Sulfur and phosphorus in the Earth's core: The Fe-P-S system at 23 GPa

Abstract

[1] Phosphorus is an important element believed to be present in the Earth's core. As a non-metal, its solubility in iron is limited, leading to the precipitation of phosphides at high concentrations. Here we present experimental results for the iron-rich portions of the Fe-P and Fe-P-S systems at 23 GPa. We find a Fe-P eutectic point occurring at 9 wt% P and 1275°C at 23 GPa. Up to 4 wt% P is soluble in the Fe-rich solid coexisting with the eutectic liquid at these conditions. The ternary Fe-P-S system displays a simple cotectic between the Fe-P and Fe-S binary eutectics with minimum melting temperature on the S-side at 1075°C. At subsolidus conditions there is complete solid solution between Fe₃P and Fe₃S. The high solubility of phosphorus in iron along with its miscibility in Fe₃S and iron-sulfur liquids at high pressures, combined with expectation of only minor concentrations in planetary cores, indicate phosphorus will not form a distinct phase within terrestrial planet cores.

1. Introduction

[2] The Earth's core is dominated by iron and nickel along with one or several light element components [McDonough, 2004]. In addition to the major elements, cosmochemical abundance ratios require minor concentrations of cobalt, manganese, chromium and phosphorous along with trace concentrations of numerous other metals to be present in the core [Allègre et al., 1995; McDonough, 2004]. The metallic components readily alloy with iron and, at their low expected concentrations, should have little effect on core phase relations. However, the non-metal phosphorus may significantly change melting temperatures and the compositions of coexisting liquids and solids. Phosphorus is a light element which is unambiguously present in the core due to its prominent deviation below the volatility trend when comparing natural chondritic abundances with the concentrations estimated for the silicate Earth [McDonough, 2004]. McDonough and Sun [1995] estimated core phosphorus concentrations of 0.50 wt% based on Ni/P ratios of 10 ± 1 as found in CI chondrites, while Allègre et al. [1995] estimated 0.369 wt% on the basis of Fe/Ca and P/Ca ratio variation within a suite of various chondrite classes. More recent examination of CI chondrite data [McDonough, 2004] resulted in the somewhat lower estimate of 0.20 wt% P. While phosphorus is present at <1 wt% levels in the core, another light element, or combination of light elements, must also be present at concentrations of several wt% to account for the observed density deficit [Anderson and Isaak, 2002; Birch, 1952]. An element accounting for at least part of the additional light component is sulfur, which is estimated to be present in the core at concentrations between 2 and 8 wt% [Anderson and Isaak, 2002; Li and Fei, 2004; Sherman, 1997].

[3] The common phosphorous bearing mineral found in meteorites is schreibersite, a solid solution between Fe₃P and Ni₃P end members [Skála and Císařová, 2005]. The crystal structure of schreibersite has been determined as I [Skála and Císařová, 2005] and, at least the Fe₃P end member, is stable to a minimum 54 GPa [Santillan et al., 2004]. Fe₃S is stable above 21 GPa [Fei et al., 2000] to pressures of at least 80 GPa [Seagle et al., 2006], but because of the minimum pressure necessary, Fe₃S has not been observed naturally. Structural determinations by Fei et al. [2000] indicate that Fe₃S belongs to the same tetragonal space group I as schreibersite. The striking similarities between the atmospheric Fe-P phase diagram [Zaitsev et al., 1995] and the high pressure Fe-S phase diagram (Figure 1) in conjunction with the isostructural relationship between Fe₃P and Fe₃S lead Fei et al. [2000] to suggest phosphorus may be easily incorporated into Fe₃S. These similarities also suggest that high pressure phase relationships in the iron-rich portion of the ternary Fe-P-S system may be relatively simple. In this study we have determined phase relations in the iron-rich portion of the Fe-P binary and Fe-P-S ternary systems, both at 23 GPa, with the aim of further constraining the nature of metallic cores in the Earth and other rocky planets and planetary bodies.

Figure 1

Simplified Fe-P phase diagram at 1 atm [Baker, 1992; Zaitsev et al., 1995] compared to the 21 GPa phase diagram of Fei et al. [2000]. Both figures have been redrawn for easier comparison.

2. Methods

[4] High-pressure experiments were performed at ETH Zurich in a 1000 ton Walker-type 6/8 multianvil press [Walker, 1991] with the 10/3.5 mm assembly described by Stewart et al. [2006]. Starting materials were prepared from mechanical mixtures of iron, sulfur and iron phosphide powders. The mixtures were ground in ethanol using an agate mortar and pestle for 30 minutes to homogenize the material and ensure fine grain sizes. Five starting materials were prepared, two for the binary phase diagram as Fe₉₅P₅ and Fe₈₈P₁₂ (in wt %) and three for the ternary experiments: Fe₉₃P₂S₅ and Fe₉₃P₅S₂ and Fe₈₂P₉S₉. These compositions were chosen to fall within the expected iron-rich and iron-poor liquidus loops in the respective systems at experimental temperatures. It is also noted here that at lower pressures (<8 GPa) solid iron undergoes two phase transitions with increased pressure, α-Fe to γ-Fe then to δ-Fe. At the conditions of our experiments only one phase transition occurs, ɛ-Fe to γ-Fe, at about 860°C [Shen et al., 1998]. At this relatively low temperature, the phase transition is expected to have no impact on the phase relations examined in this study.

[5] During the experiments, samples were contained in MgO capsules. The axial thermocouple positioning in the 10/3.5 mm assembly allows for two samples to be exposed to pressure-temperature conditions simultaneously, thereby reducing the number of necessary experimental runs [see Stewart et al., 2006]. All runs were performed at 23 GPa, with temperatures of 1000–1550°C. Samples were pressurized first over 2.5 hours, followed by isobaric heating to the experimental temperature. Conditions were held constant for up to 40 hours. Experiments were then quenched by switching off furnace power, with sample temperatures reducing to below 500°C in less than 1 second as recorded by the WRe₅–WRe₂₆ (Type C) thermocouple. The multianvil module was then decompressed and the entire octahedra recovered and mounted in epoxy. With careful polishing through the assembly, both samples were exposed simultaneously for analysis. Samples were carbon-coated and analyzed by a JOEL JXA-8200 electron microprobe at ETH Zurich with operating conditions of 20 kV and 20 nA. Iron, sulfur, and phosphorous were analyzed by wavelength dispersive spectrometry (WDS) employing analytical standards of pure iron, pyrite and apatite, respectively. Magnesium and oxygen were also directly analyzed as potential contaminants, using periclase and hematite as standards, though no significant contamination of the charges was observed. Peak and background counting times were 20 and 10 seconds respectively.

[6] Phases in quenched samples were examined by backscattered electron imaging (Figure 2). Quenched liquids were identified texturally in all our systems by typical dendritic growth crystals as described by Li et al. [2001] in the Fe-S system. Our solid iron phase commonly displays the non-uniform texture they reported and attributed to an electron channeling effect in polycrystalline samples with variable grain orientations. We assume the same process is occurring in our charges. To ensure representative compositions, 10–40 μm defocused beam sizes were used on the quenched liquids to average out the effect of quench-induced growth crystals, a focused beam was used for other phases. In some runs, a third sulfide or phosphide phase in addition to the iron-rich crystalline and liquid phases was present in the colder region of the capsule, away from the thermocouple sitting in the hot spot of the assembly. We interpret these solids as a product of the temperature gradient in our samples resulting in the cold-regions crystallizing subsolidus phases. The presence of such phases should not affect the liquid-solid equilibrium and have not been considered in this study.

Figure 2

Backscattered electron image of experimental charge 125b, into the Fe-P-S system at 1300°C and 23 GPa. The quenched melt phase and solid Fe residua are indicated. The inset shows a close up of a quench textured region of the sample. Dark straight lines in the images are scratches in the polish of the sample.

3. Results

3.1. Binary Systems

[7] A summary of our results into the iron-rich portion of the Fe-P binary is presented in Table 1. Up to 4 wt% phosphorus was found to dissolve in crystalline iron at 23 GPa. This compares to 2.6 wt% of phosphorus dissolving at atmospheric pressures [Zaitsev et al., 1995]. The eutectic liquid composition and temperature change with increasing pressure, from 10.5 wt% P and 1048°C at atmospheric pressures [Zaitsev et al., 1995] to 9 wt% P and 1275°C at 23 GPa (Figure 3a). As a consequence the liquidus loop narrows by simultaneously increasing phosphorous solubility in iron and decreasing phosphorous concentration in the eutectic liquid, this behavior is analogous to the pressure effect on the Fe-S system [Fei et al., 1997; Li et al., 2001]. Below the eutectic, the iron-rich portion of the Fe-P system crystallizes to Fe + Fe_3+xP. The origin of slight excess iron (Table 1) as compared to the ideal Fe₃P structure, which contains 15.6 wt% P, is unknown, but a similar excess is noted for the Fe₃S phase observed in the 21 GPa Fe-S study by Fei et al. [2000]. The minor changes in the Fe-P system with pressure preserve its similarities with the high pressure Fe-S system and suggest a relatively simple high pressure ternary phase diagram.

Figure 3

The Fe-P system at 23 GPa. Errors are smaller than symbols unless indicated. Open symbols are solid phases and closed symbols liquids. The pure iron melting temperature is taken from [Boehler, 1993]. Dotted phase relations are inferred from the low pressure Fe-P phase diagram and are included for completion of the diagram, but not constrained by present experiments.

Table 1. Results of Experiments in the Fe-P System at 23 GPa^aa Values in wt%.

Run	T, °C	Time, hours	P solid	P liquid	P phosphide
116	1200	14	2.30 (0.17)		13.61 (0.07)
124	1250	15	2.86 (0.24)		14.48 (0.10)
118	1300	5	3.17 (0.11)	8.75 (0.47)
137a^b	1300	5		10.84 (0.28)	14.71 (0.06)
117	1400	1	3.95 (0.24)	9.17 (0.97)
123	1500	6	3.88 (0.07)	9.78 (0.75)
136a	1550	5	3.76 (0.34)	7.76 (1.65)
136b^bb Experiments on the P-rich side of the eutectic composition.	1550	5		11.98 (0.37)

• a Values in wt%.
• b Experiments on the P-rich side of the eutectic composition.

3.2. Ternary System

[8] Results in the iron-rich portion of the ternary Fe-P-S system at 23 GPa are presented in Table 2. The ternary diagram and its isothermal sections (Figure 4) were constructed using the binary Fe-P (Table 1) and Fe-S [Fei et al., 2000; Stewart et al., 2007] systems together with the 3-component experiments (Table 2). Our data indicate a simple cotectic between the two binary eutectics reaching through the ternary diagram (Figure 4a), and complete solid solution of the Fe₃P-Fe₃S phase without new ternary phases appearing. Sulfur- and phosphorus-iron liquids appear to be completely miscible. Above 1275°C, the eutectic temperature of the Fe-P binary system, and at low bulk phosphorous and sulfur contents, the liquid coexists with an iron-rich phase (Figure 4b). At higher light element concentrations the liquid coexists with a Fe-phosphosulfide. The mineralogy of this phase depends strongly on temperature as, indicated from the binary relations of both systems, either Fe₃(P,S), Fe₂P, Fe₃S₂ or Fe(P,S) could be liquidus phases. In isothermal section at 1300°C (Figure 4b), the ternary liquid will coexist with either Fe₃(P,S) (at higher phosphorous content) or Fe(P,S) (at higher sulfur content) with the possibility that other intermediate Fe-phosphosulfides may form. However, the P,S-rich side of the eutectic line has little bearing on planetary cores and is thus not resolved with the present dataset. Between 1075-1275°C, the ternary system will crystallize the eutectic assemblage of Fe + Fe₃(S,P) + liquid, with the exact temperature for the onset of the eutectic crystallization determined by bulk composition. At subsolidus conditions, below the Fe-S eutectic temperature of 1075°C, we find complete miscibility between the endmembers Fe₃P and Fe₃S in a Fe₃(P,S) solid solution and the subsolidus phase assemblage as Fe + Fe₃(P,S) (Figure 4c).

Figure 4

The iron-rich portion of the Fe-P-S system. (a) The liquidus surface of the ternary at high iron concentrations demonstrating the cotectic relationship between the Fe-S and Fe-P binary eutectic points, the relationship between the two binary peritectic points is not constrained by the present data. Soludus phases crystallizing at the liquidus are indicated; (b) isothermal section at 1300°C through the phase diagram with binary results and experiments 125a, 125b and 137b confining the relationships; (c) isothermal section at 1000°C with results of experiments 127a and 127b confining the subsolidus phase relations. Open circles in the isothermal sections indicate the binary constraints on the ternary system. The axes of all diagrams are in weight percent.

Table 2. Experimental Results in the Ternary Fe-P-S System at 23 GPa^aa Values in wt%.

Run	T, °C	Time, hours	P solid	S solid	P liquid	S liquid	P Phosphosulfide	S Phosphosulfide
126a^bb No solid residue uncovered for analysis after multiple polishing attempts.	1400	1	n/a	n/a	5.68 (0.09)	4.26 (0.12)
126b	1400	1	1.10 (0.06)	0.42 (0.01)	1.59 (0.07)	12.10 (0.44)
125a	1300	5	2.83 (0.05)	0.19 (0.01)	6.13 (0.16)	4.85 (0.17)
125b	1300	5	1.11 (0.03)	0.41 (0.01)	1.67 (0.06)	11.66 (0.39)
137b^cc Experiment on the (P,S)-rich side of the eutectic composition.	1300	5			5.86 (0.53)	7.22 (0.24)	9.77 (0.43)	5.62 (0.39)
127a	1000	40	1.60 (0.16)	0.08 (0.03)			8.28 (3.84)	6.89 (4.11)
127b	1000	40	0.40 (0.11)	0.25 (0.09)			4.94 (2.47)	10.50 (2.68)

• a Values in wt%.
• b No solid residue uncovered for analysis after multiple polishing attempts.
• c Experiment on the (P,S)-rich side of the eutectic composition.

4. Discussion

[9] Our experiments demonstrate that phosphorus might easily be incorporated into the Fe₃S phase, in accordance with an earlier suggestion of [Fei et al., 2000]. Thus, phosphorus can indeed be incorporated into planetary cores in the Fe₃S structure (and vice-versa). With the possible exception of Mars [Stewart et al., 2007], iron-rich cores of planets and planetary bodies are expected to only begin crystallizing a sulfide phase upon evolving to the eutectic composition from more iron-rich liquids. As Fe₃S would occur only once this eutectic and thus complete solidification is reached, the behavior of phosphorus in these systems is dominated by the solubility of phosphorus in the solid iron phase, not the late-forming sulfide crystallizing once eutectic conditions are reached.

[10] The solubility of phosphorus in the solid iron residua of melting is much higher than that of sulfur, at both 1 atmosphere [Baker, 1992; Zaitsev et al., 1995] and 23 GPa [Fei et al., 2000] (also this study). Examining our data for the partitioning behavior of phosphorus between solid metal and liquid sulfide, we calculate D^solid/liquid for phosphorus from the experiments with phosphorous-poor ternary starting materials (126b and 125b) as 0.69 and 0.66, respectively. These partition values indicate only a modest preference of phosphorus for the liquid, reflecting its high solubility in solid iron. These results are directly applicable to the face centred cubic (FCC) γ-Fe [Shen et al., 1998] polymorph, stable to 60 GPa. However, the Earth's core is expected to be composed of hexagonally closest packed HCP ɛ-Fe, or possibly the so-called double HCP β-Fe [Saxena et al., 1996]. Nevertheless, as the changes in phase relations and partitioning in the low pressure Fe-P system, where iron occurs as δ-Fe (body-centred cubic structure), differ little with the results of the system at 23 GPa, it is quite likely that changes in these properties will also be minor upon the stabilization of another iron polymorph. At 23 GPa, 0.2 wt% P and 1.9 wt% S would be dissolved in solid cores of planetesimals with bulk Earth composition [McDonough, 2004], accounting for part of the inner cores density deficit. Furthermore, assuming a similar value for the phosphorus partitioning coefficient at the inner-outer core boundary of the earth, we estimate that there would be roughly twice the concentration of phosphorus in the outer liquid core than in the inner solid core.

5. Conclusions

[11] High concentrations of phosphorus are able to dissolve into iron at high pressures over a wide temperature range. Combined with low absolute phosphorus concentrations expected in planetary bodies, complete miscibility of melts in the Fe-P-S system and for the Fe₃S-Fe₃P solid-solution indicate that phosphorus will not form a distinct phase within the Earth's core. Due to the modest solid-liquid metal partition coefficient, phosphorus may concentrate only by a factor of two into the liquid over solid iron phase. Assuming continued core crystallization to the onset of sulfide saturation, phosphorus will incorporate most readily into the Fe₃S phase as a solid solution. The effect of phosphorus on the temperatures and compositions of planetary core phases and on eutectic temperatures of iron cores is thus only minor at bulk contents of ≤0.5 wt% P.

Acknowledgments

[12] This work was supported by the Swiss National Science Foundation, grants 2100-066903-01/1 and 200020-103722/1.