The prospect of partial melting of the subducted oceanic crust to produce arc magmatism has been debated for over 30 years. Debate has centred on the physical conditions of slab melting and the lack of a definitive, unambiguous geochemical signature and petrogenetic process. Experimental partial melting data for basalt over a wide range of pressures (1–32 kbar) and temperatures (700–1150°C) have shown that melt compositions are primarily trondhjemite–tonalite–dacite (TTD). High-Al (> 15% Al2O3 at the 70% SiO2 level) TTD melts are produced by high-pressure (≥ 5 kbar) partial melting of basalt, leaving a restite assemblage of garnet + clinopyroxene ± hornblende. A specific Cenozoic high-Al TTD (adakite) contains lower Y, Yb and Sc and higher Sr, Sr/Y, La/Yb and.Zr/Sm relative to other TTD types and is interpreted to represent a slab melt under garnet amphibolite to eclogite conditions. High-Al TTD with an adakite-like geochemical character is prevalent in the Archean as the result of a higher geotherm that facilitated slab melting. Cenozoic adakite localities are commonly associated with the subduction of young (<25 Ma), hot oceanic crust, which may provide a slab geotherm (≍9–10°C km−1) conducive for slab dehydration melting. Viable alternative or supporting tectonic effects that may enhance slab melting include highly oblique convergence and resultant high shear stresses and incipient subduction into a pristine hot mantle wedge. The minimum P–T conditions for slab melting are interpreted to be 22–26 kbar (75–85 km depth) and 750–800°C. This P–T regime is framed by the hornblende dehydration, 10°C/km, and wet basalt melting curves and coincides with numerous potential slab dehydration reactions, such as tremolite, biotite + quartz, serpentine, talc, Mg-chloritoid, paragonite, clinohumite and talc + phengite. Involvement of overthickened (>50 km) lower continental crust either via direct partial melting or as a contaminant in typical mantle wedge-derived arc magmas has been presented as an alternative to slab melting. However, the intermediate to felsic volcanic and plutonic rocks that involve the lower crust are more highly potassic, enriched in large ion lithophile elements and elevated in Sr isotopic values relative to Cenozoic adakites. Slab-derived adakites, on the other hand, ascend into and react with the mantle wedge and become progressively enriched in MgO, Cr and Ni while retaining their slab melt geochemical signature. Our studies in northern Kamchatka, Russia provide an excellent case example for adakite-mantle interaction and a rare glimpse of trapped slab melt veinlets in Na-metasomatised mantle xenoliths.

Most natural systems display non-linear dynamic behaviour. This should be true for magma mingling and mixing processes, which may be chaotic. The equations that most nearly represent how a chaotic natural system behaves are insoluble, so modelling involves linearisation. The difference between the solution of the linearised and ‘true’ equation is assumed to be small because the discarded terms are assumed to be unimportant. This may be very misleading because the importance of such terms is both unknown and unknowable. Linearised equations are generally poor descriptors of nature and are incapable of either predicting or retrodicting the evolution of most natural systems. Viewed in two dimensions, the mixing of two or more visually contrasting fluids produces patterns by folding and stretching. This increases the interfacial area and reduces striation thickness. This provides visual analogues of the deterministic chaos within a dynamic magma system, in which an enclave magma is mingling and mixing with a host magma. Here, two initially adjacent enclave blobs may be driven arbitrarily and exponentially far apart, while undergoing independent (and possibly dissimilar) changes in their composition. Examples are given of the wildly different morphologies, chemical characteristics and Nd isotope systematics of microgranitoid enclaves within individual felsic magmas, and it is concluded that these contrasts represent different stages in the temporal evolution of a complex magma system driven by nonlinear dynamics. If this is true, there are major implications for the interpretation of the parts played by enclaves in the genesis and evolution of granitoid magmas.

Interaction processes between acid and basic magmas are widespread in the Sardinia–Corsica Batholith. The resulting hybrid magmas are extremely variable and can be broadly divided into: (i) microgranular mafic enclaves with geochemical characteristics of both magmatic liquids and cumulates; (ii) basic gabbroic complexes with internal parts mainly formed by cumulates and with interaction zones developing only in the marginal parts; and (iii) basic septa with the form of discrete, lenticular-like bodies often mechanically fragmented in the host rock. Different styles of interaction, ranging from mixing to mingling, have been related to variations in several physicochemical parameterś, such as: (i) the initial contrast in chemical composition, temperature and viscosity; (ii) the relative mass fractions and the physical state of interacting magmas; and (iii) the static versus dynamic environment of interaction.

A model is presented for the origin and history of interaction processes between basic and acid magmas based on the geochemical characteristics of hybrid magmas. Physico-chemical processes responsible for the formation of hybrid magmas can be attributed to: (i) fractional crystallisation of basic magma and contamination by acid magma; (ii) loss of the liquid phase from the evolving basic magma by filter pressing processes; (iii) mechanical mixing between basic and acid magmas; and (iv) liquid state isotopic diffusion during the attainment of thermal equilibrium.

Plutonic complexes with interlayered mafic and silicic rocks commonly contain layers (1–50 m thick) with a chilled gabbroic base that grades upwards to dioritic or silicic cumulates. Each chilled base records the infusion of new basaltic magma into the chamber. Some layers preserve a record of double-diffusive convection with hotter, denser mafic magma beneath silicic magma. Processes of hybridisation include mechanical mixing of crystals and selective exchange of H2O, alkalis and isotopes. These effects are convected away from the boundary into the interiors of both magmas. Fractional crystallisation aad replenishment of the mafic magma can also generate intermediate magma layers highly enriched in incompatible elements.

Basaltic infusions into silicic magma chambers can significantly affect the thermal and chemical character of resident granitic magmas in shallow level chambers. In one Maine pluton, they converted resident I-type granitic magma into A-type granite and, in another, they produced a low-K (trondhjemitic) magma layer beneath normal granitic magma. If comparable interactions occur at deeper crustal levels, selective thermal, chemical and isotopic exchange should probably be even more effective. Because the mafic magmas crystallise first and relatively rapidly, silicic magmas that rise away from deep composite chambers may show little direct evidence (e.g. enclaves) of their prior involvement with mafic magma.

Anhydrite has been identified as a phenocrystic phase in some silicic volcanic magmas, but it is not commonly described in plutonic rocks. Anhydrite-bearing magmas tend to form in arc environments and to contain hydrous, low-temperature, oxidised mineral assemblages. Phenocrystic anhydrite coexists with sulphur-enriched apatite and sometimes with pyrrhotite, in silicate melt that contains from 50 ppm to 1 wt% S, depending on temperature and conditions. Vapour coexisting with anhydrite- and water-saturated magma may contain from a few tenths of a mole per cent to a few mole per cent sulphur gases (SO2 and H2S), with the exact composition and gas speciation depending on temperature and oxygen fugacity. Samples of one anhydrite-bearing magma, the 1991 Pinatubo dacite, have been experimentally crystallised to determine whether the magma retains its characteristic sulphur-rich mineral phases during solidification. Results show that anhydrite and sulphur-rich apatite are retained throughout crystallisation and vapour phase evolution. This suggests that anhydrite-bearing intrusive equivalents of the Pinatubo dacite should be present in arc plutonic complexes.

Numerical models that account for fluid flow, magmatic and metamorphic fluid production, topography and thermal expansion of the fluid following emplacement of a granitic magma in the upper crust reveal controls on the distribution of magmatic fluids during the evolution of a hydrothermal system. Initially, fluid pressures are close to lithostatic in and near an intrusion, and internally generated magmatic and metamorphic fluids are expelled. Later, fluid pressures drop to hydrostatic values and meteoric fluids circulate throughout the system. High permeabilities and low rates of fluid production accelerate this transition. Fluid production in the magma and wallrocks is the dominant mechanism elevating fluid pressures to lithostatic values. For granitic intrusions, about three to five times as much magmatic fluid is produced as metamorphic fluid. Continuous fluid release from a granitic magma with a vertical dimensions of 10 km produces a dynamic permeability of up to several tens of microdarcies.

Near the surface, topography associated with a typical volcano acts to maintain a shallow meteoric flow system and drive fluids laterally. The exponential decay with depth of the influence of topography on fluid pressures results in a persistent zone of mixing at a depth of 1-2 km between these meteoric fluids and magmatic fluids despite variations in the strength of the magmatic hydrothermal system. However, in shallow systems where fluid release is episodic, dramatic changes in the region of mixing are still possible because fluid pressure is sensitive to variations in the rates of fluid production. At depth, high rates of metamorphic fluid production in the wallrocks and low permeabilities (< 1 μD) produce elevated fluid pressures, which hinder the lateral flow of magmatic fluids. Together, these patterns are consistent with the distribution and evolution of skarns and hydrothermal ore deposits around granitic magmas.

In southwestern North America, late Palaeozoic through Cenozoic granitoids and their related mineral deposits show consistent patterns that can be interpreted in terms of combined provincial, exposure and process controls. Voluminous Cordilleran magmatism began in the Permian and continued with few major interruptions through the Mesozoic and Cenozoic, reaching maximum fluxes in the mid-Jurassic, Late Cretaceous and Oligocene. Two distinctive types of broad-scale igneous suites formed. The first type consists of calc-alkaline to alkaline suites that vary regularly with time from early intermediate-mafic centres to late felsic centres over intervals lasting 20–50 Ma. These suites formed during periods of stable convergence and compressional tectonics, most notably in the late Mesozoic and early–mid-Cenozoic. The second type is compositionally varied, but shows no obvious secular variation in composition. This type formed during neutral to extensional tectonics in the mid-Mesozoic and the mid- to late Cenozoic. Regional (west to east) and secular (old to young) changes from calcic to alkalic compositions do not correspond to basement types; they point to tectonic rather than crustal controls on magmatic evolution, although basement signatures are clearly transmitted in isotopic systematics. Contrasting types of intrusive centres formed in the same lithospheric columns, suggesting that variability reflects thermal and stress regimes, subcrustal magma flux and crustal thickness. Simple thermal and mechanical models of limits on assimilation and magma uprise are broadly consistent with these patterns.

Igneous-related mineralisation is ubiquitous where epizonal environments are preserved, thus preservation (and exposure) form the first-order filter on metallogeny. Mineralisation includes porphyry, skarn, epithermal, replacement and syngenetic deposits of widely varying styles, metal contents and links to magmatic heat and materials. Metal contents and alteration styles correlate closely with igneous compositions and are broadly independent of setting, although systematic regional variations in metal ratios are documented. Ore element suites vary from Cu–Au–Fe associated with (quartz) dioritic to monzonitic intrusive centres through Cu–Zn–Mo–Pb–Ag–W–Au associated with broadly granodioritic centres, and finally to F–Mo–Zn–W–Ag–Be associated with metaluminous to strongly peraluminous granitic centres. A model that includes both composition and process controls rationalises this igneous correlation and the lack of strong regional control. Key features are (1) mineralogical controls on fluid compositions and (2) the efficacy of magmatic processes in producing voluminous ore-forming aqueous fluids. This interpretation is supported by field relationships, igneous petrographic and isotopic data, and theoretical considerations.

Ore element ratios in intrusion-related mineralisation are in part a function of the relative oxidation state and degree of fractionation of the associated granite suite. A continuum from Cu-Au through W to Mo dominated mineralisation related to progressively more fractionated, oxidised I-type magmas can be traced within single suites and supersuites. Such systematic relationships provide strong evidence for the magmatic source of ore elements in granite-related mineral deposits and for the production of the observed ore element ratios dominantly through magmatic processes. The distribution of mineralised intrusive suites can be used to define a series of igneous metallogenic provinces in eastern Australia. In general, there is a correlated evolution in the observed metallogeny (as modelled based on the compatibility of ore elements during fractionation) with increasing degree of chemical evolution of the associated magmatic suite. This is from Cu-Au associated with chemically relatively unevolved magmas, through to Sn and Mo-rich mineralisation associated with highly evolved magmas that had undergone fractional crystallisation. Provinces recognised in that way do not necessarily correlate with the tectonostratigraphic boundaries defined by the near-surface geology, indicating that the areal distribution of some granite source regions in the deep crust is unrelated to upper crustal geology.

Comb-layered quartz is a type of unidirectional solidification texture found at the roofs of shallow silicic intrusions that are often associated spatially with Mo and W mineralisation. The texture consists of multiple layers of euhedral, prismatic quartz crystals (Type I) that have grown on subplanar aplite substrates. The layers are separated by porphyritic aplite containing equant phenocrysts of quartz (Type II), which resemble quartz typical of volcanic rocks and porphyry intrusions. At Logtung, Type I quartz within comb layers is zoned with respect to a number of trace elements, including Al and K. Concentrations of these elements as well as Mn, Ti, Ge, Rb and H are anomalous and much higher than found in Type II quartz from Logtung or in igneous quartz reported elsewhere. The two populations appear to have formed under different conditions. The Type II quartz phenocrysts almost certainly grew from a high-silica melt between 600 and 800°C (as β-quartz); in contrast, the morphology of Type I quartz is consistent with precipitation from a hydrothermal solution, possibly as α-quartz grown below 600°C. The bulk compositions of comb-layered rocks, as well as the aplite interlayers, are consistent with the hypothesis that these textures did not precipitate solely from a crystallising silicate melt. Instead, Type I quartz may have grown from pockets of exsolved magmatic fluid located between the magma and its crystallised border. The Type II quartz represents pre-existing phenocrysts in the underlying magma; this magma was quenched to aplite during fracturing/degassing events. Renewed and repeated formation and disruption of the pockets of exsolved aqueous fluid accounts for the rhythmic banding of the rocks.

The primary focus of this review is on P-T conditions, mineralogical indicators of melt or fluid composition and textural evolution; lesser treatment is given to pegmatite sources or to pegmatite-wallrock interactions. Investigations of stable and radiogenic isotopes have revealed that the source materials for pegmatites are likely to be more heterogeneous or varied than previously thought, especially for peraluminous pegmatites, but that overall pegmatites bear a clear intrusive relationship with their hosts, as opposed to an origin in situ. The P-T conditions of crystallisation of some lithium-rich pegmatites have been constrained by lithium aluminosilicate stability relations in combination with stable isotope or fluid inclusion methods. Experimental studies have elucidated the effects of components such as Li, B, P and F, which are common in some classes of pegmatites, to liquidus relations in the hydrous haplogranite system. Experimentation has also provided corroboration of an old concept of pegmatite crystallisation—that pegmatites owe their distinctive textures and mineral/chemical zonation to relatively rapid crystallisation of melt from the margins inwards at conditions far from the equilibrium (i.e. from supercooled liquids). The origin of aplites, whether alone, layered, or paired with pegmatites, remains an active area of research. Studies of fluid inclusions, crystal-vapour equilibria and wallrock alteration have helped to define the timing and compositions of vapour phases in pegmatites and to aid in the economic evaluation of deposits.

The current underlying assumption in most geochemical studies of granitic rocks is that granitic magmas reflect their source regions. However, the mechanisms by which source rocks control the intensive and compositional parameters of the magmas remain poorly known. Recent experimental data are used to evaluate the ‘source rock model’ and to discuss controls of (1) redox states and (2) the Sr isotopic compositions of granitic magmas.

Experimental studies have been performed in parallel on biotite-muscovite and tourmaline-muscovite leucogranites from the High Himalayas. Results under reducing conditions ( = FMQ – 0·5) at 4 kbar and variable suggest that the tourmaline-muscovite granite evolved under progressively more oxidising conditions during crystallisation, up to values more than four log units above the FMQ buffer. Leucogranite magmas thus provide an example of the control of redox conditions by post-segregation rather than by partial melting processes.

Other experiments designed to test the mechanisms of isotopic equilibration of Sr during partial melting of a model crustal assemblage show that kinetic factors can dominate the isotopic signature in the case of source rocks not previously homogenised during an earlier metamorphic event. The possibility is therefore raised that partial melts may not necessarily reflect the Sr isotopic composition of their sources, weakening in a fundamental way the source rock model.

The Early Proterozoic (1715 Ma) Harney Peak Granite (Black Hills, SD, U.S.A.) is a complex of hundreds of dykes and sills. Earlier studies of Nd, O and Pb isotope variations demonstrated that the complex was not derived from a single source, or even different sources of a single age. Instead, the granites can be divided into a group with sources probably dominated by Early Proterozoic sediments and a group with sources probably dominated by Archean sediments. New results on the Nd isotopic variations of many additional samples indicate that there is considerable overlap between Nd isotopic compositions within the complex. Values of εNd (1715 Ma) of the Harney Peak Granite suite (n = 20) range from −2·0, indicating an Early Proterozoic (2300-2200 Ma) crustal source, to −13·4, indicating a Middle to Late Archean (3200-3100 Ma) protolith. These results suggest that the Early Proterozoic source may have included rocks such as the c. 2200-1900 Ma metasedimentary rocks that occur in the southern Black Hills. The Archean sources might have included rocks such as those exposed on the periphery of the Black Hills. The range in Nd model ages negates the usefulness of the concept of the ‘average’ age of the crust in this part of the craton. Because such heterogeneity is present in the magmatic compositions of the Harney Peak Granite, it can be inferred that at least as much heterogeneity was present in the sources. In this granite system, melts were evidently derived from isolated, heterogeneous zones and did not have the opportunity to coalesce into large magma bodies. In systems where coalescence does occur, the evidence for such highly heterogeneous sources may be lost. These results emphasise that inferences drawn from a few samples of plutonic rocks in which magma mixing and homogenisation occurred can lead to erroneous conclusions about the age and nature of protoliths and, consequently, the development of continental crust.

The Lu-Hf and Re-Os isotope systems have been applied sparsely to elucidate the origin of granites, intracrustal processes and the evolution of the continental crust. The presence or absence of garnet as a residual phase during partial melting will strongly influence Lu/Hf partitioning, making the Lu–Hf isotope system exceptionally sensitive to evaluating the role of garnet during intracrustal differentiation processes. Mid-Proterozoic (1·1–1·5Ga ) ‘anorogenic’ granites from the western U.S.A. appear to have anomalously high εHf values, relative to their εNd values, compared with Precambrian orogenic granites from several continents. The Hf-Nd isotope variations for Precambrian orogenic granites are well explained by melting processes that are ultimately tied to garnet-bearing sources in the mantle or crust. Residual, garnet-bearing lower and middle crust will evolve to anomalously high εHf values over time and may be the most likely source for later ‘anorogenic’ magmas. When crustal and mantle rocks are viewed together in terms of Hf and Nd isotope compositions, a remarkable mass balance is apparent for at least the outer silicate earth where Precambrian orogenic continental crust is the balance to the high-εHf depleted mantle, and enriched lithospheric mantle is the balance to the low-εHf depleted mantle.

Although the continental crust has been envisioned to have exceptionally high Re/Os ratios and very radiogenic Os isotope compositions, new data obtained on magnetite mineral separates suggest that some parts of the Precambrian continental crust are relatively Os-rich and non-radiogenic. It remains unclear how continental crust may obtain non-radiogenic Os isotope ratios, and these results have important implications for Re-Os isotope evolution models. In contrast, Phanerozoic batholiths and volcanic arcs that are built on young mafic lower crust may have exceptionally radiogenic Os isotope ratios. These results highlight the unique ability of Os isotopes to identify young mafic crustal components in orogenic magmas that are essentially undetectable using other isotope systems such as O, Sr, Nd and Pb.

Granite is the final product of the high-temperature, magmatic, predominantly endogenic, chemical differentiation of the earth. Our understanding of the origin and evolution of granitoid rocks comes from a combination of direct observation, analogue experimentation and numerical modelling. A brief historical overview shows an exceptional level of such research activity over the last 50 years. The number and complexity of questions have resulted in both an absolute and a relative growth of the science since the plate tectonic revolution, largely consisting of refining the current magmatic paradigm within its overarching context. Current research activity involves large components of mineralogical–petrological–geochemical and structural–tectonic work, with much lower levels of experimental, geophysical and geochronological investigations. Many important questions concerning the thermal, physical and chemical aspects of the origin and evolution of granites remain. In keeping with the general progress of science, the complexity of the questions, the declining financial support and the revolution in information technology, directions of granite research in the foreseeable future will change from concrete and qualitative to abstract and quantitative, from expensive and active to cheap and armchair, from reductionist to holistic, and from periodic communication to continuous communication.