Establishing an SI-traceable Copper Concentration in the Candidate Reference Material MURST ISS A1 Antarctic Sediment Using Isotope Dilution Applied as a Primary Method of Measurement I. PAPADAKIS†, P. D. P. TAYLOR AND P. DE BIE` VRE Institute for ReferenceMaterials and Measurements, European Commission-JRC, B-2440 Geel, Belgium Traceability is a term that is heavily debated world-wide in the material, are the subject of dierent reports prepared by the analytical chemistry community.This paper describes an organisers of the project. attempt to obtain an SI-traceable value for the Cu According to the International Vocabulary of Basic and concentration in the Candidate Reference Material MURST- General Terms in Metrology (VIM) definition,2 traceability is: ISS A1 Antarctic Sediment. This material was collected by the ‘property of the result of a measurement whereby it can be Instituto Superiore di Sanita (ISS, Rome, Italy) and was related to stated references, usually national or international processed at the Institute for Reference Materials and standards, through an unbroken chain of comparisons all Measurements (IRMM, Geel, Belgium) into a homogeneous having stated uncertainties’.and dried powder. The analytical method used was isotope In our opinion, such ‘stated uncertainties’ should be as dilution (ID) combined with inductively coupled plasma mass complete and detailed as possible. Here, we present two spectrometry (ICP-MS).Microwave pressurised digestion and approaches to this subject using the EURACHEM and separation of Cu by ion-exchange chromatography were used. International Organisation for Standardisation (ISO)/Bureau The International Vocabulary of Basic and General Terms in International des Poids et Measures (BIPM) guides for the Metrology (VIM) definition of traceability requires ‘stated establishment of ‘full’ uncertainty. uncertainties’. Because a primary method of measurement A complete uncertainty statement is also needed for the (ID) is used, an attempt was made to make these ‘stated result obtained by primary methods of measurement.A priuncertainties’ as detailed as possible, thereby using the mary method has been defined by the Comite� Consultatif pour International Organisation for Standardization (ISO)/Bureau la Quantite� de Matie`re (CCQM)3 of the BIPM as: ‘a method International des Poids et Measures (BIPM) guide and taking having the highest metrological qualities, whose operation into account all possible sources of uncertainty (Type A and can be completely described and understood, for which a Type B).The established value for the Cu concentration is complete uncertainty statement can be written down in terms 86.7 nmol g-1 with an expanded uncertainty of 4.9 nmol g-1 of SI units, and whose results are, therefore, accepted without (coverage factor k=2). reference to a standard of the quantity being measured’.Practically, this means ‘being able to write down an explicit Keywords: T raceability; primary method; isotope dilution mass equation describing what is measured to what is intended to spectrometry; copper isotope ratio; inductively coupled plasma be measured without containing any (significant) empirical mass spectrometry; SiCl+ interference; Antarctic sediment correction factors’. The BIPM concept of primary methods, although new in chemistry, is applied in other measurement Traceability is a tool to achieve comparability of measuresciences. ments.This is important in the context of many border crossing Isotope dilution (ID or IDMS for isotope dilution mass issues related to measurement values in trade and industry. spectrometry) was recognised3 by CCQM as a method that When traceability is towards a common base such as the SI, has the potential to be primary if carried out correctly (e.g., it it is called SI-traceability. Linked with the concept of does not automatically produce correct results).Primary SI-traceability is the notion that values are stable in place and methods, and hence ID are, however, unique tools in chemical time and not dependent on the procedure used. This is, for measurements which can, from an a priori point of view, lead instance, considered important in many legal, environmental to results with small expanded uncertainty.4,5 The influence of and climatological issues.1 matrix eects caused by complicated matrices, such as for this In this paper we propose a way to achieve this traceability sediment, is far more dicult to identify and estimate quantitat- to the SI in the determination of the Cu concentration in the ively when using procedures not based on measuring isotope candidate reference material MURST ISS A1 Antarctic amount ratios.Moreover, measuring isotope amount ratios Sediment. The results presented are only the contribution of has the additional advantage of making the procedure inde- the Stable Isotope Measurement (SIM) Unit of the Institute pendent of other sources of possible uncertainties, such as for Reference Materials and Measurements (IRMM) to the analyte loss, occurring during the sample treatment (digestion, certification campaign of the material, which was carried out separation, transfer of the sample, etc.).under the responsibility of the Instituto Superiore di Sanita It is critical when using isotope mass spectrometry to verify (ISS).Other methods were also used for this purpose. Those the absence of possible isobaric interference. This problem results, as well as the certified value of the new reference exists independently of the calibration strategy used (e.g., IDMS, external calibration, standard additions) and usually causes many problems in analyses. In this work, a quadrupole † EC fellow. Journal of Analytical Atomic Spectrometry, August 1997, Vol. 12 (791–796) 791inductively coupled plasma mass spectrometry (ICP-MS) sample, Ny=known number of atoms of the element in the spike, Ry=isotope amount ratio of the chosen isotope pair in instrument was used.It is well known that the Cu mass spectrum suers from many isobaric interferences. A well the spike, Rb=isotope amount ratio of the chosen isotope pair in the blend, Rx=isotope amount ratio of the chosen isotope known case is 40Ar23Na+ interfering on 63Cu,6 when Na is present at high concentration in the samples.Ion-exchange pair in the sample, SRix=sum of the isotope amount ratios in the sample, SRiy=sum of the isotope amount ratios in the chromatography was applied to the samples in order to remove any such interferences. Since, however, other possible inter- spike, cx=unknown concentration of the element in the sample, cy=known concentration of the element in the spike, my= ferences are not well documented in the open literature, a detailed study of interferences was performed on the samples mass of spike material used for the blend preparation and mx=mass of sample material used for the blend preparation.before and after the separation. When applied to the Cu isotope ratio R=n(63Cu)/n(65Cu), where n is the amount expressed in moles, the concentration EXPERIMENTAL cx of this element in the samples can be evaluated from eqn. (2). It takes into account mass discrimination eects during Instrumentation the measurement (mass discrimination correction factor=K) For the measurements of the isotope amount ratios in the as well as the moisture content of the sediment [moisture blends (spiked samples), ICP-MS was used.The instrument correction factor w=1-(mass of water/mass of sample)]. used is manufactured by Fisons (Fisons VG PlasmaQuad 2+) and was slightly modified at IRMM. It is equipped with cx= Ry-K Rb(obs) K Rb(obs)-Rx cy my w mx Rx+1 Ry+1 (2) Balzers turbo pumps and a V-groove nebuliser.Argon was used for the operation of the plasma torch. Further technical characteristics and operating conditions are presented in Table 1. For the digestion of the samples, a microwave oven Weighing of the Samples and Blend Preparation (Milestone MLS-1200 MEGA) with a microwave digestion The weighing of the samples was performed in a humidityrotor (1000/6/100/110) (six-position, pressure up to 110 bar) conled environment (humidity stable between 50 and 60%) was used.in order to avoid absorption of water content into the sediment powder from the humidity of the atmosphere. In addition, the Sample sample was immediately weighed in the PFA digestion vessels to avoid transfer to dierent containers and eliminate possible The candidate reference material MURST ISS A1 Antarctic losses and additional uncertainties to the final result. Two Sediment originates from the Terra Nova Bay (Antarctica). dierent bottles of the material were used and three fractions The certification campaign was carried out by the ISS (Italy), of 100 mg were taken from each bottle.whose delegation performed the sampling. The material was For the spiking of the samples, the certified spike reference dispatched to IRMM where it was processed7 into a fine material IRMM-63212 enriched in 65Cu having a ratio (<150 mm) sediment powder. The moisture content of the n(63Cu)/n(65Cu)=0.0028921 with an expanded uncertainty of material was measured to be 0.35% with a standard uncertainty 0.0000086 (k=1) and concentration cy=0.0437 mmol g-1 Cu of 0.05% using Karl-Fisher titration in ten dierent bottles of with an expanded uncertainty of 0.0002 mmol g-1 (k=1) was the material.7 A homogeneity study8 for Cu was performed used.The concentration of the spike was calculated graviusing solid sampling Zeeman-eect background corrected metrically and confirmed using reverse ID, when the isotopic atomic absorption spectrometry (SS-ZAAS), carried out composition was measured by means of thermal ionisation on 20 dierent bottles of the material, which gave for Cu a mass spectrometry (TIMS).To minimise possible problems homogeneity factor HE=5.8%Ómg.9 due to multiplier dead time the blend isotope amount ratio (Rb) was kept close to unity. Procedure The spike was added to the sample material, which was weighed in the PFA vessel, before any further treatment of the The general IDMS equation [eqn. (1)] as can be found in the sample.For each blend, approximately 0.1 g of sample material literature10,11 is: and 0.08 g of IRMM-632 material were used. All the weighings were performed by the IRMM mass Nx Ny = (Ry-Rb) (Rb-Rx) S Rix S Riy metrology department using substitution measurements against operational mass standards, which are calibrated against an IRMM stainless steel mass standard, which itself is u cx = (Ry-Rb) (Rb-Rx) (cy my) mx S Rix S Riy (1) calibrated against a stainless steel BIPM mass standard.The latter is finally calibrated against the platinum–iridium primary where Nx=unknown number of atoms of the element in the mass standard at BIPM. A certificate accompanied each individual weighing where the weighed mass and its total uncertainty (taking into account buoyancy correction) was Table 1 Operating conditions of the quadrupole ICP-MS instrument given. used for the isotopic measurements Electron multiplier Galileo 4870V Digestion Procedure Dead time/ns 13±1 Nebulizer type V-groove The optimised microwave digestion procedure consisted of a Nebulizer flow/l min-1 #0.8 simplified one-step procedure to ensure homogeneity between Sample aspiration rate/ml min-1 #1 sample and spike materials.Forward power/W 1400 Plasma gas flow/l min-1 #14 The acid mixture used was 5 ml of HNO3 (#14 M) and HF Intermediate flow/l min-1 #1.5 (#20 M) in the proportion 1+1 (v/v). The heating programme Data acquisition mode Peak jump, 3 points per peak was a five-step, 7 min each, programme with 250W power Dwell time/ms 10.24 used in the odd steps and no power used in the even steps. Acquisition time/s 6×60 No pressure indication was available but maximum admissible 792 Journal of Analytical Atomic Spectrometry, August 1997, Vol. 12pressure in the vessels is 110 bar. Both acids used were of subboiled quality and prepared in the IRMM Ultra Clean Chemical Laboratory (UCCL). After heating under pressure in the digestion vessels, the samples were evaporated to dryness.Finally, the blends were dissolved in about 50 g of 0.14 M HNO3. Separation The aim of the separation was to remove the alkali and alkaline earth metals from the blends. Bio-Rad AG 1-X8 anionexchange resin in the chloride form was used. The columns were about 2 ml in volume. The separation was carried out in concentrated HCl (7.5 M) of sub-boiled quality: a 5 g fraction of each blend (in 0.14 M HNO3 matrix) was first evaporated and then redissolved in 2 ml of 7.5 M HCl prior to the separation.The column was flushed with 10 ml of 7.5 M HCl (to remove alkali and alkaline earth metals). About 10 ml of Fig. 1 Theoretical eect of SiCl+ isobaric ions on the measurement 0.14 M HNO3 were used to elute the Cu from the column. of the ion current ratio I15I2 (I1: m/z=63; I2: m/z=65) in natural and non-natural Cu (blend) and on the apparent Cu concentration (the arrows are indicate the y-axis used for each trend-line). 2 Ion current Measurements ratio on mass positions 63 and 65 in natural Cu solution (slope= 0.47). & Ion current ratio on mass positions 63 and 65 in blends The sample acquisition time was 6×60 s and between samples (slope=1.54). + Apparent Cu concentration calculated using IDMS. there was a minimum of 3 min washout time, which is adequate to avoid carry-over from one sample to the next. At the beginning of the measurement, a blank solution (0.14 M HNO3) was measured and from all subsequent measurement values, this blank value was subtracted. Every two blends, a natural Cu solution (ASARCO metal dissolved in sub-boiled 0.14 M HNO3) with an approximate concentration of 100 nmol g-1, prepared at IRMM, was measured in order to determine and apply the correction factor for mass discrimination.Interference Study Because of the possible interference of ArNa+, Na was first removed from the sample as described above. After this separation, some other possible interferences were studied in detail.The interference most dicult to detect, not widely reported in the literature, is the interference of SiCl+. Only Fig. 2 Observed eect of Si on the ratio of ion currents at mass positions 63 and 65, as measured in the blends (second-order Vanhaecke et al.13 have reported it. The eect of this interpolynomial fit). ference can be simulated. This is eected by calculating the ‘abundance’ of the SiCl+ species at the dierent mass positions problem to a level where it was no longer detectable as by multiplying the abundances of the Si isotopes by the substantiated by measurements on the separated samples.abundances of the Cl isotopes according to combination Additionally, after removal of the two influencing interferences theory. The natural abundances of Cu and the ‘theoretical (ArNa+ and SiCl+), the samples were measured for isotopic abundances’ of SiCl+ are found to be very similar.Fig. 1 composition (unspiked) and this was found to be identical (calculated from theoretical isotope abundance ratios14) clearly with the natural isotopic composition, which indicates the shows the problem of detecting this interference, which absence of other interferences (e.g., ZnH, SO2). approaches the natural Cu isotope ratio. Moreover, Fig. 1 shows that the eect of this interference is much more severe on blend (close to unity) Cu ratios (factor of 3) and it is shown RESULTS that the eect on the apparent Cu concentration is even larger.This was confirmed experimentally by spiking six blends with The Cu concentration in the material established with this procedure is 86.7 (4.9) nmol g-1 (the number in parentheses is dierent amounts of Si. The results are presented in Fig. 2. In addition, high resolution (HR)-ICP-MS was used to investigate the expanded uncertainty with k=2). The results from the measurements on each individual this problem, as has been described in a companion paper,15 in more detail.The study was carried out at the University of digestion are given in Table 2. These results are not corrected for the procedural blank. The Cu blank of the procedure used Gent, Laboratory of Analytical Chemistry, and confirmed our indirect prediction of this interference. was determined using ID as being 2.5 nmol g-1 with a standard uncertainty of 0.1 nmol g-1. This value was then subtracted The presence of SiCl+ species in the solution, after ionexchange, can be explained by the presence of Si in the from the measured Cu concentration in the sediments and its standard uncertainty added to the combined uncertainty of sediments (mostly silicate minerals) and Cl from using an ionexchange column in the chloride form.Early digests (using the result. In the following sections, the calculation of the uncertainty, dierent digestion procedures, still leaving trace amounts of Si and Cl) showed variable concentrations up to 120 nmol g-1 according to the ISO/BIPM guide,16 is described in detail.The ISO/BIPM document is a consensus document developed for Cu. The final digestion procedure as described (which uses HF in the digestion, forming volatile SiF4, which volatilises when measurements in general, which was only finalised after many years of discussion. It gives guidance on how to evaluate and the solution is evaporated to near dryness) minimised this Journal of Analytical Atomic Spectrometry, August 1997, Vol. 12 793Table 2 Results obtained from the measurement on each individual isobaric interference correction factor Kiso digestion (blend) after moisture content correction. For the relevant uncertainties consult the uncertainty budgets cx=H CRy-K Kiso Rb(obs) K Kiso Rb(obs)-Rx cy my w mx Rx+1 Ry+1D (3) Sample name Bottle No. c (Cu)/nmol g-1 As a basis for the H factor, the independent homogeneity I960620A 92.9 study was taken.8 For a sample size of 100 mg, the variability I960620B 50 89.3 I960620C 87.1 of the results obtained due to the homogeneity of the material I960620D 88.5 is expected to be 0.6%; hence, H=1.000±0.006. I960620E 75 86.6 The estimation of Kiso is more complicated.From the I960620F 90.6 HR-ICP-MS scans, the ratio of the SiCl5Cu signal was estimated (based on the same assumption as Fig. 1). Assuming that this is representative of the amount ratio, this value was then introduced into Fig. 1 and the influence on Rb was calculated. This was found to be 1.4%; hence, Kiso= express uncertainty. One of the important aspects introduced 1.000±0.014. is the distinction between Type A and Type B sources of un- With these data, a new uncertainty budget (Table 5) was certainty. ‘Type A evaluation is a method of evaluation of a produced using eqn. (3), resulting in a combined uncertainty standard uncertainty by the statistical analysis of series of similar to the original uncertainty. This confirms that the observations’ whereas ‘Type B is a method of evaluation of a magnitude of the reproducibility (Approach I) can be explained standard uncertainty by means other than by the statistical by taking into account the homogeneity and remaining varia- analysis of series of observations’ (e.g., generated by certificates, bility caused by the spectral interference of SiCl+ (Approach or when assessing the influence of parameters, as was carried II).out in this work for SiCl+ interference). The important steps in the uncertainty estimation process are: defining the measurand in a mathematical equation [in DISCUSSION our case eqn. (2)], identifying the sources of uncertainty, The somewhat vague VIM definition for traceability mentions estimating their magnitude and finally, combining the uncer- ‘stated uncertainties’. In our interpretation, to be useful, this tainty contributions. The first three steps are critical and often means a complete uncertainty budget where all uncertainty the most dicult, whereas the final step is merely a trivial sources are described is needed.Obviously, this requirement mathematical problem, for which there often exist good needs to be translated into reality, as completeness can only approximations (see Appendix). be reached asymptotically. Indeed, as an example, in this work, In the following paragraphs, the combined uncertainty for the isobaric interferences ArNa+ and SiCl+ were investigated the measured Cu concentration in this material is determined in depth, but many others, which can be found in the inter- using two dierent approaches.In the first approach, part of ference tables (e.g., ZnH, SO2), were not investigated in depth the uncertainty of the final value is simply derived from the because they seem to be negligible based on the experimental reproducibility of the six obtained IDMS results. In a second, evidence that the isotopic composition of the samples was more refined approach, all available knowledge on uncertainty found to be identical with the natural isotopic composition.sources (e.g., material inhomogeneity and variability of the This is why, in the context of the ISO/BIPM uncertainty SiCl+ interference, both of which were studied and quantified philosophy, the responsibility for the uncertainty finally as described above) were incorporated in the uncertainty remains with the person performing the analysis.Obviously, budget. the major advantage of the uncertainty budget (Tables 3–5) is to have a list of sources considered, which can be the starting Uncertainty Calculation Approach I point of discussion should problems arise. Tables 3 and 4 list the dierent uncertainty sources considered. As can be seen, an uncertainty statement is made for each of the quantities as defined in the IDMS equation [eqn. (2)]. Table 3 Uncertainty budget for Approach I using the simplified Most of these are Type A uncertainties.When these sources approach of addition of relative variances are combined, this yields an uncertainty of 2.8%, whereby the Parameter Typical value SU* RSU† (%) reproducibility of IDMS on the six digestions is by far the dominant contribution. Therefore, in an empirical, pragmatic Weighing data— approach, one could simply consider this reproducibility as mx (g) 0.106 0 0.000 5 0.5 my (g) 0.094 5 0.000 5 0.5 reflecting the uncertainty originating from all undefined Certificate data— sources. cy (Cu/mmol g-1) 0.043 7 0.000 2 0.4 Ry 0.002 9 9×10-6 0.3 Measurement data— Uncertainty Calculation Approach II Rb 0.894 3 0.001 8 0.2 Rx 2.243 6 0.002 2 0.1 In an attempt to refine the uncertainty budget further, an eort K 1.055 8 0.001 3 0.1 was made to identify the uncertainty sources as apparent in Other data— the large reproducibility (from the repeated IDMS).In the w 0.996 5 0.000 5 0.05 meantime, data from SS-ZAAS on the homogeneity of the Measurement blank subtraction <0.01 material became available.8 Furthermore, the theoretical esti- Reproducibility 2.64 Preliminary uncertainty on mate of the extent of isobaric interference was used to assess cx (Cu/mmol g-1) 0.089 2 0.002 4 2.79 the uncertainty introduced from this source.Although no Procedural blank (mmol g-1) 0.002 5 0.000 1 4.75 correction was carried out on the Cu concentration due to the Combined uncertainty uc 2.79 isobaric interference (because experimentally a change in iso- Expanded uncertainty U (k=2) 5.58 tope ratio could not be detected), an uncertainty was added to the combined uncertainty.Eqn. (2) was fine-tuned, resulting *Standard uncertainty. †Relative standard uncertainty. in eqn. (3), which includes a homogeneity factor H and an 794 Journal of Analytical Atomic Spectrometry, August 1997, Vol. 12Table 4 Uncertainty budget for Approach I, combining uncertainty contributions using the EURACHEM spreadsheet approach Parameter Reproducibility Blank Rb Ry Rx cy mx my K w Typical value 1 0.002484 0.8943 0.002892 2.2436 0.043723 0.106 0.0945 1.0558 0.9965 SU* 0.0264 0.000118 0.0018 8.6×10-6 0.0022 0.000183 0.0005 0.0005 0.0013 0.0005 Reproducibility 1.0264 1 1 1 1 1 1 1 1 1 1 Blank 0.002484 0.002602 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 Rb 0.8943 0.8943 0.8961 0.8943 0.8943 0.8943 0.8943 0.8943 0.8943 0.8943 0.8943 Ry 0.002892 0.002892 0.002892 0.002901 0.002892 0.002892 0.002892 0.002892 0.002892 0.002892 0.002892 Rx 2.2436 2.2436 2.2436 2.2436 2.2458 2.2436 2.2436 2.2436 2.2436 2.2436 2.2436 cy 0.043723 0.043723 0.043723 0.043723 0.043723 0.043906 0.043723 0.043723 0.043723 0.043723 0.043723 mx 0.106 0.106 0.106 0.106 0.106 0.106 0.1065 0.106 0.106 0.106 0.106 my 0.0945 0.0945 0.0945 0.0945 0.0945 0.0945 0.0945 0.095 0.0945 0.0945 0.0945 K 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0571 1.0558 1.0558 w 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.997 0.9965 cx 0.091518 0.089046 0.089483 0.089162 0.089071 0.089547 0.088734 0.089649 0.089359 0.089118 0.089164 SUi 0.002354 -0.00012 0.00032 -1.6×10-6 -9.3×10-5 0.000384 -0.00043 0.000485 0.000195 -4.6×10-5 (SUi)2 5.54×10-6 1.39×10-8 1.02×10-7 2.63×10-12 8.62×10-9 1.47×10-7 1.85×10-7 2.35×10-7 3.82×10-8 2.11×10-9 S(SUi2) 6.27×10-6 cx/mmol g-1 SU/mmol g-1 RSU(%) 0.089164 0.002505 2.80905 * Standard Uncertainty.Table 5 Uncertainty budget for Approach II, combining uncertainty contributions using the EURACHEM spreadsheet approach Parameter Homogeneity Blank Kiso Rb Ry Rx cy mx my K w Typical value 1 0.002484 1 0.8943 0.002892 2.2436 0.043723 0.106 0.0945 1.0558 0.9965 SU* 0.006 0.000118 0.0144 0.0018 8.6×10-6 0.0022 0.000183 0.0005 0.0005 0.0013 0.0005 Homogeneity 1.006 1 1 1 1 1 1 1 1 1 1 1 Blank 0.002484 0.002602 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 0.002484 Kiso 1 1 1.0144 1 1 1 1 1 1 1 1 1 Rb 0.8943 0.8943 0.8943 0.8961 0.8943 0.8943 0.8943 0.8943 0.8943 0.8943 0.8943 0.8943 Ry 0.002892 0.002892 0.002892 0.002892 0.002901 0.002892 0.002892 0.002892 0.002892 0.002892 0.002892 0.002892 Rx 2.2436 2.2436 2.2436 2.2436 2.2436 2.2458 2.2436 2.2436 2.2436 2.2436 2.2436 2.2436 cy 0.043723 0.043723 0.043723 0.043723 0.043723 0.043723 0.043906 0.043723 0.043723 0.043723 0.043723 0.043723 mx 0.106 0.106 0.106 0.106 0.106 0.106 0.106 0.1065 0.106 0.106 0.106 0.106 my 0.0945 0.0945 0.0945 0.0945 0.0945 0.0945 0.0945 0.0945 0.095 0.0945 0.0945 0.0945 K 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0558 1.0571 1.0558 1.0558 w 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.9965 0.997 0.9965 cx 0.089699 0.089046 0.091471 0.089483 0.089162 0.089071 0.089547 0.088734 0.089649 0.089359 0.089118 0.089164 SUi 0.000535 -0.00012 0.002307 0.00032 -1.6×10-6 -9.3×10-5 0.000384 -0.00043 0.000485 0.000195 -4.6×10-5 (SUi)2 2.86×10-7 1.39×10-8 5.32×10-6 1.02×10-7 2.63×10-12 8.62×10-9 1.47×10-7 1.85×10-7 2.35×10-7 3.82×10-8 2.11×10-9 S(SUi2) 6.34×10-6 cx/mmol g-1 SU/mmol g-1 RSU(%) 0.089164 0.002518 2.824015 * Standard Uncertainty.Journal of Analytical Atomic Spectrometry, August 1997, Vol. 12 795CONCLUSION Measurements unit for his help in the development of the separation procedure and to F.Vanhaecke from Gent This paper attempts to explain how SI-traceable values can be University for performing some very useful high resolution obtained for the measurement of the Cu concentration in a ICP-MS measurements. The whole project is in the framework sediment sample. This is established by making explicit the of the collaboration between IRMM and ISS. The MURST relationship between the measurand and the measured quantit- ISS-A1 Antarctic Sediment was prepared and certified under ies (isotope ratios, amounts), thereby including the relevant Italy’s National Programme for Research in Antarctica uncertainties in as much detail as possible.In this way, the (PNRA, Programma Nationale per la Ricerca in Antartide). CCQM definition of a primary method is applied in a real life situation. This highlights the transparency and relative ease with which uncertainty calculation can be performed, only for REFERENCES primary methods of measurement. 1 De Bie`vre, P., Kaarls, R., Peiser, H.S., Rasberry, S. D., and Reed, As is well known, ICP-MS, especially using quadrupole W. P., Accred. Qual. Assur. 1996, 1, 3. instruments, is very susceptible to isobaric interferences. In 2 International Vocabulary of Basic and General T erms in Metrology, particular, for bi-isotopic elements this can be problematic. In International Organisation for Standardisation, Gene`ve, 1993. this case, a relatively unknown interference from SiCl+ could 3 Comite� Consultatif pour la Quantite de Matie`re, Rapport de la have caused serious errors, if not properly identified.This was 1re session, 1995, E� dite� par le BIPM, Pavillon de Breteuil, F-92312 avoided by a metrological investigation, involving a careful, Se`vres Cedex, France. 4 Fassett, J. D., and Paulsen, P. J., Anal. Chem., 1989, 61, 643A. detailed and systematic study of the matrix and the subsequent 5 Moody, J. R., and Epstein, M. S., Spectrochim.Acta, Part B, 1991, removal of the interference. The uncertainty due to the possible 46, 1571. presence of undetectable amounts of SiCl+ interference was 6 Lyon, T. D. B., Fell, G. S., J. Anal. At. Spectrom., 1990, 5, 135. quantified by means of a Type B uncertainty evaluation. 7 Kramer, G., personal communication. 8 Grobecker, K. H., personal communication. 9 Kurfu� rst, U., Grobecker, K. H., and Stoeppler, M., In: T race APPENDIX Elements—Analytical Chemistry in Medicine and Biology, ed.There are several ways to combine uncertainties mathemat- Bra�ter, P., and Schramel, P., Walter de Gruyter, Berlin, 1984, Vol. 3, pp. 591–601. ically. The strictest way is a vigorous application of the 10 De Bie`vre, P., Fresenius’ J. Anal. Chem., 1990, 337, 766. uncertainty propagation law, which is cumbersome. An alterna- 11 De Bie`vre, P., Fresenius’ J. Anal. Chem., 1994, 350, 277. tive simplified approach is to add all the relative variances 12 Certificate for IRMM-632 spike isotopic reference material, in (Table 3) into a combined uncertainty uc. In this case, this preparation. approach gives a slightly overestimated combined uncertainty. 13 Vanhaecke, F., Vanhoe, H., Moens, L., and Dams, R., Bull. Soc. A third way consists of using the spreadsheet approach (Tables Chim. Belg., 1995, 104, 653. 14 International Union of Pure and Applied Chemistry, Commission 4 and 5) of the EURACHEM guide17 for uncertainty calcuon Atomic Weights and Isotope Abundances, Isotopic lation. This spreadsheet approach is a numerical method, Composition of Elements, Pure Appl. Chem., 1991, 63, 991. whereby the influence of a change in one of the uncertainty 15 Vanhaecke, F., Moens, L., Dams, R., Papadakis, I., and Taylor, components is calculated, for each of these components. P., Anal. Chem., 1997, 69, 268. Finally, these uncertainties are combined, again by adding the 16 Guide to the Expression of Uncertainty in Measurement, variances. International Organisation for Standardisation, Gene`ve, 1993. 17 Quantifying Uncertainty in Analytical Measurement, EURACHEM, London, 1995. Special thanks are due to B. Dijckmans from IRMM mass metrology for her help in the weighings, to P. Conneely and P. De Vos from the IRMM Management of Reference Paper 7/00750G Received February 3, 1997 Materials unit for their help in treating the solid sediment and its weighing, to K. Raptis from the IRMM Stable Isotope Accepted May 12, 1997 796 Journal of Analytical Atomic Spectrometry, August 199