Anal. Chem. 2002, 74, 67-73 Combined Chemical Separation of Lu, Hf, Sm, Nd, and REEs from a Single Rock Digest: Precise and Accurate Isotope Determinations of Lu-Hf and Sm-Nd Using Multicollector-ICPMS Ilka C. Kleinhanns,*,† Katharina Kreissig,‡ Balz S. Kamber,§ Thomas Meisel,| Thomas F. Na 1 gler,† and Jan D. Kramers† Gruppe Isotopengeologie, Mineralogisch-Petrographisches Institut, Universität Bern, Erlachstrasse 9a, CH-3012 Bern, Switzerland, Department of Earth Sciences, University of Bristol, Queens Road, Bristol, BS8 1RJ, U.K., Department of Earth Sciences, University of Queensland, Qld 4072 Brisbane, Australia, and Institut für allgemeine und analytische Chemie, Universität Leoben, Franz-Josef-Strasse 18, A-8700 Leoben, Austria A combined procedure for separating Lu, Hf, Sm, Nd, and rare earth elements (REEs) from a single sample digest is presented. The procedure consists of the following five steps: (1) sample dissolution via sodium peroxide sintering; (2) separation of the high field strength elements from the REEs and other matrix elements by a HF-free anion-exchange column procedure; (3) purification of Hf on a cation-exchange resin; (4) separation of REEs from other matrix elements by cation exchange; (5) Lu, Sm, and Nd separation from the other REEs by reversed-phase ion chromatography. Analytical reproducibilities of Sm-Nd and Lu-Hf isotope systematics are demonstrated for standard solutions and international rock reference materials. Results show overall good reproducibilities for Sm-Nd systematics independent of the rock type analyzed. For the Lu-Hf systematics, the reproducibility of the parent/daughter ratio is much better for JB-1 (basalt) than for two analyzed felsic crustal rocks (DR-N and an Archaean granitoid). It is demonstrated that this poorer reproducibility of the Lu/Hf ratio is truly caused by sample heterogeneity; thus, results are geologically reasonable. The combination of Sm-Nd and Lu-Hf isotope systematics with rare earth element (REE) patterns provides a powerful tool for understanding Earth’s differentiation. Single mineral phases, such as zircon and garnet, which play an important role in mantlecrust-related processes, reveal dissimilar behavior with respect to the Sm-Nd and Lu-Hf systematics and to REE pattern. For example, Hf is a major element in zircon with concentrations of up to 3 wt %, whereas the concentration of Lu in this mineral phase is 2 orders of magnitude lower, resulting in extremely low Lu/Hf ratios. In contrast, both Sm and Nd are just trace elements in * Corresponding author: (e-mail) ilka@mpi.unibe.ch; (fax) + 41 (0)31 631 49 88; (tel) +41 (0)31 631 85 33. † Universität Bern. ‡ University of Bristol. § University of Queensland. | Universität Leoben. 10.1021/ac010705z CCC: $22.00 Published on Web 11/20/2001 © 2002 American Chemical Society zircon. Differing amounts of zircon in separate powder aliquots of the same rock produce a strong variability in the measured Lu/Hf ratios of the rock but do not significantly influence the Sm/ Nd ratios. Garnet, on the other hand, has very high partition coefficients for heavy REEs, resulting in very high Lu/Hf and moderately elevated Sm/Nd ratios. Hence, garnet influences the overall budget and pattern of the REEs in rocks and effects the Lu-Hf as well as the Sm-Nd isotope systematics. The same is true for other mineral phases with high partition coefficients for REE such as pyroxene, amphibole, or monazite. Total dissolution of the rock powder aliquot is therefore essential if geological interpretations are to be made by comparison of REE pattern with Lu-Hf and Sm-Nd isotope systematics. To develop a combined technique analyzing Sm-Nd and LuHf systematics as well as REEs (unspiked except for Nd, Sm, and Lu), we adopted the method of sintering with sodium peroxide, which is already in use for REE analyses on whole-rock samples.1 Sodium peroxide sintering is not only capable of dissolving even very resistant mineral phases such as zircon, garnet, titanite, and osmiridium2 but is also a very rapid digestion method. Further, the method avoids the use of HF during sample digestion and prior to separation of high field strength elements (HFSEs) and REEs. This is favorable as problems in combined Sm-Nd and Lu-Hf isotope as well as REE analyses may result from the formation of fluorides3 or fluoride gels.4 These hinder complete sample dissolution as well as complete spike-sample homogenization and risk strongly fractionating REEs. Indeed, problems with spike-sample homogenization for the Lu-Hf isotope system were mentioned,5,6 and low reproducibilities of the Lu/Hf ratio were explained by fluoride precipitation.5 HFSE analyses problems (1) Robinson, P.; Higgins, N. C.; Jenner, G. A. Chem. Geol. 1986, 55, 121137. (2) Seelye, F. T.; Rafter, T. A. Nature 1950, 165, 317. (3) Nägler, Th. F.; Kamber, B. S. Schweiz. Mineral. Petrogr. Mitt. 1996, 76, 75-80. (4) Cohen, A. S.; O’Nions, R. K.; Siegenthaler, R.; Griffin, W. L. Contrib. Mineral. Petrol. 1988, 98, 303-311. (5) Blichert-Toft, J.; Albarède, F. Earth Planet. Sci. Lett. 1997, 148, 243-258. (6) Scherer, E. E.; Cameron, K. L.; Johnson, C. M.; Beard, B. L.; Barovich, K. M.; Collerson, K. D. Chem. Geol. 1997, 142, 63-78. Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 67 related to acid digestion were evaluated by Münker7 and Le Fèvre and Pin.8 Variabilities in present-day isotope compositions of daughter elements and parent/daughter ratios are well-known effects of sample heterogeneity.6,9-13 However, if all mineral phases are in isotopic equilibrium, the influence on initial daughter isotope ratios by sample heterogeneity is zero. Heterogeneous distribution of phases that are not in isotopic equilibrium, on the other hand, can cause inconsistency between different isotope systems and REEs when these geochemical tracers are determined on different sample powder aliquots. This is due to different parent/daughter ratios in different mineral phases of the sample and, hence, different amounts of radiogenic ingrowth. To circumvent this problem and allow direct comparison of REE pattern, Lu-Hf and Sm-Nd isotope systematics of a single rock sample, the technique presented here is based on chemical separation from a single sample digest. EXPERIMENTAL SECTION Chemicals. Ultrapure water with a resistivity of 18.2 MΩ/cm was used (Milli-Q). All acids were doubly distilled with the exception of 12 M HCl, which has been cleaned using an anionexchange column. Pure sodium peroxide was purchased from Fluka (Na2O2; purum g95%). Certified Johnson Matthey ICP standard solutions (1000 ( 3 µg/mL; Hf solution Lot No. 5011537B; Lu solution Lot No. 601782M) were used to gravimetrically prepare standard solutions with known concentrations. These solutions were used to calibrate the spike solution via multiple MC-ICPMS measurements of spiked standard solutions. Caution! HCl is toxic and corrosive even when diluted. HCl gas is toxic and highly irritating the respiratory system. Handle with great care! Caution! HF is highly toxic and corrosive even when diluted. It emits highly toxic HF gas. It irritates the skin and mucous membranes of the body. Handle with great care! Caution! Na2O2 is an oxidizing and corrosive substance. In contact with combustible material, it may cause fire. Na2O2 is strongly hygroscopic; keep container dry. Handle with great care! Step 1: Sample Digestion. Finely ground sample powder (100 mg) is weighed into glassy carbon vessels (Sigradur G) together with 176Lu-180Hf mixed spike stabilized in 1 M H2SO4 and 149Sm-150Nd mixed spike stabilized in 1.6 M HCl. (The stability of the Lu/Hf spike was tested over a three-month period. The Lu/Hf spike ratio thereby remained stable within 0.4% (2 SD), showing no indication of a variation with time). The suspension is completely dried on a hot plate, and the powder is mechanically desegregated with a spatula. Sodium peroxide is added as sintering reagent. A sodium peroxide/sample ratio of 6:1 is sufficient to ensure total decomposition of zircons in zircon-rich rock powders. In zircon-poor or finer-grained materials, lower (7) Münker, C. Chem. Geol. 1998, 144, 23-45. (8) Le Fèvre, B.; Pin, C. Anal. Chem. 2001, 73, 2453-2460. (9) Pettingill, H. S.; Patchett, P. J. Earth Planet. Sci. Lett. 1981, 55, 150-156. (10) Na, C. N.; Nakano, T.; Tazawa, K.; Sakagawa, M.; Ito, T. Chem. Geol. 1995, 123, 225-237. (11) Kane, J. S. Analyst 1997, 122, 1289-1292. (12) Meisel, T.; Moser, J.; Wegschneider, W. Fresenius J. Anal. Chem. 2001, 370, 566-572. (13) Scherer, E. E.; Cameron, K. L.; Blichert-Toft, J. Geochim. Cosmochim. Acta 2000, 64, 3413-3432. 68 Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 sodium peroxide/sample ratios are adequate, and the procedural blank is reduced. The spike-sample mixture and the sintering reagent are carefully mixed with a spatula and finally covered with a thin layer of sodium peroxide. Sintering is achieved by heating in a muffle furnace at 480 ( 10 °C for 0.5 h. After the sample is cooled to room temperature, water is carefully added in small amounts, resulting in a vigorous exothermal reaction. Water is added until the reaction has fully ceased. The result of this reaction is a suspension of precipitated cation hydroxides in strong NaOH. The suspension is centrifuged and the NaOH supernate, containing silica, is discarded. Addition of water to the precipitate followed by centrifuging is repeated, but this time the supernate liquid is transferred into a second vial. HCl is added to the decanted liquid to test whether silica gel is still evolving, which results in blurring the liquid. Water addition, centrifugation, and HCl addition to the decanted supernate liquid are repeated until no more silica gel forms (normally after one repetition). Step 2: Separation of HFSEs and REEs. The remaining precipitates from the sample digestion are dissolved in 4 mL of 12 M HCl and loaded onto an anion-exchange column (3 mL of Dowex AG1-X8, 200-400 mesh). Strong (>8 M) HCl is required for loading the sample onto the column, because Ti, Zr, and Hf do not adhere to anion-exchange resin in weaker HCl media. REEs and other matrix elements are eluted into a vial for further REE purification (step 4) using 19 mL of 12 M HCl. Ti is reduced in the HFSE fraction and washed out with 2 mL of 6 M HCl. Finally, the HSFEs are eluted together with Fe and Cr with 15 mL of 2.5 M HCl. Hf is separated from this cut in step 3. Step 3: Purification of Hf. The HFSE solution from step 2 is dried to a small droplet and redissolved in 2 mL of 2.5 M HCl. Addition of 100 µL of H2O2 to the sample solution forms yellow chromium and iron peroxide complexes, which do not adhere to cation-exchange resin. The solution is loaded onto a cationexchange column (4 mL of Dowex AG50W-X8, 200-400 mesh). Ti, Cr, and Fe are eluted with 10 mL of 2.5 M HCl or until the eluate is colorless. Purified Hf is eluted with 4 mL of 5 M HF and dried. At this point, the sample is ready for mass spectrometry. The use of HF in this step ensures high yields for Hf. There is no risk of REE fractionation by fluoride precipitation, because the REEs have already been separated from the HFSE during step 2. Step 4: Purification of REEs. REE purification is performed using a conventional cation-exchange column technique (3 mL of Dowex AG50W-X8 resin, 200-400 mesh). The REE solution from step 2 is evaporated, redissolved in 4 mL of 2.5 M HCl, and loaded onto the column. Matrix elements are washed out with 11 mL of 2.5 M HCl, and the purified REE-fraction is eluted with 20 mL of 6 M HCl. Before further processing in step 5, a 0.5-mL aliquot of the eluate is separated for determination of REE concentrations for REE patterns. Step 5: Separation of Nd, Sm, and Lu. Pure Nd, Sm, and Lu fractions are obtained by reversed-phase chromatography (modified after Richard14) (4 mL of HDEHP-coated Teflon in quartz-glass columns). The REE solution from step 4 is evaporated, redissolved in 300 µL of 0.18 M HCl, and loaded onto the column. La, Ce, and Pr are washed out with 20 mL of 0.18 M HCl. Nd is (14) Richard, P; Shimizu, N.; Allégre, C. J. Earth Planet. Sci. Lett. 1976, 31, 269-278. collected in the next 6 mL of 0.18 M HCl. The remaining Nd is fully washed out with 8 mL of 0.4 M HCl, which allows collection of a Nd-free Sm fraction in the succeeding 7 mL 0.4 M HCl. A Nd-free Sm fraction is favored to preclude any possible matrix effect. Medium to heavy REEs, including Yb, are stripped off with 73 mL of 2.5 M HCl. Lu is collected in the following 15 mL of 2.5 M HCl. Complete separation of Yb and Lu is beneficial,15,16 as it reduces corrections related to isobaric interference of 176Yb (13% of natural Yb) on 176Lu (2.6% of natural Lu). Otherwise, these corrections would be important, as Yb in geological samples is general 4-7 times more abundant than Lu. Mass Spectrometry. REE concentrations of unspiked solutions were measured on quadrupole ICPMS (HP4500, HewlettPackard) at the University of Leoben, following the procedures of Meisel et al.17 Determination of Lu, Hf, Sm, and Nd isotope ratios were carried out in static mode on Faraday collectors on the Nu instruments multicollector ICPMS at the University of Bern. Sample aspiration was performed using an Aridus desolvating nebulizer. Desolvating nebulizers have the advantage of producing “dry” aerosols, which minimizes plasma fluctuations. Apart from measuring all isotopes of a certain element simultaneously, the 12-Faraday collector array of the MC-ICPMS also allows simultaneous monitoring of other elements with possible interfering isotope masses. Isotopes monitored during Hf isotope measurements were 173Yb (to correct for the 176Yb isobaric interference on 176Hf), 175Lu (176Lu on 176Hf), 181Ta (180Ta on 180Hf), and 183W (180W on 180Hf). Monitored isotopes during Lu isotope measurements were 178Hf (176Hf on 176Lu) and 174Yb (176Yb on 176Lu), and during Nd isotopic measurements, 147Sm (144Sm on 144Nd) and 140Ce (142Ce on 142Nd) were checked. During Sm isotope measurements (masses 147, 149, and 152), 156Gd (152Gd on 152Sm) was monitored. All uncertainties given in the following text are 2σ standard deviations. Further, they comprise measurements with different sample uptake systems (Aridus and Micromist with microcyclonic spray chamber) over a long period of time. All Hf isotope ratios are internally corrected for fractionation using a value of 0.7325 for 179Hf/177Hf and the exponential law. Measured values of 176Hf/177Hf, 178Hf/177Hf, and 180Hf/177Hf ratios of an in-house Hf standard solution are 0.282 165 ( 12, 1.467 29 ( 8, and 1.8868 ( 5, respectively for the period between November 1999 and December 2000. Measured values of 176Hf/177Hf, 178Hf/177Hf, and 180Hf/177Hf ratios of the JMC 475 standard solution during the same time period are 0.282 169 ( 16, 1.467 29 ( 8, and 1.8868 ( 3, respectively. Remarkably, these two Hf standard solutions are isotopically indistinguishable. Assuming that instrumental mass bias follows the exponential law, uncorrected 178Hf/177Hf and 176Hf/177Hf ratios should fit a linear array in a ln(178Hf/177Hf) versus ln(176Hf/177Hf) plot. The theoretical slope of this linear array can be calculated with eq 1: slope ) ln(M176/M177) (M178/M177) ) -1.0070 (1) where M176, M177, and M178 are the exact atomic masses of 176Hf, (15) Patchett, P. J.; Tatsumoto, M. Contrib. Mineral. Petrol. 1980, 75, 263-267. (16) Gruau, G.; Cornichet, J.; Le Coz-Bouhnik, M. Chem. Geol. 1988, 72, 353356. 177Hf, and 178Hf, respectively. Thus, the validity of exponential fractionation can be tested by comparing the slope obtained from uncorrected Hf ratios in the ln(178Hf/177Hf) versus ln(176Hf/177Hf) plot with their respective calculated theoretical slope. Figure 1A illustrates that within uncertainties the slopes obtained for Hf in-house (i.e., -1.0085 ( 68) and JMC 475 (i.e. -1.0098 ( 68) standard solutions are indistinguishable from each other and equal to the calculated theoretical slope (see eq 1). As expected from these observations, mass bias-corrected ratios of both the Hf in-house standard solution and those of the JMC 475 standard solution plot within analytical uncertainty of the theoretical fractionation line (Figure 1B). Published mass biascorrected ln(178Hf/177Hf) and ln(176Hf/177Hf) ratios of the JMC 475 standard solution are also plotted in Figure 1B for comparison with our results. Despite some variations in the 178Hf/177Hf values, all laboratories report within uncertainties the same 176Hf/177Hf ratio. Lu has only two naturally occurring isotopes, which makes internal mass bias correction impossible. The method of element doping is applied to correct for mass bias, using W of known isotope composition as a reference element. Tungsten is ideal as doping agent for Lu measurements, since its range in isotope masses is close to that of Lu, but it does not isobarically interfere. The applicability of W doping for mass bias correction of the 175Lu/176Lu ratio was tested with the same approach that was used for internal mass bias correction of Hf isotopes. We found a strong linear correlation between uncorrected ln(175Lu/176Lu) and ln(182W/184W) ratios. Within uncertainty, the slope of the linear array of 0.5358 ( 146 equals the theoretically calculated one of 0.5217, indicating identical fractionation behavior of W and Lu. Relative to a 182W/184W ratio of 0.864 98,18 our mass bias-corrected value for 175Lu/176Lu is 37.72 ( 2. This value is 0.2% higher than the one given by Scherer et al.19 (37.63 ( 2) using Re as doping agent or Blichert-Toft et al.20 (37.65, no error cited) using Yb. However, all quoted 175Lu/176Lu ratios are directly related to the applied isotope composition of the respective doping agents. The problem of unknown absolute Lu isotope composition is however demagnified for Lu isotope dilution analyses, if spike calibration and sample measurement follow the same protocol. The relation of the W ratio to the Lu ratio remained constant on the Nu instruments multicollector ICPMS over the whole period (February to December 2000). Thus, no empirical corrections had to be applied. The measured 143Nd/144Nd and 145Nd/144Nd ratios of our inhouse Nd standard solution, between February and December 2000, were 0.511 066 ( 18 and 0.348 414 ( 26, respectively, corresponding to a 143Nd/144Nd La Jolla value of 0.511 856. All Nd isotope ratios were internally corrected for mass bias using a value of 0.7219 for 146Nd/144Nd and the exponential law. Perfect agreement between the theoretically calculated slope (see eq 1) of -1.0048 and the slope defined by uncorrected Nd ratios of -1.0024 ( 25 in the ln(143Nd/144Nd) versus ln(145Nd/144Nd) space confirms the applicability of exponential mass bias correction. (17) Meisel, T.; Schöner, N.; Paliulionyte, V.; Kahr, E. Geostand. Newsl., in press. (18) Lee, D.-C.; Halliday, A. Nature 1995, 378, 771-774. (19) Scherer, E. E.; Münker, C.; Rehkämper, M.; Mezger, K. Eos, Trans. 1999, 80, 1118. (20) Blichert-Toft, J.; Chauvel, C.; Albarède, F. Contrib. Mineral. Petrol. 1997, 127, 248-260. Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 69 Figure 1. (A) Plot of ln(176Hf/177Hf) vs ln(178Hf/177Hf) showing strong linear correlation for uncorrected ratios of the Hf in-house (filled diamonds) and JMC 475 (crosses) standard solutions. (B) Detailed view of the mass bias-corrected averages of the Hf in-house and JMC 475 standard (same symbols as in (A)). A comparison of our data with that of other laboratories for the JMC 475 standard solution shows that all laboratories measure similarly and are within uncertainty of the theoretical fractionation line. The following symbols indicate the Hf values presented in the references noted: 9,32 4,33 3,20 O,6 b,34 and ].13 RESULTS AND DISCUSSION Yield and Blank Levels. For Hf, a yield of 90% was obtained, which is comparable to published recoveries.6,15,20 Recoveries of Sm and Nd are deliberately held around 90% and for Lu around 60%. The loss occurs during REE separation (step 5), since high levels of Ce in Nd, Nd in Sm, and Yb in Lu fractions have to be avoided to minimize isobaric interferences as well as possible matrix effects during measurements. The low recovery for Lu is comparable to that of Patchett and Tatsumoto21 with 50% and that of Gruau et al.16 with 65%, who also aimed for low Yb levels in the Lu fraction. However, with complete spike-sample homogenization, the somewhat lower Lu yield does not affect the accuracy of Lu concentration determinations or Lu/Hf ratios. The chemistry blank for Nd is below 20 pg, suggesting negligible overall REE blanks. For Hf, a chemistry blank of 90 pg was determined. The total procedural blanks for Nd and Hf, including sample digestion, are 370 and 330 pg, respectively, corresponding to a blank contribution of Hf and Nd to the sample (21) Patchett, P. J.; Tatsumoto, M., Geophys. Res. Lett. 1980, 7, 1077-1080. 70 Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 analyte of below 0.1%, respectively. The main blank contribution can be related to the flux reagent sodium peroxide, which cannot be especially cleaned. However, as Scherer et al.6 pointed out, the effect of the blank on the measured Hf ratios is negligible as long as the blank contribution to the sample Hf is below 0.24%. Reproducibility of Natural Rock Samples. Two international rock reference materials (basaltic JB-1 and dioritic DR-N) were measured for Sm-Nd and Lu-Hf systematics on a MC-ICPMS (using isotope dilution for concentration determinations), as well as for REE concentrations on a quadrupole ICPMS (using external calibration for concentration determinations) to check the reliability of the developed technique for different rock types. Additionally, a sample from the Archaean Dalmein pluton, Barberton Mountain Land (South Africa) was analyzed to quantify the accuracy and reproducibility of age-corrected Nd and Hf isotope ratios and Nd and Hf mantle extraction model ages. The Dalmein pluton is a medium-grained biotite-granodiorite with a well-constrained age of 3216+2/-1 Ma (U/Pb-zircon22). Several 100-mg aliquots of JB-1 and DR-N were digested with sodium peroxide sintering. Separation of Lu, Hf, Sm, Nd, and REEs on all sample solutions was carried out with the chemical separation procedure described above. For comparison, 100-mg powder aliquots of both reference materials were digested with conventional acid attack for Sm-Nd systematic and REE. All REE measurements are given in Table 1 and Lu-Hf and Sm-Nd isotope results in Table 2. Independent of the digestion method used, Sm/Nd as well as Nd/Lu ratios of JB-1 determined by isotope dilution were consistent. The same holds true for DR-N. In addition, the obtained values of both rock reference materials agree well with those reported in the literature (Table 1). Quadrupole ICPMS results of Sm/Nd and Nd/Lu ratios of sintered powder aliquots are equal to those obtained by isotope dilution on MC-ICPMS and thus to literature values. Interestingly, Sm/Nd and Nd/Lu ratios determined by external calibration on quadrupole ICPMS from samples digested by acid attack are slightly different (Table 1). However, Eu anomalies (eq 2; N denotes chondrite-normalization) are Eu/Eu* ) EuN x(SmN × GdN) (2) identical and correspond to literature values (Table 1), independent of the digestion method and measuring technique used. It would be rather speculative to explain why acid-digested aliquots reveal a slight discrepancy from literature values in Sm/ Nd and Nd/Lu ratios when determined by external calibration; however, this is not the aim of the study. The major observation concerning the reliability of our chemical separation procedure and that of the sodium peroxide sintering is the excellent reproducibility of the Sm/Nd and Nd/Lu ratios, independent of the measuring technique. The 147Sm/144Nd and 143Nd/144Nd ratios reproduce well regardless of the rock type analyzed and the digestion technique used (Table 2). Mean Nd mantle extraction ages of JB-1 and DR-N are consistent for both digestion techniques, respectively. Additionally, Sm and Nd concentrations of DR-N using the sintering digestion (22) Kamo, S. L.; Davis, D. W. Tectonics 1994, 13, 167-192. Table 1. Comparison of the Elemental Ratios Sm/Nd, Nd/Lu, and Eu/Eu*a for Reference Materials JB-1 and DR-Nb sinter 1 sinter 2 HF-HNO3 1 HF-HNO3 2 JB-1 ID-MC-ICPMS Quad. ICPMS ID-MC-ICPMS Quad. ICPMS ID-MC-ICPMS Quad. ICPMS ID-MC-ICPMS Quad. ICPMS XRF INAA ID-MS sinter 1 sinter 3 HF-HNO3 1 HF-HNO3 2 ICP-AES ICPMS HPLC-ID-MS DR-N ID-MC-ICPMS Quad. ICPMS ID-MC-ICPMS Quad. ICPMS ID-MC-ICPMS Quad. ICPMS ID-MC-ICPMS Quad. ICPMS Sm/Nd Nd/Lu 0.192 0.192 0.193 0.197 0.194 0.202 0.193 0.201 0.196 0.188 0.189 81.2 85.3 77.1 82.1 0.223 0.228 0.223 0.226 0.224 0.236 0.224 0.236 0.228 0.224 0.211 Eu/Eu* 0.93 0.94 87.5 0.93 86.7 109.2 82.1 87.1 0.94 0.89 0.96 0.95 63.2 62.5 62.9 62.0 0.86 0.85 61.6 0.86 60.5 60.3 66.2 56.1 0.86 0.85 0.88 0.89 a See text for calculation. b ID-MC-ICPMS, this work, isotope dilution multicollector ICP-mass spectrometry, Bern; Quad. ICPMS, this work, external calibration, quadrupole ICP-mass spectrometry, Leoben; XRF,1 X-ray fluorescence spectrometry, sinter digestion; INAA, see references in ref 1, instrumental neutron activation analysis; ID-MS, see references in ref 1, isotope dilution mass spectrometry; ICP-AES,27 ICP-atomic emission spectrometry; ICPMS,25 ICP-mass spectrometry; HPLC-IDMS,28 high-performance liquid chromatography-isotope dilution-mass spectrometry. are 5.25 and 23.6 µg/g, respectively, values that are identical to those of 5.22 µg/g Sm and 23.3 µg/g Nd reported by Pin and Zalduegi.23 However, these authors reported 143Nd/144Nd ratios of 0.512 425 ( 7 and 0.512 428 ( 8 for DR-N (Table 2), values that are somewhat higher than ours. Comparison with their SmNd systematics via mantle extraction ages, on the other hand, reveals ages identical to our results (Table 2). The five sintered aliquots of the Dalmein pluton yield the highest uncertainty level for Sm-Nd systematics ((0.5 units for the 143Nd/144Nd ratio and 1.4% for the 147Sm/144Nd ratio); however, they exhibit wellconstrained mantle extraction ages (mean age, 3.35 ( 0.02 Ga). The reproducibility for age-corrected 143Nd/144Nd ratios of the Dalmein pluton is much better than that for its present-day 143Nd/144Nd ratios, which indicates a correlated variation of Nd isotope and Sm/Nd ratios that is due to sample heterogeneity. Reproducibility of Lu-Hf whole-rock data is strongly dependent on total sample decomposition as mentioned above. The ability of sodium peroxide sintering to dissolve zircon was tested using a “synthetic sample” composed of 80 mg of chalcedony powder and 20 mg of zircon crystals (30-60-µm grain size). The full decomposition of zircon crystals was verified by dissolving the remaining precipitates after centrifugation (step 1) in 12 M HCl and checking for undissolved matter under a binocular microscope. No detectable grains or gels remained even though the surface/mass ratio in the “synthetic sample” was smaller than in whole-rock powders. (23) Pin, C.; Zalduegi, J. F. S. Anal. Chim. Acta 1997, 339, 79-89. Mean values obtained for three powder aliquots of JB-1 are ) 0.0147 ( 0.8% and 176Hf/177Hf ) 0.282 974 ( 11. Our elemental Lu/Hf ratios are close to the average of published values that were determined with different analytical techniques (Figure 2) but higher than those determined by Patchett and Tatsumoto21 and Patchett.24 The latter difference is, however, positively correlated with variations in the Hf isotope composition determined during this study and in Patchett and Tatsumoto21 and Patchett,24 thus indicating sample heterogeneity. Indeed, sample heterogeneity in different batches of JB-1, caused by irregular distribution of Hf- or Lu-bearing phases, would perfectly account for the correlated variations observed, as revealed by the respective mantle extraction ages of JB-1 in both studies: These are identical within uncertainties (mean age, 0.62 ( 0.02 Ga and 0.64 ( 0.12 Ga, respectively). Compared to JB-1, the dioritic reference material DR-N shows poorer reproducibilities for 176Lu/177Hf and 176Hf/177Hf ratios of 4.1% and (2 units, respectively. However, determined concentrations for Lu and Hf agree well with reported literature values (e.g., Garbe-Schönberg:25 Lu, 0.34 µg/g; Hf, 3.56 µg/g). Since DR-N is coarser grained than JB-1 and it is rich in zircons, poorer reproducibilities for its Lu-Hf analytical data were expected due to greater sample heterogeneity. A recently published26 value for the Hf isotope composition of DR-N of 0.282 860 ( 6 (2 SE) is higher than those reported in this study. Unfortunately, the author did not report the Lu/Hf ratio, and therefore, heterogeneity within the different batches cannot be accessed via Hf isotope composition versus Lu/Hf correlations. To further evaluate the influence and dependence of sample heterogeneity with respect to Lu-Hf systematics, five powder aliquots from the Archaean Dalmein pluton were also analyzed. The effect of sample heterogeneity should increase with the age of the sample due to variable radiogenic ingrowth in different phases. The reproducibility of the 176Lu/177Hf ratio of the Dalmein pluton is (4.0%, which is similar to that of DR-N ((4.1%). The measured present-day 176Hf/177Hf ratios, on the other hand, are well reproducible. If poorer reproducibilities in 176Lu/177Hf ratios for felsic crustal rocks are truly caused by sample heterogeneity, reproducibility of agecorrected 176Hf/177Hf ratios should stay at the same level or diminish, compared to that of present-day 176Hf/177Hf ratios. Indeed, mantle extraction ages (3.34 ( 0.06 Ga) and age-corrected 176Hf/177Hf (0.280 694 ( 34) are highly reproducible (Table 2). Effect of Zircon on Sm-Nd and Lu-Hf Systematics. Re-Os is another isotope system that is highly sensitive to heterogeneous phase distribution in sample powders.12 The influence on Re-Os systematics by irregular distribution of osmiridium in powder aliquots of the same rock sample, the so176Lu/177Hf (24) Patchett, P. J. Lithos 1983, 16, 47-51. (25) Garbe-Schönberg, C.-D. Geostand. Newsl. 1993, 17, 81-97. (26) Blichert-Toft, J. Geostand. Newsl. 2001, 25, 41-56. (27) Watkins, J. P.; Nolan, J. Geostand. Newsl. 1990, 14, 11-20. (28) Verma, S. P. Geostand. Newsl. 1991, 15, 129-134. (29) Nägler, Th. F.; Kramers, J. D. Precambrian Res. 1998, 91, 233-252. (30) Nir-el, Y.; Lavi, N. Appl. Radiat. Isot. 1998, 49, 1653-1655. (31) Kramers J. D.; Kleinhanns, I. C.; Kreissig, K., Naegler, Th. F. EUG 11 Abst. Vol. 2001, p 421. (32) Patchett, P. J. Geochim. Cosmochim. Acta 1983, 47, 81-91. (33) Stevenson, R. K.; Patchett, P. J. Geochim. Cosmochim. Acta 1990, 54, 16831697. (34) Amelin, Y.; Lee, D.-C.; Halliday, A. N. Geochim. Cosmochim. Acta 2000, 64, 4205-4225. Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 71 Table 2. Sm-Nd and Lu-Hf Data for JB-1, DR-N, and the Dalmein Plutona Sm (ppm) Nd (ppm) 147Sm/ 143Nd/ 144Nd 144Nd TDM(Nd)b (Ga) sinter 1 sinter 2 sinter 3 5.25 5.34 5.20 27.32 27.69 27.09 0.1162 0.1166 0.1161 0.512754 ( 07 0.512731 ( 08 0.512742 ( 07 0.61 0.64 0.62 average 2 SD 5.27 0.14 27.37 0.61 0.1163 0.0005 0.512743 0.000023 0.62 0.04 HF-HNO3 1 HF-HNO3 2 HF-HNO3 3 5.19 5.22 5.15 26.85 27.04 26.82 0.1170 0.1168 0.1161 0.512744 ( 06 0.512754 ( 06 0.512743 ( 06 0.63 0.61 0.62 average 2 SD 5.19 0.07 26.90 0.24 0.1166 0.0009 0.512747 0.000013 0.62 0.02 Hf (ppm) 176Lu/ 176Hf/ 177Hf 177Hf TDM(Hf)c (Ga) 0.337 0.359 0.331 3.25 3.44 3.20 0.0147 0.0147 0.0146 0.282980 ( 07 0.282970 ( 07 0.282972 ( 07 0.61 0.63 0.63 0.34 0.03 3.30 0.26 0.0147 0.0001 0.282974 0.000011 0.62 0.02 0.303 0.313 0.303 3.56 3.61 3.76 0.0121 0.0123 0.0114 0.282956 ( 47 0.282895 ( 63 0.282939 ( 89 0.60 0.71 0.62 Lu (ppm) 143Nd/ 144Nd (t)d 176Hf/ 177Hf (t)d JB-1 HF-HCl24 HF-HNO321 HF-HNO321 sinter 1 sinter 2 sinter 3 sinter 4 5.26 5.25 5.23 5.27 23.63 23.48 23.46 23.70 0.1346 0.1352 0.1346 0.1344 0.512405 ( 08 0.512404 ( 08 0.512407 ( 06 0.512407 ( 07 1.33 1.34 1.33 1.33 DR-N 0.374 0.378 0.373 0.376 3.43 3.57 3.46 3.61 0.0154 0.0150 0.0153 0.0147 0.282752 ( 11 0.282809 ( 12 0.282743 ( 11 0.282755 ( 12 1.09 0.96 1.11 1.06 average 2 SD 5.25 0.04 23.57 0.23 0.1347 0.0007 0.512406 0.000003 1.33 0.02 0.375 0.004 3.52 0.17 0.0151 0.0006 0.282765 0.000060 1.06 0.13 HF-HNO3 1 HF-HNO3 2 HF-HNO3 3 5.24 5.26 5.26 23.42 23.50 23.54 0.1353 0.1352 0.1351 0.512406 ( 08 0.512403 ( 07 0.512412 ( 06 1.34 1.35 1.33 average 2 SD 5.25 0.02 23.49 0.12 0.1352 0.0002 0.512407 0.000009 1.34 0.02 HClO4-HF-HNO323 HClO4-HF-HNO323 5.22 5.22 23.3 23.3 0.1355 0.1356 0.512425 ( 07 0.512428 ( 08 1.31 1.31 sinter 1 sinter 2 sinter 3 sinter 4 sinter 5 8.26 8.73 7.96 8.74 8.51 52.44 54.06 49.34 54.56 53.48 0.0964 0.0976 0.0975 0.0968 0.0961 0.510460 ( 10 0.510483 ( 08 0.510482 ( 09 0.510481 ( 07 0.510464 ( 08 Dalmein Pluton 3.35 0.230 3.36 0.244 3.35 0.223 3.33 0.246 3.34 0.232 4.35 4.73 4.39 4.66 4.35 0.0075 0.0073 0.0072 0.0075 0.0075 0.281163 ( 09 0.281133 ( 09 0.281163 ( 13 0.281138 ( 13 0.281150 ( 10 3.32 3.35 3.29 3.36 3.35 0.508412 0.508409 0.508410 0.508423 0.508420 0.280702 0.280683 0.280720 0.280679 0.280685 average 2 SD 8.46 0.64 52.78 4.15 0.0969 0.0013 0.510474 0.000022 4.49 0.37 0.0074 0.0003 0.281149 0.000027 3.34 0.06 0.508415 0.000013 0.280694 0.000034 3.35 0.02 0.235 0.020 a In-run precisions given as 2 standard errors in the last decimal places; 143Nd/144Nd 0.511066 ( 18 corresponding to La Jolla 143Nd values of Pin and Zalduegi;23 JMC 475 176Hf/177Hf of 0.282169 ( 16 during this study; quoted values of Patchett and Tatsumoto.21 b Calculated after Nägler and Kramers.29 c Calculated with the 176Lu decay constant30 of 1.858 × 10-11 yr-1 and after Kramers et al.31 d Calculated using the zircon age of Kamo and Davis.22 Figure 2. Reproduced Lu/Hf ratios of JB-1 obtained by sodium peroxide sintering and MC-ICPMS measurements (triangles) compared to published Lu/Hf ratios: TIMS, 9,21,24 (recalculated from original isotope ratios); INAA, (; SIMS, 2; ICPSM, b. SIMS, and ICPMS data compiled by the Geological Survey of Japan (available on www.aist.go/GSJ)). Data points displaying ∼60% scatter (excluding three most extreme values ∼30%) indicating sample heterogeneity within different batches of JB-1. called “nugget effect”, is similar to the effect zircon has on the Lu-Hf systematics. Poor reproducibilities for Zr concentrations of zircon-bearing samples are explained by sample heterogeneity with respect to the distribution of zircon throughout the whole72 Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 rock powder.11 Since Hf and Zr display similar geochemical behavior, poor reproducibilities of the 176Lu/177Hf ratios in felsic crustal rocks are no surprise. Zircon, an accessory phase in felsic crustal rocks, displays Hf concentrations 4 orders of magnitude higher than whole rock, whereas the concentration difference for Lu is only 2 orders of magnitude. In conclusion, reproducibility of the 176Lu/177Hf ratio depends mainly on the distribution of Luand Hf-bearing phases throughout powdered whole-rock samples. However, 147Sm/144Nd ratios are much less affected by heterogeneous phase distribution. Sm/Nd ratios are more evenly distributed throughout different mineral phases, and differences in the order of that described for zircon with respect to the Lu-Hf system do not occur. Model calculations reveal that a 4% variation in the Lu/Hf ratio of felsic crustal rocks, as obtained for aliquots of DR-N and Dalmein pluton, can be caused by only 1.5 µg of additional zircon relative to an aliquot size of 100 mg. CONCLUSIONS Using a time-efficient analytical approach to combined Lu-Hf, Sm-Nd, and REE determinations, as presented here, demagnifies problems related to sample heterogeneity. The latter was found to be important even in homogenized international rock reference materials. The principal cause of sample heterogeneity can be traced to irregular distribution of mineral phases, with large concentration differences relative to whole rock of one of the respective elements. In particular, zircon strongly affects the Lu-Hf systematics of powder aliquots, whereas its effect on Sm-Nd systematics is much smaller. The perfect reproducibility in age-corrected 176Hf/177Hf ratios and Hf mantle extraction ages of the powdered reference rock materials is the strongest evidence that the observed variability of (4% for the 176Lu/177Hf ratio in felsic crustal rocks is due to the above-mentioned feature. This variability therefore ultimately limits the reproducibility of Lu/Hf ratios. As this type of heterogeneity is significantly more important in different batches of the same geological material, comparison of separated Lu-Hf, Sm-Nd, and REE determinations should be avoided. Direct comparison of Lu-Hf systematics with Sm-Nd systematics and REE pattern therefore requires separation and determination of these geochemical tracers to be done on a single rock powder aliquot. ACKNOWLEDGMENT This study was supported by the Swiss National Science Foundation, Grant 20-61933-00. Ronny Schoenberg, Barbara Seth, and Igor Villa are thanked for fruitful discussions and constructive criticism on an earlier version of the manuscript. Two anonymous reviewers are thanked for thoughtful comments. Received for review June 25, 2001. Accepted September 19, 2001. AC010705Z Analytical Chemistry, Vol. 74, No. 1, January 1, 2002 73