High-precision molybdenum isotope analysis by negative thermal ionization mass spectrometry

Abstract

Procedures for the separation, purification, and high-precision analysis of mass-independent isotopic variations in molybdenum (Mo) using negative thermal ionization mass spectrometry are reported. Separation and purification of Mo from silicate and metal matrices are achieved using a two-stage anion exchange chromatographic procedure. Molybdenum is ionized as the MoO₃⁻ species using a double filament assembly. The MoO₃⁻ ion beams are collected using Faraday cup detectors equipped with a mixed array of amplifiers utilizing 10¹¹ and 10¹² Ω resistors, which allows for in situ measurement and correction of oxygen isobars. The long-term external reproducibility of ⁹⁷Mo/⁹⁶Mo, the most precisely measured Mo isotope ratio, is ±5.4 ppm (2SD), based on the repeated analyses of the Alfa Aesar Specpure^® Mo plasma standard and using ⁹⁸Mo/⁹⁶Mo for fractionation correction. The long-term external reproducibilities of 92Mo/⁹⁶Mo, ⁹⁴Mo/⁹⁶Mo, ⁹⁵Mo/⁹⁶Mo, and ¹⁰⁰Mo/⁹⁶Mo are ±107, 37, 23, and 32 ppm (2SD), respectively. With this precision, smaller differences in Mo isotopic compositions can be resolved than have been previously possible.

1. Introduction

The molybdenum (Mo) isotopic system is uniquely suited to address current issues in cosmochemistry from the origin of isotopic heterogeneity in the solar nebula to the diversity of planetary bodies. Though the cosmochemical applications for Mo isotopes are the primary focus of this study, the application of mass-dependent Mo isotope effects as indicators of paleoredox conditions has been the focus of most Mo isotope studies in geochemistry (e.g., [1–3]). In cosmochemistry, “mass-independent” nucleosynthetic Mo isotope anomalies, which have been observed in a variety of cosmochemical materials, including at the bulk meteorite scale, have been used primarily to evaluate the extent of mixing and/or thermal processing in the solar nebula, and to test genetic relations amongst solar system materials (e.g., [4–7]). The use of Mo as a genetic tracing tool is important for the elucidation of nebular and planetary processes, including the characterization of the nebular feeding zones of planetary bodies. By merging groups or incorporating ungrouped or anomalous meteorites into existing groups, the accretionary histories and evolution of meteorite parent bodies can be further developed. As a corollary, by demonstrating that certain solar system objects are unrelated, the parent bodies’ evolutionary histories can be clarified.

Molybdenum has been previously analyzed by thermal ionization mass spectrometry in both positive (P-TIMS – e.g., [8–11]) and negative mode (N-TIMS – [12,5,6,13]). Molybdenum isotopes in cosmochemical and terrestrial materials have also been measured by multi-collector inductively coupled plasma mass spectrometry (MC-ICP-MS) (e.g., [14,15,4,7,16]). The current state-of-the-art precision (2SD) of the measured ⁹⁷Mo/⁹⁶Mo isotope ratio is ±13 ppm by N-TIMS [13] and ±21 ppm by MC-ICP-MS [13]. Although the method of Nagai and Yokoyama [13] produces high-precision Mo isotope ratios, 1000 ng–4000 ng of Mo were used in their study. This requires large sample masses of up to 4 g for some chondritic meteorites. Due to the valuable nature of cosmochemical materials, it is desirable to use as little material as possible.

In an effort to increase the resolving power of the Mo isotopic system when ≤1000 ng of Mo are available, we developed procedures for the separation, purification, and high-precision analysis of Mo using a Thermo-Fisher Triton Plus N-TIMS at the University of Maryland. Using N-TIMS has the advantage of more stable signal intensities and lower likelihood of isobaric interferences, as opposed to P-TIMS and MC-ICP-MS. In addition, the sensitivity of N-TIMS is typically higher than that of P-TIMS for elements that form negative ions or oxyanions, such as Mo [17,18]. The method, however, requires correction for interferences from oxides containing the heavier oxygen isotopes (e.g., [17]). One recent advance in the application of N-TIMS is the in situ measurement of certain oxygen isobars, which provides the opportunity to monitor the variability in the oxygen isotopic composition during analysis [19–24,13,25]. This study incorporates the use of 10¹² Ω resistors to measure and correct for oxygen isobars.

The methods presented here are primarily applicable to the measurement of mass-independent nucleosynthetic effects, but may also be adapted to double-spike methods for the determination of mass-dependentstable isotope effects. To demonstrate the utility of these procedures, the Mo isotopic compositions of standards, a gravimetrically prepared spike-standard mixture, the iron meteorites Toluca and Skookum, and the chondrite Allegan are reported.

2. Sample preparation

Iron meteorite samples were cut into 0.3–2 g pieces using a water-cooled Leco “Vari-cut” saw with a 12.7 cm diamond wafering blade. All iron meteorite pieces were mechanically abraded with different pieces of carborundum paper for each sample, and sonicated in ethanol three times to remove adhering rust, fusion crust, or saw blade contamination. All acids used for digestion and ion exchange chromatography were sub-boiling distilled with a quartz still (HCl and HNO₃) or a Savillex Teflon^® still (HF) in-house. High-purity water used for dilutions and cleaning was produced using a Milli-Q water purification system (deionized to a resistivity of 18.2 MΩ·cm). Iron meteorite samples were digested using 30–40 ml of 8 M HCl in Savillex Teflon^® beakers which were heated at 130–140 °C on a hotplate for at least 48 h. The chondrite sample (Allegan) was powdered in an agate mortar and pestle, and digested in 30 ml of a concentrated HF:HNO₃ mixture in 5:1 proportions. The Allegan solution was evaporated to a sludge, 5 ml of concentrated HNO₃ was added to dilute the remaining HF, and then 5 ml of concentrated HCl was added and evaporated twice to dissolve fluorides that are insoluble in HNO₃. Chondrite and iron meteorite samples were dried to a sludge, and 3 ml of concentrated HNO₃:HCl in 2:1 proportions was added to the samples and heated at 100 °C overnight or 115 °C for at least one hour in an effort to drive Mo to its highest oxidation state (MoVI). This step was used in an attempt to mitigate loss of Mo during ion exchange due to the presence of multiple valences of Mo on the column. The acid phase was evaporated to a sludge, at which point, 6 M HCl was added and dried to sludge twice to remove traces of HNO₃. Then 1 M HF was added and dried to sludge twice, and finally the samples were taken into the 1 M HF loading solution. At this point, some Mo was likely reduced to a lower oxidation state, but the successive additions of HCl and HF were necessary to ensure that the loading solution was the correct molarity of HF, with minimal traces of HCl and HNO₃.

The separation and purification of Mo from natural samples utilized a two-stage anion exchange chromatography method adapted from Pietruszka et al. [26], Scheiderich et al. [27], and Nagai and Yokoyama [28] (Table 1). Eichrom AG 1 × 8 200–400 mesh resin was pre-cleaned using reagent grade 1 M HCl and 6 M HCl, followed by quartz-distilled 6 M HCl and 1 M HCl, added in succession with Milli-Q water between stages of cleaning. The primary column generally follows the same procedure as the primary column of Nagai and Yokoyama [28]. Disposable Biorad Poly-Prep columns with 2 ml resin bed volume capacity were loaded with 1.4 ml of resin, which was further cleaned using the procedures outlined in Table 1. Samples were loaded in 1 ml of 1 M HF for each column, with 1 column per ~100 mg of iron meteorite and ~300 mg of chondrite. Major elements were eluted in an additional 3 ml of 1 M HF; Zr, Ru, Ti, and Hf were eluted in 6 ml of 9 M HCl–0.05 M HF; W was eluted in 12 ml of 9 M HCl–1 M HF; and Mo was collected in 10 ml of 6 M HNO₃–3 M HF. Before the addition of 6 M HNO₃–3 M HF, 1.5 ml of Milli-Q H₂O was added to the column to prevent the formation of aqua regia in the column.

Table 1.

The anion exchange chromatographic procedures for the separation and purification of Mo. The secondary column was repeated one time to improve sample purity.

Acid	Volume per column (ml)	Step
Primary column - 1–8 Biorad columns filled with 1.4 ml AG 1-X8 200–400 mesh
6 M HC1	10	Cleaning
Milli-Q H2O	10	Cleaning
6M HNO3 - 3 M HF	10	Cleaning
Milli-Q H2O	5	Cleaning
1 M HF	3	Equilibration
1 M HF	1	Load sample
1 M HF	3	Rinse (Fe, matrix)
9 M HC1 – 0.05 M HF	3 + 3	Rinse (Ru, Zr, Ti, Hf)
9 M HC1 – 1 M HF	3 + 3 + 3 + 3	Rinse (W, Re)
Milli-Q H2O	1.5	Remove HO
6 M HNO3 −3 M HF	10	Collect Mo
Secondary column - 1 Teflon® column filled with ~ 0.3 ml AG 1-X8 200–400 mesh
0.01 M HC1	4	Cleaning
1 M HC1	8	Cleaning
6 M HC1	4	Cleaning
6 M HC1	0.5	Load sample
6 M HC1	4	Rinse remaining matrix
0.01 M HC1 – 0.1 M HF	1	Rinse remaining matrix
0.01 M HC1	3	Rinse remaining matrix
1 M HC1	12.5	Collect Mo

The secondary column was a small version of the anion column described by Pietruszka et al. (2006), and was used to further purify the Mo. The columns used for this study were made with Teflon^® heat-shrink tubing and were fitted with Teflon^® frits. The resin bed for these columns had dimensions of 0.4 × 2.4 cm. The secondary column was filled to the base of the reservoir with Eichrom AG 1 × 8 200–400 mesh resin (~ 300 μl). The resin bed was then cleaned using 0.01 M HCl, 1 M HCl, and 6 M HCl. Prior to loading, samples were dissolved in 2:1 HNO₃:HCl and heated overnight or for one hour, as was done prior to the primary column. Again, 6 M HCl was added and dried to remove traces of HNO₃ twice, and then the sample was taken up into 0.5 ml of 6 M HCl to be loaded onto the column. Four ml of 6 M HCl was added to the column to elute remaining impurities, such as Fe, Ni, and Co, followed by 1 ml of 0.01 M HCl-0.1 M HF to remove remaining Fe, followed by 3 ml of 0.01 M HCl. Molybdenum was then collected in 12.5 ml of 1 M HCl (Table 1). This column, including the oxidation stage with 2:1 HNO₃:HCl, was repeated once to improve sample purity. Molybdenum purity and yield was monitored qualitatively using a Thermo-Fisher Element 2 single-collector ICP-MS prior to loading onto filaments for analysis. Occasional above-background signals on masses corresponding to Na, Al, V, Mn, Zn, and Sn were observed; some of which may have been generated by gas species from the plasma. The largest and most variable signals were observed for ²³Na and ²⁷Al, and the ratios of these signals to the corresponding ⁹⁸Mo signals were typically 0.02–2 (²³Na/⁹⁸Mo) and 0.2–1 (²⁷Al/⁹⁸Mo). Other species typically had signal ratios, relative to ⁹⁸Mo, significantly less than ~0.2. The occurrences of these non-analyte species were not correlated with poor ionization of Mo.

Three total analytical blanks were determined for these procedures. Each blank used either one, three, or four primary columns, and resulted in blanks of 0.92, 1.3, and 2.8 ng of Mo, respectively. The total analytical blank using one primary column is, therefore, estimated to be ~0.9 ± 0.3 ng (2SD). These blanks were sufficiently low to be negligible for the quantities of Mo used for the analyses reported.

The total yield for the chemistry described above was ~50–65%. It is possible that the sub-optimal recovery of Mo resulted in mass-dependent fractionation of the Mo isotopes. However, the Mo isotopic compositions reported here were fractionation corrected for mass-dependent isotope effects that occurred naturally and/or that resulted from the column chemistry. Although Rizo et al.[29] showed that fractionation correction of mass-dependently fractionated materials can result in apparent mass-independent anomalies in W isotopes, the agreement of the Mo data reported here with those of previous studies, as discussed below, indicate that fractionation that occurred during the column chemistry was effectively corrected for.

Loss of Mo primarily occurred in the primary column in the loading stage where some Mo was eluted with the Fe. Because the yield for the secondary column was typically 80–100%, and matrix was largely absent at this stage, it is likely that the matrix was the complicating factor in the primary column. The breakthrough of Mo in the primary column was mitigated by loading in 1 M HF, as opposed to the 0.4 M HCl–0.5 M HF loading solution used by Nagai and Yokoyama [28]. The partition coefficient between the resin and the acid phase (K_d) for Mo is relatively low in 0.4 M HCl–0.5 M HF compared to 1 M HF (K_d = 45 in 0.4 M HCl–0.5 M HF vs. K_d = 573 in 1 M HF – Nagai and Yokoyama [28]. These authors chose not to use 1 M HF due to the high K_d values they measured for Zr and Hf, but we observed that Zr was efficiently separated from Mo in the primary and secondary columns. Hafnium was not monitored by ICP-MS, but due to its similar K_d values to Zr in the acids used, it is likely that it was also efficiently separated from Mo. Although Fe does not occupy anion exchange sites in 1 M HF, overloading the column may reduce the tendency of Mo to stick to the resin when Fe is present. The sample sizes loaded onto each primary column were small in order to minimize the effect of Fe on the behavior of Mo on the column. Alternatively, the presence of redox-sensitive Fe or other major elements may change the valence of Mo, resulting in the persistence of mixed valence states in the primary column. One way to mitigate this possible effect is to load the column using H₂O₂ or another oxidizing agent to better control the oxidation state of Mo (and Fe) (e.g., [30], but this is yet to be incorporated into methods described here.

After the completion of the chromatographic procedures, the purified sample was taken into solution in 2:1 concentrated HNO₃:HCl and dried to remove resin-derived organics. This step was repeated seven times for each sample. Following the last addition of 2:1 HNO₃:HCl, concentrated HCl was added and dried three times to remove traces of HNO₃. Samples were then dissolved in 6 M HCl to load onto filaments.

3. Mass spectrometry

3.1. Sample loading procedure

Samples and standards were loaded onto outgassed Re filaments. The 99.99% pure Re ribbon used for this study was 0.76 mm wide and 0.030 mm thick and was purchased from the H. Cross Company. Rhenium filaments were outgassed at 4.2 A for 40 min at least 24 h before loading. A double filament assembly was found to increase signal intensity and duration of the ionization of Mo, compared to single filaments, in agreement with Turnlund et al. [9] and Giussani et al. [12]. Further, we found that the double filament assembly provided better control of the ionization of Mo than was achieved with single filaments. The ionization of Mo on single filaments typically would reach a point where the growth of the signal and temperature became exponential without manual heating. Soon after, both the signal and temperature would reach a maximum and then the signal would decrease precipitously. The use of an ionization filament and evaporation filament prevented the exponential growth by providing a non-direct heat source, which dramatically improved signal duration and stability. Further, Mo isotopes ionized on single filaments were often fractionated more quickly and more intensely than Mo that was ionized using a double filament assembly. This phenomenon was reflected by the absolute range in the fractionation factor, β, in the exponential law which was used for fractionation correction, discussed below. Over the course of a given analysis, β had an absolute range up to ~0.5 for single filament analyses, but a range of <0.3 for most double filament analyses, which were also ~3.5 times longer in duration.

About 400–1000 ng Mo for samples and 1000 ng Mo for standards were loaded onto the evaporation filament in two ~0.5 μl aliquots, and dried by applying a current of 0.6 A. Before the first aliquot was dry, the second was added so that distinct reservoirs of Mo did not form. The avoidance of distinct reservoirs was found to improve the ionization of Mo on single filaments in the early stages of the development of these procedures. For samples, once the deposit was dried, the current was increased slowly from 0 to ~2 A until a faint red glow was observed or until the sample transitioned from cloudy white to a colorless or bronze-colored deposit with a black rim. This was done to remove any remaining organics not eliminated by successive additions of 2:1 HNO₃:HCl prior to loading. Care was taken not to heat the filament beyond a dull glow because the sample can burn off. Ten micrograms of 5 μg/μl La(NO₃)₃ in 1 M HNO₃ were then added as an electron emitter to reduce the work function of Re and promote the ionization of MoO₃⁻ (e.g., [17,6]). The La(NO₃)₃ activator was dried at 0.6 A, but because the solution does not dry completely, the current was then increased from 0 to ~2 A until the solution was observed to boil to dryness. Ten micrograms of La(NO₃)₃ was also added to the ionization filament, using the same drying procedure.

As this loading procedure was developed, variables such as the type of loading solution, the type and amount of activator, the filament assembly, and the placement of activator on the ionization and/or evaporation filaments were systematically varied in order to optimize the ionization of Mo to facilitate stable, long lasting signals. It was found that loading Mo in a chloride form in HCl led to better ionization than when Mo was loaded in HBr or in HNO₃. In agreement with Nagai and Yokoyama [13], the optimal activator was found to be La(NO₃)₃, although we also trialed Ba(OH)₂, Ca(NO₃)₂, and a mixture of La and Gd. Nagai and Yokoyama [13] also found that a ratio of 5:1 La:Mo was most effective. The results of the initial work with single filaments in this study are consistent with this conclusion, though the optimal ratio of La:Mo for double filaments appears to be higher. Initially, 1ug of Mo was loaded onto the evaporation filament and 5ug of La(NO₃)₃ was loaded only onto the ionization filament, but it was found that adding an additional 5ug of La(NO₃)₃ on top of the sample or standard improved ionization. However, when this was done, the problem of exponential growth and subsequent rapid decrease of the Mo signals that was observed with single filaments persisted. When the amount of La(NO₃)₃ was doubled on both the ionization and evaporation filaments, the temperature of the filaments and the ionization of Mo were better controlled, and, thus, the Mo signal was more stable and long-lived.

The loading blank using this loading procedure was ~1 ng. It was suspected that the Mo was primarily sourced from the Re ribbon, as Mo was not observed in the La(NO₃)₃ activator when it was examined by ICP-MS. Because Mo was not eliminated by outgassing, and Mo and Re are chemically similar, Mo may be bound in the structure of the Re ribbon. The effect of the loading blank on the measured isotope ratios was examined using different size loads of the gravimetrically prepared spike-standard mixture (400–1000 ng), and was observed to be negligible.

3.2. Instrumental setup and measurement routine

A Thermo-Fisher Triton Plus solid source mass spectrometer was used to measure the Mo isotopic compositions of samples and standards. The instrument was equipped with nine Faraday cup detectors, seven of which were used to measure Mo isotopes. A gain calibration of the Faraday cup detectors was performed each day. Liquid nitrogen was added to a cryopump each day to maintain a stable working vacuum. Using a variable leak valve, high-purity oxygen was bled into the source can in an attempt to promote the formation and ionization of MoO₃⁻. The O₂ partial pressure was ~ 1.73 × 10⁻⁷ mbar above the baseline pressure with liquid nitrogen, such that the total pressure in the source can was typically 2.0–2.4 × 10⁻⁷ mbar. Using single filaments, standards were also analyzed with no O₂ bled into the source, and with total pressures up to 4 × 10⁻⁷ mbar. With double filaments, total pressures up to 7.4 × 10⁻⁷ mbar were tested. This testing did not definitively show that Mo ionization correlates with the amount of O₂ that is used. For their single filament method, Nagai and Yokoyama [13] observed that the best results were obtained without bleeding O₂ into the source.

The heating protocol for the filaments was consistent for all standards and samples. Because of the close proximity of the two filaments, the ionization and evaporation filaments were initially heated slowly to avoid warping the Re ribbon. The ionization filament was heated first at 50 mA/min to 1000 mA, then at 20 mA/min to 1700 mA, and then at 10 mA/min to 1930 mA, in order to gently warm the evaporation filament. When the ionization filament reached 1700 mA, the evaporation filament was heated at 30 mA/min to 500 mA, then at 20 mA/min. Both filaments were continuously heated, with the rate gradually being decreased during the heating procedure until the ionization and evaporation filaments were heated at 0.1 mA/min and 0.1–0.2 mA/min, respectively. Once the ionization filament reached ~1930 mA and the evaporation filament reached ~630 mA, a mV-level signal was typically registered on the Faraday cups. During the warm-up of samples and standards, the intensity of the signal usually went through periods during which it decreased substantially until the evaporation filament reached 1200–1500 mA. The peak shapes during this period were also usually asymmetrical, likely due to overlapping peaks from ribbon-derived Mo that was not equilibrated with the sample or standard. After the evaporation filament heating current surpassed 1200–1500 mA and the intensity of the signal on ⁹⁸Mo¹⁶O₃⁻ (m/z ~146) reached 300–400 mV, the peak shape typically became symmetrical for samples and standards. Once the signal intensity reached 800–1000 mV on ⁹⁸Mo¹⁶O₃⁻, the analyses were started. This signal intensity typically corresponded to a current of 1650–1800 mA and 2080–2140 mA applied to the evaporation filament and ionization filament, respectively. At the beginning of analyses for samples, the signal often decreased for the duration of ~50–100 ratios, but would then stabilize or begin to grow. During a single analysis, lasting ~13 h, the typical signal intensity for ⁹⁸Mo¹⁶O₃⁻ ranged between 1000 mV and 1600 mV.

Molybdenum was measured as ^xMoO₃⁻ in a static single-line measurement routine. Seven Faraday cup detectors, which were electronically connected to amplifiers equipped with 10¹¹ Ω resistors, were used to collect the ^xMoO₃⁻ beams, with ⁹⁶Mo¹⁶O₃⁻ in the center cup (m/z ~ 144–Table 2). An additional Faraday cup (H4) was electronically connected to an amplifier equipped with a 10¹² Ω resistor. This detector was used to collect the ¹⁰⁰Mo¹⁸O¹⁶O₂ beam, which was used for the in-situ oxide correction. Later in the development of the analytical method, the L3 cup and a 10¹² Ω resistor were used to collect the ⁹²Mo¹⁷O¹⁶O₂ beam, as discussed below. The amplifiers equipped with 10¹¹ Ω resistors were electronically rotated to remove amplifier bias using the virtual amplifier capability of the Triton.

Table 2.

Faraday cup configuration for Mo analyses. The resistors equipped to the amplifiers for each cup and the trioxide interferences are also shown.

	L4	L3c	L2	L1	C	H1	H2	H3	H4
Resistor (Ω)a	1011	1012	1011	1011	1011	1011	1011	1011	1012
Species	92Mo16O3	92Mo17O16O2	94Mo16O3	95Mo16O3	96Mo16O3	97Mo16O3	98Mo16O3	100Mo16O3	100Mo18O16O2
Interfering Trioxideb			92Mo50O3	92Mo51O3	92Mo52O3	94Mo51O3	94Mo52O3	96Mo52O3	98Mo52O3
				94Mo49O3	94Mo5003	95Mo50O3	95Mo51O3	97Mo51O3
					95Mo49O3	96Mo49O3	96Mo50O3	98Mo50O3
							97Mo49O3

^aOnly amplifiers equipped with 10¹¹ Ω resistors were rotated during analyses to remove amplifier bias.

^bTrioxide masses are the total mass of the O₃. For instance, ⁹²Mo⁵⁰O₃ is composed of both ⁹²Mo¹⁸O¹⁶O₂ and ⁹²Mo¹⁷O₂¹⁶O.

^cL3 was used to measure ⁹²Mo¹⁷O¹⁶O₂ from standard 35 onward, but was not used to calculate ¹⁷O/¹⁶O due to the lower internal precision, relative to the ¹⁸O/¹⁶O measurement.

Although Zr tends to form positive ions (e.g., [31]), and Ru, which forms negative trioxides, does not ionize on Re filaments at the running temperature of Mo [24], potential minor isobaric interferences from Zr and Ru were monitored by scanning the low and high masses surrounding the mass range of ^xMoO₃⁻ (140–148) using the electron multiplier. No interferences were observed corresponding to Zr or Ru. Further, no unaccounted-for signals (those not associated with MoO₂⁻ and MoO₄⁻) were observed to be greater than ~ 10–200 cps, excluding a ~1000 cps peak at mass 155 which we attribute to LaO⁻. Therefore, no corrections for isobaric interferences from species in the MoO₃ mass range were applied, excluding oxide corrections.

The procedures for incorporating the use of 10¹² resistors into the measurement routine were adapted from Koornneef et al. [32], Liu and Pearson [33], and Trinquier et al. [25]. Baseline measurement times of 1200 s, with a 30 s settling time, and idle times of 15 s were used. Due to the longer settling time of the 10¹² Ω resistor, compared to 10¹¹ Ω resistors, and the lower signal intensity of the ¹⁰⁰Mo¹⁸O¹⁶O₂ species (2–4 mV), the optimal time over which the isotope ratio measurements were integrated was 67.109 s. Much of the data for the standards and samples were collected using two 67.109 s integrations such that the isotope ratio measurements were integrated over a total of 134.218 s. It was subsequently determined, however, that this integration time was unnecessarily long, so it was halved and the number or ratios was doubled from standard 35 onward (Table 3). No difference in the measured ratios was evident with the long vs. short integration times, but the internal statistics improved slightly with the shorter integration times. Measurements were taken in 50 blocks with 10 cycles/block for a total of 500 ratios. Baselines were measured every 10 blocks and the pilot mass ⁹⁶Mo¹⁶O₃ was centered and automatically focused every 2 blocks.

Table 3.

Molybdenum isotopic compositions of repeated analyses of the Alfa Aesar Specpure^® Mo plasma standard using the in situ line by line oxide correction and fractionation correction using the molecular trioxide mass and ⁹⁸Mo/⁹⁶Mo. Measured, not fractionation corrected, ¹⁸O/¹⁶O ratios, averaged over the course of each analysis, are also shown. The 2SE of each analysis is given in ppm.

Standard #	92Mo/96Mo	2SE	94Mo/96Mo	2SE	95Mo/96Mo	2SE	97Mo/96Mo	2SE	100Mo/96Mo	2SE	18O/16O
1	0.883168	9.5	0.552441	5.9	0.953191	4.5	0.573952	4.2	0.581460	6.6	0.002067
2	0.883225	9.3	0.552453	6.8	0.953205	3.9	0.573950	3.4	0.581472	5.8	0.002072
3	0.883106	6.6	0.552430	4.4	0.953188	2.9	0.573955	2.2	0.581446	3.7	0.002063
4	0.883189	20	0.552443	8.1	0.953197	5.0	0.573953	3.6	0.581465	8.4	0.002074
5	0.883174	8.2	0.552441	4.7	0.953198	3.9	0.573954	3.0	0.581452	4.5	0.002060
6	0.883169	6.3	0.552441	3.9	0.953193	3.0	0.573950	2.3	0.581459	3.5	0.002068
7	0.883211	6.8	0.552450	3.7	0.953206	3.3	0.573951	2.7	0.581468	4.6	0.002066
8	0.883176	11	0.552443	5.0	0.953196	3.3	0.573953	2.6	0.581457	4.8	0.002060
9	0.883189	18	0.552447	8.5	0.953201	5.5	0.573952	3.9	0.581462	8.1	0.002055
10	0.883195	5.6	0.552449	3.3	0.953203	2.8	0.573954	2.2	0.581460	3.4	0.002054
11	0.883135	8.9	0.552437	4.5	0.953195	2.9	0.573953	2.1	0.581450	3.9	0.002058
12	0.883148	8.7	0.552438	5.3	0.953192	3.7	0.573953	2.6	0.581456	4.6	0.002084
13	0.883129	12	0.552434	7.8	0.953185	4.9	0.573953	3.7	0.581455	6.8	0.002089
14	0.883113	7.6	0.552432	5.3	0.953188	3.7	0.573953	2.9	0.581451	4.8	0.002071
15	0.883262	8.9	0.552461	5.4	0.953206	3.9	0.573949	3.5	0.581481	5.0	0.002044
16	0.883197	7.6	0.552448	5.1	0.953198	3.6	0.573952	5.4	0.581473	5.0	0.002062
17	0.883164	10	0.552440	6.5	0.953191	3.9	0.573953	3.8	0.581468	5.8	0.002078
18	0.883180	6.7	0.552443	4.8	0.953190	3.2	0.573952	2.7	0.581469	4.1	0.002064
19	0.883238	9.9	0.552455	5.5	0.953203	4.1	0.573952	2.8	0.581478	5.8	0.002069
20	0.883211	8.7	0.552449	6.4	0.953192	4.7	0.573948	3.3	0.581479	5.9	0.002052
21	0.883186	12	0.552438	7.6	0.953191	6.3	0.573951	5.0	0.581476	7.5	0.002056
22	0.883189	6.9	0.552443	4.8	0.953194	3.6	0.573949	3.1	0.581471	4.3	0.002072
23	0.883194	9.5	0.552441	6.4	0.953191	4.1	0.573950	3.4	0.581476	5.5	0.002053
24	0.883135	7.6	0.552432	5.1	0.953186	3.5	0.573954	3.1	0.581465	5.1	0.002055
25	0.883182	12	0.552444	9.4	0.953186	5.8	0.573952	5.5	0.581479	9.5	0.002058
26	0.883154	5.0	0.552435	3.9	0.953186	2.5	0.573952	2.4	0.581471	3.2	0.002062
27	0.883189	7.5	0.552446	4.1	0.953197	4.1	0.573954	3.7	0.581475	5.2	0.002043
28	0.883138	9.8	0.552434	5.4	0.953188	4.2	0.573952	3.0	0.581465	5.1	0.002056
29	0.883140	6.4	0.552436	3.6	0.953187	3.8	0.573953	2.9	0.581465	4.5	0.002057
30	0.883108	7.0	0.552427	4.1	0.953177	3.1	0.573952	2.4	0.581459	4.2	0.002067
31	0.883093	8.0	0.552424	4.4	0.953175	2.9	0.573954	2.5	0.581460	4.3	0.002064
32	0.883149	8.3	0.552438	5.7	0.953185	4.3	0.573953	3.1	0.581468	5.1	0.002047
33	0.883218	9.9	0.552450	7.4	0.953200	4.5	0.573951	3.9	0.581473	6.4	0.002041
34	0.883100	6.7	0.552428	4.9	0.953176	3.3	0.573954	2.5	0.581461	4.2	0.002059
35	0.883112	5.0	0.552430	3.4	0.953182	2.4	0.573953	1.9	0.581454	3.1	0.002055
36	0.883074	4.7	0.552420	3.3	0.953172	2.4	0.573954	1.8	0.581454	3.1	0.002055
37	0.883200	5.4	0.552444	3.8	0.953189	2.8	0.573951	2.2	0.581475	3.3	0.002056
38	0.883158	5.5	0.552437	3.6	0.953183	2.5	0.573953	2.0	0.581469	3.1	0.002059
39	0.883170	4.9	0.552439	3.8	0.953184	2.5	0.573953	2.0	0.581468	3.4	0.002054
40	0.883155	5.1	0.552437	3.3	0.953184	2.3	0.573952	1.8	0.581465	3.1	0.002053
41	0.883066	5.1	0.552418	3.5	0.953169	2.5	0.573953	2.0	0.581453	3.2	0.002061
42	0.883089	5.8	0.552422	3.8	0.953167	2.5	0.573952	1.8	0.581462	3.3	0.002058
43	0.883090	5.9	0.552424	4.0	0.953170	2.8	0.573951	2.5	0.581452	3.7	0.002056
44	0.883102	5.6	0.552424	4.0	0.953170	2.9	0.573951	2.2	0.581458	3.8	0.002060
45	0.883055	6.6	0.552413	4.6	0.953163	3.5	0.573951	2.9	0.581451	4.4	0.002059
46	0.883172	7.4	0.552437	5.6	0.953181	3.6	0.573950	2.7	0.581475	4.9	0.002049
47	0.883194	5.3	0.552442	3.6	0.953186	2.8	0.573950	2.2	0.581480	3.2	0.002042
48	0.883138	7.1	0.552432	4.6	0.953174	3.2	0.573950	2.3	0.581469	4.0	0.002055
Average	0.883157		0.552438		0.953188		0.573952		0.581465		0.002060
2SD (ppm)	107		37		23		5.4		32		0.96%

3.3. Fractionation correction

The isotope ratios were oxide corrected, as described below, and then fractionation corrected for natural and instrumental mass-dependent fractionation using the exponential law [34]. Instrumental mass-dependent fractionation in thermal ionization is largely due to the preferential evaporation of the lighter isotopes. Traditionally, the normalizing ratio of ⁹⁸Mo/⁹⁶Mo has been used [4,7], where ⁹⁸Mo/⁹⁶Mo = ≡1.453171 [35]. The exponential law is commonly used to correct a given isotope ratio ($R_y^x$) by normalizing to a second ratio ($R_y^x$) of two stable and non-radiogenic isotopes [34], where:

$$R_{wy}^{xy}={\left(R_y^x\times M_y^x\right)}^{-\beta }$$

and $R_{wy}^{xy}$ is the fractionation corrected ratio, $R_y^x$ is the oxide-corrected, measured ratio, $M_y^x$ is the isotope mass ratio of the isotopes being fractionation corrected, and β is the fractionation factor:

$$\beta =\frac{ln\left(R_y^w/R_{yAccepted}^w\right)}{ln\left(M_y^w\right)}$$

where $R_y^w$ is the measured, oxide-corrected ratio, $R_{yAccepted}^w$ is the accepted value (⁹⁸Mo/⁹⁶Mo = 1.453171), and $M_y^w$ is the mass ratio of the isotopes used for normalization. Because Mo is measured as a trioxide, the molecular mass ratios were used in addition to the isotope mass ratios (e.g., 140/144 vs. 92/96). These results are compared below (Table 4). The molecular mass was required to fractionation correct the ¹⁰⁰Mo¹⁸O¹⁶O₂ species because it has no corresponding isotope mass, as ¹⁰⁰Mo is the heaviest Mo isotope.

Table 4.

Comparison of the various data reduction methods. The best external precision (2SD) is obtained for the ⁹⁷Mo/⁹⁶Mo ratio using the line by line oxide correction and the molecular mass fractionation correction.

Fractionation correction	Oxygen correction	External Precision (ppm)
		92Mo/96Mo	94Mo/96Mo	95Mo/96Mo	97Mo/96Mo	100Mo/96Mo
Isotope mass	Nier	175	73	45	67	90
	Measured single value	169	50	34	16	42
	Line by line	166	56	30	6.7	45
	Line by line (FC)	158	76	21	20	76
Molecular mass	Nier	117	63	40	66	87
	Measured single value	110	31	28	15	35
	Line by line	107	37	23	5.4	32
	Line by line (FC)	100	54	17	23	55

3.4. Oxide corrections

Because Mo was measured as a trioxide, the intensity measured for each ^xMoO₃⁻, except the lowest mass ⁹²Mo¹⁶O₃⁻, must be corrected for isobaric interferences from the heavier oxygen isotope trioxides of lower mass Mo isotopes (Table 2). For example, ⁹⁴Mo¹⁶O₃⁻ at mass 142 must be corrected for the interference caused by ⁹²Mo¹⁸O¹⁶O₂⁻ at mass 142. Other oxide species which generate significant interferences are Mo¹⁷O ¹⁶O₂⁻, Mo¹⁷O₂ ¹⁶O⁻, Mo¹⁸O₂ ¹⁶O ⁻, Mo¹⁸O ¹⁷O ¹⁶O⁻, Mo¹⁷O₃⁻, and Mo¹⁷O₂¹⁸O⁻. Additional species (Mo¹⁷O¹⁸O₂⁻ and Mo¹⁸O₃⁻) were also formed, but in such low abundance that they do not generate significant interferences [36]. The relative abundances of the oxide species were calculated as in Creaser et al. [37] and Harper and Jacobsen [38] using the isotopic composition of oxygen.

Oxide corrections commonly have been applied using a single oxygen isotopic composition, such as that from Nier [39] (¹⁷O/¹⁶O = 0.0003749 and ¹⁸O/¹⁶O = 0.0020439), for every sample or standard. However, systematic variations in the isotope ratios from standard to standard are likely due to variability in the oxygen isotopic composition, as has been the case for other measurements of Mo by N-TIMS [13] and for measurements of Os, B, W, and Ru by N-TIMS (e.g., [22,26,1,37,7,39]). Therefore, 10¹² Ω resistors were installed in order to measure ¹⁰⁰Mo¹⁸O¹⁶O₂, and later ⁹²Mo¹⁷O¹⁶O₂, from which the oxygen isotope composition can be calculated. The 10¹² Ω resistor reduces the signal to noise ratio and is most advantageous for signals between ~5 and 20 mV (e.g., [32]), which is on the order of the typical signal intensity for ¹⁰⁰Mo¹⁸O¹⁶O₂ (2–4 mV). The ratio of the intensities of ¹⁰⁰Mo¹⁸O¹⁶O₂/¹⁰⁰Mo¹⁶O₃ divided by 3 gives the ¹⁸O/¹⁶O ratio. By dividing by 3, the different permutations that ¹⁸O can be configured in a trioxide are accounted for [36].

The methods for measuring oxygen in situ were adapted from Liu et al. [19] and Luguet et al. [20]. In agreement with Liu et al. [19] and Luguet et al. [20], the ¹⁸O/¹⁶O and ¹⁷O/¹⁶O ratios were observed to increase over the course of a run, and the average ¹⁸O¹⁶O and ¹⁷O/¹⁶O ratio was found to vary from standard to standard (Table 3) and sample to sample. Because of the changing oxygen isotope ratio within analyses, the oxide correction was done for each cycle (referred to as line by line oxide correction). Data were also corrected using the oxygen isotopic composition of Nier [39] (Nier oxide correction), the measured oxygen isotopic composition averaged over the entire run (measured single value oxide correction), and the line by line oxygen correction with “fractionation corrected” ¹⁰⁰Mo¹⁸O¹⁶O₂ (line by line (FC) oxide correction).

For the measured single value, line by line, and line by line (FC) oxide corrections, the data reduction required two iterations [20]. The first was to correct the measured ¹⁰⁰Mo¹⁸O¹⁶O₂⁻ intensity for the interference from ⁹⁸Mo¹⁸O ¹⁶₂O⁻ using the oxygen isotopic composition of Nier [39], termed the first-order oxide correction. The second iteration used the average oxide-corrected ¹⁰⁰Mo¹⁸O¹⁶O₂ intensity for the measured single value oxide correction, or the oxide-corrected ¹⁰⁰Mo¹⁸O¹⁶O₂ intensities for each line for the line by line oxide correction to calculate the ¹⁸O/¹⁶O ratio(s). For the line by line (FC) oxide correction, the oxide-corrected ¹⁰⁰Mo¹⁸O¹⁶O₂ was fractionation corrected as described above.

The ¹⁷O/¹⁶O was calculated from the ¹⁸O/¹⁶O ratio using the slope of the terrestrial fractionation line (TFL). The ¹⁷O/¹⁶O was also measured in situ from standard 35 onward. However, the signal intensity for the ⁹²Mo¹⁷O¹⁶O₂ species was at the limit of the effective range of the 10¹² Ωresistor (<2 mV), and the internal precision of these measurements was not as precise as the measurement of ¹⁸O/¹⁶O. A 10¹³ Ω resistor, with a lower effective range (~ 0.3–20 mV – [41]), may be advantageously employed to measure this species, but this has not yet been tested. Only negligible differences in the external reproducibility of the 14 standards were observed when the oxide correction was done using the calculated vs. measured ¹⁷O/¹⁶O, although the absolute values of the Mo isotope ratios were different (Table 5). To be consistent and to characterize the long-term external reproducibility, only Mo isotope data that was oxide corrected using the calculated ¹⁷O/¹⁶O is reported here.

Table 5.

Molybdenum isotopic compositions of standards 35–48, where the oxide corrections were done using the measured ¹⁷O/¹⁶O, as opposed to the ¹⁷O/¹⁶O calculated using the TFL. Also shown are the measured, non-fractionation corrected ¹⁷O/¹⁶O and ¹⁸O/¹⁶O, averaged over the course of the analysis and the average results of standards 35–48 calculated using the TFL-calculated ¹⁷O/¹⁶O..

Standard #	92Mo/96Mo	2SE	94Mo/96Mo	2SE	95Mo/96Mo	2SE	97Mo/96Mo	2SE	100Mo/96Mo	2SE	17O/16O	18O/16O
Measured 17O/16O
35	0.883175	5.1	0.552459	3.4	0.953203	2.4	0.573932	1.9	0.581451	3.1	0.0003872	0.002055
36	0.883136	4.7	0.552448	3.3	0.953193	2.4	0.573934	1.8	0.581451	3.1	0.0003871	0.002055
37	0.883262	5.4	0.552472	3.8	0.953210	2.8	0.573930	2.2	0.581472	3.3	0.0003870	0.002056
38	0.883219	5.6	0.552465	3.7	0.953203	2.5	0.573932	2.0	0.581466	3.1	0.0003872	0.002059
39	0.883231	4.9	0.552467	3.8	0.953205	2.5	0.573933	2.0	0.581465	3.4	0.0003869	0.002054
40	0.883218	5.1	0.552466	3.3	0.953205	2.3	0.573932	1.8	0.581462	3.1	0.0003869	0.002053
41	0.883128	5.1	0.552446	3.5	0.953190	2.5	0.573933	2.0	0.581450	3.2	0.0003876	0.002061
42	0.883152	5.9	0.552450	3.8	0.953188	2.5	0.573931	1.8	0.581459	3.3	0.0003875	0.002058
43	0.883151	5.9	0.552452	4.0	0.953191	2.8	0.573930	2.5	0.581449	3.7	0.0003872	0.002056
44	0.883163	5.6	0.552452	4.0	0.953190	2.8	0.573930	2.3	0.581455	3.8	0.0003875	0.002060
45	0.883117	6.6	0.552441	4.6	0.953184	3.4	0.573930	2.8	0.581448	4.4	0.0003875	0.002059
46	0.883233	7.5	0.552465	5.6	0.953201	3.7	0.573930	2.7	0.581471	4.9	0.0003865	0.002049
47	0.883256	5.3	0.552470	3.6	0.953207	2.8	0.573929	2.3	0.581476	3.2	0.0003858	0.002042
48	0.883201	7.2	0.552461	4.6	0.953195	3.2	0.573929	2.3	0.581466	4.0	0.0003874	0.002055
Average	0.8831S9		0.552458		0.953198		0.573931		0.581460		0.0003871	0.002055
2SD (ppm)	110		36		17		5.2		33		0.25%	0.49%
TFL calculated17O/16O
Average	0.883127		0.552430		0.953177		0.573952		0.581463
2SD (ppm)	111		36		18		5.2		33

4. Results and discussion

The standard data for all Mo isotopes using the preferred data reduction method is shown in Table 3 and Fig. 1. The external reproducibilities of the standards for the ⁹⁷Mo/⁹⁶Mo measurement, which is the Mo isotope ratio that is measured to the best statistical precision, using the various oxide and fractionation correction methods are compared in Table 4. The magnitudes of the oxide interferences for each Mo isotope are presented in Table 6. The results for spike-standard mixtures and samples are reported in μ notation (Tables 7 and 8), which gives the deviation of ^xMo/⁹⁶Mo of a sample relative to the average ^xMo/⁹⁶Mo of the repeated analyses of the Alfa Aesar Mo plasma standard over a given period of time, in parts per million. For example, for μ⁹⁷Mo:

$${\mu }^{97}M{o}_{sample}=\left[\frac{{\left(\frac{{}^{97}Mo}{{}^{96}Mo}\right)}_{sample}}{{\left(\frac{{}^{97}Mo}{{}^{96}Mo}\right)}_{standard}}-1\right]\times {10}^6$$

Fig. 1.

The μ⁹⁷Mo for repeated analyses of the Alfa Aesar Mo plasma standard and the gravimetrically prepared spike-standard mixture. The collection period was 13 months. Data were reduced using in situ ¹⁸O/¹⁶O measurements and the line by line oxide correction. Data were fractionation corrected with ⁹⁸Mo/⁹⁶Mo = 1.453171 [35] and using the molecular mass instead of the isotope mass as described in text. Dotted lines represent two standard deviations of the analyses (2SD). For standards, 2SD = ± 5.4 ppm and for spike standard mixes, 2SD = ±2.9 ppm. The spike-standard mixture was gravimetrically predicted to have a μ⁹⁷Mo value of + 16.7 ppm. This compares favorably with the measured ratio of μ⁹⁷Mo = + 15.4 ± 2.9. Error bars for the individual analyses are the internal standard error of the mean (2SE).

Table 6.

The magnitude of the oxide interference corrections for each Mo isotope, given as percent relative to the oxide-corrected signal intensity of the Mo trioxide. Also shown are the percentages of the most abundant ¹⁷O- and ¹⁸O-bearing species relative to the total oxide interference signal. Other species incorporating ¹⁷O and ¹⁸O are minor and not taken into account for these calculations, such that the totals are generally not 100%.

	94	95	96	97	98	100
Total oxide signal (% relative to oxide-corrected signal)	0.99	0.066	0.45	1.2	0.47	1.5
Total 17O16O2 signal (%)	-	99.3	23.9	16.1	9.5	-
Total 18O16O2 signal (%)	100.0	-	75.9	83.9	90.4	99.8
Total	100.0	99.3	99.7	100.0	99.9	99.8

Table 7.

μ⁹⁷Mo values for the spike-standard mixes. Mixes 6 and 7 were analyzed with only 400 ng of Mo, whereas mixes 1–5 were analyzed with 1000 ng of Mo.

	μ97Mo	2SE
16.7 ppm mix 1	16.4	2.0
16.7 ppm mix 2	13.8	2.3
16.7 ppmmix3	16.5	1.9
16.7 ppm mix 4	15.4	1.8
16.7 ppm mix 5	17.5	2.2
16.7 ppm mix 6	13.9	3.1
16.7 ppm mix 7	14.3	2.0
Average	15.4	2.9 (2SD)

Table 8.

Molybdenum isotopic compositions of natural samples (in ppm). Different pieces of the same meteorite are numbered and different analyses of the same solution from the same piece of meteorite are numbered and lettered.

Samplesa	μ92Mo	2SD	μ94Mo	2SD	μ95Mo	2SD	μ97Mo	2SD	μ100Mo	2SD
Toluca 1	−15	111	2.6	36	−9.7	18	2.0	5	3.2	33
Toluca 2	12	90	9.1	29	−5.9	15	1.3	6	0.8	24
Toluca 3	30	92	19	29	−3.9	14	0.02	5	0.4	34
Toluca mean	9	45	10	17	−6.5	6	1.1	2	1.46	3.0
Skookum la	302	92	173	29	116	14	51	5	119	34
Skookum lb	278	92	170	29	114	14	55	5	107	34
Skookum 2a	267	92	164	29	109	14	54	5	109	34
Skookum 2b	146	90	123	29	97	15	57	6	63	24
Skookum mean	248	139	157	46	109	17	54	5	100	50
Allegan	43	111	52	36	22	18	19	5.2	21	33

^aToluca 1 and Allegan were processed using the procedures described here. Toluca 2 and Skookum 2 were digested and processed using the same methods, with the exception that these pieces were loaded onto the primary column using 0.4 M HCl-0.5 M HF as the loading solution. Toluca 3 was obtained as a byproduct of the chemistry used to separate and purify W [40], and was purified using the secondary anion column described here. Skookum 1 was processed using a different primary anion column that was more similar to the secondary column [26].

The best external reproducibility (2SD) for all oxide (except Nier, as described below) and fractionation correction methods (molecular vs. isotope mass) was obtained for ⁹⁷Mo because it is bracketed by the normalizing ratio isotopes.

4.1. Oxide corrections comparison and molecular vs. isotopic fractionation correction

The average ¹⁸O/¹⁶O measured for standards was0.002060 ± 0.000020 (2SD, n = 48), which is within uncertainty of the ratio reported by Nagai and Yokoyama [13] (0.002064 ± 0.000005, n = 21). The long-term external reproducibility of ⁹⁷Mo/⁹⁶Mo using the Nier, measured single value, line by line, and line by line (FC) oxide corrections and the molecular mass fractionation correction were ± 66 ppm, 15 ppm, 5.4 ppm, and 23 ppm, respectively (Table 4). For the Nier oxide correction, ⁹⁵Mo had the best precision of the Mo isotopes (40 ppm). This effect is likely because the ⁹⁵MoO₃⁻ beam has the smallest oxide interference compared to the other Mo isotope beams, especially ⁹⁷MoO₃⁻ and ¹⁰⁰MoO₃⁻ (Table 6). Therefore, the difference between the Nier [39] and measured oxygen isotopic compositions was magnified with the Nier oxide correction on ⁹⁷MoO₃⁻, compared to ⁹⁵MoO₃⁻. All oxide corrections using the in situ measured oxygen isotopic composition resulted in improved precision relative to the Nier oxide correction, demonstrating that the variability of the oxygen isotopic composition between analyses adversely affects precision. Further, better precision for ⁹⁷Mo/⁹⁶Mo resulted from using the line by line correction vs. the measured single value correction, which demonstrates that the variability of the oxygen isotopic composition within a single analysis also affects external precision. The line by line (FC) correction exhibits worse external reproducibility for most Mo isotopes, including ⁹⁷Mo/⁹⁶Mo, relative to the non-fractionation corrected line by line method. This is likely because the fractionation correction using the ¹⁰⁰Mo¹⁸O¹⁶O₂ species does not sufficiently account for oxygen isotope fractionation that originates in the tank from which O₂ is bled into the source can, or during formation of the trioxide [20]. Therefore, the preferred oxide correction method is the line by line correction.

The fractionation correction of Mo using the isotope mass, as opposed to the molecular mass, gives a long-term external reproducibility of 6.7 ppm for ⁹⁷Mo/⁹⁶Mo. The reason for the difference between the two methods is unclear; however, the discrepancy may indicate that Mo forms the molecular trioxide before evaporation on the filament fractionates Mo. Thus, using the molecular mass for fractionation correction provides a more accurate correction.

The best external reproducibility for ⁹⁷Mo/⁹⁶Mo was obtained using the line by line oxide correction and the fractionation correction using the molecular mass. Using these correction methods, the long-term external reproducibility (2SD) over 13 months of periodic measurement campaigns (Fig. 1 and Table 3, n = 48) was ±107 ppm for ⁹²Mo, 37 ppm for ⁹⁴Mo, 23 ppm for ⁹⁵Mo, 5.4 ppm for ⁹⁷Mo, and 32 ppm for ¹⁰⁰Mo. The short-term external reproducibility for one ~2 week period (n = 8) was ± 112, 37, 17, 3.7, and 28 ppm for ⁹²Mo, ⁹⁴Mo, ⁹⁵Mo, ⁹⁷Mo, and ¹⁰⁰Mo, respectively. The internal precision (2SE) was typically <4 ppm for ⁹⁷Mo. The findings that a line by line oxide correction and fractionation correction using the molecular mass produce the most precise measurements of the Mo isotope ratios is in agreement with Nagai and Yokoyama [13], who used a single filament assembly and measured both ¹⁷O/¹⁶O and ¹⁸O/¹⁶O from the ¹⁰⁰Mo¹⁷O¹⁶O₂ and ¹⁰⁰Mo¹⁸O¹⁶O₂ species using 10¹¹ Ω resistors. However, the external reproducibility of ⁹⁷Mo/⁹⁶Mo reported in this study is improved by a factor of ~2.4 relative to Nagai and Yokoyama [13], the precision of ¹⁰⁰Mo/⁹⁶Mo is roughly equivalent to slightly better on the short term, and the other isotope ratios are less precise (by a factor of ~2.3).

The methods reported here produce more or equivalently precise ratios for the heavy Mo isotopes and less precise ratios for the lighter Mo isotopes, relative to Nagai and Yokoyama [13]. The reason for this is unknown, but may be related to the differences in fractionation of Mo on single vs. double filaments and/or to the measurement vs. calculation of the oxygen isotope ratios. Nagai and Yokoyama [13] reported an absolute range in the fractionation factor, β, of ~0.6, similar to the range we obtained using single filaments. As noted above, the range in β over the course of an analysis is ~0.3 for double filaments. Therefore, if fractionation of Mo isotopes were the dominant cause for the difference in precision between this study and Nagai and Yokoyama [13], one might expect that more strongly fractionated measurements would have poorer precision on the light Mo isotopes. This is because the light Mo isotopes are evaporated more readily than the heavier isotopes and are farther in atomic mass units from the normalizing ratio, which could cause the fractionation correction to be inadequate. However, this is not observed, likely because Mo fractionation is coupled with the effects of the fractionating oxygen isotope composition, since the trioxides evidently form before Mo is fractionated by evaporation.

The magnitude of the oxide correction is largest on ⁹⁷Mo and ¹⁰⁰Mo, and the contribution of ¹⁸O to the total oxide interference signal becomes larger relative to ¹⁷O from ⁹⁵Mo to ¹⁰⁰Mo (Table 6). Because this study utilized a 10¹² Ω resistor to measure the ¹⁰⁰Mo¹⁸O¹⁶O₂ species, the variance of the ¹⁸O/¹⁶O ratio may be better accounted for relative to measurements using 10¹¹ Ω resistors [13]. Therefore, the oxide correction applied here for ⁹⁷Mo and ¹⁰⁰Mo, which is larger in magnitude than for other isotopes, may be more accurate, resulting in higher precision for these isotope ratios.

It is possible that some of the low mass Mo isotopes (i.e., ⁹⁵Mo and ⁹⁶Mo – Table 6) are less accurately corrected, and therefore less precise, because the contribution of species incorporating ¹⁷O is larger on these isotopes than on the heavier Mo isotopes. Because ⁹⁶Mo is affected by this, and because the data are normalized to ⁹⁶Mo, the effect of an inaccurate correction on ⁹⁶Mo may be magnified on ⁹⁴Mo and ⁹²Mo, which are far from the normalizing ratio in terms of atomic mass units. The low mass Mo isotopes are measured less precisely in this study when both the measured ¹⁷O/¹⁶O and the calculated ¹⁷O/¹⁶O are used for the oxide correction. This may be because of the lower internal precision of the ¹⁷O/¹⁶O measurements or that the canonical TFL slope is not the optimal slope by which to calculate ¹⁷O/¹⁶O in this system. Bermingham et al.[7] and Trinquier [42] found that the oxygen isotopic compositions of RuO₃⁻ and U⁺/UO₂⁺, respectively, do not always fractionate along the canonical TFL slope (i.e., δ¹⁸O vs. δ¹⁷O = 0.52 or ¹⁸O/¹⁶O vs. ¹⁷O/¹⁶O = 0.0954). Instead, for RuO₃⁻, which is run in negative mode like Mo, oxygen fractionation may result from mixing isotopically distinct reservoirs of oxygen from the oxygen tank and from the filament.

To test for this possibility during analysis of Mo isotopes, a ⁹⁷Mo-¹⁰⁰Mo double spike was measured, such that sufficiently intense signals were generated for the ¹⁰⁰Mo¹⁷O¹⁶O₂⁻ (2–4 mV) and ¹⁰⁰Mo¹⁸O¹⁶O₂⁻ species (10–20 mV) when using 10¹² Ω resistors (Fig. 2). For the longer of the two analyses, the slope of ¹⁸O/¹⁶O vs. ¹⁷O/¹⁶O was 0.1128 ± 0.015 (2SD), which is marginally higher than the canonical slope of 0.0954. This indicates that mixing of two isotopically distinct reservoirs of oxygen may result in an inaccurate calculation of ¹⁷O/¹⁶O when using the TFL. This effect is evidently minor, however, as the standard data that was oxide corrected using the measured ¹⁷O/¹⁶O produces nearly identical precision. Alternatively, the mixing effect may simply not be observed when ¹⁷O/¹⁶O is measured using ⁹²Mo¹⁷O¹⁶O₂⁻ when its signal intensity is <2 mV. Because the effective range of 10¹³ Ω resistors includes signals of <2 mV, future methods using a 10¹³ Ω resistor to measure the ⁹²Mo¹⁷O¹⁶O₂⁻ or ¹⁰⁰ Mo¹⁷O¹⁶O₂⁻ species and a 10¹² Ω or 10¹³ Ω resistor to measure the ¹⁰⁰Mo¹⁸O¹⁶O₂⁻ will likely be necessary to optimize the precision on all the Mo isotope ratios. It is also worth noting that the intensity of the ⁹²Mo¹⁷O¹⁶O₂⁻ species is higher than the ¹⁰⁰Mo¹⁷O¹⁶O₂⁻ species, due to the larger abundance of ⁹²Mo, and it requires no first-order oxide correction. Thus, the use of the ⁹²Mo¹⁷O¹⁶O₂ species to calculate the ¹⁷O/¹⁶O ratio is advantageous.

Fig. 2.

In-run ¹⁸O/¹⁶O vs. ¹⁷O/¹⁶O measured in situ for two Mo double spike analyses using the ¹⁰⁰Mo¹⁷O¹⁶O₂ and ¹⁰⁰Mo¹⁸O¹⁶O₂ species and 10¹² Ω two resistors. The intensities were oxide corrected using the Nier first-order oxide correction. The average ¹⁸O/¹⁶O and ¹⁷O/¹⁶O measured in situ for 14 standards, where ¹⁷O/¹⁶O was measured using the ⁹²Mo¹⁷O¹⁶O₂ species and a 10¹² Ω resistor, is shown as the black diamond (¹⁸O/¹⁶O = 0.002055 and ¹⁷O/¹⁶O = 0.0003871–Table 5). Also shown are the oxygen isotopic compositions of [39], the “best measured” value reported by IAPC ([43]; citing [44] and [45]), and the “Mean O” value reported by Nagai and Yokoyama [13]. The black line represents the terrestrial fractionation line, which is forced through the UMD mean O value and is calculated using the canonical slope of 0.52.

In addition to the in-run oxygen isotopic compositions of the double spikes, the average measured oxygen isotopic composition of the standards for which ¹⁷O/¹⁶O and ¹⁸O/¹⁶O were measured is shown in Fig. 2 (UMD mean O). The oxygen isotopic composition of the spikes and the standards are similar to the mean oxygen isotopic composition reported by Nagai and Yokoyama [13]. The UMD mean O is not similar to the Nier [39] value, or the value recommended by IUPAC. This dissimilarity suggests that the use of the Nier [39] value for the first-order correction may be inappropriate. However, when the UMD mean O is used for the first-order correction instead, the precision of each ratio and the isotope ratios, excluding ⁹²Mo/⁹⁶Mo, are within ~1–2 ppm of those calculated with the Nier [39] value.

4.2. Gravimetrically prepared spike-standard mixture

To test the accuracy of the methods described above, an isotopically-enriched mixture was analyzed in the same way as samples and standards using the line by line oxygen correction and molecular mass fractionation correction. The mixture was gravimetrically prepared using a ⁹⁷Mo-enriched spike (Oak Ridge National Lab) and the Alfa Aesar Mo plasma standard. The concentration of the ⁹⁷Mo-enriched spike solution was calibrated prior to preparing the mixture. Once the spike and standard were combined in the appropriate proportions, the mixture was allowed to equilibrate on a hotplate at ~120 °C overnight. The mixture was then evaporated to dryness and 2:1 HNO₃:HCl was added and dried one time. Finally, the mixture was taken into solution in 6 M HCl to load onto the filaments. The predicted “anomaly” for the mixture was μ⁹⁷Mo = + 16.7. Repeated measurements of the spike-standard mixture yielded μ⁹⁷Mo = + 15.4 ± 2.9 (2 SD, n = 7), which is in excellent agreement with the predicted value (Table 7, Fig. 1). The very good reproducibility indicates that over short-term, differences in μ⁹⁷Mo < 10 ppm may be resolved. It was suspected that Mo derived from the Re ribbon resulted in relatively large Mo signals that interfered with Mo signals from the samples and standards. To test for this, five of the analyses of the spike-standard mixture were for loads of ~1000 ng and two of the analyses were for loads of ~ 400 ng Mo. The smaller loads showed μ⁹⁷Mo values within uncertainties of the larger loads, so any effect of the loading blank was not reflected in the measurements. Further, any differences in fractionation behavior caused by the different amounts of Mo on the filament were not evident. These data show that the measurement procedure and data reduction provide accurate and precise data that can be routinely used to resolve differences in μ⁹⁷Mo of ~10–15 ppm.

4.3. Natural samples

Data for natural samples are shown in Table 8 and Fig. 3. One piece of Toluca and Allegan were digested and processed using the methods described above. Two other pieces of Toluca and two pieces of Skookum were processed during the development of the method using different chemical separation procedures; resulting in variable yields for some samples (see Table 8 for details).

Fig. 3.

The μ⁹⁷Mo for the Toluca and Skookum iron meteorites, and the Allegan chondrite, compared to some literature values. For Toluca and Skookum, uncertainties are the 2SD of the separate analyses. For Allegan, the uncertainty is the 2SD external reproducibility of the standards, represented also by the grey bar (±5.4 ppm). The μ⁹⁷Mo for all three samples are in good agreement with data from meteorites of the same groups [7]. The datum for Toluca is not within uncertainty of that of Yin et al. [5].

The Mo yield for the three digestions of Toluca ranged from ~15–55%. The yield for one Toluca digestion was low (15%), potentially due to loss of Mo during the loading stage of the primary column using the 0.4 M HCl-0.5 M HF loading solution as described above. The yields for the two digestions of Skookum were similar at around 45%. The μ⁹⁷Mo values obtained for the three digestions of Toluca with variable yields are within uncertainty of one another and the Alfa Aesar Mo standard. The μ⁹⁷Mo for Toluca is not within uncertainty of the Mo isotopic composition reported by Yin et al. [5], but is in good agreement with data reported by Dauphas et al. [4] and Burkhardt et al. [7] for meteorites from the IAB group. Likewise, μ⁹⁷Mo for the two digestions of Skookum are in agreement with data reported for irons of the IVB group [5,4,7]. These data demonstrate that the analyses are not affected by variable yield or different chromatographic processing. The data for Allegan are in agreement with H5 chondrite data from Burkhardt et al. [7]. This indicates that the methods described above are applicable to metallic and silicate matrices.

5. Conclusions

The procedures for measuring Mo by N-TIMS using the Thermo-Fisher Triton Plus are shown to be accurate and precise. The use of a double filament assembly improved signal stability and duration over those achieved using single filaments. Repeated analyses of the Alfa Aesar Specpure^® Mo plasma standard using the oxygen isotope composition measured in situ and a line by line oxide correction yields a long-term external reproducibility of ±5.4 ppm for μ⁹⁷Mo. The repeated measurement of the gravimetrically prepared ⁹⁷Mo spike-standard mixture yielded μ⁹⁷Mo that is in agreement with the predicted value, which demonstrates the accuracy of the technique and that differences in μ⁹⁷Mo ≥ 10 ppm can be routinely resolved. The precision for ⁹⁷Mo/⁹⁶Mo was increased by a factor of ~3.9 compared to ICP-MS [7], and by a factor of ~2.4 compared to other TIMS methods [13]. The analyses of iron meteorite and chondrite samples show that different dissolutions and chromatographic procedures do not produce analytical artifacts related to variable yields, blanks, or matrices.