Migmatites are described from the Sillimanite-potash-feldspar Zone of the aureole around the Newer Basic suite of synorogenic intrusions. The lowest-grade migmatites are trondhjemitoid (characterized by the assemblage quartz-plagioclase-biotite) or muscovite-granitoid (quartz-plagioclase potash-feldspar-muscovite-sillimanite-biotite). With increasing grade, a transition occurs to cordierite-granitoid assemblages (quartz-plagioclase-potash-feldspar-cordierite-garnet-sillimanite-biotite), which persist to the highest grades observed, where there are also noritoid migmatites (quartz-plagioclase-orthopyroxene-cordierite-biotite). The trondhjemitoids are texturally simple because the minerals did not undergo dehydration reactions. Textural immaturity and consistently cotectic modal compositions indicate that their leucosomes originated as melts. Scatter of plagioclase compositions suggests that the partial melting occurred in small closed systems. The other migmatites have more fusible compositions, so it is deduced that they also underwent partial melting. Retrograde reaction textures are used to infer the sequence of reactions, involving muscovite and biotite, by which melting proceeded during prograde evolution. Whereas the fugacity of water probably varied among spatially associated trondhjemitoid leucosomes, in the muscovite-granitoids it was constrained to an approximately constant value, at given pressure and temperature, by the buffering effect of the mineral assemblage.

Compositions of glaucophanes and actinolite-hornblende solid solutions occurring in chemically similar metabasaltic rocks from blueschist terranes in east-central Shikoku, W. California, Valtournanche (W. Alps), and W. Liguria are compared. Chemical contrasts among coexisting Na and Ca amphibole pairs, which reflect disparate P-T histories under the presumed attendance of local equilibrium, include: Na contents are rather high among barroisitic hornblendes from the western and Ligurian Alps, as well as among high-grade tectonic blocks from California; in contrast, actinolitic amphiboles from both lower-grade Franciscan tectonic blocks and in situ schists and the blueschists of Shikoku are impoverished in Na relative to blue-green hornblendes. Sodic amphiboles contain less than 0. 5 Aliv per formula unit, whereas Alvi is very high; a situation reversed among calcic amphiboles. The Na + Ca contents ofglaucophanes are strongly clustered around the sum of 2.0 (i.e. A site vacant) whereas calcic amphiboles have a wider range with the A site variably occupied. No solvus has been detected within either sodic or calcic amphiboles under blueschist facies conditions. For low-grade metabasaltic parageneses, a miscibility gap separates these two amphibole groups; at relatively high grade such compositions have sodic calcic amphiboles of barroisitic type; this may mean that glaucophane + hornblende assemblages are metastable, accounting for textural relations indicating that the sodic amphibole typically did not grow at the same time as the barroisite. Ti, Mn, and K appear to be concentrated in calcic amphibole compared to coexisting glaucophane, probably in the M2, M4, and A sites, respectively.

Contrasts in coexisting amphibole tie lines are thought to be a consequence of the fact that the parageneses of Shikoku and California reflect high P and very high P prograde P-T paths respectively, whereas those from Valtournache and W. Liguria show evidence of decompression recrystallization (or back-reaction) of high P (i.e. eclogitic) protoliths. Comparison of the inferred physical conditions operating during the production of these four contrasting paragenetic sequences allows the provisional assignment of a P-T stability region for barroisitic amphibole in metabasaltic rocks as: P 4–5 kb at c. 350°; P 5–7 kb at c. 450 °C.

The proton microprobe has been used to determine contents of Ca, Ti, Ni, Mn and Zn in the olivine of 54 spinel lherzolite xenoliths from Australian and Chinese basalts. These data are compared with proton-probe data for Ni, Mn and Zn in the olivine of 180 garnet peridotite xenoliths from African and Siberian kimberlites. Fe, Mn, Ni and Zn contents are well-correlated; because the spinel lherzolite olivines have higher mean Fe contents than garnet peridotite olivines (average Fo89.6vs. Fo90–92) they also have lower Ni and higher Mn contents. Zn and Fe are well-correlated in garnet peridotite olivine, but in spinel peridotites this relationship is perturbed by partitioning of Zn into spinel. None of these elements shows significant correlation with temperature. Consistent differences in trace-element contents of olivines in the two suites is interpreted as reflecting the greater degree of depletion of Archean garnet peridotites as compared to Phanerozoic spinel lherzolites. Ca and Ti contents of spinel-peridotite olivine are well correlated with one another, and with temperature as determined by several types of geothermometer. However, Ca contents are poorly correlated with pressure as determined by the Ca-in-olivine barometer of Köhler and Brey (1990). This reflects the strong T-dependence of this barometer: the uncertainty in pressure (calculated by this method) which is produced by the ±50°C uncertainty expected of any geothermometer is ca ± 8 kbar, corresponding to the entire width of the spinel-lherzolite field at 900–1200°C.

The thermal dehydration of naturally occurring Ca-montmorillonite has been studied by in situ X-ray diffraction at temperatures between 60–120°C. The time-temperature-dependence of the position of the basal (001) reflection reveals that interlayer water loss on isothermal dehydration occurs in two stages. After an initial rapid decrease in interlayer spacing (on shock heating to an isothermal soak temperature) the reaction proceeds towards equilibrium more slowly. Furthermore, the width of the (001) reflection changes with time, reflecting transformation-dependent changes in homogeneity perpendicular to (001) with a maximum in peak width at the point where the rate of the reaction appears to change. This suggests that, as the interlayer spacing collapses, a local change is induced in the structure, affecting the means of movement of the water from the interlayer.

Although the thermal evolution of the mantle before c. 3.0 Ga remains unclear, since c. 3.0 Ga secular cooling has dominated over heat production—this is time's arrow. By contrast, the thermal history of the crust, which is preserved in the record of metamorphism, is more complex. Heat to drive metamorphism is generated by radioactive decay and viscous dissipation, and is augmented by the influx of heat from the mantle. Notwithstanding that reliable data are sparse before the Neoarchean, we use a dataset of temperature (T), pressure (P) and thermobaric ratio (T/P at the metamorphic ‘peak’), and age of metamorphism (t, the timing of the metamorphic ‘peak’) for rocks from 564 localities ranging in age from the Cenozoic to the Eoarchean eras to interrogate the crustal record of metamorphism as a proxy for the heat budget of the crust through time. On the basis of T/P, metamorphic rocks are classified into three natural groups: high T/P type (T/P >775°C/GPa, mean T/P ~1105°C/GPa), including common and ultrahigh-temperature granulites, intermediate T/P type (T/P between 775 and 375°C/GPa, mean T/P ~575°C/GPa), including high-pressure granulites and medium- and high-temperature eclogites, and low T/P type (T/P <375°C/GPa, mean T/P ~255°C/GPa), including blueschists, low-temperature eclogites and ultrahigh-pressure metamorphic rocks. A monotonic increase in the P of intermediate T/P metamorphism from the Neoarchean to the Neoproterozoic reflects strengthening of the lithosphere during secular cooling of the mantle—this is also time's arrow. However, temporal variation in the P of intermediate T/P metamorphism and in the moving means of T and T/P of high T/P metamorphism, combined with the clustered age distribution, demonstrate the cyclicity of collisional orogenesis and cyclic variations in the heat budget of the crust superimposed on secular cooling since c. 3.0 Ga—this is time's cycle. A first cycle began with the widespread appearance/survival of intermediate T/P and high T/P metamorphism in the Neoarchean rock record coeval with amalgamation of dispersed blocks of lithosphere to form protocontinents. This cycle was terminated by the fragmentation of the protocontinents into cratons in the early Paleoproterozoic, which signalled the start of a new cycle. The second cycle continued with the progressive amalgamation of the cratons into the supercontinent Columbia and extended until the breakup of the supercontinent Rodinia in the Neoproterozoic. This cycle represented a period of relative tectonic and environmental stability, and perhaps reduced subduction during at least part of the cycle. During most of the Proterozoic the moving means for both T and T/P of high T/P metamorphism exceeded the arithmetic means, reflecting insulation of the mantle beneath the quasi-integrated lithosphere of Columbia and, after a limited reorganisation, Rodinia. The third cycle began with the steep decline in thermobaric ratios of high T/P metamorphism to their lowest value, synchronous with the breakup of Rodinia and the formation of Pannotia, and the widespread appearance/preservation of low T/P metamorphism in the rock record. The thermobaric ratios for high T/P metamorphism rise to another peak associated with the Pan-African event, again reflecting insulation of the mantle. The subsequent steep decline in thermobaric ratios of high T/P metamorphism associated with the breakup of Pangea at c. 0.175 Ga may indicate the start of a fourth cycle. The limited occurrence of high and intermediate T/P metamorphism before the Neoarchean suggests either that suitable tectonic environments to generate these types of metamorphism were not widely available before then or that the rate of survival was low. We interpret the first cycle to record stabilisation of subduction and the emergence of a network of plate boundaries in a plate tectonics regime once the balance between heat production and heat loss changed in favour of secular cooling, possibly as early as c. 3.0 Ga in some areas. This is inferred to have been a globally linked system by the early Paleoproterozoic, but whether it remained continuous to the present is unclear. The second cycle was characterised by stability from the formation of Columbia to the breakup of Rodinia, generating higher than average T and T/P of high T/P metamorphism. The third cycle reflects colder collisional orogenesis and deep subduction of the continental crust, features that are characteristic of modern plate tectonics, which became possible once the average temperature of the asthenospheric mantle had declined to <100°C warmer than the present day after c. 1.0 Ga.

The dehydration of gypsum CaSO4.2H2O has been studied, at negligible water vapour pressure, by in situ infrared (IR) spectroscopy and by thermogravimetry to determine whether intermediate phases (CaSO4.nH2O) exist, other than the hemihydrate with n=0.5, and also to compare the mechanism of the dehydration process when measured by two techniques with very different correlation lengths. Thermogravimetry shows an apparently continuous water loss with an activation energy of 90.3 kJ.mol−1, with no changes in the activation energy as a function of the degree of dehydration. IR spectroscopy on the other hand, clearly shows the existence of three discrete phases, gypsum CaSO4.2H2O, hemihydrate CaSO4.0.5H2O and γ-CaSO4, with nucleation of each successive phase as dehydration proceeds. There is no evidence to suggest the presence of phases with any intermediate water content.

Possible relationships between groundwater arsenic concentration and alluvial sediment characteristics in a ∼19 km2 area in West Bengal have been investigated using a combination of hydrogeochemical, lithogeochemical and geophysical techniques. Arsenic hotspots, typically associated with elevated groundwater Fe and Mn, were found to be correlated to some extent with old river channels (abandoned meanders, oxbow lakes), where sandy aquifers included intercalated fine-grained overbank deposits, rich in As, Fe, Mn and Corg. Otherwise no demonstrably significant overall differences in any of lithology, grain-size distribution, mineral composition or Fe, Mn and organic C content of the sediments were found between two representative sites with contrastingly low (<50 μg 1—1) and high (>200 μg 1—1) As groundwater contents.

Our results are consistent with microbially mediated redox reactions controlled by the presence of natural organic matter within the aquifer and the occurrence of As-bearing redox traps, primarily formed by Fe and Mn oxides/hydroxides, being the most important factors which control the release of As into shallow groundwaters at the study site.

A wide range of natural K-, Na-, Ca- or (K + Li)-micas have been systematically examined by Raman spectrometry. The spectra are interpretable in terms of regular variations in peak positions and chemical parameters. Several vibrations give higher wavenumbers for Na-micas compared to K-micas, in accord with the smaller ionic size of Na+ than K+. The ≈195 cm-1 and ≈270 cm-1 peak wavenumbers and intensities vary as functions of the chemistry of the octahedral sites, i.e. the replacement of Mg2+ by Mn2+, Zn2+, Cr3+, Fe3+, Ti4+, and especially by Al3+, or by a vacancy, and the replacement of (OH)- by F-. The group of ≈700 cm-1 peaks vary in wavenumber and intensity with the replacement of Si by Al in the tetrahedra; distinct Si-O-Si and Si-O-Al vibrations can be recognized. Di- and tri-octahedral micas are distinguished on the basis of certain relative peak intensities which vary considerably with polarization direction, and of trends with increasing Al(iv), Al(vi) or Al(tot.). Calibration of these trends for the chemical analysis of mica microinclusions seems feasible once the uncertainties in the data set are resolved by the determination of further samples selected to highlight the effect of specific elements.

A manganese pan near Birness contains grains of an optically uniaxial negative mineral near (Na0·7Ca0·3)Mn7O14·2·8H2O, giving an X-ray powder pattern similar to that of synthetic materials described as ‘manganous manganite’ and δ-MnO2. Material giving a similar pattern has been described from a natural occurrence in Canada, but no mineral name was assigned; the name birnessite is now proposed. The mineral is probably formed by air-oxidation of manganous oxides under alkaline conditions.

The liquidus temperature (1198 °C) and equilibrium phase relations of a sample of Columbia River basalt from the Picture Gorge section have been determined at I atmosphere by heating in a controlled atmosphere. When this basalt is cooled from above its liquidus temperature the liquidus phase (plagioclase) may fail to crystallize depending on the degree of undercooling and the duration of the experiment. A field in temperature-time space in which plagioclase fails to crystallize on cooling is separated from another in which plagioclase always crystallizes by a third in which the nucleation of plagioclase is unpredictable in terms of temperature and time. The extent to which this basaltic liquid can be supercooled without the crystallization of plagioclase is independent of the time it is held above the liquidus or the temperature in excess of the liquidus to which it is heated.

The exceptionally long times required to ensure the nucleation of plagioclase at or near the liquidus temperature suggest that many so-called ‘equilibrium’ phase relations determined from experiments of a few hours' duration could be in serious error if the ‘equilibration’ involves a nucleation process.

It is demonstrated that, over a range of cooling rates, the temperature at which plagioclase begins to crystallize on cooling varies markedly and the temperature and times required for both possible and certain nucleation of plagioclase are calculated for a range of constant cooling rates. The range of cooling rates over which the nucleation temperature of plagioclase varies is likely to occur in nature only in certain lava flows and small minor intrusions. In such cases this could lead to changes in the order in which the minerals appear on cooling and other petrologically significant effects.

A comparison of R.P.M. Union and Amandelbult Sections reveals close geochemical and stratigraphic correlations, but the sequence at the latter is more complex. Mafic members at Union Section are consistently more magnesian than their equivalents at Amandelbult Section and (where available data allow comparisons) than at Rustenburg Section. This, taken together with regional patterns of progressive attenuation and elimination of leucocratic rocks beneath harzburgitic layers, identifies Union Section as proximally located with reference to an irruptive centre from which primitive liquids were injected along the floor/supernatant liquid interface. Successive injections led to thermal erosion of the floor, causing ‘dimpling’ and ‘potholing’ on a local scale, and elimination of noritic-anorthositic layers on a regional scale. Such erosion of the floors beneath new inputs of magmatic liquid is likely to be initiated by only partial remelting, mainly of lower-temperature phases (sodic rims to zoned plagioclase grains, and Fe-enriched pyroxenes) within intercumulus space. It is argued that resultant contamination of the primitive liquids may have led to direct superposition of anorthosites (accompanied by rare troctolite) upon harzburgite, without intervening pyroxenite members, as in the Pseudoreef Multicyclic Unit.

High-temperature geothermal fields in Iceland represent localized anomalies of hot, altered rock in the uppermost part of the crust, which coincide with points of maximum tectonic/magmatic activity. These points correspond to the intersection of oblique fault swarms to the plate boundaries. Geothermal activity under mid-ocean ridges follows probably similar tectonic/magmatic anomalies.

Due to high permeability sea-water invades the bed-rock of the Reykjanes Peninsula, Iceland, and is overlain by a variably thick lens of dilute ground water of meteoric origin. The variable degree of salinity of geothermal waters in the Reykjanes Peninsula has resulted from different degree of mixing of fresh ground water with the underlying sea-water-ground-water in the downflow zones around the geothermal fields. At Reykjanes the geothermal water represents heated sea-water without any freshwater mixing. The difference in the composition of sea-water or sea-water/fresh water mixtures and the geothermal waters is due to basalt/water interaction at elevated temperatures. The major-element chemistry of the geothermal water represents an equilibrium composition at the relevant aquifer temperatures. The activities of silica, calcium, sulphate, and carbonate are thus limited by the solubilities of quartz, anhydrite, and calcite. Fluoride activity is thought to be controlled by an ionic exchange reaction where it substitutes for hydroxyl groups in phyllosilicates. The ratios of individual cations and hydrogen ion are governed by ionic exchange equilibria with hydrothermal minerals, probably smectite and chlorite. The equilibrium pH for the Reykjanes and Svartsengi geothermal waters is 5·5 and 5·1 respectively. Sea-water will become somewhat acid upon heating to more than about 300 °C and equilibration with basalt, the acidity increasing with temperature.

Vernadite is a nanocrystalline turbostratic phyllomanganate containing Ni, and is widespread in surface environments and oceanic sediments. To improve our understanding of Ni uptake in this mineral, two series of analogues of vernadite (δ-MnO2) were prepared with Ni/Mn atomic ratios of 0.002—0.105 at pH4 and 0.002—0.177 at pH 7. Their structures were characterized using X-ray powder diffraction (XRD). The δ-MnO2 nano-crystals are essentially monolayers with coherent scattering domains sizes of ∼10 Å perpendicular to the layering and ∼55 Å within the layer plane. For Ni/Mn < 0.01, the layer charge deficit is apparently balanced entirely by interlayer Mn, Na and protons. At higher Ni/Mn, Ni occupies the same site as interlayer Mn above and below vacant sites within the MnO2 layer and at sites along the edges of the layer. However, the layer charge is balanced differently at the two pH values. At pH 4, Ni uptake is accompanied by a reduction in structural Na and protons, whereas interlayer Mn remains strongly bound to the layers. At pH 7, interlayer Mn is less strongly bound and is partially replaced by Ni. The results of this study also suggest that the number of vacant octahedral sites and multi-valent charge-copmpensating interlayer species are underestimated by the currently used structure models of δ-MnO2.

A. De Schulten published a crystallographic and optical investigation of the bismuth oxyhalides in 1900. He succeeded in preparing minute square plates capped with pyramidal planes, but evidently the difficulties of manipulation prevented the measurement of interfacial angles. He did, however, record a negative uniaxial figure observed in a direction perpendicular to the plane of the square plates of all three salts BiOCl, BiOBr, and BiOI. The present work was undertaken to extend A. de Schulten's observations, to measure the axial ratios, and to determine the crystal-structure.

This process has been reinvestigated, mainly with X-rays, using an FeO-containing, fibrous variety (nemalite), and also a nearly pure specimen. The results confirm that the transformation is oriented, the c- and a-directions of the brucite becoming the normals to (111) and (11̄0) in the periclase. There is an intermediate, spinel-like stage; the spinel a-axis is parallel to, and twice as long as, that of the periclase. This stage is especially prominent when nemalite is heated in air or nitrogen, but is shown also by the nearly pure material. The nemalite shows also a further intermediate stage under certain conditions. A new hypothesis is proposed for the dehydration mechanism, in which the number of oxygen atoms per unit volume is ahnost unchanged in those parts of the crystal that are converted into periclase. The process is essentially one of cation migration, and the occurrence of the spinel-like phase is readily explained. Similar mechanisms possibly apply to the dehydration of other lamellar hydroxides and oxy-hydroxides (e.g. gibbsite, kaolinite) and of certain carbonates and other oxy-salts.

The first broad grouping of meteorites was into irons and stones according as they consisted mainly of nickeliferous iron or of silicates. These were the two main divisions of the first really serviceable classification as applied by Gustav Rose in 1862-4 to the collection of meteorites in the, University Museum of Berlin. In this classification the division of meteoric irons included as separate groups the pallasites and the mesosiderites, in which nickel-iron "rod silicates are present in about equal amounts; and the meteo~fic stones were for the first time split up into chondrites, or stones containing those curious rounded grains (chondrules) peculiar to meteorites, and non-Chondritic stones, which were divided according to mineralogical composition into the groups of eucrites, howardites, &c., still largely recognized.

Liquid-hydrocarbon-bearing fluid inclusions have often been described associated with petroleum occurrences and diagenetic sediments. Infra-red microspectrometry allows characterization of fluid inclusions greater than 20 µm by establishing the presence of aliphatic and aromatic hydrocarbons as well as associated H2O, CO2 and CH4. Semi-quantitative analyses have been made by focussing on gaseous and liquid phases separately. Some CH2/CH3 and CO2/CH4 ratios have been determined by this method.

Fluorescence microspectrometry permits precise measurements of the fluorescence emission spectrum of chromophore-bearing organic phases (essentially aromatic hydrocarbons) in fluid inclusions greater than 10 µm. Such a spectrum is a function of both the gross composition of the trapped oil and its thermal history.

Both of these methods lead to the in situ characterization of hydrocarbon fluid inclusions. They are useful in providing a quantifiable distinction between different oil generations trapped during mineral growth in diagenetic and epigenetic minerals.

The crystal structure of whewellite [Ca(C2O4)·H2O] and the dehydration mechanism of caoxite [Ca(C2O4)·3H2O] have been studied by means of differential thermal analysis, X-ray diffraction (powder and single-crystal) analysis and infrared analysis. The first and second analyses confirmed the direct transformation of caoxite into whewellite without an intermediate weddellite [Ca(C2O4)·2H2O] stage. Infrared spectra obtained from caoxite, weddellite and whewellite emphasize the similarity of the O–H-stretching band and O–C–O-stretching band in whewellite and caoxite and the unique bands of weddellite. The structure refinement at low temperature (123 K) reveals that all the hydrogen atoms of whewellite form hydrogen bonds and the two water molecules prop up the crystal structure by the hydrogen bonds that cause a strong anisotropy of the displacement parameter.

Comparing the structural features of whewellite with those of weddellite and caoxite suggests that caoxite and whewellite have a sheet structure consisting of Ca2+ ions and oxalate ions although weddellite does not. It is additionally confirmed that the sheets of caoxite are corrugated by hydrogen bonds but whewellite has flat sheets. The corrugated sheets of caoxite would be flattened by dehydration so the direct transformation of caoxite into whewellite would not occur via weddellite. Essential for this transformation is the dehydration of interlayered water molecules in caoxite leading to the building of the crystal structure of whewellite on its intralayered water molecules. The difference in conformation of water molecules between those two crystal structures may explain the more common occurrence of whewellite than of caoxite in nature.

Two contrasting types of Hercynian U-rich granites can be distinguished by their accessory minerals: (1) subalkaline potassic granites (amphibole-biotite±pyroxene) characterized by apatite-zircon-thorite-sphene-allanite and rare ThO2-rich uraninite, in order of appearance in the magma, and (2) leucocratic granites (biotite and muscovite) which contain monazite-zircon-apatite-xenotime-ThO2-poor uraninite and Ti oxides. Intermediate accessory mineral associations may also exist.

The mineralogy, crystallization history, and geochemistry of two high K-U-Th granites from the Vosges, France, is described. Their assemblage of U- and Th-bearing accessory minerals is listed, in order of appearance in the magma, in Group 1 (above). Trace elements (U, Th, Zr, Hf) and preliminary data on REE are discussed. In a layered sequence of the Crêtes granite, thorite, with up to 15% UO2, is shown to be important during differentiation. There is a decrease of the Th/U ratio corresponding to a slight decrease of the Th content whereas there is an increase of the U content at the end of differentiation. The Crêtes and Ballons granites have REE patterns characterized by LREE enrichment with a La/Yb ratio of about 28 (Crêtes granite) and a slight negative Eu anomaly. These relationships are typical of subalkaline potassic granites of the Hercynian. The trace element associations are similar to those of the shoshonite association with high U, Th, LREE and Ba, Rb, and Sr contents. A hypothetical model for uranium mineralization associated with such granites is discussed in the light of these data.

Eudialyte-group minerals (EGM) represent the most important index minerals of persodic agpaitic systems. Results are presented here of a combined EPMA, Mössbauer spectroscopy and LA-ICP-MS study and EGM which crystallized in various fractionation stages from different parental melts and mineral assemblages in silica over- and undersaturated systems are compared. Compositional variability is closely related to texture, allowing for reconstruction of locally acting magmatic to hydrothermal processes. Early-magmatic EGM are invariably dominated by Fe whereas hydrothermal EGM can be virtually Fe-free and form pure Mn end-members. Hence the Mn/Fe ratio is the most suitable fractionation indicator, although crystal chemistry effects and co-crystallizing phases play a secondary role in the incorporation of Fe and Mn into EGM. Mössbauer spectroscopy of EGM from three selected occurrences indicates the Fe3+/ΣFe ratio to be governed by the hydration state of EGM rather than by the oxygen fugacity of the coexisting melt. Negative Eu anomalies are restricted to EGM that crystallized from alkali basaltic parental melts while EGM from nephelinitic parental melts invariably lack negative Eu anomalies. Even after extensive differentiation intervals, EGM reflect properties of their respective parental melts and the fractionation of plagioclase and other minerals such as Fe-Ti oxides, amphibole and sulphides.