Precise Determination of Mass-Dependent Variations in the Isotopic

Initial studies focused on material from the Climax mine, USA. Figure 3a presents results from an experiment in which a 5-mg MoS2 flake from this loca...
11 downloads 0 Views 82KB Size
Anal. Chem. 2001, 73, 1425-1431

Precise Determination of Mass-Dependent Variations in the Isotopic Composition of Molybdenum Using MC-ICPMS A. D. Anbar,*,†,‡ K. A. Knab,‡ and J. Barling†

Department of Earth and Environmental Sciences and Department of Chemistry, University of Rochester, Rochester, New York 14627

We present an analytical approach for the precise determination of mass-dependent differences in the isotopic composition of Mo between samples and reference standards using multiple-collector magnetic sector inductively coupled plasma mass spectrometry (MC-ICPMS). Either Zr or Ru “element spikes” are employed to correct for instrumental mass bias. Differences in 95Mo/97Mo can be determined to a precision of (0.2‰ ((2σ) using 1-10 µg of Mo. Similar precision is possible for other ratios after correction for isobaric interferences from either spike element. This approach facilitates study of mass-dependent variations in the isotopic composition of Mo in nature and in materials produced by laboratory processes. We observe fractionation of Mo isotopes of ∼1.5‰/amu during ion-exchange chromatography in the laboratory and a shift of ∼0.3‰/amu between natural MoS2 and a laboratory standard. High-precision isotopic analyses have recently revealed that the isotopes of Cu, Zn, and Fe can be chemically fractionated in the laboratory and that the isotopic compositions of these elements are variable in nature.1-5 To detect such effects requires analytical techniques capable of measuring differences in isotopic composition between samples and reference standards to a precision of better than 1 parts in 1000 (1‰). Such high-precision measurements are expected to be of broad utility in elucidating the environmental chemistry of these and other transition metals once specific fractionation mechanisms are understood. Biologically mediated isotope fractionation is of particular interest because of the critical roles of many metals in biochemistry,6 the growing realization that availability of key transition metals affects biological

activity on a global scale,7 and the potential that metal isotope fractionation may provide a signature of biological activity in the geological record.2 Here, we present methods for the routine determination of Mo isotope variations to a precision of ∼0.2‰ using multiple-collector, magnetic sector inductively coupled plasma mass spectrometry (MC-ICPMS). Variations in Mo isotope composition have not been examined to such precision previously despite the unique and critical role of this metal in nitrogen fixation and denitrification and in human metabolism. Mo is also of interest because its environmental chemistry is unusually redox-sensitive,8-10 raising the possibility that the isotopic composition of Mo in ancient sediments provides information about past redox conditions. The primary analytical challenge in high-precision isotopic studies is to compensate for mass-dependent effects arising from the mass spectrometric analysis itself (“mass bias”) without “erasing” the naturally occurring mass-dependent variations being studied. Because the magnitude of any natural variation is expected to be on the order of parts per thousand or smaller, this correction must be very precise. Precise measurements of Mo isotope compositions have been obtained in several recent studies using magnetic sector thermal ionization mass spectrometry (TIMS)11-13 and in one study using MC-ICPMS.14 In these studies, normalization for mass bias was achieved by comparing the measured value of a selected Mo isotope ratio to an assumed “true” value. This method is useful in isotope tracer studies or when searching for non-mass-dependent anomalies in isotopic composition as may result from radioactive decay or nucleosynthetic reactions. However, because it erases both the instrumental bias and any naturally occurring mass-dependent variation, this method is useless for studying Mo isotope fractionation.



Department of Earth and Environmental Sciences. Department of Chemistry. (1) Anbar, A. D.; Roe, J. E.; Barling, J.; Nealson, K. H. Science 2000, 288, 126128. (2) Beard, B. L.; Johnson, C. M.; Cox, L.; Sun, H.; Nealson, K. H.; Aguilar, C. Science 1999, 285, 1889-1892. (3) Mare´chal, C. N.; Telouk, P.; Albarede, F. Chem. Geol. 1999, 156, 251273. (4) Zhu, X.-K.; O’Nions, K.; Guo, Y.; Reynolds, B. C. Science 2000, 287, 20002002. (5) Zhu, X. K.; O’Nions, R. K.; Guo, Y.; Belshaw, N. S.; Rickard, D. Chem. Geol. 2000, 163, 139-149. (6) Frausto da Silva, J. J. R.; Williams, R. J. P. The Biological Chemistry of the Elements: The Inorganic Chemistry of Life; Clarendon Press: Oxford, 1991. ‡

10.1021/ac000829w CCC: $20.00 Published on Web 03/06/2001

© 2001 American Chemical Society

(7) Falkowski, P. G.; Barber, R. T.; Smetacek, V. Science 1998, 281, 200-206. (8) Moreford, J. L.; Emerson, S. Geochim. Cosmochim. Acta 1999, 63, 17351750. (9) Emerson, S. R.; Huested, S. S. Mar. Chem. 1991, 34, 177-196. (10) Crusius, J.; Calvert, S.; Pederson, T.; Sage, D. Earth Planet. Sci. Lett. 1996, 145, 65-78. (11) Turnlund, J. R.; Keyes, W. R.; Peiffer, G. L. Anal. Chem. 1993, 65, 17171722. (12) Qi-Lu; Masuda, A. Int. J. Mass Spectrom. Ion Processes 1994, 130, 65-72. (13) Yin, Q. Z.; Jacobsen, S. B. Lunar and Planetary Science Institute, Houston, TX, 1998. (14) Lee, D. C.; Halliday, A. N. Int. J. Mass Spectrom. Ion Processes 1995, 146, 35-46.

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001 1425

Table 1. Mass Analysis Parameters on the Plasma54 for Mo Isotope Analysis with Zr and Ru Spikesa mass 90 molybdenum zirconium ruthenium cup configuration Zr % contribution Mo % contribution cup configuration Ru % contribution Mo % contribution

51.5

L-4 100 0

91

92

94

11.2

14.8 17.2

9.25 17.4

L-3 100 0

L-2 37 63

95

96

Abundancesb 15.9 16.7 2.80 5.52

97

98

9.55

24.1 1.88

Ax 48 52

Zr Spike Method H-1 H-2 0 8 100 92

H-3 0 100

H-4 0 100

L-4 0 100

Ru Spike Method L-3 L-2 0 14 100 86

L-1 0 100

Ax 4 96

99

100

101

102

9.63 12.7

12.6

17.0

31.6

H-1 100 0

H-2 40 60

H-3 100 0

H-4 100 0

a For both Zr and Ru spike methods, the relative contributions of the spike element and of Mo to the signal measured at each mass are given, assuming that the concentration of the spike element is ∼50% of the Mo concentration. b Natural average abundances (%). Because it is not analyzed, 104Ru is omitted.

Mass-dependent variations of Mo isotopes have been examined previously only by TIMS. Using TIMS, instrumental mass bias can be corrected without erasing natural variations by carefully controlling operating conditions to limit the extent of isotope “distillation”15,16 or by using an internal standard in the form of a Mo isotopic “double spike” added to the sample.17 The doublespike approach is considerably more reliable and has yielded a precision of ∼1% on the 92Mo/100Mo ratio. However, in practice, the double-spike method is limited by the fact that spikes are never completely free of the nonspike Mo isotopes, while samples are never free of the spike isotopes. Therefore, the precision obtained depends strongly on the isotopic purity of the spike, the abundance of the spike isotopes in nature, and the precision to which the isotopic composition of the spike is known. To minimize propagation of uncertainties stemming from these complications, the spike/sample ratio must be carefully optimized. Because of these considerations, double-spike analyses are not routine. The unique characteristics of ICPMS permit compensation for instrumental mass bias to high precision using an “element spike”. This method, originally developed for the study of radiogenic Pb isotopes, takes advantage of the secular stability of instrumental mass bias in ICP systems and the observation that the magnitude of this bias per amu mass difference (f) is similar among different elements over a limited mass range.18 Therefore, when element i of unknown isotopic composition is being analyzed, fi can be determined by assuming fi ) fj, where j is a spike element of known isotopic composition added to the sample. This approach is analogous to use of an isotopic double spike but carries none of the complications as long as the sample is purified of the spike element. With MC-ICPMS systems, the element spike method has been used to obtain very precise radiogenic Pb isotope data using Tl spikes,19-23 and to study mass-dependent variations in the isotopic

compositions of a small number of transition metals.1,3,5,24 Significantly, data obtained in one laboratory using MC-ICPMS indicate that the fi * fj at high precision.3,23 Below, we demonstrate the applicability of this method to the precise determination of variations in the isotopic composition of Mo. Two elements are particularly good spike candidates: zirconium and ruthenium (Table 1). The use of both elements is explored, and we provide confirmation that fi * fj at high precision. Using these methods, we provide evidence that Mo isotopes can be fractionated by simple chemical processes. To our knowledge, this is the first demonstration of such fractionation of Mo isotopes. These results provide confidence in the use of MC-ICPMS for high-precision studies of mass-dependent isotope fractionation in general, and for Mo isotope studies in particular, and should facilitate novel research in Mo biogeochemistry, environmental chemistry, and bioinorganic chemistry.

(15) Murthy, V. R. Geochim. Cosmochim. Acta 1963, 27, 1171-1178. (16) Crouch, A. C.; Tulpin, T. A. Nature 1964, 202, 1282-1284. (17) Wetherill, G. W. J. Geophys. Res. 1964, 69, 4403-4408. (18) Longerich, H. P.; Fryer, B. J.; Strong, D. F. Spectrochim. Acta B 1987, 42, 39-48. (19) Rehka¨mper, M.; Halliday, A. M. Int. J. Mass Spectrom. Ion Processes 1998, 59, 123-133.

(20) Walder, A. J.; Furuta, N. Analy. Sci. 1993, 9, 675-680. (21) Belshaw, N. S.; Freedman, P. A.; O’Nions, R. K.; Frank, M.; Guo, Y. Int. J. Mass Spectrom. Ion Processes 1998, 181, 51-58. (22) Hirata, T. Analyst 1996, 121, 1407-1411. (23) White, W. M.; Albare`de, F.; Te´louk, P. Chem. Geol. 2000, 167, 257-270. (24) Rehka¨mper, M.; Halliday, A. N. Geochim. Cosmochim. Acta 1999, 63, 935944.

1426

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

ANALYTICAL METHODS Below, we discuss materials, reagents, and sample preparation; instrument configuration; and correction for isobaric interferences between Mo and spike elements. Evaluation of the element spike approach for Mo, and the application of this method, are presented in Results and Discussion. Materials, Reagents, and Sample Preparation. Labware used in this study was composed of polyethylene or Teflon and was cleaned prior to use with a degreasing agent followed by sequential washing in reagent-grade HNO3, HCl, and 18 MΩ H2O. All dissolutions and chemical separations and dilutions were carried out using ultrapure sub-boiling quartz-distilled HNO3 and HCl (Seastar Baseline) and 18 MΩ H2O. All manipulations were carried out using standard trace metal cleanlab protocols in laboratories with HEPA-filtered air supplies.

The spike element (either Zr or Ru) was added to samples immediately before ICPMS analysis, by pipetting from ∼100 ppm dilutions of stock solutions. The stock solutions consisted of Johnson Matthey Specpure plasma standard solutions, drawn from Lot No. 700193E (Zr) and Lot No. 701337D (Ru). It was useful to adopt a Mo isotopic standard with which to compare samples. For this purpose, we used the Johnson Matthey Specpure molybdenum plasma standard (Lot No. 7024991). This “JMC-Mo” standard is not a well-defined isotopic standard but was an adequate “working standard” for the purpose of this study. To validate the element spike approach, it was also useful to prepare standards artificially enriched in 95Mo compared to JMCMo. For this purpose, we used a spike solution prepared gravimetrically from Mo metal enriched in 95Mo (Oak Ridge National Laboratory). Mass spectrometric analysis of this spike solution indicated that it contained 61.8% 95Mo and negligible amounts of 97Mo. Aliquots of this solution were mixed with aliquots of JMC-Mo to produce two “Spike-Mo” solutions (SpikeMo A and B). If the isotopic composition of JMC-Mo is assumed to be the same as reported by Qi-Lu and Masuda,12 then the 95Mo/ 97Mo ratio in Spike-Mo A was shifted from the ratio in JMC-Mo by 3.46 ( 0.10‰ and Spike-Mo B was shifted by 1.52 ( 0.10‰. The uncertainties in these calculations reflect uncertainties in the isotopic composition and concentration of the spike solution (determined by reverse isotope dilution) and uncertainties in gravimetry. There is also uncertainty in the isotopic composition of JMC-Mo because of the possibility of mass fractionation of terrestrial Mo. However, as long as 95Mo/97Mo in JMC-Mo is within 1% of the assumed value, the contribution of this uncertainty to the expected shifts in composition is minor. Natural samples analyzed here include MoS2 (molybdenite) ore samples. MoS2 was chosen for these initial experiments because of this mineral’s high Mo content, simple stoichiometry, and ease of dissolution. As a result, the isotopic composition of Mo in MoS2 can be determined by direct MC-ICPMS analysis of dissolved MoS2, without need for preconcentration or purification of Mo. Thus, these samples were dissolved in aqua regia, dried to a ∼20-µL drop, and reconstituted in 0.05 M HNO3 to [Mo] ∼6 ppm. Because isotopic analysis of Mo from more complex natural matrixes (e.g., silicate rocks) requires chemical preconcentration of Mo, we also evaluated the fractionation of Mo during chromatographic elution from anion-exchange resin (Bio-Rad AG-1 × 8) using HCl. Instrument Configuration. All measurements reported here were made on the Plasma54 MC-ICPMS (VG Elemental) at the University of Rochester. This instrument is equipped with nine Faraday cups (four moveable low mass, four moveable high mass, one fixed axial). Analyses were conducted in static mode. The mass dispersion and maximum collector spacing of the Rochester Plasma54 permit a range of ∼8 amu at the mass values of Mo. Therefore, the only spike elements that have at least two isotopes that can be measured at the same time as Mo are Zr and Ru (Table 1). Use of other spike elements is possible if data are acquired in two cycles, with different magnetic field settings. However, precision is compromised by this approach because of the instability of the ion beam on the time scale needed to change the magnetic field (∼2 s).

The masses analyzed are illustrated in Table 1. When Mo was analyzed using a Zr spike, 100Mo was not measured; when a Ru spike was used, 92Mo was not measured. For the mass arrays shown in Table 1, there is only one pair of Zr isotopes that can be used to determine the mass bias, 90Zr and 91Zr. In the case of Ru, 99Ru, 101Ru, and 102Ru may be monitored, resulting in a choice of three different isotope pairs for determining the mass bias (104Ru cannot be monitored simultaneously with 94Mo). All other Zr and Ru masses are isobaric with isotopes of Mo, making them unacceptable as mass bias monitors. The Plasma54 was allowed to stabilize for at least 4 h prior to taking any measurements. This was found to be the minimum time needed to ensure stability in the high-voltage power supply, which ultimately affects the stability of the instrumental mass bias. Solutions (0.05 N HNO3 containing ∼1-6 ppm Mo and 0.5-3 ppm of either spike) were introduced to the instrument by free aspiration, typically through a concentric glass nebulizer (Glass Expansion) mated to a microbore Teflon uptake tube (All Tech Associates) and a Scott-type spray chamber. This system typically generated a total Mo signal of ∼2 V/ppm, at an uptake rate of ∼100 µL/min. In some runs, a microconcentric nebulizer and desolvating system (Transgenomic/CETAC Aridus I) was used in place of the standard nebulization system. This increased the sensitivity for Mo to ∼8 V/ppm. Each isotopic analysis represents the average of 30-50 ratio measurements, each of which was integrated over 5 s. These measurements were divided into 3-5 blocks of 10 cycles each. Correction for Isobaric Interferences. When either Zr or Ru spikes are used, isobaric interferences are introduced at all but three of the measured Mo isotopes (Table 1). Zr interferes at 92Mo, 94Mo, and 96Mo while Ru interferes at 96Mo, 98Mo, and 100Mo. Only 95Mo and 97Mo are free in all cases. However, we explored the possibility of obtaining high-precision data for all Mo isotopes with either spike after subtracting the contribution from the spike element. Such correction entails some tradeoffs. It is desirable for the spike element to be present in sufficient abundance so that the isotopic ratio monitored for mass bias correction can be measured to a precision similar to that of the Mo ratios. At the same time, however, the interferences caused by the addition of the element spike need to be minimized so that the error associated with the interference correction is also minimized. In practice, this means that the Ru or Zr spike is added at ∼50% of the concentration of Mo in the sample solution. However, even at this concentration, a spike isotope can contribute nearly 50% of the signal at some Mo masses (e.g., 94Zr). Because of the magnitude of these interferences, any correction for them must be extremely accurate. This necessitates correction of the assumed natural abundance of the spike isotopes for the instrumental mass bias. The accuracy of this correction is limited primarily by the accuracy with which the true isotope abundances of the element spike are known and by the assumed mass fractionation law. RESULTS AND DISCUSSION Validity of the Element Spike Approach. Using a Zr spike, Mo isotope measurements can be corrected for instrumental mass Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

1427

bias using the expressions

(9XMo/9YMo)true ≈ (9XMo/9YMo)measured(M9X/M9Y)fMo

(1)

and

fMo ) fZr ) ln[(90Zr/91Zr)measured/(90Zr/91Zr)true]/ ln(M90/M91) (2)

where MN are atomic weights. These equations assume that mass bias follows the so-called “exponential law” widely used in TIMS25,26 and shown to be valid in MC-ICPMS.3,23 The similar “power law” has also been used in MC-ICPMS.19,22 These equations can be manipulated to show that ln(9XMo/ 9YMo) 90 91 measured ) m ln( Zr/ Zr)measured + b, where

m ) ln(M9X/M9Y)/ln(M90/M91)

(3)

b ) ln(9XMo/9YMo)true - m ln(90Zr/91Zr)true

(4)

and

Analogous relationships can be derived when Ru is used in place of Zr or when the power law is assumed in place of the exponential law. Therefore, the validity of the element spike approach for Mo with either Zr or Ru as the spike element can be tested by plotting ln(9XMo/9YMo)measured vs ln(90Zr/91Zr)measured or ln(99Ru/101Ru)measured for data sets that span a range of values of instrumental mass bias (the selection of 99Ru/101Ru is arbitrary; comparable results can be seen with other Ru ratios.). The observed slope on such a plot can be compared to the slope predicted a priori from eq 3. Figure 1 illustrates the typical relationships observed between Mo and the spike elements during analysis of the JMC-Mo standard. In this figure, we consider 95Mo/97Mo in order to avoid uncertainties arising from the correction for isobaric interferences at other Mo isotopes. The internal precision of each ratio measurement was typically e(0.0040% ((2σmean). It is evident that the data exhibit only minor deviations from linearity, consistent with eqs 3 and 4. However, the observed slopes deviate from the values predicted from either the exponential or power laws. Moreover, it is found that while the observed value of m is constant during a given analytical session, typically differences of up to 10% are seen when comparing data from different sessions. The variation of m from run to run indicates that no single mass fractionation law describes the system. These observations are similar to those made for Cu and Zn isotopes.3 Mare´chal et al.3 proposed that these observations are evidence that the magnitude of instrumental mass bias of Cu and Zn isotopes in ICPMS is only approximately independent of the element involved, and White et al.23 demonstrated this to be the case also for Pb and Tl. The same may be true of Mo, Ru, and Zr. This can be expressed mathemati(25) Russell, W. A.; Papanastassiou, D. A.; Tombrello, T. A. Geochim. Cosmochim. Acta 1978, 42, 1075-1090. (26) Hart, S. B.; Zindler, A. Int. J. Mass Spectrom. Ion Processes 1989, 89, 287301.

1428

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

Figure 1. Logarithmic plots of 95Mo/97Mo against 90Zr/91Zr (a) and 99Ru/101Ru (b). Data were obtained during analysis of the JMC-Mo standard using the instrument configurations in Table 1. Observed slopes are compared to those predicted from the exponential and power mass fractionation laws (see text). Representative (2σ error bars are given in the lower right.

cally by explicitly accounting for the possibility that fMo * fZr, in which case eq 3 becomes

m ) (fMo/fZr)[ln(M9X/M9Y)/ln(M90/M91)]

(5)

Hence, the linearity seen in Figure 1 is still expected as long as fMo/fZr does not vary with time. The same is true when Ru is used. In the runs shown here, fMo/fZr ∼ 0.95 and fMo/fRu ∼ 0.90. These values are similar to fPb/fTl ∼ 0.97.23 However, at present the value of m cannot be predicted. An important implication of these findings is that systematic inaccuracy can be introduced to the determination of (9XMo/ 9YMo) 9XMo/9YMo are corrected for true when measurements of instrumental mass bias using the element spike method and either the exponential or power law expressions. However, it is critical to recognize that, as with Cu and Zn isotopes, this inadequacy need not be an obstacle for Mo isotope fractionation studies where the quantity of interest is the difference in isotopic compositions between a sample and a standard rather than the absolute isotopic compositions themselves. If we adopt

Figure 2. Logarithmic plots of 95Mo/97Mo against 90Zr/91Zr (a) and 99Ru/101Ru (b). Data were obtained during analysis of the Spike-Mo solutions (open circles) and the JMC-Mo standard (filled circles) using the instrument configurations in Table 1. δ95Mo for each analysis of the Spike-Mo solutions is determined from the distance along the ordinate axis between the measured Spike-Mo ratio and the trend line defined by the standard measurements. Representative (2σ error bars are given in the lower right.

eqs 4 and 5 as an accurate description of the system, then data obtained from materials with different values of (9XMo/9YMo)true obtained during a single analytical session should lie along parallel lines with different intercepts in ln/ln space. Therefore, the difference in (9XMo/9YMo)true between a sample and a standard can be derived from the distance along the ordinate axis between the measured isotopic composition of a sample and the regression line defined by repeated measurements of the standard under a range of instrumental mass bias (e.g., Figure 2). The accuracy and precision of this “graphical” method depends on the constancy of fMo/fZr, or fMo/fRu during the analysis and between samples and standards. Fortunately the extremely good correlations seen in Figure 1 indicate that the ratios fMo/fZr and fMo/fRu are approximately constant during a single analytical session, as in the case of fCu/fZn and fPb/fTl.3,23 Session-to-session variations in m are ascribed to small differences in gas flow rates and in the alignment between the plasma and the mass analyzer each time the plasma is lit, which may affect the relative fractionation of the elements during extraction to the mass spectrometer.

It should be emphasized that this data reduction method is independent of whether the exponential or power laws govern mass bias, as both laws will produce linear and parallel data trends on a ln/ln plot. However, it is also important to emphasize that we cannot a priori exclude the possibility that mass bias in MCICPMS only approximates such behavior; both the exponential and power laws are empirical and, hence, could prove to be inadequate models at high levels of precision. Nonlinearity and/ or nonparallelism could lead to a systematic failure of the graphical approach to properly determine differences in isotopic composition. This concern motivated preparation and analysis of the gravimetrically prepared Spike-Mo solutions, which have known deviations in 95Mo97Mo. Accuracy and Precision. We measured the difference in 95Mo/97Mo between JMC-Mo and the Spike-Mo solutions to evaluate the accuracy and precision with which variations in Mo isotopic composition can be determined. As noted above, the difference in 95Mo/97Mo between these solutions is known independently to (0.1‰. Spike-Mo A was used to assess the use of Zr as the spike; Spike-Mo B was used to assess the use of Ru. Because m may vary from session to session, measurements of each spike solution were compared only to standards measured during the same ∼10-h analytical run. Also, to minimize the influence of possible in-run fluctuations of fMo/fZr and fMo/fRu, the analysis of each sample was bracketed with analyses of standards. This procedure is modeled after the approach of Mare´chal et al.3 for Cu and Zn. Typical data (Figure 2) exhibit excellent linearity and parallelism using either Zr or Ru. Displacements between the lines defined by the JMC-Mo and Spike-Mo data arrays are obvious. In Figure 2a, the displacement of each measurement of SpikeMo from the JMC-Mo array leads to an independent determination of the difference in isotopic composition, reported as δ95Mo (δ95Mo ) [((95Mo/97Mo)sample/(95Mo/97Mo)JMC-Mo - 1) × 1000‰]). There are six such determinations in this particular experiment, with an average value of 3.42 ( 0.06‰ ((2σmean). This result is in good agreement with the expected shift of Spike-Mo A (3.46 ( 0.10‰). For Spike-Mo B (Figure 2b), we obtained δ95Mo ) 1.52 ( 0.01‰ ((2σmean), also in excellent agreement with the expected shift (1.52 ( 0.10‰). To estimate the potential external precision of this methodology, as well as the accuracy of the correction for Zr and Ru isobars, we compared the results from 100 measurements of the isotopic composition of JMC-Mo obtained during 11 sessions over a 6-month period. Roughly half of these analyses were run with Zr and half with Ru. Each measurement was translated into a δ9XMo determination by evaluating the displacement along the ordinate axis in ln/ln space between the datum and the regression line defined by the other measurements of JMC-Mo during the same session. This displacement was determined for the ratios 92Mo/ 97Mo (when using Zr),94Mo/97Mo, 95Mo/97Mo, 96Mo/97Mo, 98Mo/ 97Mo, and 100Mo/97Mo (when using Ru). Because JMC-Mo is both the sample and the standard in this experiment, δ9XMo ≡ 0; the measured δ9XMo values should equal zero within the analytical uncertainties. We obtained δ92Mo ) 0.00 ( 0.28‰, δ94Mo ) 0.02 ( 0.18‰, δ95Mo ) -0.01 ( 0.18‰, δ96Mo ) 0.00 ( 0.08‰, δ98Mo ) 0.00 ( 0.10‰, and δ100Mo ) -0.01 ( 0.16‰ ((2σ). No systematic differences were seen among δ94Mo, δ95Mo, δ96Mo, Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

1429

Table 2. Mo Isotope Compositions in Elution Fractions from a Chromatographic Experimenta fraction no.

% total Mo

δ92Mo

δ94Mo

δ95Mo

δ96Mo

δ98Mo

1 2 3 4 5

16 20 18 28 13

-2.66 -1.38 -0.08 1.42 4.59

-1.51 -1.58 -0.71 -0.71 3.27

-0.98 -0.37 0.04 0.75 1.74

-0.52 -0.24 -0.02 0.25 0.91

0.53 0.21 0.01 -0.31 -0.86

cumulative

95

0.3

0.5

0.2

0.1

0.1

aData were obtained using a Zr spike. Mo loaded on the exchange column was identical to the Mo used as a standard in the mass spectrometric analyses. Hence, δ9XMo of the loaded Mo is equal to 0. Cumulative values are integrated over all fractions. Uncertainties on individual δ9XMo are (0.20‰ ((2σ). Cumulative uncertainties are ∼(0.5‰ due to propagation of errors.

and δ98Mo data obtained using Zr and Ru. Significantly, all δ9XMo precisions are comparable in magnitude despite corrections for isobaric interferences at all masses except 95 and 97. In repeated analyses of the Spike-Mo solutions over several months, we observed similar long-term reproducibility in δ95Mo using the Zr spike. On the basis of these observations, δ95Mo of a sample can be determined with an uncertainty of better than (0.20‰ ((2σ) from a single analysis using the graphical method. This is a marked improvement over previous methods used to study Mo isotope fractionation, which were limited to errors of ∼1% on the 92Mo/ 100Mo ratio using larger samples.17 While similar uncertainties are apparently attainable for other δ9XMo determinations, as seen below, results for δ94Mo are sometimes erratic. We attribute this to a sporadic isobaric interference from an as-yet-unidentified polyatomic species. Hence, 94Mo data should be regarded cautiously. We found no evidence of interferences at any other isotopes. Comparable reproducibility of δ95Mo using the Zr spike was observed in repeated analyses of Spike-Mo solutions over many months when, instead of using the graphical approach, data were fully corrected for instrumental mass bias using the exponential law by assuming fMo/fZr ) 1 (eqs 1-4). δ95Mo was determined by comparing the corrected ratios for each sample analysis to the average of the standard analyses from the same analytical session (note that for the purpose of a δ determination, the accuracy of the 90Zr/91Zr ratio used for mass bias correction is not critical as long as the same ratio and Zr source are used for all analyses). This finding suggests that deviations from fMo/fZr ) 1, or from the exponential law, have a negligible contribution to scatter in δ9XMo at the (0.20‰ (2σ) level of precision. The reproducibility of independent δ9XMo determinations during a single session is sometimes considerably better than (0.20‰. Hence, it is possible that the stability of fMo/fZr and/or the validity of the exponential law depends on instrument tuning parameters that are difficult to control from session to session, a matter that warrants further investigation in order to improve precision. Mo Isotope Variations. To explore the application of our methodology to natural samples, and to search for evidence of variability of the isotopic composition of Mo in nature and in the laboratory, we have analyzed fractionation of Mo standards during 1430 Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

Figure 3. δ9xMo in molybdenite (MoS2) samples. The results of three replicate analyses of the same sample using a Ru spike are shown in (a). In (b), two replicate analyses of another sample are shown, one obtained with a Zr spike (triangles) and the other with a Ru spike (crosses). Representative (1σ error bars are given in the lower right.

processing by anion-exchange chromatography and also studied the variation of Mo isotopic composition in natural molybdenite (MoS2) samples from a supergene ore deposit. Anion-Exchange Chromatography. Anion-exchange chromatography is commonly used to separate Mo from natural materials, and isotope fractionation of some elements during chromatographic elution is well known,1,3,27-29 hence, determination of Mo isotope fractionation during elution serves to test whether Mo isotopes can be easily fractionated by chemical processes. In addition, it is important to quantify this fractionation to assess whether ion-exchange purification can produce fractionation artifacts. Here, we loaded a known amount (15 µg) of JMC-Mo on a 2-mL anion-exchange column (BioRad AG 1 × 8) in 6 M HCl. This Mo was then eluted using 1 M HCl and the eluate collected in a series of five fractions containing, respectively, 16, 20, 18, 28, (27) Taylor, T. I.; Urey, H. C. J. Chem. Phys. 1938, 6, 429-438. (28) Russell, W. A.; Papanastassiou, D. A. Anal. Chem. 1978, 50, 1151-1154. (29) Machlan, L. A.; Gramlich, J. W. Anal. Chem. 1988, 60, 37-39.

and 13% of the total Mo. Each fraction was spiked with Zr and analyzed for isotopic composition. The results (Table 2) indicate that Mo isotopes are indeed fractionated during elution from anion resin; the earliest fraction is enriched in the heavier isotopes by ∼0.5‰/amu, whereas the last fraction is enriched in the lighter isotopes by nearly 1‰/ amu. As required by mass balance, δ9XMo ∼ 0 when integrated over all fractions. δ94Mo data are somewhat more variable than others, presumably because of the isobaric interference noted earlier. The magnitude of the observed fractionation is similar to that during elution of Fe and Cu from anion resin,1,3 and Ca from cation resin.28 The cause of such fractionation is not certain but may relate to equilibrium isotope fractionation between dissolved and resin-bound complexes.1 MoS2. MoS2 is among the most readily analyzed natural Mobearing materials because of the high concentration of Mo and ease of acid dissolution. Initial studies focused on material from the Climax mine, USA. Figure 3a presents results from an experiment in which a 5-mg MoS2 flake from this location was divided into three roughly equal pieces which were dissolved in HNO3 and analyzed separately using the Ru spike. δ9XMo values are presented as a function of mass because mass-dependent fractionation should result in a linear array; deviations from this array would suggest isobaric interference. It is clear that the values of δ9XMo determined in these three analyses are statistically indistinguishable, even at the (1σ confidence interval. However, these values are also significantly shifted from JMC-Mo. Importantly, the isotopic shift between the sample and JMC-Mo is systematic for all masses: The abundances of masses 94, 95, 96, 98, and 100 relative to mass 97 are all shifted by ∼0.3‰/amu. This good correlation of δ9XMo with mass is strong evidence of mass-dependent fractionation of Mo isotopes, which eliminates the possibility that isobaric interferences cause these shifts in δ9XMo. Additionally, this correlation provides further evidence that corrections for isobaric interferences from the spike element are successful. Apparently, the isotopic compositions of the sample and JMC-Mo are fractionated such that the MoS2 sample is isotopically “heavier” than JMC-Mo. To confirm that similar results can be obtained using either Zr or Ru, a solution containing dissolved MoS2 from another

Climax sample was divided into two aliquots for isotopic analysis using Zr and Ru, respectively (Figure 3b). These two analyses produce the same shifts in δ95Mo, δ96Mo, and δ98Mo, which define a linear trend on which δ92Mo and δ100Mo both plot. Substantial deviation between the two analyses is seen for δ94Mo, presumably from the sporadic, unidentified interference noted earlier. However, the fact that consistent results are obtained at all other masses using Zr and Ru provides strong evidence that the element spike approach successfully corrects for instrumental mass bias in natural materials. CONCLUSIONS The results presented here demonstrate that “element spikes” can be used with MC-ICPMS to determine δ95Mo to a precision of better than (0.20‰ ((2σ), with similar precision possible for other Mo isotope ratios. We have applied this method to demonstrate that Mo isotopes can be chemically fractionated in the laboratory by ∼1.5‰/amu. This finding suggests that variations of similar magnitude may exist in nature, as has recently been observed for other elements.2-5 We have also characterized the difference in isotopic composition (∼0.3‰/amu) between molybdenite samples and our laboratory Mo standard. This difference may reflect industrial fractionation of Mo during preparation of the standard or natural variation in the isotopic composition of Mo. These results provide a foundation for the systematic exploration of Mo isotopic variations in nature using MC-ICPMS. ACKNOWLEDGMENT We thank G. Ravizza for assistance preparing the Mo spike solutions, and F. Albare`de and C. Mare´chal for discussions in the early stages of this work. J. Roe and E. Ramon provided valuable assistance in the laboratory. This work was conducted at the ICPMS Laboratory at the University of Rochester, with support from NSF-LExEn (CHE 9714282), NSF-ARI (EAR 9601929), and the NASA Astrobiology Institute.

Received for review July 19, 2000. Accepted December 4, 2000. AC000829W

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

1431