CALCIUM ISOTOPE ANALYSIS BY MASS SPECTROMETRY

Transcription

1 CALCIUM ISOTOPE ANALYSIS BY MASS SPECTROMETRY Sergei F. Boulyga* Safeguards Analytical Laboratory, Department of Nuclear Sciences and Applications, International Atomic Energy Agency, Wagramer Strasse 5, 1400 Vienna, Austria Received 15 July 2008; received (revised) 3 April 2009; accepted 3 April 2009 Published online 23 June 2009 at wileyonlinelibrary.com. DOI /mas The variations in the isotopic composition of calcium caused by fractionation in heterogeneous systems and by nuclear reactions can provide insight into numerous biological, geological, and cosmic processes, and therefore isotopic analysis finds a wide spectrum of applications in cosmo- and geochemistry, paleoclimatic, nutritional, and biomedical studies. The measurement of calcium isotopic abundances in natural samples has challenged the analysts for more than three decades. Practically all Ca isotopes suffer from significant isobaric interferences, whereas low-abundant isotopes can be particularly affected by neighboring major isotopes. The extent of natural variations of stable isotopes appears to be relatively limited, and highly precise techniques are required to resolve isotopic effects. Isotope fractionation during sample preparation and measurements and instrumental mass bias can significantly exceed small isotope abundance variations in samples, which have to be investigated. Not surprisingly, a TIMS procedure developed by Russell et al. (Russell et al., Geochim Cosmochim Acta 42: ) for Ca isotope measurements was considered as revolutionary for isotopic measurements in general, and that approach is used nowadays (with small modifications) for practically all isotopic systems and with different mass spectrometric techniques. Nevertheless, despite several decades of calcium research and corresponding development of mass spectrometers, the available precision and accuracy is still not always sufficient to achieve the challenging goals. The present article discusses figures of merits of presently used analytical methods and instrumentation, and attempts to critically assess their limitations. In Sections 2 and 3, mass spectrometric methods applied to precise stable isotope analysis and to the determination of 41 Ca are described. Section 4 contains a short summary of selected applications, and includes tracer experiments and the potential use of biological isotope fractionation in medical studies, paleoclimatic and paleoceanographic, and other terrestrial as well as extraterrestrial investigations. # 2009 Wiley Periodicals, Inc., Mass Spec Rev 29: , 2010 Keywords: calcium; stable isotopes; 41 Ca; fractionation; mass spectrometry *Correspondence to: Sergei F. Boulyga, Safeguards Analytical Laboratory, Department of Nuclear Sciences and Applications, International Atomic Energy Agency, Wagramer Strasse 5, 1400 Vienna, Austria. s.bulyha@iaea.org I. INTRODUCTION Calcium is one of the most-abundant elements in the earth crust with an abundance of more than 3%, and it plays essential role in geological and biological processes. It has five naturally stable isotopes: 40 Ca (96.941%), 42 Ca (0.647%), 43 Ca (0.135%), 44 Ca (2.086%), and 46 Ca (0.004%), and a long-lived isotope 48 Ca (0.187%) with a half-life of years. Calcium isotopic composition was studied first by Aston (1935) and later by Nier (1938), who discovered minor isotopes 46 Ca and 48 Ca when analyzing metallic Ca. The radioactive isotope of calcium 41 Ca has a half-life of years. The initial abundance of 41 Ca ( 41 Ca/ 40 Ca) in the early solar system is assumed (Alexander, Boss, & Carlson, 2001). This isotope is also produced by thermal neutrons via the reaction 40 Ca(n,g) 41 Ca, and is present in the terrestrial environment due to the reactions between the cosmic rays and the earth s surface (Johnson et al., 1991). According to Raisbeck and Yiou (1979) the production rate of 41 Ca is approximately 800 atoms/g Ca per year, and the abundance ratio of 41 Ca/Ca in the environment is now <10 14 (Fink, Klein, & Middleton, 1990). Neutron capture is also important for 41 Ca production in stony meteorites. Since Holmes (1932) first suggested the application of stable-isotope measurements in geosciences, calcium isotopic variations in nature attracted increased interest of research community. In principle, isotopic variations of calcium can evolve from different processes. For example, beta-decay of 40 K (half-life: years, modes of decay: beta-decay 40 K! 40 Ca with a branch ratio of 89.28% and electron capture 40 K! 40 Ar with a branch ratio of 10.72%) can increase the relative abundance of 40 Ca in potassium-containing minerals (Ingram et al., 1950). Altered calcium isotopic composition in meteorites as a result of nuclear reactions was also reported (Stauffer & Honda, 1962). On the other hand, isotopic fractionations of stable elements were assumed to take place under natural conditions in the heterogeneous systems that include natural ion exchangers, such as zeolites (Cook, 1943). Calcium isotope fractionation in nature can be basically caused by chemical or physical processes, like exchange reactions and diffusion, whereas such fractionation should be especially pronounced between 40 Ca and 48 Ca due to the large relative difference in the atomic masses of these isotopes. Natural variations of the isotopic composition of calcium of rocks, minerals, and biological samples represent a potential interest for geochronology, climate change studies, archaeometry, and other scientific disciplines. Calcium is also a biologically essential element and represents a common constituent of the mineralized tissues as well as soft tissue Mass Spectrometry Reviews, 2010, 29, # 2009 by Wiley Periodicals, Inc.

2 & BOULYGA and body fluids. Isotope analysis with isotope tracers, among them stable minor isotopes and the long-lived 41 Ca, finds application in biochemistry and clinical medicine to study calcium metabolism. A variety of analytical applications that involve calcium isotopic measurements specify requirements for the selection of appropriate detection methods in terms of their selectivity, precision, and accuracy. Because the natural variation of 40 Ca abundance ranges between % and % (Rosman and Taylor, 1997), highly precise methods are required to resolve isotopic differences in natural samples. For such purposes, thermal ionization mass spectrometry (TIMS) or inductively coupled plasma mass spectrometry with multiple ion collectors (MC-ICP-MS) can be used. If enriched stable isotopes are applied in tracer studies or to determine calcium concentration via isotope dilution technique, then the difference in isotopic ratios is typically larger and the requirements on the analytical precision of isotopic measurements are less strict, so that quadrupole-based ICP-MS, for instance, can be applied. Isotopic ratios 41 Ca/ 40 Ca in samples are usually extremely low, even if 41 Ca is used as a tracer. Therefore, peak tailing of 40 Ca þ in the mass spectrometer as well as isobaric interferences represent the main limitations to determine 41 Ca. To solve these problems, analytical methods that possess high elemental and isotopic selectivity are required, such as accelerator mass spectrometry (AMS) or resonance-ionization mass spectrometry (RIMS). The next sections of this article discuss figures of merit and limitations of particular mass spectrometric methods for calcium isotopic analyses. II. ANALYTICAL METHODS FOR PRECISE ISOTOPE ANALYSES OF CALCIUM Calcium isotope measurements were limited for decades by the precision of the available analytical methods that was poor compared to the range of isotopic fractionation typical for natural materials (Heumann & Luecke, 1973). Thus, most of the studies of meteorites, carbonate sediments, calciumsilicate minerals, and organic substances that were performed at that time yielded no variations in calcium isotopic composition within the error of measurements. Nevertheless, some researchers (see, e.g., Corless, 1968, Coleman, 1971) reported small but significant deviations in calcium isotope composition in natural materials. Coleman (1971) achieved precision equivalent to 0.2% in the 40 Ca/ 44 Ca ratio generally (0.03% in the best case); such a precision was considered sufficient to determine K Ca ages of mica samples. Initially, two different analytical approaches were used for isotopic analysis of calcium: mass spectrometry and activation analysis. Activation analyses were performed with several nuclear reactions, such as 48 Ca(n,g) 49 Ca, 48 Ca(d,p) 49 Ca and 40 Ca(d,a) 38 K, 44 Ca(n,a) 41 Ar and 40 Ca(n,a) 37 Ar. Although low measurement uncertainties (mainly based on counting statistics) were reported in those studies, the results were often disputed as possibly affected by non-controlled measurement errors. In general, the overall experimental error lower than 1% was difficult to achieve with activation techniques, and, therefore, these methods were inferior to mass spectrometry. On the other hand, mass spectrometric data were subject to instrumental fractionation effects, which might appear during the sample preparation and during the measurement. Because the investigations of isotope separation in Li, K, and N during elution (Taylor & Urey, 1938), it was known that the selective separation ability of ion exchangers also had an effect on different isotopes of the same element. Therefore, Heumann (1972) systematically studied with a chelating resin isotopic fractionation of calcium in heterogeneous systems. An enrichment of heavier isotope was found in solution with Dowex A1-loaded columns with a length of 67 cm. Separation factors of and were reported for 44 Ca/ 40 Ca and 48 Ca/ 40 Ca ratios, respectively. With a 12 cm-long column filled with Dowex 50-X12 resin, separation factors of and were found for 44 Ca/ 40 Ca and 48 Ca/ 40 Ca ratios, respectively, that depended on the concentration of hydrochloric acid used as the eluent (Heumann & Lieser, 1973). Later on, significant fractionation of calcium isotopes during the elution through short ion-exchange columns packed with Dowex 50W-X8 resin (the resin height in the column was 17 cm) was also observed by Russell and Papanastassiou (1978). Heumann and Lieser (1973) analyzed a range of natural carbonates and sulfates and found significant isotopic variation of Ca in only one sample ( Gipsrose ), whereas d 48 Ca was approximately two-times larger than d 44 Ca. That fractionation was attributed to potential ion-exchange or different diffusion rates of Ca isotopes in soil layers. Russell, Papanastassiou, and Tombrello (1978) performed a very detailed study of calcium isotopic composition in terrestrial and extra-terrestrial samples. It was a great importance for that work to identify ion exchangers as a source of isotope fractionation that might have been introduced during the sample preparation (Russell & Papanastassiou, 1978). Therefore, for the control of instrumental fractionations a double-isotope spike was added to samples prior to its elution through the column; it was assumed that any later fractionation will affect identically the double spike and the sample. In addition, complete elution was a precondition for accurate analysis. A mass-dependent law to correct an instrumental mass fractionation was included in the data-reduction procedure. Russell, Papanastassiou, and Tombrello (1978) demonstrated clearly resolvable differences in the 40 Ca/ 44 Ca ratios by up to 2.5% in meteorites, lunar, and terrestrial samples, and by up to 13% for industrially purified Ca. A. Thermal Ionization Mass Spectrometry (TIMS) Although calcium has a relatively low ionization potential of 6.11 ev, its thermal ionization and isotope analysis is not straightforward. In summary, the main analytical problems are stipulated by very different abundances of isotopes, wide beam dispersion due to large relative mass difference between the heaviest and the lightest isotope, isotopic mass fractionation, isobaric interferences, risk of contamination because of the ubiquitous character of this elements, and impurities. Particular care should always be taken in the handling of the samples to avoid contaminations. The low abundance of most Ca isotopes, except 40 Ca, requires, on one hand, the production of an ion beam with high intensity and stability, and, on the other hand, a 686 Mass Spectrometry Reviews DOI /mas

3 CALCIUM ISOTOPE ANALYSIS & wide dynamic range, low noise, and good linearity of ion detectors. 1. Sample Preparation and Loading Sample preparation and loading procedures control overall detection efficiency and the presence of spectral and non-spectral interferences in TIMS; and in general, these procedures are important issues that influence the accuracy of isotopic analysis. To achieve high intensity and good analytical precision, different filament arrangements (single, double, and triple filament) and different chemical compounds of Ca (e.g., Ca(NO 3 ) 2, CaCO 3, CaI 2, and CaCl 2 ) were used. An earlier review of Ca isotopic studies with TIMS was made by Platzner (1997). Very careful sample purification and loading procedures as well as isotope analysis techniques were developed by several groups that allowed reproducibility of 0.1% per mass unit (Stahl & Wendt, 1968; Heumann, Schiefer, & Spiess, 1982, Moore & Machlan, 1972). It was reported that only extremely pure samples yielded sufficiently intense and stable ion beams, whereas even the smallest impurities (e.g., iron or aluminum) reduced the ion yield of Ca by up to two orders of magnitude. Isobaric interferences exist for several Ca isotopes; 46 Ca and 48 Ca overlap with titanium isotopes, and 40 Ca overlaps with 40 Ar and 40 K. Potassium represents a problem for any mass spectrometric technique due to its presence in practically all samples, and because of a risk of contamination during sample preparation. Because of the low ionization potential of potassium (4.34 ev), 40 K is readily available in the spectra obtained by TIMS. To reduce this interference, Ca was purified by repeated precipitation as CaCO 3 (Heumann & Lieser, 1973) or by cation exchange (Moore & Machlan, 1972). Precipitation was faster, and allowed nearly complete purification from potassium, but some other potentially interfering elements could precipitate as carbonates. Cation-exchange separation was more time-consuming, but allowed the separation of most elements. Furthermore, Moore and Machlan (1972) preheated sample filaments to remove 40 K as well as other unknown interference at masses of Ca isotopes. Whenever different species have different evaporation and ionization temperatures, optimization of the filament-heating program usually helps to reduce isobaric interferences in TIMS. Thus, to separate K from Ca Fletcher et al. (1997) proposed to heat first only center (evaporation) filament in a triple-filament arrangement. That approach had the advantage of dissipating most of the residual K before Ca beams reach full strength. The temperature gap between different species can also be increased with suitable activators. The potential presence of any remaining interferences can be controlled by measuring another isotope of the interfering element; for example, 41 K þ for 40 K þ control (Heumann & Lieser, 1973). Russell, Papanastassiou, and Tombrello (1978) used highpurity, zone-refined tantalum ribbon and outgassing of the filament before sample loading to reduce K background ( 39 K signal was lower than A) during the Ca analysis (T 1,4008C). For TIMS measurements, 5 10 mg Ca were loaded either as the chloride or nitrate on a single, oxidized, V-shape tantalum filament. Calcium nitrate solutions are often loaded on zone-refined Re in a double-filament arrangement. Reaction of calcium nitrate with Re and oxygen during preheating forms high-mass calcium rhenate on the filament surface. In the ion source, calcium rhenate compounds evaporate and produce Ca þ ions by thermal ionization on the second filament to produce a constant ion beam and a lower mass fractionation in the subsequent TIMS analysis (see, e.g., Farkas et al., 2007a). Evaporation of calcium in the form of high-mass molecules reduces mass fractionation. Heuser et al. (2002) adapted a method previously used by Birck (1986) for Sr measurements. In that work, zone-refined rhenium-ribbon single filaments were used in combination with ata 2 O 5 -activator and the so called sandwich-technique. In a sequence, 0.5 ml Ta 2 O 5 activator solution, a 1 2 ml sample solution with concentrations of ng Ca/mL, and another 0.5 ml of the activator were loaded on the filament, and each solution was heated to near dryness before loading next solution. The use of an activator solution stabilized the signal intensity to produce a precise measurement of the Ca isotopic composition of the sample/spike mixture. Zhu and Macdougall (1998) and Gopalan, Macdougall, and Macisaac (2006) used a similar loading procedure with tungsten filaments (Table 1). Böhm et al. (2006) used TaCl 5 activator instead of Ta 2 O 5, but did not report if it possessed any advantages for TIMS analyses. Marshall and DePaolo (1982) loaded ca. 15 mg Ca onto a zone-refined single tantalum filament in 2 N HCl; the solution was dried and oxidized in air in the presence of phosphoric acid. In recent work (Fantle & DePaolo, 2007), that procedure was modified, so that natural carbonate samples, such as a deep-sea carbonate ooze, were analyzed with a minimal chemical preparation. Carbonate samples were dissolved in dilute acetic acid, and the solution was centrifuged. The dissolved spiked samples were loaded directly onto a tantalum filament with H 3 PO 4 for mass spectrometric measurements. Avoiding timeconsuming sample purification procedure is an attractive choice for routine calcium isotope analyses; the necessary precondition is, however, the absence of impurities, which could produce interferences. When analyzing dissolved carbonates, Fantle and DePaolo (2007) did not detect any significant amounts of 40 Kor 48 Ti, which could interfere with Ca isotopes. Heuser et al. (2002) paid particular attention to the monitoring and the elimination of any potential isobaric interferences in TIMS analysis of Ca isotopes, such as, for instance, the molecular ion MgF þ. By monitoring 26 Mg 19 F þ ions at m/z 45, detailed studies indicated no significant isobaric contributions from 24 Mg 19 F þ and 25 Mg 19 F þ on 43 Ca þ and 44 Ca þ, respectively. Interestingly, although no MgF þ has been observed; some other fluorides like CaF þ (m/z 59 67), SrF þ (m/z ), and BaF þ (m/z ) were detected. However, CaF þ ions decreased in several minutes to background levels. Interferences from Mg- and Al-oxides on mass 40u ( 24 Mg 16 O þ ) and 43u ( 27 Al 16 O þ ) were considered to be negligible because Mg/Ca and Al/Ca ratios were generally low in the analyzed samples. Interferences from 48 Ti þ and double-charged 84 Sr 2þ, 86 Sr 2þ, and 88 Sr 2þ (m/z 42, 43, and 44) were not detected. Fletcher et al. (1997) developed a procedure for routine high-precision isotopic analyses of the K Ca isotopic system. Loading samples as KI and CaI 2 on triple-filament arrays resulted in a reduced mass fractionation. The advantages of using iodides Mass Spectrometry Reviews DOI /mas 687

4 & BOULYGA TABLE 1. Measurement procedures and precision of calcium isotope ratio analysis by TIMS a Usually reported as long-term external reproducibility (2s). b Internal reproducibility for 40 Ca/ 48 Ca and 44 Ca/ 48 Ca ratios of a multi-collector measurements. c Typical in-run precision for d 44 Ca (2s). d Average precision for 44 Ca/ 40 Ca ratio in SRM 915a during a session (n ¼ 4). External precision is given as two times the standard error of the mean (2s/n 0.5 ) determined by sample repeat measurements. 688 Mass Spectrometry Reviews DOI /mas

5 CALCIUM ISOTOPE ANALYSIS & consisted in their relatively high mass and low boiling points and dissociation energies. To correct for isotope fractionation, which can occur during analyte extraction and subsequent measurement, a double spike is usually added prior to chemical separation (Russell, Papanastassiou, & Tombrello, 1978; Marshall & DePaolo, 1982). A complete chemical equilibrium of spike and sample analyte must be guaranteed for isotope dilution analysis. If pure Ca salts are measured, the mixture of spike and sample solutions can be evaporated to dryness and then redissolved again in ultrapure acid solution to achieve chemical equilibrium (Heuser et al., 2002). 2. Mass Fractionation in TIMS In principle, mass spectrometric data are always affected by fractionation. Such effects are particularly pronounced for Ca because of the large relative mass difference of isotopes. A comprehensive discussion of time- and mass-dependent evaporation-related mass-fractionation effects in TIMS can be found in work by Habfast (1998). To decrease the differences in evaporation rates for heavier and lighter isotopes, which result in time-dependent isotopic fractionation during evaporation, compounds of calcium with elements of higher masses are used so that the relative mass difference between evaporating species is reduced. Thus, Fletcher et al. (1997) observed experimentally that isotopic ratios recorded from loaded iodides varied less during the TIMS analysis than the ratios seen for other halides. However, that approach did not completely remove fractionation. In an earlier work, Heumann and Kubassek (1974) found that measured calcium isotope ratios in a double-filament thermal ion source depended on: temperature of the ionization filament; mass discrimination effect during the evaporation of the sample; salt used as sample; size of the crystals in a calcium carbonate sample. The effect of these factors on the measurement accuracy can be reduced but not eliminated with standardized sample preparation and loading techniques. A more appropriate correction can be done with internal standardization when another isotopic ratio is known. For instance, to monitor the radiogenic enrichments of 40 Ca in various geological materials, Marshall and DePaolo (1982) chose to use 40 Ca/ 42 Ca as the target ratio and 42 Ca/ 44 Ca for the mass-discrimination correction. However, if natural fractionation of Ca must be analyzed, then no isotopic ratio for internal standardization is available because all calcium isotopes are subject to fractionation. In this case, a mass-bias correction can be done by spiking the sample with two isotopes and assuming that the isotope ratio of the double spike is known. Hirt and Epstein (1964) applied a double-spike method by adding to the samples known amounts of enriched isotope spikes of 43 Ca and 46 Ca. Although enriched spikes always contain other stable isotopes and, therefore, the double-spike method involves complicated deconvolution of isotopic patterns, that approach attracted particular interest after its potential was recognized to correct the isotopic fractionation of calcium during the sample preparation and measurement. Russell, Papanastassiou, and Tombrello (1978) provided a comprehensive investigation of all significant factors that affected the accuracy of calcium isotopic measurements with TIMS, and developed a completely traceable procedure to control those factors, whereas the use of a double-spike internal standard was considered as the only reliable way to correct for instrumental mass fractionation. A double spike was prepared by a precise weighing of two isotopically enriched salts (singleisotope spikes, 42 Ca and 48 Ca), and dissolving them in weighed amount of reagent. Russell, Papanastassiou, and Tombrello (1978) estimated that the gravimetrically determined 42 Ca/ 48 Ca ratio in the spike had an uncertainty of approximately 1% with account to hygroscopic properties of the Ca salts. That uncertainty resulted in a possible constant bias of up to ca. 2% per unit mass difference when the double spike was used to measure the isotopic composition of a sample. The authors recognized that the estimated uncertainty did not consider any contribution by the uncertainty in the isotopic enrichment or stoichiometry of the salts from which the spikes were prepared. However, these uncertainties affected the mass spectrometric data equally when they were acquired with the same doublespike technique, so that relative sample-to-sample differences remained unaffected. Therefore, Russell, Papanastassiou, and Tombrello (1978) resolved differences in the 40 Ca/ 44 Ca ratio in natural samples at a level of 0.5%, which corresponded to 0.12% per mass unit. The preconditions to obtain such a good precision were (a) use the double-spike technique; (b) use an appropriate model to correct for instrumental mass fractionation; and (c) eliminate fractionation effects from differential elution of isotopes from ion-exchange resins. Russell, Papanastassiou, and Tombrello (1978) performed a sensitive test for particular mass fractionation laws, and demonstrated that neither a linear law nor a simple power law adequately described the instrumental mass-fractionation effects in Ca. The data were empirically fit best with the exponential law : R M ik ¼ m p i RC ik ð1þ where R M ik and RC ik are measured and corrected ratios of isotopes i and k, respectively; and p is the fractionation factor that represents a function of time, but is independent of mass. The time dependence of p is assumed to be the same for all ratios. That law had a relatively simple form that provided ease in computations and resulted in sufficiently time-independent ratios. Jungck, Shimamura, and Lugmair (1984) reported a small residual dependence of the corrected ratios on the varying fractionation during an isotopic analysis with the exponential law. Hart and Zindler (1989) further examined fractionation behavior with Ca. A four-isotope equal-atom mixture of 40 Ca, 42 Ca, 44 Ca, and 48 Ca was prepared, and was analyzed mass spectrometrically with a double-filament ionization mode. Of the various isotope fractionation laws proposed, the exponential law provided the best fit to the data; however, deviations of up to 0.01%/amu over the total range of fractionation were observed. m k Mass Spectrometry Reviews DOI /mas 689

6 & BOULYGA Thus, the observed deviations of the measured isotope ratio from the true ratio can only be roughly modeled with idealized fractionation laws. To more realistically describe the processes in the ion source, the models must involve additional, numerous parameters. Hart and Zindler (1989) suggested that these factors would probably provide the ultimate limit of precision for multi-collector analysis for all isotopic systems. According to Habfast (1998), the various observed deviations of the isotope ratio from the true ratio were only approximately modeled by idealized fractionation laws. Although the exponential law was the best approximation for the Rayleigh distillation law (see for detail Habfast, 1998), a small residual deviation required an empirical second order correction. Another limiting factor was represented by timeindependent (i.e., static ) mass discriminations, which were due to the mass spectrometer hardware. Thus, Habfast (1998) stated that the fundamental limit to the ability to measure the true isotope ratio in a solid sample by TIMS was determined by an incomplete knowledge of the evaporation process, insufficient reproducibility of the evaporation process, and an inability to measure with sufficient accuracy the parameters that define static discriminations. 3. Instrumental Effects on Precision and Accuracy Fletcher et al. (1997) estimated the statistical limits (expressed as 95% confidence limits for single analyses) for precision and reproducibility of the mass spectrometer-derived Ca isotopic data. In that work, several factors that affected precision were considered, such as the range of calcium isotopic abundances, weighing limitations to prepare gravimetric standards, statistical noise of the ion beam, nonlinearity of the beam-dispersion patterns, and deconvolution procedures. Isotopic tracers were assumed to be perfectly calibrated, and minor effects, such as variations in electronic gain calibrations for detector amplifiers (which were normally less than 20 ppm), were ignored. A theoretical limit of 0.007%/amu was reported for the optimized tracer/sample ratio. Fletcher et al. (1997) concluded that, even Ca data, which were somewhat less precise than the theoretical precision limit, would enable Ca isotope studies to consider the range of natural fractionation in geological samples. Differences in abundances of Ca isotopes result in low counting statistics for minor isotopes, as well as in interference from scattered of 40 Ca þ ions in the mass range of minor isotopes. The ion-beam intensities of the minor isotopes are limited by the maximum 40 Ca þ intensity, which accounted for the normal used O resistors to A. Some researchers used a O collector resistor to measure the major isotopes, and O resistors to acquire the signal of less-abundant isotopes. Problems of such an approach were associated with high beam currents in a multi-collector system (as could be done with a O resistor on one collector) because the implanted Ca was likely to affect the ion-collection efficiency of the Faraday cups, and degrade the multi-collector calibrations (Fletcher et al., 1997). Presently, the Thermo-Electron Triton TIMS possesses a 50 V dynamic range, so that the major ion beam can be higher by up to fivefold. However, increasing damage to the Faraday cups, which can affect the cup efficiency and response linearity, is again the major complication for the routine measurements at high ion beam intensities (Holmden, 2005). Calcium isotopes might be acquired with a single Faraday collector in a sequence by using the one-beam-at-a-time dataacquisition schemes or by simultaneous collecting multiple ion beams. The major disadvantages of the single-collector method are lower sample throughput and a reduced statistical precision (Heuser et al., 2002). Furthermore, larger uncertainty is caused by time-dependent fluctuations of the ion beam intensity (beam instability), so that most static mass discrimination effects are insignificant compared to measurement precision (Habfast, 1998). With multiple collector arrays for Ca isotope measurements (Price et al., 1990; Fletcher et al., 1997), precision of the acquired isotopic data increased. Thus, a multi-collector TIMS technique, developed recently by Heuser et al. (2002), improved average internal statistical uncertainty of the 44 Ca/ 40 Ca measurements by a factor of 2 4 relative to the commonly used peakjumping method. They found that the internal reproducibility for the 40 Ca/ 48 Ca and 44 Ca/ 48 Ca ratios of a multi-collector measurement was generally better than ca. 0.15%, whereas for the single-cup measurements, statistical uncertainties varied from 0.09% to 0.30%. Other researchers (DePaolo, 2004) noted that, although multi-collection did improve the precision of individual measurements, there was an unaccountable drift that worsened the reproducibility between repetitive analyses of Ca isotopic ratios. A complication of the multi-collector technique is that static effects must be corrected for with appropriate calibration procedures. Thus, calibration of differences in amplifier gains between collectors is achieved by a computer-controlled switching of a constant current into each Faraday cup amplifier circuit before beginning a measurement (so-called gain calibration ) or rotating amplifier circuits between collectors during the measurement, to eliminate subtle differences in amplifier gains (Holmden, 2005). Scattering of high-abundant 40 Ca þ ions from the mass spectrometer beam tube and into the Faraday cup caused a small, broad, reflected peak in the m/z interval from 42.5 to 49, and increased the background for minor isotopes (Russell, Papanastassiou, & Tombrello, 1978). The scattering was reduced with a series of baffles. Russell, Papanastassiou, and Tombrello (1978) measured zeroes at 0.15 amu from each mass peak, and corrected the background by averaging the zeroes for each peak. Peak-shape defects have been reported to represent a critical limitation of the quality of mass spectrometer data. Fletcher et al. (1997) observed, with a modified VG354 mass spectrometer, that the peak shapes from the outer cups were asymmetric, and limited the range of measurable calcium isotope ratios. Only fine adjustments to the magnet geometry could produce acceptably symmetric, flat-topped peaks simultaneously in collectors near both edges of the available spread. Heuser et al. (2002) also found these peak-shape defects for the outer cups of the Finnigan MAT 262 TIMS, but the asymmetry was less pronounced than reported by Fletcher et al. (1997). By narrowing the aperture slits, the asymmetry of the peak shapes could be reduced. The measurements with and without the aperture slits did not show any significant influence on the Ca isotope ratios. Therefore, Heuser et al. (2002) concluded that the small peak-shape defects observed in their TIMS were negligible. 690 Mass Spectrometry Reviews DOI /mas

7 CALCIUM ISOTOPE ANALYSIS & Wide dispersion of Ca ion beams in the mass spectrometer negates static multi-collector data acquisition over the entire mass range in commercial mass spectrometers. This problem could be solved with a sequence of two or three static multicollector measurements. Furthermore, between-sample variation in beam dispersion can also occur (Fletcher et al., 1997). This minor spectral defect could compound any problems caused by peak-shape imperfections at off-axis collectors, so that collector spacing had to be adjusted in that work before each analysis. Effect of integration and idle times on the accuracy of calcium isotope measurements was studied by Heuser et al. (2002). The integration and idle times were varied between 1 and 8 sec. The experiments with a 2 sec integration and 1 sec idle time showed the highest 44 Ca/ 40 Ca ratios, whereas longer integration and idle times resulted in lower 44 Ca/ 40 Ca ratios. Presumably, if the idle time was short (1 sec), then the 40 Ca signal was not completely decayed before the 44 Ca signal was recorded with the same cup. To guarantee complete decay of the 40 Ca signal on the Faraday cup, 4 sec of integration and idle time were chosen for both sequences. It should be also mentioned that nonlinearity in the beamdispersion pattern made it difficult to use dynamic multicollector procedures (which would allow algebraic elimination of the efficiency terms for the Faraday ion collectors), even over small parts of the spectrum (Fletcher et al., 1997). In the modern TIMS instruments, it is possible, however, to arrange moveable Faraday cups to allow the simultaneous measurement of all Ca isotopes in two sequences (Heuser et al., 2002; Böhm et al., 2006; Holmden, 2005; Farkas et al., 2007a). In general, optimizing certain instrumental conditions reduced static mass biases but did not eliminate them completely. Unfortunately, most of these parameters cannot in practice be individually identified because they are not directly accessible from a simple ion current ratio measurement. This fact results in the current use of various empirical correction methods for high precision data. Therefore, such normalized results are sometimes not comparable from instrument to instrument (or from laboratory to laboratory), as long as there is any uncertainty on the existence of static biases. 4. Isotope Dilution with Double Spike for Ca Isotope Measurements The recent review of Ca isotopic variations in natural materials (DePaolo, 2004) showed that there were only a few laboratories worldwide involved in such studies. Most of the laboratories that measured Ca isotopes with TIMS used either 42 Ca 48 Ca or 43 Ca 48 Ca double-spikes for mass-bias corrections (see Table 1). According to Russell, Papanastassiou, and Tombrello (1978), the choice of a 42 Ca 48 Ca tracer, with a 42 Ca/ 48 Ca ratio of , provided a large mass difference for the tracer isotopes, so that uncertainties in the instrumental mass fractionation per amu during analysis of a spiked sample would be reduced. The use of 46 Ca as a tracer was excluded, because Russell, Papanastassiou, and Tombrello (1978) assumed that the presence of high levels of 46 Ca could possibly interfere at a later date in a search for Ca components of distinct nucleosynthetic origin. Later on, that tracer, but with different 42 Ca/ 48 Ca ratios, was used by Skulan, DePaolo, and Owens (1997), Zhu and Macdougall (1998), and Lemarchand, Wasserburg, and Papanastassiou (2004). Three European laboratories in Germany, Switzerland, and France (Hippler et al., 2003) used the 43 Ca 48 Ca double-spike, and employed those isotopes with the lowest natural abundance beside 46 Ca. The sample-to-spike ratio was adjusted in a way that ca. 90% of 48 Ca and 43 Ca originated from the spike. Using that approach, 40 Ca/ 44 Ca and 42 Ca/ 44 Ca ratios could be both measured in natural samples. A comprehensive description of use of double isotope spikes and related error estimation can be found in a recent work by Fantle and Bullen (2009). In theory, the exponential law (Eq. 1) makes an assumption of the ruling role of the mass ratio only so that fractionation factor p is identical for all ratios of the given multiisotopic system and must not depend on mass. According to Habfast (1998) there is a small fundamental difference between the exponential law and the Rayleigh law, where fractionation factors are slightly different for two different isotope ratios in a multi-isotope system that produces a mass-dependent offset of the fractionation. Thus, a fractionation factor of all empirical laws should not only depend on time but also (to a lesser degree) on the isotope ratio itself. Gopalan, Macdougall, and Macisaac (2006) and Holmden (2005) applied, therefore, the 42 Ca 43 Ca double spike for high-precision Ca isotope analysis. The underlying reason to use these isotopes is the smaller mass difference of isotopes. Thus, the mean masses of 42 Ca 48 Ca or 43 Ca 48 Ca tracer pairs used in other works differed by 3 and 3.5 mass units, respectively, from that of the 40 Ca/ 44 Ca pair. Gopalan, Macdougall, and Macisaac (2006) assumed that such a mismatch could result in significant errors in an exponential mass fractionation correction, even larger than those reported by Jungck, Shimamura, and Lugmair (1984) and Hart and Zindler (1989). An obvious manifestation of such an error would be a pronounced residual dependence of normalized ratios on mass bias during a measurement. Gopalan, Macdougall, and Macisaac (2006) stated that a 42 Ca 43 Ca double spike should minimize errors in the corrected 44 Ca/ 40 Ca ratio, which are produced by any deviation of the actual instrumental fractionation from an exponential law because the normalizing and the corrected ratios differ in their average mass by only 0.5 mass units. It should be mentioned, however, that a fractionation factor was measured over only one mass unit with a 42 Ca 43 Ca spike and, therefore, the measurement precision of the 42 Ca/ 43 Ca ratio required to attain a given precision in the targeted 44 Ca/ 40 Ca ratio must be higher than the precision of 42 Ca/ 48 Ca or 43 Ca/ 48 Ca ratios with corresponding spikes. An important advantage of a 42 Ca 43 Ca spike is that all four isotopes ( 40 Ca, 42 Ca, 43 Ca, and 44 Ca) required for the analysis can be measured simultaneously in a modern multi-collector TIMS. Such simultaneous measurement eliminates errors in mass fractionation correction caused by rapid fluctuations or drifts in fractionation and also reduces analysis time. Gopalan, Macdougall, and Macisaac (2006) reported that these advantages led to more consistent Ca isotopic results, so that external uncertainty, or reproducibility, was in the range of 0.15% or better for a single measurement. Two groups (Fletcher et al., 1997; Gopalan, Macdougall, & Macisaac, 2007) have used 43 Ca 46 Ca double spike. Fletcher Mass Spectrometry Reviews DOI /mas 691

8 & BOULYGA et al. (1997) selected 46 Ca instead of 48 Ca to reduce the mass range of measured isotopes for multi-collector measurements. Gopalan, Macdougall and Macisaac (2007) proposed a 43 Ca 46 Ca spike to determine 48 Ca/ 42 Ca and 48 Ca/ 40 Ca ratios. Those researches stated that the 48 Ca/ 42 Ca ratio was a superior choice for studies of mass-dependent fractionation of calcium isotopes in natural materials because it allowed the massdependent fractionation to be determined free from the effects of inherited or ingrown radiogenic 40 Ca, and because this ratio increased the spread of measured isotopic masses by 50% to produce statistically better resolution of fractionation (assuming similar precision). In addition, the 48 Ca/ 42 Ca isotope ratio is close to unity so that it can be measured more precisely than very small ratios of 44 Ca/ 40 Ca. Gopalan, Macdougall, and Macisaac (2007) optimized the isotopic composition of the 43 Ca 46 Ca double spike for 48 Ca/ 42 Ca ratio measurement, and used a two-step dataacquisition procedure instead of the three-step approach, which was applied by Fletcher et al. (1997). To further improve precision of the measured 48 Ca/ 42 Ca ratio they avoided the most abundant 40 Ca, and increased current of minor isotopes by a factor of 4. Gopalan, Macdougall, and Macisaac (2007) reported a precision of 0.18% (2s) to determine the 48 Ca/ 42 Ca ratios by multi-collector TIMS. The 48 Ca/ 40 Ca or 44 Ca/ 40 Ca ratios could also be measured in the same analysis to provide complementary information on any radiogenic component. 5. Comparability and Conversion of Ca Isotope Data The direct comparison of Ca isotopic data sets obtained in different laboratories still remains problematic. Reported absolute calcium isotope ratios in different samples might show significant deviations (Table 2). In the majority of studies, only relative d-values have been measured, expressed as d x=y Cað%Þ ¼ x Ca= y Ca sample x Ca= y Ca standard 1 1; 000 Russell, Papanastassiou, and Tombrello (1978) expressed the measured Ca isotope ratios as d 40/44 Ca values by normalizing the measured 40 Ca intensity to 44 Ca. Later on, Skulan, DePaolo, and Owens (1997) proposed the d 44 Ca-notation by normalizing the ð2þ measured 44 Ca intensity to 40 Ca. In the ICP-MS analysis (described in the section below), the 40 Ca isotope is usually not measured because of a heavy interference from 40 Ar þ ions, and, therefore, normalization to 42 Ca is applied. The d-values of samples can be converted to the different reference standards as long as the conversion factor of both standards is known. Previous studies with Ca isotopes used different laboratory standards (CaF 2, CaCO 3, seawater, etc.), and reported data that were only self-consistent if the applied standards had the same isotopic composition. The apparent differences between the reference samples in use could be in the same order of magnitude as natural variations (Hippler et al., 2003). Thus, a definition of a common Ca isotopic standard was highly desirable. CaF 2 salts used by Russell, Papanastassiou, and Tombrello (1978) and Skulan, DePaolo, and Owens (1997) were supposed to represent average bulk earth; that assumption was supported by an investigation of numerous volcanic rocks and mid-ocean ridge basalts (see, e.g., DePaolo, 2004). Therefore, DePaolo (2004) normalized all Ca isotope ratios reported by different groups to the CaF 2 standard used by Skulan, DePaolo, and Owens (1997). For instance, d 44 Ca values reported by Nägler et al. (2000) and Gussone et al. (2003) were increased by 0.5, and the d 44 Ca reported by Schmitt, Stille, and Venneman (2003a), Schmitt, Chabaux, and Stille (2003b) were increased by 0.9, to make them consistent with previous results. However, those CaF 2 standards were not available to all laboratories to obtain uniform standardization of their results. To compare data from different laboratories, Zhu and Macdougall (1998) and Schmitt et al. (2001) proposed the use of seawater as a common reference material. Seawater is widely available, and has a relatively high Ca concentration of 400 mg/l; therefore, it is a very suitable reservoir to serve as the Ca standard. Zhu and Macdougall (1998) and De La Rocha and DePaolo (2000) found that the Ca isotopic composition of seawater from different parts of the world was the same within the analytical uncertainties of the available measurement techniques. On the other hand, DePaolo (2004) pointed out that the calcium isotopic composition of seawater was determined by geological processes and could change with time, whereas the bulk earth value represented the bulk of the Ca in the planet TABLE 2. Reported absolute calcium isotope ratios * 40 Ca/ 44 Ca can vary due to radiogenic 40 Ca. **Summarized data from previous publications. 692 Mass Spectrometry Reviews DOI /mas

9 CALCIUM ISOTOPE ANALYSIS & and this value remained constant over time. In addition, seawater also had a small radiogenic enrichment of 40 Ca by % relative to bulk earth calcium. Therefore, DePaolo (2004) proposed to define d 44 Ca ¼ 0 as the bulk earth value rather than that of seawater. Thus, an appropriate standard could represent a rock sample with high Ca and low K abundance, and with the mantle 40 Ca/ 42 Ca ratio (i.e., with negligible radiogenic 40 Ca). An inconvenience of seawater and rock samples as isotopic standards is that Ca isotope measurements require a calcium separation from matrix prior to analyses. Halicz et al. (1999), Coplen et al. (2002), and Hippler et al. (2003) proposed, therefore, NIST SRM 915a CaCO 3 as the reference standard. Hippler et al. (2003) proposed to define the difference between NIST SRM 915a and seawater as a reference for published and future oceanographic studies. According to Hippler et al. (2003), the d 44/x Ca Sa values in a sample normalized to NIST SRM 915a can be converted to d 44/x Ca values related to seawater in the following manner: d 44=x Ca Sa=Sw ¼½ð10 3 d 44=x Ca Sa=NISTSRM 915a þ 1Þ ð3þ ð10 3 d 44=x Ca NISTSRM 915a=Sw þ 1Þ 1Š10 3 The subscripted indices in the formula correspond to the deviation sample-seawater (d 44/x Ca Sa/Sw ) and sample-nist SRM 915a (d 44/x Ca Sa/NIST SRM 915a ). The weighted average value to describe the conversion factor between NIST SRM 915a and seawater (d 44/40 Ca NIST SRM 915a/Sw) was 1.88 (d 44/42 Ca NIST SRM 915a/Sw ¼ 0.94); and was, therefore, within the existing analytical precision: d 44=40 Ca Sa=Sw ¼ d 44=40 Ca Sa=NISTSRM 915a 1:88 d 44=42 Ca Sa=Sw ¼ d 44=42 Ca Sa=NISTSRM 915a 0:94 Since 2006, however, the Ca reference material NIST SRM 915a has been out of stock and was replaced by NIST SRM 915b (Heuser and Eisenhauer, 2008). Therefore, Heuser and Eisenhauer (2008) analyzed Ca isotopic composition (d 44/40 Ca) of NIST SRM 915b and NIST SRM 1486 relative to NIST SRM 915a to provide a precise calibration of those reference materials. ð4þ ð5þ Figure 1 presents d 44/40 Ca values of different standards relative to NIST SRM 915a. To increase the comparability between laboratories, which used different techniques and notations, Eisenhauer et al. (2004) proposed that Ca isotope data based on TIMS and MC-ICP-MS measurements were presented as d 44/40 Ca or d 44/42 Ca. They also recommended to use calculated d 44/42 Ca values. The equation used for the conversion was published by Hippler et al. (2003): d 44=40 Ca ¼ d 44=42 ð43:956 39:963Þ Ca ð6þ ð43:956 41:959Þ The conversion above assumed that isotopic differences due to radiogenic 40 Ca in the analyzed standards and seawater were negligible, and it could only be applied to samples with low potassium concentrations. Coplen et al. (2002) noted that 40 Ca might be a poor choice for the denominator in the isotope ratio measurements because it was a product of 40 K radioactive decay, in particular, if investigated rocks and minerals were old and rich in potassium, so that they could have significant enrichments in 40 Ca relative to the other Ca isotopes. Thus, Marshall and DePaolo (1982) chose to use 40 Ca/ 42 Ca as the target ratio to monitor the radiogenic enrichments of 40 Ca in various geological materials, and used 42 Ca/ 44 Ca for the mass discrimination correction. DePaolo (2004) also stated that the difference due to radiogenic Ca enrichment in standards used for normalization was only marginally significant because the typical analytical uncertainty was of approximately the same value. Figure 2 compares d 44/40 Ca and d 44/42 Ca in seawater measured with TIMS and MC- ICP-MS, respectively, with the weighted averages (d 44/40 Ca) and (d 44/42 Ca). The average d 44/ 40 Ca value calculated from the MC-ICP-MS is 1.880, which is almost identical with the corresponding value measured by TIMS ( 1.889). Figure 2 shows that this difference practically cannot be distinguished within the measurement uncertainties, and therefore, possible radiogenic enrichment of 40 Ca in seawater is difficult to determine by available measurement techniques. In a recent work, Farkas et al. (2007b) concluded, based on their experimental 40 Ca/ 44 Ca ratios, that modern and Early Paleozoic seawater should both have a radiogenic Ca excess that was indistinguishable from the mantle signature. FIGURE 1. Calcium stable isotope compositions of reference samples relative to NIST SRM 915a (based on isotopic data by Hippler et al., 2003; Heuser and Eisenhauer, 2008; Wombacher et al., 2009). Mass Spectrometry Reviews DOI /mas 693