Automation of boron chromatographic purification for δ 11 B analysis of coral aragonite

Rationale: To detect the small changes in past pH , the boron isotope ratio of coral carbonates, expressed as the δ 11 B value, needs to be both precise and accurate (2sd <<1 ‰ ). Boron measurements by Multi-Collector Inductively Coupled Plasma Mass Spectrometry (MC-ICPMS) require the boron to be carefully purified before analysis, which is time consuming, and requires specialist training. Here, we use the prepFAST-MC method that enables the automatic extraction of B (up to 25 ng load) from a CaCO 3 matrix. Methods: Samples were purified using the prepFAST-MC automated system with a ~25- μ L column of Amberlite IRA743 resin. Boron isotope measurements were performed by MC-ICPMS. The effects of matrix load, speed of sample loading onto the column, and blank contamination were tested to evaluate the effects on the purification process. The optimised protocol was tested on various standards and samples of aragonite corals. Results: The blank contribution for the approach is ~60 pg and is negligible given our sample size (<0.2% sample size). Efficiency of matrix removal is demonstrated with the addition of up to 1.6 mg of dissolved low-B calcium carbonate to NIST SRM 951 with no impact on the accuracy of δ 11 B values. The Japanese Geological Survey Porites reference material JCp-1, boric acid standard NIST SRM 951, and seawater, all processed on the prepFAST-MC system, give δ 11 B values within error of literature values ( δ 11 B JCp-1 = 24.31 ± 0.20 ‰ (2sd, n = 20); δ 11 B NIST 951 = − 0.02 ± 0.15 ‰ (2sd, n = 13) and δ 11 B Seawater = 39.50 ± 0.06 ‰ (2sd, n = 2)). Results obtained from the coral Siderastrea siderea purified with the prepFAST-MC system show an average offset from the manual ion-exchange protocols of Δδ 11 B = 0.01 ± 0.28 ‰ (2sd, n = 12). Conclusions: Our study demonstrates the capacity of the prepFAST-MC method to generate accurate and reproducible δ 11 B values for a range of materials, without fractionation, with efficient matrix removal and with negligible blank contribution.


| INTRODUCTION
Measurements of atmospheric carbon dioxide (CO 2 ) over the last century have shown a significant increase with concentrations during the pre-industrial period (pre-mid 19th century) of around 280 ppm, reaching >400 ppm in 2015. This CO 2 rise has caused a decrease in surface seawater pH of~0.1 unit on average due to the absorption of anthropogenic CO 2 into the ocean, 1 and a strengthening of the greenhouse effect causing a rise in global mean surface temperature.
The pH decrease and temperature rise have impacted the health of some marine calcifying organisms, including coral reefs, that have experienced bleaching and decrease in skeletal extension or density (e.g. [2][3][4] ). However, instrumental records of pH are scarce and only go back a few decades. One way to reconstruct environmental parameters further back in time is to use indirect measurement of pH using, for example, the boron isotope ratio pH proxy in corals. This method has been used to capture existing records of surface pH change 5 and to extend the historical pH record (e.g. 6,7 ) but often at limited spatial and temporal resolution. To accurately and precisely evaluate environmental changes, pH and climate records are needed at high temporal resolution (e.g. a millennial resolution for geological time scales and sub-annual for historic timescales). The principal reason for the scarcity in boron isotope data to date is the labourintensive laboratory processes required during sample preparation for accurate boron analysis including material collection, clay and organic matter removal and, for some methodologies, time-consuming boron purification by skilled users in a boron-free clean laboratory.
The techniques require varying quantities of boron (2-5 ng for NTIMS; 1 μg for PTIMS, and 10-20 ng for MC-ICPMS), have different sample preparation protocols, and each is associated with different B isotope ratio precisions (from 0.1 to 0.7‰, 16 roughly equivalent to 0-0.1 pH unit). The MC-ICPMS technique in general has arguably significantly improved the measurement of many isotopic systems. 17 This approach offers many advantages when measuring dual isotope systems like boron because of the stability of the mass fractionation (although it is large: e.g. 16% per m/z unit for boron 17 ) and efficient ionisation (boron: 60-90%). The B isotopic composition of a sample is commonly expressed in the palaeoceanographic literature as: Previous studies have shown that MC-ICPMS allows the precise and accurate determination of δ 11 B values, in some instances to better than 0.25‰ (at 95% confidence), on as little as 10-20 ng of B. 16 The principal requirement of the MC-ICPMS methodology, however, is the need to purify the analyte prior to analysis to avoid interference and differential ionisation or instrumental mass fractionation between samples and the bracketing standard used to correct for it (e.g. 17,18

| Samples
In order to validate the accuracy of δ 11 B values using the prepFAST-MC method, several reference materials and samples have been processed including a biogenic coral reference material (JCp-1 24 : and the other with the prepFAST-MC system using the protocol established in this study following optimisation tests on in-house standards and reference materials. All prepFAST-MC data are compared with the respective longterm average at Southampton (for the reference materials) and the δ 11 B value of the same sample processed manually using the approach detailed elsewhere 15,16 and briefly described below.

| Boron separation
Boron purification for marine carbonates has traditionally been performed manually via anion-exchange chromatography 15,16,27,32 or microsublimation. [33][34][35][36] The approach followed here involved the Considering that most boron is collected after 300 μL, the volume of eluent can be adjusted depending on the equipment used and the volume required for the measurement of the δ 11 B value by MC-ICPMS. 15,16 The principle of boron purification with the prepFAST-MC system (described and illustrated further in Figure S1a  and air to be loaded and dispensed through the column. Each solution is temporarily stored in a coil before being dispensed to the location of interest (column, probe or waste). All steps are controlled by software and parameterised by the user.

| Boron isotope and elemental analysis
Boron isotope analyses were carried out on a Neptune MC-ICPMS system (Thermo Fisher, Bremen, Germany) with 10 12 Ω amplifier resistors for Faraday cups H3 ( 11 B) and L3 ( 10 B) at the University of Southampton using bracketing standards of NIST SRM 951 following methods described extensively elsewhere. 15,16 T A B L E 2 Protocol of boron purification for the standard manual method (a) and optimised protocol for the prepFAST automated method (b).
When not indicated, the direction of flow through the column is forward. See Figure S1 (supporting information) for reference to forward and reverse direction of flow a. Step

| Matrix wash-out and matrix effect
Two configurations of matrix wash-out were examined by varying the volume, flow rate and direction of flow of Milli-Q water through the prepFAST column to determine the protocol with highest matrix wash-out efficiency ( Table 3). The direction of flow was set in the forward direction (with respect to sample loading and elution, see Figure S1a, supporting information) for configuration 1 or alternated (forward and reverse) for configuration 2. The flow rate was also varied by washing out the matrix slowly (1000 μL/min, configuration 1) or rapidly (10 000 μL/min, configuration 2). As a comparison, the flow rate on the gravity column is unidirectional and depends on the hydraulic head in the column (averaging at 100 μL/min). We tested the influence of wash-out flow rate on matrix removal on NIST 951 (with Na-acetate buffer) for configuration 1 and JCp-1 (with Naacetate buffer) for configuration 2.
In the gravity column methodology described previously 15 Here an aliquot taken of six JCp-1 samples before and after purification was also measured on the ICPMS X-series to assess the levels of Na and Ca in the purified sample and to test how reliable Na is as an indicator of matrix removal.

| Flow rate of sample loading
The influence of the flow rate during sample loading was explored through altering the sample loading speed by varying the flow rate from 100, 200 and 500 μL/min. These tests were conducted with carbonate (JCp-1) and seawater samples as well as boric acid, to evaluate the influence of matrix type on the loading speed and the measured δ 11 B value.

| Boron blank
In a similar way to the manual method, total procedural blanks (TPBs)

| Efficiency of matrix wash-out
The different configurations of matrix wash-out (Table 3)

| Level of blank contamination
TPBs processed on the prepFAST-MC system ranged between 7 and 60 pg of boron (4 replicates, Figure 5) for the standard matrix washout configuration 2 (Table 3). These values are similar to published values in the literature for the standard manual approach using gravity columns 15,32 but tend to be slightly elevated compared with the standard manual approach currently in use at Southampton (ranging from below detectability to 40 pg 27 ). Nevertheless, this level of blank represents less than 0.2% of the sample boron concentration (for a 20 ng sample) and therefore requires no correction on the δ 11 B value.
The consistent and low TPB values (interspaced by 15-20 ng B samples) demonstrate little or no effect of carryover from one sample to another.

| Optimised method for marine carbonates
Based on our experiments, the optimal method (Table 2b)

| Reproducibility and external precision
Using the optimised method, results for JCp-1 and NIST 951 (for a 20 ng B sample size) show good reproducibility ( Figure 6A) with F I G U R E 4 Effect of sample introduction flow rate on δ 11 B values (Δδ 11 B = δ 11 B sample prepFAST − δ 11 B long term ) for standards and reference materials NIST 951 (red circles), JCp-1 (blue squares) and seawater (SW, green triangles). Dotted line represents an ideal offset of 0‰ between the two methods. Data given in Table S2 (supporting information) [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 5 Total procedure blank (TPB) processed with the prepFAST method (with matrix wash-out (MWO) configuration 2, see Table 4 for details) compared with published TPBs obtained by the standard manual column protocol at Southampton 27  Our data demonstrate that the prepFAST-MC method generates highly reproducible data for JCp-1, NIST 951 and seawater, similar to the manual gravity column method (long-term average at Southampton and previous studies, Table 2), with no significant mass fractionation induced by the automated B purification.

| Accuracy
To further test the prepFAST-MC method on real samples, The Na concentrations of diluted aliquots of each of these purified samples were also measured to confirm the low level of Ca and Na contamination (from the sample and buffer, respectively). As with the tests performed with standards (with wash-out configuration 2, Table 3), the prepFAST-MC method is associated here with consistently higher levels of Na (3 vs 0.5 ng on gravity columns); however, there is no significant relationship between Na and the offset in the δ 11 B value from certified or long-term average values from gravity columns (r 2 = 0.04, p = 0.20, Figure S3, supporting information).

| Effect of sample loading speed
Tests conducted with different sample loading speeds show that samples containing a carbonate or strong ionic matrix seem to be more sensitive to the flow rate than buffered boric acid (Figure 4). Since the observed fractionation is towards isotopically lighter (i.e. more negative δ 11 B) values, it appears that the heavy isotope ( 11 B) is preferentially lost when the flow rate of the analyte over the resin bed is too fast. This fractionation occurs although our sample yield did not reveal a significantly reduced boron content at the higher flow speeds (r 2 = 0.0006, p = 0.96; Figure S4a, supporting information) within the precision of these yield measurements (±10%).

| Matrix wash-out and blank contamination
Although it does not influence the measured δ 11 B value, there is a significant difference in the efficiency of matrix wash-out between the standard gravity method and the prepFAST-MC approach, with on average~0.5 ng of Na for the standard method compared with 3 ng (with configuration 2 in Table 3)

| Matrix effect
Despite documenting a matrix effect during the sample loading stage (increasing fractionation toward lower delta values when loading complex matricese.g. seawaterat high loading flow rate,

| Resin lifetime
The analyses here were performed using two separate resin beds.
Although no systematic study of resin lifetime was attempted, based on the accuracy found here, and the consistently low amount of boron in the elution check, the resin bed lifetime is judged to be a minimum of 60 samples (for a boron and CaCO 3 load of 20 ng and 1 mg on average, respectively) and we recommend regular checking of resin performance with carbonate reference materials or standards such as JCp-1 or NIST 951 and monitoring of the elution check. This level of performance is consistent with that obtained with gravity columns.

| CONCLUSIONS
Here we show that the prepFAST-MC system can automatically separate boron from a variety of matrices without significant isotopic fractionation. We find the following principal results. (i) The δ 11 B value is accurate and reproducible for boric acid, seawater and coral CaCO 3 provided that the sample is loaded slowly (100-200 μL/min) onto the column (Figures 4 and 6). (ii) For the majority of samples examined here, the matrix was washed off less efficiently using the prepFAST-MC method than with the standard manual gravity column method (Table 3); however, the excess of matrix monitored with Na did not impact the accuracy of the δ 11 B values ( Figure 6; and Figure S3, supporting information). Furthermore, switching to a buffer that does not contain significant cations (e.g. ammonium acetate) may alleviate this contamination issue. (iii) The amount of CaCO 3 matrix loaded onto the column demonstrates no correlation with the accuracy of the δ 11 B values (within the range 0-1.5 mg; Figure 3).
(iv) The level of blank during boron purification was low (<60 pg, Further work is needed to develop a prepFAST-MC method suitable for other carbonate materials commonly used in paleoceanographic studies (in particular foraminifera). But importantly, the prepFAST-MC method requires less training than for gravity columns and makes it possible to process one sample automatically every 60 min. The methodology described here will therefore speed up the sample throughput considerably for boron isotope analysis in carbonates applied to different areas such as paleoclimatology, oceanography and environmental sciences.