A coherent method for combined stable magnesium and radiogenic strontium isotope analyses in carbonates (with application to geological reference materials SARM 40, SARM 43, SRM 88A, SRM 1B)

We undertook 87Sr/86Sr analyses for a range of carbonate bearing geological reference materials, and combined these with δ26Mg for a subset of samples. Following chemical purification in a series of chromatographic extractions, isotope ratios were measured by Multi-Collector-ICP-MS using a Plasma II (Nu instruments, Wrexham, UK). To validate efficient sample digestion procedures of carbonate fractions, total samples were treated with either 3 mol l−1 HNO3 and 0.5 mol l−1 HCl, respectively. Results of both leaching procedures are identical within reproducibility. Reference values for SRM 88A (formerly NBS 88A), SRM 1B (formerly NBS 1B), SARM 40, SARM 43, JDo-1, JLs-1, and San Carlos olivine range from 0.70292 to 0.73724 in 87Sr/86Sr and from -2.80 to -0.41 ‰ for δ26Mg, respectively. This set of geological reference materials can be used for sedimentary rock material with different carbonate mineral and matrix composition as quality control measurements of combined stable Mg and radiogenic Sr isotope analyses.• We present a protocol that facilitates the chemical separation of Mg and Sr in carbonate bearing geological reference materials including 87Sr/86Sr and δ26Mg of certified reference materials.


Introduction
Studies on stable and radiogenic isotope variations in natural materials have substantially increased over the last decades and, together with technological and scientific development, provide nowadays high precision analyses for a large number of isotopic systems (e.g., [1][2][3] ). High-precision isotope analyses have become a cornerstone of scientific research with applications in the fields of hydroand geosciences as well as e.g., forensics, archaeology or medical sciences [4][5][6] . Such analyses, however, require means of testing accuracy and precision as well as newly established methodologies in laboratories.
Among radiogenic isotope systems, 87 Sr/ 86 Sr is well-established in low-temperature marine research, in particular considering carbonate, phosphate and sulphate minerals. For the latter mineral groups, the incorporated radiogenic Sr in the bivalent ion position of the mineral structure is used as an environmental proxy and tracer. Accordingly, radiogenic Sr isotopes have been used to trace Sr sources and mixing behaviour in aquatic bodies [23][24][25] . Globally, by means of the relatively long residence times in ocean water, Sr isotopes are further considered to be almost homogeneously distributed in global oceans over a million-year time-interval, which has led to the well-established Phanerozoic seawater 87 Sr/ 86 Sr evolution curve. Recorded variation in past ocean waters from ca. 0.710 to 0.706 [26][27][28] can thus potentially be used to trace silicate weathering vs . mid-ocean ridge hydrothermal influx [29] , and through Sr chronostratigraphy may provide rough age constraints when compared with the seawater 87 Sr/ 86 Sr evolution [30] .
Combining stable Mg isotope and radiogenic Sr data has great potential within multi-proxy approaches, in particular in low temperature environments due to their high abundance in aquatic systems, solely divalent ion character and complementary stable vs . radiogenic isotope tracer behaviour. Whilst Sr isotopes can routinely be analysed through thermal ionisation mass spectrometry (TIMS), Mg isotopes can be performed to much higher efficiency with a multi-collector inductively couple plasma mass spectrometer (MC-ICP-MS). However, despite pitfalls [31] , Sr has also been analysed with MC-ICP-MS providing a much higher sample throughput [32] .
Chemical separation protocols for both Mg and Sr have been tested and optimised for different matrices, using cation exchange resin, e.g.,: AG50W-X12 (BioRad R , Hercules, USA), for Mg separation and Sr specific chromatographic resin, e.g., from Eichrom Technologies Inc. (USA) or TrisKem International (France), for Sr separation (e.g., [21 , 33] ). Some of these protocols facilitate the simultaneous separation of different elements [22 , 33 , 34] . Among these, simultaneous separation of Mg and Sr (and also Ca) is especially interesting for carbonate bearing materials. Although, the use of combined radiogenic and stable isotope investigations, e.g., in Proterozoic to Phanerozoic carbonate rock, requires reference material for quality control measurements for testing accuracy and precision, surprisingly little combined δ 26 Mg-87 Sr/ 86 Sr isotope data are available for carbonate bearing geological reference materials.
In this study, we carried out combined stable Mg -radiogenic Sr isotope analyses on natural calcareous and carbonate bearing geological reference materials using MC-ICP-MS, where separation protocols were modified after [22] . Our protocols are developed in order to facilitate the near simultaneous or coupled routine analyses of Mg-Sr isotopes in Ca-rich samples. Here, we employ two different digestion methods for carbonate rocks (e.g. limestones and dolostones), using 0.5 mol l −1 HCl and 3 mol l −1 HNO 3 , respectively. We analysed the HCl and HNO 3 soluble fraction of geological reference materials that include the carbonate minerals calcite, dolomite and/or magnesite.

Reference materials
For both isotope systems, respective Mg and Sr isotope values were determined on different types of calcareous and carbonate bearing geological reference materials and seawater. Reference materials SRM 88A, SRM 1B, SARM 40, SARM 43, JDO-1, and JLS-1 were chosen to represent carbonate materials with varying Ca/Mg ratios (0.02-127) and different bulk mineral chemistry ( Table 1 ). Additionally, for Mg isotopes, the non-certified reference material "San Carlos olivine" was analysed, a natural, forsterite-rich olivine with reported isotope values for Mg [35] . For quality control, reference materials IRMM-009 and Cambridge-1 (CAM-1) were analysed. These two reference materials are pure Mg nitrate solutions that were prepared in batch and distributed by the Institute for Reference Materials and Measurements (IRMM-009) and A. Galy (CAM-1) [36 , 37] . A brief sample description is presented in Table 1 .

Purification of chemical reagents
Sample digestion and ion (-exchange) chromatographic separation of Mg 2 + and Sr 2 + were carried out in laminar flow hoods, using Savillex R or AHF R PFA beakers. Both HNO 3 and HCl acids ( pro analyses quality) used for separation and dilution were doubly purified by sub-boiling distillation in a PFA Savillex R DST-10 0 0. The blanks of the purified acids were tested to be below the detection limits of < 50 ng l −1 for both Mg and Sr. Dilution of acids was performed with 18.2 M Ω * cm H 2 O (at 25 °C with < 5 ng ml −1 TOC; MilliQ R ). Beakers were cleaned in a two-step cleaning process involving both boiling in 5 mol l −1 HNO 3 and 6 mol l −1 HCl at 120 °C for a minimum of 24 h each. Other lab equipment (e.g., PE bottles, pipette tips) were cleaned in 0.8 mol l −1 HNO 3 at 60 °C for at least 48 h.
Ion chromatographic resins were alternately cleaned in MilliQ R water and 1 mol l −1 HNO 3 or HCl, respectively, and then stored in MilliQ R water. For Mg separation BioRad R AG50 × 12 resin was used. Columns consist of polypropylene with 5 cm length and 0.5 mm diameter loaded with 1 ml resin and an additional 5 ml reservoir. The columns -including the resin -were cleaned in several millilitres of both acids before use. In between each cleaning step the resin was rinsed with several column volumes of MilliQ R water. Strontium columns are prepared for each separation individually. Columns and polyethylene frits were cleaned in 0.45 mol l −1 HNO 3 and stored in MilliQ R water until used.

Sample digestion
The San Carlos olivine was grinded in an agate mortar and then digested in 3 ml of a 1:3 HNO 3 -HF concentrated acid mixture, sealed tight and left to boil on a hot plate at 110 °C for 24 h. The HNO 3 -HF was evaporated at 70 °C and the samples were treated with a small amount of concentrated HNO 3 and concentrated H 2 O 2 as oxidizing agents to eliminate Ca-fluoride complexes. These solutions were then dried and re-dissolved in concentrated HCl to eliminate remaining nitrates. Prior to separation, seawater was filtered through a 0.45 μm membrane acetate filter (Sartorius). For chemical separation, 10 ml of seawater was dried down and treated with small amounts of H 2 O 2 and concentrated HNO 3 to break up potential organic complexes.
For the geological reference materials SRM 88A, SRM 1B, SARM 40, SARM 43, JDO-1, and JLS-1, ca. 100 mg powdered material was digested. From these stock solutions, aliquots were taken (different in volume), each for Mg and Sr separation, aiming for a concentration of ca. 2-20 μg ml −1 for Mg depending on the Ca/Mg ratio and 20 μg ml −1 for Sr, and. Each sample was treated with the respective acid until no carbonate dissolution reaction was visible. Stammeier et al. [28] have shown that sample digestion using 0.1 and 3 mol l −1 HNO 3 with carbonate-bearing material has no significant effect on Sr isotope composition. Thus, this method was used to evaluate (external) reproducibility and repeatability employing digestion using solely 3 mol l −1 HNO 3 . As incomplete digestion is especially important for Mg isotopes, we further digested a set of samples using diluted HCl for comparison. These samples were digested in 0.5 mol l −1 HCl, in order to evaluate internal reproducibility. Aliquots were evaporated to dryness and re-digested in the respective acids (1.5 mol l −1 HNO 3 for Mg and 3 mol l −1 HNO 3 for Sr) used for the respective chemical separation protocol ( Tables 3 , 4 ). One procedural blank sample was included per ten samples per separation.

Magnesium separation
For Mg purification, a two-step ion exchange chemistry was employed using HNO 3 and HCl as eluent (after [20 , 22] ). This two-step separation is optimized for samples with a high Ca/Mg ratio. Magnesium yield during the Ca separation step. Separation was performed with two different replicates yielding identical results. After 15 ml HNO 3 more than 99% of Mg is eluted, as is required to avoid Mg fractionation during chemical separation [33] . Calcium and Sr can effectively be eluted with a higher concentrated acid. Sodium (Na) is eluted prior to Mg, however in Na-rich samples, e.g. seawater or experimental fluids, the Na elution is retarded and overlaps with the Mg peak (not shown here). The first step ensured the effective separation of Ca from the matrix and can also be employed for a simultaneous isolation of Zn, Fe and Ca ions for subsequent isotope analyses ( [22]; Table 3 , Fig. 1 ). In the second step, Mg was separated from other matrix elements, such as Na, K and Ti ( Table 3 , Fig. 2 ). Both separation steps were performed on the same columns using the BioRad R AG50-X12 resin. Between the two separation steps, the columns were cleaned with one column volume of 7 mol l −1 HNO 3 and MilliQ R water. Separation was tested with an artificial solution containing 10 μg ml −1 of Mg, Na, Sr and (i) 10 μg ml −1 of Ca, i.e., with a Ca:Mg ratio of 1:1; and (ii) 100 μg ml −1 of Ca, i.e., with a Ca:Mg ratio of 10:1. For the first Ca separation step, the columns were conditioned with 2 ml of 1.5 mol l −1 HNO 3 . The sample was subsequently loaded with 1 ml of 1.5 mol l −1 HNO 3 . After elution with 8 ml of 1.5 mol l −1 HNO 3 , Mg was recovered in 11 ml of 1.5 mol l −1 HNO 3 . The remaining divalent cations on the columns were washed off using 10 ml of 7 mol l −1 HNO 3 . In the second step the columns were

Strontium separation
For the Sr separation a single, well-established extraction ion chromatographic chemistry was employed ( Table 4 , after [38] ). A second separation step, sometimes required for high-Rb samples [39] is not required, as calcareous or carbonate bearing material can be expected to have low to negligible Rb/Sr. The columns consisted of polypropylene pipette tips (Eppendorf) and 20-60 μm polyethylene frit material (Porex Corporation, Georgia, USA) and were filled with 100 μl Sr-Specific resin (TrisKem, France). For Sr separation 3 mol l −1 HNO 3 and MilliQ water were required. Dried down samples were re-dissolved in 1 ml 3 mol l −1 HNO 3 . The columns were conditioned in 1 ml 3 mol l −1 HNO 3 and then loaded with the sample. After washing with 4 ml 3 mol l −1 HNO 3 , Sr was recovered in 3 ml MilliQ.
To test the yield during the Sr separation, an element reference solution with 10 μg ml −1 of Mg, Ca, Rb, Sr, Fe and Zn (admixed from Merck single element standard solutions) was prepared ( Fig. 3 ). The elution curve shows Rb, which forms an isobaric interference with 87 Rb on 87 Sr during measurements, is effectively separated after 3 ml of washing with 3 mol l −1 HNO 3 and Sr should be collected after 5 ml of washing (including the ml of sample loading, Fig. 3 ). Further, Mg is effectively separated from the Sr fraction, potentially facilitating a coupled separation with the Sr separation step in the reverse order. However, due to the high Ca content of most samples, this is not recommended as columns may be overloaded. In fact, with Ca-rich samples some Ca is eluted together with Sr ( Fig. 3 ), which might cause matrix effects. In these cases, the sample could be passed over the columns twice to effectively eliminate all Ca. The bulk Sr was eluted from the columns with MilliQ water.

Data acquisition and reduction
Isotope analysis was carried out on a Plasma II MC-ICP-MS (Nu instruments, Wexham, UK) at the NAWI Central Laboratory for Water, Minerals and Rocks at Graz University of Technology, Austria. The instrumental parameters and settings during measurements of the Plasma II are summarized in   Table 4 Strontium separation, modified after [62] .
Step  Table 2 for the respective isotopes. Torch position, Ar-gas flow rates and lens set up were optimized to achieve maximum signal intensity and stability of the main beam, 24 Mg and 88 Sr, for Mg and Sr, respectively. Analyses were typically performed in low resolution with a sensitivity of 15 V for 150 μg l −1 Mg and 25 V for 500 μg l −1 Sr, respectively, on the highest abundant isotopes ( 24 Mg, 88 Sr). Magnesium was measured in dry-plasma mode using a DSN 100 desolvator (Nu instruments, Wrexham, UK), whereas Sr was measured in wet plasma mode using a static cup set-up. The nebulizer flow rate was 0.1 mL/min. Data acquisition of Mg and Sr isotopes consisted of 1 block with 25 cycles with an integration time of 5 s each. The background was determined by measuring 10 s at half masses before each block. To ensure repeatability and reproducibility, repeated analysis of reference materials Cambridge-1 and IRMM-009, normalised to DSM3, during Mg isotope measurements and seawater during Sr isotope measurements, were performed. Concentration of reference materials and samples was adjusted to match within 10%, in order to avoid amplification of mass bias induced differences [40] . The total procedural blank was below 0.4 μg Mg and 1.2 ng Sr and negligible compared to analyte signals. Thus, no blank correction was performed.

Magnesium isotopes
Magnesium isotopes were collected in Faraday cups with a set-up reported in Table S1. Instrumental mass bias was corrected for using the standard sample bracketing (SSB) method normalizing to the DSM3 reference material [41] . Mass drift of two bracketing standards (DSM3) may exceed the anticipated repeatability of ±0.25 ‰ of δ 26 Mg [22] , causing samples to be artificially shifted and yield inaccurate results. To circumvent this, bracketing standards and respective enclosed samples exceeding this repeatability were discarded. Magnesium isotope ratios are reported in the δ-notation calculated relative to DSM3 reference material: with X referring to either 25 Mg or 26 Mg, respectively.

Strontium isotopes
During Sr isotope measurements each isotope was collected in an assigned cup as reported in Table S2. All measured Sr isotope ratios were in-run corrected for baseline, interferences ( 87 Rb and 86 Kr) and instrumental mass bias. The latter can be corrected by using the observed mass bias factor β of an invariant isotope ratio, in this case a 86 Sr/ 88 Sr = 0.1194 [42] , and the exponential law. Interference correction is applied by monitoring an isotope of the respective element without isobaric interference, e.g., 85 Rb and 84 Kr, 86 Kr and subtracting the mass bias corrected isotope ratios. Krypton interferences are corrected for using a value of 86 Kr/ 84 Kr = 0.3035; Rb interferences are corrected with a value of 87 Rb/ 85 Rb = 0.3857 [43] . Note that through chemical purification of Sr, Rb contents should be negligible and Rb interference correction does not affect the Sr isotope analyses.
The mass bias factor β is determined in an iterative calculation: interference of 86 Kr on 86 Sr was first subtracted, using a synthetically biased 86 Kr/ 84 Kr. For this, a β 0 value was calculated from a non-interference corrected, measured 86 Sr/ 88 Sr value and applied to 86 Kr/ 84 Kr to simulate mass bias for this ratio. The then corrected 86 Sr/ 88 Sr from this first step was used to calculate a new β 1 value and the process was repeated. We found that after ten iterations of consecutive mass bias-and interference-correction β converged to a constant value. The final β 10 value was then applied to the 87 Rb interference corrected 87 Sr/ 86 Sr. Samples were measured in blocks of 6, which were bracketed by two consecutive measurements of NBS 987. Isotope variations in radiogenic Sr isotopes were monitored and corrected for by repeated measurements of NBS 987 in each session. The correction for systematic offsets in analytical sessions was performed by normalizing the acquired data of the average of the bracketing standards measured before and after each set of samples to a reference value of 87 Sr/ 86 Sr = 0.710250 [44] .

Elemental concentrations
Analyses of element concentrations for calibrating columns and testing yields were performed using inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7500cx) at the NAWI Laboratory for Water, Minerals and Rocks, Graz University of Technology, Austria, with a measurement uncertainty generally better than ±5% on element concentrations. Samples for the elution-calibration were taken up in 0.45 mol l −1 HNO 3 . The instrument was tuned to achieve maximum sensitivity while maintaining low oxide production and doubly charged ion ratios with < 1.5% of the total concentrations. The concentration background was determined on a 0.45 mol l −1 HNO 3 blank solution and automatically subtracted from acquired data. Instrumental drift control was performed by simultaneously running an internal reference solution with a 1 ng ml −1 of Sc, Ge, and Bi.
For Sr intermediate precision, expressed as, was evaluated using seawater as a secondary reference with an uncertainty of 87 Sr/ 86 Sr = ±0.0 0 0 011 (2 sd, t = 120 days, n = 19). Repeatability of 87 Sr/ 86 Sr in seawater within each session was typically within 100 ppm and thus well within reported performances of the Plasma II MC-ICP-MS [44] . Reproducibility of the whole procedure, e.g. determined on certified reference materials (CRM) JDo-1 and JLs-1 was within 50 ppm, with the exception of JDo-1 dissolved in 0.5 mol −l HCl with a reproducibility of only 150 ppm. However, comparison of all other 87 Sr/ 86 Sr derived from sample leaching in HCl or HNO 3 shows only a small offset better than 50 ppm.

Magnesium isotopes
The geological reference material "San Carlos olivine" yielded a δ 26 Mg value of −0.41 ± 0.09 ‰ (2 sd, n = 10), which is in the range of published values (compare Table 5 Table 5 ). All results plot on a mass-dependent fractionation line with a slope of β = 0.491, similar to the slope of all previously published values with β = 0.499, both similar to a theoretically calculated β for equilibrium processes of 0.512 ( Fig. 6 ; [55] ). The deviation from the theoretical equilibrium fractionation slope is mainly caused by IRMM009 and SARM43, highlighted by a δ 25 Mg [55] of 0.07 and 0.06 for both the Mg isotope values from literature (IRMM009; [50] ) and this study. The δ 25 Mg, calculated as δ 25 Mg'-β * δ 26 Mg and quantifies the deviation from the equilibrium fractionation line, where a value < 0.04 is generally within analytical uncertainty [52] . Omitting these Mg isotope values from slope calculation the slope is β = 0.512 and thus identical to the equilibrium fractionation. Evidently those two CRM have very low δ 26 Mg, i.e., a larger difference between bracketing standard and sample, thus causing error amplification. For these CRM a different bracketing standard could be used. However, as the internationally agreed-on reference material is DSM3 this would require recalculation of these values relative to DSM3, in which case error propagation has to be considered. This would likely outweigh the observed larger δ 25 Mg when using DSM3 as a bracketing standard and thus not legitimate the extra effort using different bracketing materials.

Strontium isotopes
The lowest 87 Sr/ 86 Sr of this study was analysed for the carbonatite material SARM 40 with an average value of 0.70294 ± 0.0 0 0 04 ( n = 20, Table 6 Table 5 .   [53] . Symbols with a black margin represent sample dissolution in 3 mol l − 1 HNO 3 , with a red margin represent sample dissolution in 0.5 mol l −1 HCl. Bright grey line indicates present day seawater. Error bars represent the 2 sd variation as reported in Table 5 .

Table 6
Compiled results from Sr isotope measurements; n refers to the number of analysis. All data from this study are reported relative to NBS 987 ( 87 Sr/ 86 Sr = 0.710250 ±0.0 0 0 0 08 reported by [63] ). This study a T. Ohno, T. Hirata, Simultaneous determination of mass-dependent isotopic fractionation and radiogenic isotope variation of strontium in geochemical samples by multiple collector-ICP-mass spectrometry., Anal. Sci. 23 (2007)  respectively. 87 Sr/ 86 Sr values of JLs-1 and JDo-1 from this study were identical within analytical precision to published values by Miura et al. [56 , 57] . The dolomitic limestone SRM 88A has high 87 Sr/ 86 Sr of 0.71023 ±0.0 0 0 04 ( n = 12) close to the reference material used for internal normalization NBS 987. The highest value was measured in SARM 43, a magnesite, with an average value of 0.73725 ± 0.0 0 0 05 ( n = 11). The average 87 Sr/ 86 Sr value of seawater was used as an external control reference and was found to be 0.70920 ± 0.0 0 0 01 ( n = 19), identical to reported seawater values of 0.70924 ± 0.0 0 0 03 (e.g., [58] ).

Summary
In the present study a reliable and fast method was developed to acquire stable Mg and radiogenic Sr isotopes of carbonate bearing geological materials with a relatively high Ca/Mg. A set of combined stable Mg and radiogenic Sr isotope values for CRM SARM 43 and SRM 88A, and additionally radiogenic Sr isotope values for SARM 40 and SRM 1B are suggested, which are readily available and can be used as secondary reference material as quality control measurements. To date no such values are available and further systematic work is suggested to build up a reliable database of geological reference material values. Effective chemical separation of both elements from the same digestion was achieved using ion specific resins BioRad 50W-X12 for Mg separation and Sr specific chromatographic resin from TrisKem for Sr separation. Intermediate precision was ±0.10 ‰ for δ 26 Mg and ±0.0 0 0 01 for 87 Sr/ 86 Sr, derived from repeated measurements of CAM-1 and seawater, respectively. Whole procedural reproducibility was ±0.11 ‰ and ±0.09 ‰ for δ 26 Mg, determined on reference materials JDo-1 and IRMM 009. For Sr isotopes, the whole procedural reproducibility was determined on certified reference materials JDo-1 and JLs-1 and was within 50 ppm.
Respective average isotope values for all measured δ 25

Acknowledgments
S. Perchtholt is kindly thanked for assistance in the lab. V. Mavromatis is kindly thanked for proof-reading and scientific advice.

Declaration of Competing Interest
The authors confirm that there are no conflicts of interest.

Funding
This work was financially supported by the research project DFG-FG 736 (HI 1553/1-2, Deutsche Forschungsgemeinschaft, Germany) and NAWI Graz, Central Lab of Water Minerals and Rocks (NAWI Graz Geocentre, Austria). ON acknowledges support from the ARC (FT140101062).