Multiple episodes of melting, depletion, and enrichment of the Tethyan mantle: Petrogenesis of the peridotites and chromitites in the Jurassic Skenderbeu massif, Mirdita ophiolite, Albania

The Mesozoic ophiolites in the eastern Mediterranean region occur along nearly east-west–trending, curvilinear suture zones (Fig. 1A) separating a series of Gondwana-derived continental blocks, and represent the remnants of Tethyan oceanic lithosphere that developed in different seaways, evolved between these microcontinents (Dilek and Moores, 1990; Shallo et al., 1990; Dilek et al., 1999; Dilek, 2003; Saccani and Photiades, 2004; Saccani and Tassinari, 2015, and references therein). The Jurassic ophiolites in the Albanide-Hellenide segment of the Alpine orogenic system are the relics of the Mesozoic Pindos-Mirdita marginal basin, which developed between the Apulia and Pelagonia-Korabi microcontinents. These ophiolites are underlain by discontinuously exposed metamafic and metapelitic rocks and mélanges, which include blocks and clasts of ophiolitic and rift-drift related volcanic-sedimentary rock units in a sedimentary matrix (Shallo et al., 1990; Shallo and Dilek, 2003; Dilek et al., 2005, and references therein). The ophiolites are uncomformably overlain by supraophiolitic mélanges and middle Cretaceous and younger undeformed neritic carbonates, indicating that ophiolite emplacement in the region was completed by the Early Cretaceous (Chiari et al., 2003, 2004; Shallo and Dilek, 2003).

Early studies in the Mirdita ophiolite zone led to the recognition of two types of ophiolites with distinct geology, petrology, geochemistry, and tectonic setting of formation (Shallo et al., 1990; Beccaluva et al., 1994; Bortolotti et al., 1996; Bébien et al., 2000; Hoeck et al., 2002). The western Mirdita ophiolite (WMO) has been interpreted as the relics of the Jurassic oceanic lithosphere developed at a mid-oceanic ridge environment (MOR ophiolite), whereas the eastern Mirdita ophiolite (EMO) has been interpreted as a remnant of suprasubduction zone (SSZ) generated oceanic lithosphere within the same basin. These models therefore envision discrete and separate tectonic settings of formation for the WMO and EMO separated in time and space. However, most recent studies have shown that this twofold subdivision is not as sharp as previously thought (Bortolotti et al., 2002; Bébien et al., 1998, 2000; Insergueix-Filippi et al., 2000; Manika et al., 1997; Hoeck et al., 2002; Dilek and Flower, 2003; Beccaluva et al., 2005; Koller et al., 2006; Dilek and Morishita, 2009; Saccani and Tassinari, 2015; Saccani et al., 2017).

In the WMO, typical mid-oceanic ridge basalts (MORB) are locally interlayered with island arc tholeiite (IAT) lavas and the uppermost volcanic units include boninitic lava flows, which clearly have subduction zone geochemical affinities (Dilek and Furnes, 2011, 2014). In the southern Albanides, some of the peridotite massifs (e.g., Morava and Shpati) include both lherzolitic and harzburgitic mantle tectonites, and some ophiolites (Voskopoja, Rehove, and Morava massifs) that have been previously interpreted as MOR type contain intrusive and volcanic rocks displaying subduction zone influence in their geochemical fingerprint (Hoeck et al., 2002; Koller et al., 2006). It has become clear, therefore, that the tectonic and magmatic evolution of the Jurassic ophiolites in the Albanides was not limited to discrete and separate geodynamic settings, but was part of the continuum of oceanic lithosphere development in a marginal basin, which underwent rift drift, seafloor spreading, and subduction zone tectonics (Dilek and Flower, 2003; Saccani and Photiades, 2004; Beccaluva et al., 2005).

In this paper we present new data on the mineral chemistry, whole rock, trace element, and platinum group element (PGE) geochemistry from the upper mantle peridotites and chromitite deposits in the Skenderbeu massif in the Mirdita zone in the northern Albanides, and discuss their petrogenesis and melt evolution. We have chosen the Skenderbeu massif for this study because it is situated along the geochemically delineated boundary between the WMO and EMO within the Mirdita zone (Fig. 1B), and its peridotites reveal textural, mineralogical and geochemical evidence for transition from MOR- to SSZ- related magmatic-metasomatic processes and melt-rock interactions. The geochemistry of the Skenderbeu massif has been studied previously by Shenjetari and Beqiraj (2010) and Onuzi et al. (2007), but their data were limited to whole rock geochemical characteristics and did not focus on interpreting the melt evolution. Following our data presentation and interpretations on the melt evolution of the Skenderbeu massif, we discuss the physical conditions and processes of chromite genesis in ophiolitic peridotites.

The Mirdita ophiolite zone in the northern Albanides is nearly 30 km wide in the northwest-southeast direction and is bounded by the conjugate passive margin sequences of Apulia in the west and Korabi-Pelagonia in the east (Fig. 1B). Magnetic and gravity anomalies show a >15-km-thick mafic-ultramafic slab occupying a large synform structure between these continental blocks (Frasheri et al., 1996). Large peridotite massifs associated with ophiolitic crustal units occur on both sides of this synform structure and adjacent to the continental blocks. Adjacent to the Apulian continental margin in the west, these peridotite massifs (Krrabi, Puke, Gomsiqe, Skenderbeu) consist mainly of plagioclase lherzolites, whereas those close to Pelagonia in the east (Kukesi, Lure, Bulquiza, Shebenique) are made of depleted harzburgites, dunites, and chromitite deposits (Fig. 1B; Shallo et al., 1990; Hoxha and Boullier, 1995; Bebien et al., 1998; Shallo and Dilek, 2003; Dilek et al., 2005; Shenjatari and Beqiraj, 2010).

The ophiolitic sequences spatially associated with the western peridotite massifs are ∼3 km thick, and are composed mainly of gabbros and extrusive rocks with predominantly MORB geochemical affinities. The peridotites contain lherzolite-harzburgite, plagioclase lherzolite, plagioclase dunite, and rare amphibole peridotites (olivine hornblendite). Plutonic rocks, which are locally intrusive into and overlying the peridotites, consist of troctolite, olivine gabbro, ferrogabbro, gabbro, and rare amphibole gabbro, and generally display an olivine-plagioclase- orthopyroxene (ol-pl-opx) order of crystallization (Beccaluva et al., 1994, 2005; Saccani and Photiades, 2004; Miranda and Dilek, 2010; Shenjatari and Beqiraj, 2010). Extrusive rocks, composed mainly of massive to pillow lavas and hyaloclastites, form a nearly 600-m-thick sequence that directly overlies serpentinized peridotites and gabbroic rocks along primary contacts (Dilek et al., 2005). Isolated dikes crosscut these extrusive rocks and feed into the younger lava flows on top. The lavas are stratigraphically overlain by as much as ∼20-m-thick radiolarian cherts that are late Bajocian–early Bathonian (ca. 168–166 Ma) to late Bathonian–early Callovian (ca. 165–163 Ma) (Chiari et al., 2004; Marcucci et al., 1994).

The ophiolitic sequences that are spatially associated with the eastern peridotite massifs are generally thicker (∼10–12 km) than those in the west and display predominantly SSZ geochemical affinities (Shallo et al., 1990; Dilek and Flower, 2003; Shallo and Dilek, 2003; Saccani and Photiades, 2004; Beccaluva et al., 2005; Dilek et al., 2005). The internal structure, chemostratigraphy, and geochemical features of these eastern ophiolites were discussed in detail by Dilek and Flower (2003) and Phillips-Lander and Dilek (2008). Structural and geochemical studies have shown that these two ophiolite types are both laterally (from west to east) and vertically transitional in time and space (Dilek and Flower, 2003), although the original igneous contacts have been locally modified by late Cenozoic contractional deformation (Dilek et al., 2005).

The Skenderbeu massif occurs in the southern end of the Mirdita ophiolite zone in the northern Albanides and is unconformably (but locally tectonically along high-angle faults) by the Cenozoic terrestrial sedimentary deposits of the northwest-southeast–oriented Burreli basin (Fig. 1B). It is juxtaposed on the west against the Triassic–Jurassic carbonate rocks of the pre-Apulian platform along west-northwest–vergent thrust and transpressional fault systems (Dilek et al., 2005). The massif includes, from the bottom (in the west) to the top (in the east), harzburgites with dunite bodies, lenses and minor chromitites, lherzolite and plagioclase lherzolite with layered dunite-gabbro-troctolite on top (Fig. 2; Shenjatari and Beqiraj, 2010). Basaltic pillow lavas directly overlie the peridotites and ultramafic cumulates along east-dipping normal faults (Fig. 2). These pillow lavas are unconformably overlain by a chaotic volcanic-sedimentary unit, locally olistostromal in nature, that displays gentle to open folds. This unit represents a latest Jurassic–Cretaceous supra–ophiolitic mélange (Fig. 2; Shallo and Dilek, 2003), which likely developed during the early stages of basinwide contraction and closure (Dilek et al., 2005).

The Skenderbeu peridotites consist of harzburgite, dunite and minor chromitite (Fig. 3) with various degrees of serpentinization effects characterized by lizardite and less abundant bastite (Fig. 3). Harzburgitic rocks are coarse grained, have granular textures and consist mainly of olivine (65–80 modal%), opx (10–20 modal%), and clinopyroxene (cpx; 2–5 modal%), with minor magnesiochromite and magnetite (Figs. 3A, 3B). Olivine crystals are 0.1–0.2 mm in diameter, occur as relatively large granular grains, and show deformation lamellae, kink bands and wavy extinction. Small, euhedral olivine crystals also occur as inclusions in magnesiochromite and opx grains and along the contacts between these minerals. The opx forms large (0.5–1.5 mm), euhedral to subhedral, tabular grains showing extensive crystal flexural bends, undulatory extinction, and gliding twins. Small, granular crystals of opx also occur within microscopic shear zones and as inclusions in magnesiochromite. Some of the large opx grains have exsolution lamellae of cpx along their cleavage planes; cpx occurs as 0.3–1.5-mm-wide anhedral grains (Fig. 3A). Magnesiochromite consists of small (0.5–1 mm), subhedral to amoeboidal, interstitial grains or as euhedral inclusions in olivine and opx (Fig. 3B).

Dunitic rocks are relatively fresh with little serpentinization effects, and consist of olivine (96–99 modal%) and opx (1%–4%) with minor cpx and magnesiochromite; they display a characteristic granular texture (Fig. 3C). Olivine grains are mostly 0.1–0.5 mm in diameter, but some of them are recrystallized to form fine-grained mosaic textures. Larger olivine grains show wavy extinction, and deformation twins. Scattered magnesiochromite occurs as small (0.1–0.4 mm), subhedral to amoeboidal interstitial grains or as euhedral inclusions in olivine (Fig. 3C).

Podiform chromitites occur as massive and disseminated types in the Skenderbeu peridotites. Original silicate minerals in these chromitites are altered into serpentine, chlorite, and clay minerals. Massive chromitite are typically composed of >95 modal% magnesiochromite and ∼5 modal% olivine (Fig. 3D). Chromian spinel grains in massive chromitites are generally euhedral to subhedral or irregular in shape, 0.2–0.5 mm wide, and red. Relict olivine grains locally exist in the chromitites (Fig. 3D).

Disseminated chromitites appear as densely to moderately disseminated types, and consist of variable proportions of magnesiochromite and olivine. Magnesiochromite grains contain inclusions of olivine, opx, or cpx (Figs. 3E, 3F), all of which are altered into serpentine, chlorite, or clay minerals. Chromian spinel grains are generally euhedral to subhedral or irregular in habit, 0.2–0.5 mm to 0.3–1 mm across, and dark red and/or black.

We selected a total of 34 samples from the upper mantle units of the Skenderbeu massif, including 11 harzburgite, 4 dunite, 8 massive chromitite, 7 densely disseminated chromitite, and 4 moderately disseminated chromitite samples (Fig. 2). Sample locations are shown in Figure 2. Peridotites with a relatively low degree of serpentinization and alteration were analyzed for their whole-rock geochemistry and mineral chemistry. Mineral chemistry analyses were performed with a JEOL JXA-8100 electron microprobe at the State Key Laboratory of Continental Tectonics and Dynamics, Institute of Geology, Chinese Academy of Geological Sciences. The measurements were performed at 15 kV, 20 nA, and counting time of 10 s. A beam diameter of 5 μm was used.

Bulk rock major oxides and trace elements were analyzed at the National Research Center for Geoanalysis in Beijing, China. All of the samples were carefully cleaned, crushed, and then ground in an agate mortar to pass a 200 mesh screen. Major elements were determined by X-ray fluorescence on fused glass beads using PW4400 spectrometry. Trace elements, including the rare earth elements (REEs), were determined by inductively coupled plasma–mass spectrometry. One national standard sample (GBW07105) was measured simultaneously to ensure consistency of the analytical results. The accuracy of these results is better than 10% for most elements, with many elements better than 5% of the recommended values. Water and CO₂ were determined by gravimetric techniques in which a sample is heated in a closed container and the water vapor is collected in a separate tube, condensed, and weighed. The detection limit for H₂O and CO₂ is 0.01 wt%.

The PGEs were analyzed at the National Research Center for Geoanalysis, Chinese Academy of Geological Sciences, Beijing. Detailed analytical processes for PGE were described by Xiong et al. (2015).

Our whole-rock geochemical data of the selected samples are presented in Table 1. Although the harzburgites and dunites are altered to variable extents, as evidenced by the range of loss-on-ignition values (3.79–11.34 wt%), the analyzed samples retain their primary mineralogical and textural features, allowing us to use them to investigate their petrogenesis and geochemical evolution. Effects of hydrothermal alteration on the whole-rock major oxides of dunites are smaller than those observed in the harzburgites (Table 1). This phenomenon may have resulted from refertilization of the dunites in later stages of the petrogenesis of the Skenderbeu peridotites (see following). The low CaO contents (0.31–2.1 wt%) of all peridotites are consistent with the scarcity of cpx in them. The analyzed samples are Mg rich (36.37–45.07 wt%, average of 38.85 wt%), Al poor (0.29–1.28 wt%, average of 0.97 wt%), and Na poor (Na₂O < 0.05wt%).

In the MgO/SiO₂ versus Al₂O₃/SiO₂ diagram (Fig. 4), the bulk compositions of the peridotites display a depletion trend nearly below but parallel to the terrestrial mantle array (Hart and Zindler, 1986; Jagoutz et al., 1979). This depletion may reflect loss of magnesium or addition of SiO₂ into the system during serpentinization. The Al₂O₃/SiO₂ ratios decrease significantly, whereas MgO/SiO₂ ratios increase slightly in opx and cpx during partial melting. As melting proceeds, pyroxene, especially cpx, is consumed very quickly, and the modal contents of olivine increase, causing an increase in the MgO/SiO₂ ratio. As the partial melting degree increases, the bulk MgO/SiO₂ ratio increases. It then becomes apparent that partial melting may not be the sole reason why the observed bulk MgO/SiO₂ ratio is below the terrestrial mantle array. All harzburgite samples are represented by slightly lower MgO/SiO₂ values than the terrestrial mantle array (Fig. 4), implying some Mg loss as a result of alteration (Niu, 2004; Snow and Dick, 1995). Some samples may therefore contain lower major and trace element concentrations than those estimated for a given degree of melting due to the influence of alteration. However, some dunite samples have relatively high MgO/SiO₂ values, suggesting that their compositions were likely affected by interactions with olivine-rich melts at later stages. We have also plotted forearc and abyssal peridotites in Figure 4 for comparison. All our analyzed dunite samples plot in the forearc peridotite field, whereas the harzburgite samples plot in the abyssal peridotite field (Fig. 4).

We have examined the elemental variations in each sample with respect to its Mg content because MgO values can be used effectively as an indicator of the degree of depletion, if the rocks are not significantly altered. Any increase in the MgO content is consistent with increased depletion and higher olivine contents. On Harker diagrams, both the harzburgite and dunite samples show relatively linear decreases in CaO, Al₂O₃, and SiO₂ with increasing MgO (Figs. 5A–5C). Total iron as FeO* shows little linear increase with increasing MgO contents (Fig. 5D). Four fresh dunite samples have normalized MgO values of ∼46.0–47.8 wt% SiO₂. The CaO and Al₂O₃ values are essentially constant over this compositional range, but FeO* increases slightly with increasing MgO. All of the samples are significantly depleted relative to the primitive mantle (Fig. 5).

The total REE inventory of the Skenderbeu peridotites is low, between 0.114 and 0.646 ppm (Table 1). The harzburgites display higher REE concentrations with respect to the dunites (0.309–0.646 ppm versus 0.114–0.218 ppm, respectively).

Except for Nb, Ta, La, and Ce, all other elements correlate positively with MgO contents (Parkinson and Pearce, 1998). Cobalt behaves in a compatible manner during partial melting and correlates positively with MgO, whereas other elements that are incompatible in olivine and all REEs (except for La and Ce) correlate negatively with the MgO contents (Fig. 5). The chalcophile elements, Cu and S, behave like Ca and Al during partial melting (Luguet et al., 2003). Therefore, the Cu concentrations of the samples decrease with increasing degree of depletion (Table 1). In Figure 5 we also compare the Skenderbeu peridotites with the abyssal and forearc peridotites. Trace element characteristics of dunites are mainly comparable to those of the forearc peridotites, whereas the trace element features of the harzburgites are comparable to those of the abyssal-type peridotites.

The mantle peridotites in Skenderbeu have total REE contents well below the primitive mantle values (Fig. 6). This feature is a likely artifact of significant depletion as a result of high degrees of partial melting (Miller et al., 2003). In chondrite-normalized plots, the analyzed samples display similar depleted patterns (Fig. 6A). All harzburgites show relative enrichment in both light (L) REEs and heavy (H) REEs and depletion in middle (M) REEs (Fig. 6A). These patterns suggest weak LREE enrichment of previously depleted peridotites, and are consistent with the patterns displayed by abyssal peridotites with low REE contents and high HREE concentrations; the dunite samples display stronger depletion in comparison to the harzburgites and resemble forearc peridotites (Fig. 6A; Parkinson and Pearce, 1998).

The PGE contents of the chromitites, dunites, and harzburgites are presented in Table 2. The PGE abundances are low in all samples, averaging 146.5 ppb in chromitite, 35.6 ppb in harzburgite, and 30.87 ppb in dunite. The total PGE concentrations in the podiform chromitites are within the range of values generally reported for ophiolitic chromitites (<300 ppb; El Ghorfi et al., 2008). The Pd/Ir, Ru/Pt, and IPGE /PPGE (i.e., iridium group, Os, Ir, Ru/palladium group, Rh, Pt, Pd) ratios are 0.029, 44.5, and 11.7 for Cr-rich chromitites and 2.15, 0.83, and 0.90, respectively, for harzburgites and 1.04, 0.98, and 1.29, respectively, for dunites. Harzburgite samples from the Skenderbeu massif are enriched in PGE relative to the primitive mantle (Fig. 7; McDonough and Sun, 1995), and depleted relative to their abundances in the chromitites. The PGE concentration in dunite samples is slightly higher than that of the primitive mantle (Fig. 7; McDonough and Sun, 1995). The analyzed chromitites are ∼5 times enriched in PGE relative to their host peridotites, and show enrichment in IPGE relative to PPGE (Fig. 7).

We obtained microprobe analyses of 42 olivine grains from the harzburgite and dunite samples. The representative data are given in Table 3. Although all of the analyzed olivine is highly magnesian (forsterite, Fo 89.31–96.52), there are significant differences among the olivine grains from various host rocks.

In the peridotites and chromitites, olivine occurs as relatively large, granular crystals and as small inclusions in both magnesiochromite and opx. Granular olivine in the harzburgites has Fo values of 89.31–90.11 (average 89.74), with NiO contents of 0.28–0.41 wt% (average 0.34 wt%) and MnO contents of 0.10–0.18 wt% (average 0.13 wt%). The olivine in dunite has slightly higher Fo values (90.36–92.92, average of 91.29) and NiO contents (0.27–0.44 wt%, average of 0.37 wt%), but slightly lower MnO contents (0.05–0.16 wt%, average of 0.11 wt%) than the olivine in the harzburgites. Both the Fo values and NiO contents of olivine in the harzburgites increase near the contacts with podiform chromitites and dunite lenses.

In the moderately disseminated chromitites, the granular olivine has Fo values of 96.44–96.52, NiO contents of 0.53–0.60 wt%, and MnO contents of 0.06–0.07 wt%. In general, the olivine grains become more magnesian in the order of harzburgite® dunite® chromitite (Fig. 8). Because olivine inclusions in magnesiochromite are typically enriched in NiO and depleted in MnO in comparison to most olivine grains (Pagé et al., 2008), the NiO contents in olivine inclusions are clearly higher than those of the mantle olivine array. The MnO contents in some olivines are also low (Fig. 8). However, nickel contents show a steady increase with increasing Fo values, whereas the manganese contents show a slight decrease with increasing Fo values (Fig. 8).

Average microprobe analyses of opx in various peridotite types are given in Table 4. Most opx in the harzburgites is fresh, but in some highly altered samples it is replaced by serpentine to form bastite. Compositional variations in opx grains depend on the modal mineralogy of the peridotites. The enstatite (En) contents of opx in the depleted harzburgites range from 84.1 to 87.3, and in the dunites from 89.1 to 90.7 (Table 4). The En content of opx correlates positively with increasing depletion of the peridotites. The Al₂O₃ content of opx is variable, in the range of 2.45–3.22 wt% in the depleted harzburgites, and 0.62–1.03 wt% in the dunites (Fig. 8). Chrome contents in all peridotites are low (0.69–0.98 wt% in harzburgites, 0.14–0.44 wt% in dunites; Fig. 8). The CaO contents of opx are 1.35–2.81 wt% in the harzburgites and 0.39–1.23 wt% in the dunites, respectively (Table 4).

Cpx occurs in the harzburgites, dunites, and chromitites. Although all of the cpx is diopside (Table 5), it varies in grain size, crystal habit, and composition in different host rocks. Cpx in the harzburgites have the lowest Mg#s (90.7–92.9, average of 91.7), which increase to 93.88–96.04 (average of 95.04) in the dunites with increased degrees of depletion (Fig. 8). The Al₂O₃ content of cpx is variable, ranging from 2.8 to 4.5 wt% in the depleted harzburgites and from 0.38 to 1.0 wt% in the dunites, indicating a negative correlation (Fig. 8). Chrome contents of cpx in the harzburgites range from 0.85 to 1.49 wt% (average of 1.22 wt%), and the Cr₂O₃ contents in the dunites range from 0.13 to 0.71 (average of 0.39 wt%). Cpx in different chromitite types has a range of composition similar to that in the dunites, but a higher compositional range than those in the harzburgites. The Ti₂O contents of cpx correlate negatively with Mg#s; they are higher in the harzburgites (0.03–0.17 wt%, average of 0.08) than in the dunites (0–0.04 wt%). The Na₂O contents of cpx correlate positively with Mg#s; they are higher in dunites (0.007–0.059 wt%) than in the depleted harzburgites (0–0.047 wt%). The cpx in the chromitites has the highest Na₂O contents.

The cpx grains in the massive chromitites have Mg#s of 95.9–98.1 (average of 96.8), Al₂O₃ contents of 0.24–0.83 wt% (average of 0.57), and Cr₂O₃ contents of 0.69–1.94 wt% (average of 1.37). In densely disseminated chromitites, Mg#s are 96.0–96.9 (average of 96.4), Al₂O₃ contents 0.56–1.9 wt% (average of 0.88), and Cr₂O₃ contents 1.13–1.49 wt% (average of 1.30). In moderately disseminated chromitites, Mg#s are 95.9–96.4 (average of 96.2), Al₂O₃ contents 0.7–0.9 wt% (average of 0.8), and Cr₂O₃ contents are 1.12–1.53 wt% (average of 1.40) (Fig. 8; Table 5). Combined with the appropriate microtextural observations, such compositional fingerprints are in favor of a residual rather than metasomatic origin for the analyzed cpx grains. Their composition matches that of typical mantle-derived cpx from suboceanic peridotites, showing a relative depletion in Na (Kornprobst et al., 1981).

Selected trace element and REE compositions of the cpx from the Skenderbeu harzburgites are given in Table 1. The cpx from the harzburgites shows consistent spoon-shaped REE patterns, similar to those of cpx in abyssal peridotites (Johnson et al., 1990), with a slow decrease in HREEs and MREEs from Yb to Eu and rapid decrease in LREEs from Sm to Ce (Fig. 6C).

Trace element spider diagrams (Fig. 6D) show that the cpx grains in the harzburgites have extremely low concentrations of the more incompatible lithophile elements (Table 1), and negative anomalies in Zr relative to neighboring REE. However, cpx grains in harzburgites show generally similar MREE and HREE contents, and obvious enrichment in fluid-mobile elements (e.g., U, Pb), relative to those in primitive mantle (Fig. 6D).

Magnesiochromite occurs as residual grains in the peridotites and as primary grains in the chromitites. In the highly serpentinized peridotite samples, magnesiochromite grains have oxidized rims. We therefore analyzed only the fresh cores of the grains. Magnesiochromite compositions are given in Table 6.

Accessory magnesiochromite is common in the peridotite samples but rarely exceeds 5 modal%. In contrast to other mafic silicates, the analyzed magnesiochromite in all the samples spans a wide compositional range. Most of the grains are relatively rich in Cr, but some highly aluminous spinel varieties are also present in the harzburgites. There is a wide range of substitution of Cr and Al in the magnesiochromites, as reflected in their Cr# = [100 * Cr/(Cr + Al)]. Increasing Cr#s typically correlate with increasing degrees of partial melting in the host peridotites, and thus can be a sensitive indicator of mantle depletion (Dick and Bullen, 1984). Mg#s also indicate the degree of partial melting in upper mantle peridotites; however, both of these indices can also be affected significantly by melt-rock reactions. The residual magnesiochromites show a wide range in their crystal forms, grain sizes, and compositions, and commonly occur as large, irregular, amoeboid-shaped grains. Those in the harzburgites are low Cr with Cr#s between 41.5 and 44.7 (average of 43.9) and Mg#s of 54.8–61 (average of 59), whereas those in the dunites have much higher Cr#s (61.5–71.9, average of 69.7) and much lower Mg#s (41.3–50.2, average of 45.9) in comparison to those in the harzburgites (Table 6). The TiO₂ contents are low in the dunites and in most of the depleted harzburgites (<0.1 wt%). However, some depleted harzburgites contain magnesiochromite with high TiO₂ contents (to 0.14 wt%).

The mineral chemistry data based on 43 magnesiochromite grains obtained from 3 different orebodies are shown in Figure 8 and summarized in Table 6. Magnesiochromites in the massive chromitites have Mg#s of 60.8–64.1 (average of 63.0), Cr#s of 79.2–80.4 (average of 79.6), Al₂O₃ contents of 9.6–10.3 wt% (average of 10.13), and TiO₂ contents of 0.06–0.17 wt% (average of 0.13). In densely disseminated chromitites, Mg# range from 59.9 to 62.0 (average of 60.7), Cr#s are between 78.8 and 79.9 (average of 79.4), Al₂O₃ contents range from 9.9 to 10.5 wt% (average of 10.2), and TiO₂ contents are between 0.09 and 0.17 wt% (average of 0.12). In moderately disseminated chromitites Mg#s range from 58.1 to 59.8 (average of 59.1), Cr#s are between 75.3 and 77.3 (average of 75.9), Al₂O₃ contents are between 11.0 and 12.1 wt% (average of 11.7), and TiO₂ contents range from 0.09 to 0.18 wt% (average of 0.13) (Fig. 8; Table 6).

The modal mineralogy, mineral chemistry, and bulk-rock geochemistry can be used to evaluate the degree of partial melting of upper mantle peridotites in ophiolites. The cpx contents of ophiolitic peridotites reflect only the degree of depletion. Because the equilibrium between olivine and melt remains unchanged by H₂O input (Gaetani and Grove, 1998), the Fo composition of olivine reveals the total degree of partial melting. Based on these criteria, we infer that different peridotite types in the Skenderbeu massif that contain varying amounts of cpx (to 5 vol%) and olivine with variable Fo contents (89.3–96.5) must have undergone different degrees of partial melting.

Aluminum contents of pyroxene and spinel are sensitive to partial melting degrees of mantle rocks, and they systematically decrease as peridotites become more depleted (Dick and Natland, 1996). The Al₂O₃ contents of cpx within the Skenderbeu peridotites correlate negatively with the Cr# values of spinel with which they were equilibrated. The cpx in the harzburgites is represented by high Al₂O₃ contents, and the spinel grains in equilibrium with these minerals have low Cr# values (Fig. 9A). Olivine compositions in the dunites vary significantly, whereas the Cr# values of chromites remain relatively constant (Fig. 9A). The coarse-grained character (Fig. 3), extensive high-temperature deformation, and highly magnesian character in Skenderbeu peridotites argue against an origin as magma chamber cumulates. It is plausible that the inferred high melt flux through the Skenderbeu peridotites was so protracted that the compositions of the generations of Cr spinels were extensively homogenized (see following discussion). The Al₂O₃ versus TiO₂ contents of cpx of all the peridotite samples (harzburgites and dunites) show variations consistent with an increase in the degree of partial melting of their host rocks (Fig. 9B). The Cr# and Mg# values of spinels correlate negatively, suggesting that their host peridotites were produced by different degrees of partial melting. The correlation between the Cr#s of spinels and the Mg# values of olivines in the Skenderbeu peridotites reveals that the trend of our analyzed samples is within the olivine-spinel mantle array (Arai, 1994). This trend confirms that the Skenderbeu peridotites are the residues of various degrees of melt extraction (Fig. 9C).

To better constrain the degree of partial melting of the Skenderbeu peridotites, we have modeled the REE compositions of bulk-rock samples and cpx from the harzburgites with a nonmodal fractional melting model (Fig. 6). This model contains a four-phase assemblage of olivine + opx + cpx + spinel. We have used the starting and melting mineral modes of Niu et al. (1997) and nonmodal fractional melting equation of Johnson et al. (1990) in our modeling. The partition coefficients used in this modeling may be affected by temperature, pressure, and bulk composition of the system. However, we assume that the partition coefficients can be considered constant for the purpose of our modeling because of the complex relationships between the partition coefficients and the environment.

The variably depleted nature of the Skenderbeu peridotites, as evidenced by their mineralogical and geochemical compositions, indicates that they are mantle residues after different degrees and multiple episodes of partial melting. Some of the major mineralogical and petrological evidence in support of this inference includes: (1) the progressive depletion of the studied peridotites in magmaphile major oxides (e.g., Al₂O₃, CaO; Table 1; Fig. 5), and in incompatible trace elements (e.g., V, Sc, Y; Table 1; Fig. 5); (2) the forsteritic composition of olivine grains (Fig. 8); (3) the relatively low to moderate Al₂O₃ abundance in both opx (Fig. 8) and cpx (Fig. 8); and (4) an elevated range of Cr# values in chromian spinel (Fig. 8). The REE patterns of the Skenderbeu peridotites are compared with those predicted from petrological modeling in Figure 6. Trace elements such as Sc, V, Y, Ti, and HREEs are generally considered to be subduction-conservative elements or fluid-immobile elements, which are unaffected by mantle metasomatism (Parkinson and Pearce, 1998). Thus, we can use the HREE contents to estimate the extent of partial melting of the Skenderbeu peridotites. In general, the HREE contents of our samples are consistent with the REE modeling results. Figure 6 shows that the Skenderbeu harzburgites have low HREE contents, suggesting 10%–15% partial melting. In contrast, the dunites appear to have undergone 20%–25% melting. The partial melting degrees deduced from the REE contents generally agree with those deduced from the Cr# values of magnesiochromites and Mg# values of olivine (Fig. 9C). Although the HREE contents of the Skenderbeu peridotites can be modeled effectively, their LREE contents have much higher concentrations than can be explained by such a model; we therefore posit that these peridotites are not likely to be the simple residues after different degrees of partial melting.

Peridotites of mantle wedges above subduction zones are generally more oxidized than upper mantle peridotites in other tectonic settings (Parkinson and Arculus, 1999). Water derived from dehydration of subducted oceanic lithosphere plays a major role as an important oxidizer in the transformation of ferrous iron to ferric iron in melt (Arculus, 1994) and in the formation of subduction-related melts.

The fO₂values calculated on the basis of average mineral composition are given in Table 7, as deviations (ΔlogfO₂ FMQ) from the FMQ (fayalite-magnetite-quartz) buffer.

We interpret the observed increase in both Cr# values (41.4–44.4 versus 68.7–71.9) and the calculated fO₂values (−0.203–0.578 versus −0.086–1.143) of chromian spinel from harzburgites to dunites to be a result of melt-rock interactions. The residual mantle harzburgite, which has a MOR or backarc basin geochemical signature, represents the rock. Hydrous SSZ melts, which had higher Cr# values and calculated fO₂values (e.g., Ishii et al., 1992; Parkinson and Pearce, 1998; Pearce et al., 2000; Dare et al., 2009) represent the inferred melt. Their interactions produced the dunites and chromitites in the Skenderbeu massif (Fig. 9D). This relationship supports the notion that the boninitic melts, together with subduction-related magmas and fluids, interacted with the upper mantle rocks in a mantle wedge during their ascent. However, the harzburgites are also represented by a wide range of oxygen fugacities (from −0.203 to +0.578) and by lower and almost constant Cr#s of spinel compositions, which we attribute to fluid-rock interactions (Aldanmaz et al., 2009).

Spinel grains in the Skenderbeu harzburgites have lower Cr# values (41.5–44.7) and are represented by low Ti contents (0–0.144 wt%, average of 0.09 wt% TiO₂). However, spinels in the chromitites that crystallized from melts as a result of high-degree partial melting have a narrow range of high Cr# values (75–80) and high Ti contents (0.06–0.18 wt%, average of 0.13 wt% TiO₂) (Fig. 8H). The high-Cr (Cr# > 60) chromitites are enveloped by dunite in the field. Away from these high-Cr chromitite bodies in the field, cpx abundances in the peridotites increase in outcrops. We explain this phenomenon as a result of incongruent melting of cpx due to the interaction between the upper mantle peridotites and a boninitic melt, as suggested by Dilek and Morishita (2009) and Morishita et al. (2011). In this model, the high-Cr chromitites are the manifestations of a high degree of partial melting, producing olivine and SiO₂-rich boninitic melts. Dunitic peridotites are impregnated by irregular melt channels.

The TiO₂ contents of the spinels in the dunite samples increase toward the compositions of the Skenderbeu chromitites. This trend is well explained by a reaction that must have taken place between the preexisting MOR-type harzburgites and boninitic melts (Fig. 8H), although the spinel Cr# values of the harzburgites and dunites are clearly not continuous. It is plausible that the inferred high melt flux through the Skenderbeu peridotites was so protracted that the compositions of the analyzed Cr spinels were extensively homogenized. In this case, Al-rich spinels in the peridotites were modified by the boninitic melt, from which high-Cr chromitites were crystallized. During the course of this melt-peridotite interaction, the Cr/Al ratio of spinels was increased toward the chromitite mass. As a result, except for the cpx-rich samples, all other peridotites that interacted with boninitic melts and fluids were depleted to varying extents. This interpretation is also supported by a positive correlation between the fO₂ and Cr# of spinels (Fig. 9D) in the depleted harzburgite and dunite samples, implying that the composition of spinel was affected by a melt–wall-rock interaction (Pearce et al., 2000).

The Skenderbeu harzburgites and the cpx from these harzburgites have spoon-shaped trace element patterns, showing prominent LREE enrichment (Fig. 6). This kind of LREE enrichment signature has commonly been interpreted as a major influence of subduction-derived fluids. However, these patterns are inconsistent with the modeled REE patterns, and they are more compatible with LREE enrichment (Fig. 6) caused by interactions between ascending melts and the residual harzburgites beneath a MOR system. This process is envisioned to occur in the cold thermal boundary layer, through which ascending melts react with an earlier melt residue (Niu, 2004), resulting in both LREE and high field strength element (HFSE) enrichments in abyssal peridotites. The prominent LREE enrichment displayed by the Skenderbeu dunites was, however, a result of their reaction with hydrous SSZ melts (Parkinson and Pearce, 1998; Dilek and Furnes, 2011, 2014).

PGEs are highly siderophile elements (Lorand et al., 1999; Alard et al., 2000). On the basis of their melting temperatures, the PGEs are classified into two groups: the IPGEs (melting temperature >2000 °C) and the PPGEs (melting temperature <2000 °C). The IPGE are thought to occur within the mantle as discrete minerals or sulfides (osmiridium and laurite) that are commonly hosted in silicate grains, whereas the PPGE are more likely to occur as sulfides that are commonly interstitial in nature. The PPGEs are more easily accessed during partial melting (Woodland et al., 2002). Partial melting can substantially fractionate IPGEs from PPGEs. Common minerals in mantle peridotites have variable PGE contents, which generally increase from garnet to olivine, opx, cpx, and spinel (Mitchell and Keays, 1981). Whole-rock PGE abundances calculated from the modal mineralogy in mantle peridotites are typically lower than actually determined concentrations, indicating that 60%–80% of the PGEs occur in sulfide-rich components (Mitchell and Keays, 1981). Therefore, sulfides have been regarded as the main hosts for PGEs in the mantle peridotites.

PGE contents of the Skenderbeu peridotites show depletion of Ir relative to Os and Ru, and depletion of Pt relative to Rh and Pd (Fig. 7). The depletion of Pt relative to Rh and Pd is consistent with the PGEs characteristics of interstitial sulfides (Alard et al., 2000). As IPGEs remain in mantle residues during partial melting, the high IPGE concentrations suggest that these rocks are likely to be refractory mantle peridotites. In the Ir-Ir/(Pt + Pd) diagram (Fig. 10A), all of the analyzed peridotite samples plot in the anhydrous mantle peridotite field, indicating that they are partial melting residues of anhydrous mantle peridotites rather than ultramafic cumulates (Hattori et al., 2010). Because the interstitial sulfides (rich in Pd) are easily melted, the Pd/Ir ratios of the melt residue reflect the degree of partial melting (Woodland et al., 2002). The Skenderbeu peridotites should have much lower Pd_N/Os_N and Pd_N/Ir_N ratios than the primitive mantle because they underwent variable degrees of partial melting (Fig. 10B). The relatively low Pd_N/Os_N and Pd_N/Ir_N ratios were likely caused by precipitation of PPGE-rich sulfides as a result of melt-rock reaction processes (Alard et al., 2000). The lower Pd_N/Ir_N ratios of dunites (lower than those of the harzburgites) indicate that dunites underwent higher degrees of partial melting than the harzburgites (Fig. 10B; Woodland et al., 2002), as discussed herein.

The presence of sulfide phases in different stages of mantle melting affects the PGE abundances, which are highly compatible in sulfides (Barnes et al., 1997). Varying amounts of sulfide separation from or input into the upper mantle may be an important factor determining the PGE contents of both mantle residues and melts (Rehkämper et al., 1999; Bockrath et al., 2004). Figure 10 shows PGE (except Os) and Cu abundances in our peridotite samples against their Al₂O₃ content, which is indicative of the degree of partial melting. On the same diagram, theoretically calculated depletion trends are also shown (Marchesi et al., 2013). The Skenderbeu harzburgites appear to be depleted ∼15%–20% with respect to the primitive upper mantle, but their Pd contents with similar depletion degrees are observed above the modeled mantle trend (Fig. 10G). The Skenderbeu dunite samples are more depleted (35%–40% according to their Al₂O₃ contents) with respect to the harzburgites. However, the PGE concentrations of the dunites do not follow the melting trend and show, instead, significant enrichments, which likely resulted from the interaction of melts with the overlying mantle. We thus postulate that the dunites were formed by late-stage partial melting of harzburgites in a SSZ environment that produced boninitic magmas. In this late stage, boninitic melt would be enriched especially in incompatible PGEs (Rh, Pt and Pd) due to their addition from the subducted slab. The interaction of this boninitic melt would enrich the composition of the upper mantle in these elements (Figs. 10C–10H).

Compatible PGEs, such as Os, Ir and Ru, also partition to this boninitic melt as a result of melting of pelagic sediments derived from the subducting slab that commonly contain high concentrations of siderophile elements (e.g., Cu, Mo, and PGEs; Woodland et al., 2002). Therefore, these PGE compositions cannot be ascribed only to melt extraction by partial melting, but also provide strong evidence in support of different degrees of mantle metasomatism and melt percolation. Our analyzed samples reveal the presence of a slightly positive trend from Os to Pd (Fig. 7) that is inconsistent with the nature of the residual mantle material, but can be explained with the inferred melt reaction-enrichment process discussed here. Strong enrichment of Cu in the Skenderbeu peridotites is observed in Figure 10H, suggesting that these peridotites underwent relatively strong mantle metasomatism, resulting in the precipitation of Cu-Ni sulfides in the peridotites (Alard et al., 2000; Marchesi et al., 2013).

We have calculated the Al₂O₃ contents and FeO/MgO ratios of the parental melts for the Skenderbeu harzburgites, using the Maurel and Maurel (1982) and Augé (1987) formulae. The results are presented in Table 6. The calculated Al₂O₃ contents are in the range of 16.3–16.7 wt%, which is the typical range of variation for MORB-type rocks (Al₂O₃, 16 wt%; FeO/MgO, 1.2–1.6; Wilson, 1989) (Fig. 11). However, the Al₂O₃ contents of dunitic peridotites (Al₂O₃ 11.9–13.5 wt%) are similar to those observed from boninitic rocks (Al₂O₃, 10.6–14.4 wt%; FeO/MgO, 0.7–1.4; Wilson, 1989) (Fig. 11), as well as those from chromitites (Al₂O₃ = 10.2–11.2 wt%). The calculated FeO/MgO ratios in parental melts display a wide range of variations (Fig. 11). Nonetheless, the calculated FeO/MgO ratios are comparatively lower in the Skenderbeu chromitites (0.77–1.02) than in the harzburgites (1.28–1.71); this likely indicates derivation of chromitites from boninitic magmas and derivation of the harzburgites from MORB-type magmas. However, the FeO/MgO ratios in parental melts of the Skenderbeu dunites are high (1.48–2.35; Fig. 11). We think that this phenomenon might have resulted from subsolidus reequilibration (Augé, 1987).

It has been widely accepted that high-Cr chromitites (Cr# > 60) form as a result of interaction of boninitic melts with refractory peridotites, whereas high-Al chromitites (Cr# < 60) precipitate from less refractory, MORB-like tholeiitic magmas (Arai and Yurimoto, 1994; Zhou et al., 1996; Matsumoto et al., 1997). Our results are consistent with the findings of previous studies, and indicate that the calculated parental melts of the Skenderbeu harzburgites are compatible with their formation through MORB production, whereas the parental melts of the dunites and chromitites are compatible with the generation of boninitic melts produced in a forearc tectonic setting (Fig. 11). Experimental studies have shown that formation of Cr-rich podiform chromitites as a result of melt-rock reactions between boninitic melts and refractory peridotites requires unusually high temperatures, low pressures (0.3–1.04 Pa), and high water pressure conditions (Umino and Kushiro, 1989; Van der Laan et al., 1989; Klingenberg and Kushiro, 1996; Falloon and Danyushevsky, 2000). These physical conditions are generally met during the initiation of intraoceanic subduction zones (Van der Laan et al., 1989; Stern and Bloomer, 1992; Dilek and Flower, 2003).

We present a tectonomagmatic model for the melt evolution of the Skenderbeu peridotites based on our data and the information available from the Mirdita ophiolites (Fig. 12). The tectonic evolution of the Mirdita ophiolite and the conjugate passive margins of Apulia and Pelagonia within the regional Tethyan geodynamic framework have been discussed in detail in the literature (e.g., Dilek et al., 2007, 2008). Here we follow the geodynamic model of Dilek et al. (2008), suggesting that the Mirdita and the coeval ophiolites in the Western Hellenides evolved in a marginal basin (Pindos-Mirdita basin) between the Pelagonia and Apulia microcontinents during the Mesozoic. The upper mantle beneath this basin may have been in part subcontinental mantle from the northern edge of western Gondwana (from which Pelagonia and Apulia were derived; Dilek et al., 2007). The seafloor spreading–generated oceanic upper mantle of this basin was already depleted, however, following the extraction of MORB melts, which produced its latest Triassic–Early Jurassic oceanic crust.

Our mineralogical and geochemical data show that the Skenderbeu peridotites are variously depleted mantle residues after multiple episodes of partial melting. We infer that the Skenderbeu harzburgites developed from low-degree partial melting (10%–15%) of this already depleted MORB mantle (Fig. 12A). This episode of partial melting took place as the preexisting oceanic lithosphere of the Pindos-Mirdita basin started to sink, causing the mantle to rise and undergo decompression melting. We think that the observed LREE and HFSE enrichments in the harzburgites resulted from interactions of ascending decompressional melts with the earlier melt residue. This interpretation is consistent with our calculations of parental melts of the harzburgites suggesting their development through MORB production. The low degree of partial melting of the Skenderbeu harzburgites in turn produced the MORB lavas in the WMO extrusive sequence (Fig. 12A).

With the establishment of the intraoceanic subduction zone in the Pindos-Mirdita basin, slab-derived fluids were released into the mantle wedge, triggering a higher degree (20%–25%) of partial melting of the refractory harzburgites at shallow depths (Fig. 12B). This high-degree, low-pressure–high-temperature partial melting of the highly depleted harzburgites produced olivine- and SiO₂-rich boninitic melts (enriched in incompatible PGEs and Cu-Ni sulfides) that reacted with the peridotites, forming the podiform chromitite deposits surrounded by dunite. Partial melting of subducted pelagic sediments on the Pindos Mirdita ocean floor (Fig. 12B) provided the boninitic melts with high concentrations of Cu, Mo, and PGEs, which played a significant role in mantle metasomatism and enrichment. Percolation of these incompatible element–enriched boninitic melts through the dunites precipitated Cu-Ni sulfides and melt channels. Progressive evolution of melts at this stage of the formation of the Mirdita ophiolite generated the sheeted dikes and extrusive rocks with IAT through boninite geochemical affinities (Dilek et al., 2007, 2008).

The Skenderbeu peridotites display textural and geochemical evidence for multiple episodes of partial melting, melt extraction, depletion, and refertilization at various scales. The less depleted harzburgites represent restites of low-degree partial melting in a MOR setting. Their chromian spinels with low Cr# and low Ti contents also support this interpretation. The spoon-shaped trace element patterns of the LREE-enriched harzburgites reflect rock-melt interactions between residual abyssal peridotites and ascending asthenospheric melts.

The Skenderbeu dunites developed during high-degree partial melting and melt interactions with a previously depleted mantle (harzburgites) in a forearc setting. Cr-rich podiform chromitites formed as a result of subduction-generated fluids and melts reacting with these refractory peridotites under high-temperature, low-pressure conditions. These processes produced boninitic magmas, enriched in incompatible PGEs and Cu-Ni sulfides and infiltrated into dunitic peridotites as irregular melt channels. The Skenderbeu peridotites underwent multiple episodes of partial melting, depletion, enrichment, and refertilization processes during the evolution of the Jurassic Tethyan oceanic lithosphere at a tectonic setting that changed in time from a MOR to a forearc spreading setting with subduction initiation within intraoceanic conditions. Similar mantle melt evolution patterns through multiple episodes of partial melting have been also documented from some of the other Tethyan ophiolites in the eastern Mediterranean region (e.g., Uysal et al., 2012).

We thank Yazhou Tian, Lan Zhang, and our colleagues in Polytechnic University of Tirana, Albania, and in the Geological Survey of Albania for their assistance with the logistics of our field work in the Skenderbeu massif. We thank He Rong for his assistance with the electron probe microanalyses and the China National Research Center for the geochemical analyses. We also thank David G. Gee (Sweden) for his valuable comments on an earlier draft of the manuscript. This research was funded by grants from the International Geoscience Programme (IGCP 649, 2015–2020), the National Natural Science Foundation of China (41541017, 41641015), the Ministry of Science and Technology of China (2014DFR21270), the China Geological Survey (12120115026801, 12120115027201, 201511022, DD20160023-01), and the Fund from the State Key Laboratory of Continental Tectonics and Dynamics, China (Z1301-a20 and K201502). We thank an anonymous referee, who provided insightful and constructive reviews on our manuscript.