We present a procedure to determine the metal-silicate partitioning of siderophile elements, emphasizing techniques that suppress the formation of metal inclusions in experiments for the noble metals. The results of these experiments are used to demonstrate the effect of core-formation on the highly siderophile element composition of the mantle.
Estimates of the primitive upper mantle (PUM) composition reveal a depletion in many of the siderophile (iron-loving) elements, thought to result from their extraction to the core during terrestrial accretion. Experiments to investigate the partitioning of these elements between metal and silicate melts suggest that the PUM composition is best matched if metal-silicate equilibrium occurred at high pressures and temperatures, in a deep magma ocean environment. The behavior of the most highly siderophile elements (HSEs) during this process however, has remained enigmatic. Silicate run-products from HSE solubility experiments are commonly contaminated by dispersed metal inclusions that hinder the measurement of element concentrations in the melt. The resulting uncertainty over the true solubility and metal-silicate partitioning of these elements has made it difficult to predict their expected depletion in PUM. Recently, several studies have employed changes to the experimental design used for high pressure and temperature solubility experiments in order to suppress the formation of metal inclusions. The addition of Au (Re, Os, Ir, Ru experiments) or elemental Si (Pt experiments) to the sample acts to alter either the geometry or rate of sample reduction respectively, in order to avoid transient metal oversaturation of the silicate melt. This contribution outlines procedures for using the piston-cylinder and multi-anvil apparatus to conduct solubility and metal-silicate partitioning experiments respectively. A protocol is also described for the synthesis of uncontaminated run-products from HSE solubility experiments in which the oxygen fugacity is similar to that during terrestrial core-formation. Time-resolved LA-ICP-MS spectra are presented as evidence for the absence of metal-inclusions in run-products from earlier studies, and also confirm that the technique may be extended to investigate Ru. Examples are also given of how these data may be applied.
Terrestrial accretion is thought to have occurred as a series of collisions between planetesimals with a chondritic bulk composition, terminating in a giant-impact phase thought responsible for moon formation1,2. Heating of the proto-earth by impacts and the decay of short-lived isotopes was sufficient to cause extensive melting and the formation of a magma ocean or ponds through which dense Fe-rich metallic melts could descend. Upon reaching the base of the magma ocean, metallic melts encounter a rheological boundary, stall, and undergo final metal-silicate equilibrium before eventually descending through the solid mantle to the growing core2. Further chemical communication between metal and silicate phases as metallic melt traverses the solid portion of the mantle is thought to be precluded due to the large size and rapid descent of metal diapirs3. This primary differentiation of the Earth into a metallic core and silicate mantle is revealed today by both geophysical and geochemical observations4–6. Interpreting these observations to yield plausible conditions for metal-silicate equilibrium at the base of a magma ocean, however, requires an appropriate database of experimental results.
The primitive upper mantle (PUM) is a hypothetical reservoir comprising the silicate residue of core formation and its composition therefore reflects the behaviour of trace elements during metal-silicate equilibrium. Trace elements are distributed between metal and silicate melts during core segregation on the basis of their geochemical affinity. The magnitude of an elements preference for the metal phase can be described by the metal-silicate partition coefficient
(1)
Where and denote the concentration of element i in metal and silicate melt respectively. Values of >1 indicate siderophile (iron-loving) behaviour and those <1 lithophile (rock-loving) behavior. Estimates of the PUM composition show that siderophile elements are depleted relative to chondrites7, typically considered as representative of Earth’s bulk composition6,8. This depletion is due to sequestration of siderophile elements by the core, and for refractory elements its magnitude should directly reflect values of . Lab experiments therefore seek to determine values of over a range of pressure (P), temperature (T) and oxygen fugacity (fO2) conditions that are relevant to metal segregation from the base of a magma ocean. The results of these experiments may then be used to delineate regions of P–T–fO2 space that are compatible with the PUM abundance of multiple siderophile elements (e.g.,9–11).
The high pressures and temperatures relevant to a magma ocean scenario can be recreated in the laboratory using either a piston-cylinder or multi-anvil press. The piston-cylinder apparatus provides access to moderate pressure (~2 GPa) and high temperature (~2,573 K) conditions, but enables large sample volumes and a variety of capsule materials to be easily used. The rapid cooling rate also permits quenching of a range of silicate melt compositions to a glass, thus simplifying textural interpretation of the run-products. The multi-anvil apparatus typically employs smaller sample volumes but with suitable assembly designs can achieve pressures up to ~27 GPa and temperatures of ~3,000 K. The use of these methods has allowed partitioning data for many of the moderately and slightly siderophile elements to be gathered over a large range of P–T conditions. Predictions of the PUM composition based on these data suggest metal-silicate equilibrium occurred at average pressure and temperature conditions in excess of ~29 GPa and 3,000 K respectively, although the exact values are model dependent. In order to account for the PUM abundance of certain redox sensitive elements (e.g., V, Cr) the fO2 is also thought to evolve during accretion from ~4 to 2 log units below that imposed by co-existing iron and wüstite (FeO) at equivalent P-T conditions (the iron-wüstite buffer)12.
Although the PUM abundance of many siderophile elements can be accounted for by metal-silicate equilibrium at the base of a deep magma ocean, it has proved difficult to assess if this situation also applies to the most highly siderophile elements (HSEs). The extreme affinity of the HSEs for iron-metal indicated by low pressure (P ~0.1 MPa) and temperature (T <1,673 K) experiments suggests the silicate earth should be strongly depleted in these elements. Estimates of the HSE content for PUM, however, indicate only a moderate depletion relative to chondrite (Figure 1). A commonly posited solution to the apparent HSE excess is that Earth experienced a late-accretion of chondritic material subsequent to core-formation13. This late-accreted material would have mixed with the PUM and elevated HSE concentrations but had a negligible effect on more abundant elements. Alternatively, it has been suggested that the extremely siderophile nature of HSEs indicated by low P–T experiments does not persist to the high P-T conditions present during core-formation14,15. In order to test these hypotheses, experiments must be carried out to determine the solubility and metal-silicate partitioning of HSEs at appropriate conditions. Contamination of the silicate portion of quenched run-products in many previous studies however, has complicated run-product analysis and obscured the true partition coefficients for HSEs between metal and silicate melts.
In partitioning experiments where the HSEs are present at concentration levels appropriate to nature, the extreme preference of these elements for Fe-metal prevents their measurement in the silicate melt. To circumvent this problem, solubility measurements are made in which the silicate melt is saturated in the HSE of interest and values of are calculated using the formalism of Borisov et al.16. Quenched silicate run-products from HSE solubility experiments performed at reducing conditions, however, often display evidence for contamination by dispersed HSE±Fe inclusions17. Despite the near ubiquity of these inclusions in low fO2 experiments containing Pt, Ir, Os, Re and Ru, (e.g., 18–27), there is notable variability between studies in their textural presentation; compare for example references 22 and 26. Although it has been demonstrated that inclusions can form which are a stable phase at the run conditions of an experiment28, this does not preclude the formation of inclusions as the sample is quenched. Uncertainty surrounding the origin of inclusions makes the treatment of analytical results difficult, and has led to ambiguity over the true solubility of HSEs in reduced silicate melts. Inclusion-free run-products are required to assess which studies have adopted an analytical approach that yields accurate dissolved HSE concentrations. Considerable progress in suppressing the formation of metal-inclusions at reducing conditions has now been demonstrated in experiments using a piston-cylinder apparatus, in which the sample design was amended from previous studies by adding either Au or Si to the starting materials29–31. The addition of Au or elemental Si to the starting materials alters the sample geometry or fO2 evolution of the experiment respectively. These methods are intended to suppress metal inclusion formation by altering the timing of HSE in-diffusion versus sample reduction, and are discussed in Bennett et al.31. Unlike some previous attempts to cleanse the silicate melt of inclusions, such as mechanically assisted equilibration and the centrifuging piston-cylinder, the present protocol can be implemented without specialized apparatus and is suitable for high P-T experiments.
Described in detail here is a piston-cylinder based approach to determine the solubility of Re, Os, Ir, Ru, Pt and Au in silicate melt at high temperature (>1,873 K), 2 GPa and an fO2 similar to that of the iron-wüstite buffer. Application of a similar experimental design may also prove successful in HSE experiments at other pressures, providing the required phase relations, wetting properties and kinetic relationships persist to the chosen conditions. Existing data however, are insufficient to predict whether our sample design will be successful at pressures corresponding to a deep magma ocean. Also outlined is a general approach used to determine moderately and slightly siderophile element (MSE and SSE respectively) partitioning using a multi-anvil device. Extension of the inclusion-free dataset for HSEs to high pressure is likely to employ similar multi-anvil methods. Together, these procedures provide a means to constrain both the conditions of core-segregation and the stages of terrestrial accretion.
1) Preparation of Starting Material
2. Preparation of Assembly Components
3. Assembly of the Components
4. Running the Experiment
5. Run-product Analysis
The following examples and discussion focus on experiments to determine HSE solubility in silicate melts at low fO2. For comprehensive examples of how MSE and SSE partitioning data from multi-anvil experiments may be used to constrain the P–T–fO2 conditions of core metal segregation, the reader is referred to references9–11. Figure 7B-D displays back scattered electron images from typical experimental run-products. In experiments containing Au, the wetting properties between silicate melt, Au melt and solid HSE (Re, Os, Ir, Ru) dictate the sample geometry and result in physical separation between the silicate melt and the solid HSE. For experiments to investigate Pt, the PtIr alloy remains in direct contact with the silicate melt. Cutting power to the furnace at the end of the experiment ensures rapid cooling of the sample and quenching of the silicate melt. Run-products therefore comprise either 1 or 2 alloy phases (HSE-rich ± Au-rich) or silicate glass (providing the basalt composition of Table 1 is used).
Contamination of the silicate glass in low fO2 HSE solubility experiments is most readily identified by the presence of heterogeneity in time resolved LA-ICPMS spectra. This heterogeneity manifests as ‘peaks’ and ‘troughs’ in the spectra that result from the ablation of varying proportions of inclusion-bearing versus inclusion-free glass17. Figure 8A displays the time-resolved spectrum for a Pt solubility experiment that did not employ methods to prevent the formation of metal inclusions. For comparison, Figures 8B-F display time-resolved spectra typical for silicate run-products synthesized using the techniques outlined in the above protocol. The homogeneity of spectra b-f indicates the absence of dispersed HSE inclusions in the silicate portion of experimental run-products. Inspection of the silicate glass by scanning electron microscopy confirms the absence of visible metal-inclusions in the silicate run-products, further supporting a lack of contamination. The spectra displayed in Figure 8A-E are from run-products synthesized as part of several previous studies29–31. Figure 8F is from a Ru solubility experiment performed at 2,273 K and 2 GPa using the Au-addition technique described above. The homogeneity of this spectra suggests that this approach is also successful in avoiding the formation of metal-inclusions found in previous Ru solubility experiments performed at similarly reducing conditions (~IW+2.5)24.
Figure 1. Comparison between the estimated primitive upper mantle (PUM) composition and that predicted by the results of solubility experiments at low pressure and temperature. Data for the PUM composition from Fischer-Gödde et al.7. Partition coefficients for the HSE are at 0.1 MPa, 1573-1673 K and IW-2 from Fe-free experiments by 27(Re), 44(Os), 18(Ir), 45(Ru), 16(Pd), 46(Au), 21(Pt and Rh). Please click here to view a larger version of this figure.
Figure 2. Arrangement used to cold-press metallic powders. The lower drill blank (or shank) is initially taped to the edge of a workbench to allow easy loading of the powders into the silica glass tube. Please click here to view a larger version of this figure.
Figure 3. (A) Detailed cross-section of the piston cylinder assembly once inserted into the pressure vessel. For consistent results, the clearance between components within the resistance furnace should be within 0.025 mm of the nominal values38. BaCO3 cells should be within ~0.13 mm of the nominal inner and outer diameters. Details for the construction of a suitable die may be found in 47, although the cell inner diameter should be modified from the drawings in this reference to 7.9 mm. (B) Procedure for constructing the piston cylinder sample assembly. Either a cyanoacrylate glue or household cement are suitable to secure the graphite end plug in the BaCO3 sleeve, however, no more than ~10 mg should be applied. (C) A piston cylinder press at the University of Toronto. Please click here to view a larger version of this figure.
Figure 4. (A) Cross-section of multi-anvil assembly suitable for use with WC cubes that have an 11 mm TEL. The uppermost part of the figure is drawn to show how the thermocouple arms exit the octahedron, as viewed both perpendicular-to and down the axis of the wire as indicated. (B) Top view of cast octahedron with gasket fins. Grooves for the thermocouple arms should be cut into the areas marked in red. Note that the inner magnesia sleeve and 4-hole alumina tube shown in the figure should not be in place when the grooves are cut. Please click here to view a larger version of this figure.
Figure 5. (A) Arrangement of the WC cubes around the assembled octahedron. (B) Lower set of 1st-stage anvils and their arrangement within the retaining ring. (C) Completed experiment placed into the pressure module with 1 of the upper set of 1st-stage anvils in place. Please click here to view a larger version of this figure.
Figure 6. Dimensions of the mylar sheet described in step 3.2.10. Please click here to view a larger version of this figure.
Figure 7. (A) Experimental run-product mounted in epoxy, then ground and polished. (B) and (C) Back-scattered electron images of experimental run-products from experiments using the Si-addition (B) and Au-addition (C) techniques described in the text for experiments to determine Pt and Ru solubility respectively. Image (B) is reprinted from 31 with permission from Elsevier. (D) Enlarged view of the area outlined in red on (C) to show the detail of the AuRu bead and metal-silicate interface. Please click here to view a larger version of this figure.
Figure 8. (A) Time-resolved LA-ICP-MS spectra from a low fO2 Pt-solubility experiment that did not employ measures to suppress the formation of metal inclusions. (B-F) Typical time-resolved LA-ICP-MS spectra from experiments for Ru, Pt31, Re30, Os and Ir29 that were performed using the procedure outlined in the text. All data shown are from experiments performed at 2,273 K and 2 GPa. The vertical dashed line in each figure separates the region of ablation from the region of background acquisition. Please click here to view a larger version of this figure.
Figure 9. The change in DMet/Sil with T for experiments performed by Brenan & McDonough29 (Os, Ir, Au), Bennett & Brenan30 (Re, Au) and Bennett et al.31 (Pt) using the procedures described here. All data are from experiments done at 2 GPa. Please click here to view a larger version of this figure.
Figure 10. The solubility of Iridium in basaltic melt at 2,273 K and 2 GPa as a function of fO2 relative to the iron-wüstite (IW) buffer. Data are from Brenan & McDonough29. Please click here to view a larger version of this figure.
Before Decarbonation | After Decarbonation | ||||
wt% Oxides/Carbonates; Fe2+ Starting Composition | wt% Oxides/Carbonates; Fe2+ Starting Composition | wt% Oxide; Fe2+ Starting Composition | wt % Oxide; Fe3+ Starting Composition | ||
SiO2 | 47.92 | 47.40 | SiO2 | 51.87 | 51.26 |
Al2O3 | 9.91 | 9.80 | Al2O3 | 10.73 | 10.60 |
CaCO3 | 16.20 | 16.02 | CaO | 9.83 | 9.71 |
MgO | 14.58 | 14.42 | MgO | 15.79 | 15.60 |
FeO | 9.84 | – | FeO | 10.66 | – |
Fe2O3 | – | 10.82 | Fe2O3 | – | 11.71 |
MnO | 0.06 | 0.06 | MnO | 0.07 | 0.07 |
Na2CO3 | 1.20 | 1.19 | Na2O | 0.76 | 0.75 |
NiO | 0.28 | 0.27 | NiO | 0.30 | 0.30 |
Table 1.
The results of inclusion-free experiments performed using the protocols outlined here have previously been compared with literature data in references29 (Os, Ir, Au), 30 (Re, Au) and 31 (Pt). Pt is most instructive in demonstrating the usefulness of inclusion-free run-products. For experiments run at low fO2, Ertel et al.48 assigned inclusions to a stable origin and therefore restricted data reduction to the lowest counts-per-second region of time-resolved LA-ICPMS spectra. This approach minimizes the contribution of inclusions to the measured silicate melt concentrations. Data from Ertel et al.48 at ~IW+1 agree well with the trend between DMet/Sil and 1/T defined by inclusion-free experiments performed at similar fO2; confirming that their chosen analytical treatment is effective in determining true Pt solubilities31. Furthermore, experiments done using the inclusion-suppressing protocol outlined here are able to probe more reducing conditions, in which the spectrum filtering method becomes less effective17. In studies that assume only the presence of quench related inclusions, there is variable agreement with the inclusion-free data. For example, good agreement is observed with the results of Mann et al.49, however, the experiments of Cottrell et al.22 display systematically lower values of DMet/Sil than inclusion-free experiments31. The generation of inclusion-free experiments at a wide range of conditions is thus crucial to assess the reliability of previous inclusion-contaminated measurements.
Although the protocol described here has proven successful over a range of conditions, it is not a panacea for the problem of contamination by metal inclusions. Experiments performed using the technique of Au addition are affected by the formation of complex alloy compositions at very low fO2. In order to generate conditions significantly more reducing than the iron-wüstite buffer, elemental Si is added to the starting materials. Run products from the most reducing of these experiments contain 2 co-existing alloys which possess extensive quench related exsolution textures. Immiscibility in the alloy appears to arise due to the significant solution of Si into the metal phase at reducing conditions. A lack of suitable activity-composition data for the alloy compositions formed at very reducing conditions prevents the Fe and HSE activities in the alloy phase from being determined. This prevents accurate calculation of sample fO2 and HSE concentrations in the silicate melt at the solubility limit.
The efficacy of Si-addition as a method to prevent formation of Pt inclusions appears to decrease at lower temperatures. Bennett et al.31 noted that experiments performed at 1,873 K display evidence for contamination by metal inclusions, whereas those done at higher temperatures do not. This may be due to a change in the kinetic relationship between sample reduction and Pt in-diffusion at lower temperatures. A further consideration for experiments performed using the Si-addition technique is it’s effect on the final melt composition. Oxidation of elemental Si early in the experiment occurs via the following reaction with FeO in the melt:
Si(met) + 2FeO(sil) = SiO2(sil) + 2Fe(met) (2)
Addition of greater quantities of Si, in order to access more reducing conditions, therefore results in a more SiO2 enriched and FeO depleted melt composition. In order to conduct experiments that span a wide fO2 range, but have little variation in melt composition, suitable adjustment of the silicate starting material should be made in order to compensate for reaction 6. Broadly, this can be accomplished by adding 2 moles of FeO and removing 1 mole of SiO2 for each mole of Si added to the starting composition. It should also be noted that some studies have found the occurrence of metallic inclusions also depends upon melt composition, in particular whether the melt is Fe-bearing50.
The solubility data afforded by uncontaminated run-products allows accurate metal-silicate partition coefficients to be calculated. For experiments in which the HSE of interest is not present as a pure phase (i.e., HSE metal activity <1), concentrations measured in the silicate phase are corrected to unit activity using available thermodynamic data. With the exception of Au, a useful summary of the available activity-composition relations for HSE alloys is provided in reference49. Limited data for Au-Fe alloys may be found in references46,51. The following relation is then used to determine values of DMet/Sil from the corrected HSE concentrations in silicate melt16
(3)
Where is the HSE concentration of the silicate at saturation in the HSE phase, A is a mole to weight conversion factor and is the activity coefficient of the chosen HSE at infinite dilution in liquid Fe-metal. Figure 9 displays the change in DMet/Sil with T calculated from experiments done at 2 GPa and an fO2 close to the iron-wüstite buffer. One application of this data is to assess the ability of high temperature metal-silicate equilibrium to account for the estimated primitive upper mantle abundance of these elements. Values of DMet/Sil for the HSE must decrease to values of ~102–103 at approximately the same temperature if metal-silicate equilibrium is responsible for the PUM composition. The data from previous studies displayed in Figure 9 suggests this requirement is not met by the results of experiments performed at reducing conditions29–31.
Solubility measurements from uncontaminated silicate run-products are also useful in revealing the speciation of HSEs dissolved in silicate melt at low fO2. Information regarding the speciation of these metals over a large range of redox conditions is useful not only for guiding extrapolation of the data beyond the experimentally studied range, but also for the informed design of glasses with particular optical properties. Changes in the oxidation state of dissolved metals may be accompanied by changes in their co-ordination chemistry, from which variation in properties such as optical absorbance may arise. For example the dissolution of platinum, widely used as a container material for synthesizing glass from melt, may lead to different colored glasses depending upon the redox conditions of the melt52,53. The oxidation state of dissolved HSEs can be inferred from the change in solubility with fO2. Consider dissolution of a metal (M) as an oxide species in the melt:
(4)
Where n is the oxidation state of the dissolved metal. The equilibrium constant (K) at P and T for equation 3 is given by:
(5)
Equating lnK with the Gibbs free energy of reaction, and at saturation in the metal phase (aM = 1), equation 4 becomes:
(6)
The slope of a trend between HSE solubility and fO2 therefore yields n/4, from which the speciation may be obtained. Brenan & McDonough29 determined the solubility of Ir as a function of fO2 from experiments which used the techniques described here. The results of these experiments are displayed in Figure 10 and yield a slope of 0.2, largely consistent with a 1+ oxidation state (predicted slope of 0.25) for Ir in reduced silicate melts.
Experimentally determined metal-silicate partition coefficients can be used to establish the conditions of core-mantle equilibrium during terrestrial accretion. Results for the highly siderophile elements can also be used to assess whether the Earth experienced a late-veneer of chondritic material subsequent to core formation. Outlined here are procedures to perform metal-silicate partitioning and solubility experiments in the multi-anvil and piston-cylinder devices respectively. Techniques are also described that suppress the formation of metal inclusions in HSE solubility experiments at 2 GPa and temperatures >1,873 K. The calculated HSE partition coefficients suggest that metal-silicate equilibrium at high T does not explain the apparent excess of HSEs in primitive upper mantle. Future work remains to confirm if the HSE partitioning behavior indicated by experiments at 2 GPa persists to higher P and T. This will require testing the inclusion-suppressing protocols outlined here in a high P multi-anvil experimental design.
The authors have nothing to disclose.
This work was supported by the Natural Sciences and Engineering Research Council of Canada Equipment, Discovery and Discovery Accelerator Grants awarded to J.M.B. N.R.B acknowledges support from the Carnegie Institution of Washington post-doctoral fellowship program. Stephen Elardo is also thanked for his assistance prior to filming with the piston-cylinder press at the Geophysical Lab.
G10 Epoxy/Fiberglass Sheet | Accurate plastics, Inc. | GEES.020N.3648 | |
Powdered starting materials- -Oxides, metals, carbonates | Alfa Aesar | Specific to desired experiment | |
Castable 2-part MgO ceramic | Aremco | Ceramcast – 584 | |
PTFE Dry Lubricant | Camie-Campbell | 2000 TFE-Coat | |
Graphite resistance heaters | Carbone of America (Now owned by Mersen USA) | Custom Order | |
Barium Carbonate | Chemical Products Corporation | Custom Order | Calcined free-flowing (CFF) grade |
C-Type Thermocouple Wire (W26%Re, W5%Re) | Concept Alloys | N/A | ~0.25 mm diameter is suitable for most experiments |
Zirconia Cement | Cotronics; Resbond 940 2-part cement | N/A | Use 100 parts powder for every 25 to 28 parts activator |
Polyvinyl Acetate (PVA) Glue | e.g Bostik | N/A | Often sold as 'white glue' |
Cyanoacrylate Glue | e.g Krazy Glue/Loctite | N/A | |
Piston cylinder pressure vessel and WC piston | Hi-Quality Carbide Tooling Inc. | Custom Order | |
Silica Glass Tubing | Quartz Plus | Custom Order | |
Crushable ZrO2 tubes | Saint-Gobain | Custom Order | |
Crushable MgO rods and tubes | Saint-Gobain | Custom Order | |
WC cubes for multi-anvil experiments | Tungaloy | Custom Order | Cubes are grade-F WC alloy |
Single hole alumina tube for multi-anvil thermocouple | Vesuvius McDanel | AXS071730-04-06 | |
4-hole alumina tube for piston cylinder thermocouple | Vesuvius McDanel | AXF1159–07-12 | |
4-hole alumina tube for multi-anvil thermocouple | Vesuvius McDanel | AXF1159-04-06 |