In the Mesoproterozoic Kunene Intrusive Complex (Angola–Namibia) one of the largest massif-type anorthosite complexes in the world, two successively emplaced anorthosite varieties can be recognized. The older white anorthosite suite consists of pyroxene-bearing anorthosite and leucogabbronorite, whereas the younger dark anorthosite suite is dominated by olivine-bearing anorthosite and leucotroctolite. The oxygen isotope and trace element characteristics of plagioclase and minor phases of the two distinct anorthosite varieties are consistent with their derivation from a mantle-derived parental melt that was subject to crustal contamination during the early stages of evolution. The trace element zoning patterns of plagioclase demonstrate that the older white anorthosite was derived by simultaneous fractional crystallization of plagioclase, pyroxene and Fe–Ti oxides during and after crustal contamination. In contrast, the subsequently emplaced dark anorthosite evolved after replenishment with mantle-derived parental melts by fractional crystallization of plagioclase, olivine and Fe–Ti oxides at magmatic temperatures above 1150°C and experienced less crustal contamination.

Högbomite occurs in spinel–garnet–cordierite–sillimanite-bearing metapelites over an area of 60 km x 12 km in the northern parts of the Betroka belt, southern Madagascar. X-ray diffraction analyses indicate that the högbomites belong to the 2N2S polysome type ‘P6₃cm’ with a = 5·718–5·723 Å and c = 18·30–18·42 Å. The chemical composition of högbomite varies over a relatively wide range (in wt %): TiO₂ = 3·82–6·09, Al₂O₃ = 59·45–63·04, Cr₂O₃ ≤ 0·63, FeO^Total = 20·14–30·7, MgO = 2·19–5·76, MnO ≤ 1·35, ZnO ≤ 8·78, and CaO ≤ 0·27. The högbomites form solid solutions involving the substitutions (Mg + Fe²⁺ + Mn + Ca) = Zn, 2(Al + Cr + Fe³⁺) = Ti + (Mg + Fe²⁺ + Mn + Ca + Zn), and (Al + Cr) = Fe³⁺. The högbomite was formed in polyphase domains hosted by discrete magnetite, ilmenite, and/or hematite in close association with spinel, corundum, and/or rutile. Petrographic characteristics and the compositions of högbomite suggest that this mineral was formed by various retrogressive reactions in different microdomains. Högbomite forms at the expense of corundum ± spinel ± ilmenite ± Ti-hematite ± rutile + H₂O and exhibits systematic compositional relationships with the neighboring spinel in Fe²⁺–Mg, Fe³⁺–Al and Fe²⁺–Zn, suggesting attainment of equilibrium between these two minerals. The högbomite-bearing rocks underwent ultrahigh-temperature (UHT) crustal metamorphism following a clockwise P–T trajectory. The textural features suggest the dehydration melting reaction biotite + sillimanite + quartz = cordierite + garnet + melt. The peak mineral assemblage garnet + cordierite + spinel + sillimanite + quartz was stabilized at 880–1060°C and 4·5–6·6 kbar. Decompression cooling is evident from a reaction forming cordierite at the expense of spinel and quartz and a cordierite–spinel corona formed between garnet and sillimanite by the reaction garnet + sillimanite = cordierite + spinel. Högbomite crystallization took place after the peak metamorphic event, probably accompanied by anatectic hydrous melts formed during metamorphism or introduction of hydrous fluids along shear zones over a wide area. We propose that högbomite can evidently grow even at high metamorphic grade (T ≥ 800°C and P ≥ 4 kbar) if a source of water is present, and is an indicator of significant hydrous melt or fluid activity during the early retrogressive stage of regional metamorphism.

The build-up and eruption of crystal-rich ignimbrites, commonly referred to as ‘Monotonous Intermediates’, has been attributed to the incremental addition of new magma batches and the episodic partial melting and reactivation of rheologically locked crystal-rich magma bodies (mushes). In this study, we explore the role of volatiles exsolved from a hot intrusion underplating a crystal mush on the thermal evolution of the coupled mush–intrusion system. We solve the enthalpy conservation equation for the mush and the intrusion and investigate the exsolution of volatiles from the intrusion and their transport through the mush in one dimension. Our calculations span a range of pressures (from 1 to 4 kbar), mush composition (andesite to rhyolite) and initial water contents (from 1·5 to 6 wt %). The mobility of volatiles in the mush is controlled by the volume fraction of the pore space they occupy, and, as a consequence, by the amount of melting and the injection rate of volatiles from the intrusion to the mush. We find that volatiles affect the melting of the mush (or defrosting) in two opposite ways depending on pressure and the initial water content of the intrusion. When the intrusion volatile content is high and pressure relatively low (>4 wt % H₂O at 2 kbar), the mass transfer of volatiles from the intrusion to the mush carries enthalpy beyond the melting front and can thus enhance defrosting and possibly remobilization of the mush. For lower initial volatile contents (< 4 wt % H₂O) in the intrusion and/or higher pressure (3–4 kbar), volatiles stall at the interface between the two magma bodies and prevent defrosting as they thermally insulate the mush from the intrusion. We propose that the dual role played by volatiles during the thermal evolution of crystal-rich magmas reheated by an underplating intrusion can explain the presence of crystal-rich ignimbrites in arcs and their absence in drier hotspots or extensional regions.

The ~50 Myr old Izu–Bonin–Mariana (IBM) arc consists mostly of Oligocene middle and lower crust that underlies the upper crust; these units are in turn covered by Quaternary volcanic rocks. Seismic imaging, forearc geology, Ocean Drilling Program drilling and magnetic anomalies suggest that most IBM arc crust was created in Eocene–Oligocene times. However, remnants of this old crust have never been found at the northern end of the arc, where it is colliding with the Honshu arc (Izu collision zone). Two batholiths in this collision zone (the Tanzawa tonalites and the Kofu Granitic Complex) were emplaced during the Miocene (4–17 Ma). Major elements, Zr/Y, rare earth element ratios and normalized abundance patterns, and Sr–Nd isotopic data indicate that these plutonic bodies are compositionally similar to the Oligocene IBM volcanic rocks, and that they are dissimilar to the Miocene, Pliocene and Quaternary IBM lavas and volcaniclastic rocks. We suggest that the Miocene plutonic rocks in the Izu collision zone were derived from partially melted Oligocene middle crust. A model is proposed in which IBM arc middle crust in the collision zone was partially melted during the collision and then intruded into the overlying upper crust of the Honshu and IBM arcs. This resulted in the complete loss of chronological information related to their original source.

To shed light on the mechanisms of crustal anatexis, a detailed geochemical study has been conducted on minerals and glasses of quenched anatectic metapelitic enclaves and their host peraluminous dacites at El Hoyazo, SE Spain. Anatectic enclaves, composed of plagioclase + biotite + sillimanite + garnet + glass ± K-feldspar ± cordierite + graphite, formed during the rapid heating and overstepped melting of a greenschist-facies metapelite, and finally equilibrated at 850 ± 50°C and 5–7 kbar. Glass appears as melt inclusions within all mineral phases and in the matrix of the enclaves, and has a major element composition similar to that of peraluminous leucogranites. Melt inclusions and matrix glasses have normative quartz–orthoclase–albite compositions that plot in the vicinity of H₂O-undersaturated haplogranite eutectics. Melt inclusions show some compositional variability, with high Li, Cs and B, low Y, first row transition elements (FRTE) and rare earth elements (REE), and zircon and monazite saturation temperatures of ~665–750°C. They are interpreted as melts produced by muscovite-breakdown melting reactions at the onset of the process of rapid melting and mostly under H₂O-undersaturated conditions. Compared with melt inclusions, matrix glasses show less compositional variability, lower large ion lithophile element contents, higher Y, FRTE and REE, and higher zircon and monazite saturation temperatures (~695–815°C). They are interpreted as former melts recording the onset of biotite dehydration-melting. Matrix glasses in the dacite are compositionally different from glasses in the enclaves, hence the genetic connection between metasedimentary enclaves and dacite is not as straightforward as previous petrographic and bulk major element data suggest; this opens the possibility for some alternative interpretation. This study shows the following: (1) melt inclusions provide a window of information into the prograde evolution of anatexis in the enclaves; (2) melting occurred for the most part under H₂O-undersaturated conditions even if, because of the rapid heating, the protolith preserved most of the structurally bound H₂O contained at greenschist facies up to the beginning of anatexis, such that the excess H₂O maximized the amount of H₂O-undersaturated melt generated during anatexis; (3) although a large proportion of accessory minerals are currently shielded within major mineral phases, they have progressively dissolved to a considerable extent into the melt phase along the prograde anatectic path, as indicated by the relative clustering of accessory mineral saturation temperatures and closeness of these temperatures to those of potential melting reactions; (4) the dacite magma was probably produced by coalescence of melts derived from several compositionally distinct metasedimentary protoliths.

Graphite inclusions in diamond, K-bearing clinopyroxene, garnet and kyanite and graphite coatings around microdiamonds from the ultrahigh-pressure Kokchetav Massif, northern Kazakhstan were investigated by means of scanning electron microscopy, laser Raman spectroscopy and X-ray diffraction. Important observations include the increasing size of the graphite crystals from the diamond–graphite interface to the external part of the graphite coatings, the high volume fraction of graphite in the graphite coatings, the presence of composite crystals of successively formed diamond core–graphite mantle zone–diamond rim–external graphite coating, the spherulitic texture of the graphite coatings, the formation of both ordered and disordered graphite and the absence of any fluid associated with the graphite inclusions in diamond and other ultrahigh-pressure metamorphic (UHPM) minerals. These features combined with the presence of oriented graphite flakes in K-bearing clinopyroxene and in diamond-bearing garnet and kyanite and the presence of intergranular diamonds without graphite coatings in the quartzo-feldspathic matrix require reconsideration of the generally accepted diamond graphitization model for interpreting the formation of graphite in UHPM rocks. Metastable growth of graphite within the diamond stability field is suggested and the following crystallization scheme for the C-polymorph is proposed: graphite was the first C-polymorph formed in the diamond stability field, followed by diamond; the graphite coatings around diamond formed during the final stage of UHPM conditions.

Ranges in clinopyroxene trace element contents of 2–3 orders of magnitude occur over <2 cm distance in peridotite samples from the Atlantis II Fracture Zone on the Southwest Indian Ridge. This represents the smallest length-scale at which clinopyroxene trace element concentrations have been observed to vary in abyssal peridotites. Because of the absence of any accompanying veins or other macroscopic features of melt–rock interaction, these peridotites are interpreted as being the result of cryptic metasomatism by a low-volume melt. The small length-scale of the variations, including porphyroclastic clinopyroxene grains of 2 mm diameter with an order of magnitude variation in light rare earth elements, precludes an ancient origin for these anomalies. Calculation of diffusive homogenization time-scales suggests that for the trace element variations to be preserved, metasomatism occurred in the oceanic lithospheric mantle at 1000–1200°C and 10–20 km depth. This observation provides constraints for the on-axis thickness of the lithospheric mantle at an ultraslow-spreading ridge. Trace amounts of plagioclase are present in at least two of the metasomatized samples. Textural and trace element observations indicate that it formed following the trace element metasomatism, indicating that the mantle can be infiltrated multiple times by melt during the final stages of uplift at the ridge axis. The peridotites in this study are from two oceanic core complexes on the Atlantis II Fracture Zone. Our observations of multiple late-stage metasomatic events in the lithospheric mantle agree with current models and observations of melt intrusion into the mantle during oceanic core complex formation. These observations also indicate that heterogeneous lithospheric mantle can be created at ultraslow-spreading ridges.

Results of petrological and geochronological investigations indicate that during the Palaeoproterozoic the Mahalapye Complex of the Limpopo Belt experienced a prograde decompression-heating P–T path from 650°C/6·5 kbar to 800°C/5 kbar, followed by a retrograde decompression-cooling P–T path to 650°C/3 kbar. This P–T–time path is constrained by detailed petrographic observations in concert with P–T pseudosection calculations applied to Si-undersaturated metapelitic rocks, U–Pb ages between 2046 ± 9 and 2025 ± 18 Ma obtained by in situ dating of monazite, xenotime and zircon grains from metasediments, and a garnet Lu–Hf isochron age of c. 2040 Ma. Additionally, combined U–Pb and Lu–Hf isotope data from magmatic zircon grains indicate that the inferred metamorphic evolution was accompanied by the intrusion of granitic rocks derived from both crust and mantle sources. Diorites, which intruded the Mahalapye Complex between 2061 ± 6 and 2040 ± 18 Ma (i.e. prior to or during the metamorphic peak) are characterized by Hf_t values between –0·5 and –3·5, whereas granodiorites and granites with intrusion ages between 2039 ± 9 and 2026 ± 10 Ma have significantly lower Hf_2·04Ga between –9·4 and –12·6. These data suggest that the diorites were derived predominantly from a Palaeoproterozoic mantle source, whereas the granites were formed by re-melting of Archaean crust. This history is consistent with a model suggesting that the Mahalapye Complex was underplated and intruded by mantle-derived mafic melts between 2·06 and 2·04 Ga and that these melts caused partial melting of the adjacent continental crust. Formation and ascent of voluminous crustal melts caused heat transfer into the middle and upper crust and led to the observed high-grade metamorphic overprint. In a wider geological context, the inferred magmatic underplating at 2·06–2·04 Ga was caused by mantle melting beneath the Mahalapye Complex, which was related to the formation of the Bushveld Igneous Complex and/or to subduction processes during the Kheis–Magondi Orogeny.

Late shield-stage silicic (icelandite and rhyodacite) lavas and dikes are exposed within the caldera region of Wai‘anae Volcano, O’ahu, and comprise the most continuous silicic, subalkaline and transitional suites in Hawai‘i. The silicic rocks can be divided into two groups. Group I consists of transitional icelandite dikes and icelandite to basaltic icelandite lava flows that follow a tholeiitic liquid line of descent. Group II includes the Mauna Kuwale Rhyodacite and several intermediate-silica lavas that follow a K-rich, calc-alkaline evolutionary trend. Group II lavas are highly porphyritic, with biotite and hornblende in the rhyodacites, and are selectively enriched in Rb, U and Pb. Pb, Sr and Nd isotope ratios of a Group I dike are similar to those of other Wai‘anae basaltic samples, whereas Group II lavas have lower ²⁰⁶Pb/ ²⁰⁴Pb, ²⁰⁷Pb/ ²⁰⁴Pb, and _Nd, and high ratios of radiogenic ²⁰⁸Pb*/²⁰⁶Pb*, similar to some lavas from Ko’olau, Lana‘i, and Kaho’olawe volcanoes. Most Group I dikes define a trend consistent with fractional crystallization at ~0·5 kbar with low initial water (0·3 wt %) contents, whereas the chemical trends for Group II samples cannot be reproduced by simple crystallization models. These results, together with trace element modeling and isotopic characteristics, suggest derivation of the Group II silicic flows by hydrous melting of lower Hawaiian crust that had been altered to amphibolite. Crustal melting toward the end of the shield stage may be related to an increase in lithostatic pressure following rapid accumulation of lavas in the shield stage and ponding of basaltic magmas at depths below the level of crack closure (~9–10 km). Contemporaneously, Group I silicic magmas formed by shallow-level crystallization as a consequence of decreased magma supply to the upper crust as a result of waning activity at the end of the shield stage. Geochemical evidence indicates significant mixing between the different types of silicic magma and between basaltic and silicic magmas at this stage of volcano evolution. Although melting of recycled crustal material entrained in the upwelling Hawaiian plume has previously been proposed, this is the first documentation of a role for melting of lower Hawaiian crust in the genesis of Hawaiian volcanoes.

The Mohelno peridotite is a medium-sized ultramafic body (2 km x 4 km in size) enclosed in the Gföhl granulites in the eastern part of the Bohemian Massif. It consists mainly of coarse spinel peridotite (harzburgite and dunite); garnet peridotite occurs only in the sheared and deformed margins of the body. To decipher the origin and history of this mantle-derived peridotite, we determined the mineral chemistry by electron microprobe analysis and olivine fabric patterns by the electron backscattered diffraction method for each rock type. We found two distinct types of olivine fabric (crystal-preferred orientation; CPO) in the peridotite, which can be correlated with the mineralogy and thermal history of each. The olivine CPO in coarse-grained spinel peridotite shows a strong concentration of [100] slightly oblique to the lineation and [010] and [001] girdles normal to the lineation (which is the so-called {0kl}[100] pattern typical of medium-temperature deformation). Olivine in coarse-grained garnet peridotite, on the other hand, shows a strong concentration of [010] normal to the foliation and a concentration of [100] parallel to the lineation (which is the so-called (010)[100] pattern typical of high-temperature deformation). These fabric patterns become diffuse as the grain size is reduced for each mineralogical type. We interpret the development of these contrasting fabric patterns and mineralogical types based on the pressure–temperature history of each rock type determined by applying published geothermometers and geobarometers to the constituent minerals. Starting from a high-temperature (>1200°C) spinel peridotite, during exhumation and cooling in contact with surrounding granulites, the marginal part of the body was transformed to garnet peridotite, whereas the interior remained in the spinel-peridotite facies because cooling was slower inside the body. Because of the slow cooling and continuous deformation in the interior of the body, the original high-temperature fabric pattern in the spinel peridotite was converted to a lower-temperature type. The high-temperature fabric was preserved only at the margin of the body where cooling was more rapid. Reduction of grain size that occurred during later, low-temperature, deformation partly obliterated the high-temperature fabric patterns for both garnet and spinel peridotites. The initial rapid cooling at high temperatures associated with deformation probably occurred after the mantle peridotite was emplaced within the crustal granulites, which implies that the spinel- to garnet-peridotite transformation took place in the continental crust.

Tabular dunites in the Josephine peridotite of southwestern Oregon represent conduits for melt extraction from the upper mantle. The amount of melt trapped within these channels during cooling and exhumation of the peridotite massif was calculated using a mass-balance approach. These calculations suggest that no more than 0·1–0·2 vol. % melt could have been trapped within these dunites after crystallization. This observation indicates that the threshold porosity for melt extraction is ≤0·2 vol. %. We use this constraint to evaluate published permeability models for partially molten dunite. We then use constraints from these permeability models to calculate lower bounds on the bulk viscosity of partially molten dunite and conclude that the bulk viscosity is of the order of 10²³ Pa s at melt contents of ~0·1 vol. % and reasonable melt extraction column heights. The small amount of trapped melt in these dunites is consistent with geochemical evidence for near fractional melting in the upper mantle.

We present the results of a series of anhydrous piston cylinder experiments that illustrate the mechanisms and implications of reaction between tholeiitic melt and depleted peridotite in the uppermost mantle. To simulate infiltration–reaction processes we have applied a three-layer setup in which a layer of primitive basaltic powder (‘melt layer’) is consecutively overlain by a ‘peridotite layer’ and a layer of vitreous carbon spheres (‘melt trap’). The peridotite layer is mixed from pure separates of orthopyroxene, clinopyroxene and spinel (Balmuccia peridotite), and San Carlos olivine. Two tholeiitic melt compositions, respectively with compositions in equilibrium with lherzolitic (ol, opx, cpx) and harzburgitic (ol, opx) residues after partial melting at 1·5 GPa, were employed. Melt from the melt layer is forced to move through the peridotite layer into the melt trap. Experiments were conducted at 0·8 GPa with peridotite of variable grain size, in the temperature range 1200–1320°C and for run durations of 10 min to 92 h. In this P–T range, representing conditions encountered in the transition zone between the thermal boundary layer and the top of the asthenosphere below oceanic spreading centers, the melt is subjected to fractionation and the peridotite is partially melting (T_s ~1260°C). Modal observations indicate a strong dependence between phase relations in the melt layer and changes in the modal abundances of the peridotite layer, as a function of both temperature and melt composition. Textural and compositional evidence, as well as modeling of Fe–Mg profiles in olivine, demonstrates that reaction between percolating melt and peridotite occurs by a combination of dissolution–reprecipitation and solid-state diffusion. Dissolution–reprecipitation leads to well-equilibrated phases whereas diffusional equilibration introduces zoning at experimental timescales. We discuss the observed reaction mechanisms and the consequent compositional changes in the light of local chemical equilibria and reaction kinetics. The results have direct implications for melt migration in upper-mantle thermal boundary layers.

Peridotite xenoliths in arc volcanoes are very rare, usually small and remain poorly studied. Much of the earlier work focused on peridotites affected by re-crystallization, metasomatism and veining that took place shortly before the eruption of the host magmas; such lithologies may not be widespread in the mantle wedge. This study reports petrographic, major and trace element data for 17 large, fresh peridotite xenoliths from the active Avacha volcano and discusses the origin of supra-subduction zone lithospheric mantle, in particular the role and characteristics of partial melting and metasomatism. The xenoliths are spinel harzburgites containing interstitial cpx (1·5–3%) and amphibole (≤1%). Nearly all are medium- to coarse-grained with protogranular to granoblastic microstructures; some have fine-grained domains and thin cross-cutting veins of secondary opx and olivine. Core–rim zoning and unmixing of cpx and spinel in coarse opx indicate long-term cooling to ≤900–1000°C; Cr#_Sp and Al and Cr in opx are correlated with equilibration temperatures. The peridotites are highly refractory, with ≥44% MgO and very low Al₂O₃ and CaO (0·4–0·9%), TiO₂ (≤0.01%), Na₂O (≤0.03%), K₂O and P₂O₅ (below detection) and REE in whole-rocks, ≤2·1% Al₂O₃ in opx and ≤0·1–0·3% Na₂O in cpx. Comparisons of Mg, Al and Fe contents with melting experiments indicate 28–35% melt extraction at ≤1 to 2 GPa, in line with the absence of primary cpx and high Mg#_Ol (0·907–0·918) and Cr#_Sp (0·53–0·65). Bulk-rock Al₂O₃ is a more robust melt extraction index than Cr#_Sp, Mg#_Ol and Mg#_WR. Forearc harzburgites and certain xenoliths from the Western Pacific share many of these characteristics with the Avacha suite and may have similar origins. A distinctive feature of the Avacha harzburgites is a combination of variable but commonly high modal opx (18–30%) with very low modal cpx. At a given olivine or MgO content, they have higher opx and SiO₂, and lower cpx (as well as Al₂O₃ and CaO) than typical refractory peridotite xenoliths in continental basalts. These features may indicate fluid fluxing during melting in the mantle wedge. Alternatively, they could have been produced after partial melting by selective metasomatic enrichment in SiO₂ by fluids to transform some olivine into opx, although the latter mechanism is hard to reconcile with the very low alkalis and REE contents and the absence of silica correlation with fluid-mobile elements. Bulk-rock enrichments in silica and opx are unrelated to the presence or abundance of late-stage, fine-grained materials and are due to an ancient event rather than recent re-crystallization and veining.

Two types of olivine occur in kimberlites from Greenland, Canada and southern Africa. The first, xenocrystic olivine, displays several different forms. Most distinctive are ‘nodules’, a term we use to describe the large (1–5 mm), rounded, single crystals or polycrystalline aggregates that are a common constituent of many kimberlites. Olivine compositions are uniform within single nodules but vary widely from nodule to nodule, from Fo81 to 93. Within many nodules, sub- to euhedral asymmetric tablets have grown within larger anhedral olivine grains. Dislocation structures, particularly in the anhedral grains, demonstrate that this olivine was deformed before being incorporated into the kimberlite magma. Olivine grains in the kimberlite matrix between the nodules have morphologies similar to those of the tablets, suggesting that most matrix olivine grains are parts of disaggregated nodules. We propose that a sub- to euhedral form is not sufficient to identify phenocrysts in kimberlites and provide some criteria, based on morphology, internal deformation and composition, that distinguish phenocrysts from xenocrysts. The second type of olivine is restricted to rims surrounding xenocrystic grains. Only this olivine crystallized from the kimberlite magma. Major and trace element data for the rim olivine are used to calculate the composition of the parental kimberlite liquid, which is found to contain between about 20 and 30% MgO. The bulk compositions of many kimberlites contain higher MgO contents as a result of the presence of xenocrystic olivine. The monomineralic, dunitic, character of the nodules, and the wide range from Fo-rich to Fo-poor olivine compositions, provide major constraints on the origin of the nodules. Dunite is a relatively rare rock in the mantle and where present its olivine is persistently Fo-rich. The dunitic source of the nodules in kimberlites lacked minerals such as pyroxene and an aluminous phase, which make up about half of most mantle-derived rocks. It appears that these minerals were removed from the mantle peridotite that was to become the source of the nodules, and the Fo content of the retained olivine was modified during interaction with CO₂-rich fluids whose arrival at the base of the lithosphere immediately preceded the passage of the kimberlite magmas. Fragments of the resultant dunite were entrained into the kimberlite, where they were retained both as intact nodules and as disaggregated grains in the matrix.

To better understand the dynamical processes of partial melting, melt transport, and melt–rock reaction in the mantle, mass conservation equations for a two-porosity double lithology continuum have been developed using the method of volume averaging. Here the region of interest is treated as two overlapping continua occupied by a low-porosity peridotite matrix and a high-porosity dunite channel network. Conservation equations for the matrix and channel continua are coupled through mass transfer terms that include matrix dissolution and diffusive and advective mixing between the melt in the channel and that in the matrix. Essential features of the two-porosity double lithology model have been investigated through simple case studies. In general, the composition of the channel melt is a weighted average of the matrix melt extracted along the melting column (as a result of melt suction) and the matrix materials dissolved into the channel melt (as a result of matrix–channel transformation). Dissolution of pyroxene in the peridotite matrix into the channel melt and precipitation of olivine lowers the compatible trace element abundances in the channel melt. In the absence of matrix dissolution, the channel melt is equivalent to the aggregated or pooled matrix melt. The incompatible trace element abundance in the channel melt is dominated by the less depleted small-degree melts from the lower part of the melting column and not very sensitive to the details of how the melt suction rate and matrix–channel transformation rate vary spatially. The incompatible trace element abundances in the matrix melt and solid are very sensitive to the direction of melt flow across the matrix–channel interface, the magnitude and variation of the relative melt suction rate, and the depth of dunite channel initiation, and depend moderately on variations in channel volume fraction in the double lithology region. Percolation of the enriched channel melt into the more depleted residual matrix in the upper part of the melting column may provide a viable mechanism for late-stage melt refertilization or mantle metasomatism. Understanding the first-order characteristics of the channel and matrix melts and solids is essential in deciphering the melting and melt migration histories of residual mantle rocks and erupted basalts.

Cratonic eclogites and garnet pyroxenites from the Kaapvaal craton have heterogeneous Hf–Nd–Sr–(O) isotope ratios that define a positive Hf–Nd isotope array and a negative Nd–Sr isotope array. Isotopic variability encompasses depleted (mid-ocean ridge basalt and ocean-island basalt) to enriched mantle compositions (Group I and II kimberlites) and overlaps with that of the Kaapvaal craton garnet peridotite xenoliths. Isotopic heterogeneity at Roberts Victor is less extreme than previously reported and ranges from eclogites with a highly depleted MORB-like signature to enriched eclogites similar to Group II and transitional kimberlites and Group II megacrysts (_Hf = –32·8). Much of this similarity may well be due to partial or complete resetting during entrainment. For the majority of eclogites and garnet pyroxenites the Lu–Hf system records ‘older’ mantle events than the Sm–Nd system, but neither necessarily records the protolith age. Both the Lu–Hf and the Sm–Nd systems are prone to being reset by entrainment in high-temperature kimberlite and/or basaltic magmas (e.g. Kaapvaal) and emplacement in orogenic belts (e.g. Beni Bousera). In the case of one eclogite from Roberts Victor the Sm–Nd cpx–gt mineral isochron age (963·1 ± 42·3 Ma) differs from the Lu–Hf cpx–gt mineral isochron age (1953 ± 13 Ma) by 1 Ga and the Rb–Sr clinopyroxene model age (3·15 Ga) is 1 Ga older than the Lu–Hf age and the reconstructed whole-rock isochron age. Ironically, it may be that, in this instance, the Rb–Sr system gives a better indication of protolith age than Sm–Nd or Lu–Hf. Overall variable resetting of isotope systems between protolith formation in the Archaean (>2·5 Ga) and kimberlite and/or basalt entrainment (≤ 0·2 Ga) masks our understanding of the exact protolith age of eclogites.

The existence of different mantle domains exposed in ocean–continent transition zones provides a framework for understanding the generation of ultramafic seafloor along magma-poor rifted margins. In this study we present detailed petrological and geochemical data on peridotites from the Eastern Central Alps ophiolites in Switzerland and Italy to identify different mantle domains, to estimate the extent of refertilization, and to test whether refertilization is associated with a thermal signature that has important implications for geophysical interpretations of ocean–continent transitions. The compositions of clinopyroxene, orthopyroxene and spinel clearly reflect the different mantle domains. Relative to clinopyroxenes from spinel peridotites, clinopyroxenes from plagioclase peridotites have lower Na₂O and Sr contents, but higher middle to heavy rare earth element ratios and Zr concentrations, and different Sc–V relationships. Spinels in plagioclase peridotites have higher TiO₂ and lower Mg-numbers compared with those in spinel peridotites. Mineral–mineral trace element partitioning suggests that spinel peridotites equilibrated at substantially lower temperatures than plagioclase peridotites. The temperature difference between the spinel and plagioclase peridotites indicates an important thermal boundary between the two. The geochemical data show systematic spatial variations. A heterogeneous, ‘subcontinental domain’ with no syn-rift melt imprint is separated from a ‘refertilized domain’ that exhibits a complex history of regional-scale melt infiltration and melt–rock reaction, which has erased most of the ancient history. Simple calculations suggest that up to 12% of mid-ocean ridge basalt-type melt can be stored in plagioclase peridotite, relative to a depleted residue. Such a ‘lithospheric sponge’ provides an explanation for the fertile compositions of the peridotites and the rare occurrence of volcanic rocks in magma-poor rifted margins. We suggest that magma-poor vs magma-rich margins are largely determined by the efficiency of melt extraction and not so much by melt generation processes, given a similar initial composition of the upwelling mantle. It is proposed that refertilization increases textural diversity and chemical heterogeneity related to shallow crystallization in the mantle lithosphere.

Stromboli is known for its persistent degassing and rhythmic strombolian activity occasionally punctuated by paroxysmal eruptions. The basaltic pumice and scoria emitted during paroxysms and strombolian activity, respectively, differ in their textures, crystal contents and glass matrix compositions, which testify to distinct conditions of crystallization, degassing and magma ascent. We present here an extensive dataset on major elements and volatiles (CO₂, H₂O, S and Cl) in olivine-hosted melt inclusions and embayments from pyroclasts emplaced during explosive eruptions of variable magnitude. Magma saturation pressures were assessed from the dissolved amounts of H₂O and CO₂ taking into account the melt composition evolution. Both pressures and melt inclusion compositions indicate that (1) Ca-basaltic melts entrapped in high-Mg olivines (Fo_89–90) generate Stromboli basalts through crystal fractionation, and (2) the Stromboli plumbing system can be imaged as a succession of magma ponding zones connected by dikes. The 7–10 km interval, where magmas are stored and differentiate, is periodically recharged by new magma batches, possibly ranging from Ca-basalts to basalts, with a CO₂-rich gas phase. These deep recharges promote the formation of bubbly basalt blobs, which are able to intrude the shallow plumbing system (2–4 km), where CO₂ gas fluxing enhances H₂O loss, crystallization and generation of crystal-rich, dense, degassed magma. Chlorine partitioning into the H₂O–CO₂-bearing gas phase accounts for its efficient degassing (≥69%) under the open-system conditions of strombolian activity. Paroxysms, however, are generated through predominantly closed-system ascent of basaltic magma batches from the deep storage zone. In this situation crystallization is negligible and sulfur exsolution starts at ≤170 MPa. Chlorine remains dissolved in the melt until lower pressures, only 16% being lost upon eruption. Finally, we propose a continuum in explosive eruption energy, from strombolian activity to large paroxysmal events, ultimately controlled by variable pressurization of the deep feeding system associated with magma and gas recharges.

Macquarie Island (Southern Ocean) is a fragment of Miocene ocean crust and upper mantle formed at a slow-spreading ridge system, uplifted and currently exposed above sea-level. The crustal rocks on the island have unusually enriched compositions and the strong signature of an enriched source requires low overall degrees of melt depletion in the underlying mantle. Peridotites on the island, however, are highly refractory harzburgites that can be modeled as residues of >20–25% of near-fractional melting from which all the free clinopyroxene was melted out. The peridotites have some of the highest spinel Cr-numbers (0·40–0·49) and lowest orthopyroxene-core Al₂O₃ concentrations (2·7–3·0 wt %) reported so far for oceanic peridotites. The peridotites were subsequently modified by melt–rock reactions underneath the Miocene ridge system. The refractory character of the peridotites is inconsistent with the slow-spreading ridge setting as well as with the enriched character of the overlying crust, and must indicate a previous depletion event; the peridotites are not the source residue of the overlying ocean crust on Macquarie Island. Osmium isotopic compositions of peridotite samples are very unradiogenic (¹⁸⁷Os/¹⁸⁸Os = 0·1194–0·1229) compared with normal abyssal peridotites and indicate a long-lived rhenium depletion. Proterozoic rhenium-depletion ages indicate that these rocks have preserved a memory of an old mantle melting event. We argue that the Macquarie Island harzburgites are samples from an anciently depleted refractory mantle reservoir that may be globally important, but that is generally overlooked because of its sterility; that is, its inability to produce basalts. This reservoir may preserve key information about the history of the Earth’s mantle as a whole.

The composite intrusions of Drumadoon and An Cumhann crop out on the SE coast of the Isle of Arran, Scotland and form part of the larger British and Irish Palaeogene Igneous Province, a subset of the North Atlantic Igneous Province. The intrusions (shallow-level dykes and sills) comprise a central quartz–feldspar-phyric rhyolite flanked by xenocryst-bearing basaltic andesite, with an intermediate zone of dark quartz–feldspar-phyric dacite. New geochemical data provide information on the evolution of the component magmas and their relationships with each other, as well as their interaction with the crust through which they travelled. During shallow-crustal emplacement, the end-member magmas mixed. Isotopic evidence shows that both magmas were contaminated by the crust prior to mixing; the basaltic andesite magma preserves some evidence of contamination within the lower crust, whereas the rhyolite mainly records upper-crustal contamination. The Highland Boundary Fault divides Arran into two distinct terranes, the Neoproterozoic to Early Palaeozoic Grampian Terrane to the north and the Palaeozoic Midland Valley Terrane to the south. The Drumadoon Complex lies within the Midland Valley Terrane but its isotopic signatures indicate almost exclusive involvement of Grampian Terrane crust. Therefore, although the magmas originated at depth on the northern side of the Highland Boundary Fault, they have crossed this boundary during their evolution, probably just prior to emplacement.