2022 年 56 巻 4 号 p. 112-128
A new experimental method was developed to understand the partitioning of rare earth elements (REEs) on carbonate phases in oceanic columns. This method comprises of a mildly-oversaturated stage (Stage 1) and an undersaturated stage (Stage 2), mimicking natural oceanic columns consisting of the upper oversaturated and deeper undersaturated layers, in which carbonate particles of different sizes partition REEs with seawater. Saturation levels were adjusted by purging N2 gas with different CO2 concentrations. In this method, partitioning progress and equilibrium can be monitored via two viewpoints: differences in REE/Ca concentration among differently-sized carbonate particles and ionic activity products (IAPs) of the experimental solution.
The distribution coefficients obtained by the experiments, however, varied over three orders of magnitude. By analyzing the solution and carbonate particles, the presence of Fe in the carbonate particles (Fe/Ca) was observed to influence the partitioning of REEs. Microscopic mapping of Fe and proportionality of Fe amount to surface area strongly indicate that Fe is distributed evenly on the surface of carbonate, and Fe hydroxide is considered to be responsible for the REE enrichment. Neither the presence of Mn nor type of seed crystal (calcite or aragonite) was likely to significantly influence partitioning.
When Fe/Ca of carbonate was small, the ion activity products of the solution were found to modulate the distribution coefficient, implying that the presence of REE disturbs free relocation of carbonate ions. This may lead to overestimation and underestimation of distribution coefficients in Stages 1 and 2, respectively. We conclude that the distribution coefficient of REEs on calcium carbonate is not very high, but in natural settings carbonate is still important in controlling REE distribution by hosting Fe hydroxide, which effectively partitions REEs.
The distribution of rare earth elements (REEs) in seawater results from the interplay between particulate and dissolved phases, with dissolved phases moving laterally along an isopycnic surface. Proposed REE hosts include oxides (Sholkovitz et al., 1994), carbonates (Takebe, 2005), organic matter (Kuss et al., 2001; Sholkovitz et al., 1994), phosphates (Byrne and Kim, 1990), and silicates (including silica frustules) (Akagi, 2013; Akagi et al., 2011; Kuss et al., 2001; Rousseau et al., 2015). It is likely that no single phase can fully explain the profiles whereby REEs increase with increasing depth, referred to as recycled profiles. To explain their profiles, readily remineralizable organic matter and silica frustules seem to be the most probable hosts. However, the different profiles among REEs require a mechanism to differentiate them. Byrne and Kim (1990) proposed different constants of complex formation with organic matter, which control the release of REEs from the particles to seawater as REE-complex ions. This idea has been further refined (Schijf et al., 2015), but some different ideas, e.g., silicic acid complexation followed by diatom intake (Akagi, 2013) and/or carbonate absorption (Akagi, 2013; Nishino and Akagi, 2019) can also reproduce HREE enrichment in seawater.
A predominantly lateral process operating in conjunction with the oxygen minimum zone has also been proposed as mechanisms to characterize intermediate water (Bayon et al., 2004; German and Elderfield, 1990; Sholkovitz et al., 1994). It is however doubtful that the lateral processes can effectively remove lighter REEs (LREEs) or add heavier REEs (HREEs) to match the REE composition of intermediate water (Nishino and Akagi, 2019). We argue that the re-partitioning of released REEs from remineralized primary hosts on carbonate-particle surfaces should account for the difference, based on the varying extent of the contribution to partitioning in conjunction with carbonate saturation depths of oceans (Akagi, 2013; Nishino and Akagi, 2019). The key information must be the effective partitioning coefficients of REEs operated between carbonate surface and seawater.
The partitioning coefficient (D) of REEs between carbonate particles and seawater is expressed as
| (1) |
where [Ln] and [Ca] are concentrations of Ln, an element of REEs, and Ca in particulate carbonate or seawater. However, partitioning coefficients vary depending on how they are measured. In most experiments the partitioning coefficients were determined by precipitation methods, where calcite crystals grew in the presence of REEs under oversaturated conditions with respect to calcite (Toyama and Terakado, 2014; Zhong and Mucci, 1995). Saturation was achieved by evaporation or mixing of solutions. Some were reported by analyzing carbonate sediments and overlaying seawater (Scherer and Seitz, 1980). Reported partitioning values vary widely (Fig. 1) and consensus has not yet been reached. The partitioning values are often higher in the LREE region than the HREE region, but almost flat values across REEs have also been reported (Toyama and Terakado, 2014).
Reported partitioning patterns obtained by analyses of naturally available carbonates (broken lines in brown) and by experiments (solid lines in green).
Another disparity between experimental and natural systems is that in natural systems carbonate particles are provided at the surface and foreign ions are subject to partitioning with carbonate particles in deeper water; whereas in experimental systems partitioning takes places in an oversaturated condition during calcite growth (Tanaka and Kawabe, 2006; Toyama and Terakado, 2014; Zhong and Mucci, 1995). In natural systems growth may occur during settlement, but the growth rate is considered very small owing to a small saturation index of 2 or 3 at most (Broecker and Peng, 1982). In addition, dissolution exceeds growth in layers deeper than the carbonate saturation depth.
In this study, we have devised a new experimental system to pursue partitioning or exchange reactions of Ca-REE in calcite, in the hope that valid partitioning coefficients will be obtained. Our new system enables us to study those reactions of REEs on metastable aragonite, which has not been studied so far, as well as reactions on calcite. We also studied the influence of REEs and the major metals, Fe and Mn, on the reaction, which might be relevant in natural systems.
Commercially available surface seawater collected offshore at Ogasawara, Japan in the Pacific Ocean was used for the experiments. The seawater was stored in the dark at room temperature and was filtered by a 0.45 μm membrane filter before the experiments. The concentration of REEs in the original seawater was not measured, as it is unimportant as long as the concentrations of seawater and particles can be separately measured at the end of each REE-spike experiment. Calcite particles were prepared by pulverizing a calcite crystal, sampled at Kawaradake, Fukuoka, Japan, using an agate mortar. Aragonite particles were prepared from a piece of dried/white coral sampled offshore at Amakusa, Japan. The coral piece was crushed and washed with water. After drying at 80°C, the grains were further powdered by an agate mortar. To estimate organic content in the coral, the powder was heated at 500°C for two hours in a muffle furnace. The ignition loss of organic matter in coral skeleton was less than 3%. However, to avoid transformation to calcite, unheated powder was used in the experiments. The particle size of calcite or aragonite was arranged deliberately uneven with the largest size of particles about 100 μm and the smallest size about 1 μm. The mineral compositions of the particle powder were confirmed to be calcite and aragonite, respectively, using X-ray diffractometer (Rigaku Ultima IV) with CuKα radiation generated at 40 kV and 40 mA. Foreign metal concentration (M/Ca) in our calcite and aragonite seeds analyzed with ICP-AES or ICP-MS were: Mg/Ca = 3.9 × 10–3, Sr/Ca = 3.8 × 10–4, Fe/Ca = 4.2 × 10–4, Mn/Ca = 1.2 × 10–4, La/Ca = 5.9 × 10–7, Lu/Ca = 2.1 × 10–8 in molar/molar for calcite and Mg/Ca = 4.1 × 10–3, Sr/Ca = 4.6 × 10–3, Fe/Ca = 6.5 × 10–5, Mn/Ca = 7.2 × 10–6, La/Ca = 3.5 × 10–8, Lu/Ca = 1.1 × 10–9 in molar/molar for aragonite.
An REE spike solution containing 14 REEs at approximately 70 mg/kg was prepared by mixing laboratory-made 14 lanthanoid standard solutions, each containing a lanthanoid at about 1000 mg/kg in 1 M nitric acid. As spike solutions of Fe and Mn, corresponding atomic absorption standard solutions (1000 μg/mL or ppm) purchased from Wako Chemical Co. Ltd. were used as supplied.
Experimental procedure Partitioning experimentPartitioning reactions were carried out in a specially designed vessel shown in Fig. 2a. Seawater in the vessel was purged by a gas mixture of carbon dioxide and nitrogen, whose mixing ratio was controlled using a combination of three gas-flow controllers (SEC-400 Series, Horiba Co. Ltd.). The purging gas mixture was passed through a water trap to saturate with H2O. It was continuously agitated using a pair of PTFE blades rotating on a PTFE-covered rod.
(a) Device for carbonate partitioning experiments (b) Graduated cylinder modified for size-separating particle samplings.
Firstly, 2 liters of seawater was placed in the reaction vessel. Typically 0.20 g of calcite (or aragonite) particles were added and were continuously agitated with the stirring blades being purged with bubbles of CO2 + N2 gas (pCO2: 300 ppm, flow rate: 50 mL/min) for a day. This process allows the seawater to be saturated/equilibrated with respect to calcite (or aragonite). Equilibrium was checked by monitoring the pH of the solution: pH steadily increased for the first several hours and reached a stable value of 8.1 within 20 hours in both calcite and aragonite experiments. On the following day, a certain amount of the REEs spike solution was added to the seawater. In some experiments, an aliquot of Fe or Mn spike solutions was also added. After adding spike solution(s), the seawater was stirred for seven days at room temperature (20–25°C) with bubbling of 300 ppm-pCO2 gas at flow rate of 50 mL/min. This reaction step is referred to as “Stage 1” to distinguish it from the following “Stage 2”. Seven days later, half of the seawater (1 L) containing particles was transferred to a modified graduated polypropylene cylinder for elutriation, which will be described later. When the amount of the REE spike solution was reduced, to ensure a measurable REE concentration in seawater, the amount of carbonate particles was also reduced from 0.2 to 0.02 g.
The remaining half (1 L) of seawater containing particles left in the vessel was reacted further by bubbling of 10000 ppm-pCO2 gas at a flow rate of 50 mL/min for five days. This reaction step, where the seawater was designed to go through under-saturation with respect to carbonate, is referred to as “Stage 2”. After five days, the rest of the seawater containing particles was transferred to the modified graduated cylinder for elutriation, which is described later. At Stage 1 of some experiments, only part of the solution underwent elutriation after the first week reaction, and the reaction was allowed to continue for another two weeks to see the effect of reaction time on the partitioning of REEs.
Size separation and elemental analysisElutriation was performed using a graduated cylinder bored at the graduation lines of 950, 800, 600, 400, 200 and 100 mL where each hole was sealed with a silicone-rubber septum (Fig. 2b). Seawater containing particles transferred to the graduated cylinder was left stationary for one hour. Seawaters above the septa (100 or 200 mL depending on the septum positions) were carefully sampled separately from the top to the bottom using a syringe through the septa. The size of collected particles was represented as those of particles at the central position between septa (1.7, 7, 14, 20.5, 28 and 33 cm from the top, respectively) and was calculated using Stokes’ law for particles with specific density 2.7. Typically about one third of the carbonate particles added as seeds were recovered by the sampling after the elutriation with the uppermost layer accounting less than 1%. A portion of the seawater samples was filtered by 0.45 μm membrane filters immediately after the sampling. The calcite particles collected on the filter were washed twice with 1 mL of milli-Q water and dissolved in 0.5 mL of 6 mol/L HNO3 solution. Seawater filtrates were acidified with HNO3 and were stored as seawater samples. The samples of dissolved carbonate particles and seawater, whose concentration level of REEs were typically well above the quantification level, were diluted 10 and 100 times, respectively, to reduce the interference with matrix elements, and the REE concentrations in the samples were measured by ICP-MS (Agilent 7900) and those of Ca, Mg, Fe, Mn and Sr were measured by ICP-OES (Agilent 5100).
REE preconcentrationWhen the concentration of spiked REEs in seawater was reduced to the level of natural seawater, the filtered seawater was subject to Fe coprecipitation. To 100 mL seawater 10 mg of Fe was added and the pH of the seawater was adjusted to 8. Fe hydroxide was recovered by filtration and washed with 10 mL of Milli-Q water and then dissolved by HNO3 acid. The recovery of REEs was better than 90%.
Oxide precipitation formationIn the experiments spiked elements in seawater are desirably under-saturated. This prerequisite is met for REEs and Mn, but not for Fe, as calculated as follows. Least soluble forms of Fe and Mn are hydroxide. Fe(OH)2 (Ksp = 10–15.1) and Mn(OH)2 (Ksp = 10–12.8) may precipitate when the concentrations are above 10–3 and 0.2 mol/kg, respectively, at pH 8.1, based on the reported solubility product (Feitknecht and Schindler, 1963). The concentrations of Mn or Fe in the vessel (9 μmol/kg) were too small for them to form hydroxides of Fe(OH)2 and Mn (OH)2. However, Fe(OH)3 may precipitate in oxidizing conditions (Ksp = 10–39). The formation of REE hydroxide is negligible based on their Ksp (>10–24) (Feitknecht and Schindler, 1963). REEs in seawater could precipitate as carbonates (Byrne and Kim, 1993). The least soluble carbonate of REEs is La2(CO3)3 (K0sp = 10–33.4), (Smith and Martell, 1976). According to Zhong and Mucci (1995), the concentration at which precipitation occurs is calculated by
| (2) |
where γT is CO32– total ion activity coefficient (0.037) (Morse and Mackenzie, 1990), XF is the molar fraction ([La3+]F/[La3+]T) and γF is the activity coefficient of free La3+ (0.10) (Millero, 1992). The initial concentration of La in experiments with 2 mL REE spike (0.51 μmol/kg) is comparable with the calculated upper concentrations (Stage 1: 0.19 μmol/kg at pH8.2, Stage 2: 1.4 μmol/kg at pH 7). Once the partitioning has started, the concentration of La decreases well below the upper concentration.
A portion of the spiked REEs might be removed by coprecipitation with Fe(OH)3, but the influence of removed REEs on the calculation of D can be ignored as long as REE amounts of dissolved and particulate phases were separately measureable.
Electron micro probe analysis (EPMA)Calcite particles collected on a membrane filter were transferred to a small pit on a set epoxy resin and embedded with a dab of molten resin. After solidifying the molten resin, the resin was polished with a diamond paste to the point where some particles appeared on the surface. The resin was carbon-coated, then EPMA mapping of Ca and Fe was carried out using a JXA-8530F at 5 nA and 15 keV electron bombardment. Specific X-ray signals were accumulated for 729 seconds using EDS detector (INCA 250XT).
The concentrations of La in the carbonate particles are indicated as La/Ca ratios and are plotted against reciprocal of particle size (r). Figure 3 shows the relationship between La/Ca and 1/r observed for experiments with calcite (Fig. 3a) and aragonite (Fig. 3b) seeds [REE(2)]. Figure 3 shows that the concentrations of La in particles broadly follow the theoretical relationship (See Appendix, eqs. (A3–5)) and that they approach constant values in all experiments except Fe-doped experiments with increasing 1/r, although ΔLa/ΔCa values calculated using eq. (A6) display significant errors. When Mn was spiked, the plateau in the relationship was wider especially in the experiments with calcite seeds (Fig. 3a), indicating greater depth of partitioning layers. In the experiments with Fe, constant values of La/Ca were not seen among different sized seeds, showing the presence of independent particles or thin layer concentrating REEs, which is confirmed later by electron-microscope observations. The concentrations of La in partitioning layer were estimated simply from the values of the smallest particles. Extending the reaction time to 3 weeks did not systematically widen the plateau in a calcite experiment [REE(2)-2(3W)].
Examples of elemental ratios of particles sampled after elutriation against reciprocal of particle radius (r in μm). (a) Calcite experiments [REE(2)], (b) Aragonite experiment [REE(2)] (Upper) Closed marks: La/Ca ratio of collected particles, Open marks: calculated ΔLa/ΔCa using eq. (A6). (Lower) Sr/Ca ratio of collected particles.
Aragonite is a metastable form of calcium carbonate. There is suspicion that the surfaces of aragonite seeds may transform to calcite during the experiments. However, X-ray diffraction of particles collected after the experiments detected no calcite, only aragonite. The Sr/Ca ratio can be used as a sensitive indicator of transformation from aragonite to calcite, since D of Sr in calcite (D = 0.1) is one tenth of that in aragonite (D = 1) (Graham et al., 1982; Lear et al., 2003; Schlanger, 1988). The lower figures of Fig. 3a, b are Sr/Ca in the experiments and they show that even smallest particles possess the characteristic Sr/Ca of carbonate seeds, indicating that transformation of carbonate minerals did not take place during the partitioning experiments.
Partitioning pattern of REEsFigure 4 shows the result of an experiment obtained using aragonite seeds [REE(2)], where REE partitioning values were calculated using eq. (1). Although only REE/Ca of the smallest particulates is considered to approach equilibrium partitioning in theory (see Appendix), the data from greater particles are also compared. In most experiments, the smallest particles showed the highest partitioning coefficients or those close to the highest, which agreed with theoretical considerations shown by eq. (A3). The shapes of partitioning patterns of greater particles were similar to those of the smallest particles.
Partitioning patterns of REEs in aragonite experiment [REE(2)]. Circles are for pCO2 300 ppm and crosses are for pCO2 10000 ppm. The lines show the partitioning pattern for the smallest sizes.
The results of REE partitioning values, D, of the smallest particles at both Stages 1 (pCO2 = 300 ppm) and 2 (pCO2 = 10000 ppm) are shown in logarithmic scale (Fig. 5). The vertical positions of partitioning patterns significantly varied over two orders of magnitude. With smaller amounts of REE spikes the D value generally increased (Fig. 5b, e).
Partitioning patterns of REEs obtained by smallest particles. a, b, c) Stage 1, pCO2 300 ppm, d, e, f) Stage 2, pCO2 10000 ppm. a, b, d, e) Experiments with calcite seeds, b, e) Reduced amounts of REE spike with smaller amounts of calcite seeds, c, f) Experiments with aragonite seeds.
Regarding the shape of the partitioning patterns, the patterns generally showed higher values in the L + MREE region and lower values in the HREE region. These features were similar to those reported for partitioning of REEs between calcite and seawater (Toyama and Terakado, 2014; Zhong and Mucci, 1995) (Fig. 1). The partitioning patterns at Stage 2 (pCO2 = 10000 ppm) in Fig. 5d, e, f generally showed more reduced values with lesser contrast across REEs than those at Stage 1 (pCO2 = 300 ppm) in Fig. 5a, b, c.
The partitioning patterns obtained when Mn + REE or Fe + REE spike was added at Stage 1 (pCO2 = 300 ppm) are also shown in Fig. 5a. The partitioning patterns with Mn + REE spike [REE(2) + Mn(1) and REE(2) + Mn(0.1)] differed only slightly from those of the experiments with spike of REEs only (Fig. 5a). The partitioning pattern with Fe + REE spike [REE(2) + Fe(1)] differed markedly from those of the experiments with REE spike and Mn + REE spike, when the volume of the REE spike was 1 or 2 mL (Fig. 5a): the partitioning coefficients were higher for all REEs than those of experiments with the same amount of REE spike [REE(2) and REE(2) + Mn(0.1)]. The discrepancy was most prominent in HREE region (Fig. 5a). These features are similar in both the experiments using calcite and aragonite (Fig. 5c). At Stage 2 (pCO2 = 10000 ppm), although partitioning values are significantly lower than at Stage 1 (Fig. 5d, e, f), the addition of Fe rendered similar enhancement in partitioning values to that observed at Stage 1; that of Mn made no difference (Fig. 5d). Aragonite experiments gave slightly steeper slopes (REE and REE + Mn in Stage 2) or higher D values (REE + Fe in Stages 1 and 2) (Fig. 5c, f) in partitioning patterns than the calcite experiments.
The partitioning values were not systematically affected by the prolonged reaction time [REE(2)-2(3W), REE(0.01)-3(2W), REE(0.002)(2W) and REE(0.001)(3W)] (Fig. 5a, b, d, e). From this it was determined that one week could be used as the standard reaction time for Stage 1.
pH of experimental solutionsThe pH rapidly increased at the onset of experiments in Stage 1 and reached constant values 24 hours later and remained at almost constant values until the end of the stage. The final pHs of the experiments are summarized in Table 1. In Stage 1 pH ranged from 7.97 to 8.35, whereas in Stage 2 from 6.56 to 7.53. Experiments with aragonite tended to give a slightly lower pH than those with calcite. In experiments with calcite the addition of foreign ions such as Fe and Mn did not necessarily decrease the final pH, even though they were added as acidic solutions. The final pH of the reduced REE experiments [REE(0.02)-REE(0.001)] tended to be similar to that for unreduced REE experiments at Stage 1, but remained higher at Stage 2.
| Conditiona) | Stage 1 pCO2 = 300 ppm | Stage 2 pCO2 = 10000 ppm | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| pH | logIAP | Fe/Cac) | Mn/Cac) | La/Cac) | Mg/Cac) | pH | logIAP | Fe/Cac) | Mn/Cac) | La/Cac) | Mg/Cac) | |
| Calcite | ||||||||||||
| REE(2)-1 | 8.21 | –7.92 | — | 0.24 | 1.15 | 20 | 6.77 | –9.27 | — | 0.15 | 0.26 | 13 |
| REE(2)-2(3W) | 8.30 | –7.72 | 0.62 | 0.17 | 0.71 | — | 6.93 | –10.45 | 0.23 | 0.26 | 1.2 | — |
| REE(1) | 8.28 | –7.77 | 0.29 | 0.20 | 0.83 | 17 | 6.90 | –9.00 | 0.28 | 0.19 | 0.86 | 12 |
| REE(4) | 7.97 | –8.39 | 0.12 | 0.24 | 3.9 | 28 | 6.56 | –9.68 | 0.84 | 0.24 | 2.9 | 18 |
| REE(2) + Mn(1) | 8.29 | –7.74 | 0.85 | 1.1 | 6.2 | — | 6.92 | –8.95 | 2.7 | 0.19 | 4.7 | — |
| REE(2) + Fe(1) | 8.31 | –7.69 | 26 | 0.28 | 8.7 | — | 6.90 | –8.99 | 21 | 0.13 | 6.4 | — |
| REE(2) + Mn(0.1) | 8.29 | –7.74 | 0.70 | 0.82 | 2.37 | — | 6.93 | –8.95 | 0.13 | 0.43 | 2.3 | — |
| REE(2) + Fe(0.1) | 8.31 | –7.69 | 2.65 | 0.16 | 0.99 | — | 6.91 | –8.97 | 2.61 | 0.12 | 2.8 | — |
| REE(2) + Mn(1) + Fe(1) | 8.35 | –7.60 | 0.28 | 0.00 | 3.7 | 7.0 | 6.97 | –8.84 | 270 | 0.20 | 2.5 | 5.9 |
| REE(0.02) | 8.06 | –8.23 | 26.7 | 0.89 | 0.38 | 420 | 7.53 | –9.28 | 9.9 | 0.84 | 0.25 | 710 |
| REE(0.01)-1 | 7.06 | –10.21 | 34.0 | 0.88 | 0.30 | 700 | ||||||
| REE(0.01)-2 | 8.26 | –7.74 | 47.5 | 0.49 | 0.14 | 510 | 7.5 | –9.34 | 14.9 | 0.48 | 0.10 | 410 |
| REE(0.01)-3 | 8.20 | –7.82 | 11.5 | 0.27 | 0.12 | 490 | ||||||
| REE(0.01)-3(2W) | 8.19 | –7.93 | 12.5 | 0.46 | 0.12 | 320 | ||||||
| REE(0.01)-4 | 7.43 | –9.47 | 8.3 | 0.49 | 0.038 | 630 | ||||||
| REE(0.01)-4(2W)b) | 7.40 | –9.53 | 7.8 | 0.42 | 0.098 | — | ||||||
| REE(0.002) | 8.21 | –7.95 | 500 | 0.42 | 0.31 | — | ||||||
| REE(0.002)(2W) | 8.20 | –7.94 | 68 | 0.41 | 0.12 | 360 | 7.11 | –10.12 | 24.6 | 0.95 | 0.091 | 1850 |
| REE(0.001) | 8.31 | –7.71 | 262 | 0.37 | 0.093 | — | ||||||
| REE(0.001)(3W) | 8.26 | –7.82 | 21.6 | 0.31 | 0.050 | — | 7.08 | –10.17 | 72.7 | 1.4 | 0.32 | 1540 |
| REE(0.001)(7W)b) | 6.96 | –10.43 | 47.9 | 1.0 | 0.27 | 1770 | ||||||
| Aragonite | ||||||||||||
| REE(2) | 8.11 | –8.13 | 0.14 | 0.85 | 3.0 | — | 6.73 | –9.35 | 1.1 | 0.35 | 3.1 | — |
| REE(2) + Mn(1) | 8.12 | –8.10 | 0.24 | 0.34 | 2.7 | — | 6.74 | –9.33 | 1.9 | 0.19 | 4.2 | — |
| REE(2) + Fe(1) | 8.09 | –8.16 | 37 | — | 3.5 | — | 6.79 | –9.22 | 43 | 0.10 | 6.1 | — |
— not measured or determined,
a) Hyphenated figures are run number of a condition. Figures in parentheses stand for volume of spike solution in ml. 2W, 3W, 7W in parentheses are time lengths of Stage 1 in week. Otherwise it is a week in Stage 1 and 5 days in Stage 2. In experiments with volumes of REE spike solution <1, weight of calcite in the system was reduced from 200 mg to 20 mg, b) time length of Stage 2 c) × 10–3 (in permil)
Fe/Ca ratios of particles are listed in Table 1. When the amount of REE spike was high, Fe was detected only in the Fe-spiked experiments. In the case of the reduced REE- spike experiments, since much smaller amounts of Ca particles were introduced to the system, the Fe/Ca ratio was sometimes significantly high. This is due to accidental contamination of Fe, which cannot be overlooked in the experiments with the reduced REE spike.
The carbonic system can be described in terms of pCO2, and activities of dissolved CO2, HCO3–, CO32–, Ca2+ and H+. They are interrelated by four constants: the Henry constant of CO2 (KH), 1st and 2nd acid dissociation constants of dissolved CO2 (Ka1 and Ka2) and the solubility product of CaCO3 (Ksp). Therefore, if two of six parameters are known, the system can be described in theory. In the case of this study, however, Ksp may vary according to the surface condition (impurity, surface area, etc.) of seed crystals. We treated Ksp as an unknown parameter and calculated ion activity product, IAP, of Ca2+ and CO32– using available or measurable three parameters (pCO2, Ca2+ activity and pH) using eq. (3).
| (3) |
where γCa2+ is activity coefficient of Ca2+ at ionic strength, I, =0.7. The parameters used in the calculation are KH = 0.03405, Ka1 = 4.45 × 10–7, Ka2 = 4.69 × 10–11 and γCa2+ = 0.20 (Morse and Mackenzie, 1990). The results of calculation are summarized in Table 1. If the solution is in equilibrium with respect to calcite (or aragonite) dissolution, log IAP should be close to log Ksp [–8.3 (or –8.1)] (Morse and Mackenzie, 1990).
In Stage 1 of calcite-seeded experiments, the calculated IAP was mostly greater than the reported values of calcite, but was similar to Ksp of aragonite, indicating oversaturation. Transformation of calcite to aragonite was unlikely to have happened given the consistent Sr/Ca ratio of particles (Fig. 3). Ksp depends on surface energy or particle size and it was conceivable that Ksp became greater than the ideally-determined value in the experiments. The concomitant Mg can increase the solubility of calcite (Walter, 1984). This should partially explain the increased IAP. However, this effect is not clear, as many particles with higher Mg/Ca in the reduced REE experiments did not always exhibit higher IAP. In Stage 2 of calcite experiments, the calculated IAP was always smaller by a factor of more than 5 than the reported Ksp, indicating undersaturation. The decrease in IAP was more significant in the aragonite experiments, which showed smaller values by a factor of more than 10. This corresponds to the state where dissolution is prohibited. The formation of calcite was reported to be heavily prohibited by the presence of REEs (Akagi and Kono, 1995; Zhong and Mucci, 1995). The present study is the first to show that the dissolution of calcite/aragonite may also be prohibited by the presence of REEs.
Incidentally, in Stage 2 of the reduced REE experiments, it was suspected that the collected particles may be dolomite, less soluble than calcite, based on Mg/Ca ratio around 1 in molar (1.65 in weight). This dolomite formation may explain IAP as small as 10–10 (Sherman and Barak, 2000).
This study revealed that REE presence prohibited both dissolution and growth of carbonate. It is likely that dissolution and growth sites are identical and are blocked by REE ion attachment. In the context of precise D determination, these prohibitions can be problematic. The amount of Ca as calcite (Ca in numerator in eq. (1)) can be biased both negatively and positively by prohibition of crystal growth and dissolution, respectively, leading to overestimation and underestimation of D values, respectively. The addition of either Fe or Mn in the systems seemed not to conspicuously affect IAP or the solubility of calcite and aragonite.
Effect of foreign elementsMn and Fe are considered to be good absorbers of metals and REEs (De Carlo et al., 2000; Goldberg, 1954; Koeppenkastrop and De Carlo, 1992) and in this study the influence of their presence was studied.
Mn concentration in the finest calcite particles was higher than that in the finest aragonite particles; D of Mn in calcite was 1 and that in aragonite was 0.1 in our experiments. The recovery of Mn collected with calcite particles was 0.45%, much less than the recovery of calcite (19.7% in experiment REE(2) + Mn(1) + Fe(1)) and the amount of Mn collected with calcite from each layer was found to be almost proportional to the weight of particles rather than to the surface area of particles (e.g., Fig. 6), where surface area was estimated using eq. (A13) in Appendix. Proportionality of Mn amount to the weight of particles was less clear in the case of the aragonite experiment. MnCO3 can form a solid solution with calcite, but not aragonite (Böttcher, 1998). It is considered that this property may have something to do with the above observations with Mn. In the partitioning pattern, it is obvious that Mn exerts only minor, almost unnoticeable, influence on REE partitioning for both calcite and aragonite.
Relationship of Fe and Mn amounts collected with calcite particles against surface area (a, c) or weight (b, d) of calcite particles per 100 mL elutriation layers in experiment [REE(2) + Mn(1) + Fe(1)]. The surface area was estimated using equation (A13) in Appendix.
In the case of Fe, no plateau was obtained in the plots of Ln/Ca against 1/r (Fig. 3). The electron microprobe analysis (EPMA) of calcite particles indicates that a thin layer of Fe covers the surface (the asterisk in Fig. 7) or that Fe forms cloud-like particles (the arrows in Fig. 7) on calcite surface. No independent phase of Fe oxide was observed. Recovery of Fe collected with calcite particles was 28%, higher than the recovery of carbonate (19.7% in experiment REE(2) + Mn(1) + Fe(1)), implying that Fe tends to be enriched in finer particles [Note that in the elutriation, the greatest-sized particles that sank on the bottom of the graduated cylinder were not recovered.]. Figure 6 shows that, unlike Mn, the Fe amount in calcite particles was proportional to the total surface area of calcite particles in the Mn + Fe doped experiment [REE(2) + Mn(1) + Fe(1)]. This proportionality was always seen in any Fe doped experiments with both calcite and aragonite seeds [REE(2) + Fe(1), REE(2) + Fe(0.1)]. Fe concentration in seawater was too low to be detected. In the experimental condition it is considered that Fe2+ may have oxidized to Fe3+ and that Fe3+ is evenly distributed on the surface of carbonate particles as hydroxide. Fe oxide may exist as fine colloidal particles in seawater, which can pass through pores of the membrane filter (Nishioka et al., 2001). It was considered that most of the Fe colloidal particles were absorbed by carbonate particles in these experiments. This consideration based on the mathematics on surface area and the EPMA observation is also supported by the color of the calcite particles collected after Fe-spiked experiments being stained uniformly pale brown.
(a) Electron microscopic images of a calcite seed after an experiment with Fe spike. (b) Ca Kα image, and (c) Fe Kα. The asterisk and arrows indicate the thin layer of Fe on and attachment of Fe to the surface calcite, respectively.
Fe may change the partitioning considerably (Fig. 8). Both the experiments with calcite and aragonite gave the almost identical partitioning patterns [REE(2) + Fe(1)], when Fe was added (Fig. 5a, c, d, f). It is interesting to note that an M-type tetrad effect (Masuda et al., 1987) seems to emerge with increasing addition of Fe, which is commonly seen in Fe hydroxide phases (Bau, 1999; Liu et al., 2017; Ohta and Kawabe, 2000; Quinn et al., 2006a). Ce/La values of D increased with Fe/Ca ratios (Fig. 8b). This is the feature reported for REE absorbed by Fe hydroxide (Quinn et al., 2006b; Schijf and Marshall, 2011). It is likely that Fe hydroxide may be responsible for the partitioning in both the experiments with calcite and aragonite. Fig. 9 shows differences in D values with [REE(2) + Fe(1)] and without doped Fe [REE(2)]. The differences show almost identical patterns to D reported for Fe hydroxide (Ohta and Kawabe, 2000). They reported patterns of D at varying pH of 0.5 M NaCl + NaHCO3 aqueous solutions (I = 0.5), not seawater. Although the D data of our experiments may carry significant uncertainties in absolute value, our differences correspond very well to the reported D values at some specific pHs. We conclude that Fe forms hydroxide on calcite surface and partitions REEs with specific partitioning patterns.
Relationship of Fe/Ca of particles with (a) DLa, (b) DCe/DLa and (c) DLu/DLa for Stage 1 (red) and Stage 2 (blue). Open circles show the results of experiments with a greater volume (>1 mL) of REE spike solution and filled circles those with a smaller volume (<1 mL).
(a) Difference in Ds between Fe-spiked experiment and non-Fe experiments. Bold lines with filled circles are for calcite experiments, broken lines with open squares are for aragonite experiments. The pH range in the two experiments with and without the addition of Fe spike are shown in the figure. (b) Apparent partitioning patterns of iron hydroxide at different pHs by Ohta and Kawabe (2000).
In the case of the reduced REE-spike experiments with smaller amounts of calcite particles, accidental contamination of Fe cannot be overlooked. Absolute D values are significantly correlated with Fe/Ca values (Fig. 8a). In both Stages 1 and 2, D values display a significant correlation with Fe/Ca ratios of the carbonate particles, regardless of amounts of REE spike. This indicates that Fe content is the primary factor controlling the REE partitioning of carbonate.
Correcting for the Fe effect in REE partitioningFrom Fig. 8a, when Fe/Ca values are small enough, the influence of Fe should be small. The data chosen with a criteria of Fe/Ca <0.001 or 0.015 are then examined further against other parameters (Figs. 10 and 11). This criterion of Fe/Ca <0.001 reduces the contribution of Fe to DLa to 1/10 at most, but the possibility of overestimating DLu still remains.
The parameters which showed a good correlation with D values were IAP (Fig. 10) and pH. IAP and pH are correlated with each other and it is not certain which one of the two is more influential. The proportion of each chemical species varies nearly monotonously across REEs (e.g. Liu et al. (2017)), and we next examined if the changes in the slope of D and absolute D value were due to differences in pH, taking DLu/DLa and DLa as representatives, respectively. Based on thermodynamic calculation (Akagi et al., 2004), none of the chemical species (Ln3+, LnCO3+, Ln(CO3)2– and LnOH2+) exhibit variations in either Lu/La (Fig. 11b) or proportion (not shown) with pH change that match DLu/DLa (Fig. 11a) or absolute DLa (not shown) across the wide pH range observed in Stage 1 and 2. We thus considered that IAP was the main factor controlling D values. In terms of IAP, the higher D values can be translated by increased prohibition of carbonate formation when log (IAP) is greater than –8.3 and lower D values by increased prohibition of carbonate dissolution when log (IAP) is smaller than –8.3. This interpretation may lead to D values around log(IAP) = –8.3 as the true D values of carbonate. It is noteworthy that the slope of D across the span of REEs is almost identical when log(IAP) is closer to –8.3 between the two stages, although pHs differ by one unit [REE(4) of Stage 1, REE(2) + Mn(0.1) of Stage 2] (Fig. 10b). We consider that greater Ds around Pr (Fig. 5) is only the reflection of similarity in ionic radius between Ca2+ and Pr3+ in carbonate lattices and that they are not influenced significantly by the pH of the solution. The observed variation in slope with pH may result from the presence of Fe hydroxide and/or inhibition of carbonate growth/dissolution.
Relationship of ion activity product (IAP) of solutions with (a) DLa and (b) DLu/DLa for Stage 1 (red) and Stage 2 (blue). Particles with Fe/Ca <0.001 and <0.015 are selected in (a) and (b), respectively. Open circles show the results of experiments with a greater volume (>1 mL) of REE spike solution and filled circles those with a smaller volume (<1 mL).
(a) Relationship with pH of solutions and DLu/DLa for Stage 1 (red) and Stage 2 (blue). Particles with Fe/Ca <0.015 are selected. Open circles show the results of experiments with a higher volume (>1 mL) of REE spike solution and filled circles those with a smaller volume (<1 mL). The changes of a REE species estimated following thermodynamic calculation for Stage 1 (red line) and Stage 2 (blue line) are drawn based on (b). Note that the blue circles do not fall on the blue line, when the red circles are adjusted to fall on the red line. (b) Thermodynamic calculation of Lu/La ratios of four REE species in solutions for Stage 1 (pCO2 = 300 ppm, red line) and Stage 2 (pCO2 = 10000 ppm, blue line).
From the observed relationship between D and log(IAP) (Fig. 10) we propose true carbonate D should be obtained by interpolation as those at log(IAP) = –8.3. The D values thus calculated are summarized in Table 2 and Fig. 12. They range from 102 to 10 levels. Note that there remains the possibility of overestimating HREE due to Fe hydroxide absorption, which is difficult to correct for. This implies that the downward slope can be steeper than that shown in Fig. 12. The Mg/Ca values for those runs are around 10–2 and the influence of Mg on calcite solubility is considered negligibly small (Walter, 1984).
| La | 62 |
| Ce | 80 |
| Pr | 71 |
| Nd | 58 |
| Sm | 53 |
| Eu | 43 |
| Gd | 33 |
| Tb | 24 |
| Dy | 18 |
| Ho | 13 |
| Er | 10 |
| Tm | 9.0 |
| Yb | 8.5 |
| Lu | 7.4 |
Fig. 1 plus finally determined D in this study (gray).
We next compare reported partitioning patterns with the finally determined partitioning patterns (Fig. 12). The REE partitioning patterns were estimated from biogenic calcite in sediments (Palmer, 1985; Parekh et al., 1977; Scherer and Seitz, 1980) and also obtained by experiment (Tanaka and Kawabe, 2006; Toyama and Terakado, 2014; Zhong and Mucci, 1995). All the reported partitioning patterns are shown in Fig. 12, along with our final pattern. Our proposed partition values are lowest in the figure.
In nature Fe is always present with a carbonate phase. It is observed the acetic acid soluble fraction of settling particles collected in the Bering Sea has Fe/Ca ratios as high as 0.01 (Emoto et al., 2019). It is reported that the foraminifer tests are covered with an “FeMn-rich coating” (Palmer, 1985). Foraminifer tests have FeMn-rich crusts enriched with Nd in their cavities (Tachikawa et al., 2014). The estimated partitioning patterns from naturally-occurring biogenic calcite show a convex shape around L and MREEs with higher D values around LREEs, and much higher D values in the HREE region compared with the D values obtained by this study. These features are reproduced by our experiments with Fe spike (Fig. 5a and b). It is highly suspected that high D values observed in nature are due to a concomitant Fe phase.
Some partitioning patterns obtained experimentally (Tanaka and Kawabe, 2006; Toyama and Terakado, 2014; Zhong and Mucci, 1995) show a wide range of values. In all of these studies, partitioning was carried out under supersaturated conditions with respect to calcite. In addition, the amounts of REE spike were quite high (ppm to ppb level for each REE) with one exception (Toyama and Terakado, 2014). We consider most studies to be subject to the effect of calcite growth prohibition and likely overestimated the D value. It is interesting to note that (Toyama and Terakado, 2014) employed the smallest spike amount (ppt level) and reported the smallest values.
If the reported data are also compromised by the presence of Fe oxide, pH can influence the apparent D values. Unfortunately, none of those studies analyzed the carbonate particles for Fe. The D patterns reported by Tanaka and Kawabe (2006) as flat were actually obtained for pH 6.6; the steepest D reported by Zhong and Mucci (1995) are obtained by the highest pH of 8.3. These variations with pH are consistent features in the experiments of Fe hydroxide absorption (Fig. 9b). It is suspected that the reported D values can be doubly biased by Fe and the REE prohibition effect of calcite growth.
Comparison with observed and calculated partitioning patternsThe REE partitioning patterns estimated in this study have features similar to partitioning patterns estimated from some observational data. The observational data are not reported as the partitioning of REEs between calcite and seawater, but calculated as the ratio between REE concentration in the acetic acid soluble fraction of suspended or settling particles and seawater. [Settling particles in the Pacific Ocean from Emoto et al. (2019); Lerche and Nozaki (1998) and suspended particles in the Atlantic Ocean from Sholkovitz et al. (1994)]. The partitioning patterns shown here were obtained by averaging partitioning patterns at all measured depths, if available (Lerche and Nozaki, 1998; Sholkovitz et al., 1994). Because the exact quantity of carbonate is unknown, the normalized partitioning patterns in which partitioning coefficient of La is normalized to 1 are shown in Fig. 13. The acetic acid soluble fraction is likely to contain REEs from Fe hydroxide as well as from carbonates.
La-normalized partitioning patterns between acetic acid soluble fraction and deep seawater in the Atlantic and Pacific Oceans and Bering Sea. The gray bold line is the finally determined partitioning pattern in this study.
The observed partitioning pattern of Pacific is flatter than that of Atlantic, when the values are normalized to La. Our finally determined pattern is similar to that of Atlantic Ocean. In general, pH of deep water is lower in Pacific Ocean than in Atlantic Ocean.
Fe/Ca (w/w) ratio in the observational data of the acetic-acid-soluble fraction of settling particles is in the range of 0.005 and 0.017 (Emoto et al., 2019), which is of a similar order to our experimental Fe/Ca ratio. The distribution pattern for the Bering Sea is closer to the distribution patterns of the experiments [REE(0.01)-3], where similar amounts of Fe were detected. The suspended particles studied by Sholkovitz et al. (1994) also contain significant amount of Fe, but they did not report the amount of Ca and thus Fe/Ca values are unknown. We conclude that the acetic acid soluble fraction may contain Fe hydroxide as well as calcite and the D values obtained from acetic acid fraction may be influenced by Fe hydroxide. As the influence of pH on the slope of carbonate D is small following the discussion of this study, it may be the result of the presence of Fe hydroxide—REE distribution on which is strongly pH-dependent. Although absolute D values are unquestioned, they can be as high as 103 level due to a concomitant Fe phase.
As stated in the introduction, we argue that secondary scavenging of the REEs is responsible for the differentiating vertical profiles among REEs (Nishino and Akagi, 2019). Using an appropriate algorithm, we have estimated the partitioning patterns of the secondary scavenging. The pattern obtained by the algorithm (Nishino and Akagi, 2019) should reflect the effect of multiple carriers including organic matter, which can complex higher REEs selectively (Byrne and Kim, 1993). We also argued that the secondary scavengers are likely to be carbonate particles mainly based on the concomitant decreases of scavenged proportion with carbonate saturation depths (Nishino and Akagi, 2019). This study shows, however, that Fe hydroxide on carbonate, and not the carbonate itself, is an important scavenger of REEs. It is believed that carbonate phases of settling particles work as ballast contributing to vertical material transport (Klaas and Archer, 2002). It is likely that carbonate contributes to scavenging by transporting the Fe hydroxide phase vertically being entrained on their surface. This action should reduce the residence time of LREE in oceanic columns.
Here we have discovered several mechanisms to affect the partitioning of REEs in calcite with seawater. Firstly, calcium carbonate is a good absorber of Fe hydroxide. Fe hydroxide on carbonate is identified as the important substance to determine D. It is suspected that previously reported D values are likely to be distorted by Fe hydroxide. The partitioning pattern of REEs in calcite may not significantly vary with pH change. Fe attached to the surface of calcite in the separate form of Fe hydroxide and enhanced the partitioning and changed the pattern shape depending on pH. Secondly, REE can prohibit growth as well as dissolution of carbonate, which may lead to overestimation of D values in carbonate super-saturated conditions. These two factors may explain the wide variation in the reported D values.
To estimate partitioning coefficients of a metal, the equilibrium concentrations of the metal in partitioning layers of calcite surface are a prerequisite. In this study they were estimated by using the relationship between concentration and particle size, assuming that calcite particles reacted in seawater have a certain depth of equilibrium-partitioning layer. For the sake of simplicity, the particles are assumed to be spherical. When the concentrations of a metal in a prepared calcite particle and in the equilibrium coats are Ccalcite and Cpartitioning layer, respectively, the concentration of the metal in a calcite particle after the reaction is expressed as
| (A1) |
In the equation, d and r are thickness of partitioning layer and radius of calcite particles, respectively. When
| (A2) |
this formula is rewritten to
| (A3) |
This equation indicates that larger particles (r >> d) show a lower concentration than that of partitioning layer and that smaller particles (r ≈ d) tend to show the concentration closer to that of the partitioning layer, i.e.
| (A4) |
and
| (A5) |
It further shows, when measured concentration is plotted against the reciprocal of particle size, the equilibrium concentration appears as the asymptotic values of infinite 1/r. In theory, it allows us to work out both thickness, d, and concentration of partitioning layer by fitting the observation values to eq. (A1 or A3). The foreign metal concentration (M/Ca) in our calcite and aragonite seeds were negligibly smaller than those in partitioning layer as shown in the Method section, which provides a rationale with eq. (A3).
However, in reality, it is not so simple. First, to apply the experimental data, data pre-handling is necessary. Because the elutriation of this study started with homogenous distribution of different-sized particles in seawater throughout the graduated cylinder, particles in the upper 100 or 200 mL layer after the elutriation are supposed to be contained in the all layers below (Fig. A1). To obtain the relationship between metal concentration and particle sizes we carried out a mathematical pretreatment of data as prescribed below.
Schematic image of elutriation experiment.
The concentration of particles in the n-th layer which is corrected for the presence of the particles in the upper (n – 1)-th layer is
| (A6) |
Mn and Can are the concentrations of foreign metals and calcium in the n-th layer, respectively. The largest particle sizes of the n-th layer were estimated using Stokes’ law. Unfortunately, this handling amplifies experimental error and fitting the results to eq. (A6) is made tricky.
Moreover, the concentration can also be biased by the prohibition of dissolution/crystal growth by REEs. This means that Cpartitioning layer may fluctuate significantly. As the overall patterns of D are consistent within each experiment (see Fig. 3), we simply adopted the concentration data of the smallest particles collected from the uppermost layer as Cpartitioning layer.
Estimation of surface area in each elutriation layerAssuming the particles spherical, the total weight (W) and surface area (S) of size-separated particles is expressed as
| (A7) |
and
| (A8) |
respectively.
Eliminating n from the above two equations,
| (A9) |
Because W is proportional to weight of Ca in the size-separated layer, total surface layer of 1st layer (S1) is evaluated using total weight (W1) and size of the size-separated particles (r1) as
| (A10) |
where k is the appropriate proportionality constant. Can is the amount of Ca in n-th size-separated layer. Since elutriation was started from a homogeneous distribution with respect to particle size, the finest particles in the 1st layer should also be distributed to all layers bellow (Fig. A1). Therefore, the surface area of 2nd layer is
| (A11) |
Again the particles of these size ranges are considered to be distributed in all layers of n > 2, the surface area of the 3rd layer is
| (A12) |
To generalize the surface area of n-th layer,
| (A13) |
is obtained. In the main text, the comparisons of Sn or Wn with Fen or Mnn are made to evaluate the behavior of foreign metals, where Mn is the amount of M in n-th size-separated layer.
We would like to thank Dr Hiroshi Tsuno, Yokohama National University, for his productive discussion. This study is supported by Grant-in-Aid from MEXT (No. 17K05717).
