We developed and applied an analytical system for the simultaneous determination of magnesium/calcium, strontium/calcium, and manganese/calcium ratios in biogenic carbonates using a Perkin-Elmer 4300 Optima DV inductively coupled plasma-optical emission spectrophotometer in radial mode. Intensity ratio responses to calibration standard elemental ratios were linear over suitable working ranges. We evaluated instrumental response using solution replicates, and we analyzed planktonic foraminifera in multiple runs. Ca concentrations were kept in a relatively narrow range, typically ∼0.6–2.4 mM Ca, although we found no resolvable Ca matrix effect over a much wider range. Each analysis required a minimum sample volume of 300 μl and took 3.5 min instrument time. An analytical run of 30 samples plus standards, blanks, and control materials typically took ∼5–6 hours. Reproducibility (interrun precision, 1s %RSD) for solutions was 1.2%, 1.5%, and 1.2% for Mg/Ca, Sr/Ca, and Mn/Ca ratios, far smaller than reproducibility for foraminiferal replicates over multiple analytical runs. There was no independent reference material for assessing accuracy. Detection limits were equivalent for an 0.5 mM Ca solution to 0.058 mmol/mol for Mg/Ca, 0.015 mmol/mol for Sr/Ca, and ∼0.005–0.020 mmol/mol for Mn/Ca. Precautions included limiting length of time the peristaltic tubing for the autosampler was used, avoiding Mg contamination from plastic ware, and screening out results from samples with very low Ca concentrations (<0.25 mM Ca).
 A central goal of paleoceanography is reconstructing ocean temperature history. Studies of strontium/calcium (Sr/Ca) ratios in coral aragonite [e.g., Beck et al., 1992] and of magnesium/calcium (Mg/Ca) ratios in foraminiferal calcite [e.g., Nürnberg et al., 1996a, 1996b; Rosenthal et al., 1997; Hastings et al., 1998] documented the use of these tracers as records of past temperatures. Elemental ratios were determined by thermal ionization mass spectrometry (for coral Sr/Ca ratios) and by inductively coupled plasma-mass spectrometry (ICP-MS) and other techniques (for foraminiferal Mg/Ca ratios) because high precision on small samples was needed.
 Inductively coupled plasma-optical emission spectrometry (ICP-OES) has broad linear working ranges able to deal with the orders of magnitude concentration differences between Ca and minor elements in biogenic carbonates, and it offers advantages of high sample throughput and reasonable operating costs. Another major advantage of many ICP-OES instruments is simultaneous measurement of all analytes. Schrag  described precise measurements of Sr/Ca and Mg/Ca ratios in coral aragonite using a Jobin-Yvon-46 ICP-OES. De Villiers et al.  determined Mg/Ca and Sr/Ca ratios in foraminiferal calcite using a Varian Vista axially viewed plasma simultaneous ICP-OES with an intensity ratio calibration method to further improve accuracy and precision.
Schrag  and de Villiers et al.  found significant Ca matrix effects on their elemental ratio determinations, because Ca intensity responses were non-linear with Ca concentration and because matrix Ca concentrations affected other elemental intensity responses for these instruments. These factors required close matching of standard and sample Ca concentrations, as well as corrections, sometimes substantial, for Ca matrix differences between samples and standards.
 Technological advances in plasma technology have improved torch design and increased transfer of power from radio frequency generators (RF) to plasma gas, potentially decreasing matrix effects and improving response linearity. Brenner and Zander  reviewed different ICP-OES instruments and conditions for the impact of Ca concentrations on plasma robustness for axially and radially viewed plasmas. Ca matrix effects were minimized by increasing RF generator power (to >1.35 kW) and by decreasing argon carrier gas flow (to 0.5–0.7 l min−1). Although detection limits were typically better for axially viewed plasmas, measures of plasma robustness were generally higher for radially viewed plasmas [Brenner and Zander, 2000], making the use of radial view an advantage for minimizing matrix effects.
 The Perkin-Elmer 4300 Optima DV ICP-OES has improved RF generator and torch design compared to earlier models, making it ideal for reevaluating Ca matrix effect on minor element ratio determinations. Manganese/calcium (Mn/Ca) ratios in biogenic carbonates are indicators of diagenetic overgrowths in suboxic environments [Boyle, 1983], and are therefore of interest to complement Mg/Ca and Sr/Ca ratios. In this contribution, we developed and applied an analytical system for the simultaneous determination of Mg/Ca, Sr/Ca, and Mn/Ca ratios in biogenic carbonates using this instrument with a radially viewed plasma.
2. Instrumentation, Operating Conditions, and Materials and Methods
 We used a Perkin-Elmer Optima 4300 DV ICP-OES in radial configuration with MicroMist nebulizer (Glass Expansion) attached to a low volume cyclonic spray chamber (Cinnabar, Glass Expansion). The MicroMist nebulizer creates a narrow range of droplets allowing sensitivity to be maintained while minimizing sample volume. The low volume cyclonic spray chamber allows rapid washout, facilitating rapid sample throughput and minimizing sample equilibration time and volume. To limit dead volume in the sample uptake system, we constructed an autosampler sipper using PEEK tubing (0.01 in. interior diameter with 0.062 in. external diameter), reinforced with larger diameter PEEK tubing (0.080 in. inner diameter, 1/8 in. outer diameter), and minimized the length of all tubing connected to the nebulizer. The autosampler system uses a peristaltic pump to transport sample solutions to the nebulizer. We found that the PVC pump tubing (Watson Merlow orange red, ID 0.19 mm) has a finite lifetime of ∼6 hours on average. Tubing failure was characterized by rapidly dropping intensity responses, and this was easily monitored by tracking the Ca intensity response.
2.2. Operating Conditions
 To establish operating conditions, we used an 0.5 mM Ca solution in 0.5 N HNO3 matrix with a Mg/Ca ratio of 5 mmol/mol. We optimized RF generator voltage and sample gas flow to maximize plasma robustness, evaluating this using the Mg II 280.270 nm/Mg I 285.213 nm ionic/atomic line intensity ratio. The Mg II/Mg I ratio reflects the combined effects of plasma energy transfer, analyte residence time in the plasma, and plasma response to changes in excitation and atomization conditions, as well as changes in the chemical composition of aspirated solution [Mermet, 1989, 1991]. Higher ratios indicate more robust plasmas, with typical pure aqueous solutions having ratios >7 [Brenner and Zander, 2000].
 In optimization experiments, we obtained maximum Mg II/Mg I ratios with an RF generator voltage of 1.4 kW and a sample argon gas flow of 0.6 l min−1. A sample introduction pump rate of 0.16 (nominally 160 μl min−1) balanced maximizing the Mg II/Mg I ratio with minimizing sample consumption rate. In tests at these operating conditions, we found that the plasma maintained stable configurations, on the basis of stable Mg II/Mg I ratios, over a Ca concentration range of 0.25–4 mM (in 0.5 N HNO3 matrix), larger than our anticipated sample concentration range. Operating conditions for analytical runs were based on these optimization tests (Table 1). We monitored Mg II/Mg I variations during each analytical run, with anomalously low ratios within a run serving as indicators of analytical problems for those samples. However, we found variations in the Mg II/Mg I ratio from analytical run to analytical run, even running identical standard and control solutions (range observed from 6.7 to 4.7 over ∼4 months' time). Long-term variations in this ratio for this instrument are apparently influenced by other factors we have not identified.
Table 1. Instrumental Settings for Analytical Runs
Listed lines are recommended lines that avoid common interferences and maximize signal-to-noise ratios. Lines shown in bold are the most sensitive lines, and we used these lines in calculations. In tests we found no significant differences in using the other lines monitored for calculating results but did not routinely do so.
2.3. Reagents, Plastic Ware Cleaning, and Sample Handling
 We used glass distilled water, ACS grade acids (for cleaning; Fisher ACS grade Hydrochloric and Nitric Acid), trace metal grade nitric acid (for solutions; Fisher TraceMetal Grade Nitric Acid), and plasma grade primary standard solutions (SPEX CertiPrep Assurance). We found possible Mg contamination from plastic ware, especially at low solution concentrations of Ca and Mg and for any prolonged storage (greater than a few hours) of samples or diluted solutions of standards. Unrecognized Mg contamination at low solution Ca concentrations can be mistaken for instrumental matrix effects. Avoiding Mg contamination required careful cleaning of all containers used for samples and standards and limiting the amount of time diluted solutions were exposed to containers. Optimization tests used plastic ware cleaned by (1) rinsing in glass distilled water, (2) acid cleaning with 6 N HCl for one week, followed by rinsing with glass distilled water, (3) acid cleaning with 6 N HNO3 for one week, followed by rinsing with glass distilled water, and (4) storing in 0.5 N HCl, then rinsing with glass distilled water and drying in a laminar flow bench just prior to use.
 For sample runs, we modified the acid cleaning protocol to the following: (1) acid cleaning with 3 N HCl at 50°C for 12 hours, followed by rinsing with glass distilled water, and (2) acid cleaning for at least two weeks with 1 N HNO3, then rinsing with glass distilled water and drying in a laminar flow bench just prior to use. After initial cleaning, we used the same set of cleaned vials for preparing working standards for each analytical run.
 Foraminiferal samples (typical weights ∼0.1–1 mg) were cleaned in 0.5 ml centrifuge tubes (rinsed with distilled water before use) with the “Boyle method” protocol [Boyle and Keigwin, 1986; Boyle and Rosenthal, 1996] (see detailed description of different steps by Martin and Lea [2002, Appendix A]), including crushing, rinses, and reductive and oxidative steps. Samples were transferred to acid cleaned 1.5 ml tubes for one weak acid rinse in 0.001 N HNO3 (final step of the foraminiferal cleaning protocol), followed by dissolution in 0.5 N HNO3 (typical volume, 350–1500 μl, depending on final sample size). Minimum sample volume required for a single analysis is 300 μl. Sample solutions were run promptly and were not stored in these vials for more than six hours.
2.4. Standardization, Standards, and Typical Analytical Run
Rosenthal et al.  determined element/calcium ratios in biogenic carbonates using a sector field ICP-MS directly from intensity ratios using a single standard with matrix (Ca concentration, acid strength) matching samples. De Villiers et al.  developed an intensity ratio calibration method for ICP-OES, using a series of standards with different Mg/Ca and Sr/Ca ratios and of Ca concentrations bracketing those for samples. We applied these approaches by making a series of calibration standards at relatively high Ca concentrations (∼1050 ppm Ca, ∼26 mM Ca) with suitable Mg/Ca, Sr/Ca, and Mn/Ca ratios (Table 2).
S0 was made at a later time than the other five standards. The primary Ca solution used for S1–S5 had been entirely consumed, and we had to use a different batch of primary Ca solution for S0. Mg, Sr, and Mn concentrations measured in S0 were indistinguishable from 0.5 N HNO3 matrix blanks. We assumed this was true for the primary Ca solution used in preparing S1–S5, although we could not confirm this by measurements. We would recommend making all standards with the same Ca concentration from the same primary standard and at the same time.
We dilute S0 to the same Ca concentration as other standards for the working curves.
 Calibration standards were prepared in an air-dried, tared 500 ml volumetric flask, using glass distilled water and trace metal grade nitric acid to a final acid matrix of 0.5 N HNO3, with appropriate additions of SPEX CertiPrep Assurance Standards in HNO3 matrix (Ca, 10,000 mg/L, 5% HNO3; Mg, Sr, and Mn, 1000 mg/L, 2% HNO3). We used added weights of the Ca primary standard with the density of the primary standard solution to calculate volumes of Ca additions, and we used pipette weights delivered by multiple pipetting of distilled water and the density of distilled water to calculate volumes of Mg, Sr, and Mn additions. Final uncertainties on calibration standard elemental ratios, based on propagation of errors on weighing and pipette delivery and neglecting uncertainties in the concentrations of primary standards, were typically ≤0.2% RSD, and always ≤0.3% RSD. We assumed amounts of Mg, Sr, and Mn added with the primary Ca standard to the spiked standards were negligible relative to added amounts. Measured blanks for Mg, Sr, and Mn in our “zero standard,” made from a different bottle of Ca primary standard (see footnote to Table 2), were indistinguishable from those for 0.5 N HNO3 matrix. We also prepared a solution consistency standard (laboratory reference solution as a control material [International Union of Pure and Applied Chemistry (IUPAC), 1998]) of composition similar to that of typical foraminiferal calcite samples to use in assessing precision. This solution was not prepared with sufficient accuracy to quantitatively define its expected composition, although we recommend doing so.
 For each analytical run, we diluted a subset of at least four calibration standards (typically S0, S2, S3, and S4 for planktonic foraminiferal analyses) to working standard curves at two Ca concentrations (typically, 0.59 mM and 2.4 mM Ca, equivalent to ∼24 and ∼95 ppm Ca). These working standards were prepared fresh for each analytical run, and never used for longer than 12 hours. We ran the two different Ca concentration working curves at the beginning and end of each analytical run. Initial working curves were followed by matrix blanks, the solution consistency standard, then sample solutions. We interspersed solution consistency standard determinations throughout, typically after every ten samples. With each set of samples, we processed and ran as samples weighed aliquots of “foraminiferal consistency standards” as control materials [IUPAC, 1998]. Foraminiferal consistency standards were from single core samples from which we picked a large number of foraminifera of a given size fraction and species. Each individual ICP-OES analysis took 3.5 min, including 65 s uptake, 20 s acquisition (10 × 2 s exposure averaged), and 70 s wash (with pump rate set at 0.25 for wash, instead of 0.16 for uptake and acquisition). All lines were detected in simultaneous mode. Each analytical run, typically 30 samples, took ∼5–6 total hours, including time for plasma warm-up (∼1 hr), standardization, blank analyses, and consistency standard measurements.
 We constructed linear calibration curves of elemental intensity ratios versus elemental ratios in working standard solutions for each elemental ratio, using all working standard curves in a given analytical run to construct a single calibration line (Figure 1). For all analytical runs, we found linear calibration curves fit the working curve data, with r2 values consistently ≥0.999. We used these linear calibration curves to calculate elemental ratios for samples. We did not apply “drift correction” within or between runs. Because we calibrated to elemental ratios, uncertainties in dilutions of calibration solutions to working standards were irrelevant, and the use of multiple calibration standards minimized the effects of the uncertainty with which any one standard ratio is known.
 We did not do any corrections for Ca concentration differences between samples and standards. If a sample Ca concentration exceeded that of the highest Ca concentration working curve by more than 10%, we diluted the sample with additional 0.5 N HNO3 to bring its Ca concentration into the range of the working curves. We found that samples with Ca concentrations below 0.25 mM (10 ppm Ca, equivalent to ∼9 μg CaCO3 in 350 μl) sometimes gave unreliably high Mg/Ca ratios relative to replicates in the target Ca range, and we used this as a screening criterion for data reliability. This may result from difficulties in cleaning small samples and/or from instrumental response effects. This represents a limitation for generating Mg/Ca elemental ratio data from very small samples, such as individual foraminiferal tests, with this instrumental technique.
3. Evaluation of Analytical Responses
3.1. Analytical Runs
 We applied these procedures in two analytical runs to evaluate specific aspects of instrumental response and in multiple analytical runs determining elemental ratios in planktonic foraminiferal calcite. We used these analytical runs to assess precision as repeatability and reproducibility [IUPAC, 1998], Ca matrix effects on elemental ratio determinations, internal accuracy, and detection limits.
 The first instrumental response test evaluated precision as repeatability [IUPAC, 1998], defined as the reproducibility of measured values for a single solution in a single analytical run. We calibrated to a working curve of S0–S3 run at the beginning of the analytical run, with all working standards diluted to 1.0 mM Ca. We analyzed six vials of S1 at the same dilutions as the working standards (1.0 mM) as samples repeatedly (89 times) over a five-hour interval. We ran the six vials sequentially once each, followed by 15 sequential determinations for the first vial, 15 determinations for the second vial, 15 determinations for the third vial, sequential determinations of the six vials once each, 15 determinations for the fourth vial, 15 determinations for the fifth vial, and 2 determinations for the sixth vial (run terminated because of tubing lifetime). This treatment of “standard as a sample” also allowed an assessment of internal accuracy (i.e., how close is the measured result for a standard treated as a sample to the gravimetrically defined value), although not to an independent reference material.
 The second instrumental response test evaluated the effect of solution Ca concentration on instrumental response (Ca matrix effect) in a single analytical run. We diluted S3 (Table 2) to nine different Ca concentrations, from 26.1 mM (undiluted) to 0.25 mM. We calibrated to working curves of S0, S2, S3, and S4 at 0.59 mM and 2.38 mM Ca run at the beginning and end of the analytical run. We analyzed each S3 dilution in order of decreasing Ca concentration, running the block of nine dilutions three times sequentially with blank and consistency standard determinations interspersed between blocks. The third replicate of the 0.25 mM Ca solution (the lowest Ca concentration) was lost, however. We calculated elemental ratios as for samples from the working curves; note that the Ca concentration range covered by the working curves (0.59–2.38 mM) is significantly smaller than that for the solutions tested (0.25–26.1 mM). As in the first test, this treatment of “standard as a sample” also allows an assessment of internal accuracy.
 We did a number of analytical runs over ∼3.5 months for elemental ratios of planktonic foraminifera using this analytical system [Wara et al., 2002; Wara, 2003]. Replicate determinations of a laboratory reference solution were used to assess repeatability (intrarun precision) and reproducibility (interrun precision). We included foraminiferal consistency standards in each run to assess reproducibility for foraminiferal samples. We measured matrix blanks in each run to define blanks and detection limits.
 The repeatability test of 89 measurements of S1 dilutions over a five-hour interval (Figure 2) resulted in mean elemental ratios ±1s, with s = sample standard deviation (percentage repeatability standard deviation in parentheses) of 0.344 ± 0.001 mmol/mol (±0.39%) for Mg/Ca; 0.465 ± 0.001 mmol/mol (±0.23%) for Sr/Ca; and 0.073 ± 0.002 mmol/mol (±2.1%) for Mn/Ca. Replicate measurements of a laboratory reference solution in a number of analytical runs over ∼3.5 months resulted in excellent repeatability within runs and reproducibility between runs (Figure 3, Table 3). Repeatability, expressed as the percentage standard deviation ±1s %RSD (standard error, ±2s/√n, to correct for different n in each analytical run), was <1.4% (<1%) for Mg/Ca, typically <2.6% (<2%) for Sr/Ca, and <2.0% (<1.5%) for Mn/Ca. Multiple analytical runs resulted in percentage reproducibility standard deviations ±1s %RSD (±2s/√n) of ≤1.5% (≤0.20%) for all three ratios. Reproducibility standard deviations for each elemental ratio in the laboratory reference solution are larger than almost all of the repeatability standard deviations for those ratios in individual runs (Table 3, Figure 3), indicating that there is day-to-day variation different from the measurement error in a single analytical run. Finally, replicate measurements of three foraminiferal samples over the same ∼3.5 month interval (Figure 4, Table 4) found variability in all three ratios for foraminiferal replicates far larger than those for solution analyses. Absolute errors on foraminiferal Mg/Ca ratios were similar for the three species (0.23 to 0.32 mmol/mol), resulting in larger relative standard deviations for foraminifera with lower Mg/Ca ratios (Table 4). Foraminiferal Mn/Ca ratios had the largest percentage reproducibility relative standard deviations because mean values were close to detection limits (Table 4).
Table 3. Laboratory Reference Solution: Intrarun and Interrun Elemental Ratio Variability
Note that Mn/Ca ratio for this laboratory reference solution is significantly higher than that in the highest standard (by factor of ∼2.5) and significantly higher than those expected in foraminifera without significant Mn/Ca overgrowths (e.g., <0.2 mmol/mol [Boyle, 1983]). We recommend using a solution that more closely matches expected sample composition.
4.157 ± 0.042
1.177 ± 0.029
1.871 ± 0.018
4.068 ± 0.044
1.140 ± 0.055
1.868 ± 0.023
4.248 ± 0.029
1.196 ± 0.021
1.906 ± 0.035
4.093 ± 0.054
1.160 ± 0.022
1.844 ± 0.033
4.127 ± 0.020
1.164 ± 0.010
1.862 ± 0.007
4.129 ± 0.029
1.166 ± 0.008
1.869 ± 0.024
4.148 ± 0.039
1.167 ± 0.027
1.872 ± 0.022
4.127 ± 0.025
1.153 ± 0.008
1.866 ± 0.022
4.162 ± 0.012
1.161 ± 0.003
1.877 ± 0.009
4.130 ± 0.018
1.156 ± 0.003
1.865 ± 0.010
4.196 ± 0.006
1.156 ± 0.001
1.863 ± 0.003
4.123 ± 0.041
1.158 ± 0.012
1.857 ± 0.017
4.134 ± 0.028
1.160 ± 0.012
1.876 ± 0.017
4.013 ± 0.021
1.156 ± 0.005
1.865 ± 0.011
4.159 ± 0.008
1.155 ± 0.003
1.871 ± 0.005
4.109 ± 0.018
1.155 ± 0.006
1.852 ± 0.010
4.132 ± 0.022
1.158 ± 0.007
1.868 ± 0.015
4.212 ± 0.025
1.194 ± 0.008
1.887 ± 0.017
4.153 ± 0.038
1.163 ± 0.013
1.881 ± 0.032
4.129 ± 0.022
1.153 ± 0.015
1.858 ± 0.014
4.114 ± 0.017
1.158 ± 0.002
1.851 ± 0.016
4.122 ± 0.040
1.154 ± 0.016
1.849 ± 0.025
4.112 ± 0.023
1.158 ± 0.001
1.846 ± 0.013
4.103 ± 0.012
1.156 ± 0.003
1.846 ± 0.010
4.139 ± 0.022
1.162 ± 0.017
1.866 ± 0.019
4.126 ± 0.051
1.160 ± 0.018
1.865 ± 0.023
Table 4. Foraminiferal Consistency Standards: Elemental Ratios From Multiple Analytical Runs Over ∼3.5 Month Interval
Mean ± 1s, mmol/mol
Mean ± 1s, mmol/mol
Mean ± 1s, mmol/mol
BDL, below detection limit.
Globigerinoides sacculifer (without sacc) (V28-203, 24 cm, 355-435 μm)
3.684 ± 0.252
1.336 ± 0.025
0.018 ± 0.004
Pulleniatina obliquiloculata (V28-203, 24 cm, >150 μm)
2.391 ± 0.323
1.425 ± 0.030
0.017 ± 0.004
Neogloboquadrina pachyderma (sinistral) (80-30-14, 125 cm, >150 μm)
 The test of intrarun response to Ca concentration matrix effect using dilutions of S3 over a much wider range than typical for sample runs (Figure 5) resulted in mean elemental ratios ±1s (percentage repeatability standard deviation in parentheses) for all nine dilutions (n = 26) of 4.91 ± 0.05 mmol/mol (±0.98%) for Mg/Ca, 0.928 ± 0.011 mmol/mol (±1.14%) for Sr/Ca, and 0.726 ± 0.006 mmol/mol (±0.83%) for Mn/Ca. Using only the results from the three dilutions (n = 9) from 0.6–2.4 mM Ca bracketed by the working curve range resulted in ratios of 4.95 ± 0.02 mmol/mol (±0.39%) for Mg/Ca, 0.921 ± 0.006 mmol/mol (±0.65%) for Sr/Ca, and 0.732 ± 0.004 mmol/mol (±0.58%) for Mn/Ca, the same within error as the means over the much broader Ca range tested. Statistical tests confirmed that the differences between Mg/Ca ratios in the four lowest Ca concentration solutions compared to those in three highest Ca concentration solutions are highly significant. However, statistical tests also indicated that the differences in measured Mg/Ca ratios for the different vials of S1 at identical dilutions in the first instrumental test are highly significant, when no differences are expected, as are the differences in day-to-day measured Mg/Ca ratios for the laboratory reference solution (Table 3; Figure 3). Rather than indicating a quantifiable Ca matrix effect on Mg/Ca determinations, this indicates that, while the instrument has high precision on individual measurements, there is block-to-block variability within a run, run-to-run variability, and day-to-day variability. We concluded that any Ca concentration effect on Mg/Ca ratios in this experiment was not analytically meaningful.
 In our other investigations of sensitivity of elemental ratio determination to Ca matrix, including focusing in the much narrower range typically used for standardization and sample analysis, we were also unable to find any consistent Ca matrix effect exceeding analytical precision as defined by multiple measurements of a single solution over time. We concluded that linear calibration curves for both intensity versus elemental concentration and intensity ratio versus elemental ratio demonstrated that this instrument has no resolvable Ca matrix effect in the range tested, presumably the result of plasma characteristics and the use of radial view. Despite the lack of demonstrable Ca matrix effect, we continued to take the conservative approach of using a relatively narrow range of Ca concentrations for standards and samples.
3.2.3. Internal Accuracy
 We did not have an independent reference material to assess accuracy. We used the treatment of “standard solutions as samples” in the two instrumental tests to assess internal accuracy, defined as the difference of measured results for standard solutions treated as samples from gravimetric values for the standard composition (Table 2). For the first instrumental test running S1 repeatedly (Figure 2), the mean measured Mg/Ca ratio was 3.8% higher than gravimetric value, and Sr/Ca and Mn/Ca ratios were within +1.5% and −0.24% of the gravimetric values (Table 2). Intensity ratios were directly equivalent to elemental ratios, since no drift corrections were applied, so a comparison of the mean intensity ratio for the S1 dilutions treated as samples to the S1 intensity ratio in the initial working curve gives a direct indication of instrumental variation over the analytical run and thus potential accuracy. The mean Mg/Ca intensity ratio for all 89 measurements of S1 was 1.5% higher than the working standard S1 intensity ratio, and mean Sr/Ca and Mn/Ca intensity ratios were 0.90% and 1.8% higher than their working curve S1 intensity ratios. For the second instrumental analytical test, running various dilutions of S3 as samples (Figure 5), the measured means across the total Ca concentration range are −1.0%, +1.3%, and −0.80% of expected values for Mg/Ca, Sr/Ca, and Mn/Ca ratios (Table 2), respectively; means across the narrower concentration range of the working standards are even closer to gravimetric values (−0.1%, +0.61%, and +0.01%, respectively, for Mg/Ca, Sr/Ca, and Mn/Ca ratios).
3.2.4. Detection Limits
 The analytical system resulted in blanks and detection limits appropriate for the analysis of foraminiferal calcite at these dilutions (Table 5). Our ability to determine Mn/Ca ratios is limited to values >∼0.005–0.020 mmol/mol. Foraminiferal calcite unaffected by manganese carbonate overgrowths has Mn/Ca ratios typically <0.050 mmol/mol (and often much less). However, this analytical limitation is acceptable, as the purpose of measuring Mn/Ca is to screen for samples whose minor element composition might be adversely affected by manganese carbonate overgrowths with Mn/Ca ratios >0.100 mmol/mol.
Table 5. Blanks, Detection Limits, and Working Calibration Ranges for Planktonic Foraminiferal Analyses
Mean reagent blanks for 0.5 N HNO3 acid matrix used for sample dissolution estimated from conventional concentration versus intensity calibrations, observed over multiple batch runs. Measured blanks for dilutions of the S0 (Ca only) standard were indistinguishable from reagent blanks.
Blanks for Ca and Mg occasionally produced negative counts because the true baselines are not horizontal.
Detection limit estimated as three times the standard deviation of the reagent blank. Detection limits were used to estimate equivalent elemental ratio detection limits for samples with 0.5 mM Ca concentration. Samples at higher Ca concentrations will have higher detection limits.
Because of noisy baseline, this results in a relatively high detection limit for Mn. We also estimated Mn detection limit as three times standard deviation of a low Mn concentration solution, resulting in the lower detection limits shown.
Using S0 and S2–S4. The larger ranges in calibration standards (see Table 1) are useful for running other biogenic and diagenetic calcium carbonates.
 We developed and applied an analytical system with a PE 4300 DV ICP-OES in radial mode for the determination of Mg/Ca, Sr/Ca, and Mn/Ca ratios in biogenic and diagenetic carbonates, focusing here on its application to planktonic foraminiferal calcite. We used plasma robustness as measured by Mg ion/atom ratios to optimize operating conditions, also minimizing sample consumption. We used an intensity ratio approach to calibration with multiple calibration standards diluted to working standards at two concentrations (typically 0.6 and 2.4 mM Ca) for each analytical run. We combined all working standard curves in each run, finding linear instrumental responses for all three elemental ratios to Ca. We kept standard and sample Ca concentrations within this relatively narrow range as a precaution, although we found no resolvable Ca matrix effect on elemental ratio determinations over a much wider Ca concentration range. We did not apply any intrarun drift or matrix corrections, or any corrections for run-to-run variability. Analysis required a minimum sample volume of 300 μl (typical range 350–1500 μl, depending on foraminiferal sample size), equivalent to minimum sample weights after cleaning of ∼20–70 μg CaCO3. Each analysis required 3.5 min. In practice, we found that a complete analytical run of 30 samples, plus standards, blanks, and consistency standards, took ∼5–6 hours, including time for plasma warm-up.
 Repeatability (intrarun precision) for solutions of composition similar to foraminiferal samples (n = 3 to 17) was <1.4%, typically <2.6%, and <2.0% (1s percentage repeatability standard deviation) for Mg/Ca, Sr/Ca, and Mn/Ca ratios, with values as low as 0.13%, 0.12%, and 0.16% in individual runs, respectively. Reproducibility (interrun precision) for 249 measurements of a control solution in 25 analytical runs was 1.2%, 1.5%, and 1.2% (1s percentage reproducibility standard deviation) for Mg/Ca, Sr/Ca, and Mn/Ca ratios, respectively. Reproducibility for three foraminiferal samples measured repeatedly (n = 33 to 35) over 25 analytical runs as 1s standard deviation (percentage reproducibility standard deviation in parentheses) was 0.23–0.32 mmol/mol (6.8%–22%) for Mg/Ca, 0.03 mmol/mol (1.8%–2.2%) for Sr/Ca; and 0.004 mmol/mol (22%–24%) for Mn/Ca. We did not have an independent reference material for assessing accuracy. Treatment of two standard solutions as samples with Ca concentrations in the range of working curves, each in a separate analytical run, gave measured ratios within −0.1% to +3.8% of gravimetric values for Mg/Ca, within +0.61% to +1.5% for Sr/Ca, and within −0.01% to −0.24% for Mn/Ca. Confirmation of accuracy awaits the results of the ongoing foraminiferal Mg/Ca intercalibration exercise and the development of suitable matrix-matched reference materials. Typical detection limits, defined as three times the standard deviation of multiple measures of reagent blanks or a low concentration solution, for an 0.5 mM Ca solution were 0.058 mmol/mol for Mg/Ca, 0.015 mmol/mol for Sr/Ca, and ∼0.005–0.020 mmol/mol for Mn/Ca.
 Several precautions were necessary. First, tubing for sample transport on the peristaltic pump has a finite lifetime, with failure characterized by declining intensity responses. We changed the tubing at the start of each analytical run, and we did not use a single tubing piece for more than ∼6 hours. Second, Mg contamination from plastic ware can result in unrecognized blank effects that can be misdiagnosed as matrix effects. We used thorough acid cleaning protocols for containers, and we minimized the time solutions at sample concentration ranges were in contact with containers. Third, samples at low Ca concentrations can have apparently high Mg/Ca ratios, and we screened out samples with Ca concentrations <0.25 mM.
 This work was completed in the UCSC Plasma Analytical Facility, sponsored by UCSC Ocean Sciences, the Institute of Marine Sciences, Environmental Toxicology, and Earth Sciences. The PE Optima 4300 was purchased with funding from NSF OCE 99-77629 (MLD, principal investigator, with four co-principal investigators) and from various UCSC sources. This work received support from NSF OCE 98-19114 (MLD), NSF EAR-BE 01-20727 (MLD), an NSF Earth Sciences Post-Doctoral Research Fellowship (SAS), and NSF OCE 00-81697 (ACR). We thank J. Gille for statistical advice and two anonymous reviewers for extremely helpful and constructive reviews.