JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 187 Application of Inductively Coupled Plasma Mass Spectrometry to the Certification of Reference Materials from the Community Bureau of Reference* Luc Moens Hans Vanhoe Frank Vanhaecke Jan Goossens Michael Campbell and Richard Dams Ghent University Laboratory of Analytical Chemistry Proeftuinstraat 86 B-9000 Ghent Belgium Inductively coupled plasma mass spectrometry (ICP-MS) is widely used for applied analysis. The application of a method to the certification of reference materials is a severe test of its accuracy and precision. We have been using ICP-MS for the certification of Community Bureau of Reference (BCR) reference materials since 1988. This paper indicates that while the method proved to be potentially accurate and precise accuracy and precision cannot be taken for granted and that each analysis requires a thorough study in order to obtain results that meet BCR standards.Under these conditions however ICP-MS can compete with well established reference techniques such as neutron activation analysis. Keywords lnductively coupled plasma mass spectrometry; certification; certified reference materials Since 1983 when instruments for inductively coupled plasma mass spectrometry (ICP-MS) became commercially available the method has been rapidly accepted in both research and application laboratories. The ICP-MS technique is widely used for routine analysis and research is currently being carried out to improve the analytical performance and to extend the field of application of the method.' Most of this research is however concerned with methodological and technical novel- ties new sample introduction methods are being designed and techniques are being developed for the reduction or elimination of spectral and non-spectral interferences.2 The accuracy of the analytical results obtained seems to be taken for granted.The application of ICP-MS for the certification of reference mate- rials is a means of testing this assumption. From the 1970s onwards work has been carried out in this laboratory on the certification of trace element concentrations in the certified reference materials (CRMs) issued by the Community Bureau of Reference (BCR) of the Commission of the European Communities. Considering the 3 1 environmental and food reference materials certified in the period 1982-1990 this laboratory contributed to the certification of about 67% of all certified trace element^.^ Most of the data were obtained with neutron activation analysis (NAA).3 From 1987 onwards when the equipment for ICP-MS became available it was tested and later used for certification work.The contribution of ICP-MS applied in this laboratory to BCR certification work between 1990 and 1993 is shown in Table 1. Except for CRM 090 (doped Ti) all the materials are of biological origin or were certified for their environmental importance. The ICP-MS method was used to test the homo- geneity of the distribution of B in CRM 090 (doped Ti) and was also used in three intercomparison studies organized by BCR. Two of these studies concerned the determination of Pb in different types of wine; in a first inter-comparison only European laboratories took part whereas in a second results of European specialist laboratories were compared with data from laboratories in America and Canada.In a third study trace elements were determined in estuarine water from the river Tejo (Portugal). The analysis and certification of the reference materials mentioned above showed that ICP-MS is sufficiently accurate and precise for BCR certification work. With some exceptions the results obtained with ICP-MS were accepted for the certification on the basis of both their accuracy and precision. This general conclusion is illustrated in Fig. 1 which shows the results that were finally used for the certification of the * Paper presented at the XXVIII Colloquium Spectroscopicum Internationale (CSI) York UK June 29-July 4 1993.concentration of Na in BCR CRM 399 (Fre~hwater).~ Results obtained using ICP-MS are in good agreement with those from well established atomic absorption and emission methods and from instrumental NAA. Moreover the ICP-MS value coincides with the certified value and the precision compares favourably with the precision obtained with other techniques. In this paper the problems associated with the use of ICP-MS in BCR certification work are discussed and methods for overcoming these problems are presented. It is shown that in order to obtain the highest accuracy and precision a detailed study of the possible sources of error must be carried out.Experimental All measurements were carried out with a VG PQI ICP-MS instrument (Fisons VG Elemental Winsford UK) equipped with a Fassel torch a Gilson Minipuls-2 peristaltic pump a Meinhard type Tr-30-A3 concentric glass nebulizer and a double-pass Scott-type spray chamber with surrounding liquid jacket the temperature of which was controlled with a recircul- ating refrigerating-heating system. Sampling cones ( 1.0 mm orifice) and skimmers (0.75mm orifice) were made of nickel. Typical operation conditions are summarized in Table 2. Solutions were prepared with Milli-Q water (Millipore Bedford MA USA) and use was made of reagents of the highest available purity. The nitric acid that was used in most r 0 341 T ICP I NAA T L g 2 6 1 0 Method 0 Fig. 1. Determination of Na in BCR CRM 399 (Freshwater).Concentrations and 95% confidence intervals obtained with different methods (pg g-') and used in the final certification188 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Table 1 Contribution of ICP-MS at Ghent University Belgium to the certification of trace element concentrations in BCR reference materials CRM 101 090 303 304 398 399 403 414 422 142R 143R 145R 063R Product Spruce needles” Titanium” Human serum13 Human serum13 Freshwater (low ~ o n t e n t ) ~ Freshwater (high ~ o n t e n t ) ~ Sea water14 Plankton’’ Cod muscle16 Light sandy soil Light sandy soilt Sewage sludge-amended Sewage sludge-amended7 Sewage sludge Sewage sludget Skimmed Milk Powder Analyte Mg B Nb W Li Mg Li Mg Na Mg Al C1 Ca Mn Na Mg Al CI Ca Mn Fe Mo Mn Cu Zn Cd Pb Mn Cu Cd Pb Mn Cu Pb Mn Cu Zn Cd Pb Mn Cu Zn Cd Hg Pb Mn Cu Zn Cd Pb Mn Cu Zn Cd Hg Pb I Hg Pb Year 1990 1992 1992 1992 1992 1992 1992 1993 1993 * * * * * * * - - - - - - - *Final report in preparation.?Aqua regia soluble fraction. Table 2 Typical operation conditions of the VG PQI spectrometer Plasma r.f. power/” Forward Reflected Plasma Intermediate Nebulizer Gas flow rates/l min-’ Sample uptake rate/ml min-’ Spray chamber temperature/”C Ion sampling depth Vacuum/hPa Expansion stage Intermediate stage Analyser stage 1350 < 5 13.5 0.5-1 0.725-0.8 0.9 10 10 mm (from load coil) 2.4 10-4 4.8 x experiments was purified in the laboratory by sub-boiling distillation. Standards were prepared by dissolving high-purity metals or compounds with known stoichiometry.For isotope dilution studies isotopically enriched certified reference mate- rials were obtained from Isoflex (Consett County Durham UK) for Hg (Hg metal enriched in 201Hg) and from the National Institute of Standards and Technology (NIST Washington USA) for Pb [Pb metal enriched in 206Pb stan- dard reference material (SRM)983 Lead Radiogenic Isotopic]. The certified isotopic composition of spikes was always checked against an independent reference material or against a material with natural abundance. Results and Discussion Choice of Internal Standard It is common practice to add an internal standard to sample standard and blank solutions to correct for matrix-induced signal suppression or enhancement for signal drift and for instrumental instability.It is also known that in order to obtain accurate results the internal standard should closely match the analyte in terms of mass number and ionization potential. For the VG PQI mass matching turned out to be much more important than ionization potential matching. However mass matching between internal standard and analyte does not necessarily guarantee that the internal stan- dard will appropriately correct for matrix effects. In the BCR intercomparison study of the determination of Pb in wine the wines were analysed without any sample pre-treatment except for a 10-fold dilution with 1% nitric acid.’ Fig. 2 shows the I I I I I 0 0.5 1 .o 1.5 2.0 2.5 3.0 Ethanol concentration (%) 1.00 - Fig. 2. Determination of Pb in wine showing signal enhancement in aqueous ethanol solutions relative to 1% nitric acid solutions for different potential internal standards A ’‘’TI; B 184W; C 209Bi; and D I9’Pt.Data normalized to the enhancement for 208Pb enhancement of the signals of four potential internal standards when going from a 1% nitric acid solution to solutions containing in addition an increasing amount of ethanol. The enhancements shown have been normalized to the enhance- ment obtained for 208Pb. Similar matrix effects due to ethanol can be expected for the diluted wine samples. Only for 205Tl does the signal enhancement equal the enhancement observed for 208Pb within the experimental uncertainty. Bismuth-209 which is often used as an internal standard for Pb determi- nations was found to be unsuitable in the experimental arrangement used in this work since its use leads to an error of more than 10%.The observed effect cannot be attributed to a difference in the ionization potential since the latter is 7.4 eV for Pb which is closer to the 7.3 eV of Bi than to the 6.1 eV of T1. Similar observations were made for medium mass ranges.6 For certification work it is therefore necessary to test different internal standards to make sure that suppression or enhancement of the analyte signal is correctly compensated for by the internal standard. Both the accuracy and the precision are improved by matching the mass of the analyte and the internal standard. This is demonstrated in Fig. 3 where the precision of the Y signal is compared with the precision of the ratio of the same signal to the signal of the internal standard.Results have been plotted for different internal standards as a function of the mass of the internal standard. It is clear from Fig. 3 that whenJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 189 12 m 1 10 - 8 n m c 6 4 2 I 1 I I 0 50 100 150 200 250 Mass number of the internal standard Fig. 3. Relative standard deviations (YO) for the signal of @ "Y and a signal ratio ("Y :internal standard) as a function of the mass number of the internal standard. Results of a long-term stability experiment (mass scanning mode) the difference between the mass of the internal standard and the mass of the analyte (Y) is large the precision of the ratio can be even worse than the precision of the Y signal itself.The optimum precision was obtained when the masses were similar. Unlike the conclusions of other researcher~,~ the precision obtained with the equipment used in this work does not depend upon the similarity between the first ionization poten- tials of the analyte and the internal standard. Correction for Spectral Interferences Spectral interferences are an important problem and a major limitation of quadrupole ICP-MS. To get accurate analytical results spectral interferences must be identified and eliminated or corrected for. Different methods can be applied for this purpose. For the determination of Mo in BCR CRM 403 (Sea Water) three techniques were applied. Although Mo has several stable isotopes none of these was free of interference when measured in the sea-water matrix as can be seen in Table 3.The isotopes at m/z 92 and 94 suffer from isobaric interferences from Zr. The other isotopes suffer from interference from polyatomic ions. Because the contribution of BrOH + and BrO+ species to the peak at m/z 89 is less than 5% "Mo was selected for the certification.8 First matrix matched blanks were prepared. In the 10-fold diluted sea-water used in this analysis a Br concentration of 6.75 mg 1-1 was found. Blanks with the same Br content were prepared and the signal intensity at m/z 98 was subtracted from the signal intensity at the same mass in the sea-water spectrum. A concentration of 10.21 pg 1-1 (standard deviation 0.42 pg 1-') was found. Next the inter- ference was corrected for by calculation. A mathematical correction based on the intensity of the 95Mo peak and on the isotopic abundances of the Mo isotopes and the experimentally determined 81BrOH+:79Br0+ ratio was used.8 A value of 10.99 pg 1-1 (standard deviation 0.52 pg 1-I) was found.Finally the Br interference was eliminated by removing the Br Table3 Major spectral interferences on the Mo isotopes in a sea-water matrix ~~~ Abundance (%) 14.8 9.1 15.9 16.7 9.5 24.4 9.6 Interferences "Zr + 94Zr+ KKO+ "BrO' ArKO+ "Zr' 79BrOH+ "BrO' NaKCl+ *'BrOH+ 81Br170+ *4sr0 + from the solution with an anion-exchange resin prior to the measurement. The 98Mo peak could then be measured free from interferences. The resulting concentration was 10.83 pg 1-1 (standard deviation 0.52 pg 1-l). Within the exper- imental uncertainties the three results compared well with one another and a mean value of 10.72 pg 1-1 (standard deviation 0.52 pg 1-l) was reported to BCR.This value is in agreement with the finally certified concentration of 10.1 pgl-' with a 95% confidence interval of L2.0 pg 1-I. For most matrices the presence and nature of the polyatomic interfering species is predictable and interferences can be identified based on the known or measured matrix composi- tion. Appropriate methods for interference correction can then be applied. Occasionally however polyatomic species were observed that sometimes could not be identified. For the determination of about 3 pg I-' of Cu in BCR estuarine water only the 65Cu isotope was measured since the 63Cu isotope was strongly interfered with by ArNa'.The Cu peak at m/z 65 is also expected to suffer interference mainly by polyatomic species containing S or P. These interferences were fully eliminated by removing S and P from the solution by anion exchange. Nevertheless results were too high by a factor of 2 and were rejected for certification. To investigate this problem further the samples were analysed with high-resolution ICP-MS. A new technology developed by Finnigan MAT (Bremen Germany) and used in a new high-resolution ICP-MS instrument soon to be available on the market was applied.' The resulting spectrum obtained for 1 pg1-I Cu is shown in Fig. 4 and reveals the presence of an interfering species at m/z 64.982 that until now could not be identified. Evident possibil- ities such as 24Mg,160H+ or Na2180H+ could be ruled out because the corresponding oxides could not be observed in the spectrum.The interfering peak may be due to an organic polyatomic species. Standardization Standardization is of course an important topic in certification work. Calibration procedures must be accurate and traceable which means that in principle one should always start by weighing an appropriate amount of an element standard or a compound with known stoichiometry. Three calibration methods can be used in ICP-MS. In many cases external calibration is sufficient on condition that an appropriate internal standard is chosen. Standard additions was used when matrix induced signal suppression or enhancement was expected or turned out to be insufficiently corrected for by the internal standard.Isotope dilution was applied only exception- 160 r I40 1 ; 120 rn rn 3 100 5 80 2 60 40 20 0 v) 0. cu fi \ ? I 64.869 64.919 64.969 65.019 Masslu Fig. 4. Spectral interference in the determination of Cu in BCR CRM Estuarine Water. High resolution ICP-MS spectrum of a 1 ppb solution at a resolution of 2200 [MIAM (M is the mass of a species AM is the difference in mass between two neighbouring species) 10% valley definition]'190 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 ally. In this laboratory it is common practice to test both external calibration and standard additions in BCR certifi- cation work. If after a first set of experiments both methods yield the same result with a comparable precision the less time consuming external calibration method is then used for further analysis. For the determination of Mg in BCR human serum (CRMs 303 and 304) external calibration did not yield sufficiently precise results.Beryllium had to be used as an internal standard because other elements appropriate for internal standardization were present in the serum samples. Because the relative difference in mass between Be and Mg is large the precision of the results was poor. To check the accuracy of the standardization an NIST SRM 909 Human Serum was also analysed. Values of 28.6 [standard deviation (n= 5)= 1.2 mg 1-'1 and 28.7 mg 1-' [standard deviation (n = 6) = 0.38 mg 1-'1 were found with exter- nal calibration and standard additions respectively. Both results are in agreement with the range of 28.7-30.6 mg I-' specified by NIST.The exceptionally large uncertainty on the external calibration results was also observed for the NIST material and the better precision of standard additions is obvious. Both methods are accurate. For the certification of the BCR material standard additions was used. In the determination of Cd in BCR CRM 143R Sewage Sludge-amended external calibration resulted in much lower values (63.6 pg g-'; standard deviation 0.9 pg g-') than stan- dard additions (71.5 pg g-'; standard deviation 1.2 pg g-'). This was not to be expected since the internal standard In in this case did not reveal any signal suppression or enhance- ment and since for other elements determined in the same material agreement between results from both standardization methods was excellent.In this case standard additions was used and results later turned out to be in excellent agreement with those of other participants and with the finally certified value (71.8 pg g-'; standard deviation 1.2 pg g-'). Isotope dilution is an expensive and time consuming stan- dardization method but it can be useful when sample pre- treatment causes loss of analyte material. It then corrects appropriately- for incomplete recovery but obviously not for contamination. Therefore blanks are still necessary and unless isotope dilution is applied to the blanks also the use of an internal standard is still necessary. In practice isotope dilution has been applied to the determination of Pb in BCR CRM 414 (Plankton) BCR CRM 422 (Cod Muscle) and wine and to the determination of Hg in BCR CRM 422 (Cod Muscle).Results of the latter analysis" are presented in Fig. 5. Parallel to the c 0) 0) 700 1 - (0 2 600 .- a C $ 500 G 8 $! 400 In m u 300 0 .- 4- 2 4- 6 200 0 C 0 0 ICP-MS CVAFS ' . Certified X I € isotope dilution 2 Standard additions T T 3- External calibration 1 _L Isotope dilution 1 t External calibration 2 I 1 Method Fig.5. Determination of Hg in BCR CRM 422 (Cod Muscle). Concentrations and 95% confidence intervals (ng g-'). For Isotope dilution 1 microwave digestion was carried out immediately after spike addition; and for Isotope dilution 2 overnight room temperature digestion was carried out prior to microwave digestion ICP-MS analyses Hg was also determined with cold vapour atomic fluorescence spectrometry. For ICP-MS analyses about 0.1 g of material was digested in nitric acid in a closed vessel in a microwave oven.The recovery of Hg was greater than 96%. When using standard additions results were higher than with external calibration for which repeatability was poor as is shown in Fig. 5 for two series of analyses. The data obtained with standard additions were confirmed by the cold vapour atomic fluorescence values. Isotope dilution yielded different results depending on the details of the digestion procedure. When the Teflon bomb with the sample and the nitric acid was immediately heated in the microwave oven the results obtained were too low. When the mixture was allowed to equilibrate overnight prior to microwave digestion the results obtained compared well with those of standard additions and atomic fluorescence.A possible explanation is that organomer- cury compounds survive the microwave digestion and that the ICP-MS signal per unit mass of Hg is lower for organomercury than for inorganic Hg ions. The overnight equilibration in this case enabled the spiked mercury to be incorporated into the organic species. An overnight equilibration step was also included for the standard additions. Conclusion In general it can be concluded that BCR has principally accepted ICP-MS for certification work and that ICP-MS substantially contributes to BCR certifications in the field of biological and environmental materials. The ICP-MS technique was shown to yield accurate results only if matrix effects and spectral interferences were rigorously studied and dealt with.Accuracy and precision therefore cannot be taken for granted. Whereas in the period from 1982 'to 1990 ICP-MS contrib- uted to less than 1% of the certification work carried out in this laboratory on biological and environmental samples its contribution since then has risen to about 60%. Each certifi- cation required substantial scientific research leading to data that was usually acceptable for certification. The accuracy and precision of the results obtained by ICP-MS are comparable to or better than those of reference methods such as neutron activation analysis. 1 2 3 4 5 6 7 8 9 10 11 12 13 References Jarvis K. E. Gray A. L. and Houk R. S. Handbook of Inductively Coupled Plasma Mass Spectrometry Blackie Glasgow and London and Chapman and Hall New York 1992 p.1. Houk R. S. Shum S. C. K. and Wiederin D. R. Anal. Chim. Acta 199 250 61. Cornelis R. Dyg S. Griepink B. and Dams R. Fresenius' J. Anal. Chem. 1990 338 414. Quevauviller P. Vercoutere K. Bousfield D. and Griepink B. Report EUR 14062 EN ECSC-EEC-EAEC Brussels Luxembourg Goossens J. De Smaele T. Moens L. and Dams R. Fresenius' J. Anal. Chem. 1993 347 119. Goossens J. Vanhaecke F. Moens L. and Dams R. Anal. Chim. Acta 1993 280 137. Thomson J. J. and Houk R. S. Appl. Spectrosc. 1987 41 801. Vanhaecke F. Goossens J. Dams R. and Vdndecasteele C. Talanta 1993 40 975. Schmidt G. Finnigan MAT Bremen Germany personal communication. Campbell M. J. Vermeir G. Dams R. and Quevauviller P. J. Anal. At. Spectrom. 1992 7 61 7. Griepink B. Maier E. A. and Muntau H. Report EUR 12680 EN ECSC-EEC-EAEC Brussels Luxembourg 1990 pp. 1-82. Vandendriessche. S. Griepink B. and Marchandise H. Report EUR 14056 EN ECSC-EEC-EAEC Brussels Luxembourg 1992 De Leenheer A. P. Yeoman W. B. and Colinet E. S. Report EUR 10627 EN ECSC-EEC-EAEC Brussels Luxembourg 1992 1992 PP. 1-64. pp. 1-55. pp. 1-64.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 191 14 Quevauviller P. Kramer K. J. H. Vercoutere K. and and Griepink B. Report EUR 14557 EN ECSC-EEC-EAEC Griepink B. Report EUR 14061 EN ECSC-EEC-EAEC Brussels Luxembourg 1992 pp. 1-53. 15 Quevauviller P. Vercoutere K. Muntau H. and Griepink B. Report EUR 14558 EN ECSC-EEC-EAEC Brussels Luxembourg 1993 pp. 1-71. 16 Quevauviller P. Imbert J. L. Wagstaffe P. J. Kramer G. N. Brussels Luxembourg 1993 pp. 1-64. Paper 3/04699K Received August 4 1993 Accepted December 6 1993