Chemical Geology 274 (2010) 187–195
Contents lists available at ScienceDirect
Chemical Geology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c h e m g e o
The accuracy of δ11B measurements of foraminifers Yunyan Ni a,b, Gavin L. Foster b,⁎, Tim Elliott b a b
Research Institute of Petroleum Exploration and Development, Petrochina, Beijing 100083, China Bristol Isotope Group, Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queens Road, BS8 1RJ, Bristol, UK
a r t i c l e
i n f o
Article history: Received 10 August 2009 Received in revised form 30 November 2009 Accepted 8 April 2010 Editor: B. Bourdon Keywords: Boron isotopes MC-ICPMS NTIMS Foraminifera Total evaporation
a b s t r a c t Values of δ11B reported in the literature for Holocene samples of the same foraminiferal species show large variability (6‰) relative to cited precision (b 1‰). This is indicative of significant inter-laboratory biases and raises concerns about the accuracy of foraminiferal proxy pH records. To investigate this problem we have measured the δ11B of modern ocean carbonates using several different analytical procedures. We have used total evaporation negative thermal ionisation mass spectrometry (TE-NTIMS), with various sample preparation and loading procedures and multi-collector inductively coupled plasma mass spectrometry (MC-ICPMS). The δ11B of pure boric acid solutions measured by TE-NTIMS and MC-ICPMS agree well, demonstrating no fundamental biases between the techniques. Yet the δ11B values measured by TE-NTIMS for non-foraminiferal carbonates are about 2‰ lighter than those by MC-ICPMS whereas foraminifers measured by the same standard TE-NTIMS procedure are 2 to 6‰ heavier than those by MC-ICPMS. Significantly, we found that foraminiferal samples kept in acidic solution over several months yielded lower δ11B values (up to 5‰ lower) and better reproducibility when re-measured by TE-NTIMS. We infer that organic material, released on foraminiferal dissolution, causes biases in NTIMS measurements. No residual organics should be present in the MC-ICPMS measurements as matrix is separated from sample before analysis. Moreover we demonstrate by standard addition the absence of any matrix influence in the MC-ICPMS procedure beyond the inherent uncertainties in the standard addition approach (∼ ±0.35‰ 2sigma). Degradation of the inferred organic residue, either by long term storage in acidic solution or by loading samples in 30% H2O2, reduces but does not totally remove the (variable) bias between TE-NTIMS and MC-ICPMS δ11B measurements. Despite these problems, careful analysis of similar samples by NTIMS may permit data to be obtained with consistent relative differences that still yield valuable proxy records, but there is clearly considerable potential for inaccuracies to result in such an approach from hard to detect changes in matrix. © 2010 Elsevier B.V. All rights reserved.
1. Introduction The boron isotopic compositions of modern oceanic carbonates can be rationalised using a simple inorganic fractionation model (Vengosh et al., 1991; Hemming and Hanson, 1992). Within this framework, measurements of the boron isotopic compositions of foraminiferal shells have the potential to provide much sought after records of palaeo-oceanic pH (Spivack et al., 1993, Palmer et al., 1998) and, with some additional inferences, the concentration of past atmospheric CO2 (e.g., Sanyal et al., 1995; Pearson and Palmer, 1999; Hönisch and Hemming, 2005; Foster 2008). Despite numerous studies on modern foraminifera that have confirmed the dependence of test δ11B on ambient pH (Sanyal et al., 1996, 2000, 2001; Hönisch et al., 2003; Zeebe et al., 2003; Hönisch and Hemming, 2004; Foster 2008),
⁎ Corresponding author. Now at School of Ocean & Earth Science, National Oceanography Centre, Southampton University of Southampton, Waterfront Campus, Southampton SO14 3ZH, UK. E-mail addresses:
[email protected] (Y. Ni),
[email protected] (G.L. Foster),
[email protected] (T. Elliott). 0009-2541/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2010.04.008
the exploitation of boron isotopes as a palaeo-pH-meter remains limited. A critical review of the assumptions involved in the δ11B pHproxy (Pagani et al., 2005) cast some doubts on its reliability, although one of the key uncertainties raised by Pagani et al. (2005), namely the magnitude of boron isotopic fractionation between boric acid and borate species in seawater, has now been determined precisely (Klochko et al., 2006). An important consideration not discussed by Pagani et al. (2005) is analytical inaccuracy associated with the mass spectrometric analysis. Notably Foster et al. (2006) highlighted a 6‰ range in isotope values for the same species of Holocene foraminifers reported by different laboratories. Significantly, such a spread in δ11B is too large to reflect the very minor variability in the conditions under which the different samples lived. The limited amount of B within hard-won, painstakingly picked and cleaned, mono-specific foraminferal samples led to the widespread use of negative-ion thermal ionisation mass spectrometry (NTIMS), which has very high ion yields (typically N5%). The NTIMS approach also has the advantage that no chemical separation of B from the sample matrix is required since the dissolved test can be directly loaded onto the filament to act as an activator for B ionisation. Such
188
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
‘chemistry-free’ analyses save both a considerable amount of time and avoid the additional blank contribution that would result from column chromatography, but leave open the possibility of matrix dependent effects on the B isotope measurements. In part, this concern led Foster et al. (2006) to develop a total evaporation NTIMS technique (TE-NTIMS). By analysing the whole sample Foster et al. (2006) hoped to remove any differential mass bias, caused by different matrices, which might result over the proscribed analysis window in conventional NTIMS. Improvements in sensitivity of multi-collector plasma mass spectrometry (MC-ICPMS) now allow a different means of foraminiferal boron isotope analysis. This method has the principal advantage that instrumental mass bias can easily and accurately be corrected by bracketing unknowns with standards of known isotopic composition, as in conventional stable isotopic analysis of oxygen and carbon. Although MC-ICPMS measurements require separation of B from sample matrix, this additional step has the advantage that it should remove any potential role for sample matrix in biasing analyses. Foster (2008) showed that boric acid standards analysed by MCICPMS and TE-NTIMS (using an inorganic loading solution as an activator with a composition similar to a dissolved foraminiferal test) gave indistinguishable δ11B values. Similarly, seawater samples analysed by both methods (using the seawater matrix as the activator in TE-NTIMS) also gave the same δ11B values (Foster et al., 2006). However, analyses of recent foraminifera by MC-ICPMS (Foster 2008) gave δ11B ∼2–6‰ lighter than values for similar samples reported by Ni et al. (2007) using TE-NTIMS and by Hönisch and Hemming (2004) or Palmer and Pearson (2003) using the more standard NTIMS approach. Foster (2008) suggested that variations in sample matrices and a contrast between the matrices of loaded sample and standard in NTIMS might cause this disagreement. In contrast, Kasemann et al. (2009) have recently shown that no biases exist between NTIMS and MC-ICPMS for non-foraminiferal carbonates (e.g. coral), suggesting the problem is particular to foraminiferal analysis rather than for CaCO3 samples in general. Here we further explore the possible causes for discrepancies between MC-ICPMS and NTIMS techniques and their implications for palaeo-pH measurements. 2. Materials and methods We report the boron isotopic compositions of a number of marine carbonates determined using total evaporation TE-NTIMS (see Foster et al., 2006) and MC-ICPMS (see Foster, 2008), which we also compare with previously published, standard NTIMS analyses (see Kasemann et al., 2009). The materials analysed were foraminiferal samples from ODP Site 806 and ODP Site 664 (Globigerinoides sacculifer, Globigerinoides ruber and Neogloboquadrina dutertrei species), together with a number of non-foraminiferal carbonate samples including corals (Porites (sp) and Desmophyllum (sp)) and calcite that formed part of the study by Kasemann et al. (2009). 2.1. Samples The two open ocean sample locations used for the foraminiferal work have been used in previous boron isotope studies (i.e. Palmer and Pearson, 2003; Hönisch and Hemming, 2004, 2005; Ni et al., 2007): ODP Site 806B (0°19.1′N, 159°21.7′E, water depth 2520 m) from the Ontong Java Plateau in the equatorial Pacific and ODP Site 664C (0°07′N, 23°14′W, water depth 3806 m) in the central equatorial Atlantic. Foraminifers were separated from a Holocene core-top sample using standard approaches, and the foraminiferal species and size fractions examined are listed in the following text. Sample M93-TB-FC-1 is a piece of massive Porites (sp) coral microatoll from the north coast of Papua New Guinea. The cold water coral Desmophyllum (sp) PS69/318-1 is from the Pacific sector of the Southern Ocean from a water depth between 1480 and 1788 m, and
the inorganic calcite sample UWC-1 (University of Wisconsin Calcite Standard; Graham et al., 1998) is a piece from a 10 cm cleavage rhomb of sky blue marble from the Valentine Wollastonite Mine in the Adirondacks (sample #88-V-1; Gerdes and Valley, 1994; see Kasemann et al., 2009 for detailed information). Sample CE-95 is a Schlerosponge Ceratoporella nicholsoni (supplied by F. Böhm, GEOMAR) that is also the in-house standard for the GEOMAR laboratory and has been analysed previously by TE-NTIMS (Foster et al., 2006), and sample C1-1 is a commercially purchased piece of a coral belonging to the Sarcophyton genus. 2.2. Sample preparation Foraminiferal samples were crushed between two pre-cleaned glass plates, transferred to an acid-leached 2 ml Teflon centrifuge tube and cleaned following a clay removal protocol similar to that described by Barker et al. (2003). For all of the samples, prior to dissolution, organic material was oxidised using a treatment with sodium hypochlorite (NaClO; 5% Cl), following previous boron isotope studies (e.g. Hönisch and Hemming, 2005; Foster, 2008). Following this oxidation step NaClO was removed by pipetting and any residual reagent was removed by repeated rinsing with MQ. The preparation for non-foraminiferal samples is similar to that used for the foraminiferal samples except there was no need to go through a clay removal step. 2.3. MC-ICPMS analysis For carbonates analysed by MC-ICPMS dissolution was achieved using 100–300 μl of 0.5 M HNO3. The amount of acid added was controlled by drop-wise titration until carbonate dissolution was complete (Foster et al., 2006), which minimises excess H+. Boron was separated using micro-columns of Amberlite 743 boron specific resin and the purified boron was analysed using a sample-standard bracketing technique on a ThermoFinnigan Neptune MC-ICPMS. A full description of the protocols used can be found in Foster (2008). Full procedural replicates of an in-house coral standard M-93-TB-FC-1 (supplied by Simone Kasemann) carried out during the course of this study indicate that our external reproducibility on δ11B for this approach is better than ± 0.25‰ (95% confidence). Foster (2008) assessed the accuracy of the MC-ICPMS approach in several ways. Firstly, the more likely sources of inaccuracy, such as the effect of analytical blank and interferences (i.e. from Ar4+), were carefully investigated and dealt with at the outset. Secondly, in order to ensure the column procedure for separating boron did not cause any isotopic fractionation (e.g. Lemarchand et al., 2002). Foster (2008) processed several boric acid standards through the entire procedure. No significant isotopic fractionations were observed and column processed NIST SRM 951 gave a δ11B = −0.04 ± 0.16‰ (2 s.d.), i.e. well within uncertainty of unprocessed standard. Foster (2008) further added Ca and other trace elements (found in typical foraminiferal proportions) to NIST SRM 951 prior to chemical processing and once again no significant isotopic fractionation was observed in such ‘samples’ relative to the pure standard (δ11B = 0.01 ± 0.19‰; 2 s.d.). Importantly, all samples are screened for matrix contamination (approximated by Na intensity) prior to analysis for δ11B to ensure they are cation free. The effectiveness of chemical separation is indicated by the removal of 2 M sodium acetate buffer used in the first step of column chemistry to less than ∼ 50 ppb Na in measured samples. Foster (2008) has further demonstrated that there is no influence on measured δ11B of residual Na up to concentrations of 1 ppm, considerably in excess of the values in samples analysed. Tipper et al. (2008) noted that although measurement of ‘synthetic samples,’ as used by Foster (2008), provide a valuable indicator of the reliability of a separation procedure, not all
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
potential matrix problems of a complex natural sample are necessarily envisaged. To demonstrate further the robustness of the whole B measurement protocol, therefore, we have followed the example of Tipper et al. (2008) and undertaken a standard addition analysis. Four samples were prepared from a cleaned, mixed species Miocene foraminiferal carbonate (a fresh preparation of the “871std” of Foster et al., 2006) and mixed with variable amounts of NIST SRM 951 prior to column chemistry. The amount of foraminiferal carbonate used for each mixture was kept constant and at a value typical of our foraminiferal samples. The concentrations of the end members were determined using matrix matched standards and ICPMS. We assign a conservative uncertainty of 4% (at 95% confidence) to these boron concentration determinations. All weighing was performed on a six figure balance and we quadratically added a conservative 1% weighing uncertainty to the 4% concentration uncertainty in these final proportions to estimate the total uncertainty in these mixing proportions. 2.4. TE-NTIMS analysis The TE-NTIMS analytical approach used here largely follows that described in detail in Foster et al. (2006) and Ni et al. (2007). Briefly, foraminiferal and carbonate samples are carefully dissolved in the minimum amount of 2 M HCl and loaded directly (thus in a predominantly CaCl matrix) on an outgassed Re filament. Samples were analysed to exhaustion on a ThermoFinnigan Triton mass spectrometer. The advantages of this approach, compared to the standard NTIMS method, are that collection of the whole sample, rather than just the initial fraction, should minimise machine induced mass fractionation and enable very small amounts of boron to be analysed. Indeed, the minimum sample size is limited by loading blank. In this study, like in Ni et al. (2007) we have loading blanks of b5 pg and typically analyse around 300–400 pg of boron. Our external reproducibility for carbonate samples, assessed from repeat analysis of our in-house foraminiferal standard (871std) during the course of this study, was ±1.2‰ (2 s.d.; 95% confidence). For reasons elaborated below, it is important to note that this solution was made at the very beginning of our study and was already several months old when the earliest of these data were collected. Some modifications to this standard analysis procedure made during the course of the study are detailed in Section 3.4. 2.5. Trace element analysis Splits (∼ 10%) of the dissolved samples were measured for calcium and other trace element concentrations on a ThermoFinnigan ELEMENT 2 ICPMS following Ni et al. (2007) and Foster (2008). This provides a means to see if foram samples were cleaned effectively (e.g. Al/Ca b100 µmol/mol) as well providing an accurate measurement of solution strength to allow consistent sized loads for NTIMS and help intensity matching for sample-standard bracketing by MCICPMS. 3. Results In the following, all boron isotope data are reported in the delta notation relative to NIST SRM 951 using the following relationship: 02 11 δ B = @4 11 11
B = 10 B
3 sample
B=10 BNIST951
1
5−1A × 1000
For all TE-NTIMS data, we use 11B/10B951 = 4.0396 ± 0.0028 as determined by Foster et al. (2006) using standard addition. We note
189
that subsequently Foster (2008) determined 11B/10B = 4.0396 ± 0.0017 (2 s.d.) for NIST SRM 951 measured by TE-NTIMS using an inorganic Ca matrix (with a ‘foraminiferal’ trace element mix and b100 ppt B). For the MC-ICPMS analysis, since the samples are corrected for instrumental mass bias by bracketing with NIST SRM 951, they are reported in delta notation without recourse to further normalisation. Uncertainties in NIST SRM 951 measurements are thus automatically included in MC-ICPMS δ11B data, whereas this contribution should be separately included in NTIMS measurements but is frequently ignored and not taken into account in quoted uncertainty. For all TE-NTIMS analyses shown here we quadratically add the uncertainty of NIST SRM 951 (0.7‰) to the reproducibility of our inhouse standard (1.2‰) to give an overall uncertainty of 1.4‰ at 95% confidence. 3.1. Standard addition for MC-ICPMS The results of the standard addition experiment for MC-ICPMS are shown in Table 1 and Fig. 1. The two end members, diluted boric acid (no matrix) and B separated from foraminiferal carbonate, clearly lie on a regression line fitted to the four mixtures of the endmembers (determined using a least squares regression using Isoplot; Ludwig, 2000). It is important to note that all mixtures and the foraminiferal carbonate end member have been processed through the full chemical separation procedure but the NIST SRM 951 has not. From this good fit we therefore confirm the findings of Foster (2008) and have additional confidence that the chemical separation of B from the matrix induces no bias in the δ11B of foraminiferal and other carbonate samples. What is more, the good linear fit suggests variable B:matrix ratios also do not impact the accuracy of this approach. 3.2. Influence of different matrices on NTIMS In order to examine the role of different matrices on NTIMS boron isotope analysis, where there is no matrix removal prior to measurement, analyses were made of corals, sclerosponge, inorganic calcite, seawater (Table 2 and Fig. 2) and foraminifers (Table 3 and Fig. 3). Previously reported δ11B values of boric acid standards and seawater analysed by TE-NTIMS and MC-ICPMS are within analytical uncertainty (Foster, 2008). Likewise δ11B values for inorganic calcite, and corals measured by standard NTIMS agree with the MC-ICPMS data within analytical uncertainty (Kasemann et al., 2009; Fig. 2), but the TE-NTIMS data for these carbonates are systematically (∼ 1.8‰) lighter than those by MC-ICPMS (Fig. 2). In contrast, the δ11B values for the foraminiferal samples (G. ruber (white), G. sacculifer (with final sac-like chamber), and N. dutertrei measured by the TE-NTIMS are heavier than those measured by MC-ICPMS by 2 to 6‰ (Fig. 3).
Table 1 Boron isotopic data for standard additions by MC-ICPMS. fsamplea
Errorb
δ11B-1c (‰)
2 s.e.c
δ11B-2c (‰)
2 s.e.c
δ11B-avd (‰)
2σd
0.225 0.435 0.640 0.826 1
0.008 0.011 0.010 0.006
3.38 6.69 10.12 12.87
0.09 0.09 0.09 0.08
3.28 6.40 9.98 12.94
0.09 0.09 0.08 0.11
3.33 6.55 10.05 12.91 15.65e
0.25 0.25 0.25 0.25 0.17e
a
Fraction of sample in each mixture between sample and NIST SRM 951 standard. Uncertainty in mixing proportions based on a quadratic addition of an assumed 1% weighing uncertainty and 4% measurement uncertainty on [B] of the end members. This uncertainty is at the 95% confidence level. c First and second analysis of each mixture for δ11B. 2 s.e. is the internal precision expressed as 2 standard errors. d Average δ11B of each mixture. The uncertainty (at 95% confidence) is the external precision based on the long term reproducibility of our in-house carbonate standards. e The average and 2 s.d. of repeat (n = 4) total proceedural replicates of the foraminiferal sample. b
190
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
Fig. 1. Standard addition between an in-house foraminiferal standard and NIST SRM boric acid, plotted as fraction of sample in the mixture (fsample). The regression line and associated error envelope, determined using Isoplot (Ludwig, 2000), was fitted through the sample-standard mixtures only (grey ellipses). Uncertainties on mixing proportions, δ11B measurements, the slope and the intercept are all at the 2sigma level. For each mixture the amount of matrix was kept constant and the amount of boric acid was varied. The mixtures (with constant matrix) and in-house foraminiferal standard were processed through our full chemical separation procedure, NIST SRM 951 was not. MSWD = mean square weighted deviation.
3.3. Temporal changes in δ11B values measured by TE-NTIMS During the course of the study, an “ageing” phenomenon was found for boron isotope analyses of solutions of the same foraminiferal samples measured by TE-NTIMS over the course of several months to 1 year. Eight Holocene foraminiferal samples from ODP Site 664 and ODP Site 806 prepared in the standard manner and stored in the acidic loading solution (2 M HCl) were analysed several times over a period of approximately 1 year (Table 4). As shown in Fig. 4, the boron isotopic composition generally becomes lighter over this period of storage.
3.4. Different preparation strategies We selected one foraminiferal sample (ODP Site 806 coretop, G. sacculifer (with final sac-like chamber from the 300–355 μm size fraction) that reproduced particularly poorly using the standard protocol for further investigation. Three separate dissolutions (A–C) using standard techniques gave δ11B values of 22.7 ± 1.4‰ (A; 2 s.e.; n = 3), 23.7 ± 1.9‰ (B; 2 s.e.; n = 2), and 21.3 ± 1.4‰ (C; 2 s.e.; n = 3) with an overall average δ11B = 22.6 ± 1.4‰ (2 s.e.). Importantly, when aged for around 1 year and reanalysed the δ11B of dissolution B had decreased by ∼ 5‰ to a δ11B of 18.5 ± 0.2‰ (2 s.e.; n = 3; Table 4).
Fig. 2. Comparison of δ11B values for corals, calcite and sclerosponge measured by MCICPMS, NTIMS (blue diamonds) and TE-NTIMS (red circles). There is a good agreement between NTIMS (Kasemann et al., 2009) and MC-ICPMS but the δ11B values measured by the TE-NTIMS are constantly offset to lighter values by ∼1.8‰ (red bar) than those of MC-ICPMS. All data are plotted as the difference between NTIMS and MC-ICPMS vs. the value for the same sample measured by MC-ICPMS. Error bars are at the 2 sigma level and include a quadratic addition of both analytical uncertainties.
We tried to replicate the ageing of this sample by dissolving the sample in 12 N HCl rather 2 M HCl. We then heated the samples in sealed, screw-top Teflon vials at 150 °C for several hours. By thus refluxing the sample we aimed to hydrolyse effectively any foraminiferal derived organic material (predominantly polysaccharides; Langer, 1992) that was released from the test interior during dissolution and thus escaped initial oxidation by the NaClO (5% Cl) cleaning step. Three additional splits (1 to 3) of the same dissolved sample were heated for 5, 10 and 15 h. Fig. 5 shows the results of this reflux treatment on measured δ11B. The refluxing appears to do little to change the ratio or improve the precision of the measurements. Also, it is possible that refluxing enhances the dissolution of residual clays, with associated boron. Enhanced clay breakdown should be evident in as an increase in Al/Ca, but as can be seen in Table 5, the refluxing for various periods of time and the dissolution in 12 M HCl does not result in elevated Al/Ca. In addition to the reflux treatment each split was loaded with 1 μl of RoMil-UpA ultra-pure H2O2 (b20 ppt B, which equates to a b0.4‰ blank effect which is insignificant given our analytical precision) in an attempt to oxidise any organic matter on the filament. An average δ11B of the three refluxed and H2O2 loaded samples is 19.6 ± 0.2‰ (2 s.d.) (Table 4). This value is close to that of the ‘aged’ sample and clearly the analyses reproduce much better (Fig. 5). To confirm that loading with H2O2 is responsible for this change in δ11B two additional samples were prepared, cleaned and then dissolved in 12 M HCl and loaded with H2O2 but without refluxing. These two samples also give an average δ11B of 19.6‰ (Fig. 5 and Table 5).
Table 2 Boron isotopic data for carbonate samples measured with three different techniques. Sample
Sample ID
MC-ICPMS δ11B (‰)a
Ceratoporella nicholsoni sclerosponge Porites coral Inorganic calcite Sarcophyton coral Desmophyllum coral a
CE-95 M93-TB-FC-1 UWC-1 C1-1 PS69/318-1
19.8 24.6c 7.5c 17.2 15.5c 11
10
TE-NTIMS 2σa
δ11B (‰)a
0.25 0.25 0.25 0.25 0.25
b
19.4 22.0 6.6 14.4 12.6
NTIMS 2σa
δ11B (‰)a
2σa
0.9 1.4 1.4 1.4 1.4
b
20.4 25.0c 8.6c
0.6 0.8 0.8
15.9c
0.8
Permil deviation from NIST SRM 951. TE-NTIMS data are normalised to B/ BTE = 4.0396. NTIMS data are normalised as described in relevant paper. Uncertainty at 95% confidence based on the long term reproducibility of standards. For the TE-NTIMS data we have also propagated the uncertainty in the ratio of NIST SRM 951. b Data from Foster et al. (2006). c Data from Kasemann et al. (2009).
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
191
Table 3 Boron isotopic data for foraminifera samples measured with three different techniques. ODP Site
Details
Species
MC-ICPMS δ11B (‰)a
Site 806B Site 806B Site 664C Site 664C Site 664C Site 999A Site 999A 871 stdd
1H-1, 1H-1, 1H-1, 1H-1, 1H-1, 1H-1, 1H-1,
12–17 cm 12–17 cm 10–12 cm 10–12 cm 10–12 cm 10–12 cm 10–12 cm
G. sacculifer (w) 500–600 μm G. ruber (white) 300–355 μm G. sacculifer (w) 500–600 μm G. ruber (white) 300–355 μm N. dutertrei 425–500 μm G. ruber (white) 300–355 μm G. sacculifer (w) 500–600 μm
c
19.60 20.31c 19.82c 20.32c 16.90c 20.67c 20.28c 15.65
Normal TE-NTIMS
Oxidative TE-NTIMS
2σb
δ11B (‰)a
2σb
δ11B (‰)a
2σb
0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25
24.0 22.8 24.0 23.3 18.3
1.4 1.4 1.4 1.4 1.4
18.0
1.4
21.4 20.3 20.9 20.7 18.4 22.1 22.3 15.9
1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4
a
Permil deviation from NIST SRM 951, TE-NTIMS data are normalised to 11B/10BTE = 4.0396. b 2 sigma uncertainty based on either the long term reproducibility of relevant in-house carbonate standard or 2 standard errors (2 s.d./sqrt(n)) of repeat measurements (n) of individual samples, which ever is larger. c Data from Foster (2008). d In-house foraminiferal carbonate standard.
4. Discussion 4.1. Accuracy of MC-ICPMS As discussed above, Foster (2008) concluded on the basis of a number of experiments that δ11B by MC-ICPMS were accurate, but that study did not provide a rigorous assessment of the potential influence of residual sample matrix, post-column separation, that may cause the standard-sample routine to be inaccurate (Tipper et al., 2008). The standard addition data presented in Fig. 1 directly address this possibility. All the mixtures lie on a regression line that extrapolates through both the end member compositions. The end member compositions are thus constrained by the error on the regression, which is quite small (±0.35‰ on the lower intercept; 2 sigma) and comparable to the external precision of the MC-ICPMS technique (±0.25‰; see also Tipper et al., 2008). Thus we believe our MC-ICPMS δ11B measurements of foraminifera are accurate at this level of sample reproducibility (i.e. 0.25‰) and therefore better than typical NTIMS precision (0.8‰ at 95% confidence). Hence, we can usefully compare NTIMS and MC-ICPMS measurements to assess the accuracy of the former. 4.2. Matrix effects in δ11B measurements by NTIMS 4.2.1. Influence of bulk matrix composition Foster (2008) measured the boric acid standards NIST SRM 951, JAA and JA-B (JABA and JABB of Aggarwal et al., 2009) by TE-NTIMS and
Fig. 3. Comparison of δ11B values for foraminiferal samples measured by MC-ICPMS, TENTIMS (red circles) and NTIMS from the literature (diamond, cross, and triangle). All data are plotted as the difference between NTIMS and MC-ICPMS vs. the value for the same sample measured by MC-ICPMS. Error bars are at the 2 sigma level and include a quadratic addition of both analytical uncertainties. Foraminiferal analysed by all NTIMS techniques to date are heavier than MC-ICPMS by up to 6‰.
MC-ICPMS and found no significant differences between the δ11B values produced by the two techniques. All the non-foraminiferal carbonates in this study demonstrate a constant lower δ11B value (by ∼1.8‰) compared to those by MC-ICPMS but importantly the relative differences between samples are roughly the same. This suggests that perhaps the 11B/10B value of NIST SRM 951 used (4.0396) to generate the TE-NTIMS delta values is inappropriate for these carbonate samples. Foster (2008) used a boron-free (b100 ppt B), artificial ‘foraminiferal’ matrix as an ionisation enhancer for all TE-NTIMS measurements and so did not introduce a difference between the loading matrix of reference NIST SRM 951 and the unknowns. Notably, the absolute value obtained for NIST SRM 951 using this technique, 11 10 B/ B = 4.0396 ± 0.0017 (2 s.d.), is within error of the certified value (Catanzaro et al., 1970), implying negligible instrumental bias in these measurements. However, all the non-foraminiferal analysed here have B/Ca ratios much lower (90 to 500 μmol/mol; Kasemann et al., 2009), than the artificial ‘foraminiferal’ matrix used an ionisation enhancer by Foster (2008), which has B/Ca N2000 μmol/mol (600 pg/μl of B in with 24 mmol/L Ca solution). The larger relative amount of calcium matrix may influence the ionisation behaviour to account for the observed 1.8‰ discrepancy between TE-NTIMS and MC-ICPMS measurements. Unlike TE-NTIMS, the standard NTIMS measurements of the δ11B of non-foraminiferal carbonates agree well with MC-ICPMS (Fig. 2). This may be a consequence of the 0.5 μl of boron-free seawater added to each carbonate sample to enhance ionisation (Hemming and Hanson, 1994) and which also makes the sample matrix closer to the pure seawater matrix used for the analysis of standards (Kasemann et al., 2009). We suggest that the influence of the seawater activator over-rides any subtle sample matrix effect, such as changes in B/Ca. This approach is not used in TENTIMS as Foster et al. (2006) found that organics introduced in preparing B-free seawater made TE-NTIMS measurements unreliable. Standard NTIMS does not suffer from this problem as only a portion of the total emitted ion beam is collected, and during an initial phase of heating organic material is allowed to burn off (see Kasemann et al., 2001).
4.2.2. Potential role of “organics” The foraminifers analysed by TE-NTIMS have similar B/Ca ratios to each other and yet they yield variably heavy δ11B when compared to MC-ICPMS (Fig. 3), therefore bulk compositional differences between standard and foraminiferal bulk matrix cannot be invoked to account for this variable inaccuracy. A trace component is implicated and we suggest the organic material originally hosted within the foraminifera plays a role. Although we thoroughly oxidise the foraminifera before dissolution, internal components (such as the Primary Organic
192
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
Table 4 Boron isotopic data for foraminiferal samples analysed by TE-NTIMS over an extended time period. Sample ID
Species
664B24 664B27 806-9 806-10 806-16 806-26 806-27 806-29
N. dutertrei 425–500 μm N. dutertrei 425–500 μm G. sacculifer (w) 500–600 μm G. sacculifer (w) 500–600 μm G. ruber (white) 300–355 μm G. sacculifer (w) 500–600 μm G. sacculifer (w) 500–600 μm G.sacculifer (w) 500–600 μm
Initial measurement
After 3 months
After 1 year
δ11B (‰)a
2σb
δ11B (‰)a
2σb
19.5 20.2 23.3 20.6 24.0 21.6 22.0 23.7
1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.9
18.3 19.0
1.4 1.4
δ11B (‰)a
2σb
20.6 20.1 22.6 19.9 19.9 18.5
1.4 1.4 1.4 1.4 1.4 1.4 1.4 1.4
a
Permil deviation from NIST SRM 951 normalised to 11B/10BTE = 4.0396. 2 sigma uncertainty based on either the long term reproducibility of the in-house carbonate standard or 2 standard errors (2 s.d./sqrt(n)) of repeat measurements (n) of individual samples, which ever is larger. Uncertainty includes a propagation of the uncertainty in the normalising ratio of NIST SRM 951. b
Membrane) likely survive such processing and are only released once the carbonate matrix is dissolved. Organic contaminants can influence the measured δ11B in two ways. Firstly, molecules of 12C14N16O− have a mass of 42 and can interfere directly with the measurement of δ11B since with NTIMS boron is analysed as 11B16O16O− (mass 43) and 10B16O16O− (mass 42). The presence of such an interference may be reflected by elevated count rates at mass 26, e.g. 12C14N− (mass 26) and results in a lower than expected isotope ratio (see Hemming and Hanson, 1994; Foster et al., 2006). The second way in which organics can influence δ11B by NTIMS is by affecting ionisation. As discussed by Hemming and Hanson (1994) organic contaminants tend to retard ionisation and result in a low boron ion yield. Poor reproducibility is a further consequence as fractionation characteristics vary depending on the amount of organics loaded. Namely, a sample with a large organic load requires longer heating before the organics are burnt off and stable emission is achieved. During this time the boron is likely to have fractionated, by Rayleigh-like evaporation, more than a sample that achieves stable emission earlier with fewer organics. Total evaporation is not immune to such a problem as during the early phase of the analysis, the presence of organics will suppress ionisation of the isotopically light fraction. Thus the integrated signal will be weighted towards the heavier, residual boron that is more efficiently ionised once the organics have been burnt off. Conceptually, the heavy δ11B we see for foraminifera measured by TE-NTIMS thus seem consistent with a role of residual organics. During the course of our previous TE-NTIMS studies we found that typically 1 in 3 dissolutions exhibited worse than expected repro-
Fig. 4. δ11B values of various planktonic foraminifers from ODP Site 806 and ODP Site 664 determined by TE-NTIMS with standard protocol and reanalysed after a period of between 1 year to 3 months. All samples were kept in screw-top Teflon vials. Error bars are at the 2 sigma level.
ducibility. Crucially in some instances this poor reproducibility was not accompanied by the presence of high counts (N5000 cps) of CN−, and conversely, high counts of CN− were not always accompanied by poor reproducibility, making the monitoring of mass 26 not very definitive with respect to assessing the reliability of analysis. Moreover, the presence of 12C14N16O− from ‘organic-rich’ samples would make their apparent δ11B isotopically light rather than heavy. Further empirical support for the importance of residual organics in biasing NTIMS measurements comes from the “ageing phenomenon” and our modified sample loading techniques. Our attempts to break down any organics by hydrolysis with concentrated HCl seemed unsuccessful, whereas ageing and oxidative loading resulted in lower measured δ11B and improved reproducibility. From this we infer that oxidation, either slowly during storage or forced by H2O2 more effectively removes the organic component. Both standard NTIMS (i.e. by others) and TE-NTIMS give heavier δ11B for foraminifera than MC-ICPMS (Fig. 3), we therefore believe these experiments all indicate that for reliable boron isotope analysis by NTIMS it is first necessary to breakdown residual foraminiferal organic material released by dissolution. The results shown in Fig. 5 imply that this can be achieved quite simply by loading the dissolved sample with 1 μl of 30% H2O2. In the MC-ICPMS technique intrinsic
Fig. 5. δ11B values determined by TE-NTIMS of repeat treatments of G. sacculifer (with final sac-like chamber) from the 300–355 μm size fraction from the coretop of ODP Site 806. Each symbol represents a separate dissolution treated in a number of ways. Triangles are samples dissolved in 12 M HCl, refluxed for between 5 and 15 h and loaded directly (without H2O2). Diamonds are samples dissolved in 12 M HCl, refluxed for between 5 and 15 h and loaded with 1 μl H2O2. Circles are samples dissolved in 12 M HCl and just loaded with 1 μl of H2O2 (no refluxing). The grey bar shows the δ11B for a sample analysed by TE-NTIMS using the standard protocol and the red bar shows the δ11B analysed after a year of storage in a screw-top Teflon vial. Clearly samples loaded with 1 μl H2O2 give much better reproducibility, lighter δ11B ratios, and are close to the “aged” δ11B value.
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
193
Table 5 Boron isotopic data for samples analysed by TE-NTIMS with several kinds of pre-treatment. Sample ID
Refluxed time (hours)
806-34-1 806-34-2 806-34-3 806-45 806-46
5 10 15 0 0
δ11B (‰)a directly loaded
δ11B (‰)a loaded with 1 μl of H2O2
Mg/Ca
Al/Ca
B/Ca
Run 1
Run 2
Average
2σb
Run 1
Run 2
Average
2σb
mmol/mol
mmol/mol
mmol/mol
21.6 24.2 22.6
19.2 20.0 20.8
20.4 22.1 21.7
2.4 4.2 1.7
19.4 19.9 19.8 20.1 18.7
19.3 19.8 19.6 19.4 20.2
19.4 19.9 19.7 19.8 19.4
1.4 1.4 1.4 1.4 1.5
3.89 3.88 3.90
24.61 20.88 21.53
69.65 70.26 71.15
a
Permil deviation from NIST SRM 951, 11B/10BTE = 4.0396. 2 sigma uncertainty based on either the long term reproducibility of the in-house carbonate standard or 2 standard errors (2 s.d./sqrt(n)) of repeat measurements (n) of individual samples, which ever is larger. Uncertainty includes a propagation of the uncertainty in the normalising ratio of NIST SRM 951. b
organic matrix should be removed during B separation chemistry and any residue is likely destroyed during plasma ionisation. The standard addition results shown in Fig. 1 further demonstrate that organics cannot significantly bias the MC-ICPMS data. Fig. 6 shows a much more acceptable correspondence between TENTIMS δ11B values using H2O2 loading and MC-ICPMS, although clearly there is still a tendency for the TE-NTIMS data to be heavy (+1.2‰ on average) by slightly more than the typical reproducibility of most samples. Whilst we have only analysed foraminifera by TENTIMS, all published NTIMS Holocene foraminiferal data are similarly isotopically heavy with respect to MC-ICPMS. This observation suggests, perhaps not unsurprisingly, that the role of residual organics is generating instrumental bias relative to organic free standard matrices is a ubiquitous problem for foraminiferal analyses by NTIMS approaches.
4.2.3. Previous assessment of TE-NTIMS accuracy Foster et al. (2006) had attempted to assess the accuracy of TENTIMS using standard addition. Importantly their procedure used a constant amount of B in each analysis but a different amount of matrix, in contrast to the constant matrix but variable B used in this study. Their approach was constrained by the limited scope for varying the amount of B loaded in TE-NTIMS. Too little boron suffers from a significant loading blank correction, whereas too much requires prohibitive analysis times to run a sample to exhaustion. Foster et al. (2006) analysed mixtures of NIST SRM 951 with both variable amounts of seawater and “871std” solution of dissolved foraminifera. Both standard additions yielded δ11B of NIST SRM 951
reassuringly within error of each other and the certified value of Catanzaro et al. (1970). Moreover Foster et al. (2006) obtained a δ11B for seawater that is consistent with determinations by other analytical methods, including positive thermal ionisation mass spectrometry (Aggarwal et al., 2004 and references therein) and more recent MCICPMS measurements (Foster 2008). This suggests that measurements with seawater as an activator induced no analytical bias to TENTIMS and at face value the same might be assumed for the foraminiferal matrix. However, in the standard addition approach used by Foster et al. (2006), if an analytical bias imparted by the matrix is linearly dependent on the amount of matrix loaded, the extrapolation of the best fit of sample-standard mixes to zero matrix should return the correct value for the pure standard even though a bias is imparted under typical measurement conditions (see Fig. 7). This appears to have been the case in the study of Foster et al. (2006). Lulled into a false sense of security by the lack of matrix dependence for the seawater mix and a common value for NIST SRM 951 derived from both standard additions, Foster et al. (2006) assumed no matrix dependence for the foraminiferal matrix. As demonstrated above, this is clearly not the case and was not a robust inference from their data. 4.3. Implications for reported intra-species δ11B variability The discussion above suggests that the accuracy of the TE-NTIMS approach is questionable, but the advantage of this high sensitivity technique may still be potentially exploited if the relative differences between foraminiferal species are reproducible between different techniques. Fig. 8 shows the δ11B values by the TE-NTIMS with standard protocol (normal TE-NTIMS), with oxidative loading (refluxed and loaded with H2O2) and MC-ICPMS for the planktonic foraminifers G. sacculifer (with final sac-like chamber; 500–600 μm), G. ruber (300–355 μm) and N. dutertrei (355–425 μm) from Site 664 and Site 806 all treated similarly (Table 3). The general differences in δ11B between surface (G. sacculifer and G. ruber) and thermocline (N. dutertrei) dwelling species is evident in all three approaches (Fig. 8). Moreover, the oxidative loading TE-NTIMS approach brings the differences between the species into better agreement with MCICPMS. Thus in cases where sample size is very limited and where relatively large differences (N1‰) in δ11B are significant, this preliminary work suggests that TE-NTIMS using an oxidiative loading technique may still be of some value. 5. Conclusions
Fig. 6. δ11B values of various foraminiferal samples measured by the MC-ICPMS technique vs. TE-NTIMS with the standard protocol (circles) and with oxidative loading (TE-NTIMS with reflux treatment and loaded with H2O2, labelled as oxidative TENTIMS). All data are plotted as the difference between TE-NTIMS and MC-ICPMS vs. the value for the same sample measured by MC-ICPMS. Error bars are at the 2 sigma level and include a quadratic addition of both analytical uncertainties. Note that the oxidative loading tends to cause the TE-NTIMS values to be closer to the MC-ICPMS values but they are still heavier by around + 1.2‰.
We have demonstrated the ability of MC-ICPMS to generate accurate δ11B measurements of foraminifera, using a standard addition approach, to within the uncertainty of that procedure (±0.35‰; 2 sigma). In contrast, we show that δ11B measured by TE-NTIMS is sensitive to sample matrix with non-foraminiferal carbonates systematically ∼ 2‰ lower than MC-ICPMS and
194
Y. Ni et al. / Chemical Geology 274 (2010) 187–195
Fig. 7. Cartoon schematically illustrating the influence matrix can have on the standard addition approach. In a the mixtures of sample and standard are made such that the amount of matrix is constant. For line A there is no matrix effect and the regression line passes through the pure foraminiferal and boric acid end members. For line B the sample matrix induces an offset of + 2‰, since the amount of matrix is constant and in this case the matrix effect scales with the amount of matrix, all the mixtures and the pure sample are offset by + 2‰ and the intercept is elevated accordingly. In b the mixtures are made such that the amount of matrix scales with the fraction of sample in the mixture. For line C there is no matrix effect. For line D, as the amount of matrix decreases the offset, which is proportional to the amount of matrix in the mixture, also decreases. Unfortunately this causes the lower intercept to be correct but the value for the foraminifera to be inaccurate.
foraminiferal samples variably heavier by up to 6‰. We suspect the former is a result of the different B/Ca of sample and standard loading matrices; a problem that could likely be overcome with more careful matrix matching or via secondary normalisation. The latter appears to be a consequence of organic material released during sample dissolution. The TE-NTIMS δ11B of samples stored in dilute acid (2 M HCl) decreases over a period of months and a comparable effect can be achieved by loading the sample with 1 µl concentrated H2O2. We believe both processes oxidise this residual organic material. Oxidative loading brings TE-NTIMS δ11B measurements of foraminifera in closer agreement with MC-ICPMS analyses (b2‰) but the discordance is still greater than measurement precision. Very careful protocols are therefore required to ensure that these inaccuracies in the NTIMS approach are kept at a minimum or at least constant. This seems to be a difficult goal using TE-NTIMS, and whilst more conventional NTIMS approaches may provide more flexibility in analysis strategy, the ubiquitously heavier δ11B reported for literature NTIMS measurements of recent foraminfera relative to MC-ICPMS indicates that this is an issue that needs to be addressed.
Fig. 8. Comparison of δ11B values of G. sacculifer (with final sac-like chamber), G. ruber and N. dutertrei from ODP Site 806 and ODP Site 664 determined by the TE-NTIMS with standard protocol (standard TE-NTIMS; red squares), TE-NTIMS with reflux treatment and H2O2 (oxidative TE-NTIMS; blue open squares) and MC-ICPMS (blue filled circles).
Acknowledgements We thank B. Hönisch for long-standing open discussion on these issues, S. Kasemann and F. Böhm for supplying some of the materials used in this study. This work was funded by a NERC Advanced Fellowship to Foster, NERC grant NER/A/S/2001/01213, ORS fellowship supplemented by the Department of Earth Sciences, University of Bristol to Ni and National Science Foundation of China (grant no. 40703014).
References Aggarwal, J.K., Mezger, K., Pernicka, E., Meixner, A., 2004. The effect of instrumental mass bias on δ11B measurments: a comparison between thermal ionisation mass spectrometry and multiple-collector ICP-MS. International Journal of Mass Spectrometry 232, 259–263. Aggarwal, J.K., et al., 2009. How well do non-traditional stable isotope results compare between different laboratories: results from the interlaboratory comparison of boron isotope measurements. Journal of Analytical Atomic Spectrometry 24, 825–831. Barker, S., Greaves, M., Elderfield, H., 2003. A study of cleaning procedures used for foraminiferal Mg/Ca paleothermometry. Geochemistry, Geophysics, Geosystems 4 (9), 8407. doi:10.1029/2003GC000559. Catanzaro, E.J., et al., 1970. Boric assay; isotopic, and assay standard reference materials. US National Bureau of Standards, Special Publication 260-17 (70 pp.). Foster, G.L., 2008. Seawater pH, pCO2 and [CO2− 3 ] variations in the Caribbean Sea over the last 130 kyr: a boron isotope and B/Ca study of planktic foraminifera. Earth and Planetary Science Letters 271, 254–266. Foster, G.L., Ni, Y., Haley, B., Elliott, T., 2006. Accurate and precise isotopic measurement of sub-nanogram sized samples of foraminiferal hosted boron by total evaporation NTIMS. Chemical Geology 230, 161–174. Gerdes, M.L., Valley, J.W., 1994. Fluid flow and mass transport at the Valentine wollastonite deposit, Adirondack Mountains, N.Y. Journal of Metamorphic Geology 12, 589–608. Graham, C.M., Valley, J.W., Eiler, J.M., Wada, H., 1998. Timescales and mechanisms of fluid infiltration in a marble: an ion microprobe study. Contributions to Mineralogy and Petrology 132, 371–389. Hemming, N.G., Hanson, G.N., 1992. Boron isotopic composition and concentration in modern marine carbonates. Geochimica et Cosmochimica Acta 56, 537–543. Hemming, N.G., Hanson, G.N., 1994. A procedure for the isotopic analysis of boron by negative thermal ionization mass spectrometry. Chemical Geology 114, 147–156 (Isotope Geoscience Section). Hönisch, B., Hemming, N.G., 2004. Ground-truthing the boron isotope–paleo pH proxy in planktonic foraminifera shells: partial dissolution and shell size effects. Paleoceanography 19 (PA4010). doi:10.1029/2004PA001026. Hönisch, B., Hemming, N.G., 2005. Surface ocean pH response to variations in pCO2 through two full glacial cycles. Earth and Planetary Science Letters 236, 305–314. Hönisch, B., et al., 2003. The influence of symbiont photosynthesison the boron isotopic composition of foraminifera shells. Marine Micropalentology 928, 1–10. Kasemann, S., et al., 2001. Boron and oxygen isotope composition of certified reference materials NIST SRM 610/612 and reference materials JB-2 and JR-2. Geostandards Newsletter 25 (2–3), 405–416.
Y. Ni et al. / Chemical Geology 274 (2010) 187–195 Kasemann, S., Schmidt, D.N., Bijma, J., Foster, G.L., 2009. In situ boron isotope analysis of marine carbonates and its application for foraminifera and palaeo-pH. Chemical Geology 260, 138–147. Klochko, K., Kaufman, A.J., Yoa, W., Byrne, R.H., Tossell, J.A., 2006. Experimental measurement of boron isotope fractionation in seawater. Earth and Planetary Science Letters 248, 261–270. Langer, M.R., 1992. Biosynthesis of glycosaminoglycans in foraminifera: a review. Marine Micropaleontology 19, 245–255. Lemarchand, D., Gaillardet, J., Gopel, C., Manhes, G., 2002. An optimized procedure for boron separation and mass spectrometry analysis for river samples. Chemical Geology 182, 323–334. Ludwig, K.R., 2000. Users Manual for Isoplot/Ex v.2.3. A Geochronological Toolkit for Microsoft Excel. Berkeley Geochronological Center Special Publication No. 1a, Berkeley, California (57 pp.). Ni, Y., et al., 2007. A core top assessment of proxies for the ocean carbonate system in surface-dwelling foraminifers. Paleoceanography 22 (PA3212). doi:10.1029/ 2006PA001337. Pagani, M., Lemarchand, D., Spivack, A., Gaillardet, J., 2005. A critical evaluation of the boron isotope-pH proxy: the accuracy of ancient pH estimates. Geochimica et Cosmochimica Acta 69 (4), 953–961. Palmer, M.R., Pearson, P.N., 2003. A 23,000-year record of surface water pH and pCO2 in the Western Equatorial Pacific Ocean. Science 300, 480–482. Palmer, M.R., Pearson, P.N., Cobb, S.J., 1998. Reconstructing past ocean pH–depth profiles. Science 282, 1468–1471.
195
Pearson, P., Palmer, M.R., 1999. Middle Eocene seawater pH and atmospheric carbon dioxide concentrations. Science 284, 1824–1826. Sanyal, A., Hemming, N.G., Hanson, G.N., Broecker, W., 1995. Evidence for a higher pH in the glacial ocean from boron isotopes in foraminifera. Nature 373, 234–236. Sanyal, A., et al., 1996. Oceanic pH control on the boron isotopic composition of formanifera: evidence from culture experiments. Paleoceanography 11 (5), 513–517. Sanyal, A., Nugent, M., Reeder, R.J., Bijma, J., 2000. Seawater pH control on the boron isotopic composition of calcite: evidence from inorganic calcite precipitation experiments. Geochimica et Cosmochimica Acta 64 (9), 1551–1555. Sanyal, A., Bijman, J., Spero, H., Lea, D.W., 2001. Empirical relationship between pH and the boron isotopic composition of Globigerinoides sacculifer: implications for the boron isotope paleo-pH proxy. Paleoceanography 16 (5), 515–519. Spivack, A.J., You, C.-F., Smith, H.J., 1993. Foraminiferal boron isotope ratios as a proxy for surface ocean pH over the past 21 Myr. Nature 363, 149–151. Tipper, E.T., Louvat, P., Capmas, F., Galy, A., Gaillardet, J., 2008. Accuracy of stable Mg and Ca isotope data obtained by MC-ICP-MS using the standard addition method. Chemical Geology 257, 65–75. Vengosh, A., Y.K., Starinsky, A., Chivas, A.R., McCulloch, M.T., 1991. Coprecipitation and isotopic fractionation of boron in modern biogenic carbonates. Geochimica et Cosmochimica Acta 55, 2901–2910. Zeebe, R.E., Wolf-Gladrow, D.A., Bijma, J., Hönisch, B., 2003. Vital effects in foraminifera do not compromise the use of δ11B as a paleo-pH indicator: evidence from modeling. Paleoceanography 18 (2), 1043.