Introduction

Apatite is the most abundant naturally occurring phosphate and one of the most important halogen-bearing minerals on Earth. As a widespread accessory or minor phase in virtually all igneous and metamorphic rocks, apatite carries an essential portion of the bulk phosphorus content of 0.1–0.3 wt% P2O5 of continental and oceanic crust (Rudnick and Gao 2003; Klein 2004). Apatite is also one of the major phosphates on the moon and the terrestrial planets (Patiño Douce and Roden 2006 and references therein) and is the only OH-bearing phase currently known from the moon. Most recently, it was shown that lunar apatite may contain up to several thousand ppm H2O (McCubbin et al. 2010, Boyce et al. 2010). This finding indicates H2O contents of the moon’s interior several orders of magnitude higher than previously estimated and challenges currently accepted models for the formation and cooling of the moon. Water contents of up to 0.64 wt% were found in apatite from an SNC meteorite (Leshin 2000), making apatite an important water reservoir in the Martian interior. Because of its strongly incompatible behaviour in igneous processes, phosphorus together with Fe and Ti is concentrated during fractional crystallization, eventually resulting in the formation of apatite ± FeTi-oxide deposits (Kolker 1982). Apatite is not present in primitive terrestrial mantle rocks due to their low bulk phosphorus contents of 90–100 ppm (McDonough and Sun 1995; Workman and Hart 2005; Pearson et al. 2004; Palme and O’Neill 2004; McDonough 1990). However, it is frequently found in metasomatically altered peridotitic mantle and is often associated with phlogopite and/or calcic amphibole (e.g. Exley and Smith 1982; Zanetti et al. 1999; Wass et al. 1980; O’Reilly and Griffin 1988, 2000; Rudnick et al. 1993; Ionov et al. 1996, 2006). In these rocks, apatite carries a major portion of the whole rock LILE, HFSE and REE-budget (e.g. O’Reilly et al. 1991; O’Reilly and Griffin 2000; Exley and Smith 1982). Mantle apatite often contains high concentrations of U and Th. Hence, it was proposed that the presence of apatite may critically influence the heat production in metasomatically altered mantle (O’Reilly and Griffin 2000). The noble gas isotopic composition of apatite was used to place constraints on the source of metasomatizing fluids beneath south-eastern Australia (Matsumoto et al. 1997). The formation of apatite is often ascribed to the interaction of mantle rocks with small volumes of phosphorus-rich carbonatitic melts or fluids (Rudnick et al. 1993; Hauri et al. 1993; Morishita et al. 2003). In rare instances, vestiges of these metasomatic agents are preserved as frozen grain boundary melts (Rosenbaum et al. 1997; Zhang et al. 2007) or as fluid/melt inclusions in diamond typically associated with apatite and K–Cl-LILE-rich phases (Lang and Walmsley 1983; Guthrie et al. 1991; Tomlinson et al. 2005; Titkov et al. 2006; Klein-BenDavid et al. 2006; Wirth et al. 2009). Partial melting of deep-seated apatite ± phlogopite-bearing asthenospheric or lithospheric mantle rocks may lead to the formation of kimberlites and lamproites (Foley 1992). Apatite is the major solid Cl carrier in the mantle because hydrous silicates with extremely rare exceptions do not incorporate significant Cl due to Fe–Cl avoidance (Volvinger et al. 1985; Kullerud 1995; Oberti et al. 1993) and strong partitioning of Cl into fluids. Therefore, the breakdown of apatite contributes to the high percentage of Cl recycled in subduction zones (Straub and Layne 2003). Nevertheless, partitioning of Cl between apatite and hydrous silicates provides a potential mechanism to avoid shallow-level recycling and to transport at least small amounts of Cl to transition zone depths. Although mantle apatite may be rich in F, its overall importance as F carrier in the mantle is negligible. This is because F unlike Cl is strongly compatible, shows a crystal chemical affinity to Mg-rich hydrous silicates and can also substitute into the olivine structure (Bromiley and Kohn 2007). These factors contribute to the low recycling efficiency of F compared to Cl in subduction zones (Straub and Layne 2003). The high-P solid-state breakdown product of apatite is tuite [γ-Ca3(PO4)2], which was found in shocked chondritic and SNC meteorites (Xie et al. 2003; Ozawa et al. 2007; Greshake and Fritz 2009). Experiments by Murayama et al. (1986) have shown that pure OH- and F-apatite break down at 11–13 GPa between 1,000 and 1,500°C with a negative slope of the apatite–tuite phase boundary and a slightly lower (≤1 GPa) upper P stability limit of the OH-endmember. Due to the large spacing of experimental data points, however, no reasonable comparison of the upper T stability limits of OH- and F-apatite is possible. In MORB-type bulk compositions, tuite forms at P ≥ 7.5 GPa by reactions involving garnet and SiO2 (Konzett and Frost 2009). Like apatite, tuite contains large and high-coordinated lattice positions and, hence, is an eminently suitable host for LIL elements (Sugiyama and Tokonami 1987; Konzett and Frost 2009). Whereas apatite is by far the most important subsolidus carrier of phosphorus under crustal PT conditions, silicates—in particular garnet—start to become additional storage sites of phosphorus at P ≥ 3 GPa in a wide range of bulk compositions (e.g. Thompson 1975; Brunet et al. 2006; Konzett et al. 2008; Konzett and Frost 2009). The increasing phosphorus solubility in garnet may lead to a disappearance of apatite from garnet-bearing high-P rocks.

This study was undertaken to contribute to a better understanding of phosphorus and chlorine transport and storage in the upper mantle and transition zone. Its major aims are as follows: (1) to investigate the PT stability of OH- and Cl-apatite and its high-P breakdown product tuite in a representative upper mantle peridotite bulk composition, (2) to explore the phosphorus storage capacity of the peridotite silicate phases—in particular of garnet and the Mg2SiO4 polymorphs and (3) to place constraints on the mechanisms of upper mantle transport and storage of Cl by studying the PT stability of Cl-apatite and the partitioning of Cl between apatite, phlogopite and amphibole.

Experimental and analytical techniques

Four bulk compositions were used in this study, and all are based on a moderately fertile spinel lherzolite from Mont Briançon, French Massif Central, modified by subtraction of 30 wt% olivine (Fo91) and addition of 0.4 wt% Na2O (modBRIAN2; cf. Konzett and Ulmer 1999; Fumagalli et al. 2009). Bulk composition I consists of modBRIAN2 doped with 3 wt% synthetic hydroxyl apatite and a trace element mix containing Nb, Ta, Zr, Y, Ba, Sr, Rb, Ce, Nd and Lu with concentrations between 250 and 350 ppm (cf. Konzett and Frost 2009); bulk composition II consists of modBRIAN2 doped with 5 wt% synthetic Cl-apatite + 5 wt% synthetic phlogopite + trace element mix; bulk composition IIa corresponds to bulk composition II without any trace elements added and bulk composition III consists of modBRIAN2 doped with 5 wt% Cl-apatite + 15 wt% phlogopite + trace element mix (Table 1). Modified BRIAN2 was prepared from high-purity (≥99.9%) and finely ground SiO2, TiO2, Cr2O3, MgO, NiO, CaCO3 and Na2CO3 by mixing in ethanol for 20 min and stepwise decarbonation to 800°C with intermittent checks of the loss on ignition. Al, Fe2+ and Mn2+ were added as γ-Al2O3, Fe2SiO4 and MnO. Phlogopite was synthesized in an externally heated pressure vessel at 0.3 GPa/680°C from an oxide–carbonate mix, and Cl-apatite was prepared following the method of Bauer and Klee (1993) from ammonium dihydrogen phosphate and calcium chloride. This synthesis yielded clear idiomorphic apatite crystals of up to 4 × 0.5 mm in size. Electron microprobe analysis yields P2O5 41.6 ± 0.2; CaO 54.7 ± 0.5; and Cl 6.9 ± 0.1 (n = 15) corresponding to pure Cl-apatite. The experiments on Cl partitioning between apatite, phlogopite and amphibole faced two difficulties: Cl-loss from apatite and strongly enhanced fluid solubility of K in the presence of Cl. Although the starting materials were permanently stored at 150°C, trace amounts of free water are present during the experiments. The source of water is moisture from the air adsorbed to the starting material (Médard et al. 2008) and breakdown of phlogopite due to high potassium solubility in the free fluid. As a result of its extreme fluid compatibility, a large portion of Cl was lost from the Cl-apatite by Cl ⇔ OH exchange with the fluid (Table 2). The very high K-solubility in the Cl-rich fluid led to a complete disappearance of hydrous potassic phases from the experimental assemblages in bulk compositions II and IIa between 7 and 8.7 GPa. In order to avoid any potential effect of trace elements on the availability of K for hydrous potassic silicate stability, the experiment at 7 GPa/1,000°C was repeated with bulk composition IIa. This, however, had no effect on the phase assemblage. It was only after increasing the bulk K2O content of the starting material from 0.5 to 1.4 wt% that K-richterite was stabilized at 7 GPa/800°C.

Table 1 The composition of the starting materials
Full size table
Table 2 Representative and averaged analyses of apatite and tuite, garnet, phlogopite, Ca-amphibole and K-richterite, Mg2SiO4 polymorphs, pyroxenes, quenched melt
Full size table

Experiments (Table 3) were performed with 500t and 1000t multi-anvil presses at the Institute of Mineralogy and Petrology, University of Innsbruck (UI), and the Bavarian Research Institute of Experimental Geochemistry and Geophysics, University of Bayreuth (BGI), respectively, under conditions outlined in detail in Konzett and Frost (2009).

Table 3 Summary of experimental run conditions and products
Full size table

The composition of phases was analysed with a JEOL 8600 superprobe, and the identity of Mg2SiO4 polymorphs was verified using laser Raman spectroscopy. Variable beam currents and measurement times of the X-ray lines were used dependent on the phases analysed and the concentration of the elements measured. Quenched melt, apatite, phlogopite and calcic amphibole were analysed with 5 nA beam current in order to minimize Cl- and alkali diffusion and/or because most grains were too small to be analysed with a beam in raster mode. Resulting 2σ uncertainties for Cl in an individual analysis are 4–7% for apatite, 8–12% for phlogopite and 12% for calcic amphibole. The low X-ray count rates obtained with 5 nA also yielded comparatively high 2σ uncertainties for Sr, Ce and Nd of ~30, 25 and 20%, respectively (cf. Table 2). Phosphorus in the Mg2SiO4 polymorphs was analysed with 50 nA beam current and 50 s/25 s counting time on peak and background of the P-Kα line. Care was taken to measure only grains ≥10 μm in diameter and to avoid grains that show boundaries with garnet or Ca-phosphate. To assess the effect of secondary fluorescence from phosphorus on adjacent phases, compositional profiles were analysed across olivine grains with 15–20 μm diameter adjacent to Ca-phosphate grains. These profiles revealed a heterogeneous distribution of phosphorus concentrations and a sudden strong increase at distances of ≤3–5 μm from the olivine–Ca-phosphate grain boundary. This is in agreement with results obtained by Brunet and Chazot (2001). Phosphorus in pyroxenes was analysed using the analytical protocol described in Konzett and Frost (2009). At P > 9 GPa, the Mg2SiO4 polymorphs are too small to be analysed with a conventional electron gun. Instead, the JEOL electron microprobe JXA-8500F (Hyperprobe) equipped with a field emission gun at the GFZ German Research Centre for Geosciences Potsdam was used. Trace element measurements of phosphorus were taken with PET crystals using lower acceleration voltage and a liquid nitrogen trap to reduce the surface contamination. The analytical conditions included an acceleration voltage of 8 kV, a beam current of 30 nA and a focussed beam. The lower analytical sensitivity with decreasing acceleration voltage was compensated using three simultaneously working spectrometers. With this procedure, total counting times of 900 s (3*300 s) on the peak resulted in a reasonable detection limit of 10 ppm P. The background counting times were always set to half of the peak counting times. The CITZAF routine in the JEOL software, which is based on the Φ(ρZ) method (Armstrong 1995), was used for data processing. Multiple measurements of different reference glasses (Brunet and Chazot 2001) were taken in order to check the reproducibility of the measurements.

Results

Textures and chemical homogeneity of the run products

All starting materials recrystallized to form euhedral to subhedral mineral grains with sizes between ~5 and ~20 μm in experiments at T ≤ 1,000°C in part showing well-equilibrated textures with 120° triple grain boundaries (Fig. 1g). At higher temperatures, the grain size of garnet and orthopyroxene may increase to 50–100 μm. These larger grains often show numerous inclusions of other matrix phases, limiting the number of grains per experiment suitable for analysis. Apatite and tuite are present as lath-shaped to xenomorphic grains rarely exceeding 20 μm in diameter and also often containing inclusions (Figs. 1a, b). In many instances, the high density of inclusions and/or the small grain size made a reliable analysis difficult or even impossible. Highly variable SiO2 and MgO concentrations were taken as indication of beam overlap with adjacent or included Mg- silicate phases, and these analyses were excluded from further considerations. Quenched melt forms layers of ≤20–30 μm thickness infiltrating the graphite liner at the end of the capsule close to the thermocouple (Fig. 1d). The melt is very homogeneous when analysed with an electron beam in raster mode (Table 2). Individual mineral grains occasionally show compositional zoning especially in runs at T ≤ 1,000°C: garnet may contain diffuse cores with variable Al/Cr ratios probably due to the sluggish diffusion of Al and Cr in the garnet structure and orthopyroxenes may show patchy Fe–Mg zoning with Fe-poor rims probably due to Fe-loss in the presence of melt (Fig. 1d). Cloudy phosphorus zoning in olivine may also be observed (Fig. 1h). In the Cl-bearing bulk compositions, severe Fe-loss with variable Fe contents and Fe–Mg zoning may even occur under subsolidus conditions (Fig. 1a) in experiments of long times (e.g. MA52 and MA55).

Fig. 1
figure 1

Backscattered electron photomicrographs of experimental run products: a run MA52 at 7 GPa/1,000°C; b run B07-1 at 9 GPa/1,100°C; c run MA46 at 8.7 GPa/1,000°C; d run B06-11 at 6 GPa/1,500°C; e run B07-2 at 12.8 GPa/1,200°C; f run B08-5 at 19 GPa/1,200°C; g run MA31 at 4 GPa/1,000°C; h distribution of phosphorus in run MA31; mapped area corresponds to area shown in g; for better comparison between g and h, olivine, clinopyroxene and orthopyroxene are marked with stippled lines; for abbreviations, see Table 3

Full size image

Phase relations

Bulk composition I crystallizes Mg2SiO4 + orthopyroxene + clinopyroxene + garnet ± apatite ± tuite in the PT range 6–15 GPa and 1,050–1,300°C. Apatite is stable in runs between 6 and 8 GPa to 1,100°C, and tuite was found between 8 and 15 GPa to 1,300°C. At 8 GPa/1,050°C, apatite and tuite coexist. Olivine is replaced by coexisting olivine + wadsleyite at 12.8 GPa (Fig. 1e). At 15 and 19 GPa, wadsleyite and coexisting wadsleyite and ringwoodite, respectively, are the stable Mg2SiO4 polymorphs (Fig. 1f). Quenched melt was found in two runs at 6 GPa/1,300°C coexisting with olivine + orthopyroxene + clinopyroxene + garnet and at 6 GPa/1,500°C coexisting with olivine + orthopyroxene + garnet (Fig. 2a). In bulk composition II, apatite is stable to 8.7 GPa at 1,000°C and coexists with olivine + orthopyroxene + clinopyroxene + garnet. Additional phases are calcic amphibole + phlogopite at 2.5 GPa and phlogopite at 4–5 GPa. At 8.7 GPa, apatite and tuite coexist (Fig. 1c). No potassic silicate phase was found between 7.0 and 8.7 GPa. Instead, tiny (≤1 μm) grains of K (Nb, Ta)O3 are present, and X-ray mapping revealed diffuse potassium along grain boundaries of graphite close to the interface between experimental charge and graphite liner. The addition of extra 10% phlogopite to bulk composition II (bulk IIa) led to the formation of K-richterite coexisting with olivine + orthopyroxene + clinopyroxene + garnet + apatite at 7 GPa/800°C.

Fig. 2
figure 2

a PT diagram summarizing experimental results obtained for bulk composition I (round pie-symbols), II and III (square pie-symbols); ACMA average current mantle adiabat for a surface potential temperature of 1,280°C; CBT-solidus (F09) solidus for CO2–H2O-bearing K-enriched pyrolite according to Foley et al. (2009); b schematic arrangement of possible tuite-forming reactions based on a negative slope of diopside- and enstatite-absent reactions (see text)

Full size image

Phase compositions

Apatite and Tuite

Apatite and tuite always contain variable amounts of MgO, FeO, SrO, Na2O and RE2O3 often reaching a wt% level (Table 2), with Na and REE showing a distinct positive correlation (Fig. 3a). Whereas MgO shows a strong increase with pressure from 0.7 wt% at 2.5 GPa to 3.1 wt% at 8.7 GPa, FeO contents only slightly increase from 0.4 to 1.1 wt% (Fig. 3b). The highest FeO + MgO values correspond to 9 mol% (Mg, Fe)5(PO4)3 (OH). It is only in bulk composition III at 7 GPa/800°C that apatite shows unusually low concentrations of Mg and other minor elements. A possible reason is insufficient equilibration of these elements at the lowest temperatures attained in this study. None of the other minor elements shows a systematic correlation between concentration and pressure or temperature. The Cl content of apatite is between 0.4 and 2.4 wt% and thus significantly lower than the stoichiometric Cl content of 6.8 wt% of pure Cl-apatite. Possible reasons for this Cl-loss are discussed below. Tuite always shows significantly higher Sr and REE and lower Fe and Mg contents than apatite. This is true for experiments in which tuite is the only phosphate phase and also when apatite and tuite coexist (Fig. 3c; Table 2).

Fig. 3
figure 3

a Na and REE content of apatite (open symbols) and tuite (black symbols); numbers next to plot symbols are pressure [GPa] and temperature [°C]; b averaged FeO and MgO contents of apatite from bulk compositions II and IIa; c minor element composition of apatite (black symbols) and coexisting apatite and tuite (open symbols) at 7–8.7 GPa/1,000°C; d relation between MgO in apatite and in the bulk; II and IIa: bulk compositions II and IIa; black circle and diamond are data from Konzett and Frost (2009)

Full size image

Hydrous silicates

Phlogopite shows little compositional variation between 2.5 and 5.0 GPa with Si = 2.89–2.97 apfu, K = 0.72–0.85 apfu and K/(K + Na) = 0.85–0.97. All phlogopites contain Cl in the range 0.23 ± 0.05 to 0.41 ± 0.05 wt% corresponding to XCl-phl = 0.011–0.024. At 2.5 GPa/900°C, pargasitic calcic amphibole with Na (M4) = 0.12 apfu, Na (A) = 0.70 apfu, K/(K + Na) = 0.20 and Cl = 0.19 ± 0.05 wt% (XCl-amph = 0.023) coexists with phlogopite. The major element composition of both phlogopite and calcic amphibole is in good agreement with experimental data by Fumagalli et al. (2009). In bulk composition III, K-richterite is the stable hydrous silicate at 7 GPa/800°C. Its composition is close to endmember KNaCa (Mg, Fe)5Si8O22 (OH)2 with <0.1 wt% TiO2 and Cr2O3 and an average Cl of 0.56 ± 0.07 wt% (n = 8) (Table 2).

Garnet

Garnets are pyrope-rich pyrope-almandine-grossular solid solutions typical for peridotites with minor amounts of Cr2O3, P2O5, Na2O and TiO2 (Table 2). P2O5 and Na2O increase with increasing pressure from just above the detection limit at 2.5 GPa to 1.5–1.6 and 0.3–0.5 wt%, respectively, at P > 12.8 GPa (Figs. 4, 5). TiO2 is present in concentrations of 0.1–0.4 wt% and shows little variation with pressure but an increase with temperature. A significant decrease in P, Na, Ti and XMg along with an increase in Cr can be observed at 6 GPa between 1,100 and 1,300°C and also at 8 GPa between 1,100 and 1,200°C accompanied by the disappearance of a phosphate phase.

Fig. 4
figure 4

Averaged phosphorus contents in garnet as a function of P and T; at 15 GPa, only results from run B06-16 are shown; black stars are data from garnet from a MORB-bulk shown for comparison (Konzett and Frost 2009)

Full size image
Fig. 5
figure 5

a, b Averaged Na2O and TiO2 contents of garnets as a function of pressure and temperature; c, d P–Cr–Si systematics in garnets from bulk composition I; data scatter especially of the 6 GPa/1,100°C data is thought to be due to incomplete equilibration as a result of sluggish Cr diffusion

Full size image

Mg2SiO4 polymorphs

Under subsolidus conditions, olivine usually shows XMg values in the range 0.91–0.94. In the presence of melt (e.g. B06−11) or when long run durations were used (MA52), severe Fe-loss occurs in spite of the use of a graphite liner, resulting in XMg values of 0.96–1.00 (Table 2). Phosphorus contents in olivine range between 0.14 ± 0.03 and 0.23 ± 0.06 wt% P2O5 and do not show any significant correlation with pressure, temperature, XMg or grain size (Table 2). Run MA52 with the largest olivine single crystals or aggregates of up to 30 × 30 μm in diameter (Fig. 1a) yields 0.180 ± 0.040 (n = 15) wt% P2O5. Mapping of P2O5 concentrations of individual olivine grains in runs MA52 reveals variations in the range 0.10–0.22 wt%. In run B07-2, the olivine–wadsleyite loop is intersected. Based on Katsura and Ito (1989), XMg of the coexisting polymorphs at 1,200°C indicates a pressure of 12.8 GPa. Compared to P ≤ 9 GPa, phosphorus in both olivine and wadsleyite buffered by tuite is significantly lower with 143 ± 58 (n = 22) ppm and 125 ± 31 (n = 21) ppm, respectively, showing an equal preference for the olivine and wadsleyite structures. Similarly, low concentrations of 181 ± 77 (n = 37) ppm are found in wadsleyite at 15 GPa/1,500°C in the absence of a phosphate phase. In run B08-5 conducted at 19 GPa/1,200°C, coexisting wadsleyite and ringwoodite with XMg of 0.935 and 0.897, respectively, are present. This is consistent with a pressure of ~17.5 GPa based on the wadsleyite–olivine loop deduced by Katsura and Ito (1989). Unlike coexisting olivine and wadsleyite, coexisting wadsleyite and ringwoodite show significantly different phosphorus concentrations of 70 ± 16 (n = 11) ppm and 181 ± 65 (n = 25) ppm, respectively (Fig. 6), indicating a preference of phosphorus for the ringwoodite structure.

Fig. 6
figure 6

Range of phosphorus concentrations in Mg2SiO4 polymorphs; a olivine in run MA52 at 7 GPa/1,000°C; b coexisting olivine (a) and wadsleyite (b) in run B07-2 at 12.8 GPa/1,200°C; c wadsleyite in run B06-16 at 15 GPa/1,500°C; d coexisting wadsleyite and ringwoodite (g) in run B08-5 at 19 GPa/1,200°C

Full size image

Pyroxenes

Clinopyroxene are diopside-rich solid solutions with small amounts of jadeite, enstatite and Ca-Tschermaks pyroxene component. With the exception of a slight decrease in jadeite component with increasing pressure and temperature, there is little variation in the clinopyroxene composition. Averaged P2O5 contents are in the range of 352 ± 52 to 522 ± 86 ppm (Table 2). Orthopyroxene shows a decrease in Al with pressure constrained by coexisting garnet and an increase in Ca and Na with temperature, the former constrained by coexisting clinopyroxene. In the presence of melt, the Fe contents drop significantly and show a much stronger scatter compared to subsolidus conditions. The P2O5 contents of orthopyroxene range between 56 ± 24 ppm and 212 ± 44 ppm (Table 2).

Melt

Melts found in runs B06−11 and B06−13 in bulk composition I are carbonatitic due to oxidation of C from the graphite liner with SiO2, CaO and MgO contents of 3.6–4.6 wt%, 23.5–25.2 wt% and 20.6–22.3 wt%, respectively, corresponding to Ca/(Ca + Mg) = 0.45–0.46. Both melts are strongly enriched in phosphorus with 7.6 and 8.1 wt% P2O5, respectively. In addition and not surprisingly, high concentrations of Sr, Ce and Nd were measured in the melt from run B06−11 (Table 2).

Discussion

The stability of apatite in peridotitic compared to MORB-type lithologies

This study shows that in a peridotite apatite is stable to at least 8.7 GPa at 1,000°C, which corresponds to a depth of ~260 km. This is ~1 GPa above the upper pressure stability limit of apatite in a MORB-type bulk composition (Konzett and Frost 2009). In addition, apatite and tuite coexist at 8.0 and 8.7 GPa in the peridotite bulk composition, whereas no indication for a significant pressure–temperature interval of coexisting apatite and tuite was found in a MORB-type bulk (cf. Konzett and Frost 2009). A possible explanation for the extended pressure stability of apatite in the peridotite is the stabilizing effect of MgO and FeO. Because of the high variance of the peridotite + Ca-phosphate assemblage, mineral and bulk compositions are interdependent, leading to much higher MgO in apatite from peridotites compared to that from MORB-type bulks (Fig. 3d). In a simplified system SiO2 + MgO + CaO + Al2O3 + P2O5 + H2O, in which the lherzolite + phosphate assemblage is represented by the phase components forsterite, enstatite, diopside, pyrope, grossular, apatite, tuite and H2O-fluid, possible tuite-forming reactions are

$$ 2\,{\text{apatite}} + 3\,{\text{enstatite}} = 3\,{\text{tuite}} + {\text{forsterite}} + {\text{diopside}} + {\text{H}}_{2} {\text{O}} $$
(1)
$$ 4\,{\text{apatite}} + {\text{pyrope}} + {\text{diopside}} = 6\,{\text{tuite}} + {\text{grossular}} + 2\,{\text{forsterite}} + {\text{H}}_{2} {\text{O}} $$
(2)
$$ 6\,{\text{apatite}} + {\text{pyrope}} + 3\,{\text{enstatite}} = 9\,{\text{tuite}} + {\text{grossular}} + 3\,{\text{forsterite}} + {\text{H}}_{2} {\text{O}} $$
(3)

Based on the negative slope of the reaction grossular + 3 enstatite = pyrope + 3 diopside and experimental evidence for a negative slope of the tuite-forming reaction(s) (Fig. 2a), a possible arrangement in pressure and temperature of the reactions is shown in Fig. 2b. The stoichiometry of the apatite breakdown reaction cannot be unambiguously identified due to the absence of apatite breakdown products in addition to the phases present and the lack of significant compositional changes of garnet and pyroxene across the apatite–tuite breakdown pressure–temperature interval. In bulk composition II, increasing Ca and decreasing Mg in garnet between 7.5 and 8.7 GPa would be consistent with both reactions (2) and (3). The experiments of this study also show that apatite is stable to at least 1,100°C at 6 GPa and to 1,050–1,100°C at 8 GPa. The maximum temperature at which tuite was found is 1,300°C at 9 GPa. Within the spacing of experimental data points, there is no evidence for an extended stability of apatite or tuite coexisting with a melt. At 8 GPa for example, tuite disappears between 1,100 and 1,200°C, and at 1,200°C, the assemblage is olivine + orthopyroxene + clinopyroxene + garnet. In summary, the experimental data show that both apatite and tuite are stable in a wide range of subduction zone PT regimes and that apatite may also be present in cool lithospheric mantle to ~5–6 GPa (Fig. 2a). The absence of tuite from the experiment at 15 GPa/1,500°C further indicates that tuite is not stable in convecting asthenospheric mantle.

Phosphorus in peridotitic upper mantle phases and implications for phosphorus mobility in the upper mantle

The experiments of this study show that garnet has by far the highest phosphorus content of any solid upper mantle silicate phase. Phosphorus concentrations systematically increase with pressure from >0.1 wt% P2O5 at 2.5 GPa/900°C to 1.6 wt% P2O5 at 15 GPa/1,500°C (Fig. 4) and decrease with increasing temperature once the solidus is crossed due to preferential partitioning into coexisting melt. This increase is coupled to decreasing Si at P ≤ 12.8 GPa (Fig. 5d), confirming the extreme preference of phosphorus to substitute for [4]Si irrespective of the chemical system. The systematic increase in Na with pressure (Figs. 5a, b) indicates that the exchange [8]Na[4]P[8]M2+ −1 [4]Si−1 is operative. Due to the low bulk Na content of peridotite, however, coupled Na–P incorporation into garnet can only account for a minor portion of the observed phosphorus. If R1+ cations are not available, phosphorus can still enter the garnet structure via [6]M2+[4]P[6]M3+ −1 [4]Si−1 (Brunet et al. 2006). Whereas no correlation is obvious between Al and P, a negative correlation can be observed between P and Cr (Fig. 5c). This shows (1) that the berlinite substitution [4]Al3+[4]P[4]Si−2 is not operative and (2) that [6]M2+[4]P[6]Cr3+ −1 [4]Si−1 with a minor contribution of [8]Na[4]P[8]M2+ −1 [4]Si−1 is most likely to be responsible for phosphorus incorporation into the peridotitic garnets. Compared to MORB-type bulk compositions where [8]Na[4]P[8]M2+ −1 [4]Si−1 is dominant (cf. Konzett and Frost 2009), phosphorus concentrations in peridotitic garnets are significantly higher across the entire PT range investigated (Fig. 4).

Mechanisms for phosphorus incorporation into the Mg2SiO4 polymorphs are difficult to deduce at the levels of P2O5 contents encountered in this study. All schemes proposed so far assume that phosphorus substitutes for tetrahedral Si with charge balance maintained by octahedral vacancies possibly coupled with the introduction of trivalent cations (Milman-Barris et al. 2008 and references therein). The variability in the phosphorus concentration of individual olivine grains observed in this study (cf. Figs. 1h, 6) is consistent with results from olivine crystallization experiments by Milman-Barris et al. (2008). These authors consider two possible mechanisms for local phosphorus enrichment in olivine, both of which result from rapid crystal growth in the presence of melt: (1) phosphorus enrichment in melt boundary layers adjacent to growing olivine, which may lead to elevated phosphorus contents in olivine due to equilibrium partitioning and (2) sluggish phosphorus diffusion in the melt away from the advancing crystal face, resulting in elevated phosphorus contents of olivine due to disequilibrium partitioning. In the present study, melt–crystal boundary layer effects can only play a significant role in the experiments at 1,300 and 1,500°C where quenched melt is present. At lower temperatures (e.g. MA52 or MA31), heterogeneous phosphorus distribution on a micrometre scale in the starting material combined with sluggish solid-state diffusion involving cations on several crystallographic sites is the likely reason for the observed phosphorus zoning in olivine.

Considering all averaged phosphorus concentrations, a consistent succession of the relative enrichment of phosphorus in the peridotitic phases can be observed with Pmelt ≫ Pgrt ≫ PMg2SiO4 > Pcpx > Popx (Figs 1h, 7). Resulting values for D mineral/meltP are between 0.08 for garnet and 0.006 for orthopyroxene (Table 4). This is consistent with phosphorus partitioning amongst spinel lherzolite phases measured by Brunet and Chazot (2001) albeit with a considerably lower absolute phosphorus concentration for the silicate melt reported by these authors. Bishop et al. (1978) found slightly higher phosphorus contents in clinopyroxene than in olivine in a suite of peridotite xenoliths from South African kimberlites with Pgrt > Pcpx > Pol > Popx. The phosphorus contents of melt and coexisting clinopyroxene of this study are consistent with results on apatite solubility in carbonatitic melts at 3 GPa obtained by Baker and Wyllie (1982). The data from 6 GPa/1,300°C and 1,500°C combined with averaged garnet peridotite facies mineralogy (McDonough and Rudnick 1998) yield bulk phosphorus partition coefficients for mantle lherzolites of 0.018–0.024 (Table 4). These values are higher than those calculated by Baker and Wyllie (1992) based on their experimental data and phosphorus analyses of natural peridotite minerals due to the markedly higher phosphorus contents in olivine and garnet. Using averaged P2O5 concentrations of the peridotite silicates obtained in this study, the phosphorus storage capacity of a garnet–lherzolite at PT conditions close to the apatite ⇒ tuite reaction is ~0.2–0.3 wt% dependent on the modal proportion of phases (Table 4). This amount can be supplied by an equivalent of ~0.5–0.7 wt% apatite and is a minimum estimate because it does not take into consideration the potential presence of an additional fluid. It should be kept in mind that although garnet has by far the highest P2O5 concentrations of any solid peridotite phase, olivine is no less important as a potential phosphorus carrier due to its large modal amount in peridotites. This is true even for primitive mantle with the highest amount of garnet where olivine and garnet contribute to the bulk phosphorus partition coefficient in almost equal proportions (cf. Table 4). Because phosphorus contents in olivine show no discernable pressure dependence, it can be assumed that the high phosphorus storage capacity of olivine extends to pressures <3 GPa. In fact, Brunet and Chazot (2001) reported P2O5 contents of 0.07–0.15 wt% in olivine from an apatite-bearing spinel lherzolite, which is in good agreement with 0.16 ± 0.03 (n = 7) wt% P2O5 found at 4 GPa/1,000°C. Hence, both data from natural rocks and the experiments show that olivine alone is capable of storing the entire phosphorus budget of peridotitic mantle and that the formation of a phosphate phase cannot be expected unless phosphorus is added metasomatically.

Fig. 7
figure 7

Averaged phosphorus contents of coexisting peridotite phases from bulk composition I in the PT range 6–9 GPa and 1,100–1,500°C

Full size image
Table 4 Phosphorus partition coefficients based on averaged P2O5 contents of mineral phases from high PT experiments compared to literature data
Full size table

The experiments show that in the presence of apatite peridotitic garnet and olivine accommodate ≥0.25 and ≥0.16 wt% P2O5, respectively, as the pressure increases to ≥4 GPa (Fig. 4; Table 2). These high values are not found in garnets or olivines from orogenic peridotites. Whereas this can be simply explained by the low bulk phosphorus contents, even apatite-bearing peridotites contain garnet and olivine, which are very poor in phosphorus. For example, olivine and coexisting garnet with apatite and magnesite inclusions equilibrated at 850°C and 4.1 GPa in peridotites from the Bohemian Massif, Czech Republic, both contain <0.05 wt% P2O5 (Vrána 2009). This discrepancy indicates that substantial compositional changes of the peridotite phases take place during slow exhumation of orogenic peridotites similar to those inferred for MORB-type eclogite assemblages (Hermann and Spandler 2007; Konzett and Frost 2009).

Apatite associated with calcite/dolomite ± clinopyroxene ± phlogopite is often observed in peridotitic mantle xenoliths and taken as indication for the former (transient) presence of a phosphorus-rich carbonatitic melt (e.g. Green and Wallace 1988; Yaxley et al. 1991; Hauri et al. 1993; Coltorti et al. 1999; O’Reilly and Griffin 2000). Due to their extremely low viscosity (Dobson et al. 1996), carbonatitic melts are thought to migrate pervasively in small amounts along grain boundaries with a strong tendency of dispersing instead of pooling (Hammouda and Laporte 2000). This style of migration inhibits large-scale melt segregation and promotes solid/melt element exchange, thus making carbonatites eminently suitable agents for mantle metasomatism. The results from this study and the rare data for phosphorus in silicates from natural peridotites both indicate that even in shallow upper mantle, a small amount of phosphorus-rich carbonatitic melt would quickly get stripped of its phosphorus when travelling through previously unmetasomatized garnet + olivine-bearing mantle. Hence, transport of phosphorus via carbonatitic melts is only possible (1) if garnet and olivine are already saturated in phosphorus as a result of an earlier episode of carbonatite impregnation, (2) if the fluid/melt-rock ratio is sufficiently high to exhaust the storage capacity of garnet and olivine or (3) through a transport mechanism that is fast enough to prevent equilibration between melt and peridotitic country rocks. Extremely fast magma ascent at a speed on the order of metres per second caused by high-speed dike propagation (Sparks et al. 2006; Wilson and Head 2007; Peslier et al. 2008) thus must play an important role in the preservation of the high bulk phosphorus contents of up to several wt% P2O5 typical of kimberlites and orangeites (Le Roex et al. 2003; Mitchell 1995). The same is true for certain P2O5-rich magnesio-carbonatites and leucitites (Dalton and Presnall 1998; Bailey 1989; Humphreys et al. 2010).

Potassium and chlorine transport in peridotitic mantle—the role of apatite and hydrous silicates

In recent years, an increasing number of reports have been published on K–Cl-P-LILE-rich hypersaline brines and carbonatitic fluids or melts and associated K–Cl-rich solid inclusions in diamonds (Tomlinson et al. 2005; Schrauder and Navon 1994; Klein-BenDavid et al. 2006; Wirth et al. 2009; Izraeli et al. 2001, 2004). These exotic fluids/melts show that at least on a local scale, substantial amounts of Cl must be present in the subcontinental mantle at depths >150 km in spite of the very high recycling efficiency assumed for Cl in subduction zones (e.g. Straub and Layne 2003). Because K–Cl-carbonatitic fluids/melts are not permanently present under subduction zone PT conditions and also in cool lithospheric mantle at least to pressures ≤5 GPa (Fig. 2a), solid Cl carriers must play an important role in the long-term storage of Cl in the mantle and may become the source of Cl-rich fluids/melts during renewed heating. The present study indicates that apatite is a potential Cl carrier to a depth of ≤150–180 km under conditions of a 40 mW/m2 geotherm. At greater depths/higher temperatures, the apatite–peridotite solidus is intersected in a carbon-saturated mantle (Fig. 2a), resulting in the disappearance of apatite and the formation of P–Cl-rich carbonatitic melts/fluids (e.g. Baker and Wyllie 1992). If present in modal amounts exceeding ~1% (cf. Table 4), apatite may reach its upper pressure stability limit in subduction zones at depths between ~250 and 300 km dependent upon the thermal regime.

Hydrous silicates capable of storing Cl are stable to at least 600 km depth in the mantle wedge above subduction zones. The succession of phases is calcic amphibole (≤3 GPa) → phlogopite (≤6.5 GPa) → K-richterite (≤15 GPa) → phase X (≤20 GPa) (Konzett and Fei 2000). The very small database on Cl in hydrous mantle silicates shows that Cl contents of phlogopites and calcic amphiboles are almost always <0.1 wt% and very rarely exceed 0.5 wt%. (Zanetti et al. 1996; Naemura et al. 2009; Matson et al. 1986; Smith et al. 1981; Zaccharini and Stumpfl 2004; Hervig and Smith 1981; Peng et al. 1995; Dawson 2002; Wartho and Kelley 2003). Cl contents of apatite by comparison are usually at least one order of magnitude higher and may reach ≥5 wt% (Bonatti et al. 1986; Woodland et al. 1996). This study shows that in peridotitic bulk compositions apatite can coexist with both phlogopite and K-richterite and that significant amounts of Cl are present in both silicate phases. In the PT range 2.5–5 GPa/900–1,000°C, K ap−phlD,Cl values (cf. Zhu and Sverjensky 1992) are between 6.5 and 9.7 (Table 5). Although the Cl concentrations in apatite and phlogopite in the 4 and 5 GPa runs are different probably due to variable Cl-loss to a fluid, the KD-values are similar, indicating that Cl exchange equilibrium was at least approached. At 7 GPa/800°C, coexisting apatite and K-richterite yield K ap-KrD,Cl  = 2.0 with 0.56 ± 0.07 and 0.90 ± 0.13 wt% Cl in K-richterite and apatite, respectively. This shows that K-richterite is a suitable Cl carrier and potentially important for the crust-to-mantle transfer of Cl in subduction zone peridotites in a depth range of ~250 to 450 km.

Table 5 Chlorine partition coefficients for coexisting apatite, phlogopite and amphibole based on their averaged Cl contents
Full size table

Whereas phlogopite is stable at 4 and 5 GPa in bulk composition II, no solid potassic silicate was found at 7.0 and 7.5 GPa. By comparison, in Cl-free peridotites containing almost identical amounts of potassium and water, potassic silicates are continuously stable to ≥12 GPa (Konzett and Ulmer 1999; Konzett and Fei 2000). This indicates that Cl strongly raises the solubility of potassium in hydrous fluidsFootnote 1 at P ≥ 7 GPa. Hence, such a fluid encountering an assemblage phlogopite/K-richterite + Cl-rich apatite may become Cl-enriched by preferential partitioning of Cl into the fluid. Given high enough pressures, the Cl-enrichment of the fluid strongly raises its potassium solubility and may trigger breakdown of phlogopite/K-richterite, thereby generating small amounts of fluids extremely rich in K, Cl and other incompatible elements. Evolving in a closed system, e.g. when trapped in a diamond, such a fluid would have the potential to crystallize unusual phases such as KCl or Cl-rich hypersilicic mica in addition to apatite and carbonates as observed in diamond micro-inclusions (e.g. Klein-BenDavid et al. 2006; Wirth et al. 2009).