Vol. 35 – N 1

03 11

p.23–37

An Isotope Dilution ICP-MS Method for the Determination of Mg ⁄ Ca and Sr ⁄ Ca Ratios in Calcium Carbonate Diego P. Fernandez

(1) *,

Alex C. Gagnon and Jess F. Adkins

Geology and Planetary Science, California Institute of Technology, MS 100-23, 1200 E California Blvd, Pasadena, CA 91125, USA * Corresponding author. e-mail: [email protected] (1) Present address: Geology and Geophysics, University of Utah, 115 South 1460 East Room 383, Salt Lake City, UT 84112, USA

Mg ⁄ Ca and Sr ⁄ Ca ratios in calcium carbonate are important components of many palaeoclimate studies. We present an isotope dilution method relying on a single mixed spike containing 25 Mg, 43 Ca and 87 Sr. Dozens of samples per day, as small as 10 lg of carbonate, could be dissolved, spiked and run in an ICP-MS with a precision of 0.8% (2 RSD). Two instruments types, a sector field and a quadrupole ICP-MS, were compared. The best long term precision found was 0.4% (2 RSD), although this increased by up to a factor of two when samples of very different Mg or Sr content were run together in the same sequence. Long term averages for the two instruments concurred. No matrix effects were detected for a range of Ca concentrations between 0.2 and 2 mmol l -1 . Accuracy, tested by measuring synthetic standard solutions, was 0.8% with some systematic trends. We demonstrate the strength of this isotope dilution method for (a) obtaining accurate results for sample sets that present a broad Mg and Sr range and (b) testing solid carbonates as candidate reference materials for interlaboratory consistency. Mg ⁄ Ca and Sr ⁄ Ca results for reference materials were in good agreement with values from the literature. Keywords: carbonate, ICP-MS, quadrupole ICP-MS, sector field ICP-MS, isotope dilution, minor elements.

Les rapports Mg ⁄ Ca et Sr ⁄ Ca des carbonates de calcium sont importants pour de nombreuses études paléoclimatiques. Nous présentons une méthode de dilution isotopique s’appuyant sur un simple ajout (« spike ») constitué d’un mélange contenant 25Mg, 43 Ca et 87 Sr. Des dizaines d’échantillons par jour, aussi petit que 10 lg of carbonate, pourraient être dissous, enrichis (« spiked ») et analysés dans un ICP-MS avec une précision de 0.8% (2 RSD). Deux types d’instruments, ICP-MS à secteur magnétique et ICP-MS quadripolaire, ont été comparés. La meilleure précision à long terme constaté était de 0.4% (2 RSD), bien que cette valeur ait augmenté jusqu’à un facteur deux lorsque des échantillons caractérisés par des teneurs en Mg ou Sr très différentes ont été analysés ensemble dans une même séquence. Les moyennes à long terme pour les deux instruments sont du même ordre. Aucun effet de matrice n’a été détecté pour une gamme de concentrations de Ca comprise entre 0.2 et 2 mmol l -1 . La précision, testée par la mesure de solutions standards de synthèse, était de 0.8%, avec certaines tendances systématiques. Nous démontrons la robustesse de cette méthode de dilution isotopique pour: (1) l’obtention de résultats précis pour des ensembles d’échantillons qui présentent une large gamme de teneurs en Mg et Sr, et (2) tester des carbonates solides comme matériaux de référence candidats à des études de cohérence entre laboratoires. Les rapports Mg ⁄ Ca et Sr ⁄ Ca obtenus pour les matériaux de référence sont en bon accord avec les valeurs de la littérature.

Received 05 Jan 09 – Accepted 07 Aug 09

Mots-clés : carbonate, ICP-MS quadripolaire, ICP-MS à secteur magnétique, dilution isotopique, éléments mineurs.

Naturally occurring calcium carbonate in corals, sclerosponges, foraminifera, otoliths, speleothems, etc. contains magnesium and strontium as minor components. Notwith-

standing the complex mechanisms involved in the incorporation of these elements in calcitic or aragonitic matrices, a number of studies indicate that their spatial distribution

doi: 10.1111/j.1751-908X.2010.0031.x ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

23

reflects simple environmental changes. The best known examples, the reconstruction of past sea surface temperatures from foraminiferal Mg ⁄ Ca data (Nurnberg et al. 1996) and coralline Sr ⁄ Ca values (Smith et al. 1979, Beck et al. 1992) have opened up a wealth of opportunities in palaeoclimate research. Analytical advances in the 1990s were in great part responsible for this development. On the one hand, the widespread use of inductively coupled plasma (ICP) introduction systems allowed a high throughput of samples with very effective ionisation efficiency. In addition, new detectors possessing unsurpassed dynamic ranges made it feasible to acquire large signals from major elements together with smaller signals from minor or trace elements. Finally, samples as small as 25 lg could be analysed owing to the high sensitivity of these new instruments. The first methods presented used quadrupole mass spectrometry (Q-ICP-MS) (Lea and Martin 1996, LeCornec and Correge 1997), sector field mass spectrometry (SF-ICP-MS) (Rosenthal et al. 1999) and atomic emission spectroscopy (ICP-AES) (Schrag 1999). The determination of trace elements in addition to Mg ⁄ Ca and Sr ⁄ Ca ratios using the same method was also investigated (Lear et al. 2002, Yu et al. 2005, Marchitto 2006), and a careful comparison between emission and mass spectrometry methods was considered (Andreasen et al. 2006). Despite the high precision attained by some of these methods, specific problems hampering their accuracy were evident. Matrix effects for ICP-AES methods, being large and complex (Lehn et al. 2003), required special calibrations (de Villiers et al. 2002, Wara et al. 2003). However, for those methods using bracketing calibrators, the uncertainty associated with differences in matrix effects between the natural samples and the calibrators was not assessed. On the other hand, ICP-MS methods relying on internal standards assume the selected spike element, for example Sc or Y, is subject to the same behaviour in the plasma as Mg, Ca and Sr. Moreover, the linearity of the broad dynamic range detectors depends on the cross-calibration between different modes of operation, a requirement that cannot always be easily fulfilled (Marchitto 2006). These shortcomings, in addition to the uncertainties in composition of the calibrators used, became evident in an inter-laboratory calibration study (Rosenthal et al. 2004) revealing precisions of 3.4% and 1.8% for Mg ⁄ Ca and Sr ⁄ Ca ratios, respectively, in determinations of the synthetic standard solution distributed for the study. The need was also clear for well characterised matrix-matched reference materials (RMs), with Mg and ⁄ or Sr contents in the appropriate range,

24

and some effort has been devoted to this goal (Greaves et al. 2005, 2008, Sturgeon et al. 2005). In this work, we present an isotope dilution mass spectrometry (ID-ICP-MS) method to determine Mg ⁄ Ca and Sr ⁄ Ca ratios, and we assess the range of concentrations for which it can be applied as well as the long term precision. Isotope dilution-ICP-MS relies on the addition (spiking) of an enriched isotope (tracer) of the element of interest followed by the measurement of the ratio between this isotope and a different one from the sample (Klingbeil et al. 2001, Beauchemin 2006). In the present study, we used a mixed tracer containing mainly the isotopes 25Mg, 43Ca and 87Sr and measured the isotopic ratios 24Mg ⁄ 25Mg, 48 Ca ⁄ 43Ca and 88Sr ⁄ 87Sr using ICP-MS. The use of an isotope of the same element to be determined minimised the differences in plasma behaviour. Since isotope measurements are all made in the same sample matrix and regular calibration curves prepared from external calibrators are not used, the isotope dilution method has been recognised as an analytical tool capable of the highest accuracy. The use of a mixed tracer has an additional advantage because the elemental ratio calculation does not depend on the amount of tracer added. The dissolution of microgram amounts of the calcium carbonate sample under study and the addition and mixing of the tracer could be quickly accomplished in sub-millilitre auto-sampler vials since there was no need to measure their masses.

Experimental procedures Instrumentation A Finnigan Element ICP-MS, with standard Ni cones, quartz injector and torch without a capacitive decoupling (CD) system, was used to develop the experimental method. We also used cooled dual pass quartz (PC3 SSI Elemental Scientific; Omaha, Nebraska, USA) or PTFE Scotttype spray chambers, PTFE nebulisers with flow rates of 22, 35 and 50 ll min-1 (Elemental Scientific PFA-20 and PFA50), and an auto-sampler (ASX-100 CETAC; Omaha, Nebraska, USA). Low resolution was selected and the introduction system was tuned in the usual way (sample gas flow, torch x, y, z position and lenses) with observed sensitivities up to 300 kcps (for 88Sr) for a 1 lg l-1 Sr solution. A Thermo Scientific; West Palm Beach, Florida, USA, Neptune multi-collector ICP-MS with Ni cones and a dual pass spray chamber was used to check the isotopic composition of the Ca analytical reference solution. The method, with minor changes, was also implemented using an Agilent 7500ce quadrupole ICP-MS with a CD system (at the University of Utah) together with a CETAC ASX-520 autosampler,

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

Pt cones, cooled dual pass quartz spray chamber (Elemental Scientific PC3-SSI) and PTFE nebuliser with a 100 ll min-1 flow rate (Elemental Scientific ST-100). Sensitivities up to 140 kcps for a 1 lg l-1 Sr solution were obtained. The dynamic reaction cell usually employed to reduce the presence of molecular interferences installed in this instrument was not used. A micro-sampling device (Micro Mill; New Wave Research, Portland, Oregon, USA) was used to obtain microgram amounts of samples along bands in corals, stalagmites or soil carbonates. All solutions and samples were handled in laminar flow benches. Standard and tracer solutions were prepared gravimetrically in fluorinated ethylene propylene (FEP) bottles previously leached sequentially with hot 50% v ⁄ v HNO3, 50% v ⁄ v HCl and water, and solutions were kept in a closed humid environment in order to minimise evaporation losses. Polypropylene or FEP auto-sampler vials were leached in hot 10% v ⁄ v HCl. Pipette tips were acid washed followed by three rinses with water. Proper anti-static procedures were used when weighing FEP containers on calibrated scales, and buoyancy corrections were applied. All solid reference samples were transferred or rinsed from the weighing container into the bottle with water and dissolved by slow addition of nitric acid in order to prevent losses. Reference samples were dried or handled under argon. Samples were weighed in a stainless steel boat using a microbalance, transferred into previously cleaned polypropylene vials and dissolved with 5% v ⁄ v HNO3 (Baseline; Seastar, Sidney, British Columbia, Canada). No centrifugation was applied to samples.

cer for each lg of CaCO3. The mass of the sample was determined only to aid the rough calculation of the required amount of tracer and therefore  30% uncertainty in mass could be tolerated. Total Ca concentrations between 0.5 mmol l-1 and 1 mmol l-1 were normally obtained. Alternatively, larger samples were dissolved in a vial with enough nitric acid to obtain a 4 mmol l-1 total Ca solution and an aliquot of this solution was mixed in an autosampler vial with the mixed spike, and diluted with 5% v ⁄ v HNO3 to a total Ca concentration between 0.5 mmol l-1 and 1 mmol l-1. The isotopic ratios attained in this way for coral samples were: 24Mg ⁄ 25Mg  0.4, 48 Ca ⁄ 43Ca  0.12 and 88Sr ⁄ 87Sr  1.2, close to the optimal ratios predicted by isotope dilution error equations of 0.34, 0.04 and 1 respectively. In the case of Ca, a ratio three times larger was used in order to minimise the use of this costly isotope, although this does not entail a significant deterioration of the method precision (Hearn et al. 2005). Solutions were aspirated in aliquots of about 50 ll (for the lowest flow rate nebuliser), at least three times during the analysis sequence. The same volume of solution XG, diluted to match the chosen total Ca concentration for the samples, was run between sample test portions, and its measured ratios used to correct for mass bias. Typical wash-out times of 1 min or less for the cooled dual spray chamber were enough to reach intensity levels comparable to the background signal, usually below 5 kcps. After about 18 hr, the precision declined because of the deposition of solids on the cones, allowing a maximum throughput of about fifty samples per day. 43

Method For our full isotope dilution method we prepared two types of solution containing enriched isotopic tracers. For the first, a ‘spike’ was added to each natural sample (labelled ‘XS’). This solution required a very different isotopic ratio from the natural samples for the three elements of interest. Spiked standard solutions were used to monitor bias and drift of measured isotopic ratios in the ICP-MS during a run. This solution was labelled ‘XG’ and it was prepared to have isotopic ratios close to those present in the spiked samples, to minimise bias inherent in measurement by ICP-MS. An accurate characterisation of the isotopic ratios in XG and both the isotopic ratios and concentrations in XS was crucial to implementing the method. A minimum of  10 lg of CaCO3 was weighed using a microbalance into a stainless steel container and then transferred into an autosampler vial, where it was dissolved in 200 ll of 5% v ⁄ v HNO3 and spiked with 1 ll of XS tra-

Ca2+, 24Mg+, 25Mg+, 43Ca+, 87Sr2+, 48Ca+, 86Sr+, Sr and 88Sr+ beams were collected in low resolution mode in the sector field mass spectrometer using the following conditions: 10% mass window, 200 masses per peak, and 10 ms acquisition time per sample. The settling time for the magnet was selected automatically by the instrument’s software, although some experiments were performed with different values for this parameter. The standard deviation (1s) for fifty replicates was usually below 1.5% and the minimum intensity allowed for any of the isotopes, excluding doubly-charged ions, was 3 Mcps. Usually, this minimum intensity was achieved by using 10 lg of carbonate, although in occasions when the instrumental sensitivity was lower than usual or when the Mg content of the sample was below  2 mmol mol-1, a larger mass was required to fulfil this requirement. The largest background correction, around 0.2%, was applied to the 24 Mg+ ⁄ 25Mg+ intensity ratio, since 24Mg+ usually had the smallest intensity when analysing corals. For sclerosponges and some stalagmites, having much higher Mg content than corals, the lowest intensity corresponded to 48Ca+, in 87

+

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

25

which case the highest background correction was applied, usually  0.1%. The RM BAM RS3 (Federal Institute for Materials Research and Testing, Germany) represents a special case, where Mg ⁄ Ca is 0.8 mmol mol-1. In this case, it was necessary to dissolve a larger amount of material ( 30 lg) in order to maintain the background correction for 24Mg+ lower than 0.2%. In a previous ICP-MS method using an Element instrument (Marchitto 2006), the drift in the cross-calibration parameter between the two modes of the detector (analogue and counting) prevented precise results when acquiring ratios by using mixed detection, i.e., the less abundant isotope measured using counting mode and the more abundant using analogue mode. An example of the amount of drift in the case of our sector field instrument is shown in Figure 1, where it can be observed that the value for a ratio obtained using mixed detection varied considerably more than for a pure analogue or pure counting ratio. Even when it was possible to include a cross-calibration procedure during a sequence by measuring an isotope in the appropriate intensity range and the detector set to ‘both’ mode, it was observed that this procedure yielded a lower precision, in addition to slowing down the whole measurement. The preferred detection mode was analogue for all the isotopes measured for the samples, RMs and blanks.

8.0

24Mg+/25Mg+

7.8

7.6

7.4

7.2

0

250 500 750 Time (minutes)

1000

Figure 1. Mixed detection mode drift of raw ratios. 24

Mg + ⁄ 2 5 Mg + measured for the same solution at

three different concentrations and corresponding detection modes: open squares, both

24

Mg + and

25

Mg in analogue mode; open triangles, both

24

Mg + and

24

+

+

25

Mg + in counting mode, filled-in circles,

Mg in analogue mode and

mode.

26

25

Mg + in counting

The contribution of the doubly-charged ion 48Ca2+ to the 24Mg+ beam was estimated by measuring 43 Ca2+ ⁄ 43Ca+ in the sector field instrument. The value obtained for this ratio was about 0.7%, and assuming that this also holds for 48Ca2+ ⁄ 48Ca+, a correction of about 2% on 24Mg+ ⁄ 25Mg+ was required for a sample with Mg ⁄ Ca values close to those found in corals (around 3 mmol mol-1). Because of the magnitude of this correction, the measurement of the ratio 43Ca2+ ⁄ 43Ca+ was added to the method, and the correction applied offline. In the same fashion, the impact of 86Sr2+ on 43Ca+ was considered. The ratio 87Sr2+ ⁄ 87Sr+ was consistently 2.5% and the correction that this value entailed for 43Ca+ ⁄ 48Ca+ was about 0.5%. 87Sr2+ and 86Sr+ were thus measured and used, together with the 86Sr+ intensity, to correct 48Ca+ ⁄ 43Ca+. It was also important to monitor the existence of Kr as a contaminant of argon. Usually the correction due to its presence was unimportant because the ratio 86Kr2+ ⁄ 86Kr+ is fifty times smaller than the one for Sr and the Kr is only present in very low concentrations, although occasionally high levels were observed. A number of possible interferences other than the doubly-charged ions were identified by acquiring spectra in medium and high resolution modes when using the Element ICP-MS. The interferences [14N(16O)2H]+, [14N18O16O]+ and possibly [32S16O]+ on 48Ca+ were observed both in HNO3 and Ca solutions. Their combined intensity represented, at most, 0.2% of the signal obtained for 48Ca+ in a 1 mmol l-1 total Ca solution, so it was assumed that their contribution was stable and could be corrected with the background measurements. We observed an even smaller effect of [(14N)3H]+ on 43Ca+. The presence of 48Ti+ was also studied at the highest resolution available (about 10000, where 48Ti+ and 48Ca+ could be resolved) and its presence could not be detected for a surface coral solution. The presence of Ti was not studied for sclerosponges or stalagmites. Similarly, the presence of [48Ca40Ar]+, a possible interference on 88Sr+, was not observed. Over one year, the average mass bias (and 1s uncertainties) for 24Mg+ ⁄ 25Mg+, 43Ca+ ⁄ 48Ca+ and 88Sr+ ⁄ 87Sr+ using the sector field ICP-MS were 18 ± 6%, 13 ± 4% and 2 ± 0.6%, respectively. The mass bias drift within a run was variable from day to day and its magnitude depended on the mass calibration stability as well as the conditions of the introduction system, especially of the cones. It was observed that the peak shapes of 24Mg+ and 48 Ca+ played a crucial role in the amount of drift during a run. In general, these two isotopes presented peak tops that were less horizontal than those of the other isotopes,

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

and thus the average intensities were more sensitive to small changes in the mass calibration. These characteristic peak shapes for the different isotopes remained consistent for most of the runs and did not change substantially when the settling time for the magnet or the focus offset voltage was modified. Due to this behaviour, the amount of drift for 24 Mg+ ⁄ 25Mg+ and 43Ca+ ⁄ 48Ca+ was usually larger than that corresponding to 88Sr+ ⁄ 87Sr+ and 88Sr+ ⁄ 86Sr+. The average relative difference between consecutive ratios when comparing different aliquots of solution XG was around 0.4% for Mg and Ca ratios and 0.2% for Sr ratios. Consequently, the drift correction for Mg and Ca was an important component of the precision of the method. In order to monitor this behaviour, peak shapes were quantified by the relative standard deviation (RSD) of the twenty intensity values collected for each peak. Since, in general, these intensities changed monotonically with mass, the RSD was an indicator of the slope of the peak top. Figure 2 shows the correlation between the peak top RSD corresponding to 24Mg+ and the relative difference between consecutive values of 24Mg ⁄ 25Mg for a run with an exceptionally high amount of drift. Monitoring of the peak shape was used as a quality control tool for the method: a sample was rejected when the relative difference between the bracketing ratios corresponding to XG solutions was > 0.75%. Due to this peak shape drift, it was observed that the precision improved when the samples were run in three or more aliquots, with an XG solution run in between, rather than by using a longer single collection time interval. In spite of some deterioration of the replicate

D(24Mg/25Mg) (%)

3

0

-3

-5 -4

-2 24Mg

0 2 peak top (% RSD)

4

Figure 2. Correlation between isotopic ratio drift and peak shape. Relative

24

Mg + ⁄ 2 5 Mg + difference

between two consecutive measurements of solution XG [D( 2 4 Mg + ⁄ 25 Mg + )] versus the relative standard deviation of the peak top intensity of ( 2 4 Mg + peak top RSD).

24

Mg + (N = 20)

statistics for the shorter sample time (2 min per aliquot), the average of three or more drift corrected replicates was found to be a more reliable value. Blank, interference and mass bias corrections, using linear interpolation, were applied offline to obtain the corrected ratios RMg = 24Mg ⁄ 25Mg, RCa = 48Ca ⁄ 43Ca and RSr = 88Sr ⁄ 87Sr. The mass bias correction factor for each element was calculated as the ratio between the calibrated and the measured value of each of those isotopic ratios for the bracketing solution XG. The isotope dilution equation was then used to calculate Mg ⁄ Ca and Sr ⁄ Ca present in the sample as follows: Mg ðRMg  CMg ÞðRCa  HCa Þ C25 A 48 ¼ Ca ðRCa  CCa ÞðRMg  HMg Þ C43 A 24

ð1Þ

Sr ðRSr  CSr ÞðRCa  HCa Þ C87 A 48 ¼ Ca ðRCa  CCa ÞðRSr  HSr Þ C43 A 88

ð2Þ

where CMg = 24Mg ⁄ 25Mg, CCa = 48Ca ⁄ 43Ca and CSr = 88Sr ⁄ 87Sr are the isotopic ratios in the mixed spike; C25, C43 and C87 are the molar concentration of isotopes 25 Mg, 43Ca and 87Sr in the mixed spike respectively; QMg = 25Mg ⁄ 24Mg, QCa = 43Ca ⁄ 48Ca and QSr = 87 Sr ⁄ 88Sr are the isotopic ratios in the sample, assumed to be natural; A24, A48, A88 are the natural isotopic abundances of 24Mg, 48Ca and 88Sr respectively. See below (Calibration) for the values used for the calculation. Minor changes in the method were needed when the Q-ICP-MS Agilent 7500ce was used to implement the method. The measured beams for spiked samples and RMs were in this case 24Mg+, 25Mg+, 43Ca+, 48Ca+, 86Sr+, 87 + Sr and 88Sr+, with the following conditions: 0.12 s total integration time (0.04 s at three masses), analogue detection and hundred replicates. Instead of considering the intensity of doubly-charged ions at half masses (43Ca2+and 87 2+ Sr ), a different approach was implemented in the quadrupole instrument to correct the intensities of 24Mg+ and 43Ca+ respectively. Natural Ca and Sr standard solutions were included in the sequence and the fractions of doubly-charged ions for each of these two elements were obtained through the ratios 44Ca2+ ⁄ 44Ca+ and 88 2+ 88 + Sr ⁄ Sr . The values obtained were close to 0.5% and 1.2% respectively, and good agreement was found when other ratios were considered to estimate the fraction of doubly-charged ions, e.g., 48Ca2+ ⁄ 48Ca+ and 86 2+ 88 + Sr ⁄ Sr . The average mass bias (and 1s uncertainties) for 24Mg+ ⁄ 25Mg+, 43Ca+ ⁄ 48Ca+ and 88Sr+ ⁄ 87Sr+ were in this case 4 ± 1%, 15 ± 1% and 2 ± 0.3% respectively, while the average relative difference between consecutive ratios when comparing different aliquots of solution XG

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

27

was 0.2% for Mg and Ca ratios and 0.1% for Sr ratios. These values, about half of those observed for the Element instrument, played a lesser role in the error associated with drift. Finally, analyses of sample solutions were run in only one test portion.

Reference materials and tracer To calibrate the solutions for a given element, both isotopic and concentration calibrators (normally RMs) were required. Ideally, a single RM could be used for both purposes, but we were only able to do this for strontium. NIST SRM 987 (from the National Institute of Standards and Technology) is an isotopic certified RM (CRM). Its Ca and Mg levels reported (5 lg g-1 and < 2 lg g-1, respectively) had a very small effect on the Mg-Ca-Sr synthetic mixed standard solutions used for the accuracy test or calibration purposes (see below). The analytical and isotopic Sr standard solution was prepared by weighing this material after drying at 110 C for 2 hr and dissolving it in 5% v ⁄ v HNO3. A magnesium metal CRM (NIST SRM 980) was obtained and dissolved in 5% v ⁄ v HNO3 to prepare an isotopic standard solution. Because this material was partially oxidised and its purity not reported, distilled magnesium 99.999% (47, 475-4 Aldrich; St. Louis, Missouri, USA) was used as an analytical calibrator. Trace contamination of Ca and Sr had no effect on synthetic mixed standard solutions. The lustrous metal was weighed under argon and dissolved in dilute nitric acid. Finally, the isotopic calcium reference solution was prepared from CaF2 used previously by Russell et al. (1978). A commercial CaCO3 99.999% (Aldrich 48,180-7), dried at 125 C for 2 hr, was the source for the analytical Ca standard solution. The Sr content of this standard solution was estimated as 12 lg g-1 and this value was used to correct the Sr concentration when this solution was used to prepare synthetic mixed standard solutions. Gravimetric dilutions were prepared from all the isotopic and analytical standard solutions in 5% v ⁄ v HNO3. 25

Mg enriched MgO (Series RN Batch 217201), 43Ca enriched CaCO3 (NX Batch 169191) and 87Sr enriched SrCO3 (Series LH Batch 136990) were purchased from the Oak Ridge National Laboratory and used to prepare the mixed tracer solution XS. These materials were separately dissolved in 5% v ⁄ v HNO3 and then aliquots of the solutions mixed gravimetrically to obtain the spike solution. Our goal was to make a mixed spike that when added to natural samples yielded ratios that were optimised for the isotope dilution method assuming natural Mg ⁄ Ca and Sr ⁄ Ca ratios of 3 mmol mol-1 and 10 mmol mol-1 respectively. Finally, a solution made from a coral specimen (Porites lutea) was mixed with an aliquot of the mixed tracer XS in

28

order to obtain the spiked mass bias calibrator XG, with isotopic ratios 24Mg ⁄ 25Mg, 48Ca ⁄ 43Ca and 88Sr ⁄ 87Sr close to those of spiked samples. This solution was used to correct for mass bias during ICP-MS runs by sample bracketing. The coral was cleaned repeatedly with distilled water under ultrasound treatment, leached with dilute HNO3 and dried at 110 C for 2 hr.

Calibration Both the concentration and isotopic characterisation of the mixed tracer XS were required to implement the isotope dilution method. On the other hand, the isotopic composition of solution XG was needed to correct for mass bias. This procedure relied on the knowledge of the isotopic abundances and concentrations of individual Mg, Ca and Sr natural analytical standard solutions. Once these were known, it was possible to calibrate (a) isotopic ratios of solutions XS and XG by comparing intensity ratios measured for those solutions and for the standard solution calibrators and (b) concentrations of mixed tracer XS by obtaining the intensity ratios for gravimetric mixtures prepared from XS and the individual analytical standard solutions (reverse isotope dilution). Table 1 reports the results of the calibration. It should be noted that the isotopic ratios for the mixed tracer XS, needed in Equations (1, 2) cannot be assumed to be equal to the values reported by Oak Ridge National Laboratory since cross-contamination is important in one case (Sr, see below). To obtain the isotopic abundances of the Mg analytical standard solution, 24Mg+, 25Mg+ and 26Mg+ intensities were measured in the sector field instrument for an alternate sequence of isotopic and analytical standard solutions possessing similar total Mg concentration. The reported isotopic ratios for the NIST CRM (see Table 1) were used to correct for mass bias and drift. The range of three means for 24Mg ⁄ 25Mg and 24Mg ⁄ 26Mg spanning three different days were 0.15% and the intra-run precision was about 1% RSD. The abundances calculated from these two ratios are reported in Table 1, with errors propagated from errors in the ratios estimated as three times the range obtained. For the Ca calibration, 48Ca ⁄ 43Ca intensity ratio was initially compared for both standard solutions (isotopic and analytical) using the Element ICP-MS. In this case, the precision of the 48Ca+ ⁄ 43Ca+ ratio on ten separate days over 1 year was about twenty times larger than the Mg case, close to 2%, and on several occasions a drift over time was observed. We speculate that the difference in the matrix of the two Ca solutions (the isotopic standard solution was prepared before the start of this research in

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

Table 1. Isotopic abundances a of reference samples Isotope 24

Mg Mg 26 Mg 40 Ca 42 Ca 43 Ca 44 Ca 48 Ca 84 Sr 86 Sr 87 Sr 88 Sr 25

Mg isotopic b

Mg analytical c

78.99(3) 10.00(1) 11.01(2)

78.7(4) 10.1(3) 11.2(3)

Ca isotopic d

Ca analytical e

96.982(9) 0.64214(6) 0.13340(2) 2.0568(1) 0.18245(5)

0.1334(2)

Sr f

0.1825(6) 0.5574(15) 9.8566(34) 7.0015(26) 82.5845(66)

a Abundances are in mole percent. The number in parentheses represents the uncertainty, which corresponds to the last figure shown. The uncertainty reported for Mg analytical reference samples was propagated from the uncertainty of isotopic ratios estimated as three times the range b c NIST SRM 980, Isotopic Mg (oxide) CRM. Mg, distilled, dendritic pieces, 99.999%, Aldrich 47,475-4 Lot No (N = 3). d e Solution obtained from CaF2 (Russell et al. 1978). CaCO3 99.999 +% Aldrich 48,180-7 Lot No 11511AO. 16734LO. f NIST SRM 987, isotopic and analytical SrCO3 CRM.

1% v ⁄ v HCl) could have been in part responsible for this amount of variability by affecting some of the processes at the spectrometer introduction system, such as deposition on the cones or adsorption on the spray chamber (Andren et al. 2004). Thus, given this poor precision we could not rule out an isotopic difference between the analytical and isotopic Ca standard solutions. Moreover, commercial synthetic CaCO3 have previously been found to be isotopically distinct from natural samples (Russell et al. 1978). In order to circumvent this shortcoming, the Ca isotopic and analytical standard solutions together with two more natural abundance Ca solutions (Durango Apatite and deep sea coral) were analysed using a multi-collector ICPMS (Neptune). In addition, an aliquot of the isotopic Ca standard solution (prepared in 1% v ⁄ v HCl) was evaporated and dissolved in 5% v ⁄ v HNO3, thus matching the matrix of the other standard solutions. The intensities of 43 Ca+, 87Sr2+, 44Ca+, 48Ca+, 86Sr+, 87Sr+ and 88Sr+ were measured in medium resolution (to optically resolve interferences on 44Ca+ and 48Ca+). The non-resolvable interferences of 86Sr2+ on 43Ca+ and 88Sr2+ on 44Ca+ were corrected offline using the intensity of 87Sr2+ and the measured 86Sr+ ⁄ 87Sr+ and 88Sr+ ⁄ 87Sr+ ratios, as was explained in the previous section. Each sample was preceded by a pre-wash, wash and blank, and an offline blank subtraction was applied. Solution XG was measured every third sample; peak centring on 87Sr+ and 43Ca+ was performed just prior to measuring each standard solution. Because solution XG was designed for ID analysis, its isotopic abundance was very different from that found in natural Ca samples. To minimise a systematic memory bias due to the measurement of very different ratios, samples were

analysed in random order, with different samples following solution XG throughout the run. During the 13-hr run, solution XG was measured nine times with an overall drift in the measured 43Ca+ ⁄ 48Ca+ and 44Ca+ ⁄ 48Ca+ intensity ratios of 0.22% and 0.15%, respectively. To correct for drift in instrumental mass fractionation, a linear model was used to interpolate between the bracketing XG measurements. The 43Ca+ ⁄ 48Ca+ and 44Ca+ ⁄ 48Ca+ ratios for all the materials were the same after drift correction within 0.1%, and the values were in good agreement with the averages reported by Russell et al. (1978). The results for the Mg and Ca standard solutions, together with reported Sr reference sample abundances, are shown in Table 1. The Mg analytical reference sample was enriched in 25Mg as compared with the isotopic reference, and the difference is consistent with the distillation process used in the purification (Esat et al. 1986). Considering the narrow range of Ca fractionation in nature (De La Rocha and DePaolo 2000, Wieser et al. 2004, Steuber and Buhl 2006) and the agreement (within the experimental error of 0.1%) of isotopic ratios 48Ca ⁄ 43Ca and 44 Ca ⁄ 48Ca between analytical and isotopic Ca reference samples, their isotopic abundances were assumed to be equal for all isotopes. The uncertainty reported in Table 1 for 43Ca and 48Ca abundances were calculated assuming a 0.1% difference in the 43Ca ⁄ 48Ca isotopic ratio between the analytical and isotopic Ca reference samples and using an exponential fractionation law. The isotopic ratio calibrator XG was itself calibrated using the Mg, Ca and Sr isotopic reference samples in the same fashion. In this case, the substantial difference

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

29

between the ratios of the reference samples (close to natural abundances) and those of XG required special conditions, including wash-out times of at least 10 min and the use of HF, with concentrations  1 lmol l-1 for about a minute as an additional rinsing step. Intensity ratios of 24 Mg+ ⁄ 25Mg+, 24Mg+ ⁄ 26Mg+, 48Ca+ ⁄ 43Ca+, 88Sr+ ⁄ 86Sr+ and 88Sr+ ⁄ 87Sr+ were compared, and 48Ca2+ and 86Sr2+ interferences on 24Mg+ and 43Ca+ respectively were corrected by measuring the intensities of ions 43Ca2+ and 87 2+ Sr (see above). With the calibration of XG in hand, the isotopic ratios of the mixed spike XS were calibrated by comparison with the XG solution, which was also used to correct for mass bias and drift. Finally, gravimetric mixtures of analytical standard solutions and mixed tracer XS were prepared and their isotopic ratios measured in the sector field mass spectrometer and corrected offline in a similar way as that reported above (Method). Using the isotopic ratios for the mixed tracer and reverse isotope dilution, the concentrations for the mixed spike were obtained.

Table 2. Isotopic ratios of spiked stock solution XG and tracer XS

the Oak Ridge National Laboratory, there was a possible increase of the 24Mg and the 26Mg abundance for the mixed tracer with respect to the Oak Ridge National Laboratory values. This difference is consistent with cross-contamination, as reported Mg impurities were < 100 lg g-1 for both 43Ca and 87Sr tracers.

Table 2 shows the isotopic ratios for solutions XG and XS and Table 3 compares the abundances of individual tracers, as reported by Oak Ridge National Laboratory, with the abundances for the mixed tracer calculated from the measured ratios. In the case of Mg, the abundances in XS were calculated from the measured 24Mg ⁄ 25Mg and 24 Mg ⁄ 26Mg ratios. Since all three Mg isotopes were collected, the corresponding abundances could be calculated without any assumption. Although the abundance obtained for 24Mg agreed within uncertainty with that reported by

In the case of Ca, only the ratio 48Ca ⁄ 43Ca was measured, which agreed within experimental error with that reported by Oak Ridge National Laboratory. The cross-contamination was estimated considering the reported Ca content in 25Mg and 87Sr individual tracers and the masses and concentrations of individual tracer solutions used to make up the mixed tracer solution XS. Assuming a natural isotopic composition for the contaminant Ca and using the measured 48Ca ⁄ 43Ca value, it is possible to calculate the abundances for all Ca isotopes in the mixed

XG

XS

Average SE (2s) 24

25

Mg ⁄ Mg Mg ⁄ 26Mg 48 Ca ⁄ 43Ca 88 Sr ⁄ 86Sr 88 Sr ⁄ 87Sr 24

0.515 6.99 0.1152 7.9 1.1211

0.001 0.04 0.0004 0.04 0.0008

N

Average

SE (2s)

N

13 3 14 3 11

0.00984 4.27 0.001063 4.52 0.08849

0.00004 0.18 0.000008 0.16 0.00004

5 3 5 3 3

SE, standard error or standard deviation of the mean.

Table 3. Isotopic abundances of individual and mixed (XS) standard solutions a and concentration of mixed standard solutions Isotope

24

Mg Mg 26 Mg Mg 40 Ca 42 Ca 43 Ca 44 Ca 48 Ca Ca 84 Sr 86 Sr 87 Sr 88 Sr Sr 25

Mg b (%)

Ca c (%)

Sr d (%)

0.963(10) 98.814(20) 0.223(5) 10.13(8) 0.780(6) 83.93(10) 5.06(3) 0.090(1) 0.06(3) 88(4) 3(2) 9(3)

XS (%)

(nmol g - 1 )

0.972(2) 98.8(4) 0.227(1)

0.512(6) 52.07(17) 0.120(3) 52.70(9) 16.47(13) 1.27(1) 136.2(2) 8.21(5) 0.145(1) 162.3(1) < 0.01 1.19(5) 61.0(1) 5.40(1) 67.58 (5)

10.15(8) 0.78(1) 83.91(11) 5.06(3) 0.089(1) 0.010(5) 0.82(2) 91.26(10) 7.91(10)

0.010(6) 1.90(8) 90.12(12) 7.98(10)

a

Abundances are in mole percent. The number between parentheses represents the uncertainty which correspond to the last figures b 25 c 43 Mg enriched MgO from Oak Ridge National Laboratory Series RN Batch 217201. Ca enriched CaCO3 from Oak shown. d 87 Sr enriched SrCO3 from Oak Ridge National Laboratory Series LH Batch 136990. Ridge National Laboratory Series NX Batch 169191.

30

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

Sr abundance was assumed equal to that of the Oak Ridge tracer, a reasonable approximation considering the low abundance of this isotope. The substantial difference between individual and mixed tracer 86Sr abundance was due to the Sr content in the 43CaCO3 tracer, reported as 5 mg g-1. The isotopic composition of this impurity was measured confirming the expected enrichment in 86Sr due to the electromagnetic separation method for 43Ca tracer production, which cannot separate completely 43Ca+ ions from the 86Sr2+ ions. The correction for Sr contamination in the 25Mg tracer (reported < 100 lg g-1) was negligible. The total Mg, Ca and Sr concentrations for the mixed tracer XS and their standard errors (standard deviation of the mean) (SE 1s, N ‡ 4) are also reported in Table 3. Using the total concentration for a given element and the abundances for the different isotopes in the tracer, it was possible to calculate the isotopic concentrations, reported in nmol g-1. The larger relative uncertainties for the concentration of individual isotopes as compared with the precision of the total element concentration reflect the combined uncertainties obtained from the uncertainties in the abundances and the total concentrations.

Matrix effect and accuracy The matrix effect was studied for a solution from a deep sea coral as a function of the total Ca concentration and is shown in Figure 3. The error bar assigned to each point is the standard deviation of three 50 ll aliquots run from the same solution. The precision for the twelve different Ca concentrations measured was 0.8% (2s), for both Mg ⁄ Ca and Sr ⁄ Ca ratios, and no trend was detected. Both the Mg ⁄ Ca and the Sr ⁄ Ca values were randomly scattered around the mean over about a factor of ten of the total Ca concentration. We conclude that there was no matrix effect in our method at the 1% level. In addition, the matrix effect can also be a function of Mg and Sr concentration. In order to assess the range of application for the method and as a test for its accuracy, nine synthetic solutions were prepared gravimetrically using the analytical Mg, Ca and Sr standard solutions. The calculated Mg ⁄ Ca and Sr ⁄ Ca values spanned a decade

Mg/Ca (mmol mol-1)

84

3.11

(a)

3.06

3.01 0

10.5 Sr/Ca (mmol mol-1)

spike. As can be observed in Table 3 by comparing the values for the individual Ca tracer (reported by ORNL) and those calculated for XS, the change due to the contamination was smaller than the uncertainty for all the isotopes with the exception of 48Ca, for which the difference was comparable to the calculated uncertainty.

0.5 1 1.5 [Ca]TOTAL (mmol l-1)

2

(b)

10.3

10.1 0

0.5 1 1.5 [Ca]TOTAL (mmol l-1)

2

Figure 3. Matrix effect. (a) Mg ⁄ Ca and (b) Sr ⁄ Ca ratios as a function of total Ca concentration. The uncertainties represent the standard deviation of three replicates.

around the average values for deep sea corals. The presence of contaminant Sr in the Ca analytical standard solution, estimated as 12 lg g-1, entailed a correction for the calculated Sr ⁄ Ca value in the synthetic solution of at most 0.4%. This number can be contrasted with the correction estimated just from uncertainties in the masses and concentrations of under 0.1%. Figure 4 shows the difference between the measured and calculated Mg ⁄ Ca and Sr ⁄ Ca values for nine synthetic solutions. Measured Mg ⁄ Ca values were on average 0.5% smaller than those calculated and an improvement in the agreement could be observed for the low Mg values. The effect of Sr on the Mg ⁄ Ca values showed no clear trend. Synthetic samples with Mg ⁄ Ca contents up to  15 mmol mol-1 could be balanced within  0.8%, regardless of Sr content. However, the error for the abundance of 24Mg in the analytical Mg standard solution (0.5%) could shift the points horizontally; for instance, using and abundance of 78.3% for 24Mg the points would be scattered around the origin within 0.2% accuracy. Strontium measurements showed, on the contrary, a marked trend with the Mg content of the sample, decreas-

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

31

(Mg/Ca)meas/(Mg/Ca)true - 1

(a)

would not be modified. In this case, the agreement between measured and calculated Sr ⁄ Ca would improve for the two solutions with lower Mg content, while the point corresponding to the solution with Mg ⁄ Ca  15 mmol mol-1 would resemble the behaviour for the two solutions with higher Mg and Sr.

0.01

0.00

Results and discussion -0.01 0

5 10 15 Mg/Ca (mmol mol-1)

0

10 20 Sr/Ca (mmol mol-1)

20

(Sr/Ca)meas/(Sr/Ca)true - 1

(b) 0.01

0.00

-0.01 30

Figure 4. Accuracy test. (a) Mg ⁄ Ca ratio relative deviation of measured from calculated values for mixed synthetic standard solutions with three different Sr content: 3 mmol mol - 1 (diamonds); 9 mmol mol - 1 (circles); 28 mmol mol - 1 (triangles). (b) Sr ⁄ Ca relative deviation of measured from calculated values for mixed synthetic standard solutions with three different Mg contents: 1 mmol mol - 1 (diamonds); 4 mmol mol - 1 (circles); 15 mmol mol - 1 (triangles).

ing by almost 1% when Mg ⁄ Ca increased from 1 to 15 mmol mol-1. The change of the measured Sr ⁄ Ca ratio with the amount of Mg present in the sample limited the expected accuracy to about 0.8%, but for low Mg content solutions, that number fell under 0.3% when Sr ⁄ Ca was above 9 mmol mol-1. In summary, the bias of Sr measurements depended on the Mg content: an overall value of 0.8% could be established for the whole range of Sr and Mg values considered, but for solutions with low Mg content and high Sr, the bias was below 0.3%. In this case, the uncertainty for the Sr concentration in the synthetic solutions, due to the Sr contamination in the Ca standard solution, could modify this picture to some degree. The presence of Sr in the solid CaCO3 used as the analytical calibrator was estimated as 12 ± 8 lg g-1 using a standard addition procedure. For the higher limit of this range, the three points with a value of Sr ⁄ Ca = 3 mmol mol-1 would shift downward by 0.3%, while the points at higher Sr contents

32

A number of carbonate materials were considered in this study: corals (Desmophylum dianthus and Porites lutea), a calcitic sclerosponge from the North Atlantic, a stalagmite from Borneo, a soil carbonate from Johnson Mesa, Utah, and a tufa from Red Butte Canyon, Utah. The Mg ⁄ Ca ratio varied from 3 mmol mol-1 for the coral to about 200 mmol mol-1 for the sponge. This higher value fell outside the application range of our method, while the Sr ⁄ Ca range was more modest, from 0.4 mmol mol-1 in the soil carbonate to 10 mmol mol-1 in the corals. In addition to these, five RMs were studied: NIST SRM 8544 (Friedman et al. 1982); UN AK (Institute of Mineral Raw Materials, Czech Republic); ECRM 752-1 (Bureau of Analysed Samples Ltd, UK); BAM RS3 and CM 1767 (Metallurgical Standardisation Research Institute, China). Figure 5 shows the results obtained for a stalagmite and a soil carbonate, which were sampled with a 25 lm resolution along the axis of growth using a mill (MicroMill, New Wave Research) with a tungsten carbide tool (H21.11.009 Brasseler; Savannah, Georgia, USA.). For the stalagmite, both Mg and Sr content displayed a pronounced decrease around 0.4 mm from the origin of sampling, down to values as small as one-third of the average. Although the Mg ⁄ Ca values around 30 mmol mol-1 obtained for the stalagmite were outside the range where accuracy was tested, the increase of the uncertainty due to the isotope dilution equation was only 0.1% for this amount of departure and we expect the accuracy to be also around 0.8%, as was shown in the previous section. The soil carbonate total variation for Sr ⁄ Ca, of only 0.2 mmol mol-1, illustrates the precision of the method in the low Sr region and also suggests a lower limit of application in the lmol mol-1 range. The variation found in the stalagmite (a factor of three) was also observed for the deep sea corals Mg ⁄ Ca variation around the central bands found in their septae (Gagnon et al. 2007). In order to compare meaningfully environmental signals from different laboratories, accurate methods over a broad range of Mg and Sr content are required. For those relying on external calibration curves, a matrix effect study is essential. This was shown for an ICP-OES method with

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

3.2

(a)

Mg/Ca (mmol mol-1)

Mg/Ca (mmol mol-1)

40

30

20

10

0.6

5

3.0

Dec 09

Mar 07

Dec 09

10.6

10.4

10.2

10.0 Jun 04 0

Mar 07

(b)

1.0

Sr/Ca (mmol mol-1)

10 Sr/Ca (mmol mol-1)

0.5 1 Distance (mm) (b)

3.1

2.9 Jun 04

0 0

(a)

0.2 0

0.5 1 Distance (mm)

Figure 6. (a) Mg ⁄ Ca and (b) Sr ⁄ Ca ratio long-term precision tests for a deep sea coral solution. Values

Figure 5. (a) Mg ⁄ Ca and (b) Sr ⁄ Ca ratios for a stalagmite (circles) and a soil carbonate (diamond) sampled from an arbitrary origin along the axis of growth.

sub-percent accuracy over Mg ⁄ Ca and Sr ⁄ Ca ranges similar to the one shown in the present work (de Villiers et al. 2002). The matrix effect for this method, due to the Ca selfabsorption and the increased background on Mg and Sr emission intensities, was as large as 15% and corrected using an intensity ratio calibration. In this way, it was possible to obtain precise results by utilising a number of RMs or standard solutions with the same Ca concentrations as the samples and a broad range of Mg and Sr concentrations. However, the impact of Mg content on the Sr ⁄ Ca ratio was not dealt with by these authors. In order to study the long term behaviour of the present ID-ICP-MS method and to monitor the trustworthiness for batches of samples run on different days, three solutions were prepared by dissolving  50 mg of deep sea coral (Desmophylum dianthus), calcitic sclerosponge and stalagmite from Borneo; each had a total Ca content of 4 mmol l-1. The long term reproducibility of the method was studied by measuring the consistency of the standard solutions over a period of at least 1 year. Figure 6 shows Mg ⁄ Ca and Sr ⁄ Ca values obtained for the deep sea coral solution over 5 years with two different instruments: sector

obtained with the sector field ICP-MS are shown by diamond symbols, and values obtained with the quadrupole ICP-MS are shown by triangles. Values obtained with the quadrupole ICP-MS for runs in which samples with very different Mg ⁄ Ca and Sr ⁄ Ca ratios were also measured are shown by circles.

field and quadrupole ICP-MS. Each point represents the average of at least six measurements of the coral solution diluted to a Ca concentration of about 1 mmol l-1 and the uncertainty shown is two standard deviations (2s). In addition, two points are shown associated with runs in which samples with very different Mg ⁄ Ca and Sr ⁄ Ca ratios were also included: NIST SRM 8544 (limestone) (Mg ⁄ Ca  18 mmol mol-1, Sr ⁄ Ca  0.2 mmol mol-1) and UN AK (Mg ⁄ Ca  2.4 mmol mol-1, Sr ⁄ Ca  3 mmol mol-1). Excluding these two points, the inter-run precision (2s) for both Mg ⁄ Ca and Sr ⁄ Ca was close to 0.4% for the sector field instrument values, while the quadrupole instrument values were 2.6% and 0.8%, respectively. Intra-run precision was as high as 4.6% for the sector field instrument, while the quadrupole instrument’s worst case was 1.4%. The averages obtained for the two sets corresponding to the different instruments (excluding the two points mentioned previously) agreed within 0.1% for Mg ⁄ Ca and 0.2% for Sr ⁄ Ca.

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

33

The effect of running samples possessing Mg ⁄ Ca and Sr ⁄ Ca ratios with marked differences in the same batch is illustrated by the two points shown in Figure 6 corresponding to the measurements of RMs NIST SRM 8544 and UN AK. In this case, in the face of a normal drift in the mass bias for the isotopic ratios for solution XG, similar to the value routinely obtained, a clear trend with run time was observed for the Mg ⁄ Ca and Sr ⁄ Ca replicate values. In fact, the increased intra-run precision compared to the case when only the deep sea coral solution was measured is associated with this trend. The effect is also exemplified in Figure 7, which displays the Sr ⁄ Ca reproducibility found for the sclerosponge reference sample using the sector field ICP-MS. Again, two different types of runs were considered: those for which only the sclerosponge reference sample was included, showed an inter-run precision of 0.4% (2s), comparable to that found for the deep sea coral solution. When both reference samples and unknown samples were present in the batch, the inter-run precision increased to 0.8%, and despite a deterioration of the intra-run precision, the average agreed well with the first case (just reference sample measured). We believe this effect could be related to the fact that the mass of test portion dissolved was less accurately known in the sclerosponge case, and consequently the isotopic ratios obtained after spiking differed markedly among samples. Regardless of the rinsing time used, long enough to reduce the background intensities to less than 0.2% of the values measured for the samples, the reproducibility experienced a significant deterioration. Probably, the intensity ratio values obtained for each reference sample depended on the ratio of the previous sample run, deteriorating the precision. This signals the importance of

Sr/Ca (mmol mol-1)

1.49

1.46

1.43

1.40 12 Jul 2004

26 Aug 2005

10 Oct 2006

Figure 7. Long-term precision for the sclerosponge reference sample. Filled-in diamonds, runs for which only the reference sample was run; open circles: runs for which both the reference sample and unknown samples were measured.

34

both intensity and ratio matching between spiked samples and solution XG in order to obtain the best possible precision. Consequently, in cases such as those shown for the stalagmite in Figure 5, with marked differences between subsamples, batches of samples with a narrow range of Mg ⁄ Ca and Sr ⁄ Ca values should be considered in order to achieve the best possible result. In addition to the memory effect on ratios affecting the precision of the method, it is important to assess the influence of inadequately characterised isotopic abundances and impurities of the RMs used. This is especially true for mass spectrometry methods using calcium isotopes as internal standards (Rosenthal et al. 1999, Marchitto 2006, Shen et al. 2007). In these cases, high purity materials were used as calibrators, although no mention was made of their isotopic composition. Russell et al. (1978) found a large fractionation for a commercial high-purity Ca RM. Differences in isotopic abundance between samples and calibrators can affect the outcome of the method in a significant way. For instance, a calibrator prepared from commercially available high purity metal calcium, probably purified by distillation, having a d(40Ca ⁄ 44Ca) of  10& (Russell et al. 1978) would have 40Ca ⁄ 43Ca, 40Ca ⁄ 46Ca and 40Ca ⁄ 48Ca ratios differing by about 1%, 1.5% and 2% respectively from a sample with natural abundance. Assuming that the Ca calibrator has natural abundance would then entail a systematic error. The same holds for commercial Mg RMs and for our case, it is conceivable that the accuracy test for Mg is mostly limited by the uncertainty in the abundance of the distilled material used to prepare the synthetic solution. Furthermore, the use of mixed synthetic standard solutions to assess the accuracy of a method is questionable unless the impurities are carefully evaluated. For example, as was mentioned in the previous section, the uncertainty in the content of Sr in the Ca reference sample used to prepare the synthetic solutions obscures the evaluation of the accuracy for the low Sr content range. Detrital material can also be important when assessing the method accuracy. A detailed study of the limestone ECRM 752-1 (Bureau of Analysed Samples Ltd, Newmal Hall, Middlesbrough, UK) made it possible to propose this material as a Mg ⁄ Ca RM (Greaves et al. 2005). Removal of aluminosilicate mineral phases by centrifugation was demonstrated as an essential step to improve the precision of the measurement. As was suggested by these authors, the existence of well characterised solid RMs together with a basic protocol for dissolution and separation of detrital minerals would be ideal in order to establish the accuracy of a given method and to facilitate the interpretation of

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

inter-laboratory studies. Reference materials ECRM 752-1, BAM RS3 and CM 1767, were also considered in a recent interlaboratory study (Greaves et al. 2008). In this case, centrifuging produced up to a 3% decrease in the Mg ⁄ Ca ratio average between laboratories (BAM RS3), with a smaller impact on the interlaboratory standard deviation. The Mg ⁄ Ca value for ECRM 752-1 (3.75 mmol mol-1) fell within the range found for foraminifera and corals; however the Sr ⁄ Ca value, 0.19 mmol mol-1, was an order of magnitude lower than that usually found in aragonitic carbonates. This was also the case for every CRM considered by Greaves et al. (2005), with the exception of UN AK with calculated values close to 2.8 mmol mol-1 for both Mg ⁄ Ca and Sr ⁄ Ca. This reference sample is also attractive due to its relatively low content of Al, suggesting a small contribution of detrital materials to its Mg ⁄ Ca and Sr ⁄ Ca values. Table 4 gives the Mg ⁄ Ca and Sr ⁄ Ca values for the non-centrifuged solutions prepared from NIST SRM 8544, UN AK, BAM RS3, ECRM 752-1 and CM 1767. Our results for NIST SRM 8544 can be compared with literature values obtained by an ICP-MS method using an in-house reference sample for external calibration (Sano et al. 2005). Their value for Sr ⁄ Ca (0.199 ± 0.006 mmol mol-1) is in good agreement with our value; however, for Mg ⁄ Ca, for which Sano et al. (2005) reported 13.04 ± 0.64 mmol mol-1, a disagreement is manifest. Mg ⁄ Ca results obtained for BAM RS3, ECRM 752-1 and CM 1767 agree within combined uncertainty with the mean values reported by Greaves et al. (2008) for non-centrifuged samples. These were: 3.82 ± 0.03 mmol mol-1 for ECRM 752-1; 0.78 ± 0.04 mmol mol-1 for BAM RS3 and 5.73 ± 0.06 mmol mol-1 for CM 1767. The uncertainties considered for these averages represent the interlaboratory standard deviation of the mean (SE, 2s). Greaves et al. (2008) only report Sr ⁄ Ca for CM 1767, and their value (1.506 ± 0.016 mmol mol-1) is significantly lower than that found by us. As an additional check on accuracy, filtered

and diluted seawater collected from the Bermuda Atlantic Time Series (BATS) at 1200 m was measured using the ID method on a multi-collector ICP-MS (Neptune). The highresolution mode was required to resolve 48Ca from a suspected 32S16O interference in these samples where column chemistry was not conducted. The value of 8.56 ± 0.001 mmol mol-1 from our analysis is about 1% lower than the 8.64 mmol mol-1 result reported by de Villiers (1999) for North Atlantic seawater from a similar depth but different location. We believe that the isotope dilution method presented here can be used to obtain precise and accurate high-resolution Mg ⁄ Ca and Sr ⁄ Ca ratio measurements in carbonates with environmental significance. Inter-instrument precision has been demonstrated for the method. It would also be a valuable tool for testing candidate materials to be used as RMs with an accuracy within 0.8%. Furthermore, a better characterisation of impurities and isotopic composition of calibrators could deepen the understanding of the systematic differences found between measured and calculated values for synthetic test solutions. In summary, we present a method with sub-percent accuracy in the determination of Mg ⁄ Ca and Sr ⁄ Ca ratio values over the broad range found in calcium carbonates found in nature. Although the Mg range was restricted to carbonates with relatively low Mg values for our present method, it could be extended easily by preparing a second mixed tracer with a higher 25Mg content. This new tracer would allow Mg ⁄ Ca values in the range 40–300 mmol mol-1 to be obtained.

Conclusions An isotope dilution method based on a Mg-43Ca-87Sr mixed tracer was developed for the determination of Mg and Sr in natural calcium carbonates containing concentrations of these elements of less than 25

Table 4. Mg ⁄ Ca and Sr ⁄ Ca values for reference materials (not centrifuged) Mg ⁄ Ca

NIST SRM 8544 AKb ECRM 752-1c BAM RS3d CM 1767e a c e

a

Sr ⁄ Ca

Average

2s

N

Average

2s

N

17.4 2.42 3.86 0.80 5.67

0.5 0.08 0.03 0.02 0.05

6 4 6 4 6

0.201 3.06 0.186 0.196 1.534

0.002 0.03 0.003 0.002 0.015

6 4 6 6 6

b NIST SRM 8544 (NBS 19 limestone, NIST, USA). UN AK (Institute of Mineral Raw Materials, Czech Republic). d ECRM 752-1 (Bureau of Analysed Samples Ltd, UK). BAM RS3 (Federal Institute for Materials Research and Testing, Germany). CM 1767 (Metallurgical Standardisation Research Institute, China).

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

35

30 mmol mol-1. Microgram amounts of carbonate could be dissolved, volumetrically spiked with a mixed tracer and run in a sequence, together with a solution of known isotopic ratio used to correct for mass bias. A throughput of fifty samples per day with a precision of 0.8% (2s) or better was possible. Two types of ICP-MS instrument, a sector field and a quadrupole, gave consistent results in a long term study, within 0.14% for Mg ⁄ Ca and 0.11% for Sr ⁄ Ca, although the inter-run precision found for the sector field instrument was significantly better than that obtained with the quadrupole for Mg measurement. For both instruments, the precision deteriorated by a factor of two when samples with markedly different Mg and Sr content were run sequentially. Bias was tested for the sector field instrument using synthetic solutions prepared by mixing gravimetrically selected reference samples. An overall agreement of 0.8% between measured and calculated values was found, although systematic trends for these differences were found also. Mg ⁄ Ca results were on average 0.5% lower than the calculated values regardless of the Sr content. On the contrary, the Sr ⁄ Ca values depended on both the amount of Sr and Mg present. Synthetic solutions with a Sr ⁄ Ca ratio > 9 mmol mol-1 and a Mg ⁄ Ca ratio < 4 mmol mol-1 could be balanced within 0.3% of their calculated values, while for lower Sr contents or higher Mg contents the accuracy deteriorated to 0.8%. A more precise characterisation of the impurities and the isotopic composition of materials to be used as calibrators should be accomplished in order to gain a better understanding of the method at the sub-percent level. The method is suitable to obtain Mg ⁄ Ca and Sr ⁄ Ca ratios with an accuracy of 0.8% for sets of samples differing by a factor of three in their Mg or Sr content. Mg ⁄ Ca and Sr ⁄ Ca ratios for five RMs were measured: Mg ⁄ Ca values were found to be in good agreement with previously reported results, while the agreement for Sr ⁄ Ca was marginal.

Acknowledgements This work was supported by the National Science Foundation, Grants OCE-0096373 and OCE-0502642, and the Comer Science and Education Foundation, Grant CM113. We acknowledge two anonymous reviewers for comments and suggestions than improved this manuscript.

References Andreasen D.H., Sosdian S., Perron-Cashman S., Lear C.H., DeGaridel-Thoron T., Field P. and Rosenthal Y. (2006) Fidelity of radially viewed ICP-OES and magnetic-sector ICP-MS measurement of Mg ⁄ Ca and Sr ⁄ Ca ratios in

36

marine biogenic carbonates: are they trustworthy together? Geochemistry Geophysics Geosystems, 7, Q10P18. doi:1029/2005GC001124. Andren H., Rodushkin I., Stenberg A., Malinovsky D. and Baxter D.C. (2004) Sources of mass bias and isotope ratio variation in multicollector ICP-MS: optimization of instrumental parameters based on experimental observations. Journal of Analytical Atomic Spectrometry, 19, 1217–1224. Beauchemin D. (2006) Inductively coupled plasma mass spectrometry. Analytical Chemistry, 78, 4111–4135. Beck J.W., Edwards R.L., Ito E., Taylor F.W., Recy J., Rougerie F., Joannot P. and Henin C. (1992) Sea-surface temperature from coral skeletal strontium calcium ratios. Science, 257, 644–647. De La Rocha C.L. and DePaolo D.J. (2000) Isotopic evidence for variations in the marine calcium cycle over the Cenozoic. Science, 289, 1176–1178. Esat T.M., Spear R.H. and Taylor S.R. (1986) Isotope anomalies induced in laboratory distillation. Nature, 319, 576–578. Friedman I., O’Neil J. and Cebula G. (1982) Two new carbonate stable-isotope standards. Geostandards Newsletter, 6, 11–12. Gagnon A.C., Adkins J.F., Fernandez D.P. and Robinson L.F. (2007) Sr ⁄ Ca and Mg ⁄ Ca vital effects correlated with skeletal architecture in a scleractinian deep-sea coral and the role of Rayleigh fractionation. Earth and Planetary Science Letters, 261, 280–295. Greaves M., Barker S., Daunt C. and Elderfield H. (2005) Accuracy, standardization, and interlaboratory calibration standards for foraminiferal Mg ⁄ Ca thermometry. Geochemistry Geophysics Geosystems, 6, Q02D13. doi:10.1029/2004GC000790. Greaves M., Caillon N., Rebaubier H., Bartoli G., Bohaty S., Cacho I., Clarke L., Cooper M., Daunt C., Delaney M., deMenocal P., Dutton A., Eggins S., Elderfield H., Garbe-Schoenberg D., Goddard E., Green D., Groeneveld J., Hastings D., Hathorne E., Kimoto K., Klinkhammer G., Labeyrie L., Lea D.W., Marchitto T., Martinez-Boti M.A., Mortyn P.G., Ni Y., Nuernberg D., Paradis G., Pena L., Quinn T., Rosenthal Y., Russell A., Sagawa T., Sosdian S., Stott L., Tachikawa K., Tappa E., Thunell R. and Wilson P.A. (2008) Interlaboratory comparison study of calibration standards for foraminiferal Mg ⁄ Ca thermometry. Geochemistry Geophysics Geosystems, 9, Q08010. doi:10.1029/2008gc001974. Hearn R., Evans P. and Sargent M. (2005) Development of a high ratio isotope dilution mass spectrometry calibration technique for chemical metrology. Journal of Analytical Atomic Spectrometry, 20, 1019–1023.

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

references Klingbeil P., Vogl J., Pritzkow W., Riebe G. and Muller J. (2001) Comparative studies on the certification of reference materials by ICP-MS and TIMS using isotope dilution procedures. Analytical Chemistry, 73, 1881–1888. Lea D.W. and Martin P.A. (1996) A rapid mass spectrometric method for the simultaneous analysis of barium, cadmium, and strontium in foraminifera shells. Geochimica et Cosmochimica Acta, 60, 3143–3149. Lear C.H., Rosenthal Y. and Slowey N. (2002) Benthic foraminiferal Mg ⁄ Ca-paleothermometry: a revised core-top calibration. Geochimica et Cosmochimica Acta, 66, 3375–3387. LeCornec F. and Correge T. (1997) Determination of uranium to calcium and strontium to calcium ratios in corals by inductively coupled plasma mass spectrometry. Journal of Analytical Atomic Spectrometry, 12, 969–973. Lehn S.A., Warner K.A., Huang M. and Hieftje G.M. (2003) Effect of sample matrix on the fundamental properties of the inductively coupled plasma. Spectrochimica Acta Part B, 58, 1785–1806. Marchitto T.M. (2006) Precise multielemental ratios in small foraminiferal samples determined by sector field ICP-MS. Geochemistry Geophysics Geosystems, 7, Q05P13. doi:10.1029/2005GC0001018. Nurnberg D., Bijma J. and Hemleben C. (1996) Assessing the reliability of magnesium in foraminiferal calcite as a proxy for water mass temperatures (vol 60, pg 803, 1995). Geochimica et Cosmochimica Acta, 60, 2483. Rosenthal Y., Field M.P. and Sherrell R.M. (1999) Precise determination of element ⁄ calcium ratios in calcareous samples using sector field inductively coupled plasma mass spectrometry. Analytical Chemistry, 71, 3248–3253. Rosenthal Y., Perron-Cashman S., Lear C.H., Bard E., Barker S., Billups K., Bryan M., Delaney M.L., deMenocal P.B., Dwyer G.S., Elderfield H., German C.R., Greaves M., Lea D.W., Marchitto T.M., Pak D.K., Paradis G.L., Russell A.D., Schneider R.R., Scheiderich K., Stott L., Tachikawa K., Tappa E., Thunell R., Wara M., Weldeab S. and Wilson P.A. (2004) Interlaboratory comparison study of Mg ⁄ Ca and Sr ⁄ Ca measurements in planktonic foraminifera for paleoceanographic research. Geochemistry Geophysics Geosystems, 5, Q04D09. doi:10.1029/2003GC000650. Russell W.A., Papanastassiou D.A. and Tombrello T.A. (1978) Ca isotope fractionation on Earth and other solar-system materials. Geochimica et Cosmochimica Acta, 42, 1075–1090.

Sano Y., Shirai K., Takahata N., Hirata T. and Sturchio N.C. (2005) Nano-SIMS analysis of Mg, Sr, Ba and U in natural calcium carbonate. Analytical Sciences, 21, 1091–1097. Schrag D.P. (1999) Rapid analysis of high-precision Sr ⁄ Ca ratios in corals and other marine carbonates. Paleoceanography, 14, 97–102. Shen C.C., Chiu H.Y., Chiang H.W., Chu M.F., Wei K.Y., Steinke S., Chen M.T., Lin Y.S. and Lo L. (2007) High precision measurements of Mg ⁄ Ca and Sr ⁄ Ca ratios in carbonates by cold plasma inductively coupled plasma quadrupole mass spectrometry. Chemical Geology, 236, 339–349. Smith S.V., Buddemeier R.W., Redalje R.C. and Houck J.E. (1979) Strontium-calcium thermometry in coral skeletons. Science, 204, 404–407. Steuber T. and Buhl D. (2006) Calcium-isotope fractionation in selected modern and ancient marine carbonates. Geochimica et Cosmochimica Acta, 70, 5507–5521. Sturgeon R.E., Willie S.N., Yang L., Greenberg R., Spatz R.O., Chen Z., Scriver C., Clancy V., Lam J.W. and Thorrold S. (2005) Certification of a fish otolith reference material in support of quality assurance for trace element analysis. Journal of Analytical Atomic Spectrometry, 20, 1067–1071. de Villiers S. (1999) Seawater strontium and Sr ⁄ Ca variability in the Atlantic and Pacific oceans. Earth and Planetary Science Letters, 171, 623–634. de Villiers S., Greaves M. and Elderfield H. (2002) An intensity ratio calibration method for the accurate determination of Mg ⁄ Ca and Sr ⁄ Ca of marine carbonates by ICP-AES. Geochemistry Geophysics Geosystems, 3, doi:10.1029/2001GC000169. Wara M.W., Anderson L.D., Schellenberg S.A., Franks R., Ravelo A.C. and Delaney M.L. (2003) Application of a radially viewed inductively coupled plasma-optical emission spectrophotometer to simultaneous measurement of Mg ⁄ Ca, Sr ⁄ Ca, and Mn ⁄ Ca ratios in marine biogenic carbonates. Geochemistry Geophysics Geosystems, 4, doi:10.1029/2003GC000525. Wieser M.E., Buhl D., Bouman C. and Schwieters J. (2004) High precision calcium isotope ratio measurements using a magnetic sector multiple collector inductively coupled plasma mass spectrometer. Journal of Analytical Atomic Spectrometry, 19, 844–851. Yu J.M., Day J., Greaves M. and Elderfield H. (2005) Determination of multiple element ⁄ calcium ratios in foraminiferal calcite by quadrupole ICP-MS. Geochemistry Geophysics Geosystems, 6, Q08P01. doi:10.1029/2005GC000964.

ª 2010 The Authors. Geostandards and Geoanalytical Research ª 2010 International Association of Geoanalysts

37