The reproducibility (RSD) of three to four independent determinations (different spik- ing/digestions, Appendix B) of different reference materials is, in most cases, better than 1% for elements determined in the LR mode and better than 2% for elements determined in the HR mode. These findings are in good agreement with the long-term reproducibility obtained from the analysis of BHVO-1 (see above). In Figures 12, 13, and 14, relative deviations (mean value/reference value) of concentration data for silica-rich rock sam- ples, basalts, and glasses are shown. Most data agree within 3-4% with reference values that have been determined by ID and by external calibration using ID values as internal standards (filled symbols; Raczek et al. [2001]; Jochum et al. [2001]; Baker et al. [2002]).
Exceptions are the Pb concentrations of BCR-2 and BHVO-2 (Fig. 13) that are up to 14% lower than the values of Jochum et al. [2001]. Assuming our values to be correct, the lower Pb concentrations in new USGS reference materials (BCR-2, BHVO-2, AGV- 2) compared to higher Pb concentrations in original samples (BCR-1, BHVO-1, AGV- 1) can be explained by Pb contamination during the sample preparation of the original samples, a finding which is consistent with different Pb isotopic compositions between both generations [Woodhead and Hergt, 2000]. In addition, the U and Th concentrations in BCR-1 and the Th concentration in BCR-2 are up to 10% lower than the respective values from Jochum et al. [2001]. Yet, the Th/U ratios of 3.22 and 3.21 in BCR-1 and BCR-2 of this study are in better accordance than the respective values of 3.30 and 3.48 in Jochum et al. [2001].
Reference materials containing refractory minerals or having low concentrations of trace elements have significantly larger uncertainties of about 4 to 9% for some elements (e.g., Zr, Hf, Pb, U in G-2 or Sm, Ta in PCC-1). They also show larger relative deviations from published values (up to 60%). This can be related to (1) variable digestion yields especially of refractory mineral phases (e.g., zircon in G-2), (2) a too small test sample
Rb Sr Y Zr Nb Cs Ba La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Hf Ta Pb Th U 0.9 1.0 1.1 1.2
Concentration normalized to reference value
KL2-G ML3B-G NIST 612
Figure 14: Comparison of concentration data of geological reference glasses KL2-G, ML3B-G, NIST SRM 612 with reference values (Appendix B). Note that confidence intervals (68%) also include uncertainties of reference values that range between 1 to 8%.
size especially if the reference material was produced from a coarse grained rock and the powder itself is not fine grained enough, (3) the low concentration levels of trace elements in the analyzed sample (e.g., 30 - 100 ng g−1 for REE in PCC-1), or a combination of 2) and 3).
In Fig. 15, trace element data for the new USGS reference glasses BCR-2G, BHVO-2G, and BIR-1G are compared with those of their original rock powders. Most data agree within the given uncertainty levels of both determinations (ca. 3% relative deviation). The elevated U and Pb concentrations in BIR-1G (almost 300% and 7% higher compared to original powder) are significant features and are probably related to U and Pb con- tamination during the preparation of the BIR-1G reference glass by the USGS. There is a good agreement for Pb concentrations in BHVO-2G and BHVO-2 (ca. 2% depletion of Pb in BHVO-2G). However, a significant depletion of Pb in BCR-2G relative to BCR-2 is observed (ca. 10% relative deviation). We relate this to the loss of Pb during the preparation of BCR-2G. Similar depletion of Pb during the preparation of melt beads for laser ablation ICP-MS has been observed in our laboratory. This is probably related to increasing volatility of Pb with increasing melting temperature and time of melting.
In NIST SRM 612 the concentration levels for trace elements range between 20 and 80 µg g−1 [Pearce et al., 1997] leading to large spectroscopic interferences that are commonly
not encountered in natural samples. These interferences are 69Ga16O on85Rb,74Ge16O on
90Zr (affecting both the Zr and Nb concentration; see Table 1), 162Dy16O on 178Hf, and 165Ho16O on181Ta. In addition, inadequate spiking leads to increased error magnification
Rb Sr Y Zr Nb Ba La Ce Pr Nd Sm Eu Gd Tb Ho Er Tm Yb Lu Hf Ta Pb Th U 0.8 0.9 1.0 1.1 1.2
USGS reference glass / original rock powder
Dy
BCR-2G BIR-1G BHVO-2G
Figure 15: Comparison of trace element concentrations for new USGS reference glasses BCR-2G, BIR-1G, BHVO-2G with the results obtained for the original USGS reference materials (Appendix B).
for Pb and U due to the high concentrations of these elements in NIST SRM 612 that strongly contrast those in natural samples. We therefore have omitted concentration data for Rb, Zr, Nb, Hf, Ta, Pb and U for NIST SRM 612 in Appendix B.
6
Overall performance of multi-element ID-SF-ICP-MS
In the following, we discuss the advantages and disadvantages of ID-SF-ICP-MS in the light of the most prominent obstacles in ICP-MS and TIMS analysis. As a result, we pro- pose that multi-element ID-SF-ICP-MS fulfills the requirements of advanced geochemical investigations for data with low uncertainty and short analysis time.6.1 Matrix effects
Matrix effects can considerably diminish accuracy of ICP-MS analyses if external calibra- tion is applied. It has been shown that only matrix-matched standardization (e.g., the use of reference materials as calibration standards) can account for variation in instrument sensitivity due to matrix effects [Cheatham et al., 1993]. However, if only one reference material is used for standardization (e.g. BHVO-1) the accuracy can be diminished due to the uncertainty of concentrations in the used reference material [Eggins et al., 1997].
Matrix effects can be eliminated by the use of ID because the sensitivity of the in- strument is determined within the sample matrix itself by the added tracer [Heumann,
1986]. In addition, matrix effects on concentration data that are determined by RSF can be significantly reduced by the use of several internal standards if there is a close mass matching between the element of interest and the internal standard [Longerich et al., 1990; Field and Sherrell, 1998]. Accordingly, matrix effects are eliminated or largely avoided by the use of multi-element ID-SF-ICP-MS.