Transport and Storage of Potassium in the Earth’s Upper Mantle and Transition Zone: an Experimental Study to 23 GPa in Simplified and Natural Bulk Compositions

Abstract

INTRODUCTION

The hydrous potassic phases (HPP) phlogopite and K-amphibole are major storage sites for potassium in the Earth’s upper mantle. Both host the incompatible trace elements with large ionic radii (Rb, Ba or Pb), which can occupy large and highly coordinated lattice positions in phlogopite and K-amphibole (Basu, 1978; Foley et al., 1995; Ionov et al., 1997). Phlogopite is an important component in sources of kimberlites, lamproites, lamprophyres and K-rich basaltic rocks (e.g. Wilkinson & LeMaitre, 1986; Esperança & Holloway, 1987; Rogers, 1992; Mitchell, 1995, and references therein; Sato, 1997). K-richterite has been suggested as a possible source of K, Ti and high field strength elements (HFSE) for kimberlites and ultrapotassic rocks (Foley, 1992; Taylor et al., 1994). Phlogopite is stable in Al-rich metasedimentary and peridotitic bulk compositions. In Al-rich bulk compositions phlogopite breaks down at relatively low P (and T) to form phengite ± K-feldspar and possibly K,Mg-rich fluid (Massonne & Schreyer, 1989; Massonne, 1992). In peridotitic bulk compositions, phlogopite is stable to at least 6 GPa and 1100°C with orthopyroxene + clinopyroxene + olivine (Konzett & Ulmer, 1999; see also Wendlandt & Eggler, 1980; Mengel & Green, 1989, for experiments at lower P) and to at least 12 GPa and 1350°C with phlogopite + clinopyroxene (Luth, 1997). Phlogopite is stable above the solidus of natural lherzolite at P ≤ 3 GPa (Wendlandt & Eggler, 1980). Experiments with phlogopite + enstatite (Sato et al., 1997) suggest above-solidus stability of phlogopite through continuous melting reactions between 4 and 8 GPa that form olivine + pyrope. Although it is difficult to distinguish hydrous melts from fluids at high P, K-richterite is likely to be unstable above the solidus in either subalkaline or peralkaline bulk compositions (Konzett et al., 1997; Konzett & Ulmer, 1999). However, K-richterite probably controls the P–T location of the solidus by incongruent melting (Gilbert & Briggs, 1974; Foley, 1991).

K-amphibole with a composition close to KNaCaMg₅Si₈O₂₂(OH)₂ can form as a high-P breakdown product of phlogopite in peridotitic bulk compositions at P > 6 GPa (Trønnes et al., 1988; Luth, 1997; Konzett & Ulmer, 1999). In Na-free systems the K-amphibole KKCaMg₅Si₈O₂₂(OH)₂ is present (Sudo & Tatsumi, 1990; Luth, 1997; Inoue et al., 1998; Yang et al., 1999). In peralkaline bulk compositions [mica–amphibole–rutile–ilmenite–diopside (MARID), lamproites] phase relations from natural rocks and from high-P experiments suggest a continuous stability of K-amphibole + phlogopite from 0·1 MPa to at least 8·5 GPa (Mitchell & Bergman, 1991, and references therein; Konzett et al., 1997).

Experimental studies of the KCMSH and KCMASH systems (Luth, 1997; Inoue et al., 1998) show that K-amphibole breaks down at high pressures to a hydrous K-rich silicate that is capable of transporting alkalis and water to even greater depths than does amphibole. The structure, stoichiometry and compositional variability of this phase—termed phase X (Luth, 1995)—are still unknown and its stability field is poorly constrained. The aims of our study are to (1) better constrain the stability field of phase X, especially its high-pressure stability limit; and (2) determine the chemical variability of phase X and coexisting phases. This will permit us to assess the potential of phase X as a storage site for water and alkalis in the mantle transition zone and to trace mechanisms of potassium and water recycling into the mantle.

COMPOSITION OF STARTING MATERIALS

Our subalkaline and peralkaline starting materials are mixes of high purity (≥99·95% purity) synthetic oxides or silicates and carbonates, and cover the full range of bulk compositions that can stabilize K-amphibole and its breakdown products (Table 1). They represent peridotitic and MARID-type (Dawson & Smith 1977) or lamproitic bulk composition, respectively. The phase relations of these bulk compositions at P < 10 GPa have been described by Konzett et al. (1997) and Konzett & Ulmer (1999). A simplified K₂O–Na₂O–CaO–MgO–Al₂O₃–SiO₂–H₂O (KNCMASH) system was examined to avoid the potential influence of fO₂ on phase relations in an Fe-bearing system at very high pressures as a result of stabilization of phases by Fe³⁺ partitioning. Despite this simplification, the system is still complex enough to permit all important exchange reactions (MgSiAl₋₂, NaSiCa₋₁Al₋₁, CaMg₋₁, KNa₋₁) among the silicate phases. Additional experiments were conducted with KLB-1 peridotite (Takahashi, 1986) doped with 10 wt % synthetic K-richterite, to locate the amphibole → phase X transition in a natural peridotitic bulk composition.

EXPERIMENTAL AND ANALYTICAL TECHNIQUES

Experiments (Table 2) were performed with Walker-type and split-sphere MA-8 multianvil devices at the Geophysical Laboratory (GL) and at the Bayerisches Geoinstitut (BG), respectively, using prefabricated pyrophyllite gaskets and MgO octahedra. Assembly sizes and furnace materials are as follows: GL: 10/5 (10 mm edge length of octahedra/5 mm truncated edge length of WC cubes) and 8/3 assemblies to 15 and 23 GPa using Re heaters; BG: 14/7 assemblies using stepped LaCrO₃ heaters. Pre-dried starting materials were placed in 1·55 mm or 1·00 mm (GL, for 8/3 assemblies) outer diameter Pt₁₀₀ capsules and welded shut immediately. For the KLB-1 starting material, an additional inner graphite capsule (approximate dimensions after runs: wall thickness 300 μm, bottom and lid 200 μm) was added. To minimize T gradients and phase separation as a result of thermal diffusion, the length of experimental charges ranged between 200 and 600 μm. T gradients were ∼20°C/100 μm for 8/3 assemblies and <10°C/100 μm for 10/5 assemblies (Bertka & Fei, 1997, and unpublished data, 1999). Temperatures were measured with W3%Re–W25%Re thermocouples without correcting for the pressure effect on e.m.f. Both pressure and temperature were computer controlled during the runs. Detailed descriptions of the GL and BG experimental and calibration procedures have been given by Bertka & Fei (1997) and Rubie et al. (1993), respectively.

Sample capsules from completed experiments were embedded in epoxy resin and ground to expose the center of the charges. Most phase compositions were analyzed with an electron microprobe at analytical conditions of 15 kV and 20 nA. Phase X, which was found to be extremely susceptible to beam damage and loss of alkalis, was analyzed with 5 nA beam current and a rastered electron beam as large as the size of phase X grains permitted (typically 10–20 μm). Counting times of 20 s on peaks and 10 s on backgrounds of the X-ray lines were ratioed to a combination of synthetic oxide (Si, Mg, Al), synthetic mineral (Na) and natural mineral (Ca, K) standards. Data were corrected on-line using the PRZ correction procedure. After standardization, no peak search procedures were performed on phase X grains, to minimize residence time of the electron beam. Microprobe analyses of phlogopite and amphibole–pyribole were recalculated assuming stoichiometric OH. No recalculation was attempted for phase X because H₂O was not determined quantitatively.

To search for structural OH in phase X, Raman spectra were recorded at the GL with a Dilor XY confocal micro Raman spectrometer equipped with a cryogenic Wright Model CCD. The excitation source was the 514 nm line of a Coherent Innova Model 90-5 Ar⁺ laser operating at 150 mW power using an integration time of 600 s.

PREVIOUS EXPERIMENTAL WORK

Phase X was described as a breakdown product of K-amphibole at P > 14 GPa between 1100 and 1400°C in the KCMSH system by Inoue et al. (1995a, 1998). Luth (1995, 1997) observed phase X in the system phlogopite–diopside at P ≥ 11 GPa. Based on secondary ionization mass spectrometry (SIMS) measurements of OH combined with microprobe analyses, Inoue et al. (1995a) proposed a formula of K₄Mg₈Si₈O₂₅(OH)₂ for phase X. Experimental results of Luth (1997) and Inoue et al. (1998) show a wide range in K₂O contents (10·2–19·4 wt % K₂O) and oxide totals (91·5–98·2 wt %) at relatively constant Si:Mg or Si:(Mg + Al) ratios of 1:1. Phase X may be the ‘amphibole-like mineral’ reported by Trønnes (1990) as a breakdown product of phlogopite between 11 and 12 GPa. The formula given by Trønnes (1990)—K_3·3Mg_6·5Al_0·5Si₇O₂₂(OH)₂—is similar to the composition of phase X reported by Luth (1997) and Inoue et al. (1998).

RESULTS

Petrography and chemical homogeneity of the phases

All starting materials readily recrystallize to mineral grains ∼10–50 μm in size at T ≥ 1100°C and to grains ≤10 μm in size in lower T runs (Fig. 1). HPPs are typically euhedral to subhedral but phase X also forms irregular grains that contain numerous olivine, clinopyroxene, or amphibole inclusions. Many phase X grains show irregularly spaced cleavage (Fig. 1b). A mixed-chain hydrous pyribole (sensuVeblen, 1981) was present in three runs, along with K-richterite. The former occurs as lath-shaped to needle-like crystals up to 100 μm × 20 μm in size. In high-P runs, K-hollandite appears as small (≤20 μm × 5 μm) needle-like crystals, and Ca-perovskite forms irregular patches up to 20 μm in size dispersed in the matrix or is present as inclusions in garnet. The phase distribution within individual capsules is inhomogeneous, and melt or quenched fluid is most abundant in the hotter part of the capsule. In experiments on the peralkaline bulk compositions, evidence for quench crystallization is present over a T interval of ≥200°C. The modal amounts of quench vary from <3% at 10 GPa and 1100°C to ∼20% at 10 GPa and 1350°C. Near the hot ends of the capsules, along the solid–liquid and quench interface, garnet and clinopyroxene often display larger grain sizes compared with cooler parts of the capsule. Inhomogeneities in grain size and phase distribution can be ascribed to grain maturation and chemical diffusion in a temperature gradient (Lesher & Walker, 1988) aided by the water content of the systems. With the exceptions of large garnets and phase X, neither systematic zoning of individual mineral grains nor differences in phase compositions are observed between tops and bottoms of the capsules. In low-T runs, diffuse Al-rich cores in garnets can be ascribed to incomplete equilibration. Phase X may show strong zoning with respect to K/(K + Na). This zoning is fairly regular, with K-rich cores and Na-rich rims, or is patchy and irregular. These compositional inhomogeneities are independent of run duration and temperature.

Fig. 1.

Open in new tabDownload slide

Back-scattered electron photomicrographs of phase X-bearing assemblages taken from the centers of experimental charges (tops of the images point to the hotter end of the capsules). (a) Run JKW7 at 14 GPa and 1300°C showing euhedral crystals of phase X in part with heterogeneity in K/(K + Na): light areas are K rich, whereas dark areas are more Na rich. (b) Run JKW9 at 12 GPa and 1300°C with coexisting phase X and K-richterite. Phase X shows irregular cleavage and numerous inclusions of Kr and hiCapx. Holes in the sample surface are due to mechanical abrasion during polishing. (c) Run JKW47 at 20 GPa and 1500°C showing poikiloblastic phase X coexisting with majoritic garnet and γ-Mg₂SiO₄; abbreviations are given in Table A1.

Phase relations

Peralkaline KNCMASH

In the peralkaline KNCMASH system, HPPs are stable to at least 20 GPa and 1300°C, along with garnet + Mg₂SiO₄ + high-Ca clinopyroxene ± low-Ca clinopyroxene ± K-hollandite ± Ca-perovskite (Fig. 2). With increasing pressure, the first HPP to disappear is phlogopite, which is stable at 10 GPa between 1100°C and 1350°C along with K-richterite, but absent from run JKW17 at 11 GPa and 1300°C. In runs at 11 GPa and 1300°C, and 13 GPa and 1100°C K-richterite is the only stable HPP. At higher pressures, amphibole is joined by phase X as a result of continuous amphibole breakdown. At ≥15 GPa and 1100°C, and 14 GPa and 1300°C, the upper pressure stability limit of K-richterite is reached and amphibole is replaced by phase X as the HPP. Within the spacing of experimental data points, both phase X-in and K-richterite-out reactions have negative slopes with 5 MPa/K < dP/dT < 15 MPa/K for phase X-in and ≤10 MPa/K for K-richterite-out. At 20 GPa and 1300°C, K-hollandite appears as the first anhydrous potassic phase as a result of continuous phase X breakdown. Between 18 and 20 GPa, high-Ca clinopyroxene breaks down, and its diopside and jadeite components form Ca-perovskite and sodium-garnet solid solution, respectively.

Fig. 2.

Open in new tabDownload slide

P–T diagram of experimental results in the peralkaline KNCMASH system (Tables 1 and 2). Phases present in the experimental charges are represented by black sectors within the run symbol; phases not detected are denoted by white sectors (inset upper left); abbreviations are given in Table A1; ACMA average current mantle adiabat (see text); α- to β-Mg₂SiO₄ and β- to γ-Mg₂SiO₄ transitions according to Morishima et al. (1994) and Katsura & Ito (1989), respectively; experimental data at P < 10 GPa from Konzett et al. (1997) for comparison.

The high-temperature stability limit of all HPPs is reached between 1300 and 1400°C. At conditions of 8 GPa and 1400°C, and 13 GPa and 1400°C the stable assemblage is garnet + high-Ca clinopyroxene + Mg₂SiO₄ + quench (the term ‘quench’ is used to denote a mixture of unidentified and mostly K-rich phases that crystallized from a solute-rich fluid or a hydrous melt upon quenching). The spacing of experimental data points (Fig. 2) precludes discussion of the slope of the K-phase-out reaction(s), but in accordance with results at P ≤ 8·5 GPa (Konzett et al., 1997), we chose a positive slope. Because of the difficulty in distinguishing quenched melts from solute-rich fluids at high pressures based on textural evidence (see Konzett et al., 1997) we did not attempt to locate the position of the solidus. The mixed-chain hydrous pyribole was found in runs at 10 GPa and 15 GPa (Table 3), either with K-richterite + phlogopite or K-richterite alone (see Finger et al., 1998; Konzett & Fei, 1998).

Subalkaline KNCMASH and KLB-1

In the subalkaline system, K-richterite is stable to 13 GPa along with olivine + garnet + high-Ca clinopyroxene ± low-Ca clinopyroxene in an assemblage resembling a metasomatized lherzolite (Fig. 3). The apparent lack of low-Ca clinopyroxene in run Ma95sB is probably due to small grain sizes and the sparse occurrence of this phase in subalkaline runs. At P ≥ 14 GPa, K-richterite is replaced by phase X. The slope of the K-richterite-out reaction was assumed to be slightly negative, in accordance with the results of K-amphibole breakdown in the KCMSH system obtained by Inoue et al. (1998). Phase X is stable to at least 20 GPa. At this pressure, phase X may coexist with K-hollandite as a result of H₂O partitioning constraints (see below), and with Ca-perovskite produced by the breakdown of high-Ca clinopyroxene. At 23 GPa, phase X is absent and K-hollandite carries the K in the system. The upper T stability limit of phase X is between 1400 and 1600°C at 14 GPa, in an assemblage without a solid K-rich phase: high-Ca clinopyroxene + garnet + low-Ca clinopyroxene + quench. At 20 GPa, the K-phase-out reaction must occur at T > 1700°C, which is at least 200°C above an average current mantle adiabat (ACMA) as defined by a β → γ transition in Mg₂SiO₄ at 17·9 GPa and 1475°C (Katsura & Ito, 1989) and a γ → perovskite + wüstite transition at 23·2 GPa and 1530°C (Ito & Takahashi, 1989).

Fig. 3.

Open in new tabDownload slide

P–T diagram of experimental results in the subalkaline KNCMASH (pie symbols) and the KLB-1 (square symbols) systems (Tables 1 and 2). Meaning of symbols, ACMA and α → β → γ transitions in Mg₂SiO₄ as in Fig. 2; K-amph out I98 and L97 indicate K-amphibole-out reactions in the KCMSH and KCMASH systems according to Inoue et al. (1998) and Luth (1997), respectively; filled arrows bracket position of H₂O-saturated solidus for KLB-1 peridotite according to Kawamoto et al. (1996) (K96).

In the K-richterite-doped peridotite KLB-1, HPPs coexist with garnet + low-Ca clinopyroxene + high-Ca clinopyroxene + Mg₂SiO₄. K-richterite is stable at 12 GPa and 1200°C after a run duration of 72 h. X-ray mapping also showed that a K-rich phase is dispersed within graphite, along the interface between experimental charge and inner graphite capsule (see Konzett & Ulmer, 1999). At 13 GPa and 1200°C, after an identical run time, no HPP could be identified and X-ray mapping showed that all K was concentrated at the charge–graphite interface in diffuse grain boundary films. Because of their extremely small grain size, the K-carrying phase(s) could not be identified. Small amounts of alkaline and possibly CO₂-rich fluid that probably formed by amphibole breakdown and/or graphite oxidation evidently destabilized phase X with increasing run durations. In a run that lasted only 48 h, phase X was stabilized at 14 GPa and 1200°C (see Table 9, below). Thus, the amphibole breakdown for starting material KLB-1 was placed between 12 and 13 GPa at 1200°C, which is 1 GPa below the amphibole → phase X transition in the subalkaline KNCMASH system (Fig. 3).

Mineral chemistry

K-amphibole and hydrous pyribole

Fig. 4.

Open in new tabDownload slide

Selected mineral chemical parameters of K-richterite as a function of P at constant T (1100°C) for peralkaline and subalkaline KNCMASH systems; each point represents an average of 5–13 individual amphibole analyses (Table 3).

In the KLB-1 starting composition, amphibole at its upper P stability limit is still close to K-richterite end-member composition (see Table 9, below) with 1·07–1·13 K p.f.u. and 0·11–0·13 Al p.f.u. The X_Mg for amphibole is 0·95 with X_Mg^Kr > X_Mg^loCapx > X_Mg^ol > X_Mg^hiCapx > X_Mg^ga.

Fig. 5.

Open in new tabDownload slide

Compositional variation of mixed-chain hydrous pyribole in terms of Na–Ca and Mg–Al; □, hypothetical pyribole end-member compositions (see text for explanation).

Phlogopite

Phlogopite is near the end-member composition, with a small excess of 0·02–0·04 Si p.f.u. (Table 4), indicating limited solid-solution with K(Mg_2·5□_0·5)Si₄O₁₀(OH)₂ (Seifert & Schreyer, 1971).

Phase X

Phase X is a disilicate (Libau, 1982), containing corner-sharing SiO₄ tetrahedra and has a stoichiometry A_2–xB₂Si₂O₇H_x with A = K or Na, B = Mg, Al or Ca, and x = 0–1, ranging from anhydrous A₂B₂Si₂O₇ to A□B₂Si₂O₇H with a maximum possible H₂O content of 3·51 wt % and a vacancy on the A-position (H. Yang & J. Konzett, unpublished data, 1999; Konzett & Yang, 1998). Variable H₂O contents have been observed in phase E (Kudoh et al., 1993), phase D (Frost & Fei, 1998), hydrous wadsleyite (Kudoh et al., 1996) or hydrous modified spinel (Inoue et al., 1995b). The presence of OH in phase X in both subalkaline and peralkaline systems was confirmed by laser Raman spectroscopy. Raman spectra show strong peaks in the OH-stretching region at 3600 cm⁻¹ (Fig. 6). This is consistent with low average analytical oxide totals in the range of 96–98 wt % in most runs (Table 5). In the silicate stretching region, phase X spectra are characterized by major peaks at 635–640 cm⁻¹ and 895–903 cm⁻¹ (Fig. 6). The composition of phase X may be inhomogeneous in K/(K + Na), which may range between 0·15 and 0·88 in an individual run (Table 5). This compositional inhomogeneity is restricted to the K/(K + Na) ratio and occurs even if the coexisting phases are well equilibrated. Other chemical characteristics such as Al or Ca contents and Mg/Si ratios show no correlation with K/(K + Na). A decrease in the degree of inhomogeneity can be observed with increasing pressure and K₂O contents. Because the inhomogeneity of the phase X composition is independent of run duration and temperature, it cannot be easily explained by a failure of the runs to attain equilibrium. It might instead be a quench effect promoted by high Na contents of phase X. This effect may result from the difference of the ionic radii of K and Na,which leads to a lattice collapse during pressure release. Excluding Na-rich rims, an increase in K₂O and a decrease in Na₂O with increasing P (and T) is observed (Fig. 7). In the subalkaline bulk composition, the highest K₂O contents of phase X (20–25 wt %) are correlated with the highest oxide totals of >99 wt % (Fig. 8). Although oxide totals can be affected by the analytical technique (i.e. variable electron beam raster size adjusted to grain size of phase X) this correlation indicates decreasing H₂O contents of phase X with increasing P. The effect may be explained by an increasing ability of coexisting phases—especially Mg₂SiO₄—to incorporate H₂O (Kohlstedt et al., 1996), which changes H₂O partitioning and leads to a continuous dehydration of phase X. The increase in K of phase X with pressure is consistent with other studies (Luth, 1997; Inoue et al., 1998) and parallels results for phlogopite and K-richterite (Konzett & Ulmer, 1999).