High‐precision determination of lithium and magnesium isotopes utilising single column separation and multi‐collector inductively coupled plasma mass spectrometry

Rationale Li and Mg isotopes are increasingly used as a combined tool within the geosciences. However, established methods require separate sample purification protocols utilising several column separation procedures. This study presents a single‐step cation‐exchange method for quantitative separation of trace levels of Li and Mg from multiple sample matrices. Methods The column method utilises the macro‐porous AGMP‐50 resin and a high‐aspect ratio column, allowing quantitative separation of Li and Mg from natural waters, sediments, rocks and carbonate matrices following the same elution protocol. High‐precision isotope determination was conducted by multi‐collector inductively coupled plasma mass spectrometry (MC‐ICPMS) on the Thermo Scientific™ NEPTUNE Plus™ fitted with 1013 Ω amplifiers which allow accurate and precise measurements at ion beams ≤0.51 V. Results Sub‐nanogram Li samples (0.3–0.5 ng) were regularly separated (yielding Mg masses of 1–70 μg) using the presented column method. The total sample consumption during isotopic analysis is <0.5 ng Li and <115 ng Mg with long‐term external 2σ precisions of ±0.39‰ for δ7Li and ±0.07‰ for δ26Mg. The results for geological reference standards and seawater analysed by our method are in excellent agreement with published values despite the order of magnitude lower sample consumption. Conclusions The possibility of eluting small sample masses and the low analytical sample consumption make this method ideal for samples of limited mass or low Li concentration, such as foraminifera, mineral separates or dilute river waters.

the instrumental mass bias, it is necessary to analyse purified monoelemental solutions (e.g. [22][23][24]. This requires a multi-step sample preparation, including the separation of Li and Mg from sample matrix through cation-exchange chromatography. Lithium has previously been separated by using between one and four separate column procedures (e.g. 23,[25][26][27] whereas Mg is eluted in two or three columns (e.g. 4,18,28 ).
The objective of analysing both Li and Mg on the same sample would thus require between three and seven separate column procedures. This approach is time-consuming and increases sample blanks and the risk of incomplete sample recovery with associated isotopic fractionation.
In this paper we present a single column, one-step elution method to separate small masses of Li and Mg from multiple sample matrices.
Seawater, river water, sediment, foraminifera and rock standards with  Samples used in this study were prepared as follows. Seawater and river water samples were evaporated to dryness at 80°C and then refluxed with 1-2 mL of concentrated aqua regia at 100°C overnight to oxidise organic matter. The samples were then evaporated to dryness and taken up in 0.7 N HCl to be loaded onto the ion-exchange column. Sediment and rock powders were baked at 950°C for 8 h in ceramic crucibles to destroy organic matter, and then dissolved in a mixture of concentrated HNO 3 , HCl and HF (1:1:1) in Savillex© (Eden Prairie, MN, USA) screw-top beakers on a hotplate at 110°C. Post dissolution (typically a few hours), the samples were dried down and taken up in 6 N HCl. If fluoride residues were present the sample was refluxed with concentrated HNO 3 until a clear solution was obtained. An aliquot, generally containing 1-10 ng Li, was then diluted to 200 μL with 0.7 N HCl and loaded onto the columns. The synthetic foraminifera standard was made from pure concentrated stock solutions (ROMIL-SpS™ super purity standards, Waterbeach, UK).
The column loads of different elements for each sample matrix that were utilised to generate the column elution/calibration curves ( Figure 1) are presented in Table 1.
However, the distribution coefficients for Li and Na, and those of Mg, Fe and Mn, for different strength acids and the AG50W resin are similar, especially with increasing acid strength (Table 2). 29 Therefore, dilute acids and/or a combination of several columns are commonly utilised to fully separate both Li and Mg from other matrix elements (e.g. 4,27,28 ). Alternatively, a mixture of dilute HCl or HNO 3 and an organic solvent also increases the separation especially between Li and Na (e.g. 25,26,31,32 ). However, organic solvents and HNO 3 may cause: (1) rapid degradation of the resin resulting in non-quantifiable migration of element peaks; (2) early breakthrough of Na into the Li fraction; 30,31 and (3) rapid volatilisation of methanol, which has been hypothesised to cause element peak migration and cross contamination of Li between columns. 32 Other strategies include initial removal of Fe from high Fe matrices by eluting through an anion-exchange column, reducing the total matrix load. 15,18,33 The Elution curves for various sample matrices: G2 granite (short-dashed lineonly Li), BCR-1 basalt (solid line), foraminifera calcite standard (dashed dotted line), river water (dotted line) and river sediment (dashed line). For samples with high Al and Ti load, the initial 3 mL are eluted in 0.5 N HF, and the elution volume (x-axis) denotes the volume of HCl added. The grey boxes mark the cuts which are collected for Li and Mg isotope analysis, with the 1-mL pre-and post-cuts in grey stripes. The calibration was carried out volumetrically by collection of each millilitre of the elution. Ca, Sr and Ba elute after 100 mL. (For sample composition, see Table 1) peak separation of Mg, especially from Fe and Mn, and that of Li from Na, is significantly larger in the macro-porous equivalent of the gel-type resin -AGMP-50 (BioRad™, Hercules, CA, USA) 34,35 ( Table 2). Utilising a 3-mL Savillex® Teflon™ ion-exchange column with high aspect ratio (25 cm height and inner diameter of 4 mm), quantitative separation of both Li and Mg from multiple matrices is achieved in a single-step elution.

| Elution protocol
The resin was backwashed using a handheld pump and allowed to settle under gravity between each elution. This enables the resin to fully expand and uniformly distribute with homogeneous porosity between each sample elution. The columns were then conditioned with 9 mL (three column volumes) of 0.7 N HCl before being loaded with the sample (typically 2 ng Li yielding Mg masses between 1 and 70 μg). Samples were loaded in <200 μL of 0.7 N HCl, and then eluted with 9 mL of 0.7 N HCl, with the first 1 mL added incrementally in steps of 200 μL to ensure that the sample is properly loaded onto the resin. Li was then eluted in 0.7 N HCl and collected as a 13-mL cut. A 1-mL pre-and a post-Li cut were collected to ensure that there was no Na breakthrough and that the Li

| Mass spectrometry
High-precision isotope ratio determination of both Li and Mg was performed by MC-ICPMS at the University of Cambridge (Cambridge, UK) on the Thermo Scientific™ NEPTUNE Plus™ fitted with a Jet ion extraction pump. We adopted a concentration matched standardsample bracketing technique to correct for instrumental drift and mass bias. Each standard and sample were followed by a background instrumental blank measurement in 2% HNO 3 matrix. A typical TABLE 1 Loaded masses on column for calibration seen in Figure 1 Element unit  and an uptake time of 60 s.

| Li isotopic measurements
Li isotopic ratios were determined with respect to the NIST L-SVEC standard 37 and each analytical session included the measurement of secondary standards spiked with 6 Table 4. The key feature of the δ 7 Li determination method was the use of 10 13 Ω amplifiers (Thermo Scientific) with ultra-low electronic noise that allowed determination of precise 7 Li/ 6 Li ratios with 6 Li and 7 Li beam sizes of ≤35 mV and ≤0.51 V, respectively. The low baseline noise of the 10 13 Ω amplifiers (±0.9 μV, 1σ, n = 900) gave a 4-to 5-fold higher signal-to-noise ratio for 6 Li beam intensities of 15-35 mV than when using 10 11 Ω amplifiers (±4.2 μV, 1σ, n = 900). Prior to each analytical session a long baseline of 900 cycles was performed. A Savillex® C-flow self- FIGURE 2 Separation of elements in the AGMP-50 resin utilising our elution protocol, with concentration in log-scale to magnify tailing of neighbouring elements for the river water matrix (A, B) and the river sediment matrix (C, D). Lithium is separated from Na with no peak-tail overlap. There is a minor (although insignificant given the high Mg concentrations) tailing of Na into the Mg peak. Samples that are not eluted with initial HF (e.g. water samples and foraminifera carbonate) have Ti and Al eluting after Mg (B), compared with prior to Li when HF is used (C). Ti concentrations are low in water samples and the tailing into Mg is negligible. Note that in (A) the average Na blanks from the pre-Na peak are plotted. Mg is clearly separated from Mn, Fe, and K (both Fe and K elute at >75 mL) Our strategy was to minimise the deposition of Li by pre-coating the cones with alkali or alkaline earth elements. 39 Prior to sample analysis the cones were conditioned by aspirating a 10 ppm Na solution for 10 min. Using this "coating" technique, the Li background generally ranged from <0.5 to 3 mV, approximately 0.1-0.75% of the sample signal intensity. Without utilising the Na wash, the Li backgrounds could increase to~100 mV, rendering it impossible to measure Li at the desired low concentration. In addition, nickel cones were preferred over platinum cones in the present study, as the latter suffered from higher and more rapidly increasing background levels, possibly due to less efficient Na-coating on the platinum.

| Mg isotopic measurements
The ratio of the three isotopes of Mg (viz: 24 Mg, 25 Mg and 26 Mg) were determined and bracketed against the DSM-3 standard. 40 Figure 2B). However, the concentration of Ti in river water is negligible and there are thus no detectable amounts of Ti, especially after further dilution for Mg isotope analysis. There is a slight asymmetry in the Na-peak visible on a log-scale (Figures 2A and 2B). The Na tailing does not drastically change between sample matrices, with similar magnitudes observed for river water samples and seawater with Na/Mg ratios~0.25 and~8, respectively. The Na tailing leads to co-elution of a few ng of Na in a~30 μg Mg peak, which is insignificant as it is further diluted by a factor of <150 before isotopic analysis on the Neptune mass spectrometer. The tailing can, however, be pronounced if the resin is not properly cleaned between successive sample passes. The isotopic ratios of Li and Mg are expressed in the δ-notation (‰) by the convention:

| Isotope ratio determination by MC-ICPMS: Analyses and reproducibility of standards
where X is either Li or Mg, H is the heavy isotope and L the light   Figure 5). The CN interference sits on the right-hand shoulder of the 26 Mg-peak and an offset of the H1 cup towards a higher mass, combined with peak-centering on 25   n is the number of analyses, which equals the number of column separations for this study. Studies using MC-ICPMS are preferentially referenced for comparison. In addition, for commonly used standards, studies with 10 or more analyses are included (for a more comprehensive list of references, see http://georem.mpch-mainz.gwdg.de).
Concentrations of samples and standards were therefore matched to within ±10% of each other in this study. Especial care was taken for Li as the 7

| Matrix element effects
The presence of matrix elements in the analyte may degrade the accuracy of Mg isotope ratio determinations in dry plasma conditions (e.g. 22,41,42 ), although instrumental mass bias in wet plasma appears to be less sensitive. 15,51,52 Matrix-induced mass bias is similarly recognised for Li isotope ratio determinations, especially when dry plasma is generated using an Aridus® membrane-containing desolvator. 31 Figure 9), and Ca concentrations up to 200 ppb did not cause systematic interference     53 The solid black line is the best-fit linear regression through the data set (slope = 0.5127, R 2 = 0.9997) and the dashed line is the theoretical equilibrium fractionation line 53 (slope = 0.521). DSM-3 solutions doped with matrix elements (circles) do not show observable deviation from the regression line but do, however, suffer from larger instrumental uncertainty than other purified samples