LA-ICP-MS depth profiling perspective on cleaning protocols for elemental analyses in planktic foraminifers

Authors


Abstract

[1] Measurements of trace metal ratios in foraminiferal calcite are routinely used to reconstruct paleoceanographic conditions. Analyses using solution-based inductively coupled plasma mass spectrometry (ICP-MS) require dissolution of the entire foraminifer shell. The potential exists for contamination from adherent clays, mineralized coatings, and other diagenetic components that confound the biogenic trace metal signal. We present results from a cleaning experiment on fossil specimens of the planktic foraminifer Orbulina universa that were cracked into several shell fragments and subjected to different cleaning protocols. We use laser ablation ICP-MS (LA-ICP-MS) depth profiling to evaluate the effects of reductive, oxidative, and chelating (diethylene-triamine pentaacetic acid) cleaning protocols on shell Mg/Ca and Ba/Ca ratios. Using the natural pattern of intrashell Mg/Ca heterogeneity exhibited by O. universa, we demonstrate that reductive and oxidative cleaning can dissolve shell calcite from available surfaces, although intrashell Mg/Ca minima and maxima are unaffected. High-resolution depth profiles can be used to identify areas of heterogeneous intrashell Ba/Ca, which can be excluded from computations of whole-shell Ba/Ca. The size and density of shell pores plays a major role in the degree of contamination from sedimentary material. We demonstrate an approach for computing whole-shell Me/Ca ratios from LA-ICP-MS depth profiles that accounts for potential contamination and diagenetic overprinting.

1. Introduction

[2] Trace metal ratios in the calcite shells of fossil foraminifers have been used in paleoclimate studies to evaluate paleotemperature (Mg/Ca) [Anand et al., 2003; Dekens et al., 2002; Lea et al., 1999; von Langen et al., 2005], pH and carbonate chemistry (U/Ca, B/Ca) [Allen et al., 2011; Russell et al., 2004; Sanyal et al., 1997], redox conditions (V/Ca, Mn/Ca, U/Ca) [Boiteau et al., 2012; Glock et al., 2012; Hastings et al., 1996; Russell et al., 1996], and salinity (Mg/Ca, Ba/Ca) [Arbuszewski et al., 2010; Ferguson et al., 2008; Hall and Chan, 2004; Hoogakker et al., 2009; Weldeab et al., 2007]. Reconstructions from trace metal ratios are based on the assumption that analyses are measuring primary biogenic calcite, but diagenetic overprinting under different depositional conditions may complicate measurement of Me/Ca ratios in primary calcite using solution inductively coupled plasma mass spectrometry (ICP-MS). Foraminifers with high-Mg overgrowths of inorganic calcite have been found in well-oxygenated depositional environments [Boussetta et al., 2011; Regenberg et al., 2007; van Raden et al., 2011]. Suboxic to anoxic sediments are high in Mn, Fe, and other redox-sensitive metals [Froelich et al., 1979; Klinkhammer, 1980] and may promote the formation of diagenetic phases with high concentrations for these elements. In addition to in situ overgrowths, clays and other sedimentary material can form adherent coatings that act as contaminants when measuring a biogenic signal in foraminifer calcite, if not removed before analysis [Benway et al., 2003; Haley and Klinkhammer, 2002; Pena et al., 2005, 2008; Torres et al., 2010].

[3] Several cleaning protocols are used to isolate primary calcite, including sonication in methanol to remove adherent fine clays and other sedimentary particles; a reductive step to remove oxide coatings rich in trace elements [Boyle, 1981]; an oxidative step to remove organic matter [Boyle and Keigwin, 1985]; and a final weak acid leach. An additional cleaning step using diethylene-triamine pentaacetic acid (DTPA) is sometimes employed as an alkaline chelating agent prior to Ba/Ca and Cd/Ca analyses [Boyle and Keigwin, 1985; Lea and Boyle, 1989]. Several cleaning experiments have sought to quantify the effects of each of these steps on measured trace metal ratios [Barker et al., 2003; Bian and Martin, 2010; Martin and Lea, 2002; Yu et al., 2007], and some studies show unequivocal evidence of diagenetic coatings with high concentrations of diagenetic trace elements [Haley and Klinkhammer, 2002; Pena et al., 2005].

[4] Due to the widespread use of Mg/Ca paleothermometry, many studies have focused on the relative effects of different cleaning treatments on Mg/Ca ratios [Barker et al., 2003; Weldeab et al., 2006]. Cleaning experiments demonstrate that whole-shell Mg/Ca ratios systematically decrease during the reductive cleaning step [Barker et al., 2003; Benway et al., 2003; Bian and Martin, 2010; Klinkhammer et al., 2004; Martin and Lea, 2002; Yu et al., 2007]. A comparison of the oxidative cleaning step on sediment trap and live-cultured material shows a decrease in Mg/Ca ratios between specimens cleaned with sodium hypochlorite [Lea et al., 1999] and buffered hydrogen peroxide [Pak et al., 2004; Russell et al., 2004]. High-Mg coatings have been documented on outer [Pena et al., 2005] and inner shell surfaces [Ferguson et al., 2008], and high-Mg components are removed from shell material during progressive acid leaching [Benway et al., 2003; Klinkhammer et al., 2004; Sadekov et al., 2010]. Proposed explanations vary for the decrease in Mg/Ca with increasingly aggressive cleaning treatments, including the removal of authigenic or diagenetic coatings that are high in Mg [Benway et al., 2003], preferential dissolution of high-Mg phases within the calcite lattice [Brown and Elderfield, 1996; Rosenthal et al., 2004; Yu et al., 2007], or active dissolution and removal of primary calcite during cleaning [Barker et al., 2003; Sadekov et al., 2010].

[5] In contrast to Mg/Ca, the use of Ba/Ca as a paleosalinity proxy [Hall and Chan, 2004; Plewa et al., 2006; Sprovieri et al., 2008; Weldeab et al., 2007] is not as widespread, so cleaning techniques vary. The DTPA chelating step has traditionally been used to remove both contaminant coatings and layers of diagenetic calcite for ICP-MS and ICP-optical emission spectroscopy analyses [Lea and Boyle, 1989-1991] as well as flow-through time-resolved analyses [Haley and Klinkhammer, 2002]. However, cleaning experiments show that the DTPA chelating step has variable effects on measured Ba/Ca [Martin and Lea, 2002]. In recent paleoceanographic applications of Ba/Ca, some studies have continued to include the DTPA cleaning step [Hall and Chan, 2004; Plewa et al., 2006], while others have omitted this step [Weldeab et al., 2007].

[6] Conventional solution-based ICP-MS and flow-through approaches analyze bulk samples of foraminifer shells, and thus produce average geochemical information from multiple individuals. Laser ablation ICP-MS (LA-ICP-MS) depth profiling through the shell walls of individual foraminifers can be used to visualize the intrashell distribution of trace elements within shell calcite [Eggins et al., 2003; Hathorne et al., 2003] and to identify potential diagenetic overgrowths and contaminant phases in foraminifer shell walls [Hathorne et al., 2009; Pena et al., 2005, 2008]. Sadekov et al. [2010] used LA-ICP-MS to quantify changes in intrashell Mg/Ca patterns in foraminifers that were partially dissolved in the laboratory to simulate seafloor dissolution, and suggested that variation in Mg/Ca is due to preferential removal of shell calcite from surfaces exposed during dissolution.

[7] A direct comparison between fragments of the same shell that have undergone different treatments is needed in order to evaluate the effects of different cleaning protocols. The planktic foraminifer Orbulina universa grows a multichambered, trochospiral juvenile shell, and then secretes a single, spherical adult chamber that calcifies on an initial primary organic membrane (POM) and thickens over the course of several days [Spero, 1988]. Mg/Ca heterogeneity is a natural characteristic of O. universa [Eggins et al., 2003; Spero et al., 2012], and LA-ICP-MS depth profiles on multiple fragments of the same O. universa shell demonstrate that the intrashell Mg/Ca pattern is reproducible between fragments [Sadekov et al., 2010]. These observations indicate that multiple fragments from the same individual O. universa shell can be compared directly in different experiments, and provide the opportunity to evaluate Mg/Ca changes due to reagent cleaning. In O. universa, Ba is not biologically active; the Ba/Ca is only dependent on the Ba/Ca of seawater, and does not vary due to salinity, temperature, pH, [CO32−], or other oceanographic factors [Hönisch et al., 2011]. LA-ICP-MS depth profiles on live-cultured specimens of O. universa demonstrate that intrashell Ba/Ca is constant across the shells of individuals calcified under controlled conditions [Vetter et al., 2013]. Measured Ba/Ca depth profiles on samples from sediment cores thus enable us to examine potential intrashell heterogeneity in fossil O. universa due to diagenetic overprinting.

[8] In this paper, we present the results of a cleaning experiment on Mg/Ca and Ba/Ca in specimens of the planktic foraminifer O. universa from both anoxic and oxic depositional settings, where different types of diagenetic overprinting may occur. Multiple fragments of each individual O. universa shell were subjected to different cleaning protocols (methanol and water sonication and rinses only, methanol sonication plus reductive plus oxidative cleaning, and methanol sonication, reductive plus oxidative cleaning, and a DTPA step). We use LA-ICP-MS depth profiling to quantify the intrashell spatial distribution of Mg and Ba in each shell fragment. Our results demonstrate that although reductive and oxidative cleaning techniques remove portions of primary calcite, LA-ICP-MS analyses may be used accurately to compute whole-shell Me/Ca ratios.

2. Experimental Approach and Methodology

[9] We evaluated Mg/Ca and Ba/Ca ratios in fossil O. universa from two different depositional environments where mineralized diagenetic overprinting may differ. We use specimens from the western Caribbean Sea, where bottom waters are relatively high in O2, to represent normal depositional conditions, and samples from the Orca Basin in the Gulf of Mexico to represent anoxic bottom waters. The Orca Basin is bounded by salt diapirs [Shokes et al., 1977] and episodically contains anoxic, hypersaline bottom waters on glacial/interglacial timescales [Leventer et al., 1983]. During these anoxic intervals, the concentration of redox-sensitive elements in bottom-water brines poses the possibility of in situ diagenetic coatings rich in Mn, Fe, Ba, and other trace elements. For this study, 10 specimens were selected from Orca Basin core EN32–PC6 (26°56.8′N; 91°20.0′W) from a laminated, anoxic deglacial interval (583–585 cm, >350 μm size fraction) [Leventer et al., 1983]. Ten specimens were selected from Caribbean core ODP 999A (12°44.64′N; 78°44.36′W, 2838 m; 110–112 cm core depth) to represent open-ocean oxygenated conditions.

[10] Each individual shell was split into three separate fragments using a scalpel. All shell fragments from each individual were ultrasonicated two times in Optima methanol and then rinsed three times in 18.2 MΩ Milli-Q® water. One shell fragment from each specimen received no further cleaning treatment and was set aside for LA-ICP-MS analysis. The other two fragments from each specimen were cleaned with both reductive and oxidative cleaning treatment using the protocols in Martin and Lea [2002], modified for cleaning very small samples. Shell fragments were placed in a reducing anhydrous hydrazine/ammonium citrate solution and maintained at 70–80°C for 30 min. The hydrazine solution was removed, and the samples were rinsed four times in 18.2 MΩ Milli-Q water. Each fragment was then cleaned in an oxidizing solution of 30% H2O2 buffered in 0.1 N NaOH at 80–90°C for 10 min, and then rinsed four times in 18.2 MΩ Milli-Q water. Following this step, one of the two fragments subjected to the reduction + oxidation cleaning step was then set aside for LA-ICP-MS analysis. The remaining fragment from each specimen was additionally cleaned with a chelating treatment with DTPA, using the protocol of Lea and Spero [1994] modified for very small samples. Each fragment was cleaned in 10 μL of a 0.002 mol L−1 solution of DTPA buffered in 0.1 N NaOH, and maintained at 100°C. After 5 min, the DTPA was neutralized by the addition of 0.01 N NaOH, and then the samples were rinsed three times in 18.2 MΩ Milli-Q water. All fragments from each individual shell were subjected to a final “hot rinse” step in Milli-Q and maintained at 100°C for 30 min. A final leaching step in dilute acid is typically conducted to remove additional surface-adhered particles prior to analyses using solution ICP-MS. We omitted the final dilute acid leaching step because with LA-ICP-MS, surface-adhered particles are removed in the first 1–2 s of laser ablation, and this portion of collected data can be manually excluded from computations. Specimens were air dried and mounted on double-stick carbon tape with trace metal-clean Optima methanol. Because of the limitations of the size of individual O. universa tests and the high probability of sample loss during cleaning of small shell fragments, we did not evaluate the separate effects of reductive and oxidative cleaning during this study.

[11] Shell fragments were analyzed for Ba/Ca and Mg/Ca in the Department of Geology, Stable Isotope Laboratory at the University of California, Davis using a Photon Machines pulsed ArF excimer laser, with HeLex dual-volume sample cell, coupled to an Agilent 7700x quadrupole ICP-MS. Each sample was ablated in depth profile, from shell interior to exterior, at 5 Hz using a 30 μm spot and a laser fluence of 1.46 J cm−1. Ablated material is transported to the ICP-MS for analysis in an Ar-He gas mixture through a 10 path distributed delay manifold (a “squid”) to smooth the signal [Müller et al., 2009]. Isotopes and elements measured included 24Mg, 25Mg, 27Al, 43Ca, 44Ca, 55Mn, 88Sr, and 138Ba, with a total acquisition time of ∼250 ms. In this paper, we present computed Mg/Ca and Ba/Ca ratios. Typical ablated depth profiles required 30–70 s. The operating conditions of the laser and ICP-MS during analyses are summarized in Table 1. Following laser ablation analyses, the inner surfaces of shell fragments were imaged using a Hitachi tabletop Scanning Electron Microscope (SEM) (Figure 1).

Table 1. Operating Conditions of the LA-ICP-MS System During Collection of Depth Profiles
ICP-MS: Agilent 7700x
RF power1500 W
Argon gas flow1.05 L/min
Coolant gas flow15 L/min
Auxiliary gas flow1 L/min
Dwell time per mass20–60 ms
Monitored masses (m/z)24Mg, 25Mg, 27Al, 43Ca, 44Ca, 55Mn, 88Sr, 138Ba
Laser Ablation System: ArF Excimer Laser
 
Energy density (fluence)1–3 J/cm2
He gas flow1.05 L/min
Laser repetition rate5 Hz
Laser spot size30 µm
ThO+/Th+<0.5%
Figure 1.

SEM images of (a–c) the interior of O. universa shell fragments from a Caribbean specimen (specimen 8) and (d-f) an Orca Basin specimen (specimen 20). Fragments were sonicated-only in Figures 1a and 1d, sonicated + reductive + oxidatively cleaned in Figures 1b and 1e, and sonicated + reductive +  oxidative + DTPA cleaned in Figures 1c and 1f. Images illustrate differences in porosity between specimens, and increasing dissolution with cleaning techniques between fragments of the same specimen. Two to four holes ablated during analysis are visible on each fragment.

[12] Internal standardization was performed using measured 43Ca intensity. Every 10–20 samples, an identical ablation protocol on National Institute of Standards and Technology (NIST 610) standard glass was performed, and measured elemental intensities were used for standardization of bracketed samples. Data reduction consisted of a despiking routine to remove outliers, followed by subtraction of measured background intensities for each element. Three to four ablation depth profiles were collected on each shell fragment from shell interior to exterior. Data collected by the ICP-MS before and after ablation of shell material are excluded (Figure 2, shaded regions). To compute Me/Ca ratios in each depth profile, points were selected starting from where measured Al (in counts per second) decreased in the depth profile (Figure 2a). Some depth profiles contained a high-Ba region at the start of the profile; we excluded these nonhomogeneous high-Ba regions from computations of whole-shell Ba/Ca. When the laser is initially turned on and ablation commences, 2–4 s of high Mg, Ba, and Al concentrations were observed; these portions of the depth profile were also excluded from computed Me/Ca ratios (Figures 2a and 2b). The conclusion of each depth profile was determined by monitoring a decrease in 44Ca and 43Ca, along with rising counts of 27Al which indicate an increasing proportion of material from the carbon tape used for sample mounting. Laser ablation ICP-MS depth profiles in spar calcite ablate at a constant rate per laser pulse under similar analytical conditions [Eggins et al., 2003], so we use the time component of collected depth profiles to estimate distance ablated into the shell.

Figure 2.

(a) Raw data from a typical laser ablation depth profile through an O. universa shell wall. This specimen (specimen 3, Caribbean) was sonicated in methanol and ablated from shell interior to exterior. After an initial burst of high 24Mg, 27Al, and 138Ba, Me/Ca ratios decrease to a stable signal through the rest of the shell. Data collection on shell material ends when the 44Ca signal (counts per second) reaches an inflection point on the log scale, and the 27Al signal rises. During data reduction, the areas shaded in gray are omitted from intrashell depth profiles and computed whole-shell ratios. (b) A depth profile through a fragment of the same shell in Figure 2a that went through reductive and oxidative cleaning. High counts are observed in 27Al and 138Ba immediately after ablation commences, even though shell material has been aggressively cleaned.

[13] The mean Ba/Ca ratio for each depth profile is computed as the robust statistics H15 mean. This approach reduces the weight of outlier values, and provides a more reliable estimate of mean and standard deviation in analytical chemistry data sets, where error distribution may be asymmetrical [Analytical Methods Committee, 1989]. For Mg/Ca, where the natural variability exceeds instrument-related error, we report the arithmetic mean. The internal precision on each depth profile is reported as ±2 standard error (SE) of all data points included in the computation of the mean. Because Mg/Ca is naturally heterogeneous across an O. universa depth profile, the internal precision on individual depth profiles is not a true measurement of instrument error, so we only report external precision for computed whole-shell Mg/Ca ratios. For each shell fragment, we compute an average whole-shell Me/Ca by calculating the arithmetic average of mean Me/Ca ratios from two to four repeat, adjacent depth profiles on a single fragment. The precision of a computed whole-shell Me/Ca ratio is 2 standard deviation (SD) of this average (n = 2–4) [Marr et al., 2011]. The volumetric difference between innermost and outermost layers of shell calcite is less than 5%, so we weight all calcite layers equally when computing whole-shell Me/Ca ratios.

3. Intrashell Mg/Ca and Ba/Ca Measurements

3.1. Mg/Ca Ratios

[14] Measured Mg/Ca ratios for all shell fragments from all treatments are summarized in Table S1.1 In sonicated-only shell fragments from ODP999A, computed whole-shell Mg/Ca ratios range from 5.2 to 8.6 mmol mol−1. On fragments that received reductive + oxidative cleaning treatments, the change in computed whole-shell Mg/Ca ratios is variable, ranging from a 19% decrease to a 30% increase (Table S1). On the same individuals, shell fragments that received reductive + oxidative + DTPA treatments also show variable changes in Mg/Ca, ranging from 28% decrease to 30% increase in comparison to sonicated-only shell fragments. Specimens from the Orca Basin also exhibit diverse responses to cleaning treatments. Computed whole-shell Mg/Ca ratios in sonicated-only shell fragments range from 3.6 to 6.6 mmol mol−1, with reductive + oxidatively cleaned fragments exhibiting 16% decrease to 22% increase in Mg/Ca ratios, and reductive + oxidative + DTPA-cleaned fragments exhibiting 40% decrease to 11% increase (Table S1). The numerical values of measured Mg/Ca ratios, when evaluated alone, do not display a consistent pattern of increase or decrease between cleaning treatments.

[15] Depth profiles show the patterns of intrashell Mg/Ca heterogeneity and provide some insight into the variable responses of individual shells to different cleaning treatments. Figure 3 shows an example of the Mg/Ca depth profiles collected in this study, as computed from raw data shown in the example in Figure 2. In Figure 3a, three separate Mg/Ca depth profiles are shown from one shell fragment that received sonication in methanol treatment. The process of sonicating shell material in methanol is intended to clean the shell of loose adherent clays and other sedimentary material but is not intended to remove mineralized coatings. An initial 2–4 s of high-count readings on 27Al, 24Mg, and 138Ba is observed in the raw data for depth profiles through sonicated-only fragments (Figure 2a), potentially due to remaining clays not removed by sonication; this portion of the data is excluded from the computed Mg/Ca depth profiles (Figure 3a).

Figure 3.

Mg/Ca ratios (reduced data) from multiple fragments of the same shell (specimen 3, Caribbean), showing three different ablated depth profiles through a (a) sonicated-only, (b) reductive + oxidatively cleaned, and (C) reductive + oxidative + DTPA cleaned fragment. (d) Combined depth profiles from Figures 3a (blue) and 3b (green), shown as 5 point running means, with original Mg/Ca data in gray. Depth profiles from the shell fragment that received reductive and oxidative cleaning (green) are offset by 15 s of collection time to align spatial pattern of Mg/Ca with depth profiles from Figure 3a. Tie point used to identify offset is indicated by a black triangle.

[16] The spatial positions of Mg/Ca minima and maxima from different depth profiles in Figure 3a are in excellent agreement and demonstrate reproducibility between multiple ablation holes through the same shell fragment. Depth profiles through a second fragment of the same individual shell, which received reductive and oxidative cleaning treatments, are shown as both raw data (Figure 2b) and computed intrashell Mg/Ca (Figure 3b). The 2–4 s initial burst of high counts on 24Mg, 27Al, and 138Ba is also present in shell fragments that have undergone rigorous cleaning treatments (Figure 2b). The Mg/Ca minima and maxima are in excellent spatial agreement and demonstrate good reproducibility between repeat depth profiles (Figure 3b). In Figure 3c, the profiles from the reductively and oxidatively cleaned shell fragment are superimposed on depth profiles from the methanol-sonicated shell fragment, using the Mg/Ca minimum as a tie point. The ablation conditions were identical during analysis of the two fragments, so we assume the same rate of removal of shell calcite during ablation of depth profiles. The Mg/Ca spatial patterns in the two shell fragments are very similar and can be spatially correlated, with the exception of missing material on the inner and outer portions of the ablation profiles. This suggests that material from the inner and outer surfaces of the shell is removed during the reductive and oxidative cleaning, rather than the selective dissolution of a specific contaminant phase.

[17] Superimposing Mg/Ca depth profiles from both methanol-sonicated-only and sonicated + reductive + oxidative cleaning treatments on fragments of the same shell reveals patterns of dissolution induced by different cleaning treatments on both Caribbean (Figures 4a–4e) and Orca Basin specimens (Figures 4f–4j). Mg/Ca depth profiles through reductively + oxidatively cleaned shell fragments are shorter in duration (seconds of collection time) but retain the spatial pattern of intrashell heterogeneity exhibited by fragments that were only sonicated in methanol. Using Mg/Ca minima and maxima as tie points between cleaned and uncleaned fragments, the depth profiles from cleaned shells clearly show that shell material is missing from the inner shell surface after reductive and oxidative cleaning (Figure 4). This dissolution pattern is consistent between multiple samples from different depositional environments. However, the amount of missing material differs for each individual shell, so the effect of reductive + oxidative cleaning on whole-shell Mg/Ca is not consistent between different individual shells (Table S1). On shell fragments where the DTPA cleaning step was performed after the reductive and oxidative cleaning step, additional shell material is not removed, and computed whole-shell Mg/Ca ratios are not significantly affected (Table S1).

Figure 4.

Intrashell Mg/Ca depth profiles through multiple shell fragments from the same individual O. universa specimens from (a–e) the Caribbean and (f–j) the anoxic Orca Basin, Gulf of Mexico. Shell fragments with reductive and oxidative cleaning (light green lines and text) have been offset to align intrashell Mg/Ca patterns with sonication-only treatment (dark blue lines and text). Tie point(s) used to identify offset are indicated by black triangles.

3.2. Ba/Ca Ratios

[18] Figure 5a shows an example of three depth profiles through a methanol-sonicated-only shell fragment, with computed Ba/Ca of 1.1 μmol mol−1 and external (spot-to-spot) reproducibility of 0.08 μmol mol−1 (2 SD of repeat depth profiles). We highlight intrashell Ba/Ca homogeneity by calculating a 15 point running mean average of Ba/Ca for each depth profile. Figures 5b and 5c show three different depth profiles from the fragment of the same shell that received reductive + oxidative and reductive + oxidative + DTPA cleaning treatments (computed whole-shell Ba/Ca = 1.0 ± 0.03 μmol mol−1 and 0.9 ± 0.09 μmol mol−1, respectively). To compare computed Ba/Ca ratios from different fragments, we superimpose 15 point running mean averages of depth profiles on the same specimen from the two cleaning protocol end-members: sonicated-only and reductive + oxidative + DTPA-cleaned (Figure 5d).

Figure 5.

Intrashell Ba/Ca depth profiles for multiple fragments of an O. universa shell (specimen 15, Orca Basin) that received different cleaning treatments. Raw data (gray) and 15 point running means (bold, colored lines) are shown for (a) sonicated-only, (b) sonicated + reductive + oxidatively cleaned, and (c) sonicated + reductive + oxidative + DTPA cleaned fragments. (d) Depth profiles from the two end-members (Figure 5a; blue) and (Figure 5c; red) are superimposed.

[19] The comparison of depth profiles from methanol-sonicated fragments and shell fragments that received the full suite of cleaning treatments (including DTPA) is repeated for specimens from both the oxic Caribbean site (Figures 6a–6e) and the anoxic Orca Basin (Figures 6f–6j). Reductive + oxidative cleaning protocols decrease computed whole-shell Ba/Ca by an average of 21% for Caribbean samples and 10% for Orca Basin samples, with respect to measured Ba/Ca for sonicated-only fragments. On individuals that received the full reductive + oxidative + DTPA cleaning treatment, whole-shell Ba/Ca decreased an average of 46% for Caribbean specimens and 13% for Orca Basin specimens (Table S2). On depth profiles that contain a nonhomogeneous, high-Ba region at the start of the profile, this portion is excluded from computed whole-shell Ba/Ca (Table S2; excluded profiles are marked with a “*”).

Figure 6.

Comparison of intrashell Ba/Ca depth profiles through sonicated-only (blue) and sonicated + reductive + oxidative + DTPA cleaned fragments (red) for O. universa specimens from (a–e) the Caribbean and (f–j) the anoxic Orca Basin. Raw data are in gray; bold, colored lines are 15 point running means.

[20] The concentration of Ba in foraminiferal calcite (∼0.5 to 2 μmol mol−1) is ∼103 times less abundant than Mg, and measured Ba is much closer to background levels measured by the ICP-MS. As a result, the intrashell Ba/Ca profiles exhibit much more “noise” (Figures 5 and 6), which can be quantified by the SE of all the raw data points in a depth profile (Table S2). Shell fragments of O. universa from the Caribbean tend to exhibit greater differences between sonicated-only fragments and shell fragments that received increasingly aggressive cleaning treatments, both in the SD between repeat depth profiles and the SE (“noisiness”) of individual depth profiles (Table S2). In specimens from the Caribbean, computed Al/Ca is higher and more variable in sonicated-only fragments (average 0.6 mmol mol−1) and cleaned specimens (average 0.1 mmol mol−1) than in specimens from the Orca Basin (Table S2). For most shell fragments, the average external (spot-to-spot) reproducibility for computed Ba/Ca is <0.2 μmol mol−1 (Table S2). On some specimens, the intrashell Ba/Ca pattern is not reproducible between repeat depth profiles through the same shell fragment (e.g., Figure 6d). Generally, these depth profiles can be visually identified; for these specimens, we have excluded portions of depth profiles with clear Ba/Ca heterogeneity from calculations of whole-shell Ba/Ca.

[21] We performed an Analysis of Variance (ANOVA) F-test for each specimen, which compares both the error on individual cleaning treatments (±2 SD on repeat profiles through the same fragment) and computed mean Ba/Ca between different treatments. On individuals for which calculated F < Fcritical, the calculated mean Ba/Ca is not statistically different between different cleaning treatments (Table S2). On individuals that do not pass the F-test and exhibit statistically significant differences in measured Ba/Ca between fragments, analyses that produce anomalously high-Ba/Ca ratios can be identified by a combination of other factors. The computed Ba/Ca ratios from these shell fragments tend to have elevated Al/Ca ratios, external spot-to-spot reproducibility >0.2 μmol mol−1 (2 SD), noisier depth profiles (higher SE) compared to other samples from the same basin, depth profiles with regions of visually obvious Ba/Ca heterogeneity, or some combination of all these traits.

4. Dissolution During Cleaning Treatments

[22] The LA-ICP-MS depth profiling technique allows us to evaluate what drives changes in measured Mg/Ca ratios between different cleaning treatments. Each individual O. universa shell exhibits a unique intrashell pattern of Mg/Ca variability that is reproducible between different fragments of the same spherical shell, so we are able to compare directly the results of different cleaning protocols on different fragments of the same shell. We observe clear patterns of dissolution and material removal from inner shell surfaces after reductive and oxidative cleaning (Figures 3 and 4), as well as textural evidence from SEM images (Figure 1). By superimposing offset depth profiles on methanol-sonicated-only and sonicated + reductive + oxidatively cleaned fragments, we demonstrate that maxima and minima in intrashell Mg/Ca bands do not change during cleaning steps. These patterns are observed in specimens from both anoxic, hypersaline environments (Orca Basin) and fully-oxygenated bottom waters (Caribbean; Figure 4).

[23] Our results are consistent with the results of Sadekov et al. 2010, whose experiments demonstrate preservation of Mg/Ca spatial patterns during simulated seafloor dissolution and suggest that material is dissolved from available shell surfaces. Of the specimens we analyzed, only 44% of individual O. universa show a decrease in computed whole-shell Mg/Ca ratios with additional cleaning treatments (Table S1). The removal of portions of the shell with variable Mg/Ca accounts for these results (Figures 3 and 4). In shells with high-Mg diagenetic overgrowths, the removal of these outer shell layers during reductive and oxidative cleaning may explain the clear decrease in Mg/Ca during cleaning observed in many studies [Barker et al., 2003; Benway et al., 2003; Haley and Klinkhammer, 2002; Martin and Lea, 2002; Yu et al., 2007]. However, data presented here do not support preferential dissolution of high-Mg layers from within the calcite lattice.

[24] Live-cultured O. universa exhibit a region of high-Mg calcite inside the POM and an increase in Mg/Ca toward the outer shell surface, measurable using both LA-ICP-MS and electron probe micro-analyzer (EPMA) [Eggins et al., 2004]. The inner high-Mg region is composed of multiple thin bands of high- and low-Mg calcite, which appear as a region of elevated Mg/Ca when measured by LA-ICP-MS if ablation profiles are not perfectly orthogonal to the direction of shell growth [Spero et al., 2012]. The inner layer of calcite with elevated Mg/Ca appears to be dissolved preferentially during reductive and oxidative cleaning, in addition to the outer surface of the shell on some samples (Figure 4). Similar preferential dissolution of the innermost layer of shell calcite was attributed by Benway et al. [2003] to selective dissolution of the innermost, dissolution-prone microgranular layer of shell calcite [Bé et al., 1975]. It is also possible that the innermost layer is removed during reagent cleaning along a plane of preferential weakness in the shell created by the organic-rich POM that is easily resolvable in SEM images [Bé et al., 1975; Spero, 1988].

[25] Some researchers have also suggested that the degree of mechanical cracking may affect the efficacy of cleaning procedures, because of the increased ease of removing clay contaminants from shells that are cracked open [Barker et al., 2003]. Mechanical crushing of shells has been shown to facilitate the separation and removal of sedimentary material from the inside of foraminifer shells. During chemical cleaning, reagents do not immediately penetrate into the interior of the shell through pores and apertures, which vary depending on species and shell morphology. Chemical dissolution still occurs inside of an unbroken shell, but can be limited by slow flow of reagents through shell pores. In mechanically cracked shells, inner shell surfaces have increased exposure to reagents during cleaning.

[26] Our data suggest that the inner layer of calcite [Bé et al., 1975] is a preferential zone of dissolution and removal of shell material during cleaning, so increased exposure to reagents has the potential for a strong impact on the amount of shell material dissolved during cleaning. Pena et al. [2005] demonstrated that diagenetic coatings on shell interiors are loosened during reductive cleaning via dissolution of oxide layers holding the inner diagenetic coatings in place. Barker et al. [2003] also demonstrated that ultrasonication during reagent cleaning steps increases the effects of cleaning on Mg/Ca ratios, in part because of the removal of silicate clays. A shell filled with sedimentary material may have its inner calcite surfaces occluded or blocked, so that inner surfaces are not available for dissolution, as discussed by Sadekov et al. [2010]. Hence, the degree of mechanical cracking of shells may play an important role in the effects of reductive and oxidative cleaning on whole-shell Mg/Ca.

[27] Our results unequivocally demonstrate that material is dissolved and removed from the interior surface of the shell during reagent cleaning. The effect of dissolution on computed whole-shell Mg/Ca ratios is dependent on the amount of shell calcite dissolved during reagent cleaning and the proximity of intrashell high- and low-Mg bands to shell surfaces accessible to cleaning reagents and available for dissolution. While it is currently common practice to conduct a reductive cleaning step prior to Mg/Ca analyses, the reductive step may instead dissolve a considerable proportion of primary calcite and bias results. A full LA-ICP-MS depth profile characterization that isolates the individual effects of reductive and oxidative cleaning protocols, or the role of hydrazine in the reductive step, is beyond the scope of this study. Interspecies differences in the internal distribution of high- and low-Mg calcite may explain some of the complexity inherent in directly comparing cleaning experiments conducted on different planktic and benthic species. This also highlights the importance of selecting the same cleaning protocol for Mg/Ca measurements as used during calibration of the Mg/Ca paleothermometer selected.

[28] The primary applications of LA-ICP-MS depth profiling in planktic foraminifer shells to date have addressed questions about biological controls [Bolton et al., 2011; Eggins et al., 2003; Marr et al., 2011] and the effects of diagenetic alteration [Hathorne et al., 2009; Pena et al., 2008] and dissolution [Sadekov et al., 2010]. Studies that have used stable isotope (δ18O, δ13C) analyses of multiple individual foraminifers from single-core intervals [Killingley et al., 1981; Schiffelbein and Hills, 1984] capitalize on the geochemical variability between individuals, which is a function of ecological parameters such as seasonality and depth habitat. The same range of environmental conditions is reflected in the Me/Ca ratios recorded by a suite of individual foraminifers from a single sediment sample, which could be used as an additional dimension of information in paleoceanographic reconstructions. The average Me/Ca from multiple measured individuals thus represents the species mean temperature and is similar to the mean temperature obtained using traditional bulk analyses of multiple shells. However, removing a portion of shell material from all individuals in a sample during reagent cleaning will ultimately affect the final measured Me/Ca ratios. When applying LA-ICP-MS analyses of individual foraminifers to paleoceanographic questions, an assessment of inter-individual variation in each depositional setting will enable researchers to identify, and decide how to account for, the intershell ecological variability recorded in Me/Ca ratios.

5. Evaluating Sediment Coatings and Diagenetic Phases

[29] Measurements of foraminiferal Me/Ca using solution-based ICP-MS can be complicated by the presence of adhering sedimentary particles and mineralized diagenetic overgrowths. Flow-through analyses with progressive dissolution steps have isolated components with Me/Ca ratios an order of magnitude higher than in biogenic calcite, which are interpreted as contaminant phases [Benway et al., 2003; Haley and Klinkhammer, 2002; Klinkhammer et al., 2004]. EPMA analyses have confirmed that diagenetic alteration of Me/Ca ratios is often confined to outer shell coatings and the interior surfaces of shell pores [Pena et al., 2005]. The Me/Ca depth profiles we present in this paper do not show thick Mn- and Mg-rich diagenetic coatings, despite the range of depositional environments where diagentic coatings might be expected.

[30] Studies using LA-ICP-MS depth profiles to analyze foraminifer shells also report a trace-element-rich surface veneer on samples within the first 2–4 s of ablation time [Eggins et al., 2003; Sadekov et al., 2008, 2010]. Some studies interpret these high counts as an indicator of a diagenetic coating [Hathorne et al., 2003] or biogenic calcite layer [Bolton et al., 2011]. However, the initial high counts present immediately after ablation commences are different from high-Me/Ca regions identified as mineralized diagenetic layers of shell calcite [Pena et al., 2008]. We observe the same initial high-Me/Ca ratios as soon as ablation begins (Figure 2a), even at the start of depth profiles through specimens that are missing surface shell material due to dissolution (Figure 2b). On these more aggressively cleaned shell fragments, material has been removed from shell surfaces since deposition (Figures 3 and 4), so it does not represent a sedimentary contaminant phase. The initial high Mg, Al, Mn, and Ba (with respect to Ca) is also present on a low-energy (<2 J cm−2) ablated depth profile in NIST 612 standard glass, which has concentrations of these elements comparable to foraminiferal calcite [Jochum et al., 2011]. Based on these results, we suggest that the initial high-Me/Ca may be an artifact of the laser ablation depth profiling process at low energy.

[31] Experiments with live O. universa have empirically determined the relationship between Ba/Caseawater and Ba/Cashell and demonstrate that biogenic Ba/Cashell in O. universa is very low. Cultured O. universa specimens exhibit intrashell Ba/Ca homogeneity [Vetter et al., 2013], and thus any high-Ba phases observed in LA-ICP-MS depth profiles likely represent a nonbiogenic signal, either from adherent sedimentary particles or from postdepositional diagenetic overgrowths. Previous studies demonstrate that the use of full reductive + oxidative + DTPA cleaning on multiple species of planktic foraminifers reduced Ba/Ca from 6 to 8 μmοl mol−1 (sonicated-only) to 0.5–1 μmol mol−1 (full cleaning treatment) [Lea and Boyle, 1991]. The latter range is consistent with biogenic-only Ba/Cashell as measured in cultured O. universa [Hönisch et al., 2011; Lea and Spero, 1994]. Haley and Klinkhammer [2002] used flow-through analyses to confirm that Ba is released into solution in concentrations of 10–15 ppb during the DTPA cleaning step (solution [Ca] ∼10 ppm), showing clear removal of a high-Ba contaminant phase. While we observe some portions of intrashell Ba/Ca depth profiles with high Ba (Table S2 and Figures 5 and 6), we did not observe Ba/Ca ratios in any of the specimens we analyzed that are comparable in magnitude to the results of Haley and Klinkhammer [2002].

[32] In our data set, we directly compare individual depth profiles and computed whole-shell Ba/Ca ratios from different fragments of the same O. universa shell (Figure 6). Multiple repeat depth profiles through individual shell fragments show generally constant intrashell Ba/Ca and reproducible patterns. Ba/Ca is homogeneous in cultured O. universa specimens that have never been in sediments [Vetter et al., 2013], and so we exclude portions of the shell that contain heterogeneous Ba/Ca ratios from the computed whole-shell Ba/Ca ratios (e.g., Figure 6d) on the assumption that these phases are not biogenic. In 66% of our samples, the Ba/Ca ratio is statistically lower in shell fragments that have received different cleaning treatments when compared to sonicated-only shell fragments. The use of increasingly aggressive cleaning techniques also appears to decrease the “noisiness” (SE) of measured intrashell Ba (Figure 6 and Table S2). Notably, sonicated-only fragments of specimens from the Caribbean have much more variable Ba/Ca ratios, and Caribbean specimens have higher spot-to-spot SD and “noisier” profiles (higher SE) overall.

6. Cleaning Techniques and Porosity

[33] The size and density of shell pores differs between the O. universa specimens we analyzed from the Caribbean and the Orca Basin (Figure 1). Recent genomic evidence suggests that the textural differences we observe between specimens from our two localities represent two cryptic species of O. universa: the “Caribbean” and “Sargasso” morphotypes [Morard et al., 2009]. Additionally, pore density and distribution in planktic foraminifers varies between species [, 1968], and may vary as a function of latitude, temperature, and dissolved oxygen [, 1968; Bé et al., 1976; Frerichs et al., 1972]. For each shell fragment, the possibility exists that sedimentary material is still present on the interiors of pore voids in samples that received less aggressive cleaning treatments (e.g., Figure 1d). In this study, we collected LA-ICP-MS depth profiles using a 30 μm spot that averages material both from ablated calcite and through several pores, so sedimentary material that remains in pores after cleaning has the potential to affect measured Me/Ca ratios.

[34] Electron microprobe mapping of trace element distribution in planktic foraminifer shells demonstrates that sediment and other contaminant phases are often present in pore spaces [Pena et al., 2008]. These findings, in combination with our results, suggest that both pore size and pore density may play an important role in the efficiency of individual cleaning protocols. While sonication can visibly remove sedimentary material from pores, thin veneers or coatings may remain on the interior walls of pores. In addition to dissolving material from shell interiors, SEM images demonstrate that dissolution during reagent cleaning also dissolves pore walls and widens individual pores [Benway et al., 2003; Figures 1b, 1c, 1e, and 1f]. This dissolution of pore walls may explain some of the reduction in “noisiness” observed in the Ba signal of shell fragments that received progressively more rigorous cleaning treatments as well as differences in SE observed between Orca Basin and Caribbean specimens.

7. Summary

[35] In this paper, we use LA-ICP-MS depth profiles to identify patterns of intrashell Mg/Ca in multiple fragments of fossil O. universa shells that were sonicated in methanol, reductively and oxidatively cleaned, or received a full (methanol + reductive + oxidative + DTPA) cleaning treatment. We demonstrate that the reductive and oxidative cleaning steps remove a significant portion of biogenic calcite from shell surfaces available for dissolution, and the removal of material may alter computed whole-shell Mg/Ca ratios. However, we do not observe evidence for high-Mg phases only being preferentially dissolved from within shell calcite.

[36] Measured intrashell Ba/Ca patterns demonstrate Ba/Ca homogeneity in fossil O. universa from multiple depositional and oceanographic settings. These results highlight the potential for LA-ICP-MS depth profiling to identify either diagenetic overgrowths or contamination from sedimentary material. The addition of reductive + oxidative and reductive + oxidative +  DTPA cleaning steps can decrease computed whole-shell Ba/Ca, but results vary between individuals. The size and density of shell pores has a pronounced effect on the amount of contamination from sedimentary material. Areas of intrashell Me/Ca depth profiles with high internal SE, a high external spot-to-spot reproducibility (SD), should be excluded from computations of whole-shell Ba/Ca ratios. We conclude that LA-ICP-MS depth profiling on fossil foraminifers sonicated in methanol, without using additional reagent cleaning, can be used to compute whole-shell Me/Ca ratios for use in paleoceanographic reconstructions, if quality control is employed for each depth profile.

Acknowledgments

[37] The authors gratefully acknowledge conversations with Steve Eggins and Steve Shuttleworth regarding laser ablation ICP-MS analyses. This work was supported by National Science Foundation grants OCE-1061676 and EAR-0946297 (H.J.S. and A.D.R).