are unlikely to reveal any statistically significant deviations from the certified value because this reagent will not digest resistant Zr-containing minerals such as zircon. This study demonstrates that proficiency testing schemes may not, by themselves, highlight specific problems with a method of analysis. This is especially true when the participating laboratories use a variety of methods of sample preparation, e.g., aqua regia and HF. The authors acknowledge the valuable discussions and data provided by B. Jones (CEFAS, Burnham-on-Crouch). This paper is published with the permission of the Director, British Geological Survey (Natural Environment Research Council). References 1 Houk, R. S., Fassel, V. A., Flesch, G. D., Svec, H. J., Gray, A. L., and Taylor, C. E., Anal. Chem., 1980, 52, 2283. 2 Evans, E. H., and Giglio, J. J., J. Anal. At. Spectrom., 1993, 8, 1. 3 Cook, J. M., Gardner, M. J., Griffiths, A. H., Jessep, M. A., Ravenscroft, J. E., and Yates, R., Mar. Pollut. Bull., 1997, in the press. 4 Jones, B., Personal communication. 5 Venth, K., Danzer, K., Kundermann, G., and Blaufuß, K.-H., Fresenius J. Anal. Chem., 1996, 354, 811. 6 Wu, S., Zhao, Y.-H., Feng, X., and Wittmeier,A., J. Anal. At. Spectrom., 1996, 11, 287. Paper 7/05973F Received August 14, 1997 Accepted October 10, 1997 1210 Analyst, November 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a705973f
出版商:RSC
年代:1997
数据来源: RSC
|
6. |
Measurement of Trace Element Distributions in Soils and Sediments Using Sequential Leach Data and a Non-specific Extraction System With Chemometric Data Processing† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1211-1221
Mark R. Cave,
Preview
|
|
摘要:
Measurement of Trace Element Distributions in Soils and Sediments Using Sequential Leach Data and a Non-specific Extraction System With Chemometric Data Processing† Mark R. Cave and Joanna Wragg Analytical and Regional Geochemistry Group, British Geological Survey, Keyworth, Nottingham, UK NG12 5GG A chemometric mixture resolution procedure suitable for determining the number and composition of physico-chemical components in data derived from soil leachates is described. The procedure is used to determine the number of components in sequential leachate data obtained for a NIST certified soil (SRM 2710) using a widely employed leaching scheme.The resulting data show that the sequential leaching media are not specific for their designated target fractions and that erroneous identification of fractions occurs. A scoping study in which a new non-specific extraction method is tested is described. The experimental design varies the concentration of nitric acid, the reaction time and the ratio of sample to extractant.The resulting solutions were analysed by ICP-AES for major and trace metals and the data obtained from 34 experiments subjected to the chemometric resolution procedure. Four components are identified and the effects of the three variables on each component are modelled using multiple linear regression, allowing the conditions which favour dissolution of each component to be identified. Calculated element compositions of the components identified in the non-specific extraction trial are compared with those identified in the sequential extraction data.Significant correlations between the two sets of components are noted and tentative identification of the source of the components is made. In particular, there is evidence that the Tessier method extracts both Fe and Mn oxides simultaneously, whereas the non-specific method has resolved the Fe and Mn oxides as separate entities. Keywords: Sequential leaching; chemometric mixture resolution; soils analysis; trace element partitioning; inductively coupled plasma The determination of potentially toxic inorganic substances (e.g., heavy metals) in soils and sediments is an important tool for monitoring environmental pollution.Although the total concentrations of these potentially toxic substances provides broad evidence for possible contamination, it has been recognised that quantification of the chemical forms of metals in soils is essential for estimating the mobility and bioavailability of the metals in the environment.1,2 Metals in soils may be present in several different geochemical phases that act as reservoirs or sinks of trace elements in the environment.3,4 The chemical phases considered to be important are divided up into a series of broad categories usually consisting of: exchangeable; specifically adsorbed; carbonate; Fe and Mn-oxides; organic matter; mineral lattice.To obtain information on the distributions of trace metals between these soil/sediment phases, a number of workers5–8 have developed extraction schemes in which phases are selectively dissolved with carefully chosen reagents.By subsequent chemical analysis of the extraction media the concentration of trace elements associated with the target phase can be determined. The method of Tessier et al.5 has been widely adopted in a number of applications.9–15 Despite the widespread use of these selective extraction methods and the insight they have given to understanding the geochemical processes governing trace metal distributions, selective extraction techniques have been demonstrated to have a number of weaknesses,16–20 the two most important being: 1 The so-called ‘selective extraction’ reagents are not specific for one mineral phase; therefore the associated analysis is not a true representation of the amount of trace elements from a single phase. 2 The design of the selective extraction schemes leads to a methodological definition of the distribution of trace elements between solid phases which may not reflect the actual distribution within the test samples. A recent study by Cave and Harmon21 investigated the trace elements associated with the iron oxide phase of red-bed sandstones and showed that chemometric processing of the iron oxide phase data from related samples could identify the presence of more than one component being mobilised by the so-called ‘selective extractive reagent’.This work confirmed the limitations noted earlier and also suggests an alternative approach to the study of speciation that should overcome some of the problems of the traditional methods. If the chemical composition of the products of individual extraction steps within a sequential extraction procedure are considered to be mixtures of the physico-chemical components of the soil or sediment, each containing different proportions of each component, then a similar procedure to that described by Cave and Harmon21 could be used to identify and quantify these components.Further, if this chemometric mixture resolution is viable for the sequential extraction data, a new extraction method could be applied in which a relatively simple nonspecific reagent is used to extract the different physico-chemical phases from the target soil or sediment. The resulting solution would be made up of a mixture of different proportions of each physico-chemical phase.By producing a series of these mixed phase solutions with different proportions of each phase present, chemometric methods could be used resolve the composition of each phase. The method used to produce the solutions containing the different proportions of each phase could be: (1) time series extractions, because different phases should dissolve at different rates; (2) variation of rock/ extractant ratio; (3) a series of different extractant concentrations.The main advantages of this approach would be the simplicity of the extraction procedures, as only one extractant † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1211–1221) 1211would be required, and the partitioning of trace metals between the different phases would not be methodologically defined. In this study, a mixture resolution approach to interpreting a traditional sequential extraction scheme was carried out and a feasibility study on the use of a simple single extraction approach combined with chemometric data processing is reported.Experimental Test Material To test the new methodologies, a well characterised material was required. At present there are very few, if any, SRMs certified for leaching purposes. Li et al.,15 however, have carried out a rigourous Tessier style5 sequential extraction Fig. 1 Relationship between the leachate concentration matrix, the matrix containing the physico-chemical component compositions and the matrix containing the proportions of each physico-chemical component in each leachate solution. Table 1 Central composite experimental design for the non-specific extraction trial and total solids extracted for each of the 4 identified components Levels (see below for actual values) Total extracted solids/mg kg21 Sample : Nitric acid extractant Component Component Component Component Replicate Time concentration ratio 1 2 3 4 1 21 21 21 8689 4485 7749 3105 1 21 1 1 7438 7279 10151 2830 1 1 21 1 6191 690 10839 3629 1 1 1 21 4704 16634 15746 3717 1 0 0 0 6653 4456 11153 3696 1 21 21 1 6495 854 6815 3911 1 21 1 21 7792 8601 9464 2842 1 1 21 21 7442 8359 13292 3200 1 1 1 1 4258 13686 15624 3656 1 0 0 0 6490 4442 10981 3765 1 21.67 0 0 7524 1113 4404 3712 1 1.67 0 0 5844 5277 13003 3859 1 0 21.67 0 9 0 138 3905 1 0 1.67 0 5448 13904 15187 3430 1 0 0 21.67 7851 10618 11185 3826 1 0 0 1.67 6242 191 10126 3851 1 0 0 0 6497 4333 10669 3633 2 21 21 21 8757 4446 7663 2999 2 21 1 1 7939 7331 10138 2896 2 1 21 1 6302 831 11248 3765 2 1 1 21 4432 16333 15328 3994 2 0 0 0 6690 4186 10862 3550 2 21 21 1 6747 838 6680 3768 2 21 1 21 8184 8363 9250 2624 2 1 21 21 6974 8326 13364 3314 2 1 1 1 4256 12983 15337 3670 2 0 0 0 6543 4233 10862 3538 2 21.67 0 0 7101 1030 4089 3487 2 1.67 0 0 5853 5226 13053 3852 2 0 21.67 0 2 0 138 3882 2 0 1.67 0 5754 12406 13979 2956 2 0 0 21.67 7982 10169 10921 3603 2 0 0 1.67 7498 0 11116 5013 2 0 0 0 6745 4242 10947 3493 Nitric Level Time acid/m Sample : extractant ratio 21.67 0.5 0.01 0.01 21 18.3 0.21 0.05 0 44.75 0.50 0.12 1 71.79 0.80 0.20 1.67 90.00 1.0 0.25 1212 Analyst, November 1997, Vol. 122procedure on an NIST certified soil (SRM 2710) in which they looked at extending the number of trace metals under study by modifying the method of Tessier et al.5 This particular SRM is highly contaminated soil from pasture land along Silver Bow Creek in the Butte, Montana, area.Li et al.15 noted that this sample was unusual in that it had a high level of heavy metals in the exchangeable fraction, which would make it a suitable material for studying the mobility and bioavailabilty of metals in contaminated soils. Because of the availability of data and the nature of the sample, the published data15 for this soil was used for the mixture resolution exercise and a sample of the soil was used for the non-specific extraction trial.Chemometric Mixture Resolution Procedure Chemometric strategies for mixture resolution have been used widely in analytical chemistry22,23 and have also been used for studying environmental data sets.21,24,25 The procedure developed here is a combination of the methods of Thurston and Spengler,25 Gamp et al.26 and Cave and Harmon.21 This method is based on the assumption that the sample (soil or sediment) is made up of a number of physico-chemical components each of which has its own chemical composition (e.g., carbonate component, iron oxide component).By leaching the sample, under certain conditions, a proportion of these components is leached into solution. The concentration of an element in a particular leach solution can be described as a linear sum of the amounts leached from each physico-chemical sources present, such that: E En n n n c tot = = =åa 1 (1) where Etot is the total concentration of an element E in a given leach solution, Ec the concentration of element E in component n, ac the proportion of element E leached from component n, and c the number of components.In this study there is more than one element being considered and a number of leaches have been carried out. In this instance, eqn. (1) can be expressed in matrix form which is shown pictorially in Fig. 1 showing the dimensions of each matrix with a description of the contents of each matrix. In order to be able to tell which elements are associated with which each physico-chemical component it is necessary to find matrices B and C given matrix A. The first stage in determining B and C is to use principal component analysis (PCA) of matrix A to estimate the number of physico-chemical components (c) present and to give a first estimate of the proportions of each component in each leach (i.e., matrix B).This was carried out in a similar manner to that proposed by Thurston and Spengler25 but required a number of modifications to make the method suitable to the leachate composition data. Firstly, PCA is normally carried out on a matrix with the compositional data (in this case elemental concentrations) in columns and the different samples in rows (as shown by matrix A in Fig. 1) using a pre-scaling procedure in which each column is scaled by subtracting the column mean and dividing by the column standard deviation. This ensures that all elements whether present at high or low concentration contribute equally to the PCA model.In the case of the leachate data, however, the total amount of material in each leachate can vary considerably (for example, in the Tessier method differences between the amount leached in the exchangeable and the residual fractions can vary by 1–2 orders of magnitude) and scaling down the element columns followed by PCA would have produced a model which was dominated by the leachate samples in which the largest amount of material was extracted.Table 2 Eigenvalues and percentage variance explained for the Varimax rotated principal components for the Tessier and non-specific extraction data Extraction method Tessier method Non-specific Principal component Variance Variance no. Eigenvalue (%) Eigenvalue (%) 1 1.5 30.6 16.3 47.9 2 1.2 24.9 5.1 15.1 3 1.1 21.2 3.8 11.2 4 1.0 20.6 6.9 20.3 5 0.13 2.7 0.14 0.41 6 — — 0.07 0.20 Table 3 Varimax rotated scores the Tessier and non-specific extraction data (numbers in bold indicate elements with the highest score within a given PC) Tessier method Non-specific method Principal component Principal component Element 1 2 3 4 Element 1 2 3 4 Al 0.15 3.26 20.67 20.49 Al 0.78 20.98 20.69 1.26 Ca 20.48 0.24 20.75 3.07 Ba 20.74 20.48 20.32 20.34 Cd 20.64 20.51 20.40 20.51 Ca 20.20 2.60 1.64 20.18 Co 20.64 20.51 20.39 20.53 Cd 20.79 20.48 20.31 20.29 Cu 1.67 20.53 20.34 20.67 Cu 20.46 20.77 1.63 1.34 Fe 20.31 1.05 2.53 20.79 Fe 3.03 20.52 20.37 21.78 K 0.19 0.74 20.57 0.63 K 20.17 0.87 0.09 20.95 Mn 20.01 20.53 2.20 1.07 Mg 20.30 0.07 20.46 20.33 Ni 20.64 20.51 20.39 20.53 Mn 1.06 0.82 21.07 2.56 P 20.58 20.48 20.34 20.54 Na 20.66 0.30 20.35 20.63 Pb 2.95 20.45 20.19 20.05 Ni 20.80 20.49 20.32 20.29 Sr 20.63 20.50 20.40 20.46 P 20.76 20.49 20.28 20.32 Ti 20.64 20.34 20.41 20.52 Pb 0.97 21.28 2.53 0.23 V 20.64 20.51 20.39 20.53 Si 0.07 0.10 20.88 20.18 Zn 0.24 20.39 0.52 0.83 Ti 20.77 20.49 20.33 20.31 V 20.79 20.48 20.32 20.30 Zn 0.51 1.68 20.18 0.49 Analyst, November 1997, Vol. 122 1213This could lead to an underestimate of the number of physicochemical components being leached. It was therefore decided to transpose matrix A (AT) and scale the matrix over each leachate, therefore making each leachate contribute equally to the PCA model. PCA of the AT matrix was carried out followed by Varimax27 rotation.According to Thurston and Spengler,25 this procedure provides abstract PC score and loadings matrices which should be closely related to the true proportion and component concentration matrices (matrices B and C, respectively, in Fig. 1). The number of physico-chemical components (c) present was found by the number of PCs with eigenvalues greater than one after Varimax rotation.25 The next stage is to use the abstract PCs derived from the PCA of AT to provide a first approximation of the proportions of matrix B.Thurston and Spengler25 used a procedure where they derived absolute PC scores which could be used as the first estimate. This procedure was not successful for the leachate data which gave proportions with negative values. This step was therefore replaced by that used by Cave and Harmon21 in which each of the significant (c) PCs within the Varimax rotated scores matrix was examined. The elements which have the highest scores within each of these PCs are assumed to be the most highly correlated to the rotated PCs and hence to the true physico-chemical components.The concentration of these elements in each leachate should be linearly related to the columns in matrix B. Regressing the concentrations of each of these identified elements against the total extractable solids for each leachate solution, using multiple linear regression (MLR), gives estimates of the coefficients which convert the element concentrations into physico-chemical component mass contributions (in mg kg21) for each leachate solution, allowing a first approximation of B to be calculated. In the final stage matrix A is scaled so that each element concentration is expressed as a fraction of the total extracted mass for each leach solution (AA).Similarly, the first approximation of B is scaled so that each mass contribution is expressed as a fraction of the total extracted mass for each leach solution (BA).Using the pseudoinverse calculation22 ( in other words an MLR regression of BA against AA) a first estimate of CA (the scaled physico-chemical component concentration matrix) can be calculated. The first estimate of CA is further refined by setting any negative values to 0 and by scaling each row so that the relative element contributions for each component adds up to 1. Using the pseudoinverse calculation with the refined CA and matrix AA a second approximation for BA is calculated.The second approximation for BA is refined by setting negative values to 0 and by scaling each column so that the relative mass contribution for each leachate adds up to 1. Further estimates of BA and CA are iteratively calculated using this procedure until no further improvement in their values is obtained. This follows the method first used by Gamp et al.26 and more recently by Cave and Harmon.21 Finally the refined matrices BA and CA are rescaled to absolute concentration values.The PCA, Varimax rotation and MLR calculations were carried out using Statistica (Version 5.1) and the iterative pseudo-inverse calculation routine was written in MathCad Plus (Version 6.0). Non-specific Extraction Procedure A central composite experimental design procedure was used to investigate the effects of three variables (time, nitric acid concentration and sample to extractant ratio) on the chemical composition of the resulting acid extracts from SRM 2710.Using this formal experimental design, the relative effects and interactions of each parameter could be measured and MLR modelling of the response surface was carried out. The five levels used for each variable and the experimental design are shown in Table 1. Each measurement was carried out Fig. 2 Relative proportions of the each resolved component in the sequential extraction data. 1214 Analyst, November 1997, Vol. 122in duplicate resulting in a total of 34 test solutions for analysis.The extractions were carried out in 30 ml polycarbonate screw top tubes which were mixed on an ‘end-over-end’ rotating shaker. All experiments were carried out in an air conditioned laboratory with the temperature nominally maintained at 20 °C. The experimental design matrix and the analysis of the results was carried out using the experimental design module of Statistica Version (5.1). Analysis of the Acid Leachate The 34 leachate solutions generated by the experimental design procedure were analysed by ICP-AES for Al, Ba, Ca, Cd, Cu, Fe, K, Mg, Mn, Na, Ni, P, Pb, Si, Ti, V and Zn.The instrumentation and wavelengths used have been previously described.21 Results and Discussion Sequential Leach Data In addition to the average concentration of each element extracted in the sequential extraction scheme described by Li et al.,15 uncertainty measurements for each determinand were also supplied. In order to include this information in the chemometric data processing, an additional nine sets of data were generated with random amounts of uncertainty, within the reported limits, and were added to the average values.The original data matrix of 15 elements (columns) and 5 leachates (rows) was combined with the additional nine data sets to produce a single matrix of 150 columns and 5 rows. This new matrix was then processed as a single data matrix using the mixture resolution procedure described in the experimental section.The eigenvalues for the Varimax rotated PCA model for the Tessier extraction data are shown in Table 2. This clearly shows that there are four components with eigenvalues greater than one and a step change in eigenvalue between the fourth and fifth PC. In addition, Table 3 shows the scores for the first four significant PCs identifying the elements Pb, Al, Fe and Ca as having the highest scores. Using these element concentration profiles and four components as being significant, the composition of each component was calculated using the chemometric procedure previously described.The relative contributions of each identified component for the ten sets of data were then recombined by taking an average. The standard deviation of the data was used to give a measure of the uncertainty for the relative proportion of any component in a given extraction step. Fig. 2 shows a plot of the calculated average relative proportions of each component in each of the 5 sequential leaches with error bars representing twice the standard deviation on each value (n = 10).Fig. 3 shows the chemical compositions of each of the four component represented as a pie diagram. Component 1 makes up a significant proportion of the first four extracts with particularly high levels in the designated carbonate and organic/sulfide extracts. This component is made up principally from the heavy metals Pb, Cu and Zn. The origin of this component is not clear but may be related to an organic rich component or possibly a fine clay fraction.The reasons for this will be discussed later. Component 2 only appears in the last two extracts and is more than 50% Al. This can be interpreted as the alumino-silicate matrix of the soil. Component 3 appears predominantly in the Fe/Mn oxide extracted fraction although the designated organic/sulfide and residual fractions contain approximately 10% of this component which is made up predominantly from Mn and Fe.This indicates an Fe/Mn oxide component. Component 4 appears predominantly in the designated exchangeable and carbonate fractions and is dominated by Ca, Mn, Zn, K and Pb. The high proportion of this component in the first fraction indicates this is the easily exchangeable fraction. Although components 2–4 appear predominantly in a single extract and are consistent with the designation of the extract in which they occur, significant proportions of each component ‘spill’ over into preceding or subsequent fractions.Component 1, however, is predominant in two designated fractions and does not clearly fit into either the carbonate or the organic/sulfide designations. If the data processing method used is valid, the problems of non-specificity of extraction reagents have been demonstrated. In addition, possible mis-identification of extracted fractions has also been shown. By multiplying the proportion of each component by its chemical composition and re-scaling to the total extracted solids for a given fraction, an analogous table to that previously produced15 can be calculated (Table 4).In this instance, however, it is not the methodologically defined fractions that are reported but the composition of each of the resolved Table 4 Resolved component compositions for NIST 2710 in mg g21 with ±2 s (n = 10) Component 1 2 3 4 S CTV Element Value Error Value Error Value Error Value Error Value Error Value Error Al 322 65.2 63 700 5600 608 227 0.00 0 64 600 5600 64 400 800 Ca 0.00 0 10 000 883 0.00 0 2510 1330 12 500 1600 12 500 300 Cd 1.89 0.38 0.00 0 3.06 1.14 14.8 7.86 19.8 7.95 21.8 0.2 Co 1.56 0.31 2.24 0.2 3.22 1.2 0.51 0.27 7.53 1.28 (10) Cu 2020 408 586 51.5 515 192 0.00 0 3120 454 2950 130 Fe 71.8 14.5 22 700 1990 10 600 3950 0.00 0 33 400 4420 33 800 1000 K 379 76.6 17 000 1494 0.00 0 893 473 18 300 1570 21 100 1100 Mn 751 152 0.00 0 9860 3680 1240 657 11 900 3740 10 100 400 Ni 0.99 0.2 8.65 0.76 0.13 0.05 0.08 0.04 9.85 0.79 14.3 1 P 65.5 13.2 791 69.6 227 84.8 0 0 1080 111 1060 150 Pb 3040 615 15.4 1.35 1210 452 523 277 4790 812 5532 80 Sr 6.76 1.37 263 23.1 7.87 2.94 48.6 25.8 326 34.8 240 Ti 0.00 0 2530 222 0.00 0 10.8 5.72 2540 222 2830 100 V 3.33 0.67 53.3 4.69 17.9 6.69 0.00 0 74.5 8.2 76.6 2.3 Zn 789 160 1055 92.7 3760 1410 1030 547 6630 1520 6952 91 * S represents the sum of the four components.† CTV is the certified total concentration.Analyst, November 1997, Vol. 122 1215components and their associated uncertainties that is given. This is a useful check to show that at the end of the processing procedure the total amount of each element extracted is still comparable with the certified values for the soil. Non-specific Extraction Trial The requirements for the non-specific extraction experiment were: (i) to produce a series of leachate solutions containing varying proportions of the different physico-chemical components of the soil, to allow the chemometric procedure to identify and quantify each component; and (ii) to carry out the experiments in such a way that analysis of data would allow the effects and interactions of the three variables (acid concentration, sample to extractant ratio and time) on the dissolution of the physico-chemical components of the soil to be studied.In order to achieve these objectives a formal experimental design was required.The central composite design was chosen as being more suitable than a two level design as it allows curvature of effects to be investigated and the results can be more readily visualised as 3 dimensional surface plots. The design outlined in Table 1 was chosen from a menu of designs given within the Statistica software package. The total extracted solids value for each experiment is shown in Table 1 and the chemical compositions of each leachate solution are shown in Table 5.The eigenvalues for the Varimax rotated PCA model for this data are shown in Table 2. This clearly shows that there are four components with eigenvalues greater than 1 and a step change in eigenvalues between the fourth and fifth PC. In addition, Table 3 shows the scores for the first four significant PCs identifying the elements Fe, Ca, Pb and Mn as having the highest scores. Using these element concentration profiles and four components as being significant the composition of each component was calculated using the chemometric procedure previously described.The total extracted solids in each sample due to each component were modelled separately using a standard MLR method. The total extracted solids for a given component for each sample is regressed against the main effects of acid concentration (A), the time (T), the sample to extractant ratio (S) and all of their linear interactions (AT, AS and TS) and the quadratic effects (A2, T2 and S2).The model takes the form: Ec = k + x1A + x2T + x3S + x4AT + x5AS + x6TS + x7A2 + x8T2 + x9S2 (2) where Ec is the total extracted solids for component c; k a constant term; and x1···x9 the linear regression coefficients. The k term and the x1···x9 coefficients are calculated by the MLR algorithm. For each component model a regression coefficient (R2) is calculated that gives a measure of how well the model fits the data (R2 = 1 is a perfect fit). Initially, all main effects, interactions and quadratic effects are used to form the MLR model.This model is examined and effects which are shown to be insignificant (only t values significant at the 95% confidence interval were retained), are removed one by one (lowest t value first) until the model consists only of effects that are significant. The final MLR models for each component are shown in Table 6. As an additional check for the significance of each effect ANOVA was also carried out on the total extractable solids for each Table 5 Chemical composition of each leach solution obtained using the experimental design shown in Table 1 (results in mg kg21 and in the same order as the experimental design in Table 1) Al Ca Fe Mg P K Si Na Ti Ba Cd Cu Mn Ni V Zn Pb 1932 3346 2744 537 51.6 1134 1217 288 8.29 41.9 19.0 2447 3433 2.12 11.0 2725 4091 2517 3422 4293 622 61.7 1128 849 300 17.2 34.5 19.4 2535 4442 2.02 15.2 3146 4294 2066 3214 1003 591 11.1 868 953 290 1.01 7.38 18.4 2274 4618 1.82 6.07 3128 2299 4633 3374 8085 1466 66.3 1809 2482 385 78.3 125.0 18.5 2579 6153 4.15 20.1 4343 5180 2454 3387 2753 656 35.8 1060 1443 305 3.33 21.0 18.8 2471 4746 2.67 10.0 3324 3268 1537 3149 1071 456 16.2 842 730 270 1.01 8.42 18.0 2206 3084 2.33 6.07 2425 2253 2590 3265 4585 652 67.3 1316 1750 292 33.2 94.0 18.6 2453 4031 2.21 14.8 2900 4633 3092 3423 4381 849 48.9 1269 2378 343 7.37 49.4 18.8 2506 5480 2.77 14.2 3791 4640 4048 3479 7395 1197 60.7 1439 764 354 30.4 27.8 18.8 2574 6557 3.34 17.9 4652 4606 2439 3362 2732 651 35.0 1082 1432 309 3.33 20.6 18.9 2444 4685 2.50 9.92 3269 3185 993 3133 1253 380 31.6 889 351 243 5.83 16.7 18.3 2209 2314 1.83 7.66 2108 2798 2721 3423 3148 764 31.7 1087 1465 333 2.50 20.8 18.2 2394 5584 2.50 12.4 3690 3287 20.0 1152 0.49 188 4.2 436 234 234 0.83 2.09 6.3 175 729 1.17 4.17 839 25.4 4013 3516 7346 1129 70.0 1590 1200 433 39.2 76.2 19.3 2657 6318 3.33 18.8 4447 5093 3197 3422 5069 912 67.7 1588 3299 367 35.9 193.2 18.7 2594 4484 2.99 15.3 3303 4910 1928 3284 823 544 8.00 784 848 307 4.00 5.91 18.0 2337 4351 4.40 2.40 3044 2117 2341 3308 2673 629 3.4 1051 1406 298 3.33 21.5 18.4 2373 4599 2.83 9.91 3181 3213 1906 3314 2710 533 52.5 1125 1235 283 7.37 40.8 19.2 2414 3395 2.39 10.6 2684 4131 2651 3536 4303 639 67.8 1067 869 293 18.2 35.0 20.4 2569 4320 2.33 15.5 3379 4520 2125 3326 1094 617 10.1 898 969 313 1.01 7.23 18.4 2323 4835 2.33 6.27 3213 2387 4479 3398 7963 1447 67.3 1810 2425 380 77.4 123.0 17.9 2515 6115 3.87 20.1 4226 5017 2348 3289 2597 648 33.3 1073 1378 307 3.33 22.5 18.8 2395 4644 2.50 10.6 3244 3275 1519 3153 1061 451 17.2 830 736 259 2.02 8.3 18.2 2204 3019 1.92 5.77 2385 2362 2551 3234 4464 640 66.3 1328 1716 290 33.2 92.1 18.1 2465 3907 2.86 14.7 2856 4744 3026 3369 4374 845 47.0 1260 2354 347 7.37 50.4 17.9 2461 5574 2.85 14.2 3766 4462 3857 3472 7088 1133 58.7 1365 749 346 30.4 27.0 18.4 2535 6523 3.64 18.0 4552 4470 2341 3263 2618 640 34.1 1056 1371 305 3.33 22.3 18.5 2369 4651 2.33 10.6 3236 3236 925 2925 1147 352 30.8 880 337 237 5.00 17.7 17.1 2046 2169 1.67 7.17 1953 2656 2718 3412 3120 761 30.0 1098 1484 333 2.50 20.9 18.2 2403 5598 2.50 12.7 3696 3276 20.8 1147 0.97 188 0.83 429 227 233 0.83 2.08 6.3 170 724 0.75 4.17 843 26.1 3708 3297 6560 1035 65.8 1502 1142 335 37.5 73.4 18.2 2486 5764 3.83 18.2 4128 4923 3103 3349 4866 901 66.1 1560 3161 354 36.0 195.0 18.8 2541 4370 2.20 15.0 3235 4902 2176 4092 758 696 4.0 924 1004 307 4.00 6.45 22.4 2708 4960 2.00 4.40 3488 2469 2374 3309 2627 652 30.8 1056 1392 301 3.33 21.6 18.8 2387 4678 2.42 10.3 3243 3322 1216 Analyst, November 1997, Vol. 122Fig. 3 Chemical compositions of the four resolved components found in the sequential extract data. Table 6 Multiple linear regression models for the four components in the non-specific extraction trial data* Regression coefficient SE† t value p 295% CL‡ +95% CL‡ R2 Variable Component 1— k 2915.6539 776.9068 3.752901 0.000749 1328.999 4502.309 0.548 A 18974.789 3399.415 5.581781 4.51E-06 12032.26 25917.32 A2 214660.055 3137.452 2 4.672599 5.86E-05 221067.59 28252.523 AT 264.284855 18.04548 2 3.562379 0.001251 2101.1387 227.43106 Component 2— k 10723.18 1731.21 6.194038 1.27E-06 7171.031 14275.33 0.919 A 212730.264 4479.041 2 2.842185 0.008428 221920.5 23540.031 A2 14723.282 3623.775 4.062968 0.000375 7287.91 22158.65 S 299468.687 18487.88 2 5.38021 1.1E-05 2137402.7 261534.69 S2 186638.66 60949.61 3.06218 0.004931 61580.39 311696.9 AS 39170.263 17685.89 2.214775 0.035402 2881.821 75458.71 AT 140.48036 20.36927 6.896682 2.07E-07 98.68607 182.2746 Component 3— k 2336.1316 1219.135 2 0.275713 0.784659 22825.937 2153.674 0.788 A 18251.044 4489.926 4.064887 0.000319 9081.391 27420.7 A2 29202.4231 4267.909 2 2.15619 0.039211 217918.66 2486.1903 T 99.33633 13.53768 7.337764 3.58E-08 71.68869 126.984 Component 4— k 3288.867 145.6462 22.58121 2.2E-20 2991.418 3586.316 0.488 A2 21206.386 271.995 24.435324 0.000114 21761.874 2650.8981 S 2371.3722 877.2696 2.703128 0.0112 579.7487 4162.996 AT 16.166035 4.259985 3.794857 0.000669 7.465985 24.86608 * Data significant at the 95% confidence limit (i.e., p < 0.05).† SE, standard error. ‡ CL, confidence limit. Analyst, November 1997, Vol. 122 1217component with the ‘F’ statistic being used to test for significance. The ANOVA tables for the significance the factors is summarised in Table 7.Each MLR model is plotted as a surface plot of the two most significant factors against total extracted solids (Fig. 4) and the chemical composition of each component is given in pie diagrams in a similar manner to the sequential leach data (Fig. 5). Component 1 is predominantly made up of Pb, Ca and Cu and with smaller amounts of Fe, K, Zn, Fe, Si and Al. Table 6 shows that the most important factor controlling its dissolution is acid concentration (significant A and A2 coefficients) with a small but significant interaction effect between time and acid concentration.The ANOVA analysis confirms these findings (A2 and T effects significant at the 95% confidence level). The surface plot (Fig. 4) shows that the optimum acid concentration for dissolution is approximately 0.3–0.7 m with highest concentrations at very short reaction times. This suggests this component dissolves very quickly as long as there is a reasonable acid concentration, but that its concentration decreases slightly with time possibly indicating a re-adsorption effect.Such behaviour and chemical composition could reflect the dissolution of very fine particulate or clayey material. Component 2 is predominantly Fe, the MLR analysis shows the most important factors controlling its dissolution are sample to extractant ratio and acid concentration (significant A, A2, S, S2 and AS coefficients) with an additional time and acid concentration interaction effect. The ANOVA analysis confirms this (significant A, A2, S, S2, T, AS and AT effects). Its dissolution is favoured by low sample to extractant ratios, high Table 7 ANOVA tables for the significance of the factors in the non-specific extraction trial data* Sums of Degrees Mean square Variable squares (SS) of freedom value F ratio p Component 1— A 3931968 1 3931968 2.121374 0.158215 A2 2.7E + 07 1 2.7E + 07 14.7554 0.00079 S 4913781 1 4913781 2.651081 0.116538 S2 7796433 1 7796433 4.206329 0.051344 T 1.8E + 07 1 1.8E + 07 9.95551 0.00428 T2 1477269 1 1477269 0.797016 0.380852 AS 967137.7 1 967137.7 0.52179 0.477059 AT 6455185 1 6455185 3.482699 0.074281 ST 243634.1 1 243634.1 0.131445 0.720111 Error 44484014 24 1853501 Total SS 1.28E + 08 33 Component 2— A 4.2E + 08 1 4.2E + 08 177.125 1.4E-12 A2 3.9E + 07 1 3.9E + 07 16.3547 0.00047 S 1.6E + 08 1 1.6E + 08 66.9972 2.1E-08 S2 2.3E + 07 1 2.3E + 07 9.46522 0.00517 T 9E + 07 1 93 + 07 37.4768 2.5E-06 T2 271800.5 1 271800.5 0.113712 0.738888 AS 1.2E + 07 1 1.2E + 07 5.0374 0.03429 AT 2.6E + 07 1 2.6E + 07 11.0809 0.00281 ST 8025314 1 8025314 3.357519 0.079336 Error 57366030 24 2390251 Total SS 8.18E + 08 33 Component 3— A 1.9E + 08 1 1.9E + 08 51.682 2E-07 A2 12767880 1 12767880 3.440736 0.075933 S 1848968 1 1848968 0.498267 0.487059 S2 7284935 1 7284935 1.963172 0.173973 T 1.9E + 08 1 1.9E + 08 51.5757 2E-07 T2 1996555 1 1996555 0.538039 0.470352 AS 2793707 1 2793707 0.752859 0.394166 AT 558913 1 558913 0.150618 0.701367 ST 1164110 1 1164110 0.313709 0.580604 Error 89059172 24 3710799 Total SS 5.05E + 08 33 Component 4— A 566192 1 566192 5.46793 0.02803 A2 343451.7 1 343451.7 3.316845 0.081063 S 755136 1 755136 7.29263 0.01249 S2 58748.29 1 58748.29 0.567355 0.458638 T 849040 1 849040 8.1995 0.00856 T2 83393.03 1 83393.03 0.805358 0.378411 AS 401611.7 1 401611.7 3.878518 0.060556 AT 877586 1 877586 8.47518 0.00766 ST 101024.5 1 101024.5 0.975633 0.333134 Error 2485145 24 103547.7 Total SS 6669169 33 * Bold values indicate correlations significant at the 95% confidence limit (i.e., p < 0.05). 1218 Analyst, November 1997, Vol. 122acid concentrations (Fig. 4) and longer dissolution times. This component is probably derived from iron oxide dissolution. Component 3 is predominantly Mn: the MLR and ANOVA analysis shows the most important factors controlling its dissolution are acid concentration and time (significant A, A2 and T coefficients for the MLR and significant A, A2 and T effects for the ANOVA). Its dissolution is favoured by longer reaction times and high acid concentrations (Fig. 4). This component is probably derived from Mn oxide dissolution. Component 4 is predominantly made up of Ca, Zn and Mn: the MLR analysis shows most important factors controlling its dissolution are acid concentration and ratio (significant A2 and R coefficients) with an additional time and acid concentration interaction effect.The ANOVA analysis confirms this but shows T to be a significant effect on its own (significant A, S, and T effects). Its dissolution is favoured by low acid concentration and high sample to extractant ratio (Fig. 4). The composition of this component does not intuitively point to its origin but the mild conditions which favour its dissolution suggests that this is an easily extractable component, possibly the exchangeable fraction. Comparison of the Two Extraction Methods From a practical point of view, the non-specific extraction method has a number of analytical advantages.The simple nitric acid leaching solution does not cause analytical matrix problems and is likely to have lower blank values than those found in the Tessier extraction scheme. In addition, it allows Mg and Na to be determined; these are masked by the extraction media used in the Tessier method. Comparison of the results of the two chemometric data sets reveals a number of distinct similarities between the two sets of components.Table 8 shows the correlation between the chemical compositions of the components identified in each extraction method data set. The compositions of component 1 from both methods are significantly correlated. This fraction is dominated by the Fig. 4 Surface plots of the MLR models of the four components identified in the non-specific extraction trial. Analyst, November 1997, Vol. 122 1219metals Cu, Pb, Mn and Zn but also has significant quantities of Al and K.This component could be a fine clay material which adsorbs heavy metals. Alternatively, this could be an organic material. Further work is required to identify the source of this extracted fraction. Component 2 from the Tessier method shows no significant correlation with the non-specific method. This is not surprising as this is the silicate matrix component which is unlikely to be attacked by the relatively mild dissolution conditions of the non-specific extraction method.Component 3 from the Tessier method has a low but significant correlation with both components 2 and 3 of the nonspecific method. This component is predominant in the Fe/Mn oxide designated fraction and the components identified in the non-specific method are dominated by Fe and Mn respectively. The sum of the compositions of components 2 and 3 in the nonspecific method give a high and significant correlation with component 3 of the Tessier method.This suggests that the Tessier method extracts both Fe and Mn oxides simultaneously, whereas the non-specific method has resolved the Fe and Mn oxides as separate entities. The composition of component 4 from both methods is significantly correlated. The predominance of this component in the designated exchangeable fraction in the Tessier scheme and the fact that it is extracted under very mild extraction conditions suggests that this is the exchangeable fraction.Conclusions The application of a chemometric mixture resolution procedure to a well established sequential leach method and to a new nonspecific leach procedure has produced data that are geochem- Fig. 5 Chemical compositions of the four resolved components found in the non-specific extraction trial data. Table 8 Correlation coefficients between the component compositions found in the Tessier sequential leach data and the non-specific extraction trial data* Tessier method Component 1 Component 2 Component 3 Component 4 Non-specific method Correlation p value Correlation p value Correlation p value Correlation p value Component 1 0.7933 0.001 20.1003 0.745 20.1186 0.7 0.3552 0.234 Component 2 0.2117 0.487 0.3268 0.276 0.5664 0.044 20.1752 0.567 Component 3 0.0961 0.755 0.1702 0.578 0.5831 0.036 0.4069 0.168 Component 4 20.0368 0.905 20.1186 0.7 0.1768 0.563 0.9471 < 0.0001 Component 2 + 3 0.2422 0.425 0.3891 0.189 0.8759 < 0.0001 0.1398 0.649 * Effects in bold are significant at the 95% confidence limit. 1220 Analyst, November 1997, Vol. 122ically consistent with the material being studied. It has revealed a certain lack of specificity in the Tessier method for some phases and has been shown to be a potentially powerful method for studying the fate of heavy metals in soils and sediments. The non-specific extraction trial scoping study has demonstrated considerable promise. The results are comparable with the data independently obtained by the Tessier scheme15 and suggest that the new method has more flexibility and selectivity in identifying the presence of different physico-chemical components within a soil material and the trace elements associated with it.The method has considerable potential for application to environmental pollution studies and to geochemical exploration work. This paper is published with the approval of Director, British Geological Survey (NERC).References 1 Leschber, R., Davis, R. D., and L’Hermite, P., Chemical Methods for Assessing Bio-available Metals in Sludges and Soils, Elsevier, London, 1985. 2 Broekaert, J. A. C., G�uçer, S., and Adams, F., Metal Speciation in the Environment, Springer, Berlin, 1990. 3 Jenne, E. A., in Proceedings of the Symposium on Molybdenum in the Environment, ed. Chappell, W., and Peterson, S. K., Marcel Dekker, New York, 1977, pp. 425–552. 4 Kramer, J. R., and Allen, H. E., Metal Speciation: Theory, Analysis and Application, Lewis, Chelsea, Michigan, 1988. 5 Tessier, A., Campbell, P.G. C., and Bisson, M., Anal. Chem., 1979, 51, 844. 6 Breward, N., and Peachey, D., Sci. Total Environ., 1983, 29, 155. 7 Lake, D. L., Kirk, P. W. W., and Lester, J. N., J. Environ. Qual., 1984, 13, 175. 8 Schuman, L. M., Soil Sci., 1985, 140, 11. 9 Harrison, R. M., Laxen D. P. H., and Wilson, S. J., Environ. Sci. Technol., 1981, 15, 1378. 10 Hickey, M. G., and Kitterick, J.A., J. Environ. Qual., 1984, 13, 372. 11 F�orstner, U., in Chemical Methods for Assessing Bio-available Metals in Sludges and Soils, ed., Leschber, R., Davis, R. D., and L’Hermite, P., Elsevier, London, 1985, pp. 1–31. 12 Xian, X., Environ. Pollut., 1989, 57, 127. 13 Clevinger, T. E., Water, Air Soil Pollut., 1990, 50, 241. 14 Li, X., Coles, B. J., Ramsey, M. H., and Thornton, I., Chem. Geol., 1995, 124, 109. 15 Li, X., Coles, B. J., Ramsey, M. H., and Thornton, I., Analyst, 1995, 120, 1415. 16 Jouanneau, J. M., Latouche, C., and Pautrizel, F., Environ. Technol. Lett., 1983, 4, 509. 17 Tipping, E., Hetherington, N. B., Hilton, J., Thompson, D. W., Bowles, E., and Hamilton-Taylor, J., Anal. Chem., 1985, 57, 1944. 18 Khebonian, C., and Bauer, C., Anal. Chem., 1987, 59, 1417. 19 Sholkovitz, E. R., Chem. Geol., 1989, 77, 47. 20 Bermond, A., Environ. Technol., 1992, 23, 1175. 21 Cave, M. R., and Harmon K., Analyst, 1997, 122, 501. 22 Malinowski, E. R., Factor Analysis in Chemistry, Wiley, New York, 2nd edn., 1991. 23 Brereton, R. G., Analyst, 1995, 120, 2313. 24 Hopke, P. K., in Chemometrics in Environmental Chemistry— Applications, ed. Einax, J., Springer, Berlin, 1995, vol. 2, part H, pp. 47–86. 25 Thurston, G. D., and Spengler, J. D., Atmos. Env., 1985, 19, 9. 26 Gamp, H., Maeder, M., Meyer, C. J., and Zuberbuhler, A. D., Talanta, 1985, 32, 1133. 27 Kaiser, H. F., Psychometrika, 1958, 23, 187. Paper 7/05163H Received July 18, 1997 Accepted October 8, 1997 Analyst, November 1997, Vol. 122 1221 Measurement of Trace Element Distributions in Soils and Sediments Using Sequential Leach Data and a Non-specific Extraction System With Chemometric Data Processing† Mark R. Cave and Joanna Wragg Analytical and Regional Geochemistry Group, British Geological Survey, Keyworth, Nottingham, UK NG12 5GG A chemometric mixture resolution procedure suitable for determining the number and composition of physico-chemical components in data derived from soil leachates is described.The procedure is used to determine the number of components in sequential leachate data obtained for a NIST certified soil (SRM 2710) using a widely employed leaching scheme. The resulting data show that the sequential leaching media are not specific for their designated target fractions and that erroneous identification of fractions occurs. A scoping study in which a new non-specific extraction method is tested is described.The experimental design varies the concentration of nitric acid, the reaction time and the ratio of sample to extractant. The resulting solutions were analysed by ICP-AES for major and trace metals and the data obtained from 34 experiments subjected to the chemometric resolution procedure. Four components are identified and the effects of the three variables on each component are modelled using multiple linear regression, allowing the conditions which favour dissolution of each component to be identified.Calculated element compositions of the components identified in the non-specific extraction trial are compared with those identified in the sequential extraction data. Significant correlations between the two sets of components are noted and tentative identification of the source of the components is made. In particular, there is evidence that the Tessier method extracts both Fe and Mn oxides simultaneously, whereas the non-specific method has resolved the Fe and Mn oxides as separate entities.Keywords: Sequential leaching; chemometric mixture resolution; soils analysis; trace element partitioning; inductively coupled plasma The determination of potentially toxic inorganic substances (e.g., heavy metals) in soils and sediments is an important tool for monitoring environmental pollution. Although the total concentrations of these potentially toxic substances provides broad evidence for possible contamination, it has been recognised that quantification of the chemical forms of metals in soils is essential for estimating the mobility and bioavailability of the metals in the environment.1,2 Metals in soils may be present in several different geochemical phases that act as reservoirs or sinks of trace elements in the environment.3,4 The chemical phases considered to be important are divided up into a series of broad categories usually consisting of: exchangeable; specifically adsorbed; carbonate; Fe and Mn-oxides; organic matter; mineral lattice.To obtain information on the distributions of trace metals between these soil/sediment phases, a number of workers5–8 have developed extraction schemes in which phases are selectively dissolved with carefully chosen reagents. By subsequent chemical analysis of the extraction media the concentration of trace elements associated with the target phase can be determined. The method of Tessier et al.5 has been widely adopted in a number of applications.9–15 Despite the widespread use of these selective extraction methods and the insight they have given to understanding the geochemical processes governing trace metal distributions, selective extraction techniques have been demonstrated to have a number of weaknesses,16–20 the two most important being: 1 The so-called ‘selective extraction’ reagents are not specific for one mineral phase; therefore the associated analysis is not a true representation of the amount of trace elements from a single phase. 2 The design of the selective extraction schemes leads to a methodological definition of the distribution of trace elements between solid phases which may not reflect the actual distribution within the test samples.A recent study by Cave and Harmon21 investigated the trace elements associated with the iron oxide phase of red-bed sandstones and showed that chemometric processing of the iron oxide phase data from related samples could identify the presence of more than one component being mobilised by the so-called ‘selective extractive reagent’.This work confirmed the limitations noted earlier and also suggests an alternative approach to the study of speciation that should overcome some of the problems of the traditional methods. If the chemical composition of the products of individual extraction steps within a sequential extraction procedure are considered to be mixtures of the physico-chemical components of the soil or sediment, each containing different proportions of each component, then a similar procedure to that described by Cave and Harmon21 could be used to identify and quantify these components.Further, if this chemometric mixture resolution is viable for the sequential extraction data, a new extraction method could be applied in which a relatively simple nonspecific reagent is used to extract the different physico-chemical phases from the target soil or sediment. The resulting solution would be made up of a mixture of different proportions of each physico-chemical phase.By producing a series of these mixed phase solutioith different proportions of each phase present, chemometric methods could be used resolve the composition of each phase. The method used to produce the solutions containing the different proportions of each phase could be: (1) time series extractions, because different phases should dissolve at different rates; (2) variation of rock/ extractant ratio; (3) a series of different extractant concentrations.The main advantages of this approach would be the simplicity of the extraction procedures, as only one extractant † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1211–1221) 1211would be required, and the partitioning of trace metals between the different phases would not be methodologically defined. In this study, a mixture resolution approach to interpreting a traditional sequential extraction scheme was carried out and a feasibility study on the use of a simple single extraction approach combined with chemometric data processing is reported.Experimental Test Material To test the new methodologies, a well characterised material was required. At present there are very few, if any, SRMs certified for leaching purposes.Li et al.,15 however, have carried out a rigourous Tessier style5 sequential extraction Fig. 1 Relationship between the leachate concentration matrix, the matrix containing the physico-chemical component compositions and the matrix containing the proportions of each physico-chemical component in each leachate solution. Table 1 Central composite experimental design for the non-specific extraction trial and total solids extracted for each of the 4 identified components Levels (see below for actual values) Total extracted solids/mg kg21 Sample : Nitric acid extractant Component Component Component Component Replicate Time concentration ratio 1 2 3 4 1 21 21 21 8689 4485 7749 3105 1 21 1 1 7438 7279 10151 2830 1 1 21 1 6191 690 10839 3629 1 1 1 21 4704 16634 15746 3717 1 0 0 0 6653 4456 11153 3696 1 21 21 1 6495 854 6815 3911 1 21 1 21 7792 8601 9464 2842 1 1 21 21 7442 8359 13292 3200 1 1 1 1 4258 13686 15624 3656 1 0 0 0 6490 4442 10981 3765 1 21.67 0 0 7524 1113 4404 3712 1 1.67 0 0 5844 5277 13003 3859 1 0 21.67 0 9 0 138 3905 1 0 1.67 0 5448 13904 15187 3430 1 0 0 21.67 7851 10618 11185 3826 1 0 0 1.67 6242 191 10126 3851 1 0 0 0 6497 4333 10669 3633 2 21 21 21 8757 4446 7663 2999 2 21 1 1 7939 7331 10138 2896 2 1 21 1 6302 831 11248 3765 2 1 1 21 4432 16333 15328 3994 2 0 0 0 6690 4186 10862 3550 2 21 21 1 6747 838 6680 3768 2 21 1 21 8184 8363 9250 2624 2 1 21 21 6974 8326 13364 3314 2 1 1 1 4256 12983 15337 3670 2 0 0 0 6543 4233 10862 3538 2 21.67 0 0 7101 1030 4089 3487 2 1.67 0 0 5853 5226 13053 3852 2 0 21.67 0 2 0 138 3882 2 0 1.67 0 5754 12406 13979 2956 2 0 0 21.67 7982 10169 10921 3603 2 0 0 1.67 7498 0 11116 5013 2 0 0 0 6745 4242 10947 3493 Nitric Level Time acid/m Sample : extractant ratio 21.67 0.5 0.01 0.01 21 18.3 0.21 0.05 0 44.75 0.50 0.12 1 71.79 0.80 0.20 1.67 90.00 1.0 0.25 1212 Analyst, November 1997, Vol. 122procedure on an NIST certified soil (SRM 2710) in which they looked at extending the number of trace metals under study by modifying the method of Tessier et al.5 This particular SRM is highly contaminated soil from pasture land along Silver Bow Creek in the Butte, Montana, area.Li et al.15 noted that this sample was unusual in that it had a high level of heavy metals in the exchangeable fraction, which would make it a suitable material for studying the mobility and bioavailabilty of metals in contaminated soils. Because of the availability of data and the nature of the sample, the published data15 for this soil was used for the mixture resolution exercise and a sample of the soil was used for the non-specific extraction trial.Chemometric Mixture Resolution Procedure Chemometric strategies for mixture resolution have been used widely in analytical chemistry22,23 and have also been used for studying environmental data sets.21,24,25 The procedure developed here is a combination of the methods of Thurston and Spengler,25 Gamp et al.26 and Cave and Harmon.21 This method is based on the assumption that the sample (soil or sediment) is made up of a number of physico-chemical components each of which has its own chemical composition (e.g., carbonate component, iron oxide component). By leaching the sample, under certain conditions, a proportion of these components is leached into solution.The concentration of an element in a particular leach solution can be described as a linear sum of the amounts leached from each physico-chemical sources present, such that: E En n n n c tot = = =åa 1 (1) where Etot is the total concentration of an element E in a given leach solution, Ec the concentration of element E in component n, ac the proportion of element E leached from component n, and c the number of components.In this study there is more than one element being considered and a number of leaches have been carried out. In this instance, eqn. (1) can be expressed in matrix form which is shown pictorially in Fig. 1 showing the dimensions of each matrix with a description of the contents of each matrix. In order to be able to tell which elements are associated with which each physico-chemical component it is necessary to find matrices B and C given matrix A. The first stage in determining B and C is to use principal component analysis (PCA) of matrix A to estimate the number of physico-chemical components (c) present and to give a first estimate of the proportions of each component in each leach (i.e., matrix B).This was carried out in a similar manner to that proposed by Thurston and Spengler25 but required a number of modifications to make the method suitable to the leachate composition data. Firstly, PCA is normally carried out on a matrix with the compositional data (in this case elemental concentrations) in columns and the different samples in rows (as shown by matrix A in Fig. 1) using a pre-scaling procedure in which each column is scaled by subtracting the column mean and dividing by the column standard deviation.This ensures that all elements whether present at high or low concentration contribute equally to the PCA model. In the case of the leachate data, however, the total amount of material in each leachate can vary considerably (for example, in the Tessier method differences between the amount leached in the exchangeable and the residual fractions can vary by 1–2 orders of magnitude) and scaling down the element columns followed by PCA would have produced a model which was dominated by the leachate samples in which the largest amount of material was extracted. Table 2 Eigenvalues and percentage variance explained for the Varimax rotated principal components for the Tessier and non-specific extraction data Extraction method Tessier method Non-specific Principal component Variance Variance no. Eigenvalue (%) Eigenvalue (%) 1 1.5 30.6 16.3 47.9 2 1.2 24.9 5.1 15.1 3 1.1 21.2 3.8 11.2 4 1.0 20.6 6.9 20.3 5 0.13 2.7 0.14 0.41 6 — — 0.07 0.20 Table 3 Varimax rotated scores the Tessier and non-specific extraction data (numbers in bold indicate elements with the highest score within a given PC) Tessier method Non-specific method Principal component Principal component Element 1 2 3 4 Element 1 2 3 4 Al 0.15 3.26 20.67 20.49 Al 0.78 20.98 20.69 1.26 Ca 20.48 0.24 20.75 3.07 Ba 20.74 20.48 20.32 20.34 Cd 20.64 20.51 20.40 20.51 Ca 20.20 2.60 1.64 20.18 Co 20.64 20.51 20.39 20.53 Cd 20.79 20.48 20.31 20.29 Cu 1.67 20.53 20.34 20.67 Cu 20.46 20.77 1.63 1.34 Fe 20.31 1.05 2.53 20.79 Fe 3.03 20.52 20.37 21.78 K 0.19 0.74 20.57 0.63 K 20.17 0.87 0.09 20.95 Mn 20.01 20.53 2.20 1.07 Mg 20.30 0.07 20.46 20.33 Ni 20.64 20.51 20.39 20.53 Mn 1.06 0.82 21.07 2.56 P 20.58 20.48 20.34 20.54 Na 20.66 0.30 20.35 20.63 Pb 2.95 20.45 20.19 20.05 Ni 20.80 20.49 20.32 20.29 Sr 20.63 20.50 20.40 20.46 P 20.76 20.49 20.28 20.32 Ti 20.64 20.34 20.41 20.52 Pb 0.97 21.28 2.53 0.23 V 20.64 20.51 20.39 20.53 Si 0.07 0.10 20.88 20.18 Zn 0.24 20.39 0.52 0.83 Ti 20.77 20.49 20.33 20.31 V 20.79 20.48 20.32 20.30 Zn 0.51 1.68 20.18 0.49 Analyst, November 1997, Vol. 122 1213This could lead to an underestimate of the number of physicochemical components being leached. It was therefore decided to transpose matrix A (AT) and scale the matrix over each leachate, therefore making each leachate contribute equally to the PCA model.PCA of the AT matrix was carried out followed by Varimax27 rotation. According to Thurston and Spengler,25 this procedure provides abstract PC score and loadings matrices which should be closely related to the true proportion and component concentration matrices (matrices B and C, respectively, in Fig. 1). The number of physico-chemical components (c) present was found by the number of PCs with eigenvalues greater than one after Varimax rotation.25 The next stage is to use the abstract PCs derived from the PCA of AT to provide a first approximation of the proportions of matrix B.Thurston and Spengler25 used a procedure where they derived absolute PC scores which could be used as the first estimate. This procedure was not successful for the leachate data which gave proportions with negative values. This step was therefore replaced by that used by Cave and Harmon21 in which each of the significant (c) PCs within the Varimax rotated scores matrix was examined. The elements which have the highest scores within each of these PCs are assumed to be the most highly correlated to the rotated PCs and hence to the true physico-chemical components.The concentration of these elements in each leachate should be linearly related to the columns in matrix B. Regressing the concentrations of each of these identified elements against the total extractable solids for each leachate solution, using multiple linear regression (MLR), gives estimates of the coefficients which convert the element concentrations into physico-chemical component mass contributions (in mg kg21) for each leachate solution, allowing a first approximation of B to be calculated.In the final stage matrix A is scaled so that each element concentration is expressed as a fraction of the total extracted mass for each leach solution (AA). Similarly, the first approximation of B is scaled so that each mass contribution is expressed as a fraction of the total extracted mass for each leach solution (BA). Using the pseudoinverse calculation22 ( in other words an MLR regression of BA against AA) a first estimate of CA (the scaled physico-chemical component concentration matrix) can be calculated.The first estimate of CA is further refined by setting any negative values to 0 and by scaling each row so that the relative element contributions for each component adds up to 1. Using the pseudoinverse calculation with the refined CA and matrix AA a second approximation for BA is calculated.The second approximation for BA is refined by setting negative values to 0 and by scaling each column so that the relative mass contribution for each leachate adds up to 1. Further estimates of BA and CA are iteratively calculated using this procedure until no further improvement in their values is obtained. This follows the method first used by Gamp et al.26 and more recently by Cave and Harmon.21 Finally the refined matrices BA and CA are rescaled to absolute concentration values.The PCA, Varimax rotation and MLR calculations were carried out using Statistica (Version 5.1) and the iterative pseudo-inverse calculation routine was written in MathCad Plus (Version 6.0). Non-specific Extraction Procedure A central composite experimental design procedure was used to investigate the effects of three variables (time, nitric acid concentration and sample to extractant ratio) on the chemical composition of the resulting acid extracts from SRM 2710.Using this formal experimental design, the relative effects and interactions of each parameter could be measured and MLR modelling of the response surface was carried out. The five levels used for each variable and the experimental design are shown in Table 1. Each measurement was carried out Fig. 2 Relative proportions of the each resolved component in the sequential extraction data. 1214 Analyst, November 1997, Vol. 122in duplicate resulting in a total of 34 test solutions for analysis. The extractions were carried out in 30 ml polycarbonate screw top tubes which were mixed on an ‘end-over-end’ rotating shaker. All experiments were carried out in an air conditioned laboratory with the temperature nominally maintained at 20 °C. The experimental design matrix and the analysis of the results was carried out using the experimental design module of Statistica Version (5.1).Analysis of the Acid Leachate The 34 leachate solutions generated by the experimental design procedure were analysed by ICP-AES for Al, Ba, Ca, Cd, Cu, Fe, K, Mg, Mn, Na, Ni, P, Pb, Si, Ti, V and Zn. The instrumentation and wavelengths used have been previously described.21 Results and Discussion Sequential Leach Data In addition to the average concentration of each element extracted in the sequential extraction scheme described by Li et al.,15 uncertainty measurements for each determinand were also supplied.In order to include this information in the chemometric data processing, an additional nine sets of data were generated with random amounts of uncertainty, within the reported limits, and were added to the average values. The original data matrix of 15 elements (columns) and 5 leachates (rows) was combined with the additional nine data sets to produce a single matrix of 150 columns and 5 rows. This new matrix was then processed as a single data matrix using the mixture resolution procedure described in the experimental section.The eigenvalues for the Varimax rotated PCA model for the Tessier extraction data are shown in Table 2. This clearly shows that there are four components with eigenvalues greater than one and a step change in eigenvalue between the fourth and fifth PC. In addition, Table 3 shows the scores for the first four significant PCs identifying the elements Pb, Al, Fe and Ca as having the highest scores.Using these element concentration profiles and four components as being significant, the composition of each component was calculated using the chemometric procedure previously described. The relative contributions of each identified component for the ten sets of data were then recombined by taking an average. The standard deviation of the data was used to give a measure of the uncertainty for the relative proportion of any component in a given extraction step. Fig. 2 shows a plot of the calculated average relative proportions of each component in each of the 5 sequential leaches with error bars representing twice the standard deviation on each value (n = 10).Fig. 3 shows the chemical compositions of each of the four component represented as a pie diagram. Component 1 makes up a significant proportion of the first four extracts with particularly high levels in the designated carbonate and organic/sulfide extracts. This component is made up principally from the heavy metals Pb, Cu and Zn.The origin of this component is not clear but may be related to an organic rich component or possibly a fine clay fraction. The reasons for this will be discussed later. Component 2 only appears in the last two extracts and is more than 50% Al. This can be interpreted as the alumino-silicate matrix of the soil. Component 3 appears predominantly in the Fe/Mn oxide extracted fraction although the designated organic/sulfide and residual fractions contain approximately 10% of this component which is made up predominantly from Mn and Fe.This indicates an Fe/Mn oxide component. Component 4 appears predominantly in the designated exchangeable and carbonate fractions and is dominated by Ca, Mn, Zn, K and Pb. The high proportion of this component in the first fraction indicates this is the easily exchangeable fraction. Although components 2–4 appear predominantly in a single extract and are consistent with the designation of the extract in which they occur, significant proportions of each component ‘spill’ over into preceding or subsequent fractions.Component 1, however, is predominant in two designated fractions and does not clearly fit into either the carbonate or the organic/sulfide designations. If the data processing method used is valid, the problems of non-specificity of extraction reagents have been demonstrated. In addition, possible mis-identification of extracted fractions has also been shown.By multiplying the proportion of each component by its chemical composition and re-scaling to the total extracted solids for a given fraction, an analogous table to that previously produced15 can be calculated (Table 4). In this instance, however, it is not the methodologically defined fractions that are reported but the composition of each of the resolved Table 4 Resolved component compositions for NIST 2710 in mg g21 with ±2 s (n = 10) Component 1 2 3 4 S CTV Element Value Error Value Error Value Error Value Error Value Error Value Error Al 322 65.2 63 700 5600 608 227 0.00 0 64 600 5600 64 400 800 Ca 0.00 0 10 000 883 0.00 0 2510 1330 12 500 1600 12 500 300 Cd 1.89 0.38 0.00 0 3.06 1.14 14.8 7.86 19.8 7.95 21.8 0.2 Co 1.56 0.31 2.24 0.2 3.22 1.2 0.51 0.27 7.53 1.28 (10) Cu 2020 408 586 51.5 515 192 0.00 0 3120 454 2950 130 Fe 71.8 14.5 22 700 1990 10 600 3950 0.00 0 33 400 4420 33 800 1000 K 379 76.6 17 000 1494 0.00 0 893 473 18 300 1570 21 100 1100 Mn 751 152 0.00 0 9860 3680 1240 657 11 900 3740 10 100 400 Ni 0.99 0.2 8.65 0.76 0.13 0.05 0.08 0.04 9.85 0.79 14.3 1 P 65.5 13.2 791 69.6 227 84.8 0 0 1080 111 1060 150 Pb 3040 615 15.4 1.35 1210 452 523 277 4790 812 5532 80 Sr 6.76 1.37 263 23.1 7.87 2.94 48.6 25.8 326 34.8 240 Ti 0.00 0 2530 222 0.00 0 10.8 5.72 2540 222 2830 100 V 3.33 0.67 53.3 4.69 17.9 6.69 0.00 0 74.5 8.2 76.6 2.3 Zn 789 160 1055 92.7 3760 1410 1030 547 6630 1520 6952 91 * S represents the sum of the four components.† CTV is the certified total concentration. Analyst, November 1997, Vol. 122 1215components and their associated uncertainties that is given. This is a useful check to show that at the end of the processing procedure the total amount of each element extracted is still comparable with the certified values for the soil. Non-specific Extraction Trial The requirements for the non-specific extraction experiment were: (i) to produce a series of leachate solutions containing varying proportions of the different physico-chemical components of the soil, to allow the chemometric procedure to identify and quantify each component; and (ii) to carry out the experiments in such a way that analysis of data would allow the effects and interactions of the three variables (acid concentration, sample to extractant ratio and time) on the dissolution of the physico-chemical components of the soil to be studied. In order to achieve these objectives a formal experimental design was required. The central composite design was chosen as being more suitable than a two level design as it allows curvature of effects to be investigated and the results can be more readily visualised as 3 dimensional surface plots.The design outlined in Table 1 was chosen from a menu of designs given within the Statistica software package. The total extracted solids value for each experiment is shown in Table 1 and the chemical compositions of each leachate solution are shown in Table 5.The eigenvalues for the Varimax rotated PCA model for this data are shown in Table 2. This clearly shows that there are four components with eigenvalues greater than 1 and a step change in eigenvalues between the fourth and fifth PC. In addition, Table 3 shows the scores for the first four significant PCs identifying the elements Fe, Ca, Pb and Mn as having the highest scores. Using these element concentration profiles and four components as being significant the composition of each component was calculated using the chemometric procedure previously described.The total extracted solids in each sample due to each component were modelled separately using a standard MLR method. The total extracted solids for a given component for each sample is regressed against the main effects of acid concentration (A), the time (T), the sample to extractant ratio (S) and all of their linear interactions (AT, AS and TS) and the quadratic effects (A2, T2 and S2).The model takes the form: Ec = k + x1A + x2T + x3S + x4AT + x5AS + x6TS + x7A2 + x8T2 + x9S2 (2) where Ec is the total extracted solids for component c; k a constant term; and x1···x9 the linear regression coefficients. The k term and the x1···x9 coefficients are calculated by the MLR algorithm. For each component model a regression coefficient (R2) is calculated that gives a measure of how well the model fits the data (R2 = 1 is a perfect fit).Initially, all main effects, interactions and quadratic effects are used to form the MLR model. This model is examined and effects which are shown to be insignificant (only t values significant at the 95% confidence interval were retained), are removed one by one (lowest t value first) until the model consists only of effects that are significant. The final MLR models for each component are shown in Table 6. As an additional check for the significance of each effect ANOVA was also carried out on the total extractable solids for each Table 5 Chemical composition of each leach solution obtained using the experimental design shown in Table 1 (results in mg kg21 and in the same order as the experimental design in Table 1) Al Ca Fe Mg P K Si Na Ti Ba Cd Cu Mn Ni V Zn Pb 1932 3346 2744 537 51.6 1134 1217 288 8.29 41.9 19.0 2447 3433 2.12 11.0 2725 4091 2517 3422 4293 622 61.7 1128 849 300 17.2 34.5 19.4 2535 4442 2.02 15.2 3146 4294 2066 3214 1003 591 11.1 868 953 290 1.01 7.38 18.4 2274 4618 1.82 6.07 3128 2299 4633 3374 8085 1466 66.3 1809 2482 385 78.3 125.0 18.5 2579 6153 4.15 20.1 4343 5180 2454 3387 2753 656 35.8 1060 1443 305 3.33 21.0 18.8 2471 4746 2.67 10.0 3324 3268 1537 3149 1071 456 16.2 842 730 270 1.01 8.42 18.0 2206 3084 2.33 6.07 2425 2253 2590 3265 4585 652 67.3 1316 1750 292 33.2 94.0 18.6 2453 4031 2.21 14.8 2900 4633 3092 3423 4381 849 48.9 1269 2378 343 7.37 49.4 18.8 2506 5480 2.77 14.2 3791 4640 4048 3479 7395 1197 60.7 1439 764 354 30.4 27.8 18.8 2574 6557 3.34 17.9 4652 4606 2439 3362 2732 651 35.0 1082 1432 309 3.33 20.6 18.9 2444 4685 2.50 9.92 3269 3185 993 3133 1253 380 31.6 889 351 243 5.83 16.7 18.3 2209 2314 1.83 7.66 2108 2798 2721 3423 3148 764 31.7 1087 1465 333 2.50 20.8 18.2 2394 5584 2.50 12.4 3690 3287 20.0 1152 0.49 188 4.2 436 234 234 0.83 2.09 6.3 175 729 1.17 4.17 839 25.4 4013 3516 7346 1129 70.0 1590 1200 433 39.2 76.2 19.3 2657 6318 3.33 18.8 4447 5093 3197 3422 5069 912 67.7 1588 3299 367 35.9 193.2 18.7 2594 4484 2.99 15.3 3303 4910 1928 3284 823 544 8.00 784 848 307 4.00 5.91 18.0 2337 4351 4.40 2.40 3044 2117 2341 3308 2673 629 3.4 1051 1406 298 3.33 21.5 18.4 2373 4599 2.83 9.91 3181 3213 1906 3314 2710 533 52.5 1125 1235 283 7.37 40.8 19.2 2414 3395 2.39 10.6 2684 4131 2651 3536 4303 639 67.8 1067 869 293 18.2 35.0 20.4 2569 4320 2.33 15.5 3379 4520 2125 3326 1094 617 10.1 898 969 313 1.01 7.23 18.4 2323 4835 2.33 6.27 3213 2387 4479 3398 7963 1447 67.3 1810 2425 380 77.4 123.0 17.9 2515 6115 3.87 20.1 4226 5017 2348 3289 2597 648 33.3 1073 1378 307 3.33 22.5 18.8 2395 4644 2.50 10.6 3244 3275 1519 3153 1061 451 17.2 830 736 259 2.02 8.3 18.2 2204 3019 1.92 5.77 2385 2362 2551 3234 4464 640 66.3 1328 1716 290 33.2 92.1 18.1 2465 3907 2.86 14.7 2856 4744 3026 3369 4374 845 47.0 1260 2354 347 7.37 50.4 17.9 2461 5574 2.85 14.2 3766 4462 3857 3472 7088 1133 58.7 1365 749 346 30.4 27.0 18.4 2535 6523 3.64 18.0 4552 4470 2341 3263 2618 640 34.1 1056 1371 305 3.33 22.3 18.5 2369 4651 2.33 10.6 3236 3236 925 2925 1147 352 30.8 880 337 237 5.00 17.7 17.1 2046 2169 1.67 7.17 1953 2656 2718 3412 3120 761 30.0 1098 1484 333 2.50 20.9 18.2 2403 5598 2.50 12.7 3696 3276 20.8 1147 0.97 188 0.83 429 227 233 0.83 2.08 6.3 170 724 0.75 4.17 843 26.1 3708 3297 6560 1035 65.8 1502 1142 335 37.5 73.4 18.2 2486 5764 3.83 18.2 4128 4923 3103 3349 4866 901 66.1 1560 3161 354 36.0 195.0 18.8 2541 4370 2.20 15.0 3235 4902 2176 4092 758 696 4.0 924 1004 307 4.00 6.45 22.4 2708 4960 2.00 4.40 3488 2469 2374 3309 2627 652 30.8 1056 1392 301 3.33 21.6 18.8 2387 4678 2.42 10.3 3243 3322 1216 Analyst, November 1997, Vol. 122Fig. 3 Chemical compositions of the four resolved components found in the sequential extract data. Table 6 Multiple linear regression models for the four components in the non-specific extraction trial data* Regression coefficient SE† t value p 295% CL‡ +95% CL‡ R2 Variable Component 1— k 2915.6539 776.9068 3.752901 0.000749 1328.999 4502.309 0.548 A 18974.789 3399.415 5.581781 4.51E-06 12032.26 25917.32 A2 214660.055 3137.452 2 4.672599 5.86E-05 221067.59 28252.523 AT 264.284855 18.04548 2 3.562379 0.001251 2101.1387 227.43106 Component 2— k 10723.18 1731.21 6.194038 1.27E-06 7171.031 14275.33 0.919 A 212730.264 4479.041 2 2.842185 0.008428 221920.5 23540.031 A2 14723.282 3623.775 4.062968 0.000375 7287.91 22158.65 S 299468.687 18487.88 2 5.38021 1.1E-05 2137402.7 261534.69 S2 186638.66 60949.61 3.06218 0.004931 61580.39 311696.9 AS 39170.263 17685.89 2.214775 0.035402 2881.821 75458.71 AT 140.48036 20.36927 6.896682 2.07E-07 98.68607 182.2746 Component 3— k 2336.1316 1219.135 2 0.275713 0.784659 22825.937 2153.674 0.788 A 18251.044 4489.926 4.064887 0.000319 9081.391 27420.7 A2 29202.4231 4267.909 2 2.15619 0.039211 217918.66 2486.1903 T 99.33633 13.53768 7.337764 3.58E-08 71.68869 126.984 Component 4— k 3288.867 145.6462 22.58121 2.2E-20 2991.418 3586.316 0.488 A2 21206.386 271.995 24.435324 0.000114 21761.874 2650.8981 S 2371.3722 877.2696 2.703128 0.0112 579.7487 4162.996 AT 16.166035 4.259985 3.794857 0.000669 7.465985 24.86608 * Data significant at the 95% confidence limit (i.e., p < 0.05).† SE, standard error. ‡ CL, confidence limit. Analyst, November 1997, Vol. 122 1217component with the ‘F’ statistic being used to test for significance.The ANOVA tables for the significance the factors is summarised in Table 7. Each MLR model is plotted as a surface plot of the two most significant factors against total extracted solids (Fig. 4) and the chemical composition of each component is given in pie diagrams in a similar manner to the sequential leach data (Fig. 5). Component 1 is predominantly made up of Pb, Ca and Cu and with smaller amounts of Fe, K, Zn, Fe, Si and Al. Table 6 shows that the most important factor controlling its dissolution is acid concentration (significant A and A2 coefficients) with a small but significant interaction effect between time and acid concentration.The ANOVA analysis confirms these findings (A2 and T effects significant at the 95% confidence level). The surface plot (Fig. 4) shows that the optimum acid concentration for dissolution is approximately 0.3–0.7 m with highest concentrations at very short reaction times. This suggests this component dissolves very quickly as long as there is a reasonable acid concentration, but that its concentration decreases slightly with time possibly indicating a re-adsorption effect.Such behaviour and chemical composition could reflect the dissolution of very fine particulate or clayey material. Component 2 is predominantly Fe, the MLR analysis shows the most important factors controlling its dissolution are sample to extractant ratio and acid concentration (significant A, A2, S, S2 and AS coefficients) with an additional time and acid concentration interaction effect.The ANOVA analysis confirms this (significant A, A2, S, S2, T, AS and AT effects). Its dissolution is favoured by low sample to extractant ratios, high Table 7 ANOVA tables for the significance of the factors in the non-specific extraction trial data* Sums of Degrees Mean square Variable squares (SS) of freedom value F ratio p Component 1— A 3931968 1 3931968 2.121374 0.158215 A2 2.7E + 07 1 2.7E + 07 14.7554 0.00079 S 4913781 1 4913781 2.651081 0.116538 S2 7796433 1 7796433 4.206329 0.051344 T 1.8E + 07 1 1.8E + 07 9.95551 0.00428 T2 1477269 1 1477269 0.797016 0.380852 AS 967137.7 1 967137.7 0.52179 0.477059 AT 6455185 1 6455185 3.482699 0.074281 ST 243634.1 1 243634.1 0.131445 0.720111 Error 44484014 24 1853501 Total SS 1.28E + 08 33 Component 2— A 4.2E + 08 1 4.2E + 08 177.125 1.4E-12 A2 3.9E + 07 1 3.9E + 07 16.3547 0.00047 S 1.6E + 08 1 1.6E + 08 66.9972 2.1E-08 S2 2.3E + 07 1 2.3E + 07 9.46522 0.00517 T 9E + 07 1 93 + 07 37.4768 2.5E-06 T2 271800.5 1 271800.5 0.113712 0.738888 AS 1.2E + 07 1 1.2E + 07 5.0374 0.03429 AT 2.6E + 07 1 2.6E + 07 11.0809 0.00281 ST 8025314 1 8025314 3.357519 0.079336 Error 57366030 24 2390251 Total SS 8.18E + 08 33 Component 3— A 1.9E + 08 1 1.9E + 08 51.682 2E-07 A2 12767880 1 12767880 3.440736 0.075933 S 1848968 1 1848968 0.498267 0.487059 S2 7284935 1 7284935 1.963172 0.173973 T 1.9E + 08 1 1.9E + 08 51.5757 2E-07 T2 1996555 1 1996555 0.538039 0.470352 AS 2793707 1 2793707 0.752859 0.394166 AT 558913 1 558913 0.150618 0.701367 ST 1164110 1 1164110 0.313709 0.580604 Error 89059172 24 3710799 Total SS 5.05E + 08 33 Component 4— A 566192 1 566192 5.46793 0.02803 A2 343451.7 1 343451.7 3.316845 0.081063 S 755136 1 755136 7.29263 0.01249 S2 58748.29 1 58748.29 0.567355 0.458638 T 849040 1 849040 8.1995 0.00856 T2 83393.03 1 83393.03 0.805358 0.378411 AS 401611.7 1 401611.7 3.878518 0.060556 AT 877586 1 877586 8.47518 0.00766 ST 101024.5 1 101024.5 0.975633 0.333134 Error 2485145 24 103547.7 Total SS 6669169 33 * Bold values indicate correlations significant at the 95% confidence limit (i.e., p < 0.05). 1218 Analyst, November 1997, Vol. 122acid concentrations (Fig. 4) and longer dissolution times. This component is probably derived from iron oxide dissolution. Component 3 is predominantly Mn: the MLR and ANOVA analysis shows the most important factors controlling its dissolution are acid concentration and time (significant A, A2 and T coefficients for the MLR and significant A, A2 and T effects for the ANOVA).Its dissolution is favoured by longer reaction times and high acid concentrations (Fig. 4). This component is probably derived from Mn oxide dissolution. Component 4 is predominantly made up of Ca, Zn and Mn: the MLR analysis shows most important factors controlling its dissolution are acid concentration and ratio (significant A2 and R coefficients) with an additional time and acid concentration interaction effect.The ANOVA analysis confirms this but shows T to be a significant effect on its own (significant A, S, and T effects). Its dissolution is favoured by low acid concentration and high sample to extractant ratio (Fig. 4). The composition of this component does not intuitively point to its origin but the mild conditions which favour its dissolution suggests that this is an easily extractable component, possibly the exchangeable fraction.Comparison of the Two Extraction Methods From a practical point of view, the non-specific extraction method has a number of analytical advantages. The simple nitric acid leaching solution does not cause analytical matrix problems and is likely to have lower blank values than those found in the Tessier extraction scheme. In addition, it allows Mg and Na to be determined; these are masked by the extraction media used in the Tessier method.Comparison of the results of the two chemometric data sets reveals a number of distinct similarities between the two sets of components. Table 8 shows the correlation between the chemical compositions of the components identified in each extraction method data set. The compositions of component 1 from both methods are significantly correlated. This fraction is dominated by the Fig. 4 Surface plots of the MLR models of the four components identified in the non-specific extraction trial.Analyst, November 1997, Vol. 122 1219metals Cu, Pb, Mn and Zn but also has significant quantities of Al and K. This component could be a fine clay material which adsorbs heavy metals. Alternatively, this could be an organic material. Further work is required to identify the source of this extracted fraction. Component 2 from the Tessier method shows no significant correlation with the non-specific method. This is not surprising as this is the silicate matrix component which is unlikely to be attacked by the relatively mild dissolution conditions of the non-specific extraction method.Component 3 from the Tessier method has a low but significant correlation with both components 2 and 3 of the nonspecific method. This component is predominant in the Fe/Mn oxide designated fraction and the components identified in the non-specific method are dominated by Fe and Mn respectively. The sum of the compositions of components 2 and 3 in the nonspecific method give a high and significant correlation with component 3 of the Tessier method.This suggests that the Tessier method extracts both Fe and Mn oxides simultaneously, whereas the non-specific method has resolved the Fe and Mn oxides as separate entities. The composition of component 4 from both methods is significantly correlated. The predominance of this component in the designated exchangeable fraction in the Tessier scheme and the fact that it is extracted under very mild extraction conditions suggests that this is the exchangeable fraction.Conclusions The application of a chemometric mixture resolution procedure to a well established sequential leach method and to a new nonspecific leach procedure has produced data that are geochem- Fig. 5 Chemical compositions of the four resolved components found in the non-specific extraction trial data. Table 8 Correlation coefficients between the component compositions found in the Tessier sequential leach data and the non-specific extraction trial data* Tessier method Component 1 Component 2 Component 3 Component 4 Non-specific method Correlation p value Correlation p value Correlation p value Correlation p value Component 1 0.7933 0.001 20.1003 0.745 20.1186 0.7 0.3552 0.234 Component 2 0.2117 0.487 0.3268 0.276 0.5664 0.044 20.1752 0.567 Component 3 0.0961 0.755 0.1702 0.578 0.5831 0.036 0.4069 0.168 Component 4 20.0368 0.905 20.1186 0.7 0.1768 0.563 0.9471 < 0.0001 Component 2 + 3 0.2422 0.425 0.3891 0.189 0.8759 < 0.0001 0.1398 0.649 * Effects in bold are significant at the 95% confidence limit. 1220 Analyst, November 1997, Vol. 122ically consistent with the material being studied. It has revealed a certain lack of specificity in the Tessier method for some phases and has been shown to be a potentially powerful method for studying the fate of heavy metals in soils and sediments. The non-specific extraction trial scoping study has demonstrated considerable promise. The results are comparable with the data independently obtained by the Tessier scheme15 and suggest that the new method has more flexibility and selectivity in identifying the presence of different physico-chemical components within a soil material and the trace elements associated with it. The method has considerable potential for application to environmental pollution studies and to geochemical exploration work. This paper is published with the approval of Director, British Geological Survey (NERC). References 1 Leschber, R., Davis, R. D., and L’Hermite, P., Chemical Methods for Assessing Bio-available Metals in Sludges and Soils, Elsevier, London, 1985. 2 Broekaert, J. A. C., G�uçer, S., and Adams, F., Metal Speciation in the Environment, Springer, Berlin, 1990. 3 Jenne, E. A., in Proceedings of the Symposium on Molybdenum in the Environment, ed. Chappell, W., and Peterson, S. K., Marcel Dekker, New York, 1977, pp. 425–552. 4 Kramer, J. R., and Allen, H. E., Metal Speciation: Theory, Analysis and Application, Lewis, Chelsea, Michigan, 1988. 5 Tessier, A., Campbell, P. G. C., and Bisson, M., Anal. Chem., 1979, 51, 844. 6 Breward, N., and Peachey, D., Sci. Total Environ., 1983, 29, 155. 7 Lake, D. L., Kirk, P. W. W., and Lester, J. N., J. Environ. Qual., 1984, 13, 175. 8 Schuman, L. M., Soil Sci., 1985, 140, 11. 9 Harrison, R. M., Laxen D. P. H., and Wilson, S. J., Environ. Sci. Technol., 1981, 15, 1378. 10 Hickey, M. G., and Kitterick, J. A., J. Environ. Qual., 1984, 13, 372. 11 F�orstner, U., in Chemical Methods for Assessing Bio-available Metals in Sludges and Soils, ed., Leschber, R., Davis, R. D., and L’Hermite, P., Elsevier, London, 1985, pp. 1–31. 12 Xian, X., Environ. Pollut., 1989, 57, 127. 13 Clevinger, T. E., Water, Air Soil Pollut., 1990, 50, 241. 14 Li, X., Coles, B. J., Ramsey, M. H., and Thornton, I., Chem. Geol., 1995, 124, 109. 15 Li, X., Coles, B. J., Ramsey, M. H., and Thornton, I., Analyst, 1995, 120, 1415. 16 Jouanneau, J. M., Latouche, C., and Pautrizel, F., Environ. Technol. Lett., 1983, 4, 509. 17 Tipping, E., Hetherington, N. B., Hilton, J., Thompson, D. W., Bowles, E., and Hamilton-Taylor, J., Anal. Chem., 1985, 57, 1944. 18 Khebonian, C., and Bauer, C., Anal. Chem., 1987, 59, 1417. 19 Sholkovitz, E. R., Chem. Geol., 1989, 77, 47. 20 Bermond, A., Environ. Technol., 1992, 23, 1175. 21 Cave, M. R., and Harmon K., Analyst, 1997, 122, 501. 22 Malinowski, E. R., Factor Analysis in Chemistry, Wiley, New York, 2nd edn., 1991. 23 Brereton, R. G., Analyst, 1995, 120, 2313. 24 Hopke, P. K., in Chemometrics in Environmental Chemistry— Applications, ed. Einax, J., Springer, Berlin, 1995, vol. 2, part H, pp. 47–86. 25 Thurston, G. D., and Spengler, J. D., Atmos. Env., 1985, 19, 9. 26 Gamp, H., Maeder, M., Meyer, C. J., and Zuberbuhler, A. D., Talanta, 1985, 32, 1133. 27 Kaiser, H. F., Psychometrika, 1958, 23, 187. Paper 7/05163H Received July 18, 1997 Accepted October 8, 1997 Analyst, November 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a705163h
出版商:RSC
年代:1997
数据来源: RSC
|
7. |
Application of Laser Induced Plasma Spectroscopy to the Analysis of Rock Samples† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1223-1227
Yoon Yeol Yoon,
Preview
|
|
摘要:
Application of Laser Induced Plasma Spectroscopy to the Analysis of Rock Samples† Yoon Yeol Yoon*a, Tae Sam Kima, Kang Sup Chunga, Kil Yong Leea and Gae Ho Leeb a Analytical Research Division, Korea Institute of Geology, Mining and Materials (KIGAM), 30 Kajung-dong, Yusong-ku, Taejeon, 305-350, Korea b Chemistry Department, Chungnam National University, Gung-dong, Yusong-ku, Taejeon, 305-764, Korea Laser induced plasma spectroscopy has been applied to the analysis of element distribution mapping of polished rock sections.The plasma was generated by focusing a frequency-doubled second harmonic 532 nm Nd :YAG laser on the target under atmospheric conditions. The experimental parameters, such as laser energy, atomic emission line and time profile of the plasma spectrum, were characterized to obtain optimum experimental conditions and estimate the element composition of the target surface. For the element mapping of samples, an X–Y stage was used to move the sample and an element image of 50 3 50 mm could be made in 30 min.Using this technique, the element concentration distribution of Ba, Cu, Fe, Mn, Pb, Si and Sr in polished rock sections were obtained. Quantitative analysis was achieved by analyzing standard rock samples. Calibrated concentration versus plasma intensity was used for the color grading for the mapping of element concentration distribution. Keywords: Laser ablation; elemental mapping; geological sample Laser induced plasma spectroscopy (LIPS) is a well established instrumental analysis technique in analytical chemistry.1–7 In LIPS, a short, pulsed, high power laser beam is focused onto the solid target and a plasma spot is created.In the laser induced plasma, a portion of the material instantaneously explodes into vapor. Ablated elements are excited in the plasma, and emit spectral excitation lines. These lines are recorded and analyzed for the identification and quantification of elements.Recently, LIPS has been applied to many kinds of samples such as liquid samples,6,7 direct analysis of solid samples1–4 and depth profiling of metal film.5 Laser ablation coupled with ICPAES and ICP-MS8,9 for the direct analysis of solid samples has also been reported. Plasma characterization experiments have been performed10–12 to discover the optimum ablation conditions. LIPS has various advantages over more conventional methods of atomic emission spectroscopy. The most relevant characteristics are direct analysis of solid samples (conducting and non-conducting materials), no sample preparation, a small sample requirement, the possibility on-line analysis, localized microanalysis, a single step of vaporization and excitation and simultaneous multielement analysis.However, in spite of the above mentioned advantages, LIPS has several drawbacks for the analysis of inhomogeneous samples or when localized microanalysis is desired. For the quantification of samples, the standard used has to show similar physical and chemical matrix effects and provide similar ablation behavior to the sample. To overcome the problems with laser ablation analysis and obtain precise calibration results, matrix effects have been studied.13,14 These studies were based on the intensity of the emission lines as a function of the vaporized mass and the plasma excitation temperature.The analytical procedure commonly used for the correction of matrix effects is based on normalization of the analyte signal by a reference signal.15 In the present work, LIPS was applied as a point analysis technique for obtaining the element distribution mapping of polished rock sections.The optimum analytical conditions have been studied. This mapping technique can be utilized in various fields; interpretation of ore formation history in mineralogy, investigation of segregation phenomena in metallurgy and examination of conditions in an electric circuit board, etc.† Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Fig. 1 Schematic diagram of the experimental set-up for measuring emission intensity from LIPS. Fig. 2 Emission spectra of rock sample with different delay times. Analyst, November 1997, Vol. 122 (1223–1227) 1223Experimental The experimental set-up is shown in Fig. 1. A frequency doubled Nd : YAG Q-switched laser (Spectron SL802G, FWHM 10 ns, Rugby, UK) emitting at 532 nm was used as the ablation source for rock samples.The laser induced plasma is generated by focusing the laser on the sample surface (60° relative to the normal incidence) using a planoconvex 3 cm focal length lens. The pulse repetition rate was 5 Hz and the laser energy was 30 mJ. The crater resulting from the laser pulse was about 10 mm diameter. The analysed sample was positioned about 10 mm above the focal point.This configuration was found to produce a stable plasma and minimize air break-down. Emission from the plasma was collected by a set of combination lenses (Minolta, y = 49 mm, f = 50 mm) for transmitting a planar image of the plasma plume into the collecting planoconvex quartz lens (f = 350 mm) and imaged on to the entrance slit (20 mm slit width) of a spectrometer (SPEX 1404, 0.85 m). The He–Ne laser was used as a pointer to obtain an exact point analysis. The spectra were recorded using a Boxcar integrator (Standford Research, SR250, Palo Alto, CA, USA) coupled with a 1024 pixel photodiode OMA III (spectral range = 20 nm, EG&G Princeton Applied Research, Princeton, NJ, USA; Model 1420) system and a photomultiplier tube (PMT).The collection signal was triggered by a laser pulse signal. To synchronize the laser pulse and emitting signal, the laser pulse signal was delayed for 4 ms using a digital delay generator (Standford Research, DG 535).The sample holder was designed to move in the X and Y direction by a stepping motor and the movement was controlled with a personal computer. Manganese nodule and granite rock samples were analyzed using this set-up. The element concentration was obtained by analyzing a standard granite rock sample. The standard samples were prepared at our Institute a few years ago. To overcome matrix matching properties, standard samples were pelletized at a pressure of 1 MPa.Results and Discussion Temporal Characteristics of the Plasma Emission Emission spectra of rock in air at atmospheric pressure were recorded at different delay and gate width times. Because of the transient nature of the laser induced plasmas, the atomic and ionic populations present in the plasma rapidly evolved with time and position. At the moment of plasma formation, the spectrum was dominated by ionic emission and the plasma emission contributed to both collision of electrons with ions and atoms and recombination of electrons with ions.These emitted lines were broadened by Stark effect because of the high electron density. Subsequently, neutral atom emission lines dominate and emission due to the radiative decay of excited species are dominant. As a consequence, the emitted spectra vary with observation time after the impact of the laser beam. Fig. 3 Emission spectra of rock sample with different gate width times. Fig. 4 Calibration curves for elements in standard rock samples.Table 1 Major element composition of standard rock samples (% m/m) Sample Element KB-1 KD-1 KG-1 KG-2 KGB-1 KT-1 SiO2 48.42 57.89 76.09 74.91 55.97 62.35 Al2O3 15.68 16.68 12.67 13.39 17.04 18.35 Fe2O3 11.20 7.65 1.18 1.08 8.55 3.86 MgO 9.29 3.32 0.10 0.21 4.40 0.31 CaO 8.46 6.51 0.71 1.13 7.59 2.71 Na2O 3.52 3.57 3.62 3.77 2.90 6.42 K2O 1.54 2.38 4.79 4.58 1.99 4.33 MnO 0.16 0.14 0.05 0.06 0.13 0.24 P2O5 0.33 0.32 0.03 0.03 0.25 0.09 TiO2 1.65 0.93 0.10 0.14 0.95 0.51 Ba* 225 556 130 572 513 1210 Cu* 54.5 36.6 5.26 4.22 45.1 3.09 Sr* 518 444 41.4 155 429 718 Pb* — 12.5 25.5 24.2 9.5 6.9 * ppm.Table 2 Detection limit for each element Ba Cu Fe Mn Si Sr Wavelength/nm 493.4 510.55 404.73 403.07 390.55 407.77 Detection limit (ppm) 100 5 50 10 2000 50 1224 Analyst, November 1997, Vol. 122Fig. 2 shows the temporal evolution of the plasma in the rock samples and the variation in the emission signal intensity with delay time.In order to obtain a good S/N ratio and emission intensity, an appropriate choice of time delay is required for the measurement of emission lines. In this experiment emission signals were evaluated with a Boxcar integrator and an average of three plasma emission signals was taken. The delay time was varied from 1 to 10 ms to study the emission signal variation. The emission signal was most intense after a 1 ms delay time but the background signal was also very high.To obtain a maximum S/N ratio, a 3 ms delay time was chosen. The emitted signal was also dependent on the sample collection time. Fig. 3 shows the temporal evolution of plasma with different gate width times. In order to discriminate against continuum and to restrict the observed species to neutral atoms, a 3 ms delay time and a 9 ms gate width time was chosen. Experimental Condition for the Measurement In the experiment the whole plasma image was collected in the entrance slit to obtain the maximum signal intensity, avoid plasma inhomogeneity and irregular distribution of atoms in the plasma. The sample holder was tilted at 60° to the monochromator entrance slit for the collection of plasma emission without hindrance of sample movement.To reduce statistical errors, measurements were performed by averaging the emission signals of 3 shots. Quantitative Analysis Sample matrix problems are significant in LIPS and ablated elements vary with different sample matrices.If standards have the same matrix, the ablated elements are the same in the samples. In this experiment, calibration curves, were obtained for pelletized standard rock samples of a similar matrix (to the samples) but the actual significance and the physical and chemical properties of the samples remained uncertain. The major composition of the standard rock samples and the content of determined elements are shown in Table 1. Fig. 4 Fig. 5 Element distribution patterns for Ba, Pb, Sr and Fe in a polished granite rock section. Fig. 6 Element distribution patterns for Ba, Pb and Sr in a sliced rock section. Analyst, November 1997, Vol. 122 1225shows the calibration curves obtained for each analyzed element. To obtain a reliable calibration result, each standard sample was analyzed fifty times. The emission lines were selected by analyzing pure (99.99%, Johnson Matthey, Royston, UK) metal and metal oxide samples.The selected lines in this analysis were free from interferences and were well isolated. The calibration curves for each element, Ba (493.4 nm), Cu (510.55 nm), Mn (403.07 nm) and Sr (407.77 nm), are approximately linear up to 0.1%. The Fe (404.73 nm) and Si (390.55 nm) lines are also straight up to 10 and 30%, respectively. The curves for Pb (405.78 nm) were not linear because the Pb content of the standard was not a certified value but an averaged value. Therefore, Pb content was only obtained semi-quantitatively.The RSD of 50 consecutive measurements of Cu, Fe and Si was 5% for the selected analysis line and Ba, Mn, Sr was 7%. The limit of detection is a function of the element studied. In this experiment, the limit of detection for each element is given in Table 2. Elemental Mapping Analysis Ore veins within existing samples were selected to identify the different compositions of ore and surface element distribution. The analyzed area was 50 3 50 mm and element distribution differences were represented by color grading.The analytical results for different kinds of samples are shown in Figs. 5–8. In these figures, the upper first line represents the color scale. A granite sample is shown in Fig. 5, in which a lode was crossed during the analysis. This region is rich in Pb and Sr but the Ba content is low. Iron does not show differences and is nearly uniformly distributed across the sample. Another granite sample is shown in Fig. 6. In this sample, a dark spot was also analysed. A thin section of this sample was prepared using a diamond cutter and Ba, Pb and Sr were analyzed. In this sample, Ba and Pb showed inhomogeneous distribution but Sr did not show any marked concentration difference across the sample. Two kinds of Mn nodule were also analyzed and the results are Fig. 7 Element distribution patterns for Cu, Fe, Mn and Sr in a polished Mn nodule. Fig. 8 Element distribution patterns of Cu, Fe, Mn and Si in a polished Mn nodule section. 1226 Analyst, November 1997, Vol. 122shown in Figs. 7 and 8. In the nodule shown in Fig. 7, there was no significant element content difference for Cu, Fe, Mn and Sr. In the nodule shown in Fig. 8, a white lode exists. In this area, the content of Fe and Si were not different across the sample but Cu and Mn were differently distributed in the lode and main body of the sample. Conclusion Laser induced plasma spectroscopy was applied as a point analysis technique for element distribution mapping of polished rock samples.This technique can be utilized for highly sensitive and relatively rapid quantitative analysis of rock samples. The mapping technique can be applied to various fields and has a potential advantage over electron probe microanalysis (EPMA). By using LIPS, a large analysis area is possible and the sensitivity is greater than EPMA. In spite of its merits, however, the LIPS method still has calibration problems.For point analysis, the sample and standard matrices were not exactly matched. Therefore, the analytical result of the point analysis is semi-quantitative. References 1 Wisbrun, R., Schechter, I., Schroder, H., and Kompa, K. L., Anal. Chem., 1994, 66, 2964. 2 Sabsabi, M., and Cielo, P., Appl. Spectrosc., 1995, 49, 499. 3 Russo, R. E., Appl. Spectrosc., 1995, 49, 14A. 4 Kurniawan, H., Nakajima, S., Batubara, J. E., Marpaung, M., Okamoto, M., and Kagawa, K., Appl.Spectrosc., 1995, 49, 1067. 5 Anderson, D. R., McLeod, C. W., English, T., and Smith, A. T., Appl. Spectrosc., 1995, 49, 691. 6 Poulain, D. E., and Alexander, D. R., Appl. Spectrosc., 1995, 49, 569. 7 Nakamura, S., Ito, Y., Sone, K., Hiraga, H., and Kaneko, K., Anal. Chem., 1996, 68, 2981. 8 Lichte, F. E., Anal. Chem., 1995, 67, 2479. 9 Allen, A., Leach, J. J., Pang, H.-M., and Houk, R. S., J. Anal. At. Spectrom., 1997, 12, 171. 10 Fernandez, A., Mao, X. L., Chan, W.T., Shannon, M. A., and Russo, R. E., Anal. Chem., 1995, 67, 2444. 11 Mao, X. L., Shannon, M. A., Fernandez, A. J., and Russo, R. E., Appl. Spectrosc., 1995, 49, 1054. 12 Bulatov, V., Xu, L., and Schechter, I., Anal. Chem., 1996, 68, 2966. 13 Gunter, D., Cousin, H., Magyar, B., and Leopold, I., J. Anal. At. Spectrom.,, 1997, 12, 165. 14 Chaleard, C., Mauchien, P., Andre, N., Uebbing, J., Lacour, J. L., and Geertsen, C., J. Anal. At. Spectrom., 1997, 12, 183. 15 Wisbrun, R., Niessner, R., and Schroder, H., Anal.Meas. Instrum., 1993, 1, 17. Paper 7/04782G Received July 7, 1997 Accepted October 17, 1997 Analyst, November 1997, Vol. 122 1227 Application of Laser Induced Plasma Spectroscopy to the Analysis of Rock Samples† Yoon Yeol Yoon*a, Tae Sam Kima, Kang Sup Chunga, Kil Yong Leea and Gae Ho Leeb a Analytical Research Division, Korea Institute of Geology, Mining and Materials (KIGAM), 30 Kajung-dong, Yusong-ku, Taejeon, 305-350, Korea b Chemistry Department, Chungnam National University, Gung-dong, Yusong-ku, Taejeon, 305-764, Korea Laser induced plasma spectroscopy has been applied to the analysis of element distribution mapping of polished rock sections.The plasma was generated by focusing a frequency-doubled second harmonic 532 nm Nd :YAG laser on the target under atmospheric conditions. The experimental parameters, such as laser energy, atomic emission line and time profile of the plasma spectrum, were characterized to obtain optimum experimental conditions and estimate the element composition of the target surface.For the element mapping of samples, an X–Y stage was used to move the sample and an element image of 50 3 50 mm could be made in 30 min. Using this technique, the element concentration distribution of Ba, Cu, Fe, Mn, Pb, Si and Sr in polished rock sections were obtained. Quantitative analysis was achieved by analyzing standard rock samples. Calibrated concentration versus plasma intensity was used for the color grading for the mapping of element concentration distribution.Keywords: Laser ablation; elemental mapping; geological sample Laser induced plasma spectroscopy (LIPS) is a well established instrumental analysis technique in analytical chemistry.1–7 In LIPS, a short, pulsed, high power laser beam is focused onto the solid target and a plasma spot is created. In the laser induced plasma, a portion of the material instantaneously explodes into vapor.Ablated elements are excited in the plasma, and emit spectral excitation lines. These lines are recorded and analyzed for the identification and quantification of elements. Recently, LIPS has been applied to many kinds of samples such as liquid samples,6,7 direct analysis of solid samples1–4 and depth profiling of metal film.5 Laser ablation coupled with ICPAES and ICP-MS8,9 for the direct analysis of solid samples has also been reported. Plasma characterization experiments have been performed10–12 to discover the optimum ablation conditions.LIPS has various advantages over more conventional methods of atomic emission spectroscopy. The most relevant characteristics are direct analysis of solid samples (conducting and non-conducting materials), no sample preparation, a small sample requirement, the possibility on-line analysis, localized microanalysis, a single step of vaporization and excitation and simultaneous multielement analysis.However, in spite of the above mentioned advantages, LIPS has several drawbacks for the analysis of inhomogeneous samples or when localized microanalysis is desired. For the quantification of samples, the standard used has to show similar physical and chemical matrix effects and provide similar ablation behavior to the sample. To overcome the problems with laser ablation analysis and obtain precise calibration results, matrix effects have been studied.13,14 These studies were based on the intensity of the emission lines as a function of the vaporized mass and the plasma excitation temperature.The analytical procedure commonly used for the correction of matrix effects is based on normalization of the analyte signal by a reference signal.15 In the present work, LIPS was applied as a point analysis technique for obtaining the element distribution mapping of polished rock sections. The optimum analytical conditions have been studied. This mapping technique can be utilized in various fields; interpretation of ore formation history in mineralogy, investigation of segregation phenomena in metallurgy and examination of conditions in an electric circuit board, etc.† Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Fig. 1 Schematic diagram of the experimental set-up for measuring emission intensity from LIPS. Fig. 2 Emission spectra of rock sample with different delay times.Analyst, November 1997, Vol. 122 (1223–1227) 1223Experimental The experimental set-up is shown in Fig. 1. A frequency doubled Nd : YAG Q-switched laser (Spectron SL802G, FWHM 10 ns, Rugby, UK) emitting at 532 nm was used as the ablation source for rock samples. The laser induced plasma is generated by focusing the laser on the sample surface (60° relative to the normal incidence) using a planoconvex 3 cm focal length lens.The pulse repetition rate was 5 Hz and the laser energy was 30 mJ. The crater resulting from the laser pulse was about 10 mm diameter. The analysed sample was positioned about 10 mm above the focal point. This configuration was found to produce a stable plasma and minimize air break-down. Emission from the plasma was collected by a set of combination lenses (Minolta, y = 49 mm, f = 50 mm) for transmitting a planar image of the plasma plume into the collecting planoconvex quartz lens (f = 350 mm) and imaged on to the entrance slit (20 mm slit width) of a spectrometer (SPEX 1404, 0.85 m).The He–Ne laser was used as a pointer to obtain an exact point analysis. The spectra were recorded using a Boxcar integrator (Standford Research, SR250, Palo Alto, CA, USA) coupled with a 1024 pixel photodiode OMA III (spectral range = 20 nm, EG&G Princeton Applied Research, Princeton, NJ, USA; Model 1420) system and a photomultiplier tube (PMT). The collection signal was triggered by a laser pulse signal. To synchronize the laser pulse and emitting signal, the laser pulse signal was delayed for 4 ms using a digital delay generator (Standford Research, DG 535).The sample holder was designed to move in the X and Y direction by a stepping motor and the movement was controlled with a personal computer. Manganese nodule and granite rock samples were analyzed using this set-up. The element concentration was obtained by analyzing a standard granite rock sample.The standard samples were prepared at our Institute a few years ago. To overcome matrix matching properties, standard samples were pelletized at a pressure of 1 MPa. Results and Discussion Temporal Characteristics of the Plasma Emission Emission spectra of rock in air at atmospheric pressure were recorded at different delay and gate width times. Because of the transient nature of the laser induced plasmas, the atomic and ionic populations present in the plasma rapidly evolved with time and position.At the moment of plasma formation, the spectrum was dominated by ionic emission and the plasma emission contributed to both collision of electrons with ions and atoms and recombination of electrons with ions. These emitted lines were broadened by Stark effect because of the high electron density. Subsequently, neutral atom emission lines dominate and emission due to the radiative decay of excited species are dominant. As a consequence, the emitted spectra vary with observation time after the impact of the laser beam.Fig. 3 Emission spectra of rock sample with different gate width times. Fig. 4 Calibration curves for elements in standard rock samples. Table 1 Major element composition of standard rock samples (% m/m) Sample Element KB-1 KD-1 KG-1 KG-2 KGB-1 KT-1 SiO2 48.42 57.89 76.09 74.91 55.97 62.35 Al2O3 15.68 16.68 12.67 13.39 17.04 18.35 Fe2O3 11.20 7.65 1.18 1.08 8.55 3.86 MgO 9.29 3.32 0.10 0.21 4.40 0.31 CaO 8.46 6.51 0.71 1.13 7.59 2.71 Na2O 3.52 3.57 3.62 3.77 2.90 6.42 K2O 1.54 2.38 4.79 4.58 1.99 4.33 MnO 0.16 0.14 0.05 0.06 0.13 0.24 P2O5 0.33 0.32 0.03 0.03 0.25 0.09 TiO2 1.65 0.93 0.10 0.14 0.95 0.51 Ba* 225 556 130 572 513 1210 Cu* 54.5 36.6 5.26 4.22 45.1 3.09 Sr* 518 444 41.4 155 429 718 Pb* — 12.5 25.5 24.2 9.5 6.9 * ppm.Table 2 Detection limit for each element Ba Cu Fe Mn Si Sr Wavelength/nm 493.4 510.55 404.73 403.07 390.55 407.77 Detection limit (ppm) 100 5 50 10 2000 50 1224 Analyst, November 1997, Vol. 122Fig. 2 shows the temporal evolution of the plasma in the rock samples and the variation in the emission signal intensity with delay time. In order to obtain a good S/N ratio and emission intensity, an appropriate choice of time delay is required for the measurement of emission lines. In this experiment emission signals were evaluated with a Boxcar integrator and an average of three plasma emission signals was taken.The delay time was varied from 1 to 10 ms to study the emission signal variation. The emission signal was most intense after a 1 ms delay time but the background signal was also very high. To obtain a maximum S/N ratio, a 3 ms delay time was chosen. The emitted signal was also dependent on the sample collection time. Fig. 3 shows the temporal evolution of plasma with different gate width times. In order to discriminate against continuum and to restrict the observed species to neutral atoms, a 3 ms delay time and a 9 ms gate width time was chosen.Experimental Condition for the Measurement In the experiment the whole plasma image was collected in the entrance slit to obtain the maximum signal intensity, avoid plasma inhomogeneity and irregular distribution of atoms in the plasma. The sample holder was tilted at 60° to the monochromator entrance slit for the collection of plasma emission without hindrance of sample movement. To reduce statistical errors, measurements were performed by averaging the emission signals of 3 shots.Quantitative Analysis Sample matrix problems are significant in LIPS and ablated elements vary with different sample matrices. If standards have the same matrix, the ablated elements are the same in the samples. In this experiment, calibration curves, were obtained for pelletized standard rock samples of a similar matrix (to the samples) but the actual significance and the physical and chemical properties of the samples remained uncertain.The major composition of the standard rock samples and the content of determined elements are shown in Table 1. Fig. 4 Fig. 5 Element distribution patterns for Ba, Pb, Sr and Fe in a polished granite rock section. Fig. 6 Element distribution patterns for Ba, Pb and Sr in a sliced rock section. Analyst, November 1997, Vol. 122 1225shows the calibration curves obtained for each analyzed element. To obtain a reliable calibration result, each standard sample was analyzed fifty times.The emission lines were selected by analyzing pure (99.99%, Johnson Matthey, Royston, UK) metal and metal oxide samples. The selected lines in this analysis were free from interferences and were well isolated. The calibration curves for each element, Ba (493.4 nm), Cu (510.55 nm), Mn (403.07 nm) and Sr (407.77 nm), are approximately linear up to 0.1%. The Fe (404.73 nm) and Si (390.55 nm) lines are also straight up to 10 and 30%, respectively.The curves for Pb (405.78 nm) were not linear because the Pb content of the standard was not a certified value but an averaged value. Therefore, Pb content was only obtained semi-quantitatively. The RSD of 50 consecutive measurements of Cu, Fe and Si was 5% for the selected analysis line and Ba, Mn, Sr was 7%. The limit of detection is a function of the element studied. In this experiment, the limit of detection for each element is given in Table 2.Elemental Mapping Analysis Ore veins within existing samples were selected to identify the different compositions of ore and surface element distribution. The analyzed area was 50 3 50 mm and element distribution differences were represented by color grading. The analytical results for different kinds of samples are shown in Figs. 5–8. In these figures, the upper first line represents the color scale. A granite sample is shown in Fig. 5, in which a lode was crossed during the analysis.This region is rich in Pb and Sr but the Ba content is low. Iron does not show differences and is nearly uniformly distributed across the sample. Another granite sample is shown in Fig. 6. In this sample, a dark spot was also analysed. A thin section of this sample was prepared using a diamond cutter and Ba, Pb and Sr were analyzed. In this sample, Ba and Pb showed inhomogeneous distribution but Sr did not show any marked concentration difference across the sample.Two kinds of Mn nodule were also analyzed and the results are Fig. 7 Element distribution patterns for Cu, Fe, Mn and Sr in a polished Mn nodule. Fig. 8 Element distribution patterns of Cu, Fe, Mn and Si in a polished Mn nodule section. 1226 Analyst, November 1997, Vol. 122shown in Figs. 7 and 8. In the nodule shown in Fig. 7, there was no significant element content difference for Cu, Fe, Mn and Sr. In the nodule shown in Fig. 8, a white lode exists. In this area, the content of Fe and Si were not different across the sample but Cu and Mn were differently distributed in the lode and main body of the sample.Conclusion Laser induced plasma spectroscopy was applied as a point analysis technique for element distribution mapping of polished rock samples. This technique can be utilized for highly sensitive and relatively rapid quantitative analysis of rock samples. The mapping technique can be applied to various fields and has a potential advantage over electron probe microanalysis (EPMA). By using LIPS, a large analysis area is possible and the sensitivity is greater than EPMA. In spite of its merits, however, the LIPS method still has calibration problems. For point analysis, the sample and standard matrices were not exactly matched. Therefore, the analytical result of the point analysis is semi-quantitative. References 1 Wisbrun, R., Schechter, I., Schroder, H., and Kompa, K. L., Anal. Chem., 1994, 66, 2964. 2 Sabsabi, M., and Cielo, P., Appl. Spectrosc., 1995, 49, 499. 3 Russo, R. E., Appl. Spectrosc., 1995, 49, 14A. 4 Kurniawan, H., Nakajima, S., Batubara, J. E., Marpaung, M., Okamoto, M., and Kagawa, K., Appl. Spectrosc., 1995, 49, 1067. 5 Anderson, D. R., McLeod, C. W., English, T., and Smith, A. T., Appl. Spectrosc., 1995, 49, 691. 6 Poulain, D. E., and Alexander, D. R., Appl. Spectrosc., 1995, 49, 569. 7 Nakamura, S., Ito, Y., Sone, K., Hiraga, H., and Kaneko, K., Anal. Chem., 1996, 68, 2981. 8 Lichte, F. E., Anal. Chem., 1995, 67, 2479. 9 Allen, A., Leach, J. J., Pang, H.-M., and Houk, R. S., J. Anal. At. Spectrom., 1997, 12, 171. 10 Fernandez, A., Mao, X. L., Chan, W. T., Shannon, M. A., and Russo, R. E., Anal. Chem., 1995, 67, 2444. 11 Mao, X. L., Shannon, M. A., Fernandez, A. J., and Russo, R. E., Appl. Spectrosc., 1995, 49, 1054. 12 Bulatov, V., Xu, L., and Schechter, I., Anal. Chem., 1996, 68, 2966. 13 Gunter, D., Cousin, H., Magyar, B., and Leopold, I., J. Anal. At. Spectrom.,, 1997, 12, 165. 14 Chaleard, C., Mauchien, P., Andre, N., Uebbing, J., Lacour, J. L., and Geertsen, C., J. Anal. At. Spectrom., 1997, 12, 183. 15 Wisbrun, R., Niessner, R., and Schroder, H., Anal. Meas. Instrum., 1993, 1, 17. Paper 7/04782G Received July 7, 1997 Accepted October 17, 1997 Analyst, November 1997, Vol. 122 1227
ISSN:0003-2654
DOI:10.1039/a704782g
出版商:RSC
年代:1997
数据来源: RSC
|
8. |
Analysis of Metals in Condensates and Naphtha by Inductively Coupled Plasma Mass Spectrometry |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1229-1234
S. D. Olsen,
Preview
|
|
摘要:
Analysis of Metals in Condensates and Naphtha by Inductively Coupled Plasma Mass Spectrometry S. D. Olsena, S. Westerlunda and R. G. Visserb a RF - Rogaland Research, P.O. Box 2503, Ullandhaug, N-4004 Stavanger, Norway b Institute for Interlaboratory Studies, P.O. Box 8204, 3301 CE Dordrecht, The Netherlands Condensates and naphtha are petroleum samples with largely gasoline-range components (C5–C10). Metal organic complexes are source inherited components of oils which are associated with the polar components and asphaltenes of oils.Because of the very low levels of biomarkers such as hopanes and steranes in condensates, they present a special correlation problem. Geochemically significant metals, such as V and Ni, can be measured in some condensates using conventional Meinhardt–Scott chamber inductively coupled plasma mass spectrometry (ICP-MS). In order to increase the sensitivity and the range of condensates which can be analysed, the Cetac U-6000 AT Ultrasonic nebuliser–desolvation system was tested.The Cetac system improved the sensitivity for V and Ni by a factor of about 40. Valuable information for relating these difficult samples to each other and to heavier oils in a basin can thus be supplied to exploration geochemists. The influence of organic matrix on the analytical signal was investigated. Matrix effects are more pronounced for the Cetac system than for the conventional ICP-MS system. For both systems, the use of an In internal standard was found to be necessary for compensating for differences in nebulisation and combustion behaviour caused by variations in the nature of the sample.The use of ICP-MS for the analysis of toxic metals such as Hg and Pb in condensates and naphthas is also discussed. The results are compared with those obtained using electrothermal atomic absorption spectrometry and cold vapour atomic absorption spectrometry in an interlaboratory study of naphthas.Conventional Meinhardt–Scott chamber ICP-MS provided good accuracy and precision of analysis compared with the other techniques during this study. Volatile species such as Et4Pb and Me2Hg were lost in the desolvation unit when using the Cetac system. Volatilisation effects were not observed when using the conventional Meinhardt–Scott chamber ICP-MS system. Keywords: Trace metals, naphtha; condensate oil; ICP-MS; interlaboratory study ‘Fingerprinting’ of oils is important for delineating families of oils in a basin during exploration.Light oils and condensates are examples of a special correlation problem. This is because many condensates have largely gasoline-range components. The higher molecular mass components such as biomarkers and metal porphyrins are therefore low or absent in many condensate samples.1 It is therefore geochemically of great importance to obtain reliable values of source inherited metals such as V and Ni in order to characterise these difficult samples.A second reason for the importance of trace metal analysis in these samples is that some condensates contain traces of toxic metals such as Hg, Pb and As. In the petroleum industry, the presence of Hg has attracted much attention because of the corrosion problems associated with this metal. The aluminium heat exchanger in the Skikda facility, Algeria, was completely destroyed due to small amounts (1 ng to 100 mg m23) of Hg in natural gas. Similar problems have been reported in Groningen, Holland, and from fields in the North Sea (Bingham2).The presence of organic arsenic compounds in petroleum and natural gas has been measured by Irgolic and Puri.3 Concern for the release of toxic metals into the environment during refining and combustion of fuel produced from condensates, naphtha and gas has led to monitoring of the contents of metals such as Hg, Pb and As.4,5 The performance of conventional nebulisation ICP-MS will be described and compared with that obtained when using an accessory that improves the detection power of ICP-MS.Because metals occur at such low concentrations in these samples, the Cetac 6000 ultrasonic nebuliser–desolvator system was tested in order to improve the sensitivities of the analyses and thus the range of condensates for which geochemical data can be provided. Experimental Sample Preparation, Solvents and Standards Samples (2.0 g) were diluted 1 + 6.5 in xylene.A Sartorius MC 1 balance was used for weighing the condensates, naphtha and xylene. Indium internal standard (in xylene) was added so that the diluted samples (15 g) and standards contained 100 ng g21 In. Polyethylene scintillation tubes (22 ml, from Zinsser Analytic Gmbh, Frankfurt, Germany) were used for the dilutions. Samples were vigorously shaken to dissolve the oil in xylene in the proportions mentioned above. Experience has shown that these ‘one-time use’ scintillation bottles did not contribute any trace element contamination and could be used without prior acid washing. The time between making up of samples and standards and their use in analysis was limited as much as possible.Samples were stored in a refrigerator if it was unavoidable to wait a few days before analysis. The operating companies that supplied the condensate samples shipped them as soon as they had been taken. Where a delay occurred, samples were kept in a refrigerator.These practices were done to minimise the effect of metals like Hg being deposited on the sample vessel. At present, sampling protocols for petroleum samples received are not within the control of our laboratory. Calibration standard solutions of metals were prepared from Conostan elemental standards (Conoco, Ponca City, OK, USA), i.e., As, Cd, Hg, Ni, Pb and V. These were diluted to give standards with concentrations between 10 and 100 ng g21. Each sample dilution was analysed using 3 runs.Blanks were analysed after every sample. In addition, the integrals which were printed out after each analysis, were monitored. Where high levels of metal showing memory effect were observed in the blank, standard practice was a 3 min rinse using 5% † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1229–1234) 1229suprapure HNO3 followed by a 3 min xylene rinse.After this rinse, the blank levels were checked with a fresh xylene solution before proceeding with the next analysis. Inductively Coupled Plasma Mass Spectrometer and Cetac U-6000AT Ultrasonic Nebuliser The VG PlasmaQuad PQ2+ ICP-MS system were the same as previously described.6,7 The instrumental conditions for operating the ICP-MS are summarised in Table 1. Argon (99.998%) was used as the plasma gas. Oxygen (99.998%) was used to reduce carbon build-up on the outer Pt cone inlet to the mass spectrometer.The Cetac U-6000AT consists of an ultrasonic nebuliser with a special membrane desolvator for removing the solvent vapour load. It results in more efficient nebulisation as well as concentrating analyte molecules in the desolvator. A sweep gas flow of 0.3 l min21 and sample uptake rate of 1.5 ml min21 was used. An ultrasonic heater temperature of 140 °C was used whereas that of the cooler was 3 °C and the desolvator temperature was set to 160 °C.An oxygen flow of 0.12 l min21 and a nebuliser gas flow of 0.75 l min21 was used. Results and Discussion Performance of the Meinhardt–Scott Chamber System During the Analysis of Condensates and Naphtha The nature of petroleum samples analysed in our laboratory varies widely. Fig. 1 illustrates the variation in samples from viscous, heavy oils to light oils, condensates of varying shades of yellow depending on the content of polar molecules.Results for the analysis of these oils are shown in Table 2. The yellow condensate in this table produced an aerosol that did not burn as well as that of the brown light oil or that of the very light, colourless condensate. The GC traces of these samples are compared in Fig. 2. From this figure it can be seen that the high level of butane in the yellow condensate appears to be the reason why this sample has a much more noisy ICP-MS signal as seen from the standard deviation of the metal results in Table 2.These very light, volatile organic compounds cause an increase in the reflected power and can lead to the plasma being extinguished. Addition of more oxygen only increases this reflected power. In order to analyse these samples, a larger than normal dilution factor is used, the dilution making the already low levels of metal even more difficult to analyse using the Meinhardt–Scott chamber system. The Cetac U-6000AT+ Ultrasonic nebuliser system was found to be much more tolerant of the volatile organic components.However, as discussed below, severe loss of volatile metal species due to selective volatilisation occurs with the Cetac system. Table 2 illustrates the elements analysed in condensates for geochemical (V, Ni) as well as environmental reasons (As, Hg, Pb). In order to give an indication of contamination due to drilling mud, Ba is monitored during each analysis. Owing to an interference by a carbonaceous species, the Ni values obtained using the 60Ni isotope are too high.A Ni concentration of 90 ng g21 was obtained for the brown oil using 60Ni, whereas that obtained using 58Ni was 15 ng g21. This interference becomes noticeable only when low concentrations of Ni are analysed. In aqueous solution, on the other hand, the 60Ni isotope appears to have less interference than the 58Ni. The contents of V and Ni in condensates are often close to or below the detection limits of the Meinhardt–Scott system.Samples in Table 2 were diluted by a factor of 7. As the V/(V+Ni) ratio is geochemically so important, it is desirable to have the 58Ni value for the colourless sample in Table 2 substantially higher than the detection limit of 0.2 ppb than what is measured (3.9/7 = 0.5 ppb) for this sample. The calibration curves for some of the above elements are shown in Fig. 3. This figure illustrates the different sensitivities of elements analysed by ICP-MS. A mono-isotopic element such as 51V has a steeper calibration curve than elements such as Ni and Pb which have the signal distributed amongst four isotopes.Higher background due to molecular ions offsets this advantage in the mass range of these elements. Mono-isotopic 75As illustrates an element which is less sensitive because of a higher ionisation potential. Mercury and As ion production is Table 1 ICP-MS (VG PlasmaQuad 2+) operating conditions for xylene diluted oils ICP-system Plasma RF power 1.5 kW Reflected power < 10 W Torch VG/Fassel Torch injector id 1.5 mm Coolant argon flow 17 l min21 Auxiliary argon flow 0.8 l min21 Nebulizer argon flow 0.8 l min21 Nebulizer oxygen flow 0.12 l min21 Double pass spray chamber temperature 215 °C Peristaltic pump Gilson minipuls 3 Iversinic (fluoroelastomer) tubing 0.3 mm Sample uptake rate 0.5 ml min21 Nebulizer Meinhard TR-30-A3 Upgraded interface— Pt sampling cone orifice 1.0 mm Ni skimmer cone orifice 0.75 mm Resolution 0.7 m/z at 5% of peak height Maximum scan speed 2500 m/z s21 Fig. 1 Variation in nature of petroleum samples. Condensates and naphtha samples of Table 2 are shown. Left to right: viscous, heavy oil; brown light oil; yellow condensate; light condensate; naphtha; wax. Oil source rock in foreground. Table 2 Showing the range of metal concentrations determined in three condensates and a naphtha sample. Concentrations in ppb (ng g21) Condensates Metal Yellow condensate Brown light oil Light condensate Naphtha 51V 117 ± 24 183 ± 2 22 ± 1 71 ±7 58Ni 19 ± 7 15 ± 0.1 3.9 ± 0.1 40 ± 2 66Zn 146 ± 37 110 ± 3 40 ± 2 91 ± 12 75As 46 ± 9 < 1 < 1 40 ± 9 111Cd < 2 < 2 < 2 25 ± 1 138Ba 0.4 ± 0.15 0.05 ± 0.02 0.05 ± 0.04 42 ± 1 200Hg 47 ± 20 4 ± 1.8 7 ± 4 37 ± 6 208Pb < 0.3 3.9 ± 0.6 1.5 ± 0.2 36 ± 6 1230 Analyst, November 1997, Vol. 122less efficient and partly explains the lower sensitivities shown by the calibration curves of these elements.Performance of the Cetac U-6000AT+ Ultrasonic Nebuliser System A Cetac U-6000AT+ Ultrasonic nebuliser system was tested in order to evaluate the possible improvements that this system would bring to our analysis of condensates and naphthas. There is a considerable improvement in signal intensity obtained with this system compared to that of the conventional Meinhardt– Scott chamber system. The signal for 50 ng g21 V increases from about 2000 to 80 000 counts s21 and that for Pb from about 1000 to 40 000 counts s21.The calculated detection limits for these two systems are compared in Table 3. Generally, the intensity of the background also increases for the Cetac system and the improvement in detection limit is not as high as would be expected from the increased sample signal. It is interesting to note that the intensity of the blank or background on 51V Fig. 2 Gas chromatograms (GC–FID) showing the hydrocarbon composition of the samples in Table 2.(a) Brown light oil; (b) yellow condensate; (c) light condensate; and (d) naphtha. Fig. 3 Calibration curves of metals determined in condensates and naphtha illustrating the difference in sensitivity and specially the lower sensitivity and reproducibility of Hg. Analyst, November 1997, Vol. 122 1231decreased substantially using the Cetac system. This demonstrates the mechanism by which the desolvation removes solvent which, in the Meinhardt chamber system, has a larger relative contribution from the 40Ar12C+ , which itself has such an intense and broad peak that its wing enters the 51V channel.The Cetac system is of great advantage for the analysis of low levels of Ni and V. Unfortunately, the sensitivity advantage of the Cetac system is offset for the analysis of Hg and Pb as these metals often occur as volatile species in petroleum. Fig. 4 compares the spectra obtained with the two nebuliser systems when aspirating the Pb and Hg as two different species.On the left the spectra are shown when 25 ppb Hg as alkyl dithiocarbamate and 25 ppb Pb as alkyl aryl sulfonate complexes are aspirated into the plasma. The spectra in the right hand boxes were obtained when 25 ppb Hg as Me2Hg and 46 ppb Pb as Et4Pb were aspirated into the plasma. From Fig. 4(a), it can be seen that the light molecular species (Me2Hg and Et4Pb) are lost during the desolvation stage of the Cetac Ultrasonic–desolvator system. The spectra in the lower right hand boxes of Fig. 4 illustrate our findings that selective volatilisation is minimal for light molecular species when using the conventional Meinhardt nebulisation system. The conventional system is thus used for analysing toxic metals in petroleum samples. Memory effects of Hg are particularly troublesome for samples with low concentrations of this element. For the Meinhardt–Scott system with a detection limit is 0.13 ng g21, the blank value increased to 0.8 ng g21 following aspiration of the 53 ppb standard solution.When this was repeated on the Cetac system with a detection limit for Hg of 0.02 ng g21, it was difficult to get the blank value below 0.2 ng g21 following a xylene rinse after aspirating the 53 ppb Hg standard. Table 3 Comparison of the performance of the two nebulizer systems used Meinhardt–Scott Chamber Cetac Ultrasonic–Desolvator Metal Integrals blank/ counts s21 Detection limit Integrals blank/ counts s21 Detection limit Improvement 51V 10 043 0.39 5 988 0.009 43 58Ni 2 529 0.22 15 600 0.006 37 75As 35 0.2 380 nd* nd 111Cd 64 0.4 90 0.004 100 200Hg 11 0.13 370 0.02 6 208Pb 63 0.03 690 0.004 7 * nd = not determined Fig. 4 Spectra for Hg and Pb obtained with the (a) Meinhardt–Scott chamber system and (b) the Cetac Ultrasonic–desolvator system when analysing volatile and non-volatile species of these elements. 1232 Analyst, November 1997, Vol. 122Analysis of Trace Metals in Naphtha Samples It was found that the influence of naphtha on the plasma properties was very similar to that of many condensates; signal suppression and the plasma stability were similarly affected.In order to have some measure of the accuracy of our analyses of condensates, the round robin of the Institute for Interlaboratory Studies (IIS) was joined. From the IIS naphtha round robin of 1994 and from individual observations, it was found that the analysis of low levels of Pb or Hg in naphtha is very problematic.Fifteen laboratories in 14 different countries (US, Europe, Middle East and Singapore) participated in the 1997 round robin programme. 8 Our laboratory was the only one using ICP-MS. Other methods used were spectrometric (IP 224), cold vapour atomic absorption spectrometry (CVAAS) and electrothermal atomic absorption spectrometry (ETAAS). The five solutions supplied (9701 to 9705) had been spiked with Pb in the form of tetraethyllead (Et4Pb) and dithiocarbamate mercury.Only 8 of the 15 laboratories reported Hg results. Most participating laboratories used different pretreatment procedures before the samples were analysed. For ICP-MS analysis, no other pretreatment than dilution with xylene was made. The ICP-MS results for the five naphtha samples are compared with those of the round robin averages in Table 4. Samples 9702 and 9704 were duplicates. Fig. 5 shows the two sample Youden plots for these samples and illustrates the spread of analytical results.It can be seen that the results for Hg were scattered around the target value to a much greater extent than those for Pb. Because of the loss of the volatile Et4Pb, the Cetac result of for example 9705 was < 5 instead of the 55.7 ppb in Table 4 obtained with the Meinhardt–Scott chamber system. Table 4 Average round robin values (Av) for the determination of Hg and Pb in naphtha compared with our ICP-MS results. Concentrations given in ng g21 (ppb).The uncertainty given is the standard deviation Sample 9701 Sample 9702/4 Sample 9703 Sample 9705 Metal Av ICP-MS Av ICP-MS Av ICP-MS Av ICP-MS Hg 7.8 ± 4.2 8.8 ± 6.4 41.8 ± 7.6 35.5 ± 7.6 2.8 < 5 6.4 ± 4.3 < 5 Pb 17.4 ± 6.6 10.9 ± 1.2 31.8 ± 9.4 32.8 ± 5.7 8.9 ± 4.5 11.5 ± 0.5 57.6 ± 16.2 55.7 ± 7.2 Table 5 Comparison of the data obtained using the conventional Meinhardt–Scott ICP-MS system on the same standard in different matrices (315 ng g21 of V, Mo, Cd and Pb and 133 ng g21 of Co ) Xylene Naphtha Base oil Integrated signal Conc.Integrated signal Conc. Integrated signal Conc. Metal Blank Std* No In In Blank Std* No In In Blank Std* No In In 51V 3674 529 310 468 312 18 500 488 260 432 319 5988 453 070 400 314 59Co 46 179 480 132 133 58 162 350 120 134 63 151 800 112 133 98Mo 25 87 946 315 316 41 79 500 284 315 32 74 330 265 315 111Cd 23 20 557 313 315 46 18 800 283 315 38 17 600 269 319 208Pb 447 179 000 314 315 74 160 000 282 315 153 153 700 271 315 115In 300 263 274 273 254 105 * Standard = 315 ng g21 of V, Mo, Cd and Pb and 133 ng g21 of Co.Table 6 Comparison of data obtained using the Cetac system on the same standard in different matrices (108 ng g21 V, Mo, Cd and Pb and 67 ng g21 Co) Xylene Naphtha Base oil Integrated signal Conc. Integrated signal Conc. Integrated signal Conc. Metal Blank Std* No In In Blank Std* No In In Blank Std* No In In 51V 2534 994 340 108 109 2764 649 365 70 108 7215 336 092 37 127 59Co 46 530 000 67 67 42 364 000 46 69 426 164 850 21 73 98Mo 24 152 430 108 107 10 102 314 72 109 34 51 151 36 124 111Cd 21 14 465 107 108 21 9 946 73 112 9 4 696 34 120 208Pb 32 169 625 103 108 24 106 700 65 102 467 38 344 30 106 115In 177 787 117 590 514 052 0 0 * Standard = 108 ng g21 of V, Mo, Cd and Pb and 67 ng g21 of Co.Fig. 5 Two sample Youden plots of duplicate samples 9702 and 9704. Means of the results are presented by the dotted lines; the intersection being the target value.Continuous lines for Pb are the mean ± reproducibility. Analyst, November 1997, Vol. 122 1233Effect of Variation of the Sample Matrix Standards were made up in three different matrices; xylene, naphtha and oil, respectively. A naphtha sample with very low levels of metals as well as the NIST 1083 Base Oil were used for this purpose. Ten-fold dilutions of the standards in these different matrices were then analysed using both conventional Meinhardt–Scott ICP-MS and the Cetac system.The results in Table 5 show the integrated signals for the blanks and for a standards with 315 ng g21 V, Mo, Cd and Pb (S21) and 133 ng g21 Co, respectively. The signal for the In internal standard is also shown. The concentrations calculated using these signals with and without reference to the internal standard are also shown. It can be seen that the signal intensity of standard in naphtha was less than that of xylene. The signal suppression was highest in the base oil standards. From the concentrations calculated with reference to internal standard, it can be seen that this matrix effect has been corrected for.The effect of the matrix can be seen to be much more severe for the Cetac system in Table 6. The integrated signals for the blanks and a 108 ng g21 V, Mo, Cd and Pb and 67 ng g21 Co standards in the different matrices can be seen in Table 6. The integrated signal for the In internal standard is shown.Once again, the signal intensity decreases in the order: xylene > naphtha > oil. The base oil standard signal was 84% of that in the xylene for the conventional system. For the oil matrix measured by the Cetac system, the signal was suppressed to 30% of what it was in the xylene matrix. The concentrations calculated without reference to the internal standard can be seen to be seriously in error. Table 6 also illustrates that use of matrix matching is very important when using the Cetac system.However, when analysing light oils, one faces a dilemma. On the one hand, matrix does play a role in suppressing the signal. On the other hand, if one uses a base oil such as the NIST 1083 base, then one has a problem with accurately determining low levels of Ni and V seeing that this base oil contains for example 85 ng g21 Ni and 3 ng g21 V. The use of analytical grade xylene with appropriate internal standard addition has been found to provide the best compromise in this dilemma in our laboratory where the most important requirement is a correct V/(V+Ni) ratio.Conclusion The round robin has shown that conventional Meinhardt–Scott chamber ICP-MS performed well compared with other techniques such as ETAAS and CVAAS for the analysis of low levels of Pb and Hg in naphtha. ICP-MS has advantages over these techniques in that samples can be analysed directly without pretreatment and elements are analysed ‘simultaneously’. The fact that there is less sample handling and no need for addition of chemicals such as matrix modifiers and hydrides means that ICP-MS analyses are more rapid to carry out and less prone to problems such as contamination and influence of chemical species.ICP-MS provides the high sensitivity needed for the geochemical analysis of V and Ni in many condensates. However, reliable values of V and Ni cannot be determined in all light oils and condensates. The Cetac Ultrasonic nebuliser– desolvation system has been found to be essential for extending the range of condensates which can be analysed for fingerprinting or oil–oil correlation.By removing a large proportion of the solvent molecules, the problem of interference due to carbonacious species such as ArC+ on 51V is substantially decreased. The use of the Cetac Ultrasonic nebuliser–desolvation system is not recommended for the analysis of the environmentally important elements such as Pb and Hg in condensates and naphtha until the behaviour of the different species is better understood.Volatile species such as Et4Pb and Me2Hg were lost during the analysis with the Cetac system, but not with the conventional Meinhardt–Scott chamber ICP-MS system. Elf Petroleum funded and made possible work on metals in petroleum using ICP-MS as part of the application of new methods to a large multidisciplinary reservoir geochemistry study. A. K. Bjerklund of Løvland AS provided the Cetac U- 6000 AT Ultrasonic nebuliser–desolvation system on loan in order that the tests on petroleum samples could be carried out.West-Lab AS performed the GC analyses for this work. Permission given by the Institute for Interlaboratory Studies to report the naphtha results is gratefully acknowledged. References 1 Peters, K. E., and Moldowan, J. M., The Biomarker Guide, Prentice- Hall, Englewood Cliffs, New Jersey, 1993, p. 363. 2 Bingham, M. D., Field Detection and Implications of Mercury in Natural Gas, SPE Production Engineering, May 1990, pp. 120-124. 3 Irgolic, K. J., and Puri, B. K., in Metal Speciation in the Environment, ed. Broekaert, J. A. C., G�uçer, S., and Adams, F., NATO ASI Series G23 , Springer, Berlin, 1990, pp. 377-389. 4 Kucha, H., Slupczynski, K., and Prochaska, W., Nature, 1993, 363, 680. 5 Gijselman, P. B., in The First International Conference on Health, Safety and Environment in Oil and Gas Exploration and Production, The Hague, The Netherlands, 1991, pp. 123-130. 6 Filby, R.H., and Olsen, S. D., J. Radioanal. Nucl. Chem., 1994, 180, 285. 7 Olsen, S. D., Filby, R. H., Brekke, T., and Isaksen, G. H., Analyst, 1995, 120, 1379. 8 Visser, R.G., Results of Method Evaluating Interlaboratory Study Pb and Hg in Naphtha, Report No. ISS97N01, Institute for Interlaboratory Studies, Dordrecht, The Netherlands, 1997. Paper 7/04017B Received June 9, 1997 Accepted September 12, 1997 1234 Analyst, November 1997, Vol. 122 Analysis of Metals in Condensates and Naphtha by Inductively Coupled Plasma Mass Spectrometry S.D. Olsena, S. Westerlunda and R. G. Visserb a RF - Rogaland Research, P.O. Box 2503, Ullandhaug, N-4004 Stavanger, Norway b Institute for Interlaboratory Studies, P.O. Box 8204, 3301 CE Dordrecht, The Netherlands Condensates and naphtha are petroleum samples with largely gasoline-range components (C5–C10). Metal organic complexes are source inherited components of oils which are associated with the polar components and asphaltenes of oils.Because of the very low levels of biomarkers such as hopanes and steranes in condensates, they present a special correlation problem. Geochemically significant metals, such as V and Ni, can be measured in some condensates using conventional Meinhardt–Scott chamber inductively coupled plasma mass spectrometry (ICP-MS). In order to increase the sensitivity and the range of condensates which can be analysed, the Cetac U-6000 AT Ultrasonic nebuliser–desolvation system was tested.The Cetac system improved the sensitivity for V and Ni by a factor of about 40. Valuable information for relating these difficult samples to each other and to heavier oils in a basin can thus be supplied to exploration geochemists. The influence of organic matrix on the analytical signal was investigated. Matrix effects are more pronounced for the Cetac system than for the conventional ICP-MS system. For both systems, the use of an In internal standard was found to be necessary for compensating for differences in nebulisation and combustion behaviour caused by variations in the nature of the sample.The use of ICP-MS for the analysis of toxic metals such as Hg and Pb in condensates and naphthas is also discussed. The results are compared with those obtained using electrothermal atomic absorption spectrometry and cold vapour atomic absorption spectrometry in an interlaboratory study of naphthas.Conventional Meinhardt–Scott chamber ICP-MS provided good accuracy and precision of analysis compared with the other techniques during this study. Volatile species such as Et4Pb and Me2Hg were lost in the desolvation unit when using the Cetac system. Volatilisation effects were not observed when using the conventional Meinhardt–Scott chamber ICP-MS system. Keywords: Trace metals, naphtha; condensate oil; ICP-MS; interlaboratory study ‘Fingerprinting’ of oils is important for delineating families of oils in a basin during exploration. Light oils and condensates are examples of a special correlation problem. This is because many condensates have largely gasoline-range components.The higher molecular mass components such as biomarkers and metal porphyrins are therefore low or absent in many condensate samples.1 It is therefore geochemically of great importance to obtain reliable values of source inherited metals such as V and Ni in order to characterise these difficult samples.A second reason for the importance of trace metal analysis in these samples is that some condensates contain traces of toxic metals such as Hg, Pb and As. In the petroleum industry, the presence of Hg has attracted much attention because of the corrosion problems associated with this metal. The aluminium heat exchanger in the Skikda facility, Algeria, was completely destroyed due to small amounts (1 ng to 100 mg m23) of Hg in natural gas.Similar problems have been reported in Groningen, Holland, and from fields in the North Sea (Bingham2). The presence of organic arsenic compounds in petroleum and natural gas has been measured by Irgolic and Puri.3 Concern for the release of toxic metals into the environment during refining and combustion of fuel produced from condensates, naphtha and gas has led to monitoring of the contents of metals such as Hg, Pb and As.4,5 The performance of conventional nebulisation ICP-MS will be described and compared with that obtained when using an accessory that improves the detection power of ICP-MS.Because metals occur at such low concentrations in these samples, the Cetac 6000 ultrasonic nebuliser–desolvator system was tested in order to improve the sensitivities of the analyses and thus the range of condensates for which geochemical data can be provided. Experimental Sample Preparation, Solvents and Standards Samples (2.0 g) were diluted 1 + 6.5 in xylene.A Sartorius MC 1 balance was used for weighing the condensates, naphtha and xylene. Indium internal standard (in xylene) was added so that the diluted samples (15 g) and standards contained 100 ng g21 In. Polyethylene scintillation tubes (22 ml, from Zinsser Analytic Gmbh, Frankfurt, Germany) were used for the dilutions. Samples were vigorously shaken to dissolve the oil in xylene in the proportions mentioned above. Experience has shown that these ‘one-time use’ scintillation bottles did not contribute any trace element contamination and could be used without prior acid washing.The time between making up of samples and standards and their use in analysis was limited as much as possible. Samples were stored in a refrigerator if it was unavoidable to wait a few days before analysis. The operating companies that supplied the condensate samples shipped them as soon as they had been taken. Where a delay occurred, samples were kept in a refrigerator.These practices were done to minimise the effect of metals like Hg being deposited on the sample vessel. At present, sampling protocols for petroleum samples received are not within the control of our laboratory. Calibration standard solutions of metals were prepared from Conostan elemental standards (Conoco, Ponca City, OK, USA), i.e., As, Cd, Hg, Ni, Pb and V. These were diluted to give standards with concentrations between 10 and 100 ng g21. Each sample dilution was analysed using 3 runs.Blanks were analysed after every sample. In addition, the integrals which were printed out after each analysis, were monitored. Where high levels of metal showing memory effect were observed in the blank, standard practice was a 3 min rinse using 5% † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1229–1234) 1229suprapure HNO3 followed by a 3 min xylene rinse.After this rinse, the blank levels were checked with a fresh xylene solution before proceeding with the next analysis. Inductively Coupled Plasma Mass Spectrometer and Cetac U-6000AT Ultrasonic Nebuliser The VG PlasmaQuad PQ2+ ICP-MS system were the same as previously described.6,7 The instrumental conditions for operating the ICP-MS are summarised in Table 1. Argon (99.998%) was used as the plasma gas. Oxygen (99.998%) was usedo reduce carbon build-up on the outer Pt cone inlet to the mass spectrometer.The Cetac U-6000AT consists of an ultrasonic nebuliser with a special membrane desolvator for removing the solvent vapour load. It results in more efficient nebulisation as well as concentrating analyte molecules in the desolvator. A sweep gas flow of 0.3 l min21 and sample uptake rate of 1.5 ml min21 was used. An ultrasonic heater temperature of 140 °C was used whereas that of the cooler was 3 °C and the desolvator temperature was set to 160 °C.An oxygen flow of 0.12 l min21 and a nebuliser gas flow of 0.75 l min21 was used. Results and Discussion Performance of the Meinhardt–Scott Chamber System During the Analysis of Condensates and Naphtha The nature of petroleum samples analysed in our laboratory varies widely. Fig. 1 illustrates the variation in samples from viscous, heavy oils to light oils, condensates of varying shades of yellow depending on the content of polar molecules.Results for the analysis of these oils are shown in Table 2. The yellow condensate in this table produced an aerosol that did not burn as well as that of the brown light oil or that of the very light, colourless condensate. The GC traces of these samples are compared in Fig. 2. From this figure it can be seen that the high level of butane in the yellow condensate appears to be the reason why this sample has a much more noisy ICP-MS signal as seen from the standard deviation of the metal results in Table 2.These very light, volatile organic compounds cause an increase in the reflected power and can lead to the plasma being extinguished. Addition of more oxygen only increases this reflected power. In order to analyse these samples, a larger than normal dilution factor is used, the dilution making the already low levels of metal even more difficult to analyse using the Meinhardt–Scott chamber system. The Cetac U-6000AT+ Ultrasonic nebuliser system was found to be much more tolerant of the volatile organic components.However, as discussed below, severe loss of volatile metal species due to selective volatilisation occurs with the Cetac system. Table 2 illustrates the elements analysed in condensates for geochemical (V, Ni) as well as environmental reasons (As, Hg, Pb). In order to give an indication of contamination due to drilling mud, Ba is monitored during each analysis. Owing to an interference by a carbonaceous species, the Ni values obtained using the 60Ni isotope are too high.A Ni concentration of 90 ng g21 was obtained for the brown oil using 60Ni, whereas that obtained using 58Ni was 15 ng g21. This interference becomes noticeable only when low concentrations of Ni are analysed. In aqueous solution, on the other hand, the 60Ni isotope appears to have less interference than the 58Ni. The contents of V and Ni in condensates are often close to or below the detection limits of the Meinhardt–Scott system. Samples in Table 2 were diluted by a factor of 7.As the V/(V+Ni) ratio is geochemically so important, it is desirable to have the 58Ni value for the colourless sample in Table 2 substantially higher than the detection limit of 0.2 ppb than what is measured (3.9/7 = 0.5 ppb) for this sample. The calibration curves for some of the above elements are shown in Fig. 3. This figure illustrates the different sensitivities of elements analysed by ICP-MS.A mono-isotopic element such as 51V has a steeper calibration curve than elements such as Ni and Pb which have the signal distributed amongst four isotopes. Higher background due to molecular ions offsets this advantage in the mass range of these elements. Mono-isotopic 75As illustrates an element which is less sensitive because of a higher ionisation potential. Mercury and As ion production is Table 1 ICP-MS (VG PlasmaQuad 2+) operating conditions for xylene diluted oils ICP-system Plasma RF power 1.5 kW Reflected power < 10 W Torch VG/Fassel Torch injector id 1.5 mm Coolant argon flow 17 l min21 Auxiliary argon flow 0.8 l min21 Nebulizer argon flow 0.8 l min21 Nebulizer oxygen flow 0.12 l min21 Double pass spray chamber temperature 215 °C Peristaltic pump Gilson minipuls 3 Iversinic (fluoroelastomer) tubing 0.3 mm Sample uptake rate 0.5 ml min21 Nebulizer Meinhard TR-30-A3 Upgraded interface— Pt sampling cone orifice 1.0 mm Ni skimmer cone orifice 0.75 mm Resolution 0.7 m/z at 5% of peak height Maximum scan speed 2500 m/z s21 Fig. 1 Variation in nature of petroleum samples. Condensates and naphtha samples of Table 2 are shown. Left to right: viscous, heavy oil; brown light oil; yellow condensate; light condensate; naphtha; wax. Oil source rock in foreground. Table 2 Showing the range of metal concentrations determined in three condensates and a naphtha sample. Concentrations in ppb (ng g21) Condensates Metal Yellow condensate Brown light oil Light condensate Naphtha 51V 117 ± 24 183 ± 2 22 ± 1 71 ±7 58Ni 19 ± 7 15 ± 0.1 3.9 ± 0.1 40 ± 2 66Zn 146 ± 37 110 ± 3 40 ± 2 91 ± 12 75As 46 ± 9 < 1 < 1 40 ± 9 111Cd < 2 < 2 < 2 25 ± 1 138Ba 0.4 ± 0.15 0.05 ± 0.02 0.05 ± 0.04 42 ± 1 200Hg 47 ± 20 4 ± 1.8 7 ± 4 37 ± 6 208Pb < 0.3 3.9 ± 0.6 1.5 ± 0.2 36 ± 6 1230 Analyst, November 1997, Vol. 122less efficient and partly explains the lower sensitivities shown by the calibration curves of these elements.Performance of the Cetac U-6000AT+ Ultrasonic Nebuliser System A Cetac U-6000AT+ Ultrasonic nebuliser system was tested in order to evaluate the possible improvements that this system would bring to our analysis of condensates and naphthas. There is a considerable improvement in signal intensity obtained with this system compared to that of the conventional Meinhardt– Scott chamber system. The signal for 50 ng g21 V increases from about 2000 to 80 000 counts s21 and that for Pb from about 1000 to 40 000 counts s21.The calculated detection limits for these two systems are compared in Table 3. Generally, the intensity of the background also increases for the Cetac system and the improvement in detection limit is not as high as would be expected from the increased sample signal. It is interesting to note that the intensity of the blank or background on 51V Fig. 2 Gas chromatograms (GC–FID) showing the hydrocarbon composition of the samples in Table 2.(a) Brown light oil; (b) yellow condensate; (c) light condensate; and (d) naphtha. Fig. 3 Calibration curves of metals determined in condensates and naphtha illustrating the difference in sensitivity and specially the lower sensitivity and reproducibility of Hg. Analyst, November 1997, Vol. 122 1231decreased substantially using the Cetac system. This demonstrates the mechanism by which the desolvation removes solvent which, in the Meinhardt chamber system, has a larger relative contribution from the 40Ar12C+ , which itself has such an intense and broad peak that its wing enters the 51V channel. The Cetac system is of great advantage for the analysis of low levels of Ni and V.Unfortunately, the sensitivity advantage of the Cetac system is offset for the analysis of Hg and Pb as these metals often occur as volatile species in petroleum. Fig. 4 compares the spectra obtained with the two nebuliser systems when aspirating the Pb and Hg as two different species.On the left the spectra are shown when 25 ppb Hg as alkyl dithiocarbamate and 25 ppb Pb as alkyl aryl sulfonate complexes are aspirated into the plasma. The spectra in the right hand boxes were obtained when 25 ppb Hg as Me2Hg and 46 ppb Pb as Et4Pb were aspirated into the plasma. From Fig. 4(a), it can be seen that the light molecular species (Me2Hg and Et4Pb) are lost during the desolvation stage of the Cetac Ultrasonic–desolvator system.The spectra in the lower right hand boxes of Fig. 4 illustrate our findings that selective volatilisation is minimal for light molecular species when using the conventional Meinhardt nebulisation system. The conventional system is thus used for analysing toxic metals in petroleum samples. Memory effects of Hg are particularly troublesome for samples with low concentrations of this element. For the Meinhardt–Scott system with a detection limit is 0.13 ng g21, the blank value increased to 0.8 ng g21 following aspiration of the 53 ppb standard solution.When this was repeated on the Cetac system with a detection limit for Hg of 0.02 ng g21, it was difficult to get the blank value below 0.2 ng g21 following a xylene rinse after aspirating the 53 ppb Hg standard. Table 3 Comparison of the performance of the two nebulizer systems used Meinhardt–Scott Chamber Cetac Ultrasonic–Desolvator Metal Integrals blank/ counts s21 Detection limit Integrals blank/ counts s21 Detection limit Improvement 51V 10 043 0.39 5 988 0.009 43 58Ni 2 529 0.22 15 600 0.006 37 75As 35 0.2 380 nd* nd 111Cd 64 0.4 90 0.004 100 200Hg 11 0.13 370 0.02 6 208Pb 63 0.03 690 0.004 7 * nd = not determined Fig. 4 Spectra for Hg and Pb obtained with the (a) Meinhardt–Scott chamber system and (b) the Cetac Ultrasonic–desolvator system when analysing volatile and non-volatile species of these elements. 1232 Analyst, November 1997, Vol. 122Analysis of Trace Metals in Naphtha Samples It was found that the influence of naphtha on the plasma properties was very similar to that of many condensates; signal suppression and the plasma stability were similarly affected.In order to have some measure of the accuracy of our analyses of condensates, the round robin of the Institute for Interlaboratory Studies (IIS) was joined. From the IIS naphtha round robin of 1994 and from individual observations, it was found that the analysis of low levels of Pb or Hg in naphtha is very problematic.Fifteen laboratories in 14 different countries (US, Europe, Middle East and Singapore) participated in the 1997 round robin programme. 8 Our laboratory was the only one using ICP-MS. Other methods used were spectrometric (IP 224), cold vapour atomic absorption spectrometry (CVAAS) and electrothermal atomic absorption spectrometry (ETAAS). The five solutions supplied (9701 to 9705) had been spiked with Pb in the form of tetraethyllead (Et4Pb) and dithiocarbamate mercury.Only 8 of the 15 laboratories reported Hg results. Most participating laboratories used different pretreatment procedures before the samples were analysed. For ICP-MS analysis, no other pretreatment than dilution with xylene was made. The ICP-MS results for the five naphtha samples are compared with those of the round robin averages in Table 4. Samples 9702 and 9704 were duplicates. Fig. 5 shows the two sample Youden plots for these samples and illustrates the spread of analytical results.It can be seen that the results for Hg were scattered around the target value to a much greater extent than those for Pb. Because of the loss of the volatile Et4Pb, the Cetac result of for example 9705 was < 5 instead of the 55.7 ppb in Table 4 obtained with the Meinhardt–Scott chamber system. Table 4 Average round robin values (Av) for the determination of Hg and Pb in naphtha compared with our ICP-MS results. Concentrations given in ng g21 (ppb). The uncertainty given is the standard deviation Sample 9701 Sample 9702/4 Sample 9703 Sample 9705 Metal Av ICP-MS Av ICP-MS Av ICP-MS Av ICP-MS Hg 7.8 ± 4.2 8.8 ± 6.4 41.8 ± 7.6 35.5 ± 7.6 2.8 < 5 6.4 ± 4.3 < 5 Pb 17.4 ± 6.6 10.9 ± 1.2 31.8 ± 9.4 32.8 ± 5.7 8.9 ± 4.5 11.5 ± 0.5 57.6 ± 16.2 55.7 ± 7.2 Table 5 Comparison of the data obtained using the conventional Meinhardt–Scott ICP-MS system on the same standard in different matrices (315 ng g21 of V, Mo, Cd and Pb and 133 ng g21 of Co ) Xylene Naphtha Base oil Integrated signal Conc. Integrated signal Conc.Integrated signal Conc. Metal Blank Std* No In In Blank Std* No In In Blank Std* No In In 51V 3674 529 310 468 312 18 500 488 260 432 319 5988 453 070 400 314 59Co 46 179 480 132 133 58 162 350 120 134 63 151 800 112 133 98Mo 25 87 946 315 316 41 79 500 284 315 32 74 330 265 315 111Cd 23 20 557 313 315 46 18 800 283 315 38 17 600 269 319 208Pb 447 179 000 314 315 74 160 000 282 315 153 153 700 271 315 115In 300 263 274 273 254 105 * Standard = 315 ng g21 of V, Mo, Cd and Pb and 133 ng g21 of Co.Table 6 Comparison of data obtained using the Cetac system on the same standard in different matrices (108 ng g21 V, Mo, Cd and Pb and 67 ng g21 Co) Xylene Naphtha Base oil Integrated signal Conc. Integrated signal Conc. Integrated signal Conc. Metal Blank Std* No In In Blank Std* No In In Blank Std* No In In 51V 2534 994 340 108 109 2764 649 365 70 108 7215 336 092 37 127 59Co 46 530 000 67 67 42 364 000 46 69 426 164 850 21 73 98Mo 24 152 430 108 107 10 102 314 72 109 34 51 151 36 124 111Cd 21 14 465 107 108 21 9 946 73 112 9 4 696 34 120 208Pb 32 169 625 103 108 24 106 700 65 102 467 38 344 30 106 115In 177 787 117 590 514 052 0 0 * Standard = 108 ng g21 of V, Mo, Cd and Pb and 67 ng g21 of Co.Fig. 5 Two sample Youden plots of duplicate samples 9702 and 9704. Means of the results are presented by the dotted lines; the intersection being the target value.Continuous lines for Pb are the mean ± reproducibility. Analyst, November 1997, Vol. 122 1233Effect of Variation of the Sample Matrix Standards were made up in three different matrices; xylene, naphtha and oil, respectively. A naphtha sample with very low levels of metals as well as the NIST 1083 Base Oil were used for this purpose. Ten-fold dilutions of the standards in these different matrices were then analysed using both conventional Meinhardt–Scott ICP-MS and the Cetac system.The results in Table 5 show the integrated signals for the blanks and for a standards with 315 ng g21 V, Mo, Cd and Pb (S21) and 133 ng g21 Co, respectively. The signal for the In internal standard is also shown. The concentrations calculated using these signals with and without reference to the internal standard are also shown. It can be seen that the signal intensity of standard in naphtha was less than that of xylene.The signal suppression was highest in the base oil standards. From the concentrations calculated with reference to internal standard, it can be seen that this matrix effect has been corrected for. The effect of the matrix can be seen to be much more severe for the Cetac system in Table 6. The integrated signals for the blanks and a 108 ng g21 V, Mo, Cd and Pb and 67 ng g21 Co standards in the different matrices can be seen in Table 6. The integrated signal for the In internal standard is shown.Once again, the signal intensity decreases in the order: xylene > naphtha > oil. The base oil standard signal was 84% of that in the xylene for the conventional system. For the oil matrix measured by the Cetac system, the signal was suppressed to 30% of what it was in the xylene matrix. The concentrations calculated without reference to the internal standard can be seen to be seriously in error. Table 6 also illustrates that use of matrix matching is very important when using the Cetac system.However, when analysing light oils, one faces a dilemma. On the one hand, matrix does play a role in suppressing the signal. On the other hand, if one uses a base oil such as the NIST 1083 base, then one has a problem with accurately determining low levels of Ni and V seeing that this base oil contains for example 85 ng g21 Ni and 3 ng g21 V. The use of analytical grade xylene with appropriate internal standard addition has been found to provide the best compromise in this dilemma in our laboratory where the most important requirement is a correct V/(V+Ni) ratio.Conclusion The round robin has shown that conventional Meinhardt–Scott chamber ICP-MS performed well compared with other techniques such as ETAAS and CVAAS for the analysis of low levels of Pb and Hg in naphtha. ICP-MS has advantages over these techniques in that samples can be analysed directly without pretreatment and elements are analysed ‘simultaneously’.The fact that there is less sample handling and no need for addition of chemicals such as matrix modifiers and hydrides means that ICP-MS analyses are more rapid to carry out and less prone to problems such as contamination and influence of chemical species. ICP-MS provides the high sensitivity needed for the geochemical analysis of V and Ni in many condensates. However, reliable values of V and Ni cannot be determined in all light oils and condensates. The Cetac Ultrasonic nebuliser– desolvation system has been found to be essential for extending the range of condensates which can be analysed for fingerprinting or oil–oil correlation. By removing a large proportion of the solvent molecules, the problem of interference due to carbonacious species such as ArC+ on 51V is substantially decreased. The use of the Cetac Ultrasonic nebuliser–desolvation system is not recommended for the analysis of the environmentally important elements such as Pb and Hg in condensates and naphtha until the behaviour of the different species is better understood. Volatile species such as Et4Pb and Me2Hg were lost during the analysis with the Cetac system, but not with the conventional Meinhardt–Scott chamber ICP-MS system. Elf Petroleum funded and made possible work on metals in petroleum using ICP-MS as part of the application of new methods to a large multidisciplinary reservoir geochemistry study. A. K. Bjerklund of Løvland AS provided the Cetac U- 6000 AT Ultrasonic nebuliser–desolvation system on loan in order that the tests on petroleum samples could be carried out. West-Lab AS performed the GC analyses for this work. Permission given by the Institute for Interlaboratory Studies to report the naphtha results is gratefully acknowledged. References 1 Peters, K. E., and Moldowan, J. M., The Biomarker Guide, Prentice- Hall, Englewood Cliffs, New Jersey, 1993, p. 363. 2 Bingham, M. D., Field Detection and Implications of Mercury in Natural Gas, SPE Production Engineering, May 1990, pp. 120-124. 3 Irgolic, K. J., and Puri, B. K., in Metal Speciation in the Environment, ed. Broekaert, J. A. C., G�uçer, S., and Adams, F., NATO ASI Series G23 , Springer, Berlin, 1990, pp. 377-389. 4 Kucha, H., Slupczynski, K., and Prochaska, W., Nature, 1993, 363, 680. 5 Gijselman, P. B., in The First International Conference on Health, Safety and Environment in Oil and Gas Exploration and Production, The Hague, The Netherlands, 1991, pp. 123-130. 6 Filby, R.H., and Olsen, S. D., J. Radioanal. Nucl. Chem., 1994, 180, 285. 7 Olsen, S. D., Filby, R. H., Brekke, T., and Isaksen, G. H., Analyst, 1995, 120, 1379. 8 Visser, R.G., Results of Method Evaluating Interlaboratory Study Pb and Hg in Naphtha, Report No. ISS97N01, Institute for Interlaboratory Studies, Dordrecht, The Netherlands, 1997. Paper 7/04017B Received June 9, 1997 Accepted September 12, 1997 1234 Analyst, November 1997, Vol. 1
ISSN:0003-2654
DOI:10.1039/a704017b
出版商:RSC
年代:1997
数据来源: RSC
|
9. |
Determination of Iodide in Brines by Membrane Permeation Flow Injection Analysis† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1235-1238
J. T. Håkedal,
Preview
|
|
摘要:
Determination of Iodide in Brines by Membrane Permeation Flow Injection Analysis† J. T. Håkedal* and P. K. Egeberg Faculty of Mathematics and Sciences, Agder College, Tordenskjolds gate 65, Postuttak, N-4604 Kristiansand, Norway. E-mail: john.hakedal@hia.no A flow injection method is described for the determination of iodide based on the oxidation of iodide to iodine, which after permeation through a PTFE membrane is detected spectrophotometrically. The sample is injected into a carrier stream of water and merged with acidic dichromate reagent, oxidising iodide to iodine.The iodine permeates through the membrane into a collector stream containing iodide. The iodine reacts with iodide forming triiodide, and is measured spectrophotometrically in a flow cell at 350 nm. The repeatability of the technique was better than a relative standard deviation of 2% at 5 mg l21 level, with a throughput of 50 samples h21. The detection limit was 0.2 mg l21 iodide. The method is suitable for analysing high saline waters.No significant matrix effect is found for sea-water up to 32‰ S, or sodium chloride concentrations up to 50 g l21. The method is tested for potential interfering substances (Br2, COOH, CH3COOH, HCO32, Fe2+, Mn2+ and S22). The only interfering compound found was sulfide, which can be removed by preheating the samples after acidification. The method was tested by analysing marine pore water samples from Ocean Drilling Program Leg 164; the results compared well with data obtained by the manual method.Keywords: Iodide; spectrophotometry; flow injection analysis; membrane permeation; saline water The major source of iodide in pore fluids is from decomposition of organic matter.1 Because iodide does not form diagenetic minerals, it has been used to constrain the source of subsurface fluids,2 to study the ability of geological membranes to retard ion transport in ground water,3 and to estimate rates of pore water advection.4 Thus, iodide is an important tool in the study of subsurface waters.The high Cl : I ratio of natural brines (103 to 106) and complex matrixes, however, make the accurate determination of iodide problematic. Manual methods for iodide determination are usually tedious; however, flow injection analysis represents an interesting method for automation.5 Several flow injection methods for determination of iodide based on spectrophotometric,6–8 potentiometric9,10 and catalytic spectrophotometric detection11,12 have been reported.Motomizu and Yoden13 reported a tubular microporous PTFE membrane that was applied to the permeation of chlorine, bromine and iodine. By separation of the analyte from the sample and collecting it into a clean recipient stream matrix problems can be avoided and the selectivity of the method can be enhanced. The purpose of this study was to develop an automatic and rapid flow injection method for the determination of iodide in small sample volumes of saline waters.The method should not suffer from systematic errors caused by high salinity and dissolved organic matter content. Coloured samples should be analysed without interferences, necessitating a separation technique like membrane permeation. Experimental Apparatus The flow injection system is shown in Fig. 1. The instrument used was a FIAstar 5010 Analyzer (Tecator, H�ogan�as, Sweden) with a Chemifold V gas-diffusion cell, using a permeable membrane (Tecator).The (Teflon) membrane used was supplied with the unit. The detector consisted of a Hellma (Jamaica, NY, USA) flow cell with either a 10 or 40 mm light path installed in a Model 8625 UV/VIS spectrometer (Pye Unicam; now Philips Analytical, Cambridge, UK). All connections were 0.5 mm id Teflon tubes. The detector output was recorded on a Model R 100 A strip-chart recorder (Perkin-Elmer, Norwalk, CT, USA). For continuous variation of flow rate, a Minipuls 3 (Gilson, Villiers-le-Bel, France) peristaltic pump was used.The flow rate was measured with a Flocsoai Liquid Flowmeter (Phase Separations, Queensferry, Clwyd, UK). Reagents All solutions were prepared using analytical-reagent grade chemicals and de-ionized water. Iodide standard solution, 1 g l21. A stock solution was prepared by dissolving 1.308 g of potassium iodide (Merck, Darmstadt, Germany) in water made up to 1 l. Working standard solutions were prepared by diluting the stock solution with water.Potassium dichromate solution, 0.15 mol l21. This oxidation solution was prepared by dissolving 44.1 g of potassium dichromate (Merck) in 500 ml of de-ionized water containing about 55 ml of concentrated sulfuric acid and diluting the resulting solution to 1 l. The final concentration of sulfuric acid in this solution is approximately 1 m. Potassium iodide solution, 0.016 mol l21. The acceptor– carrier solution was prepared by dissolving 2.62 g of potassium iodide with water to give a 1 l solution.Sea-water. IAPSO Standard Seawater (Ocean Scientific International, Surrey, UK). All other solutions were prepared by dissolving an appropriate amount of salt or reagents in water. Procedure The sample (30–300 ml) was injected into a carrier stream of water, and merged with acidic oxidative solution for oxidation † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Fig. 1 Flow injection manifold: C, carrier; R1, oxidation reagent; R2, potassium iodide; P, peristaltic pump; I, sample injection; M, membrane; D, detector; W, waste. Analyst, November 1997, Vol. 122 (1235–1237) 1235of iodide to iodine. The liberated iodine permeated the PTFE membrane into an acceptor stream containing iodide, and the triiodide formed was detected spectrophotometrically in a flow cell at 350 nm. Quantifications are based on peak height measurement from the recorder output calibrated against absorbance readings from the spectrophotometer.When optimising the method, peak areas were measured and quantified by the cut-and-weigh method. A manual method used for comparative determinations was adapted from Pedersen,14 as described by Gieskes et al.15 The method determines the sum of iodide and iodate, but only iodide is present in the reduced samples analysed here. In brief, the method involves oxidation of iodide to iodate by means of Br2(aq) in an acetate buffer, removal of excess Br2(aq) by boiling, and spectrophotometric quantification of I32 formed by addition of excess amounts of iodide under acidic conditions: 2IO32 + 8I2 + 6H+"3I32 + 3H2O The maximum absorbance at 355 nm occurs approximately 3 min after addition of iodide.Results and Discussion Optimisation Tryzell and Karlberg16 found that, for gas diffusion manifolds, the acceptor flow rate is more significant to peak height than the donor flow rate.In this study, both peak area and peak height were measured for different acceptor flow rates (Fig. 2). Decreasing peak areas indicated lower transfer efficiency through the membrane when the flow rate was increased. The peak height reached a maximum at a flow rate of approximately 3 ml min21. The decrease in peak heights towards lower flow rates is probably due to dispersion in the tube connecting the collector side of the membrane and the flow cell. Iodide in the acceptor stream is necessary for detection of the transferred iodine as triiodide.The increase in peak height with increasing iodide concentration (Fig. 3) is in accordance with displacement of the equilibrium reaction: I2(aq) + I2"I32 K = 73102 The peak height is stabilised at an iodide concentration of 2000 mg l21, which was recommended for further operations. Three different oxidation agents were investigated for oxidation of iodide to iodine prior to permeation: iodate, dichromate and permanganate.The half-reactions together with reduction potential in acid solution are: IO32 + 6H+ + 5e2"1 2 I2(aq) + 3H2O E0 = 1.18 V Cr2O7 22 + 14H+ + 6e2"2Cr3+ + 7H2O E0 = 1.33 V MnO42 + 8H+ + 5e2"Mn2+ + 4H2O E0 = 1.51 V Iodate and permanganate gave high effects even at low concentrations, while the output for dichromate increased with increasing concentration (Table 1). Iodate it recommended because it may produce iodine by reaction with reduced compounds in the sample, causing interferences. Permanganate, being the strongest oxidation agent, oxidises iodide easily, but potassium permanganate solutions are unstable and difficult to handle. Dichromate is more selective for oxidation of iodide, and gave the lowest within-sample variance.Thus, dichromate was found to be appropriate for this method. Potassium dichromate solutions are stable, and do not react with hydrochloric acid.Interferences and Matrix Effects Potential interfering substances were studied. As can be seen from Table 2, sulfide interferes seriously, even at very low concentrations. The sulfide could be removed from the sample by heating the sample solution after acidification. A 100 ml volume of concentrated hydrochloric acid was added to 1.0 ml of sample, and the solution heated to near boiling for approximately 10 min. After cooling, the solution was diluted to 2.0 ml in a calibrated flask.This pre-treatment is adequate for removal of this interference. The other species commonly occurring in brines did not interfere significantly, the effects being less than ±5% at the 50 mg l21 iodide level. The effect of high salt content was investigated by measuring standard solutions of 40 mg l21 iodide, containing different concentrations of sodium chloride and standard sea-water. The peak heights were not significantly altered by increasing salt content (Fig. 4). Calibration The sensitivity of the method can be altered by varying the injection volume and change of flow cell to a different light Fig. 2 Effect of acceptor stream flow rate on peak height and peak area. Flow cell path length, 40 mm. Fig. 3 Effect of acceptor stream iodide concentration on peak height. Flow cell path length, 40 mm. Table 1 Alternative oxidation agents for oxidation of iodide to iodine. All peak heights are measured after injection of 100 ml of 50 mg l21 iodide solution.All oxidation solutions contained 1 mol l21 H2SO4. Within-sample variances are based on duplicate injections of the same sample at the various concentrations of oxidation agent Within- Peak height (arbitrary units) sample at various oxidant concentrations variance, s2 Oxidation 0.001 0.01 0.1 0.2 agent mol l21 mol l21 mol l21 mol l21 IO32 0.264 0.284 0.269 0.281 231025 Cr2O7 22 0.002 0.114 0.203* 0.219 431026 MnO42 0.230 0.215 0.242 — 431025 * 0.15 mol l21. 1236 Analyst, November 1997, Vol. 122path. The calibration graph was linear up to 400 mg l21 (3 mm) iodide, when using a 30 ml injection volume and a 1 cm light path. The detection limit was 0.2 mg l21 (2 mm) iodide, when 300 ml was injected and a 4 cm flow cell was used. The repeatability, with successive injections of a 5 mg l21 iodide standard solution, was better than a relative standard deviation of 2%. The sample throughput was 50 samples h21. Application The proposed flow injection method was applied to determine iodide in high saline water samples. Marine pore water samples from Ocean Drilling Program Leg 164 were coloured and contained high concentrations of both total organic carbon (TOC) and bromide. Concentrations of TOC in these samples ranged between 16 and 654 mg l21, while those for bromide ranged between 81 and 239 mg l21.Samples containing 2–200 mg l21 iodide were analysed by both flow injection and the manual method described by Gieskes et al.15 The six samples containing least iodide were sulfidic and treated accordingly, by removing the interfering sulfide.The other samples were injected directly for FI determination. A statistical treatment by linear regression was applied, and the equation y = 1.0008x + 1.675 (r = 0.9992) describing the line has a 95% confidence limit of 0.02126 for the slope and 2.3389 for the intercept. The value one and zero are included in the confidence limits for the slope and the intercept, respectively.Thus, there is no evidence of significant difference between the results obtained with the two analytical procedures.17 Conclusions The proposed method is suitable for analysing high salinity waters and brines. The sensitivity is not as good as other flow injection methods, e.g., a catalytic spectrophotometric flow injection method has a reported detection limit of 0.1 mg l21 iodide.11 The method described here is, however, highly selective to iodide, as no matrix effects or interferences other than sulfide were found.The flow injection method is simpler, easier to use and more rapid than the manual batch methods. The time required for calibration is short and the sampling frequency is high. Problems with analysing polluted water samples are avoided by the use of separation by membrane permeation. Furthermore, the proposed method can easily be adapted to different concentration ranges by simply modifying the light path of the flow cell or changing sample injection volume.Finally, the apparatus used is standard flow injection equipment that is commercially available and the standard and reagent solutions are easily prepared and have prolonged stability. References 1 Pedersen, T. F., and Price, N. B., J. Mar. Res., 1980, 38, 397. 2 Martin, J. B., Gieske, J. M., Torres, M., and Kastner, M., Geochim. Cosmochim. Acta, 1993, 57, 4377. 3 Kharaka, Y. K., and Berry, F. A., Geochim. Cosmochim.Acta, 1973, 37, 2577. 4 Egeberg, P. K., and Dickens, G. R., in preparation. 5 R°u�zi�cka, J., and Hansen, E. H., Flow Injection Analysis, Wiley, New York, 2nd edn., 1988. 6 Kamson, O. F., Anal. Chim. Acta, 1986, 179, 475. 7 Yaqoob, M., and Masoom, M., Anal. Chim. Acta, 1991, 248, 219. 8 Oguma, K., Kitada, K., and Kuroda, R., Mikrochim. Acta, 1993, 110, 71. 9 Davey, D. E., Mulcahy, D. E., and O’Connell, G. R., Analyst, 1992, 117, 761. 10 Najib, F. M., and Othman, S., Talanta, 1992, 39, 1259. 11 Yonehara, N., Kozono, S., and Sakamoto, H., Anal. Sci., 1991, 7, 229. 12 Chandrawanshi, C. K., Chandrawanshi, S. K., and Patel, K. S., J. Autom. Chem., 1996, 18, 181. 13 Motomizu, S., and Yoden, T., Anal. Chim. Acta, 1992, 261, 461. 14 Pedersen, T. F., PhD Thesis, University of Edinburgh, 1979. 15 Gieskes, J. M., Gamo, T., and Brumsack, H., Chemical Methods for Interstitial Water Analysis Aboard JOIDES Resolution, ODP Technical Note, 15, 1991. 16 Tryzell, R., and Karlberg, B., Anal.Chim. Acta, 1995, 308, 206. 17 Miller, J. C., and Miller, J. N., Statistics for Analytical Chemistry, Ellis Horwood PTR Prentice Hall, New York, 3rd edn., 1993. Paper 7/04014H Received June 9, 1997 Accepted September 1, 1997 Table 2 Effect of coexisting compounds. All test solutions contained 50 mg l21 iodide Ion/ Concentration/ Interference Compound Added as mmol l21 effect (%) Mn2+ MnSO4·H2O 10 +0.4 Fe2+ Fe(NH4)2(SO4)2·6H2O 10 +1.0 100 22.6 HCO32 NaHCO3 10 20.4 Br2 NaBr 10 +0.4 COOH COOH 10 21.8 CH3COOH CH3COOH 10 0.0 S22 Na2S 0.5 225.2 10 273.6 20 negative peaks 20* 24.4 * This sample was pre-treated by acidification and warming for removal of H2S.Fig. 4 Effect of increasing sea-water and sodium chloride concentrations in standard solutions containing 50 mg l21 iodide. Flow cell path length, 40 mm. Analyst, November 1997, Vol. 122 1237 Determination of Iodide in Brines by Membrane Permeation Flow Injection Analysis† J.T. Håkedal* and P. K. Egeberg Faculty of Mathematics and Sciences, Agder College, Tordenskjolds gate 65, Postuttak, N-4604 Kristiansand, Norway. E-mail: john.hakedal@hia.no A flow injection method is described for the determination of iodide based on the oxidation of iodide to iodine, which after permeation through a PTFE membrane is detected spectrophotometrically. The sample is injected into a carrier stream of water and merged with acidic dichromate reagent, oxidising iodide to iodine.The iodine permeates through the membrane into a collector stream containing iodide. The iodine reacts with iodide forming triiodide, and is measured spectrophotometrically in a flow cell at 350 nm. The repeatability of the technique was better than a relative standard deviation of 2% at 5 mg l21 level, with a throughput of 50 samples h21. The detection limit was 0.2 mg l21 iodide. The method is suitable forlysing high saline waters. No significant matrix effect is found for sea-water up to 32‰ S, or sodium chloride concentrations up to 50 g l21.The method is tested for potential interfering substances (Br2, COOH, CH3COOH, HCO32, Fe2+, Mn2+ and S22). The only interfering compound found was sulfide, which can be removed by preheating the samples after acidification. The method was tested by analysing marine pore water samples from Ocean Drilling Program Leg 164; the results compared well with data obtained by the manual method. Keywords: Iodide; spectrophotometry; flow injection analysis; membrane permeation; saline water The major source of iodide in pore fluids is from decomposition of organic matter.1 Because iodide does not form diagenetic minerals, it has been used to constrain the source of subsurface fluids,2 to study the ability of geological membranes to retard ion transport in ground water,3 and to estimate rates of pore water advection.4 Thus, iodide is an important tool in the study of subsurface waters.The high Cl : I ratio of natural brines (103 to 106) and complex matrixes, however, make the accurate determination of iodide problematic. Manual methods for iodide determination are usually tedious; however, flow injection analysis represents an interesting method for automation.5 Several flow injection methods for determination of iodide based on spectrophotometric,6–8 potentiometric9,10 and catalytic spectrophotometric detection11,12 have been reported.Motomizu and Yoden13 reported a tubular microporous PTFE membrane that was applied to the permeation of chlorine, bromine and iodine. By separation of the analyte from the sample and collecting it into a clean recipient stream matrix problems can be avoided and the selectivity of the method can be enhanced. The purpose of this study was to develop an automatic and rapid flow injection method for the determination of iodide in small sample volumes of saline waters. The method should not suffer from systematic errors caused by high salinity and dissolved organic matter content.Coloured samples should be analysed without interferences, necessitating a separation technique like membrane permeation. Experimental Apparatus The flow injection system is shown in Fig. 1. The instrument used was a FIAstar 5010 Analyzer (Tecator, H�ogan�as, Sweden) with a Chemifold V gas-diffusion cell, using a permeable membrane (Tecator). The (Teflon) membrane used was supplied with the unit.The detector consisted of a Hellma (Jamaica, NY, USA) flow cell with either a 10 or 40 mm light path installed in a Model 8625 UV/VIS spectrometer (Pye Unicam; now Philips Analytical, Cambridge, UK). All connections were 0.5 mm id Teflon tubes. The detector output was recorded on a Model R 100 A strip-chart recorder (Perkin-Elmer, Norwalk, CT, USA). For continuous variation of flow rate, a Minipuls 3 (Gilson, Villiers-le-Bel, France) peristaltic pump was used.The flow rate was measured with a Flocsoai Liquid Flowmeter (Phase Separations, Queensferry, Clwyd, UK). Reagents All solutions were prepared using analytical-reagent grade chemicals and de-ionized water. Iodide standard solution, 1 g l21. A stock solution was prepared by dissolving 1.308 g of potassium iodide (Merck, Darmstadt, Germany) in water made up to 1 l. Working standard solutions were prepared by diluting the stock solution with water. Potassium dichromate solution, 0.15 mol l21.This oxidation solution was prepared by dissolving 44.1 g of potassium dichromate (Merck) in 500 ml of de-ionized water containing about 55 ml of concentrated sulfuric acid and diluting the resulting solution to 1 l. The final concentration of sulfuric acid in this solution is approximately 1 m. Potassium iodide solution, 0.016 mol l21. The acceptor– carrier solution was prepared by dissolving 2.62 g of potassium iodide with water to give a 1 l solution. Sea-water.IAPSO Standard Seawater (Ocean Scientific International, Surrey, UK). All other solutions were prepared by dissolving an appropriate amount of salt or reagents in water. Procedure The sample (30–300 ml) was injected into a carrier stream of water, and merged with acidic oxidative solution for oxidation † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Fig. 1 Flow injection manifold: C, carrier; R1, oxidation reagent; R2, potassium iodide; P, peristaltic pump; I, sample injection; M, membrane; D, detector; W, waste.Analyst, November 1997, Vol. 122 (1235–1237) 1235of iodide to iodine. The liberated iodine permeated the PTFE membrane into an acceptor stream containing iodide, and the triiodide formed was detected spectrophotometrically in a flow cell at 350 nm. Quantifications are based on peak height measurement from the recorder output calibrated against absorbance readings from the spectrophotometer.When optimising the method, peak areas were measured and quantified by the cut-and-weigh method. A manual method used for comparative determinations was adapted from Pedersen,14 as described by Gieskes et al.15 The method determines the sum of iodide and iodate, but only iodide is present in the reduced samples analysed here. In brief, the method involves oxidation of iodide to iodate by means of Br2(aq) in an acetate buffer, removal of excess Br2(aq) by boiling, and spectrophotometric quantification of I32 formed by addition of excess amounts of iodide under acidic conditions: 2IO32 + 8I2 + 6H+"3I32 + 3H2O The maximum absorbance at 355 nm occurs approximately 3 min after addition of iodide.Results and Discussion Optimisation Tryzell and Karlberg16 found that, for gas diffusion manifolds, the acceptor flow rate is more significant to peak height than the donor flow rate. In this study, both peak area and peak height were measured for different acceptor flow rates (Fig. 2).Decreasing peak areas indicated lower transfer efficiency through the membrane when the flow rate was increased. The peak height reached a maximum at a flow rate of approximately 3 ml min21. The decrease in peak heights towards lower flow rates is probably due to dispersion in the tube connecting the collector side of the membrane and the flow cell. Iodide in the acceptor stream is necessary for detection of the transferred iodine as triiodide.The increase in peak height with increasing iodide concentration (Fig. 3) is in accordance with displacement of the equilibrium reaction: I2(aq) + I2"I32 K = 73102 The peak height is stabilised at an iodide concentration of 2000 mg l21, which was recommended for further operations. Three different oxidation agents were investigated for oxidation of iodide to iodine prior to permeation: iodate, dichromate and permanganate.The half-reactions together with reduction potential in acid solution are: IO32 + 6H+ + 5e2"1 2 I2(aq) + 3H2O E0 = 1.18 V Cr2O7 22 + 14H+ + 6e2"2Cr3+ + 7H2O E0 = 1.33 V MnO42 + 8H+ + 5e2"Mn2+ + 4H2O E0 = 1.51 V Iodate and permanganate gave high effects even at low concentrations, while the output for dichromate increased with increasing concentration (Table 1). Iodate is not recommended because it may produce iodine by reaction with reduced compounds in the sample, causing interferences.Permanganate, being the strongest oxidation agent, oxidises iodide easily, but potassium permanganate solutions are unstable and difficult to handle. Dichromate is more selective for oxidation of iodide, and gave the lowest within-sample variance. Thus, dichromate was found to be appropriate for this method. Potassium dichromate solutions are stable, and do not react with hydrochloric acid. Interferences and Matrix Effects Potential interfering substances were studied.As can be seen from Table 2, sulfide interferes seriously, even at very low concentrations. The sulfide could be removed from the sample by heating the sample solution after acidification. A 100 ml volume of concentrated hydrochloric acid was added to 1.0 ml of sample, and the solution heated to near boiling for approximately 10 min. After cooling, the solution was diluted to 2.0 ml in a calibrated flask. Thie-treatment is adequate for removal of this interference.The other species commonly occurring in brines did not interfere significantly, the effects being less than ±5% at the 50 mg l21 iodide level. The effect of high salt content was investigated by measuring standard solutions of 40 mg l21 iodide, containing different concentrations of sodium chloride and standard sea-water. The peak heights were not significantly altered by increasing salt content (Fig. 4). Calibration The sensitivity of the method can be altered by varying the injection volume and change of flow cell to a different light Fig. 2 Effect of acceptor stream flow rate on peak height and peak area. Flow cell path length, 40 mm. Fig. 3 Effect of acceptor stream iodide concentration on peak height. Flow cell path length, 40 mm. Table 1 Alternative oxidation agents for oxidation of iodide to iodine. All peak heights are measured after injection of 100 ml of 50 mg l21 iodide solution. All oxidation solutions contained 1 mol l21 H2SO4. Within-sample variances are based on duplicate injections of the same sample at the various concentrations of oxidation agent Within- Peak height (arbitrary units) sample at various oxidant concentrations variance, s2 Oxidation 0.001 0.01 0.1 0.2 agent mol l21 mol l21 mol l21 mol l21 IO32 0.264 0.284 0.269 0.281 231025 Cr2O7 22 0.002 0.114 0.203* 0.219 431026 MnO42 0.230 0.215 0.242 — 431025 * 0.15 mol l21. 1236 Analyst, November 1997, Vol. 122path. The calibration graph was linear up to 400 mg l21 (3 mm) iodide, when using a 30 ml injection volume and a 1 cm light path.The detection limit was 0.2 mg l21 (2 mm) iodide, when 300 ml was injected and a 4 cm flow cell was used. The repeatability, with successive injections of a 5 mg l21 iodide standard solution, was better than a relative standard deviation of 2%. The sample throughput was 50 samples h21. Application The proposed flow injection method was applied to determine iodide in high saline water samples.Marine pore water samples from Ocean Drilling Program Leg 164 were coloured and contained high concentrations of both total organic carbon (TOC) and bromide. Concentrations of TOC in these samples ranged between 16 and 654 mg l21, while those for bromide ranged between 81 and 239 mg l21. Samples containing 2–200 mg l21 iodide were analysed by both flow injection and the manual method described by Gieskes et al.15 The six samples containing least iodide were sulfidic and treated accordingly, by removing the interfering sulfide.The other samples were injected directly for FI determination. A statistical treatment by linear regression was applied, and the equation y = 1.0008x + 1.675 (r = 0.9992) describing the line has a 95% confidence limit of 0.02126 for the slope and 2.3389 for the intercept. The value one and zero are included in the confidence limits for the slope and the intercept, respectively. Thus, there is no evidence of significant difference between the results obtained with the two analytical procedures.17 Conclusions The proposed method is suitable for analysing high salinity waters and brines.The sensitivity is not as good as other flow injection methods, e.g., a catalytic spectrophotometric flow injection method has a reported detection limit of 0.1 mg l21 iodide.11 The method described here is, however, highly selective to iodide, as no matrix effects or interferences other than sulfide were found. The flow injection method is simpler, easier to use and more rapid than the manual batch methods.The time required for calibration is short and the sampling frequency is high. Problems with analysing polluted water samples are avoided by the use of separation by membrane permeation. Furthermore, the proposed method can easily be adapted to different concentration ranges by simply modifying the light path of the flow cell or changing sample injection volume. Finally, the apparatus used is standard flow injection equipment that is commercially available and the standard and reagent solutions are easily prepared and have prolonged stability.References 1 Pedersen, T. F., and Price, N. B., J. Mar. Res., 1980, 38, 397. 2 Martin, J. B., Gieske, J. M., Torres, M., and Kastner, M., Geochim. Cosmochim. Acta, 1993, 57, 4377. 3 Kharaka, Y. K., and Berry, F. A., Geochim. Cosmochim. Acta, 1973, 37, 2577. 4 Egeberg, P. K., and Dickens, G. R., in preparation. 5 R°u�zi�cka, J., and Hansen, E. H., Flow Injection Analysis, Wiley, New York, 2nd edn., 1988. 6 Kamson, O. F., Anal. Chim. Acta, 1986, 179, 475. 7 Yaqoob, M., and Masoom, M., Anal. Chim. Acta, 1991, 248, 219. 8 Oguma, K., Kitada, K., and Kuroda, R., Mikrochim. Acta, 1993, 110, 71. 9 Davey, D. E., Mulcahy, D. E., and O’Connell, G. R., Analyst, 1992, 117, 761. 10 Najib, F. M., and Othman, S., Talanta, 1992, 39, 1259. 11 Yonehara, N., Kozono, S., and Sakamoto, H., Anal. Sci., 1991, 7, 229. 12 Chandrawanshi, C. K., Chandrawanshi, S. K., and Patel, K. S., J. Autom. Chem., 1996, 18, 181. 13 Motomizu, S., and Yoden, T., Anal. Chim. Acta, 1992, 261, 461. 14 Pedersen, T. F., PhD Thesis, University of Edinburgh, 1979. 15 Gieskes, J. M., Gamo, T., and Brumsack, H., Chemical Methods for Interstitial Water Analysis Aboard JOIDES Resolution, ODP Technical Note, 15, 1991. 16 Tryzell, R., and Karlberg, B., Anal. Chim. Acta, 1995, 308, 206. 17 Miller, J. C., and Miller, J. N., Statistics for Analytical Chemistry, Ellis Horwood PTR Prentice Hall, New York, 3rd edn., 1993. Paper 7/04014H Received June 9, 1997 Accepted September 1, 1997 Table 2 Effect of coexisting compounds. All test solutions contained 50 mg l21 iodide Ion/ Concentration/ Interference Compound Added as mmol l21 effect (%) Mn2+ MnSO4·H2O 10 +0.4 Fe2+ Fe(NH4)2(SO4)2·6H2O 10 +1.0 100 22.6 HCO32 NaHCO3 10 20.4 Br2 NaBr 10 +0.4 COOH COOH 10 21.8 CH3COOH CH3COOH 10 0.0 S22 Na2S 0.5 225.2 10 273.6 20 negative peaks 20* 24.4 * This sample was pre-treated by acidification and warming for removal of H2S. Fig. 4 Effect of increasing sea-water and sodium chloride concentrations in standard solutions containing 50 mg l21 iodide. Flow cell path length, 40 mm. Analyst, November 1997, Vol. 122
ISSN:0003-2654
DOI:10.1039/a704014h
出版商:RSC
年代:1997
数据来源: RSC
|
10. |
Single Zircon Evaporation Thermal Ionisation Mass Spectrometry: Method and Procedures† |
|
Analyst,
Volume 122,
Issue 11,
1997,
Page 1239-1248
U. S. Klötzli,
Preview
|
|
摘要:
Single Zircon Evaporation Thermal Ionisation Mass Spectrometry: Method and Procedures† U. S. Kl�otzli Laboratory for Geochronology, Department of Geology, University of Vienna, Geozentrum, Althanstrasse-14, A-1090 Vienna, Austria Zircon evaporation thermal ionisation mass spectrometry (TIMS) is used in geochronology to determine absolute 207Pb*/206Pb* ages and Th/U ratios of single zircon crystals. The process involves the breakdown of zircon (ZrSiO4) to porous baddeleyite (ZrO2) along a reaction front which progresses into the interior of the crystal.Evaporation of high quality zircons thus allows us to distinguish between crystal rim (overgrowth) and core, providing precise information about the time of magmatic crystal growth, partial dissolution, and/or metamorphic overgrowth. Derived Th/U ratios complement age data interpretation and provide valuable petrogenetic implications. A double Re-filament ion source is used. The zircon is encased in the evaporation filament and heated step-wise to 1200–1300 °C to strip off unsupported common and radiogenic Pb components.After cleaning, evaporation proceeds in temperature steps of ~ 20 °C. The evaporate (SiO2, Pb, REEs, and U from the zircon and Re from the evaporation filament) of each step is deposited for 45 min on the cold ionisation filament and subsequently analysed. Lead isotopic composition is determined using a dynamic secondary electron multiplier ion counter or static Faraday cup data acquisition schemes.Lead ratios are corrected for fractionation using correction factors derived from standard measurements of a 1 ng NBS SRM 982 sample. The precision on 207Pb/206Pb ratios is < 1%. Only high temperature steps ( > 1300 °C) with 206Pb/204Pb > 5000 are used for age calculations. The ages reported (single temperature step, multi-temperature step means) are weighted means calculated from at least 20 measured 207Pb*/206Pb* ratios with 2 standard errors of the mean.Precision of ages is strongly dependent on age range and varies between 0.1 and 10%. Keywords: Thermal ionisation mass spectrometry; zircon evaporation; geochronology; lead–lead dating; absolute dating The uranium–thorium–lead and lead–lead dating techniques for zircon and other U-bearing mineral phases (i.e., monazite, xenotime, allanite, sphene) are standard geochronological methods used in laboratories throughout the world. In particular, zircon dating has proven to be a very powerful tool, due to the large resistance of zircon against thermal and mechanical alteration, in establishing absolute geochronological information about magma generation and intrusion events and/or periods of high temperature metamorphic overprinting.Other commonly used geochronometers normally do not allow the geoanalyst to directly date such high temperature events. The U–Th–Pb and Pb–Pb dating techniques are based on the radioactive decay of 238U to 206Pb*, 235U to 207Pb*, and 232Th to 208Pb*, respectively (for fundamentals see Hunziker and J�ager,1 Faure,2 and Geyh and Schleicher3). Conventional U–Th–Pb dating principally comprises the chemical dissolution of the mineral under investigation, the purification and determination of the concentrations of U, Th, Pb (using isotope dilution techniques), and the determination of the isotopic composition of Pb.Therefore, conventional U–Th–Pb dating requires an ultra-clean laboratory environment, sophisticated isotope dilution thermal ionisation mass spectrometry (TIMS), complicated fractionation corrections, and error evaluation.Additionally, the different age information present in different domains of one zircon crystal is often lost due to the ‘integration’ effect of crystal dissolution, thus providing less age information than actually present. These restrictions can be overcome by applying either even more sophisticated analytical methods [i.e., partial dissolution experiments (Mattinson4), mechanical separation of rim and core (Steiger et al.5), vapour digestion combined with cathodoluminescence investigations (Wendt and Todt,6 Poller et al.7) or spot analysis using ion microprobes (Compston et al.,8 Wiedenbeck and Goswami9) and laser ablation mass spectrometric techniques (e.g., Jackson et al.,10 Hirata and Nesbitt11)].This paper presents an alternative zircon dating method: single zircon evaporation TIMS comprises the direct determination of 207Pb*/206Pb* ages (so called lead–lead ages).This method circumvents the need for the determination of elemental concentrations of the conventional U–Th–Pb dating techniques. Thus, the laborious and complicated work involved in the ultraclean laboratory and the isotope dilution TIMS can be completely avoided. Direct evaporation TIMS of finely ground zircon for Pb isotope analysis was first suggested by Kosztolanyi12 as early as 1965 and was later applied by Sunin and Malyshev13 to zircons from various rocks.But these early applications were severely hampered by analytical difficulties (for discussion see Kober.14) Gentry et al.15 and Kober14 suggested a mounting procedure for unground zircon crystals in a single-filament ion-source. To further enhance beam stability and duration Kober14 used a double-filament ion-source arrangement and a stepwise heating procedure resulting in a major improvement in the precision of the resulting evaporation 207Pb*/206Pb* ages.The method described in this contribution is partly based on the method of Kober,14,16 but advances analytical precision, accuracy and the possibilities of age data interpretation. Geochronological Background A fundamental prerequisite of absolute age determination is the assumption of a closed system behaviour of mother and daughter elements and the absence of any internal isotope fractionation since the time of closure with respect to diffusion of the elements under consideration.Thus, if one assumes that no spatial fractionation between U, Th, and Pb has taken place in the crystal lattice since the time of closure the 207Pb*/206Pb* ratio can directly be recalculated to an apparent lead–lead age: † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997. Analyst, November 1997, Vol. 122 (1239–1248) 1239206 235 207 206 235 238 1 1 1 1 235 238 Pb = U (e and Pb U ( e Pb Pb U e U e 238 207 238 235 * ) * ) * * ( ) ( ) l l l l t t t t - = - ® = - - Assuming that the U isotopic composition is constant (238U/ 235U = 137.88) and that the decay constants of 238U and 235U are accurately and precisely known (Steiger and J�ager17) the transcendental equation can iteratively be solved for t, providing an age-estimate for the isotopic system since its closure.Additionally the 208Pb*/206Pb* ratio is directly proportional to the Th/U ratio, an important mineralo- and petrogenetic indicator which is helpful for the interpretation of single zircon age spectra (Kl�otzli18).Process of Zircon Evaporation The process of evaporation involves the breakdown of the accessory silicate zircon (ZrSiO4, normally < 200 mm in size) to the porous oxide baddeleyite (ZrO2) and the associated loss of mainly Si, SiO2, Pb, REEs, U and Th along a reaction front which progresses into the interior of the grain (Ansdell and Kyser,19 Roddick and Chapman,20 Roddick21 and Kl�otzli18). Continuous evaporation or evaporation with stepwise increasing evaporation temperatures ideally results in a ‘depth’ profile through the zircon crystal from the outermost rim to the innermost core (Kober,14 Kl�otzli,18 Kl�otzli and Parrish22).The stepwise evaporation on a ‘cold’ ionisation filament, as described below, makes use of the silica-gel effect (routinely used in conventional Pb TIMS, Cameron et al.23) and the accumulation of Pb on the ionisation filament resulting in better ionisation efficiency and ion beam stability (Kober14; Kl�otzli18).Mass Spectrometer Parameters and Maintenance A double Re-filament arrangement in a conventional extendedgeometry thermal ionisation mass spectrometer (Finnigan MAT 262, Bremen, Germany) equipped with an 120 MHz ion-counting secondary electron multiplier (SEM-IC) is used.Faraday System For Pb isotopic analysis the Faraday system is used in a conventional way and is not described in detail. SEM-IC System The SEM-IC system is heavily used in zircon evaporation TIMS (90% of all analyses) and therefore its operation and performance is described in more detail. Additionally, the use of dynamic data acquisition and the chronic instability of SEM systems requires precise monitoring of the SEM-IC system parameters. The peak shape and the deflection voltages of the ion beam in the SEM are controlled routinely before and after a zircon evaporation session using a ~ 8 mV Pb+ or Re+ ion beam.The optimum HV for the counting efficiency of the SEM is adjusted once a week using a Finnigan-MAT automatic calibration routine. It is set ~100 V higher than the beginning of the SEM saturation plateau. Since July 1991 the optimum HV has increased from 1950 to 2950 V (at present), demonstrating the effect of ageing of the SEM through heavy usage.Count gains are in the range of 90–95% relative to the axial Faraday cup. The precise count gain relative to the Faraday system is not determined because no mixed acquisition schemes are used in zircon evaporation TIMS. This is due to nonsufficient intensity overlap between the Faraday and SEM which leads to inconsistent and temporarily strongly varying count gains. Proper focusing has to be checked more frequently than when using the Faraday system. The focus point of the very small evaporate deposit (compared with conventionally loaded samples) tends to move during evaporation due to thermal relaxation and/or stress release of the evaporation filament.Additionally, the gain across the conversion dynode of the SEM is not constant leading to fluctuations in the ion beam intensity if the ion beam moves around. Dead time correction for the counter is accomplished on-line using Finnigan MAT software routines. The applied correction is correct up to ca. 500 000 counts s21, about half of the maximum allowable beam intensity. At higher beam intensities counting problems arise with the more abundant Pb masses 206 and 208 leading to noticeably to high 207Pb/206Pb and to low 208Pb/206Pb. Because dead time effects are mass and intensity dependent, an appropriate dead time correction for higher ion beam intensities would be rather complicated to accomplish and would incorporate additional uncertainty to the analysis. To completely exclude any dead time effects of the SEM-IC system ion beam intensities above 500 000 counts s21 are avoided, both for standard and zircon evaporation analyses.Experimental Procedures Zircon Preparation The zircons used for evaporation analyses are washed in warm 3 m HNO3 for half an hour, rinsed with distilled water and dried. Criteria for choosing zircon crystals for analysis are: absence of cracks, micro-fissures and inclusions, no visible turbidity, suitable size (30 to 250 mm along c-axis, elongation maximum 1 : 10), and colour.In special cases (e.g., direct analysis of an inherited core) abrasion techniques are employed (Krogh;24 Kl�otzli25). Filament Preparation Rhenium filaments used for encasing zircons are preheated and cleaned at 4.5 A for 4–8 h. This long outgassing duration softens the filaments substantially, thus making it easier to bend the filament and to encase the zircon. Softened filaments are first pre-bent using a modified jig of Finnigan MAT and then formed by tweezers for zircon encasing. Ionisation filaments used for the analysis of the evaporated deposit are outgassed at 4.5 A for 45 min.Zircon Evaporation Zircon cleaning procedures Zircon evaporation analyses follows modified procedures originally described by Kober14 and modified by Kl�otzli.25,26 Fig. 1 gives a schematic representation of the evaporation procedures. Encased zircons are stepwise heated to 1200–1300 °C in order to strip off unsupported common and radiogenic Pb components with low activation energies.Such Pb is weakly bound to metamict zircon domains, cracks, micro-cleavages, and to Pb bearing inclusions within the crystal. Especially in metamict crystal domains activation energies for Pb are very low (0.1–0.4 eV, Tilton27) and Pb is thus readily mobilised. Non-metamict domains are much more retentive with Pb activation energies between 2.2–2.5 eV (Shestakov28). Pure common Pb is characterised by the presence of 204Pb and low 206Pb/204Pb (see below). 1240 Analyst, November 1997, Vol. 122Evaporation-filament with zircon Ionisation-filament with evaporated deposit zircon evaporation and deposition on 'cold' ionisation-filament filament current/mA vs. temperature/°C Pb isotopic analysis and filament cleaning procedure 4500/1800 3000/1450 2200/1300 2000/1200 1500/950 1200/900 Ionisation-filament cleaning cycle Zircon cleaning cycle 1st evaporation/ analysis cycle 2nd evaporation/ analysis cycle 3rd evaporation/ analysis cycle 4th evaporation/ analysis cycle Check for zircon presence and Hf isotopic analysis time T4 T3 T2 T1 The progress of cleaning is always monitored with the EM-IC by mass scans from 203 to 209 (1 s integration time with 0.1 u steps).Additionally to common Pb, masses 204 (94Zr2 16O, 204Hg), 206 (94Zr2 18O), 207 (206Pb1H, organic material), 208 (207Pb1H, 96Zr2 16O, 176Hf16O2) can be occupied by isobaric molecules.The mass resolution of the MAT 262 is not sufficient to resolve these potential isobars. Therefore, before data acquisition can start, the absence of any isobars on the Pb isotope masses has to be checked. Masses 196 (90Zr2 16O), 202 (202Hg, 138BaP16O17O), 203 (203Tl), 205 (205Tl, 138BaP18O2, organic material) and 209 (208Pb1H, 209Bi, 177Hf16O2) are monitored to recognise the isobaric overlaps. In particular the isobar on mass 207 is critical because no constant ratio with any other non-Pb mass has been recognised which would allow an appropriate correction. Empirically the presence of 207 is always accompanied by 205 (with 207/205 < 15).Thus it is assumed that no isobaric 207 is present as soon as no 205 is present any more. Isobars of Zr2O and HfO2 normally do not pose any problems because the ionisation energy of the compound is far higher than the energies needed for Pb ionisation. Problems with isobars of BaPO2, Tl and Hg can arise when Clerici-solution (TlHCO2–Tl2C3H2O4) or Hg bearing heavy liquids (Thoulet-, Rohrbach-solution) were used for heavy mineral separation.PbH is sometimes present at very low temperatures ( < 850 °C) but disappears above 900 °C. Once no ‘low-temperature’ Pb or isobaric overlaps are present (0 min–2 h) the zircon temperature is raised by approximately 20 °C and, depending on ion beam intensities and experiment design, the isotopic composition of the evaporated Pb is either measured directly or the evaporate is deposited on the cold ionisation filament.Evaporation steps of 20 °C have been found to be a good compromise between spatial/temporal resolution (preferably small temperature steps) and sufficient ion beam intensity (preferably large temperature steps). The material evaporated from the evaporation filament consists mainly of a mixture of material from the zircon (70%) and Re from the filament (30%). From the zircon SiO2, most Pb, REEs, and about 50% of U and Th are quantitatively evaporated (Roddick and Chapman;20 Kl�otzli18).In particular, the amount of SiO2 evaporated depends strongly on crystal quality and to a somewhat lesser degree on evaporation temperature (Kl�otzli- Chowanetz et al.29). The amount of SiO2 deposited on the ionisation filament has a major influence on the quality of the subsequent Pb isotopic analysis. If only minor amounts of SiO2 are evaporated the silica-gel effect cannot function properly resulting in unstable ion beams and thus low analytical precision, although suitable amounts of Pb are present on the ionisation filament.This silica-gel effect sometimes jeopardises the analysis of highest quality zircons and leads to the situation that slightly metamics can be more easily analysed than completely non-metamict zircons, and that the absolute amount of Pb present in a crystal is not necessarily the dominant factor for the quality of a single zircon evaporation analysis. Total amounts of lead available are in the range of 1 pg to 1 ng for normal zircon crystals resulting in the analysis of sub-pg to subng amounts of Pb per evaporation step.The Pb+ ion yield is in the range of 1 3 1023 equivalent to the efficiency of the conventional silica-gel technique (Cameron et al.23). Evaporation procedures After a deposition step (15–45 min) Pb is analysed using either static Faraday or dynamic SEM-IC data acquisition procedures.During analyses, the evaporation filament is set to 1.2 A in order to prevent Pb being evaporated from the ionisation filament to be re-deposited on the evaporation filament (Ansdell and Kyser;30 Kl�otzli and Parrish22). After the Pb analyses the ionisation filament is raised to 4.5 A for some seconds, stripping off all remaining material deposited during the evaporation step. Then the next evaporation–analysis cycle can start. Evaporation temperatures are raised by approximately 20 °C from step to step until Pb evaporation from the zircon is complete.Depending on crystal quality, size, age and U and Pb content of the zircons, 2 to 8 evaporation–analysis cycles can be made. To check whether the Pb evaporation really is complete Fig. 1 Schematic representation of a single zircon evaporation analysis. T1–T4 symbolise increasing evaporation temperatures (+20 °C per evaporation step). See text for discussion. Analyst, November 1997, Vol. 122 1241or whether the zircon has just fallen off the evaporation filament, the evaporation filament is slowly raised to 3.5–3.8 A.The ionisation filament is set to 5 A. Presence of the zircon is controlled by monitoring the Zr+ ion beam on mass 90. Knowing whether Pb evaporation is complete or not is critical for the later interpretation of the evaporation data. High quality zircons can be heated to ~ 2000 °C. At these temperatures very stable Hf+ ion beams are attained thus making it possible to additionally determine the Hf isotopic composition of the zircon.The Hf isotopic composition gives important petrogenetic information, complementary to the Nd isotope system (e.g., crustal residence times, model ages, see Kl�otzli25 and references cited therein). Data Acquisition For compatibility reasons the acquisition schemes for the dynamic SEM-IC and static Faraday cup measurements are kept as similar as possible. In both modes and for normal acquisition ion beam intensities are measured in blocks of 10 scans with 4 s integration time and 2 s delay time each. Peak centering and intensity monitoring is done at the beginning of each block on mass 206.The background is measured on half-masses every 5 blocks with 15 s delay time and 32 s integration time. The background correction is made on-line during data acquisition. Acquisition schemes with masses analysed, integration times, and respective succession for the SEM-IC procedures are given in Table 1.For the SEM-IC analyses peaks are measured in succession of increasing mass in order to avoid problems with magnetic field stability (hysteresis). Data acquisition comprises 2 to 20 blocks, depending mostly on the durability of the ion beam. Acquisition is started or interrupted at a minimum ion beam intensity on mass 206 of 10 000 counts per s21 for the SEM-IC procedures and 5 mV for the Faraday cup procedures, respectively. Because of insufficient counting statistics, too low ion beam intensities can lead to unrealistic low 204Pb/206Pb ratios seemingly proving the absence of common Pb.Using the relatively high threshold intensities allows the recognition of 204Pb with appropriate precision. During the process of evaporation sometimes low to intermediate ion beam intensities ( < 5000 counts s21 on 206) normally exhibiting large intensity fluctuations can be registered. Special acquisition procedures with integration times of 16 s allows the analysis of these ion beams providing further information about the progress of evaporation.The same acquisition procedures are used to analyse low quality evaporation deposits from which only minor and unstable ion beams can be achieved. Age data derived from such procedures can then be compared to good age data found for other evaporation steps thus providing additional information for age data interpretation. Data Reduction and Fractionation Correction The subsequent data reduction is completely the same for both kinds of data acquisition schemes.Lead ratio calculation and statistical test with outlier elimination are performed using modified off-line software from Finnigan MAT. In Pb isotopic analysis no direct correction of machine discrimination and time dependent fractionation (internal normalisation) and other fractionation phenomena is possible. Natural lead is formed by a mixture of 4 different isotopes (204Pb, 206Pb, 207Pb, 208Pb), three of which are the stable daughter products of radioactive decay series (206Pb, 207Pb, 208Pb). So the amount of these 3 isotopes compared to 204Pb depends solely on the amount of radioactive mother isotope present and the time elapsed since the accumulation of the lead isotopes started.This then means that not one of all the possible ratios is constant and thus none can be used for the above mentioned corrections. This is the major drawback of the U–Pb and Pb–Pb dating techniques compared to other isotope systems (e.g., Rb/Sr, Sm/Nd), where the presence of at least two nonradiogenic isotopes resulting in one constant ratio allows the necessary corrections to be made.Therefore, 207Pb/206Pb and 208Pb/206Pb ratios are corrected using correction factors derived from NBS SRM 982 standard measurements with 1 ng of Pb loaded and using the conventional Si-gel technique (using the modified values given by Todt et al.31) These correction factors are determined individually for Faraday and SEM-IC procedures before and after the zircon analyses using the same data acquisition procedures as for zircon analyses. Two different sets of fractionation factors can be determined using both 208Pb/ 206Pb and 207Pb/206Pb values.The within error consistency of the two independently derived fractionation factors further proves the linearity and thus the correctness of the applied fractionation correction scheme.Compared to the overall analytical error of a single age the difference found in the 2 independently calculated fractionation factors is negligible. Because the 207Pb/206Pb ratio is the one of most interest, the correction factor derived from 207Pb/206Pb is used for the correction. Mass discrimination for Faraday procedures is about 50% of the mass discrimination of the SEM-IC procedures. 204Pb/206Pb ratios are not corrected. Table 2 gives the appropriate NBS SRM 982 data. Recommended standard values used are from Todt et al.31 Fig. 2 shows plots of the NBS SRM 982 data for the SEM-IC and the Faraday procedures. Table 3 gives a compilation of the derived correction factors, while Table 4 shows the influence of the applied correction on final 207Pb*/206Pb* ages using the SEM-IC Table 1 Peak succession and position used for zircon evaporation analysis and accompanying NBS SRM 982 standard measurements for both the SEM-IC and the Faraday collector systems Integration Detector time/delay system Mass succession/position time* Ziron cleaning SEM-IC 203–209 mass scan 1/0.1 u Common lead and isobars SEM-IC 206, 207, 208, 209, 202, 4/2 present 203, 204, 205 Common lead present SEM-IC 206, 207, 208, 204 4/2 (206Pb204Pb < 50 000) Faradays 204 = 7, 206 = 5, 207 = 4, 208 = 3 4 Radiogenic lead only SEM-IC 206, 207, 208 4/2 (206Pb204Pb > 50 000) Faradays 206 = 5, 207 = 4, 208 = 3 4 Radiogenic lead with very SEM-IC 206, 207, 208 16/2 low intensity, unstable * Measured in seconds. 1242 Analyst, November 1997, Vol. 1220.4650 0.9965 0.9970 0.9975 0.9980 0.9985 0.9990 0.9995 1.0000 1.0005 0.4655 0.4660 0.4665 0.4670 0.4675 0.4680 1 ng NBS 982 ICm mean 1 ng NBS 982 ICm NBS 982 Standard val207Pb/206Pb 207Pb/206Pb 208Pb/206Pb 208Pb/206Pb 0.460 0.462 0.464 0.466 0.468 0.470 0.472 0.474 0.476 0.478 0.480 1.015 1.013 1.011 1.009 1.007 1.005 1.003 1.001 0.999 0.997 0.995 ( a) ( b) 1 ng NBS 982 Far mean 1 ng NBS 982 Far NBS 982 Standard values system fractionation correction.In most cases, the applied correction is well within the analytical error of the raw data. Relative corrections range from ~ 10% for Tertiary age to < 0.1% for Archean ages. The absolute age shift is between 25 Ma and 23 Ma, respectively (1 Ma = 106 y). For the Faraday system corrections are about half the size and in the opposite direction. Time dependent fractionation (Rayleigh-type fractionation) cannot be accounted for using this simple correction calculation.But regarding the very small relative mass differences involved, the effect is assumed to be negligibly small and must thus not be accounted for. The calculated correction factors thus account for all biases of the SEM-IC and Faraday systems together with mass discrimination of the mass spectrometer and other fractionation effects, but not for time dependent fractionation effects. Bias characteristics of the SEM-IC compared to the Faraday cups are not independently incorporated in the data reduction because no mixed acquisition schemes (SEM-IC and Faraday cups used at the same time) are used in zircon evaporation analysis.Fractionation effects due to variation of heating procedures and ionisation temperatures can only be overcome by keeping the procedures and ionisation temperatures as similar and as Table 2 NBS SRM 982 standard values for the SEM-IC and Faraday collector systems (period 28.11.96 to 4.03.97) NBS 982 SEM-IC NBS 982 Faraday No.of analyses 25 2SE* % 25 2SE* % 208/206 measured 1.0052 ± 0.0030 0.30 0.9979 ± 0.0003 0.03 207/206 measured 0.4679 ± 0.0024 0.52 0.4665 ± 0.0002 0.04 208/204 measured 35.76 ± 0.70 1.96 36.72 ± 0.33 0.91 207/204 measured 17.62 ± 0.46 2.60 17.14 ± 0.18 1.07 206/204 measured 36.01 ± 0.70 1.95 36.51 ± 0.35 0.95 * Errors are 2 standard errors of the mean. Fig. 2 NBS SRM 982 standard measurements. Standard amount loaded for all measurements is 1 ng.(a) Plot of 25 dynamic SEM-IC collector measurements (period 02.12.96 to 04.03.97). (b) Plot of 19 static Faraday collector measurements (period 28.11.96 to 14.02.97). All error bars are 2sm. Recommended standard values of Todt et al.31 are given for comparison. Errors for recommended standard values are smaller than plotting symbol. Analyst, November 1997, Vol. 122 1243constant as possible for both standard measurements and for zircon evaporation analyses. For the applied acquisition schemes ionisation temperatures are constant within ±20 °C.After fractionation correction all relevant information from the acquired data is written as ASCII files for further evaluation. Pb Blank Assessment As no chemical treatment is used no severe problems with external Pb blank contributions are encountered in zircon evaporation analysis. But problems with Pb blank contributions from former zircon evaporation analyses could arise by the reactivation of deposits in the ion source coming especially from the first shielding plate.In order to avoid sample to sample cross contamination the shielding plate is changed and cleaned on a regular basis. Additionally an ion source blank is measured before each zircon analysis using a blank Re-single filament which is heated step-wise up to 5.5 A. The ion source can effectively be cleaned by using a defocused Re+ ion beam of 1–2 V intensity for 5 min duration. The process of cleaning can be monitored with mass 39K.Mass scans from 203 to 209 with 1 s integration time and 0.1 u steps are employed for monitoring the Pb blank. After cleaning, ion source Pb blank levels are in the range of < 10 counts s21 at 2200 °C and < 0.1 counts s21 at 1400 °C. The Pb blank of the used Re-filament material is negligibly small and thus poses no problems. Common Pb Correction The addition of common Pb to the radiogenic Pb poses one of the most significant problems in U–Th–Pb and Pb–Pb geochronology.Different methods are used to establish appropriate corrections needed to achieve geologically meaningful age Table 3 Lead mass fractionation values derived from NBS SRM 982 standard measurements (Table 2). Mass fractionation is independently calculated for 207Pb/206Pb and 208Pb/206Pb for both the SEM-IC and the Faraday collector systems. See text for discussion Mass Fraction* discrimim/ t 2SE† % nation/u 2SE† % 208/206 SEM-IC (25) 1.0050 ± 0.0030 0.301 20.00251 ± 0.0000076 0.301 207/206 SEM-IC (25) 1.0020 ± 0.0052 0.523 20.00200 ± 0.0000104 0.523 208/206 Far (19) 0.9978 ± 0.0003 0.031 0.00112 ± 0.0000004 0.031 207/206 Far (19) 0.9989 ± 0.0004 0.037 0.00107 ± 0.0000004 0.037 * m/t = measured value/true value.† Errors are 2 standard errors of the mean. Table 4 Influence of Pb isotope fractionation correction on final 207Pb*/206Pb* ages for the SEM-IC system. For comparison 3 different age ranges are shown. See text for discussion Age/ 207Pb/206Pb 2SE* % Ma 2SE* %† Tertiary age— Measured 0.04700 ± 0.00047 1.00 49.2 ± 23.9 48.5 Corrected 0.04691 ± 0.00053 1.13 44.5 ± 27.0 60.7 Age difference 24.8 10.7 Palaeozoic age— Measured 0.05800 ± 0.00058 1.00 530 ± 22 4.14 Corrected 0.05788 ± 0.00065 1.13 525 ± 25 4.71 Age difference 24.4 0.84 Archean age— Measured 0.45000 ± 0.00450 1.00 4085 ± 15 0.36 Corrected 0.44910 ± 0.00507 1.13 4082 ± 17 0.41 Age difference 23.0 0.07 Fractionation factor: 207/206 SEM-IC: 1.00200 ± 0.00524 (0.52%) * Errors are 2 standard errors of the mean.† Relative shift in corrected age. Table 5 Pb isotopic and age data for the zircon evaporation analysis of sample 2E92-B Weinsberg granite (cf. lower plot of Fig. 5). For discussion see text Evaporation 207/206 Block temperature*/ 2SE‡ age§/ 2SE†/ 2SE† No. °C 207/206† 2SE‡ % Ma Ma (%) 208/206† 2SE† Th/U¶ 2SE† Sample 2E92-B 2E92BC01 10 1398 0.05360 0.00022 0.4 354 9 2.6 0.0198 0.0049 0.060 0.015 2E92BC02 10 1443 0.05350 0.00019 0.4 350 8 2.3 0.0181 0.0047 0.055 0.014 2E92BC03 10 1463 0.05816 0.00025 0.4 536 10 1.8 0.0808 0.0051 0.243 0.015 2E92BC04 10 1482 0.05820 0.00018 0.3 537 7 1.3 0.0790 0.0053 0.238 0.016 2E92BC05 10 1504 0.05808 0.00028 0.5 533 11 2.0 0.0779 0.0054 0.235 0.016 Mean C01-C02, rim 0.05355 0.00006 0.1 352 3 0.7 Mean C03-C05, core 0.05815 0.00021 0.4 535 8 1.5 * Error on evaporation temperature is estimated to be ±10 °C.† Mean from individual scan ratios.‡ All errors reported are 2 standard errors of the mean. § Mean ages derived from individual scan ratios and not from individual scan ages. ¶ Th/U at apparent 207Pb/206Pb age. 1244 Analyst, November 1997, Vol. 122information. Fig. 3 shows the influence of the addition of common Pb to the radiogenic Pb for different age ranges. It is evident that common Pb can very dramatically change the apparent 207Pb/206Pb age. Any reasonable correction procedures must thus rely on precise knowledge of the common Pb isotopic composition, a fact very difficult to establish.Using wrong or non-precise common Pb compositions will unavoidably lead to wrong 207Pb*/206Pb* ages. If for the age calculation correct error assessment procedures are used the precision of the common Pb isotopic will have a large and negative impact on the achievable final age precision (see below). Thus the only reliable method to gain sound age data is to avoid common Pb contamination completely.In this respect, zircon evaporation analysis is superior to conventional U–Th– Pb dating. The sites of unsupported common Pb within a zircon crystal are metamict crystal domains, Pb bearing inclusion, fissures, and cracks. The common Pb can very effectively be removed from the crystal as described above. In order to minimise additionally the contribution of common Pb only high temperature steps ( > 1300 °C) with 204Pb/206Pb < 0.0002 (206Pb/204Pb > 5000) are used for age calculations.Therefore, the influence of common Pb (at least for zircons older than 500 Ma) is negligible and no common Pb correction has to be applied at all. During routine analysis 204Pb/ 206Pb is normally < 0.00001. As the 207Pb/206Pb ‘age’ of common Pb tends to be higher than the purely radiogenic 207Pb/206Pb age admixture of common Pb can be recognised by decreasing 207Pb/206Pb ages with increasing evaporation temperatures at low temperature steps.Rarely, evaporation data from zircons exhibiting a large common Pb contribution to the total Pb can be corrected for this common Pb component (Kl�otzli et al.33). Fig. 3 Influence of common Pb correction on measured 207Pb/206Pb ratios with varying 206Pb/204Pb and different age ranges. (a) Plot of the apparent 207Pb*/ 206Pb* age after common Pb correction. (b) Plot of the relative age difference between the apparent 207Pb/206Pb age and the 207Pb*/206Pb* age after common Pb correction. Dashed vertical lines designates 206Pb/204Pb = 5000.Common Pb composition for both plots are 206Pb/204Pb = 18.700 and 207Pb/ 204Pb = 15.628, respectively (mean crustal Pb, Stacey and Kramers.32) For discussion see text. Analyst, November 1997, Vol. 122 1245Age Calculation Age calculation and statistics are made using Microsoft Excel spreadsheets closely following evaluation routines given by York,34 Ludwig35 and Roddick et al.36 and with ISOPLOT of Ludwig.37 Decay constants used are from Steiger and J�ager.17 Reported ages are weighted-mean ages calculated from at least 20 measured 207Pb*/206Pb* ratios.Weighting factors for the individual ratios are derived from counting statistics. Errors reported are either 2 standard deviations (2s) or 2 standard errors of the mean (2sm). Correlation between 206Pb and 207Pb during data acquisition is assumed to be 0, so the correlation coefficient equals 0 for error calculation on 207Pb*/206Pb* ratios and ages.Bootstrap analysis and Monte Carlo simulations of measured 207Pb*/206Pb* ratio spectra are used to check whether or not obtained mean ratios and errors are statistically meaningful. Chi squared tests are used to check for proper Gaussian distribution of the 207Pb*/206Pb* ratios. It is assumed, that the variation of the 207Pb*/206Pb* of a single evaporation step of an analysis of mono-aged Pb should follow a Gaussian normal distribution (variation derived from counting statistics alone).Data sets not following a Gaussian distribution probably exhibit a mixture of different Pb components, which then precludes any significant age information. Variations in the 208Pb*/206Pb* cannot be checked in this respect because they primarily reflect changing Th/U of the evaporated zircon domain (Kl�otzli.18) Basic statistics have to be made using the raw data because of the non-linear age transformation obscuring any relevant non-Gaussian distributions.Reports are in the form of plots giving the most important parameters: evaporation temperature, 207Pb*/206Pb* ratios and ages, number of ages, mean values, 208Pb*/206Pb* and Th/U ratios (see examples). Error Assessment Proper error assessment is one of the major problems in analytical geochronology and is often not rigorously done (York,34 Ludwig,35,37 Mattinson,4 Roddick et al.,36 Kl�otzli25). Very often ages reported from single zircon evaporation analysis include errors which are simply derived from the internal analytical scatter of individual ages excluding any external error sources.Such a simple approach to error assessment is neither justified nor correct and should be avoided completely in as much as the involved mathematics for the correct error calculation are rather simple. In the present report all errors (except errors on the decay constants of the U isotopes) are propagated into the final mean ages. Error propagation is done using the standard Gaussian error propagation formula.Errors incorporated in calculations are: errors on individual ratios from counting statistics, errors derived from fractionation correction factors, errors from standard measurements and from the recommended standard values, and the weighted errors from individual temperature steps. If a common Pb correction is applied to the age data the precision of the common Pb isotopic composition is incorporated as well.Precision and Accuracy The accuracy of the method is demonstrated by a number of studies comparing single zircon evaporation data with conventional U/Pb data or with ion probe data (i.e., Ansdell and Kyser,30 Kl�otzli,26 Kl�otzli and Parrish,22 Kober,14 Kr�oner and Seng�or,38 Kr�oner et al.,39 M�uller et al.,40 Peindl and H�ock,41 Kl�otzli-Chowanetz et al.,29 Kl�otzli et al.33). The internal precision of the method is defined above, but is not of significant interest in a geological context.The more important external precision of the method can only be assessed by a complete error propagation scheme as shown above and comparison with the external reproducibility of individual zircons with the same age. At present no internationally recommended standard zircon is available for zircon evaporation analysis. In house reproducibilities of 27 analyses from a zircon population from an Ordovician alkali gneiss are in the range of 487.2 ± 9.7 Ma (2%).It is thus assumed that under normal circumstances the external precision of the method is in the range of 1–5%, depending mostly on the age range investigated (Kl�otzli,25 Bernhard et al.42). Zircon evaporation age data is interpreted to be significant if at least 3 crystals of a sample population exhibit (each within at least 2 temperature steps) within error concordance. It is then assumed that such age data reflects true concordant 207Pb*/206Pb* ages which can then be interpreted as geologically meaningful.Examples Age data derived from zircon evaporation analysis is often reported in the form of histograms showing age range versus number of ages or mass scans or blocks of the highest temperature steps from a number of zircons. The amount of information provided with such diagrams is rather scarce, sometimes even misleading. For instance, no individual errors are incorporated into a simple histogram and the reported mean age does not necessarily correspond to the most frequent 207Pb*/ 206Pb* ratio or age.All age information from lower temperature steps and from 208Pb*/206Pb* ratios is lost or omitted. If such a compilation is shown it should be done in the form of probability density plots of ages and not as histograms with arbitrary class widths. The examples presented here show all major aspects of single zircon evaporation dating, their possible representation and interpretation. Example 1 (Fig. 4) presents two plots of evaporation temperature versus 207Pb*/206Pb* age for two zircons from the paragneiss–migmatite boundary of the Winnebach migmatite in the Upper Austroalpine � Otztal–Stubai nappe of the Eastern Alps, Austria (Kl�otzli-Chowanetz et al.29).Width of plotted boxes is the estimated error on an evaporation temperature of ±10 °C. The height of the boxes is calculated as 2s of the mean age of individual temperature steps. Assigned errors of mean ages reported are all 2sm.The examples clearly demonstrate the complementary ‘behaviour’ of zircons during evaporation analysis. Evaporation of crystal 8830-E (Fig. 4) resulted in a perfect age plateau of 484 ± 6 Ma over 8 evaporation steps ranging from 1350 to 1490 °C. Conversely, evaporation of zircon 8830-C (Fig. 4) resulted in a staircase of increasing ages with increasing evaporation temperature. Both zircons were evaporated to completeness. The age plateau of 8830-E is interpreted to represent the crystallisation age of a core-free zircon (at least in respect to Pb isotopic systematics).Based on zircon typology and additional conventional zircon dating, the crystallisation event is attributed to the migmatite formation during the Ordovician. Zircon 8830-C exhibits (within ion steps) the same age for the migmatite formation event at 480 ± 6 Ma. The single step ages of 561 and 632 Ma possibly represent older metamorphic events. Similar but better defined ages were found in other zircons from the same locality, further supporting this interpretation.The higher temperature staircase ( > 1460 °C) is interpreted as representing a mixture of an old, possibly Archean Pb component with Pb of Cambrian or Proterozoic age. If no additional information for the two older metamorphic events would exist, the age staircase must be interpreted as representing a mixture between the Archean and the Ordovician Pb component.The age of 2355 ± 85 Ma is interpreted as representing a minimum age estimate for the formation or recrystallisation of an inherited zircon core. The evaporation temperature profile of zircon 8830-C directly proves within one zircon crystal the existence of differently old 1246 Analyst, November 1997, Vol. 122Pb components with individual levels of activation energy which can very effectively be separated by evaporation analysis (Kl�otzli18). Example 2 (Fig. 5) presents two plots of block number versus 207Pb*/206Pb* and 208Pb*/206Pb* for two zircons from granitoids of the South Bohemian Pluton, Austria (Kl�otzli and Parrish,22 Kl�otzli et al.33) The block number gives the progress of evaporation with steps of increasing temperature as indicated by labelled evaporation temperatures (in °C).Each block represents the mean of 10 mass scans. For easier reading errors on individual blocks are not shown. They are in the range of the symbol size.Assigned errors of mean ages reported are 2sm. Both zircons demonstrate the presence of an inherited core and a later overgrowth. Inherited cores and overgrowth could be analysed by at least two evaporation steps thus providing plateau ages which can be interpreted as being geologically meaningful. For zircon 4690-A both ages (336 ± 4 Ma and 635 ± 12 Ma, respectively) are interpreted to represent magmatic growth events. This interpretation is further supported by the large intra- and inter-evaporation step variation in 208Pb*/206Pb* at constant 207Pb*/206Pb* indicative for magmatic Th/U variation in the evaporated zircon domain.The analysis of the deposit of the third evaporation step of 4690-A (at 1515 °C) shows the typical pattern of a reversed deposit (Kl�otzli18), direct evidence for the mixing of differently old zircon domains during evaporation. The 208Pb*/ 206Pb* spectrum obtained for the inherited core of 2E92-B (535 ± 8 Ma) does not show significant variation.This reflects constant Th/U ratios throughout the inherited core. This is interpreted as reflecting crystal homogenisation during a high temperature metamorphic overprint leading to the formation of charnockitic rocks (Kl�otzli et al.32). The exact meaning of the Variscan overgrowth (352 ± 3 Ma) is still a matter of debate. Outlook To further enhance the precision of 207Pb*/206Pb* ages from zircon evaporation analysis static SEM-IC data acquisition using multi-collector SEM-IC systems has to be established.It should be possible to achieve the same precision for Pb isotopic analysis as is routinely found for multi-collector Faraday systems (i.e., 10 times better in precision as at present). One possible way of upgrading is by substituting the Faraday cups of a MAT 262 with ion-counting channeltrons as demonstrated by Fig. 4 Plots of evaporation temperature versus 207Pb*/206Pb* age for two zircons from the paragneiss–migmatite boundary of the Winnebach migmatite in the Upper Austroalpine � Otztal-Stubai nappe of the Eastern Alps, Austria (Kl�otzli-Chowanetz et al.29).Width of plotted boxes is estimated error on evaporation temperature of ±10 °C. Height of boxes is calculated as 2s of the mean age of individual temperature steps. For discussion see text. Fig. 5 Plots of block number versus 207Pb*/206Pb* and 208Pb*/206Pb* for two zircons from granitoids of the South Bohemian Pluton, Austria (Kl�otzli and Parrish,22 Kl�otzli et al.33) Block number gives progress of evaporation with steps of increasing temperature as indicated by labelled evaporation temperatures (in °C).Each block represents the mean of 10 mass scans. For easier reading errors on individual blocks are not shown. They are in the range of the symbol size. Assigned errors of mean ages reported are 2sm. For discussion see text. Analyst, November 1997, Vol. 122 1247Richter et al.43 or by using a newly designed Wien-filter TIMS (Laue et al.44).Additionally, the zircon mounting and encasing procedures can substantially be improved by using preformed Re filaments and micro-manipulators. The author thanks the following for spending tedious hours at the MS, for discussion, criticism, advice, a steady hand during zircon mounting, and critical reviews of an earlier version of this paper: F. Bernhard, E. Chowanetz, W. Frank, V. H�ock, G. Hoinkes, M. Jelenc, S. Meli, B. M�uller.Financial support by the Austrian Science Foundation is also acknowledged. References 1 Lectures in Isotope Geology, ed. Hunziker, J. C., and J�ager, E., Springer Verlag, Berlin, 1979. 2 Faure, G., Principles of Isotope Geology, 2nd edn., Wiley, New York, 1986. 3 Geyh, M., and Schleicher, H., Absolute Age Determination: Physical and Chemical Dating Methods and Their Application, Springer Verlag, Berlin, 1990. 4 Mattinson, J. M., Chem. Geol., 1987, 66, 151. 5 Steiger, R., Bickel, R.A., and Meier, M., Terra Abstr., 1993, 1–5, 395. 6 Wendt, J. J., and Todt, W., Terra Abstr., 1991, 3, 507. 7 Poller, U., Liebetrau, V., and Todt, W., in Proceedings of the V. M. Goldschmidt Conference, Heidelberg, 1996, volume 1, p. 119. 8 Compston, W., Williams, I. S., and Clement, S. W., in Proceedings of the 30th American Society of Mass Spectrometry Conference, 1982, p. 593. 9 Wiedenbeck, M., and Goswami, J. N., Geochim. Cosmochim Acta, 1994, 58, 2135. 10 Jackson, S.E., Longerich, H. P., Horn, I., and Dunning, G. R., in Proceedings of the V. M. Goldschmidt Conference, Heidelberg, 1996, vol. 1, p. 283. 11 Hirata, T., and Nesbitt, R. W., Geochim. Cosmochim. Acta, 1995, 59, 2491. 12 Kosztolanyi, C., Compt. Rend. Acad Sci., 1965, 261, 5849. 13 Sunin, L. V., and Malyshev, V. I., Geochem. Int., 1983, 20, 34. 14 Kober, B., Contrib. Mineral. Petrol., 1997, 93, 482. 15 Gentry, R. V., Sworski, T. J., McKown, H. S., Eby, R. E., and Christie, W.H., Science, 1982, 216, 296. 16 Kober, B., Contrib. Mineral. Petrol, 1996, 96, 63. 17 Steiger, R. H., and J�ager, E., Earth Planet. Sci. Lett., 1977, 36, 359. 18 Kl�otzli, U. S., Chem. Geol., 1997, submitted for publication. 19 Ansdell, K. M., and Kyser, T. K., Am. Min., 1993, 78, 36. 20 Roddick, J. C., and Chapman, H. J., EOS, Trans. Am. Geophys. Union, 1991, 72, 531. 21 Roddick, J. C., USGS Circular, 1994, 1107, 269. 22 Kl�otzli, U. S., and Parrish, R. R., Mineral. Petrol., 1996, 58, 197. 23 Cameron, A. E., Smith, D. E., and Walker, R. L., Anal. Chem., 1969, 41, 525. 24 Krogh, T. E., Geochim. Cosmochim. Acta, 1982, 46, 637. 25 Kl�otzli, U. S., Mitt. � Osterr. Miner. Ges., 1994, submitted for publication. 26 Kl�otzli, U. S., Mitt. � Osterr. Miner. Ges., 1993, 138, 123. 27 Tilton, G. R., J. Geophys. Res., 1960, 65, 2933. 28 Shestakov, G. I., Geochim. Int., 1972, 9, 801. 29 Kl�otzli-Chowanetz, E., Kl�otzli, U. S., and Koller, F., Schw. Miner. Petrol. Mitt., 1997, in the press. 30 Ansdell, K.M., and Kyser, T. K., Geology, 1991, 19, 518. 31 Todt, W., Cliff, R. A., Hanser A., and Hofmann, A. W., Geophys. Monograph, 1996, 95, 429. 32 Stacey, J. S., and Kramers, J. D., Earth Planet. Sci. Lett., 1975, 26, 207. 33 Kl�otzli, U. S, Koller, F., Scharbert, S., and H�ock, V., Chem. Geol., 1997, submitted for publication. 34 York, D., Earth Planet. Sci. Lett., 1969, 5, 320. 35 Ludwig, K. R., Earth Planet. Sci. Lett., 1980, 46, 212. 36 Roddick, J.C., Loveridge, W. D., and Parrish, R. R., Chem. Geol., 1987, 66, 111. 37 Ludwig, K. R., USGS Open-file Report, 1992, 91-445. 38 Kr�oner, A., and Seng�or, A. M. C., Geology, 1990, 18, 1186. 39 Kr�oner, A., Todt, W., Humbrian Res., 1992, 59, 15. 40 M�uller, B., Kl�otzli, U. S., and Flisch, M., Geol. Rundschau, 1995, 84, 457. 41 Peindl, P., and H�ock, V., Terra Abstr., 1993, 1/5, 392. 42 Bernhard, F., Kl�otzli, U. S., Hoinkes, G., and Th�oni, M., Mineral.Petrol., 1996, 58, 171. 43 Richter, S., Ott, U., and Begemann, F., Int. J. Mass Spec. Ion. Proc., 1994, 136, 91. 44 Laue, H. J., Tegtmeyer, A., and Wegener, M., Spectromat Inf. Sheet, 1995, 1. Paper 7/04114D Received June 12, 1997 Accepted September 1, 1997 1248 Analyst, November 1997, Vol. 122 Single Zircon Evaporation Thermal Ionisation Mass Spectrometry: Method and Procedures† U. S. Kl�otzli Laboratory for Geochronology, Department of Geology, University of Vienna, Geozentrum, Althanstrasse-14, A-1090 Vienna, Austria Zircon evaporation thermal ionisation mass spectrometry (TIMS) is used in geochronology to determine absolute 207Pb*/206Pb* ages and Th/U ratios of single zircon crystals.The process involves the breakdown of zircon (ZrSiO4) to porous baddeleyite (ZrO2) along a reaction front which progresses into the interior of the crystal. Evaporation of high quality zircons thus allows us to distinguish between crystal rim (overgrowth) and core, providing precise information about the time of magmatic crystal growth, partial dissolution, and/or metamorphic overgrowth.Derived Th/U ratios complement age data interpretation and provide valuable petrogenetic implications. A double Re-filament ion source is used. The zircon is encased in the evaporation filament and heated step-wise to 1200–1300 °C to strip off unsupported common and radiogenic Pb components. After cleaning, evaporation proceeds in temperature steps of ~ 20 °C.The evaporate (SiO2, Pb, REEs, and U from the zircon and Re from the evaporation filament) of each step is deposited for 45 min on the cold ionisation filament and subsequently analysed. Lead isotopic composition is determined using a dynamic secondary electron multiplier ion counter or static Faraday cup data acquisition schemes. Lead ratios are corrected for fractionation using correction factors derived from standard measurements of a 1 ng NBS SRM 982 sample.The precision on 207Pb/206Pb ratios is < 1%. Only high temperature steps ( > 1300 °C) with 206Pb/204Pb > 5000 are used for age calculations. The ages reported (single temperature step, multi-temperature step means) are weighted means calculated from at least 20 measured 207Pb*/206Pb* ratios with 2 standard errors of the mean. Precision of ages is strongly dependent on age range and varies between 0.1 and 10%. Keywords: Thermal ionisation mass spectrometry; zircon evaporation; geochronology; lead–lead dating; absolute dating The uranium–thorium–lead and lead–lead dating techniques for zircon and other U-bearing mineral phases (i.e., monazite, xenotime, allanite, sphene) are standard geochronological methods used in laboratories throughout the world.In particular, zircon dating has proven to be a very powerful tool, due to the large resistance of zircon against thermal and mechanical alteration, in establishing absolute geochronological information about magma generation and intrusion events and/or periods of high temperature metamorphic overprinting. Other commonly used geochronometers normally do not allow the geoanalyst to directly date such high temperature events.The U–Th–Pb and Pb–Pb dating techniques are based on the radioactive decay of 238U to 206Pb*, 235U to 207Pb*, and 232Th to 208Pb*, respectively (for fundamentals see Hunziker and J�ager,1 Faure,2 and Geyh and Schleicher3).Conventional U–Th–Pb dating principally comprises the chemical dissolution of the mineral under investigation, the purification and determination of the concentrations of U, Th, Pb (using isotope dilution techniques), and the determination of the isotopic composition of Pb. Therefore, conventional U–Th–Pb dating requires an ultra-clean laboratory environment, sophisticated isotope dilution thermal ionisation mass spectrometry (TIMS), complicated fractionation corrections, and error evaluation.Additionally, the different age information present in different domains of one zircon crystal is often lost due to the ‘integration’ effect of crystal dissolution, thus providing less age information than actually present. These restrictions can be overcome by applying either even more sophisticated analytical methods [i.e., partial dissolution experiments (Mattinson4), mechanical separation of rim and core (Steiger et al.5), vapour digestion combined with cathodoluminescence investigations (Wendt and Todt,6 Poller et al.7) or spot analysis using ion microprobes (Compston et al.,8 Wiedenbeck and Goswami9) and laser ablation mass spectrometric techniques (e.g., Jackson et al.,10 Hirata and Nesbitt11)].This paper presents an alternative zircon dating method: single zircon evaporation TIMS comprises the direct determination of 207Pb*/206Pb* ages (so called lead–lead ages). This method circumvents the need for the determination of elemental concentrations of the conventional U–Th–Pb dating techniques.Thus, the laborious and complicated work involved in the ultraclean laboratory and the isotope dilution TIMS can be completely avoided. Direct evaporation TIMS of finely ground zircon for Pb isotope analysis was first suggested by Kosztolanyi12 as early as 1965 and was later applied by Sunin and Malyshev13 to zircons from various rocks. But these early applications were severely hampered by analytical difficulties (for discussion see Kober.14) Gentry et al.15 and Kober14 suggested a mounting procedure for unground zircon crystals in a single-filament ion-source. To further enhance beam stability and duration Kober14 used a double-filament ion-source arrangement and a stepwise heating procedure resulting in a major improvement in the precision of the resulting evaporation 207Pb*/206Pb* ages.The method described in this contribution is partly based on the method of Kober,14,16 but advances analytical precision, accuracy and the possibilities of age data interpretation.Geochronological Background A fundamental prerequisite of absolute age determination is the assumption of a closed system behaviour of mother and daughter elements and the absence of any internal isotope fractionation since the time of closure with respect to diffusion of the elements under consideration. Thus, if one assumes that no spatial fractionation between U, Th, and Pb has taken place in the crystal lattice since the time of closure the 207Pb*/206Pb* ratio can directly be recalculated to an apparent lead–lead age: † Presented at Geoanalysis 97: 3rd International Conference on the Analysis of Geological and Environmental Materials, Vail, CO, USA, June 1–5, 1997.Analyst, November 1997, Vol. 122 (1239–1248) 1239206 235 207 206 235 238 1 1 1 1 235 238 Pb = U (e and Pb U ( e Pb Pb U e U e 238 207 238 235 * ) * ) * * ( ) ( ) l l l l t t t t - = - ® = - - Assuming that the U isotopic composition is constant (238U/ 235U = 137.88) and that the decay constants of 238U and 235U are accurately and precisely known (Steiger and J�ager17) the transcendental equation can iteratively be solved for t, providing an age-estimate for the isotopic system since its closure.Additionally the 208Pb*/206Pb* ratio is directly proportional to the Th/U ratio, an important mineralo- and petrogenetic indicator which is helpful for the interpretation of single zircon age spectra (Kl�otzli18).Process of Zircon Evaporation The process of evaporation involves the breakdown of the accessory silicate zircon (ZrSiO4, normally < 200 mm in size) to the porous oxide baddeleyite (ZrO2) and the associated loss of mainly Si, SiO2, Pb, REEs, U and Th along a reaction front which progresses into the interior of the grain (Ansdell and Kyser,19 Roddick and Chapman,20 Roddick21 and Kl�otzli18). Continuous evaporation or evaporation with stepwise increasing evaporation temperatures ideally results in a ‘depth’ profile through the zircon crystal from thhe innermost core (Kober,14 Kl�otzli,18 Kl�otzli and Parrish22).The stepwise evaporation on a ‘cold’ ionisation filament, as described below, makes use of the silica-gel effect (routinely used in conventional Pb TIMS, Cameron et al.23) and the accumulation of Pb on the ionisation filament resulting in better ionisation efficiency and ion beam stability (Kober14; Kl�otzli18).Mass Spectrometer Parameters and Maintenance A double Re-filament arrangement in a conventional extendedgeometry thermal ionisation multi-collector mass spectrometer (Finnigan MAT 262, Bremen, Germany) equipped with an 120 MHz ion-counting secondary electron multiplier (SEM-IC) is used. Faraday System For Pb isotopic analysis the Faraday system is used in a conventional way and is not described in detail. SEM-IC System The SEM-IC system is heavily used in zircon evaporation TIMS (90% of all analyses) and therefore its operation and performance is described in more detail. Additionally, the use of dynamic data acquisition and the chronic instability of SEM systems requires precise monitoring of the SEM-IC system parameters. The peak shape and the deflection voltages of the ion beam in the SEM are controlled routinely before and after a zircon evaporation session using a ~ 8 mV Pb+ or Re+ ion beam.The optimum HV for the counting efficiency of the SEM is adjusted once a week using a Finnigan-MAT automatic calibration routine. It is set ~100 V higher than the beginning of the SEM saturation plateau. Since July 1991 the optimum HV has increased from 1950 to 2950 V (at present), demonstrating the effect of ageing of the SEM through heavy usage. Count gains are in the range of 90–95% relative to the axial Faraday cup. The precise count gain relative to the Faraday system is not determined because no mixed acquisition schemes are used in zircon evaporation TIMS.This is due to nonsufficient intensity overlap between the Faraday and SEM which leads to inconsistent and temporarily strongly varying count gains. Proper focusing has to be checked more frequently than when using the Faraday system. The focus point of the very small evaporate deposit (compared with conventionally loaded samples) tends to move during evaporation due to thermal relaxation and/or stress release of the evaporation filament.Additionally, the gain across the conversion dynode of the SEM is not constant leading to fluctuations in the ion beam intensity if the ion beam moves around. Dead time correction for the counter is accomplished on-line using Finnigan MAT software routines. The applied correction is correct up to ca. 500 000 counts s21, about half of the maximum allowable beam intensity. At higher beam intensities counting problems arise with the more abundant Pb masses 206 and 208 leading to noticeably to high 207Pb/206Pb and to low 208Pb/206Pb. Because dead time effects are mass and intensity dependent, an appropriate dead time correction for higher ion beam intensities would be rather complicated to accomplish and would incorporate additional uncertainty to the analysis.To completely exclude any dead time effects of the SEM-IC system ion beam intensities above 500 000 counts s21 are avoided, both for standard and zircon evaporation analyses.Experimental Procedures Zircon Preparation The zircons used for evaporation analyses are washed in warm 3 m HNO3 for half an hour, rinsed with distilled water and dried. Criteria for choosing zircon crystals for analysis are: absence of cracks, micro-fissures and inclusions, no visible turbidity, suitable size (30 to 250 mm along c-axis, elongation maximum 1 : 10), and colour. In special cases (e.g., direct analysis of an inherited core) abrasion techniques are employed (Krogh;24 Kl�otzli25).Filament Preparation Rhenium filaments used for encasing zircons are preheated and cleaned at 4.5 A for 4–8 h. This long outgassing duration softens the filaments substantially, thus making it easier to bend the filament and to encase the zircon. Softened filaments are first pre-bent using a modified jig of Finnigan MAT and then formed by tweezers for zircon encasing. Ionisation filaments used for the analysis of the evaporated deposit are outgassed at 4.5 A for 45 min.Zircon Evaporation Zircon cleaning procedures Zircon evaporation analyses follows modified procedures originally described by Kober14 and modified by Kl�otzli.25,26 Fig. 1 gives a schematic representation of the evaporation procedures. Encased zircons are stepwise heated to 1200–1300 °C in order to strip off unsupported common and radiogenic Pb components with low activation energies. Such Pb is weakly bound to metamict zircon domains, cracks, micro-cleavages, and to Pb bearing inclusions within the crystal.Especially in metamict crystal domains activation energies for Pb are very low (0.1–0.4 eV, Tilton27) and Pb is thus readily mobilised. Non-metamict domains are much more retentive with Pb activation energies between 2.2–2.5 eV (Shestakov28). Pure common Pb is characterised by the presence of 204Pb and low 206Pb/204Pb (see below). 1240 Analyst, November 1997, Vol. 122Evaporation-filament with zircon Ionisation-filament with evaporated deposit zircon evaporation and deposition on 'cold' ionisation-filament filament current/mA vs.temperature/°C Pb isotopic analysis and filament cleaning procedure 4500/1800 3000/1450 2200/1300 2000/1200 1500/950 1200/900 Ionisation-filament cleaning cycle Zircon cleaning cycle 1st evaporation/ analysis cycle 2nd evaporation/ analysis cycle 3rd evaporation/ analysis cycle 4th evaporation/ analysis cycle Check for zircon presence and Hf isotopic analysis time T4 T3 T2 T1 The progress of cleaning is always monitored with the EM-IC by mass scans from 203 to 209 (1 s integration time with 0.1 u steps).Additionally to common Pb, masses 204 (94Zr2 16O, 204Hg), 206 (94Zr2 18O), 207 (206Pb1H, organic material), 208 (207Pb1H, 96Zr2 16O, 176Hf16O2) can be occupied by isobaric molecules. The mass resolution of the MAT 262 is not sufficient to resolve these potential isobars. Therefore, before data acquisition can start, the absence of any isobars on the Pb isotope masses has to be checked.Masses 196 (90Zr2 16O), 202 (202Hg, 138BaP16O17O), 203 (203Tl), 205 (205Tl, 138BaP18O2, organic material) and 209 (208Pb1H, 209Bi, 177Hf16O2) are monitored to recognise the isobaric overlaps. In particular the isobar on mass 207 is critical because no constant ratio with any other non-Pb mass has been recognised which would allow an appropriate correction. Empirically the presence of 207 is always accompanied by 205 (with 207/205 < 15).Thus it is assumed that no isobaric 207 is present as soon as no 205 is present any more. Isobars of Zr2O and HfO2 normally do not pose any problems because the ionisation energy of the compound is far higher than the energies needed for Pb ionisation. Problems with isobars of BaPO2, Tl and Hg can arise when Clerici-solution (TlHCO2–Tl2C3H2O4) or Hg bearing heavy liquids (Thoulet-, Rohrbach-solution) were used for heavy mineral separation.PbH is sometimes present at very low temperatures ( < 850 °C) but disappears above 900 °C. Once no ‘low-temperature’ Pb or isobaric overlaps are present (0 min–2 h) the zircon temperature is raised by approximately 20 °C and, depending on ion beam intensities and experiment design, the isotopic composition of the evaporated Pb is either measured directly or the evaporate is deposited on the cold ionisation filament. Evaporation steps of 20 °C have been found to be a good compromise between spatial/temporal resolution (preferably small temperature steps) and sufficient ion beam intensity (preferably large temperature steps).The material evaporated from the evaporation filament consists mainly of a mixture of material from the zircon (70%) and Re from the filament (30%). From the zircon SiO2, most Pb, REEs, and about 50% of U and Th are quantitatively evaporated (Roddick and Chapman;20 Kl�otzli18). In particular, the amount of SiO2 evaporated depends strongly on crystal quality and to a somewhat lesser degree on evaporation (Kl�otzli- Chowanetz et al.29).The amount of SiO2 deposited on the ionisation filament has a major influence on the quality of the subsequent Pb isotopic analysis. If only minor amounts of SiO2 are evaporated the silica-gel effect cannot function properly resulting in unstable ion beams and thus low analytical precision, although suitable amounts of Pb are present on the ionisation filament.This silica-gel effect sometimes jeopardises the analysis of highest quality zircons and leads to the situation that slightly metamict zircons can be more easily analysed than completely non-metamict zircons, and that the absolute amount of Pb present in a crystal is not necessarily the dominant factor for the quality of a single zircon evaporation analysis. Total amounts of lead available are in the range of 1 pg to 1 ng for normal zircon crystals resulting in the analysis of sub-pg to subng amounts of Pb per evaporation step.The Pb+ ion yield is in the range of 1 3 1023 equivalent to the efficiency of the conventional silica-gel technique (Cameron et al.23). Evaporation procedures After a deposition step (15–45 min) Pb is analysed using either static Faraday or dynamic SEM-IC data acquisition procedures. During analyses, the evaporation filament is set to 1.2 A in order to prevent Pb being evaporated from the ionisation filament to be re-deposited on the evaporation filament (Ansdell and Kyser;30 Kl�otzli and Parrish22).After the Pb analyses the ionisation filament is raised to 4.5 A for some seconds, stripping off all remaining material deposited during the evaporation step. Then the next evaporation–analysis cycle can start. Evaporation temperatures are raised by approximately 20 °C from step to step until Pb evaporation from the zircon is complete. Depending on crystal quality, size, age and U and Pb content of the zircons, 2 to 8 evaporation–analysis cycles can be made.To check whether the Pb evaporation really is complete Fig. 1 Schematic representation of a single zircon evaporation analysis. T1–T4 symbolise increasing evaporation temperatures (+20 °C per evaporation step). See text for discussion. Analyst, November 1997, Vol. 122 1241or whether the zircon has just fallen off the evaporation filament, the evaporation filament is slowly raised to 3.5–3.8 A.The ionisation filament is set to 5 A. Presence of the zircon is controlled by monitoring the Zr+ ion beam on mass 90. Knowing whether Pb evaporation is complete or not is critical for the later interpretation of the evaporation data. High quality zircons can be heated to ~ 2000 °C. At these temperatures very stable Hf+ ion beams are attained thus making it possible to additionally determine the Hf isotopic composition of the zircon. The Hf isotopic composition gives important petrogenetic information, complementary to the Nd isotope system (e.g., crustal residence times, model ages, see Kl�otzli25 and references cited therein).Data Acquisition For compatibility reasons the acquisition schemes for the dynamic SEM-IC and static Faraday cup measurements are kept as similar as possible. In both modes and for normal acquisition ion beam intensities are measured in blocks of 10 scans with 4 s integration time and 2 s delay time each. Peak centering and intensity monitoring is done at the beginning of each block on mass 206.The background is measured on half-masses every 5 blocks with 15 s delay time and 32 s integration time. The background correction is made on-line during data acquisition. Acquisition schemes with masses analysed, integration times, and respective succession for the SEM-IC procedures are given in Table 1. For the SEM-IC analyses peaks are measured in succession of increasing mass in order to avoid problems with magnetic field stability (hysteresis).Data acquisition comprises 2 to 20 blocks, depending mostly on the durability of the ion beam. Acquisition is started or interrupted at a minimum ion beam intensity on mass 206 of 10 000 counts per s21 for the SEM-IC procedures and 5 mV for the Faraday cup procedures, respectively. Because of insufficient counting statistics, too low ion beam intensities can lead to unrealistic low 204Pb/206Pb ratios seemingly proving the absence of common Pb.Using the relatively high threshold intensities allows the recognition of 204Pb with appropriate precision. During the process of evaporation sometimes low to intermediate ion beam intensities ( < 5000 counts s21 on 206) normally exhibiting large intensity fluctuations can be registered. Special acquisition procedures with integration times of 16 s allows the analysis of these ion beams providing further information about the progress of evaporation. The same acquisition procedures are used to analyse low quality evaporation deposits from which only minor and unstable ion beams can be achieved.Age data derived from such procedures can then be compared to good age data found for other evaporation steps thus providing additional information for age data interpretation. Data Reduction and Fractionation Correction The subsequent data reduction is completely the same for both kinds of data acquisition schemes. Lead ratio calculation and statistical test with outlier elimination are performed using modified off-line software from Finnigan MAT.In Pb isotopic analysis no direct correction of machine discrimination and time dependent fractionation (internal normalisation) and other fractionation phenomena is possible. Natural lead is formed by a mixture of 4 different isotopes (204Pb, 206Pb, 207Pb, 208Pb), three of which are the stable daughter products of radioactive decay series (206Pb, 207Pb, 208Pb).So the amount of these 3 isotopes compared to 204Pb depends solely on the amount of radioactive mother isotope present and the time elapsed since the accumulation of the lead isotopes started. This then means that not one of all the possible ratios is constant and thus none can be used for the above mentioned corrections. This is the major drawback of the U–Pb and Pb–Pb dating techniques compared to other isotope systems (e.g., Rb/Sr, Sm/Nd), where the presence of at least two nonradiogenic isotopes resulting in one constant ratio allows the necessary corrections to be made.Therefore, 207Pb/206Pb and 208Pb/206Pb ratios are corrected using correction factors derived from NBS SRM 982 standard measurements with 1 ng of Pb loaded and using the conventional Si-gel technique (using the modified values given by Todt et al.31) These correction factors are determined individually for Faraday and SEM-IC procedures before and after the zircon analyses using the same data acquisition procedures as for zircon analyses.Two different sets of fractionation factors can be determined using both 208Pb/ 206Pb and 207Pb/206Pb values. The within error consistency of the two independently derived fractionation factors further proves the linearity and thus the correctness of the applied fractionation correction scheme. Compared to the overall analytical error of a single age the difference found in the 2 independently calculated fractionation factors is negligible.Because the 207Pb/206Pb ratio is the one of most interest, the correction factor derived from 207Pb/206Pb is used for the correction. Mass discrimination for Faraday procedures is about 50% of the mass discrimination of the SEM-IC procedures. 204Pb/206Pb ratios are not corrected. Table 2 gives the appropriate NBS SRM 982 data. Recommended standard values used are from Todt et al.31 Fig. 2 shows plots of the NBS SRM 982 data for the SEM-IC and the Faraday procedures.Table 3 gives a compilation of the derived correction factors, while Table 4 shows the influence of the applied correction on final 207Pb*/206Pb* ages using the SEM-IC Table 1 Peak succession and position used for zircon evaporation analysis and accompanying NBS SRM 982 standard measurements for both the SEM-IC and the Faraday collector systems Integration Detector time/delay system Mass succession/position time* Ziron cleaning SEM-IC 203–209 mass scan 1/0.1 u Common lead and isobars SEM-IC 206, 207, 208, 209, 202, 4/2 present 203, 204, 205 lead present SEM-IC 206, 207, 208, 204 4/2 (206Pb204Pb < 50 000) Faradays 204 = 7, 206 = 5, 207 = 4, 208 = 3 4 Radiogenic lead only SEM-IC 206, 207, 208 4/2 (206Pb204Pb > 50 000) Faradays 206 = 5, 207 = 4, 208 = 3 4 Radiogenic lead with very SEM-IC 206, 207, 208 16/2 low intensity, unstable * Measured in seconds. 1242 Analyst, November 1997, Vol. 1220.4650 0.9965 0.9970 0.9975 0.9980 0.9985 0.9990 0.9995 1.0000 1.0005 0.4655 0.4660 0.4665 0.4670 0.4675 0.4680 1 ng NBS 982 ICm mean 1 ng NBS 982 ICm NBS 982 Standard values 207Pb/206Pb 207Pb/206Pb 208Pb/206Pb 208Pb/206Pb 0.460 0.462 0.464 0.466 0.468 0.470 0.472 0.474 0.476 0.478 0.480 1.015 1.013 1.011 1.009 1.007 1.005 1.003 1.001 0.999 0.997 0.995 ( a) ( b) 1 ng NBS 982 Far mean 1 ng NBS 982 Far NBS 982 Standard values system fractionation correction.In most cases, the applied correction is well within the analytical error of the raw data.Relative corrections range from ~ 10% for Tertiary age to < 0.1% for Archean ages. The absolute age shift is between 25 Ma and 23 Ma, respectively (1 Ma = 106 y). For the Faraday system corrections are about half the size and in the opposite direction. Time dependent fractionation (Rayleigh-type fractionation) cannot be accounted for using this simple correction calculation. But regarding the very small relative mass differences involved, the effect is assumed to be negligibly small and must thus not be accounted for.The calculated correction factors thus account for all biases of the SEM-IC and Faraday systems together with mass discrimination of the mass spectrometer and other fractionation effects, but not for time dependent fractionation effects. Bias characteristics of the SEM-IC compared to the Faraday cups are not independently incorporated in the data reduction because no mixed acquisition schemes (SEM-IC and Faraday cups used at the same time) are used in zircon evaporation analysis. Fractionation effects due to variation of heating procedures and ionisation temperatures can only be overcome by keeping the procedures and ionisation temperatures as similar and as Table 2 NBS SRM 982 standard values for the SEM-IC and Faraday collector systems (period 28.11.96 to 4.03.97) NBS 982 SEM-IC NBS 982 Faraday No.of analyses 25 2SE* % 25 2SE* % 208/206 measured 1.0052 ± 0.0030 0.30 0.9979 ± 0.0003 0.03 207/206 measured 0.4679 ± 0.0024 0.52 0.4665 ± 0.0002 0.04 208/204 measured 35.76 ± 0.70 1.96 36.72 ± 0.33 0.91 207/204 measured 17.62 ± 0.46 2.60 17.14 ± 0.18 1.07 206/204 measured 36.01 ± 0.70 1.95 36.51 ± 0.35 0.95 * Errors are 2 standard errors of the mean.Fig. 2 NBS SRM 982 standard measurements. Standard amount loaded for all measurements is 1 ng. (a) Plot of 25 dynamic SEM-IC collector measurements (period 02.12.96 to 04.03.97). (b) Plot of 19 static Faraday collector measurements (period 28.11.96 to 14.02.97).All error bars are 2sm. Recommended standard values of Todt et al.31 are given for comparison. Errors for recommended standard values are smaller than plotting symbol. Analyst, November 1997, Vol. 122 1243constant as possible for both standard measurements and for zircon evaporation analyses. For the applied acquisition schemes ionisation temperatures are constant within ±20 °C. After fractionation correction all relevant information from the acquired data is written as ASCII files for further evaluation.Pb Blank Assessment As no chemical treatment is used no severe problems with external Pb blank contributions are encountered in zircon evaporation analysis. But problems with Pb blank contributions from former zircon evaporation analyses could arise by the reactivation of deposits in the ion source coming especially from the first shielding plate. In order to avoid sample to sample cross contamination the shielding plate is changed and cleaned on a regular basis.Additionally an ion source blank is measured before each zircon analysis using a blank Re-single filament which is heated step-wise up to 5.5 A. The ion source can effectively be cleaned by using a defocused Re+ ion beam of 1–2 V intensity for 5 min duration. The process of cleaning can be monitored with mass 39K. Mass scans from 203 to 209 with 1 s integration time and 0.1 u steps are employed for monitoring the Pb blank.After cleaning, ion source Pb blank levels are in the range of < 10 counts s21 at 2200 °C and < 0.1 counts s21 at 1400 °C. The Pb blank of the used Re-filament material is negligibly small and thus poses no problems. Common Pb Correction The addition of common Pb to the radiogenic Pb poses one of the most significant problems in U–Th–Pb and Pb–Pb geochronology. Different methods are used to establish appropriate corrections needed to achieve geologically meaningful age Table 3 Lead mass fractionation values derived from NBS SRM 982 standard measurements (Table 2).Mass fractionation is independently calculated for 207Pb/206Pb and 208Pb/206Pb for both the SEM-IC and the Faraday collector systems. See text for discussion Mass Fraction* discrimim/ t 2SE† % nation/u 2SE† % 208/206 SEM-IC (25) 1.0050 ± 0.0030 0.301 20.00251 ± 0.0000076 0.301 207/206 SEM-IC (25) 1.0020 ± 0.0052 0.523 20.00200 ± 0.0000104 0.523 208/206 Far (19) 0.9978 ± 0.0003 0.031 0.00112 ± 0.0000004 0.031 207/206 Far (19) 0.9989 ± 0.0004 0.037 0.00107 ± 0.0000004 0.037 * m/t = measured value/true value.† Errors are 2 standard errors of the mean. Table 4 Influence of Pb isotope fractionation correction on final 207Pb*/206Pb* ages for the SEM-IC system. For comparison 3 different age ranges are shown. See text for discussion Age/ 207Pb/206Pb 2SE* % Ma 2SE* %† Tertiary age— Measured 0.04700 ± 0.00047 1.00 49.2 ± 23.9 48.5 Corrected 0.04691 ± 0.00053 1.13 44.5 ± 27.0 60.7 Age difference 24.8 10.7 Palaeozoic age— Measured 0.05800 ± 0.00058 1.00 530 ± 22 4.14 Corrected 0.05788 ± 0.00065 1.13 525 ± 25 4.71 Age difference 24.4 0.84 Archean age— Measured 0.45000 ± 0.00450 1.00 4085 ± 15 0.36 Corrected 0.44910 ± 0.00507 1.13 4082 ± 17 0.41 Age difference 23.0 0.07 Fractionation factor: 207/206 SEM-IC: 1.00200 ± 0.00524 (0.52%) * Errors are 2 standard errors of the mean.† Relative shift in corrected age.Table 5 Pb isotopic and age data for the zircon evaporation analysis of sample 2E92-B Weinsberg granite (cf. lower plot of Fig. 5). For discussion see text Evaporation 207/206 Block temperature*/ 2SE‡ age§/ 2SE†/ 2SE† No. °C 207/206† 2SE‡ % Ma Ma (%) 208/206† 2SE† Th/U¶ 2SE† Sample 2E92-B 2E92BC01 10 1398 0.05360 0.00022 0.4 354 9 2.6 0.0198 0.0049 0.060 0.015 2E92BC02 10 1443 0.05350 0.00019 0.4 350 8 2.3 0.0181 0.0047 0.055 0.014 2E92BC03 10 1463 0.05816 0.00025 0.4 536 10 1.8 0.0808 0.0051 0.243 0.015 2E92BC04 10 1482 0.05820 0.00018 0.3 537 7 1.3 0.0790 0.0053 0.238 0.016 2E92BC05 10 1504 0.05808 0.00028 0.5 533 11 2.0 0.0779 0.0054 0.235 0.016 Mean C01-C02, rim 0.05355 0.00006 0.1 352 3 0.7 Mean C03-C05, core 0.05815 0.00021 0.4 535 8 1.5 * Error on evaporation temperature is estimated to be ±10 °C.† Mean from individual scan ratios. ‡ All errors reported are 2 standard errors of the mean. § Mean ages derived from individual scan ratios and not from individual scan ages.¶ Th/U at apparent 207Pb/206Pb age. 1244 Analyst, November 1997, Vol. 122information. Fig. 3 shows the influence of the addition of common Pb to the radiogenic Pb for different age ranges. It is evident that common Pb can very dramatically change the apparent 207Pb/206Pb age. Any reasonable correction procedures must thus rely on precise knowledge of the common Pb isotopic composition, a fact very difficult to establish. Using wrong or non-precise common Pb compositions will unavoidably lead to wrong 207Pb*/206Pb* ages.If for the age calculation correct error assessment procedures are used the precision of the common Pb isotopic will have a large and negative impact on the achievable final age precision (see below). Thus the only reliable method to gain sound age data is to avoid common Pb contamination completely. In this respect, zircon evaporation analysis is superior to conventional U–Th– Pb dating.The sites of unsupported common Pb within a zircon crystal are metamict crystal domains, Pb bearing inclusion, fissures, and cracks. The common Pb can very effectively be removed from the crystal as described above. In order to minimise additionally the contribution of common Pb only high temperature steps ( > 1300 °C) with 204Pb/206Pb < 0.0002 (206Pb/204Pb > 5000) are used for age calculations. Therefore, the influence of common Pb (at least for zircons older than 500 Ma) is negligible and no common Pb correction has to be applied at all.During routine analysis 204Pb/ 206Pb is normally < 0.00001. As the 207Pb/206Pb ‘age’ of common Pb tends to be higher than the purely radiogenic 207Pb/206Pb age admixture of common Pb can be recognised by decreasing 207Pb/206Pb ages with increasing evaporation temperatures at low temperature steps. Rarely, evaporation data from zircons exhibiting a large common Pb contribution to the total Pb can be corrected for this common Pb component (Kl�otzli et al.33).Fig. 3 Influence of common Pb correction on measured 207Pb/206Pb ratios with varying 206Pb/204Pb and different age ranges. (a) Plot of the apparent 207Pb*/ 206Pb* age after common Pb correction. (b) Plot of the relative age difference between the apparent 207Pb/206Pb age and the 207Pb*/206Pb* age after common Pb correction. Dashed vertical lines designates 206Pb/204Pb = 5000. Common Pb composition for both plots are 206Pb/204Pb = 18.700 and 207Pb/ 204Pb = 15.628, respectively (mean crustal Pb, Stacey and Kramers.32) For discussion see text.Analyst, November 1997, Vol. 122 1245Age Calculation Age calculation and statistics are made using Microsoft Excel spreadsheets closely following evaluation routines given by York,34 Ludwig35 and Roddick et al.36 and with ISOPLOT of Ludwig.37 Decay constants used are from Steiger and J�ager.17 Reported ages are weighted-mean ages calculated from at least 20 measured 207Pb*/206Pb* ratios.Weighting factors for the individual ratios are derived from counting statistics. Errors reported are either 2 standard deviations (2s) or 2 standard errors of the mean (2sm). Correlation between 206Pb and 207Pb during data acquisition is assumed to be 0, so the correlation coefficient equals 0 for error calculation on 207Pb*/206Pb* ratios and ages. Bootstrap analysis and Monte Carlo simulations of measured 207Pb*/206Pb* ratio spectra are used to check whether or not obtained mean ratios and errors are statistically meaningful.Chi squared tests are used to check for proper Gaussian distribution of the 207Pb*/206Pb* ratios. It is assumed, that the variation of the 207Pb*/206Pb* of a single evaporation step of an analysis of mono-aged Pb should follow a Gaussian normal distribution (variation derived from counting statistics alone). Data sets not following a Gaussian distribution probably exhibit a mixture of different Pb components, which then precludes any significant age information.Variations in the 208Pb*/206Pb* cannot be checked in this respect because they primarily reflect changing Th/U of the evaporated zircon domain (Kl�otzli.18) Basic statistics have to be made using the raw data because of the non-linear age transformation obscuring any relevant non-Gaussian distributions. Reports are in the form of plots giving the most important parameters: evaporation temperature, 207Pb*/206Pb* ratios and ages, number of ages, mean values, 208Pb*/206Pb* and Th/U ratios (see examples).Error Assessment Proper error assessment is one of the major problems in analytical geochronology and is often not rigorously done (York,34 Ludwig,35,37 Mattinson,4 Roddick et al.,36 Kl�otzli25). Very often ages reported from single zircon evaporation analysis include errors which are simply derived from the internal analytical scatter of individual ages excluding any external error sources.Such a simple approach to error assessment is neither justified nor correct and should be avoided completely in as much as the involved mathematics for the correct error calculation are rather simple. In the present report all errors (except errors on the decay constants of the U isotopes) are propagated into the final mean ages. Error propagation is done using the standard Gaussian error propagation formula. Errors incorporated in calculations are: errors on individual ratios from counting statistics, errors derived from fractionation correction factors, errors from standard measurements and from the recommended standard values, and the weighted errors from individual temperature steps.If a common Pb correction is applied to the age data the precision of the common Pb isotopic composition is incorporated as well. Precision and Accuracy The accuracy of the method is demonstrated by a number of studies comparing single zircon evaporation data with conventional U/Pb data or with ion probe data (i.e., Ansdell and Kyser,30 Kl�otzli,26 Kl�otzli and Parrish,22 Kober,14 Kr�oner and Seng�or,38 Kr�oner et al.,39 M�uller et al.,40 Peindl and H�ock,41 Kl�otzli-Chowanetz et al.,29 Kl�otzli et al.33).The internal precision of the method is defined above, but is not of significant interest in a geological context. The more important external precision of the method can only be assessed by a complete error propagation scheme as shown above and comparison with the external reproducibility of individual zircons with the same age.At present no internationally recommended standard zircon is available for zircon evaporation analysis. In house reproducibilities of 27 analyses from a zircon population from an Ordovician alkali gneiss are in the range of 487.2 ± 9.7 Ma (2%). It is thus assumed that under normal circumstances the external precision of the method is in the range of 1–5%, depending mostly on the age range investigated (Kl�otzli,25 Bernhard et al.42).Zircon evaporation age data is interpreted to be significant if at least 3 crystals of a sample population exhibit (each within at least 2 temperature steps) within error concordance. It is then assumed that such age data reflects true concordant 207Pb*/206Pb* ages which can then be interpreted as geologically meaningful. Examples Age data derived from zircon evaporation analysis is often reported in the form of histograms showing age range versus number of ages or mass scans or blocks of the highest temperature steps from a number of zircons.The amount of information provided with such diagrams is rather scarce, sometimes even misleading. For instance, no individual errors are incorporated into a simple histogram and the reported mean age does not necessarily correspond to the most frequent 207Pb*/ 206Pb* ratio or age. All age information from lower temperature steps and from 208Pb*/206Pb* ratios is lost or omitted. If such a compilation is shown it should be done in the form of probability density plots of ages and not as histograms with arbitrary class widths.The examples presented here show all major aspects of single zircon evaporation dating, their possible representation and interpretation. Example 1 (Fig. 4) presents two plots of evaporation temperature versus 207Pb*/206Pb* age for two zircons from the paragneiss–migmatite boundary of the Winnebach migmatite in the Upper Austroalpine � Otztal–Stubai nappe of the Eastern Alps, Austria (Kl�otzli-Chowanetz et al.29).Width of plotted boxes is the estimated error on an evaporation temperature of ±10 °C. The height of the boxes is calculated as 2s of the mean age of individual temperature steps. Assigned errors of mean ages reported are all 2sm. The examples clearly demonstrate the complementary ‘behaviour’ of zircons during evaporation analysis.Evaporation of crystal 8830-E (Fig. 4) resulted in a perfect age plateau of 484 ± 6 Ma over 8 evaporation steps ranging from 1350 toation of zircon 8830-C (Fig. 4) resulted in a staircase of increasing ages with increasing evaporation temperature. Both zircons were evaporated to completeness. The age plateau of 8830-E is interpreted to represent the crystallisation age of a core-free zircon (at least in respect to Pb isotopic systematics). Based on zircon typology and additional conventional zircon dating, the crystallisation event is attributed to the migmatite formation during the Ordovician.Zircon 8830-C exhibits (within error for the two first evaporation steps) the same age for the migmatite formation event at 480 ± 6 Ma. The single step ages of 561 and 632 Ma possibly represent older metamorphic events. Similar but better defined ages were found in other zircons from the same locality, further supporting this interpretation. The higher temperature staircase ( > 1460 °C) is interpreted as representing a mixture of an old, possibly Archean Pb component with Pb of Cambrian or Proterozoic age.If no additional information for the two older metamorphic events would exist, the age staircase must be interpreted as representing a mixture between the Archean and the Ordovician Pb component. The age of 2355 ± 85 Ma is interpreted as representing a minimum age estimate for the formation or recrystallisation of an inherited zircon core.The evaporation temperature profile of zircon 8830-C directly proves within one zircon crystal the existence of differently old 1246 Analyst, November 1997, Vol. 122Pb components with individual levels of activation energy which can very effectively be separated by evaporation analysis (Kl�otzli18). Example 2 (Fig. 5) presents two plots of block number versus 207Pb*/206Pb* and 208Pb*/206Pb* for two zircons from granitoids of the South Bohemian Pluton, Austria (Kl�otzli and Parrish,22 Kl�otzli et al.33) The block number gives the progress of evaporation with steps of increasing temperature as indicated by labelled evaporation temperatures (in °C).Each block represents the mean of 10 mass scans. For easier reading errors on individual blocks are not shown. They are in the range of the symbol size. Assigned errors of mean ages reported are 2sm. Both zircons demonstrate the presence of an inherited core and a later overgrowth.Inherited cores and overgrowth could be analysed by at least two evaporation steps thus providing plateau ages which can be interpreted as being geologically meaningful. For zircon 4690-A both ages (336 ± 4 Ma and 635 ± 12 Ma, respectively) are interpreted to represent magmatic growth events. This interpretation is further supported by the large intra- and inter-evaporation step variation in 208Pb*/206Pb* at constant 207Pb*/206Pb* indicative for magmatic Th/U variation in the evaporated zircon domain.The analysis of the deposit of the third evaporation step of 4690-A (at 1515 °C) shows the typical pattern of a reversed deposit (Kl�otzli18), direct evidence for the mixing of differently old zircon domains during evaporation. The 208Pb*/ 206Pb* spectrum obtained for the inherited core of 2E92-B (535 ± 8 Ma) does not show significant variation. This reflects constant Th/U ratios throughout the inherited core. This is interpreted as reflecting crystal homogenisation during a high temperature metamorphic overprint leading to the formation of charnockitic rocks (Kl�otzli et al.32).The exact meaning of the Variscan overgrowth (352 ± 3 Ma) is still a matter of debate. Outlook To further enhance the precision of 207Pb*/206Pb* ages from zircon evaporation analysis static SEM-IC data acquisition using multi-collector SEM-IC systems has to be established. It should be possible to achieve the same precision for Pb isotopic analysis as is routinely found for multi-collector Faraday systems (i.e., 10 times better in precision as at present).One possible way of upgrading is by substituting the Faraday cups of a MAT 262 with ion-counting channeltrons as demonstrated by Fig. 4 Plots of evaporation temperature versus 207Pb*/206Pb* age for two zircons from the paragneiss–migmatite boundary of the Winnebach migmatite in the Upper Austroalpine � Otztal-Stubai nappe of the Eastern Alps, Austria (Kl�otzli-Chowanetz et al.29).Width of plotted boxes is estimated error on evaporation temperature of ±10 °C. Height of boxes is calculated as 2s of the mean age of individual temperature steps. For discussion see text. Fig. 5 Plots of block number versus 207Pb*/206Pb* and 208Pb*/206Pb* for two zircons from granitoids of the South Bohemian Pluton, Austria (Kl�otzli and Parrish,22 Kl�otzli et al.33) Block number gives progress of evaporation with steps of increasing temperature as indicated by labelled evaporation temperatures (in °C). Each block represents the mean of 10 mass scans. For easier reading errors on individual blocks are not shown. They are in the range of the symbol size. Assigned errors of mean ages reported are 2sm. For discussion see text. Analyst, November 1997, Vol. 122 1247Richter et al.43 or by using a newly designed Wien-filter TIMS (Laue et al.44). Additionally, the zircon mounting and encasing procedures can substantially be improved by using preformed Re filaments and micro-manipulators. The author thanks the following for spending tedious hours at the MS, for discussion, criticism, advice, a steady hand during zircon mounting, and critical reviews of an earlier version of this paper: F. Bernhard, E. Chowanetz, W. Frank, V. H�ock, G. Hoinkes, M. Jelenc, S. Meli, B. M�uller. Financial support by the Austrian Science Foundation is also acknowledged. References 1 Lectures in Isotope Geology, ed. Hunziker, J. C., and J�ager, E., Springer Verlag, Berlin, 1979. 2 Faure, G., Principles of Isotope Geology, 2nd edn., Wiley, New York, 1986. 3 Geyh, M., and Schleicher, H., Absolute Age Determination: Physical and Chemical Dating Methods and Their Application, Springer Verlag, Berlin, 1990. 4 Mattinson, J. M., Chem. Geol., 1987, 66, 151. 5 Steiger, R., Bickel, R. A., and Meier, M., Terra Abstr., 1993, 1–5, 395. 6 Wendt, J. J., and Todt, W., Terra Abstr., 1991, 3, 507. 7 Poller, U., Liebetrau, V., and Todt, W., in Proceedings of the V. M. Goldschmidt Conference, Heidelberg, 1996, volume 1, p. 119. 8 Compston, W., Williams, I. S., and Clement, S. W., in Proceedings of the 30th American Society of Mass Spectrometry Conference, 1982, p. 593. 9 Wiedenbeck, M., and Goswami, J. N., Geochim. Cosmochim Acta, 1994, 58, 2135. 10 Jackson, S. E., Longerich, H. P., Horn, I., and Dunning, G. R., in Proceedings of the V. M. Goldschmidt Conference, Heidelberg, 1996, vol. 1, p. 283. 11 Hirata, T., and Nesbitt, R. W., Geochim. Cosmochim. Acta, 1995, 59, 2491. 12 Kosztolanyi, C., Compt. Rend. Acad Sci., 1965, 261, 5849. 13 Sunin, L. V., and Malyshev, V. I., Geochem. Int., 1983, 20, 34. 14 Kober, B., Contrib. Mineral. Petrol., 1997, 93, 482. 15 Gentry, R. V., Sworski, T. J., McKown, H. S., Eby, R. E., and Christie, W. H., Science, 1982, 216, 296. 16 Kober, B., Contrib. Mineral. Petrol, 1996, 96, 63. 17 Steiger, R. H., and J�ager, E., Earth Planet. Sci. Lett., 1977, 36, 359. 18 Kl�otzli, U. S., Chem. Geol., 1997, submitted for publication. 19 Ansdell, K. M., and Kyser, T. K., Am. Min., 1993, 78, 36. 20 Roddick, J. C., and Chapman, H. J., EOS, Trans. Am. Geophys. Union, 1991, 72, 531. 21 Roddick, J. C., USGS Circular, 1994, 1107, 269. 22 Kl�otzli, U. S., and Parrish, R. R., Mineral. Petrol., 1996, 58, 197. 23 Cameron, A. E., Smith, D. E., and Walker, R. L., Anal. Chem., 1969, 41, 525. 24 Krogh, T. E., Geochim. Cosmochim. Acta, 1982, 46, 637. 25 Kl�otzli, U. S., Mitt. � Osterr. Miner. Ges., 1994, submitted for publication. 26 Kl�otzli, U. S., Mitt. � Osterr. Miner. Ges., 1993, 138, 123. 27 Tilton, G. R., J. Geophys. Res., 1960, 65, 2933. 28 Shestakov, G. I., Geochim. Int., 1972, 9, 801. 29 Kl�otzli-Chowanetz, E., Kl�otzli, U. S., and Koller, F., Schw. Miner. Petrol. Mitt., 1997, in the press. 30 Ansdell, K. M., and Kyser, T. K., Geology, 1991, 19, 518. 31 Todt, W., Cliff, R. A., Hanser A., and Hofmann, A. W., Geophys. Monograph, 1996, 95, 429. h Planet. Sci. Lett., 1975, 26, 207. 33 Kl�otzli, U. S, Koller, F., Scharbert, S., and H�ock, V., Chem. Geol., 1997, submitted for publication. 34 York, D., Earth Planet. Sci. Lett., 1969, 5, 320. 35 Ludwig, K. R., Earth Planet. Sci. Lett., 1980, 46, 212. 36 Roddick, J. C., Loveridge, W. D., and Parrish, R. R., Chem. Geol., 1987, 66, 111. 37 Ludwig, K. R., USGS Open-file Report, 1992, 91-445. 38 Kr�oner, A., and Seng�or, A. M. C., Geology, 1990, 18, 1186. 39 Kr�oner, A., Todt, W., Hussein, I. M., Mansour, M., and Rashwan, A. A., Precambrian Res., 1992, 59, 15. 40 M�uller, B., Kl�otzli, U. S., and Flisch, M., Geol. Rundschau, 1995, 84, 457. 41 Peindl, P., and H�ock, V., Terra Abstr., 1993, 1/5, 392. 42 Bernhard, F., Kl�otzli, U. S., Hoinkes, G., and Th�oni, M., Mineral. Petrol., 1996, 58, 171. 43 Richter, S., Ott, U., and Begemann, F., Int. J. Mass Spec. Ion. Proc., 1994, 136, 91. 44 Laue, H. J., Tegtmeyer, A., and Wegener, M., Spectromat Inf. Sheet, 1995, 1. Paper 7/04114D Received June 12, 1997 Accepted September 1, 1997 1248 Analyst, Nov
ISSN:0003-2654
DOI:10.1039/a704114d
出版商:RSC
年代:1997
数据来源: RSC
|
|