The high-pressure stability of chlorite and other hydrates in subduction mélanges: experiments in the system Cr₂O₃–MgO–Al₂O₃–SiO₂–H₂O

Abstract

The solubility of chromium in chlorite as a function of pressure, temperature, and bulk composition was investigated in the system Cr₂O₃–MgO–Al₂O₃–SiO₂–H₂O, and its effect on phase relations evaluated. Three different compositions with X _Cr = Cr/(Cr + Al) = 0.075, 0.25, and 0.5 respectively, were investigated at 1.5–6.5 GPa, 650–900 °C. Cr-chlorite only occurs in the bulk composition with X _Cr = 0.075; otherwise, spinel and garnet are the major aluminous phases. In the experiments, Cr-chlorite coexists with enstatite up to 3.5 GPa, 800–850 °C, and with forsterite, pyrope, and spinel at higher pressure. At P > 5 GPa other hydrates occur: a Cr-bearing phase-HAPY (Mg_2.2Al_1.5Cr_0.1Si_1.1O₆(OH)₂) is stable in assemblage with pyrope, forsterite, and spinel; Mg-sursassite coexists at 6.0 GPa, 650 °C with forsterite and spinel and a new Cr-bearing phase, named 11.5 Å phase (Mg:Al:Si = 6.3:1.2:2.4) after the first diffraction peak observed in high-resolution X-ray diffraction pattern. Cr affects the stability of chlorite by shifting its breakdown reactions toward higher temperature, but Cr solubility at high pressure is reduced compared with the solubility observed in low-pressure occurrences in hydrothermal environments. Chromium partitions generally according to $X_{\text{Cr}}^{\text{spinel}}$ ≫ $X_{\text{Cr}}^{\text{opx}}$ > $X_{\text{Cr}}^{\text{chlorite}}$ ≥ $X_{\text{Cr}}^{\text{HAPY}}$ > $X_{\text{Cr}}^{\text{garnet}}$. At 5 GPa, 750 °C (bulk with X _Cr = 0.075) equilibrium values are $X_{\text{Cr}}^{\text{spinel}}$ = 0.27, $X_{\text{Cr}}^{\text{chlorite}}$ = 0.08, $X_{\text{Cr}}^{\text{garnet}}$ = 0.05; at 5.4 GPa, 720 °C $X_{\text{Cr}}^{\text{spinel}}$ = 0.33, $X_{\text{Cr}}^{\text{HAPY}}$ = 0.06, and $X_{\text{Cr}}^{\text{garnet}}$ = 0.04; and at 3.5 GPa, 850 °C $X_{\text{Cr}}^{\text{opx}}$ = 0.12 and $X_{\text{Cr}}^{\text{chlorite}}$ = 0.07. Results on Cr–Al partitioning between spinel and garnet suggest that at low temperature the spinel- to garnet-peridotite transition has a negative slope of 0.5 GPa/100 °C. The formation of phase-HAPY, in assemblage with garnet and spinel, at pressures above chlorite breakdown, provides a viable mechanism to promote H₂O transport in metasomatized ultramafic mélanges of subduction channels.

Introduction

Although chromium is a minor constituent of the Earth mantle, it is incorporated in almost all major mantle phases, i.e., spinels, garnets, and pyroxenes, and it is expected to substantially modify phase equilibria in both mafic and ultramafic rocks. Natural spinel or garnet lherzolites contain 2,600 ppm chromium on average (O’Neill and Palme 1998); pyroxenites can host a variable amount of chromium, from up to 3,400 ppm on average of chromium in the pyroxenites of Beni Boussera (Pearson et al. 1993; Kumar et al. 1996) to 2,400 ppm in Balmuccia pyroxenites (Mukasa and Shervais 1999; Sinigoi et al. 1983; Voshage et al. 1988). In the mantle, at moderate pressures, chromium is mainly hosted in spinel, containing 10–32 wt% of Cr₂O₃, followed by 3–5 wt% in garnet, 0.2–0.6 wt% in orthopyroxene, and up to 0.02 wt% in olivine (Hervig and Smith 1982). At higher pressure, garnet becomes one of the major hosts of chromium, as testified by the occurrence of garnet with up to 22 wt% Cr₂O₃ (up to 65 % mole fraction of knorringite—Mg₃Cr₂Si₃O₁₂) in deep mantle xenoliths (Stachel and Harris 2008) and in inclusions in diamonds (Sobolev et al. 2004; Meyer 1975).

Chromium has also a strong influence on the spinel to garnet transition, usually taken as reference for the pressure evolution of mantle rocks (Klemme 2004).

The simple system MgO–Al₂O₃–SiO₂–H₂O (MASH) has been widely used not only to model hydrated mantle rocks (Ulmer and Trommsdorff 1999), but also to investigate a variable range of lithologies at the slab–mantle interface, where mass transfer in mélange zones generates hybrid rocks. Particular emphasis has been given to the role of chlorite-rich schists (Marschall and Schumacher 2012) and therefore to the Al-rich compositional space in the system MASH. Although chlorite is the most important hydrate mineral in this system, a new Hydrous Al-bearing PYroxene-like phase (phase-HAPY—Mg_2.1Al_0.9(OH)₂Al_0.9Si_1.1O₆, Gemmi et al. 2011) has been recently recognized at pressures beyond clinochlore stability and found to be relevant to the transport of water to depth exceeding 150 km in locally Al-enriched hybrid rocks in subduction environments (Bebout 2007; Spandler et al. 2008). Nonetheless, this Al-rich and H₂O-rich part of the system remains only poorly investigated.

Phase equilibria in the system MASH are expected to be strongly modified, due to the uneven fractionation of Cr³⁺ among major mantle phases (Chatterjee and Terhart 1985). Experiments so far investigated the effect of chromium in dry systems, carefully focusing on high-temperature partitioning between spinel and garnet (Klemme and O’Neill 2000) or dealing with the effect of chromium on Al solubility in orthopyroxene (Klemme 2004). The role of hydrates, additional candidates for hosting Cr into their structure in “metasomatized” ultramafics, has not been evaluated yet.

The stability of chlorite in the model system MASH is critical in the modelling of the hydration/dehydration sequence in “altered” ultramafic rocks, which include both serpentinized peridotites and metasomatized either on the ocean floor or along outer-rise normal faults at subduction trenches (Peacock 2001). Other hybrid rocks occur in various subduction settings, both at the top of the slab (Spandler et al. 2008; Bebout 2007) and in the wedge (Smith 2010). In hydrated ultramafic systems, chlorite (containing 13 wt% of H₂O) is a good candidate for transferring water beyond the stability field of antigorite, being stable up to 4–6 GPa, 700–800 °C. Chromium solubility in chlorite may further extend its stability and likely influence phase relations. Chromian chlorites of hydrothermal origin (Erzincan Chlorite, Turkey) have been known to contain up to 9.3 wt% Cr₂O₃ (Brown and Bailey 1963; Welch and Crichton 2005). High-pressure mantle chlorite, found in natural peridotites, hosts up to 2.0 wt% Cr₂O₃ (Ravna 2006); intermediate contents of Cr₂O₃ (up to 5–6 wt%) have been reported for chromian chlorite found in veins in association with chromite deposits and ultramafics (e.g., Niuggihalli Schist Belt, India—5.18 wt% Cr₂O₃). Grove et al. (2006) and Till et al. (2012) synthesized chlorites containing up to 1.47 wt% Cr₂O₃ at 3.6 GPa, 800 °C, in a bulk composition representing a model primitive mantle peridotite.

Phase equilibria studies in peridotites modelled in the Cr-free system Na₂O–CaO–FeO–MASH (Fumagalli and Poli 2005) reveal that, as a results of preferential partitioning of Fe into garnet, chlorite breaks down to garnet + olivine + pyroxenes at approx. 780 °C and at 3 GPa. In experiments performed in fluid-saturated Cr-bearing peridotite systems, Grove et al. (2006) reported the assemblage olivine + orthopyroxene + clinopyroxene + garnet + chlorite + ilmenite + “glass” at 840 °C and 2.8 GPa.

The goals of this work are (1) to evaluate the influence of Cr on the high-temperature stability of chlorite at high pressure; (2) to estimate the Cr partitioning between chlorite and coexisting anhydrous phases, and, thereby, offer further constraints on the pyrope + forsterite = spinel + enstatite divariant loop; and (3) to explore phases that form as a consequence of chlorite breakdown at high pressure.

Phase equilibria in the MASH system: previous experimental constraints

Phase relations in the system MASH relevant for hydrated mantle bulk compositions have been reviewed by Ulmer and Trommsdorff (1999) based on available experimental data. Figure 1 shows phase equilibria in the system MASH for pressures higher than 2 GPa. Phases involved in the four invariant points are forsterite, enstatite, spinel, pyrope, diaspore/corundum, and the two hydrous silicates chlorite and Mg-sursassite. Invariant point I involves spinel, enstatite, forsterite, pyrope, and chlorite, with four main reactions. The first represents the extensively studied spinel to garnet transition (the high-temperature assemblages are on the right side of the equations)

$${\text{pyrope}} + {\text{forsterite}} = {\text{spinel}} + {\text{enstatite}}$$

(1)

The other three control the thermal stability of chlorite and are the terminal reaction in the spinel stability field

$${\text{chlorite}} = {\text{spinel}} + {\text{enstatite}} + {\text{forsterite}} + {\text{H}}_{2} {\text{O}},$$

(2)

occurring within the stability field of spinel, extensively investigated and constrained by Staudigel and Schreyer (1977), Jenkins (1981), and Jenkins and Chernosky (1986); the terminal reaction within the garnet stability field

$$2\,{\text{chlorite}} = 3\,{\text{forsterite}} + 1\,{\text{pyrope}} + 1\,{\text{spinel}} + 8\,{\text{H}}_{2} {\text{O}}$$

(3)

investigated by Staudigel and Schreyer (1977) and Fockenberg (1995); and the breakdown reaction for chlorite + enstatite assemblages

$$1\,{\text{chlorite}} + 1\,{\text{enstatite}} = 2\,{\text{forsterite}} + 1\,{\text{pyrope}} + 4\,{\text{H}}_{2} {\text{O}} .$$

(4)

The thermal stability of chlorite on its own composition has been suggested to be very similar to the thermal stability of chlorite + enstatite (Ulmer and Trommsdorff 1999). However, subsequent reversal experiments (Pawley 2003) shifted the location of reaction (4) toward lower temperatures, increasing the thermal difference between the two-phase boundaries.

Fig. 1

Phase relations in the MgO–Al₂O₃–SiO₂–H₂O system as reported by Ulmer and Trommsdorff (1999), experimentally constrained by Fockenberg (1995). chl clinochlore, en enstatite, fo forsterite, sp spinel, py pyrope, Mg-surs Mg-sursassite, dia diaspore, cor corundum. Star on numbers marks the location of reactions after experimental results of Pawley (2003) and Bromiley and Pawley (2002)

Reaction (3) intersects the invariant point II that involves chlorite, enstatite, pyrope, diaspore/corundum, and forsterite. The location of invariant point II is not well constrained, as it is based on just a few experiments on the stability of the forsterite + diaspore assemblage (Fockenberg 1995) controlled by

$${\text{forsterite}} + {\text{corundum}} = {\text{pyrope}} + {\text{spinel}}$$

(5)

As a result of Schreinemakers’ analysis, the reaction

$${\text{chlorite}} = {\text{forsterite}} + {\text{pyrope}} + {\text{diaspore}}/{\text{corundum}} + {\text{H}}_{2} {\text{O}}$$

(6)

has been located between invariant point II and invariant point III.

Invariant point III implies the stability of Mg-sursassite, previously known as MgMgAl-pumpellyite, Schreyer et al. (1986), Artioli et al. (1999), Fockenberg (1995), and subsequently re-named Mg-sursassite, Gottschalk et al. (2000). This stability is mutually controlled by reactions that stabilize the pyrope + H₂O join. Mg-sursassite would form as a product of the chlorite breakdown reaction

$${\text{chlorite}} = {\text{forsterite}} + {\text{Mg-sursassite}} + {\text{diaspore}} + {\text{H}}_{2} {\text{O}},$$

(7)

constrained by experiments of Fockenberg (1995) who found the assemblage forsterite, diaspore, and Mg-sursassite starting from 5.4 GPa, 700 °C, and thermally constrained by the reaction

$${\text{forsterite}} + {\text{Mg-sursassite}} = {\text{pyrope}} + {\text{diaspore}} + {\text{H}}_{2} {\text{O}} .$$

(8)

At the lowest temperature investigated, the invariant point IV involves chlorite, Mg-sursassite, forsterite, enstatite, and pyrope, and controls the location of the reaction

$${\text{forsterite}} + {\text{enstatite}} + {\text{Mg-sursassite}} = {\text{pyrope}} + {\text{H}}_{2} {\text{O}},$$

(9)

bracketed by Bromiley and Pawley (2002) and shown as reaction 9* in Fig. 1, and of the reaction

$${\text{chlorite}} + {\text{enstatite}} = {\text{forsterite}} + {\text{Mg - sursassite}} + {\text{H}}_{2} {\text{O}}.$$

(10)

At pressures higher than 7 GPa, reaction (9) has been bracketed by Fockenberg (2008) who found pyrope plus H₂O at temperatures as low as 630 °C, at 6.4 GPa, a temperature considerably lower than that determined by Bromiley and Pawley (2002).

This study is mainly aimed to evaluate the effect of Cr on the thermal stability of Cr-chlorite related to reaction (4), to draw further constraints on the reactions controlling the occurrence of other hydrous silicates beyond the chlorite stability field, and to evaluate the Cr partitioning at temperatures characteristic of subduction environments.

Experimental procedures

Starting materials

Experiments were performed in the system Cr₂O₃–MgO–Al₂O₃–SiO₂–H₂O. Starting from a clinochlore composition Mg₅Al₂Si₃O₁₀(OH)₈, three different bulk compositions were prepared: bulk A with Cr/(Cr + Al) = 0.075; bulk B with Cr/(Cr + Al) = 0.25, and bulk C with Cr/(Cr + Al) = 0.5 (Table 1; Fig. 2). Three additional bulk compositions (AT, JK and JO) were prepared to investigate poorly explored parts of the MASH system.

Table 1 Compositions of starting materials in wt% of oxides and atomic proportions

Fig. 2

Compositions (in mol%) of the starting materials investigated (see also Table 1) plotted in the system MgO–Al₂O₃–SiO₂ with H₂O in excess (MASH), as compared to bulk compositions investigated in previous studies. Most relevant phases in the MgO–SiO₂–H₂O system are also shown; chl clinochlore, en enstatite, fo forsterite, sp spinel, ta talc, 10Å 10Å phase, atg antigorite, phA phase A, py pyrope, Mg-surs Mg-sursassite, per periclase, dia diaspore, cor corundum, 11.5 11.5 Å phase, HAPY phase-HAPY—High-Aluminum PYroxene (Gemmi et al. 2011)

Bulk composition JK is close to the average Cr-bearing phase-HAPY, first detected at 5.2 GPa, 700 °C in bulk A (Gemmi et al. 2011); bulk AT lies on the join pyrope–spinel; bulk JO lies on the join pyrope–phase A, and simulates the composition of an additional phase, characterized by its first diffraction peak at 11.5 Å, revealed in bulk A at 6.0 GPa, 650 °C. To promote equilibrium at subsolidus temperatures, gels were prepared using tetraethylorthosilicate as the silica source, and pure Cr-, Mg-, and Al-nitric solutions, following the method of Hamilton and Henderson (1968). In order to promote crystal growth rather than nucleation, gels were seeded with 5 wt% of natural monoclinic clinochlore (Val Malenco, Italy; see Table 1). The seeds contain about 5 wt% FeO introducing up to 0.25 wt% of FeO in the bulk; this causes a slight deviation from the nominally Fe-free model chemical system Cr–MASH. Nonetheless, FeO is a minor component, and its influence is assumed to be negligible.

Experimental techniques

Experiments were performed at the Dipartimento di Scienze della Terra (University of Milan—UMI, Italy) at pressures ranging from 1.5 to 6.5 GPa and temperatures from 650 to 900 °C (Table 2). Pressures lower than 3.4 GPa were investigated in both single-stage (up to 1.8 GPa) and end-loaded (up to 3.4 GPa) piston cylinders using full-salt and MgO-salt assemblies, respectively. Temperature was measured by K-type and S-type thermocouples and was considered to be accurate to ±5 °C. A Walker type multi-anvil apparatus was used for higher pressures using tungsten carbide cubes of 32-mm-edge length and 17 mm truncation, and pressure cells made of prefabricated MgO–Cr₂O₃ octahedra with a 25-mm-edge length. Multi-anvil pressure calibration was performed both at room temperature, using the phase transitions Bi I–II, Bi III–V (respectively, at 2.55 and 7.7 GPa), and, at 1,000 °C, using the coesite–stishovite and the CaGeO₃ garnet–perovskite transitions at 8.7 GPa (Zhang et al. 1996) and 6.1 GPa (Suzaki et al. 1985). Pressure uncertainties were assumed to be ±3 %. Further details are given by Fumagalli and Poli (1999). Temperature was measured by S-type thermocouples and was considered to be accurate to ±20 °C without taking into account any effect of pressure on the electromotive force. Seeded gels were loaded into gold capsules of 3 mm of outer diameter. All experiments were performed at fluid-saturated conditions. About 20 wt% of distilled H₂O was introduced by a microsyringe. The capsules were welded and after quenching tested for the presence of a free fluid phase. Runs lasted up to 289 h.

Table 2 Run conditions and run products

Although the oxidation state within the capsule was not buffered directly, we assume that in all phases Cr was present in its most stable oxidation state, i.e., Cr³⁺. Furthermore, the occurrence of significant Cr²⁺ is ruled out on the basis of the lack of Cr in olivine, as discussed by Klemme and O’Neill (2000).

Run products were identified by laboratory and synchrotron X-ray powder diffraction (XRPD), performed with a PanAlytical X’Pert instrument (Dipartimento di Scienze della Terra, UMI), and at ESRF—European Synchrotron Radiation Facility (Grenoble, France) with data collected at BM08, ID09A, and ID11 beamlines (BM08: wavelength 0.6878 Å, Fuji image plate detector; ID09: wavelength 0.4142 Å, mar555 flat panel detector; ID11: 0.4200 Å, Frelon area detector). The integration of the raw 2D data was performed with the Fit2D software (Hammersley et al. 1996), and Rietveld analysis was performed with the GSAS software (Larson and Von Dreele 1994). Mineral chemistry and BSE images were collected by electron microprobe using a JEOL JXA 8200 equipped with five WDS wavelength-dispersive spectrometers (Dipartimento di Scienze della Terra, UMI). Analyses were performed at 15 kV and 5 nA; natural and synthetic minerals were used as standards. All standards were calibrated within 0.5 % at one standard deviation. Raw data were corrected using a Phi–Rho–Z quantitative analysis program. Representative WDS analyses of mineral phases are reported in the Supplementary Material.

Approach to equilibrium

The attainment of equilibrium in complex systems and for continuous reactions is difficult to be demonstrated. Thus, we performed synthesis unreversed experiments. However, the approach to equilibrium was evaluated for each run through the following observations: (1) the growth of compositionally homogeneous, chemically unzoned minerals, likely favored by long run durations; (2) systematic and consistent variations in mineral chemistry and coherent element partitioning at different pressure and temperature conditions; and (3) constant bulk composition as gained by mass balance calculations.

Results: phase assemblages and textures

Experimental results are summarized in Table 2. Phase proportions, where available, are reported in parentheses and have been derived by XRPD Rietveld refinement. Phase relationships in the Cr–MASH system are reported in Fig. 3.

Fig. 3

Phase assemblages in Cr–MASH (bulk A, B, and C) and MASH (bulk JK, AT, and JO) systems. Reaction numbers as in Fig. 1; F95 Fockenberg 1995, P03 Pawley 2003; opx orthopyroxene, gar garnet, guy guyanaite; other abbreviations as in Fig. 2

Cr-chlorite only occurs in bulk A (X _Cr = Cr/(Cr + Al) = 0.075, squares in Fig. 3). Chlorite coexists with orthopyroxene up to 3.5 GPa, 800–850 °C, and with pyrope, forsterite, and spinel at higher pressure (5.0 GPa, 670 °C; and 5.5 GPa, 750 °C). Chlorite-bearing assemblages at 2 GPa, 800 °C and 3.5 GPa, 850 °C appear as intergrowth of flakes of Cr-chlorite coexisting with elongated prismatic orthopyroxene (Fig. 4a). Large euhedral Cr-bearing pyrope and approx. 1–2 μm size isometric spinels and forsterite form at higher pressure (5.0 GPa, 750 °C and 5.5 GPa, 670 °C) in assemblage with Cr-chlorite (Fig. 4b). The 5-phase, Cr-chlorite–garnet–forsterite–spinel–fluid, divariant assemblage, unfeasible in the H₂O-saturated MASH system, is possible here because of the addition of a further component (i.e., Cr₂O₃).

Fig. 4

Representative back-scattered electron images of run products; a chlorite and enstatite in bulk A at 3.5 GPa, 800 °C; b chlorite, pyrope, orthopyroxene, spinel in bulk A at 5.0 GPa, 750 °C; c phase-HAPY in assemblage with pyrope, forsterite, and spinel in bulk A at 5.2 GPa, 700 °C; bulk at, d the newly synthesized 11.5 Å phase in the MASH system, bulk JO, at 6.5 GPA, 700 °C; e 11.5 Å phase in the Cr–MASH system with garnet (+guyanite + diaspore) at 6.5 GPa, 700 °C; f enstatite, forsterite, spinel assemblage in bulk C at 2.5 GPa, 800 °C

The anhydrous assemblage garnet, forsterite, and spinel was found in bulk A at 4.8 GPa, 780 °C; 4.2 GPa, 840 °C; and 3.5 GPa, 900 °C, and in the MASH system in bulk JK at 6.5 GPa, 700 °C and 6 GPa, 800 °C. In bulk A, beyond the stability field of chlorite, isometric euhedral garnets are embedded in subhedral forsterite and in up to few μm size spinels, which are homogenously and randomly distributed.

At P ≥ 5.4 GPa, a phase-HAPY with up to 3.21 wt% Cr₂O₃ was detected. It is homogenously distributed in assemblage with forsterite, pyrope, and spinel (Fig. 4c), suggesting the attainment of the Cr-clinochlore terminal reaction. X-ray powder diffraction of the multiphase assemblage containing phase-HAPY (run A10) is shown in Fig. 5a.

Fig. 5

a Low-angle portion of synchrotron X-ray powder pattern (λ = 0.4142 Å), and Rietveld refinement (circles observed intensity; lines calculated profile) of run at 5.4 and 700 °C (bulk A); HAPY phase-HAPY, ol olivine, gar garnet, dia diaspore; b Low-angle region of synchrotron X-ray powder pattern (λ = 0.4142 Å) and Rietveld refinement of the sample synthesized at 6.5 GPa and 700 °C (bulk A); labeled with stars are the diffraction peaks of the unknown 11.5 Å phase

At 6.0 GPa, 650 °C (run A7) Mg-sursassite was found in assemblage with forsterite and spinel. At 6.5 GPa and 700 °C (run A8), the XRPD indicates the presence of pyrope, forsterite, CrO(OH) guyanaite, and diaspore. In addition, in runs A7 and A8, the diffraction patterns show also a few unknown diffraction peaks, which may indicate the presence of a new phase. Synchrotron high-resolution diffraction data clearly reveal several peaks that could not be matched with known phases in the Cr–MASH system, with the first unindexed peak at 11.5 Å (Fig. 5b). This is confirmed by SEM images, which show the presence of homogenously distributed large flaky minerals, which are several tens of micrometers of diameter often arranged in fan-like aggregates (Fig. 4d, e). Runs A8 and A10 represent univariant phase assemblages at fluid saturation. However, whether the phase assemblage indicates fluid undersaturation, or the occurrence of diaspore in these runs is the result of a metastable precipitation upon quenching is unclear. Both phase-HAPY and the 11.5 Å phase were successfully reproduced in synthesis experiments in the simple MASH system. Phase-HAPY was found in bulk JK at 5.2 GPa, 700 °C in assemblage with pyrope, diaspore, and an MgO-rich fluid as testified by precipitation of brucite during quenching (Stalder et al. 2001). The 11.5 Å phase was detected together with forsterite in bulk JO at 6.5 GPa and 700 °C. A detailed crystallographic investigation of phase-HAPY is reported by Gemmi et al. (2011), and further characterizations are in progress. At 6.5 GPa, 700 °C, and at 6.0 GPa, 800 °C, pyrope, forsterite, and spinel indicate the breakdown of phase-HAPY, in agreement with the reaction in the MASH system.

$${\text{phase-HAPY}} = 7\,{\text{forsterite}} + 5\,{\text{pyrope}} + 13\,{\text{spinel}} + 20\,{\text{H}}_{2} {\text{O}}$$

Results on bulk AT further confirm the assemblage phase-HAPY, pyrope, and diaspore at 5.4 GPa, 720 °C.

No Cr-chlorite was found in bulk compositions B and C (circles); the anhydrous phase assemblage orthopyroxene, forsterite, spinel was found instead from 1.5 to 5.0 GPa and 750–850 °C. At higher pressure and lower temperatures (i.e., 6.0 GPa, 650 °C), guyanaite—CrO(OH)—was first detected by X-ray diffractometry and then analyzed at the microprobe. This phase coexisted with orthopyroxene, forsterite, and diaspore. In bulk compositions B and C, phase assemblages result in intergrowths of large elongated prismatic crystals of orthopyroxene and forsterite together with aggregates of smaller and randomly distributed spinels (Fig. 4f).

Chemical and structural properties of mineral phases synthesized in Cr–MASH system

Chlorite

Chlorite is compositionally close to clinochlore, Mg₅Al₂Si₃O₁₀(OH)₈; a slight deviation from the Cr–MASH system due to iron in chlorite crystal seeds accounts for the occurrence of up to 0.25 wt% of FeO in chlorite. In agreement with the coexistence with garnet, X _Mg in chlorite markedly increases with pressure, reaching values of <0.1 wt. FeO at 5.5 GPa, 670 °C.

The maximum Cr₂O₃ content in chlorite reaches 2.18 wt% corresponding to 0.343 a.p.f.u., at 5.5 GPa, 670 °C, P–T conditions close to the upper pressure stability (see Supplementary data). This Cr₂O₃ content is very close to the limit of 2.0 wt% reported by Lapham (1958) above which optical properties of chlorites would change (Lapham 1958; Phillips et al. 1980).

In order to gain crystal-chemical insights of chlorites, we performed structural analyses, with high-resolution synchrotron diffraction data (Fig. 6). The powder patterns of the chlorite-bearing samples are indexed by a triclinic unit cell (a = 5.3237 Å, b = 9.2215 Å, c = 14.3785 Å, α = 89.88°, β = 97.08°, γ = 89.99°, V = 700.5 Å³ at 2 GPa, 650 °C), and the diffraction intensities are well reproduced using the triclinic Cr-chlorite model (Brown and Bailey 1963; Phillips et al. 1980). The sharp hk0 and hkl peaks suggest a rather ordered stacking sequence. We noticed also several very weak diffraction peaks in the d-range 4.5–3.5 Å, which may be indexed if a 2-layer chlorite model (c ≈ 28.7 Å) is adopted. However, the attempt to build a 2-layer structural model resulted in a poor matching of diffraction intensities. Moreover, if a triclinic symmetry is assumed, as indicated by the high-resolution synchrotron angular position, the 003 and 005 diffractions (present in the 10–5.5 Å d-range, non-overlapping with any other diffraction lines) would present a nonzero intensity, but this is not the case. A reasonable explanation would be the presence of a second chlorite phase with 2-layer periodicity. Refinements with two structural models, the triclinic Cr-bearing one, and a monoclinic 2-layer clinochlore, result in a phase fraction of 93 % of triclinic polymorph and 7 % of monoclinic one. This phase fraction indicates the possible presence of unreacted clinochlore seeds, even if a complex intergrowth of monoclinic and triclinic chlorite cannot be excluded. The predominance of one-layer triclinic chlorite, however, allowed a structural refinement, which confirmed the presence of Cr³⁺ in the octahedral interlayer. In fact, the unconstrained refinement of occupancies for the four octahedral sites converged with higher electron density in octahedral brucitic interlayer sites.

Fig. 6

Rietveld refinement of Cr-chlorite, using the structural model of Brown and Bailey (1963). Synchrotron X-ray diffraction, λ = 0.4142 Å

As a result, chlorite mineral chemistry is described in terms of two exchange vectors: Mg₋₁AlSi₋₁Al (Tschermak component, active in the 2:1 layer) and CrAl₋₁ (active in the brucite layer, preferentially in the M4 site). In Fig. 7a, the relation between Mg + Fe and Al + Cr a.p.f.u. is shown. The Tschermak component, represented by the arrow pointing toward the amesite end member (Mg₄Al₄Si₂O₁₀(OH)₈), decreases with pressure and increases with temperature. This observation is in agreement with the experimental results obtained by Pawley (2003, her Fig. 2), as predicted by the model of Baker and Holland (1996) along the univariant reaction (4). In Fig. 7b, the effect of the CrAl₋₁ exchange is plotted and compared with available experimental data in complex peridotite systems and with natural occurrences (Brown and Bailey 1963; Phillips et al. 1980; and references therein). The diagram shows that chlorite in equilibrium with a mantle assemblage has a limited content of Cr₂O₃, that is below 2 wt%.

Fig. 7

Mineral chemistry of chlorite: a (Mg + Fe) versus (Al + Cr) in a.p.f.u. The arrow from clinochlore toward amesite component represents the Tschermak exchange vector; b Cr versus Al diagram of experimental results of this study as compared to natural occurrences and previous experimental results in complex peridotite systems. Natural occurrences include chlorite found in mantle rocks, i.e., dunites, in intergrowths of olivine and chromite, e.g., Day Book Body, North Carolina), and from hydrothermal environments, i.e., vugs, veins, and other openings in serpentine associated with chromite ore (e.g., Erzincan, Turkey; Niggihalli Belt shists, India)

Orthopyroxene

Orthopyroxenes, normalized on the basis of 6 oxygens, are Mg₂Si₂O₆ enstatites with a small component of Al and Cr (see Supplementary data). Orthopyroxene accepts Al through the substitution of the Tschermak component (MgAl₂SiO₆); Cr is incorporated via CrAl₋₁ in the octahedral sites, according to crystal field theory and simple crystal-chemical considerations (Klemme and O’Neill 2000; Canil and O’Neill 1996, Burns 1993). The results of this experimental study have been plotted in the diagram Al versus Cr (Fig. 8) and are compared with previous experimental results (Klemme and O’Neill 2000; Brey et al. 1999). Present data are consistent in terms of Al/Cr ratio and comparable as a function of bulk X _Cr with previous results.

Fig. 8

Mineral chemistry of orthopyroxene. Al versus Cr in a.p.f.u. compared with previous experimental studies

Spinel and pyrope

Spinels are solid solutions between magnesiochromite (MgCr₂O₄) and spinel s.s. (MgAl₂O₄), X _Cr (Cr/(Cr + Al)) describing their mineral chemistry. Overall, Cr decreases with increasing pressure and temperature. However, different bulk compositions and, above all, different coexisting phases strongly affect the composition of spinels (see Supplementary data). In bulk compositions B and C, where spinel coexists with enstatite, X _Cr reveals a strong Cr fractionation into spinel, only limited variations as a function of P and T are observed; in bulk A, where spinel coexists with garnet, large variations occur as a function of P and T.

Pyrope contains very low knorringite component, with Cr/(Cr + Al) from 0.05 to 0.03 even at the highest temperature investigated (see Supplementary data). Knorringite component in garnets is markedly lower than in both experimental studies at higher temperatures and natural occurrences in ultramafic xenoliths or natural diamonds form kimberlites (Nixon and Hornung 1968; Stachel and Harris 2008).

Mg-sursassite, phase-HAPY, and 11.5 Å phase

Chemical analyses of Mg-sursassite are normalized to 16 cations and 28 oxygens, according to the formula Mg₅Al₅Si₆O₂₁(OH)₇. Following Gottschalk et al. (2000), who re-named MgMgAl-pumpellyite first synthesized by Schreyer et al. (1986) and further investigated by Fockenberg (1998) and Artioli et al. (1999), Mg-sursassite is characterized by the isostructural phase Mn-sursassite (Mn₄Al₂Al₄Si₆O₂₂(OH)₆), in which not only Mg substitutes for Mn but also Al is replaced by Mg + H. Previous results, however, also suggest a continuously variable composition with Mg:Al:Si ratios ranging from 5.28:5.02:5.81 (Bromiley and Pawley 2002) to 4.87:5.17:5.93 (Fockenberg 1998). The addition of Cr might imply the additional substitution Al = Cr assuming all Cr to occur as Cr³⁺. As a result, Mg-sursassite is expected to have a complex solid solution. Chemical analysis of the Mg-sursassite of the present study (Mg_5.5Al_4.4Cr_0.2Si_5.8 O₂₁(OH)₇) reveals a ratio Mg:(Al + Cr):Si = 5.55:4.61:5.82, enriched in Mg and depleted in Al + Cr. Cr has previously never been reported in the crystal structure of Mg-sursassite. Present results suggest that at 6 GPa, 650 °C Mg-sursassite can host up to 1.9 wt% of Cr₂O₃ (0.22 Cr a.p.f.u.). However, chemical data have large uncertainties due to the analytical difficulties in fine-grained samples, and further experiments are required to better characterize the crystal chemistry of Mg-sursassite in the presence of chromium.

Phase-HAPY (Mg_2.1Al_0.9(OH)₂ Al_0.9Si_1.1O₆) has been very recently investigated and its structure defined on the basis of run products obtained in the system MASH. Gemmi et al. (2011) determined its structure by the automated electron diffraction tomography method and refined its structure by synchrotron X-ray powder diffraction. Phase-HAPY is a single-chain hydrous silicate, with a topology closely related to pyroxenes. The difference consists in the presence of an additional cation site in the octahedral layer, which expands the structure perpendicular to the alignment of the tetrahedral chains and, consequently, a negative charge balanced by the presence of hydrogen in OH groups (Gemmi et al. 2011). Likely substitutions expected in phase-HAPY further indicate that this phase might be of relevance in a variety of ultramafic bulk compositions. In the Cr-bearing system, phase-HAPY has been identified by X-ray powder diffraction. Microprobe analysis shows that phase-HAPY can host from 3.21 (5.2 GPa, 700 °C) to 2.67 (5.4 GPa, 720 °C) wt% of Cr₂O₃, values that translate, after a normalization based on 5 cations and 7 oxygens, to 0.11–0.09 a.p.f.u. (see Supplementary data). As compared to the previously defined hydrate (Gemmi et al. 2011) found in the MASH system, Cr-bearing phase-HAPY—Mg_2.2Al_1.5Cr_0.1Si_1.1O₆(OH)₂—is enriched in Mg, with 2.2 compared with 2.1 a.p.f.u., while Al is lower, with 1.5 compared to 1.8 a.p.f.u..

The chemical analysis of the 11.5 Å phase suggests a quite homogeneous mineral chemistry both in the Cr–MASH and in the MASH system. Cr₂O₃ ranges from 1.76 to 1.87 wt% at 6 GPa, 650 °C and 6.5 GPa, 700 °C, respectively. Totals suggest a particularly large amount of water, estimated to be 14–15wt%. Microprobe analysis suggests a very reproducible mineral chemistry with MgO = 46.5 wt%, Al₂O₃ = 10.8 wt%, SiO₂ = 26.3 wt%, and Cr₂O₃ = 1.76 wt%, at 6.0 GPa, 650 °C; and MgO = 46.3 wt%, Al₂O₃ = 11.0 wt%, SiO₂ = 26.8 wt%, and Cr₂O₃ = 1.87 wt%, at 6.5 GPa, 700 °C, corresponding approximately to a composition with Mg/Al/Si ratios of 6.3:1.2:2.4.

Discussion

The effect of Cr on phase equilibria as compared to the simple MASH system

The addition of Cr into the simple MASH system adds new constraints on phase equilibria relevant for modeling transformations in a hydrated lithospheric mantle and in Al-rich compositions of subduction mélanges. No additional Cr-bearing phases were found within the stability field of chlorite, indicating that phase equilibria are mostly affected by continuous reactions. Univariant reactions in the system MASH change to divariant fields in the more complex Cr-bearing system. All phases can host Cr into their structure although to different extents. The P–T location and the width of divariant fields are strongly related to the Cr partitioning among all phases. Present results suggest that for bulk A (X _Cr = 0.075), Cr–Al partitioning follows the order: spinel > orthopyroxene > chlorite ≥ phase-HAPY > garnet. In particular, at 5 GPa, 750 °C (bulk with X _Cr = 0.075) equilibrium values are X ^spinel_Cr  = 0.27, X ^chlorite_Cr  = 0.08, X ^garnet_Cr  = 0.05; at 5.4 GPa, 720 °C the values are X ^spinel_Cr  = 0.33, X ^HAPY_Cr  = 0.06, and X ^garnet_Cr  = 0.04; and at 3.5 GPa, 850 °C X ^opx_Cr  = 0.12 and X ^chlorite_Cr  = 0.07.

Cr solubility in chlorite is limited to less than 2 wt%, and its effect is to shift reaction (4) to slightly higher temperature, e.g., between 850 and 900 °C, at 3.5 GPa, as compared to 830 °C as suggested by Pawley (2003) for the system MASH (Fig. 3). Therefore, the divariant field representing the transformation of chlorite + enstatite to the anhydrous assemblage garnet, spinel, and forsterite is limited in temperature. The presence of Cr strongly favors the stability of spinel, which is an almost ubiquitous phase. Spinel coexists with garnet, forsterite, and chlorite up to 5.5 GPa, 670 °C and 5 GPa, 700 °C, suggesting that chlorite persists to slightly higher pressure compared with the simple system MASH (see Fig. 1).

Experimental results obtained here were used to complete a chemographic analysis projecting coexisting phases from H₂O, i.e., fluid-saturated conditions in the system MgO–Al₂O₃–SiO₂–Cr₂O₃ (Fig. 9). From 2 to 3.5 GPa and 800 to 850 °C, the chlorite + orthopyroxene join is stable; therefore, as a function of bulk compositions, the 4-phase assemblages chlorite–orthopyroxene–spinel–forsterite and chlorite–orthopyroxene–spinel–garnet are possible. At 5.5 GPa and 670 °C, we observe the assemblage chlorite–garnet–spinel–forsterite (blue tetrahedron in Fig. 9) which implies an overstepping of the reaction

$${\text{chlorite}} + {\text{orthopyroxene}} = {\text{garnet}} + {\text{spinel}} + {\text{forsterite}} + {\text{H}}_{2} {\text{O}},$$

the latter assemblage is represented by the light-blue plane in Fig. 9.

Fig. 9

Chemographic analysis of coexisting phases in the system Cr–MASH at fluid-saturated conditions. See text for further details

Despite the addition of Cr₂O₃ to the model system MASH, the terminal reaction that indicates the ultimate breakdown of chlorite appears still related to the reaction (3),

$${\text{chlorite}} = {\text{forsterite}} + {\text{garnet}} + {\text{spinel}} + {\text{H}}_{2} {\text{O}},$$

as a result of compositional degeneracy between garnet, olivine, chlorite, and spinel as shown by the yellow plane in Fig. 9. The stoichiometry of this reaction is consequently affected by chemical variations of the phases involved with P and T.

The stability of Cr-bearing chlorite is not fully constrained and based only on synthesis experiments. Nonetheless, our results show that the stability of chlorite is driven by the disappearance of the chlorite + orthopyroxene assemblage (of relevance for mantle rock compositions). The stability of chlorite on its own bulk composition is driven by the terminal reaction that is very close in P–T space. A more rigorous prediction of the pressure temperature stability of the 4-phase assemblage chlorite–garnet–spinel–forsterite prior to the terminal reaction of the hydrous phase would be strongly dependent on the Cr–Al partitioning of all phases involved and on solid solution models, still needing to be comprehensively investigated, especially for chlorite solid solutions.

The role of the newly synthesized phase-HAPY: implications for high-pressure phase equilibria in MASH and Cr–MASH systems

Phase relationships in the high-pressure region of the P–T space investigated here are strongly affected by the occurrence of the newly synthesized phase-HAPY. Figure 10 shows a possible topology of the reactions in the model system MASH, with emphasis on the location of invariant points involving phase-HAPY. Schreinemakers’ analysis is consistent also with experimental results in the Cr-bearing system: colored phase fields in Fig. 10 represent the stable assemblage in the Cr–MASH system.

Fig. 10

Schreinemakers’ analysis showing the possible topology of phase relations in the MASH system involving phase-HAPY. Colored phase fields in chemographies represent experimental assemblages found in the Cr–MASH system at the indicated P–T conditions with which the topology is consistent

With respect to phase equilibria usually taken as a reference frame in the MASH system (Fig. 1), invariant point IV is slightly modified here to take into account more recent experimental constraints of Bromiley and Pawley (2002). Invariant points II and III should be reconsidered in the light of the addition of phase-HAPY. Three new invariant points A, B, and C are required to reconcile the new experimental results presented here. They are expected to be stable within a restricted range of pressure and temperature, approaching invariant point I. The invariant point A (chlorite, forsterite, pyrope, spinel, and phase-HAPY) is generated by the intersection of chlorite terminal reaction (3) with the stability field of phase-HAPY. The invariant point B describes the relations involving Mg-sursassite, as generated by the reaction forsterite + Mg-sursassite = chlorite + pyrope (B3), departing from invariant point IV, and, again, the stability field of phase-HAPY. Invariant point C, only partially constrained experimentally, is required in order to predict the assemblage forsterite + diaspore observed in the present study at 6.5 GPa, 700 °C although this implies that reaction (5) (see Fig. 1), marking the stability of forsterite + diaspore/corundum assemblages, should be shifted toward lower temperature with respect to previous reports by Fockenberg (1995). The Schreinemakers’ analysis combined with the present data suggests that beyond the stability of chlorite, a restricted field of stability of phase-HAPY is found. The appearance of phase-HAPY above the stability of chlorite (Gemmi et al. 2011) is controlled, in the MASH system, by the reaction

$$13\,{\text{chlorite}} = 16\,{\text{forsterite}} + 4\,{\text{pyrope}} + 10\,{\text{phase-HAPY}} + 55\,{\text{H}}_{2} {\text{O}}\, ( {\text{A}}2 ).$$

The thermal stability of phase-HAPY is governed, within the stability field of forsterite + diaspore join, by the reaction

$$30\,{\text{phase}}\,{\text{HAPY}} = 30\,{\text{forsterite}} + 1\,{\text{pyrope}} + 52\,{\text{diaspore}} + 4{\text{H}}_{2} {\text{O}}\,({\text{C}}1),$$

and at higher temperatures, by the reaction

$$20\,{\text{phase-HAPY}} = 7\,{\text{forsterite}} + 5\,{\text{pyrope}} + 13\,{\text{spinel}} + 20\,{\text{H}}_{2} {\text{O}}\,({\text{A}}1).$$

Cr–Al partitioning and the spinel- to garnet-peridotite transition: low-temperature constraints

Our results provide additional information on the spinel- to garnet-peridotite transition in the mantle, adding important constraints at low temperatures. Indeed, although the effect of Cr on the spinel- to garnet-peridotite transition has been extensively investigated both experimentally and theoretically by thermodynamic calculations (see Klemme 2004 and references therein, Ziberna et al. 2013), its pressure at relatively low temperatures is still poorly constrained. Klemme (2004) investigated the reaction spinel + orthopyroxene = garnet + olivine in the Cr–MASH system from 1,200 to 1,600 °C over a range of pressure between 4.5 and 16 GPa and confirmed conclusions reached by Webb and Wood (1986): in fertile compositions, i.e., in bulk compositions with 0 < X _Cr < 0.2, the divariant field where spinel and garnet coexist is narrow. In more depleted bulk compositions, i.e., for X _Cr > 0.2, the effect of Cr becomes of greater importance. Cr–Al partitioning data between spinel and garnet are shown in Fig. 11 and compared with the loop for 1,100 °C reported by Webb and Wood (1986) and Klemme (2004). At low temperatures, the spinel- to garnet-peridotite divariant field is largely displaced to higher pressure, because of the relatively low chromium content present in both spinel and garnet. As a result, we expect that the spinel phase boundary should have a markedly negative P–T slope. Although Klemme (2004) defined experimentally a negative slope in the Al-free system MgO–Cr₂O₃–SiO₂ (MCR), consistent with results derived by thermodynamic extrapolations by Doroshev et al. (1997) and Girnis and Brey (1999), a positive slope is predicted in the MASH system. On the other hand, Niida and Green (1999) in a synthetic model “MORB-lherzolite” with X _Cr = 0.065 found that the transformation from spinel- to garnet-bearing assemblages occurred at approx. 2 GPa between 925 and 1,050 °C with a slightly negative slope for this transition. A negative dP/dT slope for the spinel- to garnet-peridotite transition at low temperatures has been envisaged also by Wood and Yuen (1983) and further discussed by Webb and Wood (1986) to account for the observed uplift of old oceanic lithosphere as a result of a lowering of density due to garnet transforming into spinel on cooling at temperature of less than 900 °C (i.e., on aging of the lithosphere). Furthermore, in agreement with experimental results of this study, Ziberna et al. (2013) have recently evaluated the effect of bulk composition on the spinel- to garnet-peridotite transition and suggested that the depth and the width of the transition in a multi-component system are strongly related to the thermal state of the lithosphere. In the present study, the negative P–T slope is obvious and was calculated to be 0.5 GPa/100 °C, similar to the value given by Klemme (2004) in the Al-free MgO–Cr₂O₃–SiO₂ system. However, the dependence of the dP/dT slope on knorringite and chromite components in garnet and spinel, and its dependence on temperature, remains to be resolved experimentally.

Fig. 11

X _Cr versus pressure in experimental spinels and garnets. Present results are compared with high-temperature loop (1,100 °C) reported by Webb and Wood (1986). Triangles refer to X _Cr as derived by X-ray diffraction refinements of lattice parameter a and applying the equation: a (Å) = 8.086 + 0.25 X _Cr spinel (Webb and Wood 1986)

Phase assemblages in subduction mélanges and H₂O transfer to deep mantle

We use the phase boundaries from spinel peridotite, to spinel–garnet peridotite and to garnet peridotite, as modeled by Ziberna et al. 2013 for two bulk compositions (gray shaded field in Fig. 12, bounded by X _Cr = 0.04 and X _Cr = 0.32), and combine these data with hydrous phase stabilities as determined in this study (Fig. 12) to infer phase assemblages and H₂O transfer in a subduction zone mélange. Phase equilibria are compared with a schematic subarc P–T profile (dashed–dotted line in Fig. 12) retrieved from numerical modelling of subduction zones (Gerya and Meilick 2011) and with P–T paths for slab surfaces (orange area), at variable mechanical decoupling settings as modeled by Arcay et al. (2007).

Fig. 12

Schematic phase stability field of chlorite and phase-HAPY are compared with the vertical pressure–temperature profile (dashed–dotted line) for a subduction zone beneath the volcanic arc after numerical modeling of Gerya and Meilick (2011) and with P–T paths for the slab surface (orange area) from Arcay et al. (2007); different paths take into account different water-weakening effect on mantle rocks; solid line represents a model reference for a “dry” rock; dashed line refers to a moderate rock strength reduction due to water, while narrow dashed line to a relevant strength reduction. The stability field of chlorite at low pressure is taken from Ulmer and Trommsdorff (1999). Schematic fields for spinel-peridotite, spinel + garnet-peridotite and garnet-peridotite are inferred from thermodynamic modeling of Ziberna et al. 2013 consistent with the results of this study (gray lines)

Although the mantle wedge is entirely in the garnet stability field, spinel is expected to be the major Cr-bearing phase within the subduction mélange at the slab–mantle interface, both in the chlorite and in phase-HAPY stability fields. Spinel in subduction mélanges should, therefore, not be necessarily regarded as a retrogression product of high-pressure assemblages, but we expect spinel to be a prograde phase coexisting with hydrous phases at subarc depth.

Weighted mass balance calculations were performed in order to constrain the water budget stored in phase-HAPY-bearing assemblages. Note that the H₂O component and the fluid phase have not been used in the calculations, and Monte Carlo simulations have been employed for error propagation (Fumagalli et al. 2009). At 5.4 GPa, 720 °C, bulk composition A gives the following phase abundances: phase-HAPY = 37.2 wt% (±5.5), garnet = 28.3 wt% (±5.8), olivine = 32.2 wt% (±2.1), and spinel = 2.3 wt% (±0.6). Similar results have been obtained at 5.4 GPa, 700 °C. Representative P–T paths for a slab–mantle interface show that chlorite schists in subduction mélange move from the chlorite field to a phase-HAPY + olivine + spinel + garnet field, liberating approximately 10.5 wt% H₂O at chlorite breakdown, but still conserving 2.6 (±0.4) wt% H₂O, therefore possibly promoting a link to the so-called alphabet phases, notably phase A, above 6 GPa (Ulmer and Trommsdorff 1999).

Conclusions

Experimental results in the simple Cr–MASH system of the present study reveal that the Cr solubility in chlorite in equilibrium with a mantle assemblage is limited to 2.2 wt% Cr₂O₃. Nonetheless, as a result of addition of Cr, the stability field of chlorite is enhanced by approximately 50 °C per 0.5 GPa as compared to the MASH system. Cr–Al partitioning between spinel and garnet suggests a negative dP/dT slope (approx. 0.5 GPa/100 °C) for the spinel- to garnet-peridotite transition at T = 670–900 °C. At high pressures, and T ≤ 750 °C the breakdown reactions of chlorite involve the formation of phase-HAPY, a hydrous Al-bearing pyroxene-like structure (Gemmi et al. 2011) synthesized here for the first time in the Cr–MASH system. We find a new hydrous phase that coexists with pyrope and spinel at pressures exceeding the chlorite stability. This may be important for the storage of H₂O in the mantle and for transport of H₂O in subduction zones.