Subsolidus phase relations in a mantle pyroxenite: an experimental study from 0.7 to 1.5 GPa

Pyroxenites are considered as diffuse heterogeneities in the upper mantle (e.g., Bodinier & Godard, 2014). Despite being volumetrically subordinated with respect to peridotites, pyroxenites play a key role in mantle melting processes (e.g., Hirschmann & Stolper, 1996; Phipps Morgan, 2001; Shorttle & Maclennan, 2011; Lambart et al., 2016) and in mantle rheology (e.g., Hidas et al., 2013; Henry et al., 2017). Indeed, they have been invoked as diffuse components in mantle sources of basalts in several magmatic environments (e.g., Sobolev et al., 2005, 2007; Lambart et al., 2013) and recognized as catalyst for lithosphere softening (Hidas et al., 2013).

Pyroxene-rich veins or layers embedded in mantle peridotites commonly occur in ophiolitic and orogenic ultramafic massifs (e.g., Bodinier & Godard, 2014) and in mantle xenoliths (e.g., Gonzaga et al., 2010; Aulbach & Jacob, 2016) and their origin has been related to high-pressure magmatic segregation, metamorphism and melt-rock reaction processes. Some of these pyroxenites represent long-lived deep mantle heterogeneity that experienced the same metamorphic evolution as the host peridotites at lithospheric mantle levels. However, pyroxenites are expected to develop sensibly different phase assemblages at fixed pressure (P) and temperature (T) conditions as a result of different bulk composition (e.g., Schmädicke, 2000). In particular, the P–T stability of aluminous phases is strongly affected by major-element composition, such as Al₂O₃, Cr₂O₃, CaO and Na₂O contents. Experimental studies have suggested that garnet appears at significantly lower pressure in pyroxenites than in peridotites (e.g., Irving, 1974; Adam et al., 1992), explaining the widespread occurrence of garnet-pyroxenites layers in spinel-bearing peridotites (e.g., Bodinier et al., 1987a, b; Garrido & Bodinier, 1999; Takazawa et al., 1999; Morishita & Arai, 2001; Montanini et al., 2006, 2012; Van et al., 2010; Gysi et al., 2011, Montanini & Tribuzio, 2015).

Although subsolidus phase relations in peridotites have been experimentally investigated in both simplified and complex chemical systems (e.g., Fumagalli & Klemme, 2015), very few experimental works have been focused on subsolidus phase equilibria in pyroxenites and the stability of aluminous phases is still only barely known (e.g., Irving, 1974; Adam et al., 1992).

In this paper, we present and discuss the results of subsolidus experiments performed on a natural secondary-type pyroxenite at pressure from 0.7 to 1.5 GPa and 1100–1250 °C. Specific aim is to investigate the stability of plagioclase, spinel and garnet in this pyroxenite bulk in order to provide useful geobarometric information for the subsolidus evolution of exhumed ultramafic mantle sectors.

As starting material we selected an olivine-websterite (GV10), sampled in the ophiolitic mantle sequences from the External Liguride Unit (EL, Northern Apennines, Italy; Borghini et al., 2013, 2016). Pyroxenites from this mantle section occur as cm-thick layers subparallel to the mantle foliation of host peridotite (Rampone et al., 1995; Borghini et al., 2013, 2016). Field, microstructural and geochemical observations indicated that they originated by pyroxenite-melt-peridotite interactions, thus supporting its secondary origin (Lambart et al., 2012, 2013). Trace-element compositions strongly suggest that many of these pyroxenite layers contained garnet in the primary mineral assemblage, thus indicating crystallization at relatively deep mantle level (Borghini et al., 2016). After the pyroxenite emplacement, inferred to have occurred at Ordovician ages (Borghini et al., 2013), this veined mantle sector experienced a subsolidus spinel- to plagioclase-facies decompressional tectonic evolution, related to its Jurassic exhumation. This is testified by the development of plagioclase-bearing neoblastic assemblages in both peridotite and pyroxenite (Borghini et al., 2011, 2016). Pyroxenite GV10 well preserves microstructural and geochemical features related to its deep melt-rock reaction origin and, among the EL pyroxenites, it better records chemical imprinting inherited by a garnet-bearing primary assemblage (Borghini et al., 2016).

The major-element bulk-rock composition of GV10 and other pyroxenite bulk compositions investigated in previous subsolidus experiments are reported in Table 1. The melting relations of GV10 at 1.0 and 1.5 GPa have been recently investigated (Borghini et al., 2017), thus the phase relations at solidus conditions are fully constrained.

A glass has been prepared by complete melting of rock powder at 1500 °C in a gas mixing vertical furnace operating at fayalite–magnetite–quartz (FMQ) fO₂ and quenched in dry ice (Borghini et al., 2017). To promote the nucleation of the minor phases in subsolidus experiments, powdered glass was seeded with 1 wt% of a 1:1 mixture of synthetic pure spinel (Al₂O₃ = 71.67 wt%, MgO = 28.33 wt%) and Dora-Maira pyrope (SiO₂ = 44.71 wt%, Al₂O₃ = 25.29 wt%, MgO = 29.99 wt%), i.e. a bulk seed composition of 22.36 wt% SiO₂, 48.48 wt% Al₂O₃, 29.16 wt% MgO. The very low amount of seeds in the starting material (1 wt%) and the occurrence of large unreacted seed relics in the experiments indicate that seeds addition did not significantly affect the final bulk composition.

Experiments were performed at pressures from 0.7 to 1.5 GPa, and temperatures from 1100 °C to 1250 °C (Table 2), at the Laboratorio di Petrologia Sperimentale, Dipartimento di Scienze della Terra, University of Milano. Experiments up to 1.0 GPa were carried out in a single-stage piston cylinder; for experiments at higher pressure an end-loaded piston cylinder was used, both with MgO–Pyrex–Salt assemblies. Experiments lasted from 94 to 495 h (Table 2). Approximately 20 mg of starting material was loaded into a graphite inner capsule (outer diameter 2.8 mm), and then welded into an outer Pt capsule (outer diameter 3.0 mm, length about 7–8 mm). Graphite is used to isolate the sample from the Pt capsule and avoid Fe-loss (Kinzler, 1997; Walter, 1998). The graphite–Pt assembly, combined with pre-conditioning of the starting material at FMQ, maintained the oxygen fugacity close to the C–CO/CO₂ equilibrium (Ulmer & Luth, 1991). In order to ensure nominally anhydrous conditions, the platinum–graphite capsule with the starting material was dried in an oven at 250 °C over a night before being rapidly welded shut. The thermocouple tip was separated from the platinum capsule by a 0.5-mm thick hard alumina disc. Assemblies were kept in oven at about 200 °C for several hours before running the experiments. Temperature was measured by K-type and S-type thermocouples and is considered to be accurate to ±5 °C. According to piston-cylinder calibration, pressure uncertainties are assumed to be ±3%.

An initial pressure of 0.25 GPa was applied, then the sample was first heated to 400 °C for 10 min in order to soften the Pyrex, pressure was raised to the experimental value before reaching the desired temperature.

After experiment, capsules were enclosed in epoxy, sectioned lengthwise, polished and carbon-coated. Run products were inspected by back-scattered electron (BSE) images and microanalyses were performed using a JEOL JXA 8200 Superprobe equipped with five wavelength-dispersive spectrometers (WDS) and one energy-dispersive spectrometer (EDS) at the Dipartimento di Scienze della Terra, University of Milano. Both images and X-ray element maps were extremely useful in textural examination of the experimental charges. Analyses on mineral phases were performed using 1 μm beam size and beam conditions of 15 kV and 5 nA. Counting time was 30 s for peak and 10 s for background.

Run products and experimental conditions are reported in Table 2 and summarized in Fig. 1. Textural observations show grain size varying between 2 and 25 µm on average. As expected from the higher temperature, slightly coarser textures have been found in melt-bearing experiments (GV10-83-1 and GV10-83-9, Table 2).

A plagioclase-bearing assemblage composed of clinopyroxene, othopyroxene, olivine, plagioclase and Cr-spinel is stable from 0.7 to 0.9 GPa at 1100 °C. Pyroxenes form the larger crystals (up to 25 µm), usually showing prismatic habit (Fig. 2a). Olivine occurs as small sized crystals (≤5 µm), with a rounded habit (Fig. 2a). Plagioclase is homogeneously distributed in the run charges and occurs as subhedral crystals ranging in size from 2 to 8 µm (Fig. 2a). Cr-bearing spinel forms thin rims on large relics of spinel seeds, or occurs as small grains, around 2–3 µm, homogeneously distributed within the charge (Fig. 2a). The extremely fine grain size of olivine and spinel in experiment at 0.7 GPa and 1100 °C prevented satisfactory chemical analysis.

A spinel-bearing assemblage made of clinopyroxene, orthopyroxene, olivine and spinel is stable from 1 GPa at 1100–1180 °C, to 1.4 GPa at 1150 °C. In these experiments, clinopyroxene occurs as large grains of up to 20 µm showing triple junctions (Fig. 2b). Orthopyroxene is easily recognizable in BSE images by its dark grey contrast and shows large prismatic habit (Fig. 2b). Rounded crystals of olivine up to 10 µm in size have grey tone intermediate between clinopyroxene and orthopyroxene (Fig. 2b). Spinel is present as small rounded crystals up to 5 µm in size.

A garnet-bearing assemblage composed of clinopyroxene, orthopyroxene, olivine, garnet and spinel is stable at 1.5 GPa and 1150–1230 °C. These experiments are characterized by textures with grains size ranging between 2 and 10 µm. Garnet neoblasts occur as rounded crystals either on or far from the garnet seeds (Fig. 2c), as revealed by Ca–Al X-ray mapping (Fig. 3). Rare spinel is present in garnet-bearing experiments as very small rounded grains (1–3 µm), mostly recognized with the support of X-ray mapping (Fig. 3).

Pyroxenes display significant chemical changes as a function of pressure. Clinopyroxene has an X_Mg value [X_Mg = Mg/(Mg + Fe^tot)] between 0.83 and 0.85. In the garnet-bearing experiments (P > 1.4 GPa), systematic higher X_Mg values reflect the coexistence with garnet having much lower X_Mg. The Al content progressively increases with pressure at 1100 °C, in the plagioclase-bearing experiments and across the plagioclase-out curve (Fig. 4a). It ranges from 0.216 atom per formula unit (apfu) at 0.7 GPa, to 0.355 apfu at 1.0 GPa, and is rather constant within the spinel facies (0.351–0.377; Table 3). A slight Al decrease is also observed as garnet appears at 1.5 GPa, with Al content increasing from 1150 to 1230 °C (Table 3 and Fig. 4a). This is in agreement with results reported in experiments on peridotites in both the CaO–MgO–Al₂O₃–SiO₂ system (Obata, 1976; Herzberg, 1978; Gasparik, 1984) and more complex chemical systems (Borghini et al., 2010). The Na content in clinopyroxene increases with pressure from 0.7 to 1.0 GPa at 1100 °C, from 0.016 to 0.030 apfu, and varies between 0.025 and 0.030 apfu in spinel- and garnet-bearing experiments up to 1.5 GPa (Table 3). The Ti contents are rather variable, in the range 0.009–0.015 apfu, without any dependence on P and T. The Cr abundance is always lower than 0.006 apfu (Table 3), reflecting the very low Cr content of the bulk composition (Table 1).

Orthopyroxene presents X_Mg values from 0.84 to 0.85, with only slight variations as observed in clinopyroxene. The Al content shows a positive correlation with pressure within the stability of plagioclase, increasing from 0.179 to 0.303 apfu at 0.8 and 1 GPa, respectively. It slightly decreases across the garnet-in curve at 1150 °C (Table 4 and Fig. 4b). The Ca contents vary between 0.049 and 0.071 apfu, and Ti abundance is as low as 0.003–0.007 apfu (Table 4). As in clinopyroxene, Cr is very low (0.002–0.004 apfu).

Olivines have homogeneous compositions with X_Mg value [X_Mg = Mg/(Mg + Fe^tot)] around 0.83, and very low TiO₂ (≤0.03 wt%) and Cr₂O₃ (≤0.05 wt%; Table 5).

Reliable spinel compositions have been obtained mostly by combining WDS analysis with X-ray mapping on small rounded neoblasts isolated in the mineral matrix. In cases of very small grains, spinel composition has been derived by removing the chemical contribution of the contaminating host minerals, usually when spinel was included in clinopyroxene. All iron was considered as Fe²⁺ and Fe³⁺ was not calculated, because oxygen fugacity was not buffered. Spinel has X_Mg values between 0.63 and 0.68 without any correlation with P or T (Table 6). Significant variations of X_Cr are observed, with the highest X_Cr values in experiments within the plagioclase stability field, i.e. 0.079 at 0.8 GPa and 0.062 at 0.9 GPa (Table 6). On the contrary, X_Cr is very low in spinel-facies experiments, ranging over 0.021–0.023 apfu, and increases in garnet-bearing experiments at 1.5 GPa. Furthermore a negative correlation with temperature from 1150 to 1230 °C is also shown (Table 6). The Ti content is usually lower than 0.006 apfu (Table 6).

Despite the small grain size (usually <10 µm), we obtained several good plagioclase analyses with negligible contamination (MgO < 0.2 wt%). Plagioclase is characterized by anorthite contents [An = Ca/(Ca + Na)], which is negatively correlated with pressure (Table 7). At 1100 °C, An decreases from 0.70 to 0.61 for pressure increasing from 0.7 to 0.9 GPa, in excellent agreement with the An pressure dependence documented by subsolidus experiments on lherzolite bulks (Fig. 5; Borghini et al., 2010, 2011; Fumagalli et al., 2017).

Garnet is pyrope-rich with X_Mg values of 0.75–0.76 and Ca contents positively correlated with temperature, ranging from 0.481 apfu at 1150 °C to 0.525 apfu at 1230 °C (Table 7).

Demonstration of equilibrium through reversal experiments is difficult in case of complex chemical systems and for continuous reactions. However, the approach to equilibrium is carefully assessed through the following observations: (1) the growth of compositionally homogeneous (Fig. 3), chemically unzoned minerals, likely enhanced by long-time duration of experimental runs (Table 2), (2) systematic and consistent variations in mineral chemistry with varying P–T conditions (Figs. 4 and 5), (3) maintenance of constant bulk composition, as supported by mass-balance calculations (see below). The long duration of experiments led to well-developed textures with mineral phases homogeneously distributed in the charges. Furthermore, the behaviour of seeds can be taken into account to demonstrate a close approach to equilibrium; when not completely resorbed, seeds promoted growth rather than nucleation (Fig. 3).

Phase abundances were derived by mass-balance calculations using a weighted least-squares minimization procedure (Table 8). We included in the mass-balance calculations eight major oxides, SiO₂, TiO₂, Al₂O₃, Cr₂O₃, FeO, MgO, CaO and Na₂O. As the ferric/ferrous ratio in the run charges is unknown, all iron has been assumed as Fe²⁺, although moderate Fe³⁺ amount can be contained in spinel. Standard errors on the chemical analyses have been propagated to the phase abundances by the Monte Carlo method. Up to 10 000 simulations normally distributed around the average, value of each component of each phase have been evaluated and then statistically treated (Fumagalli et al., 2009).

This reaction is similar to that experimentally derived for mantle lherzolites (Borghini et al., 2010), but the higher plagioclase coefficient indicates that more modal plagioclase is involved in the pyroxenite, as expected by the higher bulk Al₂O₃ and CaO contents. Plagioclase modal abundance reaches 14 wt% at 0.7 GPa, 1100 °C (Table 8); this amount is more than the double of that developed at the same P–T conditions in fertile lherzolite (5–6 wt%, Borghini et al., 2010).

Additionally, relatively high bulk Al₂O₃ in the pyroxenite results in a spinel mode up to 4.7 wt% (Table 8), almost twice the spinel amount derived from experiments (2.5 wt%, Borghini et al., 2010) and thermodynamic calculations (about 2.0 wt%, Ziberna et al., 2013) on fertile lherzolite bulk compositions. Spinel mode is rather constant within the spinel-bearing experiments (4.1–4.7 wt%), with no appreciable dependence on pressure (Fig. 6).

Chromium-bearing rims on spinel seeds in garnet-bearing experiments have suggested that spinel is still stable at 1.5 GPa, together with garnet. However, quantification of spl₄ in reaction (2) is made difficult by the high uncertainty on spinel composition due to its very small size. The reaction coefficient of spl₄ obtained by balancing reaction (2) is well below 0.01 (around 0.002). Ziberna et al. (2013) argued that very low spinel modes in many garnet peridotite xenoliths result from spinel overlooking. This effect is amplified in GV10 garnet pyroxenite having Cr₂O₃ content (X_Cr = Cr/(Cr + Al) = 0.01) much lower than lherzolites (X_Cr = 0.07–0.10, Borghini et al., 2010).

Given the large chemical variability of pyroxenites, the bulk composition is relevant in depicting the phase assemblage stable at fixed P and T.

which defines the spinel to garnet facies transition (e.g., Kushiro & Yoder, 1966; O’Hara et al., 1971; Herzberg, 1978; Fumagalli & Klemme, 2015). It is well established that Cr strongly affects the location of the spinel-to-garnet transition, stabilizing spinel at higher pressure and resulting in spinel–garnet coexistence (Klemme, 2004). Similarly, the persistence of plagioclase to higher pressure has been established to be sensitive to the normative Ab/Di ratio and X_Cr of the bulk (Borghini et al., 2010).

determines the lowest possible pressure whereby garnet is stable (e.g., Kushiro & Yoder, 1966; Herzberg, 1976). In this case a pyrope-grossular garnet appears at 1.3–1.5 GPa, 1000 °C (Kushiro & Yoder, 1966). Variations in bulk X_Mg would however lead to the stability of an almandine-grossular garnet as breakdown product of olivine + plagioclase assemblages at much lower pressure, i.e. 0.7 GPa, 1000 °C (Green & Hibberson, 1970).

The present study further shows the effect of bulk composition on the stable assemblage. Phase assemblages at 1 GPa, 1000–1100 °C are plotted into the compositional space (Mg,Fe)O–CaO*–SiO₂–(Al,Cr)₂O₃ for the system Cr-FNCMAS, constructed projecting mineral compositions from the exchange vectors CaAlNa₋₁Si₋₁, MgFe₋₁ and AlCr₋₁; CaO* includes the contribution of Ca and Na through the vector CaAlNa₋₁Si₋₁.

The bulk compositions are indicated together with the amount of normative plagioclase (Pl_CIPW). Pyroxenite GV10 shows the same phase assemblage, olivine, orthopyroxene, clinpyroxene and spinel as a fertile lherzolite (FLZ). The higher normative Pl_CIPW results in a higher modal abundance of clinopyroxene (and spinel) as the bulk approaches the orthopyroxene–clinopyroxene–spinel plane. At further increase of Pl_CIPW as observed in pyroxenite DR9734 (Pl_CIPW = 39.1), the stable assemblage (orange star in Fig. 7) at the same P–T conditions is plagioclase, spinel, clinopyroxene and garnet (Adam et al., 1992). It should be noted that the stable garnet in pyroxenite DR9734 has, as expected, slightly lower X_Mg (53 mol% pyrope) and higher Ca (19 mol% grossular) with respect to garnets stabilized in GV10 pyroxenite at higher pressure (at 1.5 GPa, 1150 °C, 64 mol% pyrope, 14 mol% grossular). Its occurrence is related to reaction (4). In general, the clinopyroxene–orthopyroxene–spinel plane acts as a barrier separating olivine-bearing from olivine-free assemblages, the latter being able to stabilize associations of garnet and spinel, as observed in websterite R394 of Irving (1974), or garnet, spinel and plagioclase, as in clinopyroxenites DR9734 (Adam et al., 1992) and R392 (Irving, 1974). The GV10 pyroxenite bulk plots near this barrier and, within the spinel-facies, low-mode olivine is stabilized (approx. 5 wt%, Table 8). This confutes the lack of olivine in the spinel-bearing assemblage estimated by using the composition of the natural mineral phases (Borghini et al., 2016), indicating that, for GV10 bulk, olivine is stable at least from 0.7 to 1.5 GPa. In the following sections, we will discuss the plagioclase-out boundary, i.e. reaction (1), and the garnet-in curve, i.e. reaction (2), as derived by present experiments in relation with peridotite bulk compositions.

Defining the pressure limit of plagioclase stability in pyroxenites is useful to obtain information on geobarometric evolution of mantle rocks, because pyroxenites are commonly associated to peridotites in ultramafic massifs.

Experimental studies in the simplified system CaO–MgO–Al₂O₃–SiO₂ (CMAS) have indicated that the stability of plagioclase lherzolite is limited at pressure below 1 GPa (e.g., Kushiro & Yoder, 1966; Obata, 1976; Gasparik, 1984). Addition of Na to the CMAS system makes the plagioclase-in reaction divariant and extends the stability of plagioclase to pressures higher than 1 GPa (e.g., Walter & Presnall, 1994).

Subsolidus experiments on peridotite modelled in complex chemical systems revealed that the pressure of plagioclase-out curve is strongly influenced by the bulk Ab/Di ratio and X_Cr (e.g., Green & Hibberson, 1970; Green & Falloon, 1998; Borghini et al., 2010). Higher bulk Ab/Di ratio leads to crystallization of a more albitic plagioclase, expanding the plagioclase stability towards higher pressure (e.g., Green & Hibberson, 1970; Walter & Presnall, 1994; Green & Falloon, 1998; Falloon et al., 1999; Borghini et al., 2010; Till et al., 2012; Laporte et al., 2014; Fumagalli et al., 2017). At near-solidus temperatures, plagioclase persists up to 1.3 GPa in fertile lherzolites (e.g., Falloon et al., 1999; Laporte et al., 2014). Moreover, the plagioclase-out boundary is also sensitive to the bulk X_Cr (or the chromite/anorthite normative ratio), which acts in favour of spinel expanding its stability toward lower pressure at the expense of plagioclase (Borghini et al., 2010).

In Fig. 8, the plagioclase-out boundary derived for pyroxenite GV10 is compared with equilibria determined for different peridotite bulks in complex chemical systems, at the same temperatures. The high-pressure limit of plagioclase stability in GV10 occurs within the pressure range of the boundaries determined for these lherzolites. In particular, at 1100 °C the plagioclase-out curve of GV10 is very close to that of the Na-rich lherzolite HNa-FLZ recently investigated by Fumagalli et al. (2017) (Fig. 8). This further supports that the effect of the much lower bulk X_Cr of pyroxenite (0.01) is counterbalanced by its Ab/Di ratio (0.11) significantly lower than those of some mantle peridotites (Ab/Di = 0.15–0.30; Green & Falloon, 1998; Borghini et al., 2010).

As a result, plagioclase-facies recrystallization is expected to occur at very close depths or, possibly, simultaneously in fertile lherzolites and in high-Mg pyroxenites, such as pyroxenite GV10 and R934 (Irving, 1974). This is consistent with the observations in ultramafic massifs, where both pyroxenites and peridotites may be partially re-equilibrated at plagioclase-facies conditions. This is the case of mantle sequences from EL ophiolites (Rampone et al., 1995; Montanini et al., 2006; Borghini et al., 2010, 2011, 2013, 2016), as well as mantle tectonites in Ronda (e.g., Hidas et al., 2013). In agreement with experimental results, thermodynamically based estimates of plagioclase-facies recrystallization in Ronda mantle tectonites are within the pressure range of 1.0–0.5 GPa (Hidas et al., 2013). As documented for lherzolites (e.g., Green & Falloon, 1998; Borghini et al., 2010; Fumagalli et al., 2017), the increase of bulk Ab/Di ratio (at fixed bulk X_Mg and X_Cr) is expected to move toward higher pressure the plagioclase-out curve also in Mg-rich pyroxenites, although it needs to be confirmed by further experiments. In pyroxenites with bulk composition approaching those of mafic rocks (i.e. higher X_Mg and Ab/Di ratio), plagioclase stability is further moved to higher pressure, e.g., up to about 2 GPa in olivine-free pyroxenite with an Ab/Di ratio of 0.74 (e.g., pyroxenite R130 from Irving, 1974).

Thermodynamic calculations of the plagioclase-out curve have been performed using the Perple_X package (Connolly, 1990; Connolly & Petrini, 2002) in the chemical system Cr-NCFMAS, adopting the updated version of Holland & Powell (1998) database and solid solution models for Cr-bearing pyroxenes, spinel and garnet (http://www.perplex.ethz.ch). At 1100 °C, the plagioclase-out curve computed by thermodynamic modelling is located at pressure of about 0.1 GPa lower than the experimentally-derived curve (Fig. 8). As discussed in Borghini et al. (2010), calculations are sensitive to the bulk Ab/Di ratio but do not consider the effect of bulk X_Cr on the stability of plagioclase-lherzolite assemblages. Pyroxenite GV10 has a very low X_Cr that is expected to move the plagioclase-out curve towards higher pressure (Borghini et al., 2010). On the contrary, the computed curve is predicted at slightly lower pressure as an effect of the low bulk Ab/Di (Table 1). Moreover, we found that the thermodynamic modelling overestimates the solubility of Al in pyroxenes, mostly in clinopyroxene (Fig. 4), and this could further contribute to reduce the stability field of plagioclase.

Results of this work suggest that, at 1100 °C, anorthite content in plagioclase decreases with pressure, as observed in experiments on mantle peridotites (Fig. 5). In a recent paper, Fumagalli et al. (2017) proposed a geobarometer for plagioclase lherzolites based on the equilibrium Forsterite + Anorthite = CaTschermak + Enstatite (FACE geobarometer). Figure 5 shows that the relation between anorthite in plagioclase and X_Al^M1 in clinopyroxene observed in experiments on GV10 pyroxenite is consistent with data on lherzolites. However, application of FACE geobarometer on experiments of the present study provides equilibrium pressures (0.65 ± 0.05, 0.68 ± 0.05, 0.75 ± 0.05 GPa) systematically lower than experimental pressures (0.7, 0.8, 0.9 GPa), with the highest deviation at the highest pressure. More experimental data on plagioclase-bearing pyroxenites are needed to extend the applicability of the FACE geobarometer to mafic compositions. Noteworthy, few chemical analyses on plagioclase cores from GV10 natural pyroxenite provided an averaged anorthite of 0.58 (Borghini et al., 2016), which is slightly lower than anorthite in experimental samples (Table 7). This is presumably ascribable to the lower temperature of equilibration at plagioclase-facies conditions inferred for this mantle sequence (870–930 °C; Borghini et al., 2011), with respect to temperature of experiments (1100 °C; Table 2), in agreement with the positive correlation between temperature and anorthite in plagioclase (e.g., Fumagalli et al., 2017).

The spinel- to garnet-facies transition is one of the major phase boundaries in the Earth's upper mantle and is relevant to investigate chemical heterogeneity in the lithospheric mantle. A number of experimental studies have been dedicated to this transition in simplified chemical system, such as MAS and CMAS (for a review see Fumagalli & Klemme, 2015). In the CMAS system, the spinel–garnet boundary is univariant and, at near-solidus pressure, occurs at pressure of 1.8–2.2 GPa (e.g., Milholland & Presnall, 1998; Klemme & O’Neill, 2000; Walter et al., 2002). Addition of Fe to the system tends to shift the garnet-in reaction towards lower pressure (O’Neill, 1981; Irifune et al., 1982; Nickel, 1986; Webb & Wood, 1986; Brey et al., 1999, Girnis et al., 2003) whereas the presence of Cr stabilizes spinel relative to garnet (e.g., Klemme, 2004). In particular, in Cr-bearing chemical systems spinel and garnet coexist until spinel breaks down, which is expected to occur at rather high pressure with depletion degree of peridotite (i.e. higher bulk X_Cr, Ziberna et al., 2013). Experimental studies on the stability of hydrated phases in peridotite complex system also documented the spinel–garnet transition at T < 1100 °C (e.g., Niida & Green, 1999; Fumagalli & Poli, 2005).

As discussed above, the garnet-in curve for GV10 pyroxenite is located just below 1.5 GPa, in agreement with the garnet appearance in pyroxenites with similar bulk composition (Fig. 7a), at lower pressure than garnet-in curves for all the peridotites (Fig. 9). However, experimental data in anhydrous peridotite system are scarce (Fumagalli & Klemme, 2015), making difficult the comparison with experiments on pyroxenites. Expansion of garnet stability toward lower pressure is promoted by higher FeO and Al₂O₃ contents, as well as the lower X_Cr of the pyroxenites compared to the peridotites. Experimental results of this study confirm that garnet-bearing assemblages in GV10-type pyroxenite bulks are stable within a rather broad interval of pressure in which most of lherzolites are garnet-free, i.e. from 1.5 to about 2.0 GPa. Moreover, phase relations defined by our experiments indicate that the occurrence of garnet in the primary mineral assemblages, inferred for many pyroxenite layers from EL mantle sequences (Borghini et al., 2016), points to pyroxenite emplacement at pressure higher than 1.5 GPa.

The garnet-in curve calculated by Perple_X for the GV10 pyroxenite has a positive slope and, at near-solidus temperature of 1200–1330 °C, is located about 0.4 GPa above the curve constrained by experiments (Fig. 9). As we discussed in the previous section, the modelling overestimates the Al solubility in pyroxenes and does not consider Ca in orthopyroxene (Fig. 4). Calculations also tend to underestimate Cr solubility in pyroxenes, as previously documented applying the same thermodynamic calculation on various lherzolites (Borghini et al., 2010; Ziberna et al., 2013; Fumagalli et al., 2017). As a consequence, spinel composition derived by thermodynamic modelling is much Cr-richer than the spinel observed in experiments. This could increase the predicted stability of spinel over garnet and plagioclase, expanding the spinel-facies field in Mg-rich pyroxenites, as well as in peridotites, both toward higher and lower pressure (Figs. 8 and 9).

• We have experimentally investigated the subsolidus phase relations in secondary-type pyroxenite GV10 from 0.7 to 1.5 GPa, 1100–1230 °C. Spinel, olivine and pyroxene are stable within the whole pressure range studied here. Plagioclase is observed from 0.7 to 0.9 GPa and garnet occurs solely at 1.5 GPa.

• Plagioclase composition is influenced by pressure, with anorthite decreasing from An = Ca/(Ca + Na) = 0.70 at 0.7 GPa to An = 0.61 at 0.9 GPa. At 0.7 GPa (1100 °C), the mode of plagioclase produced in the studied pyroxenite by metamorphic reaction is more than twice that found in fertile lherzolites. Spinel modal abundance decreases at decreasing pressure in plagioclase-bearing experiments and at decreasing temperature where it coexists with garnet at 1.5 GPa. Garnet has pyrope-rich compositions and its modal abundance increases with decreasing temperature, up to 10 wt% at 1150 °C.

• Pyroxenites having bulk composition characterized by high X_Mg and relatively low normative plagioclase, such as the secondary-type pyroxenite GV10, have subsolidus phase relations similar to fertile lherzolite. Plagioclase-out boundary is within the pressure range of many lherzolites, and spinel is the unique aluminum phase stable at intermediate pressure (1–1.4 GPa). However, in pyroxenites garnet appears at significantly lower pressure than in lherzolite.

We gratefully acknowledge reviewers A. Montanini and D. Laporte for their constructive criticism and suggestions. We thank E. Rampone for the editorial handling. A. Risplendente is thanked for his technical assistance during the electron microprobe work. This work was financially supported by the Italian Ministry of Education, University and Research (MIUR) [PRIN-2015C5LN35] “Melt rock reaction and melt migration in the MORB mantle through combined natural and experimental studies”.